There are real differences between 16 bit and 24 bit files, just as there are real differences between uncompressed and mp3. As with the differences between wav and mp3, 16 bit vs 24 bit is often difficult to identify by listening.
Test signals are one thing, but has anyone found any 24 bit music recording that can be successfully identified via ABX testing against a properly resample to 16 bit version of same?
I'd imagine ABX would be possible if you found a sample with a near silent portition in the middle and listened with the volume cranked up uncomfortably loud. The 16 bit version should have noise down at ~-100dB. Assuming your equipment was good enough, the room quiet enough, and you didn't deafen yourself on the preceeding section (all big assumptions), you could probably ABX it.
I recall a test here some years back to try to see if people could ABX a 16-bit original from the same clip reduced to 15- 14- and 12-bit depth with proper dithering. The result was that very few people could ABX the 15-bit file (only one as I recall) and many could not ABX the 12-bit file. Being able to distinguish 24-bit from 16-bit should be even more difficult.
If the clip is properly dithered, this doesn't really prove much of anything other than the dithering did its job. The whole point of dithering is to make the resultant 16-bit file sound as much like the 24-bit version as possible. A better real-world test would be to record two different multi-track sessions, one in 16 bit and one in 24 bit. You would still have to dither the final mixes, but the difference, especially if there was some degree of processing, should be quite distinguishable.
As with the differences between wav and mp3, 16 bit vs 24 bit is often difficult to identify by listening.
Really? With a good set of speakers, I think most mp3's are relatively easy to identify...
If the clip is properly dithered, this doesn't really prove much of anything other than the dithering did its job. The whole point of dithering is to make the resultant 16-bit file sound as much like the 24-bit version as possible. A better real-world test would be to record two different multi-track sessions, one in 16 bit and one in 24 bit. You would still have to dither the final mixes, but the difference, especially if there was some degree of processing, should be quite distinguishable.
Very few people question that 24bit is superior to 16bit for mastering and processing purposes. The question is really whether it's useful to distribute content to the customer in 24bit over 16bit. Dithering a 24bit signal down to 16bit is, in my opinion, the right approach to make samples to test the hypothesis that end-users need 24bit formats.
As with the differences between wav and mp3, 16 bit vs 24 bit is often difficult to identify by listening.
Really? With a good set of speakers, I think most mp3's are relatively easy to identify...
Welcome to HA. I see this is your first post. Can I suggest that you read the Terms of Service (http://www.hydrogenaudio.org/forums/index.php?showtopic=3974), especially number 8. If you do indeed find the majority of mp3s easy to identify in a proper double blind test, then you should contribute to some of the encoder development and testing that goes on around here.
Andy: Can I suggest a classical sample like Ravel's Bolero or Copland's Appalacian Spring - a good recording of both are very dynamic with very quiet sections and very loud sections. I have tried 16 versus 24 before and failed, but I think samples like these might be useful.
I agree that a good sample, if it exist in music, would be something that has very low level passages, perhaps one or both of those suggested. Unfortunately I have nothing of that sort on which to test. The only sources I have to record at 24 bit are already existing recordings on vinyl, tape, and CD. Those are not good candidates for the obvious reasons. The differences exist at every signal level of course, even from a CD going D to A followed by A to D, but detecting those differences with human hearing is the task.
Just to reiterate, this post is not about theory. It is an inquiry about what people have experienced. Has anyone been able to do a valid ABX test? Is so, on what?
Welcome to HA. I see this is your first post. Can I suggest that you read the Terms of Service (http://www.hydrogenaudio.org/forums/index.php?showtopic=3974), especially number 8.
Ah, sorry about that. No, I've not participated in any double-blind studies.
If you do indeed find the majority of mp3s easy to identify in a proper double blind test, then you should contribute to some of the encoder development and testing that goes on around here.
I'd be happy to.
The only sources I have to record at 24 bit are already existing recordings on vinyl, tape, and CD. Those are not good candidates for the obvious reasons.
Why not record a musical selection from a CD that does not have a wide dynamic range (i.e. rock / electronica / contemporary jazz), but record it peaking at around -24dB? It's not an ideal solution, but if those three are your only sources, it would minimize the effect of the original recording being at 16 bit, because (I believe) 16bit and 24bit have been shown to be fairly indistinguishable at high recorded levels. Alternately, just recording an acoustic guitar or a piano at 24bit seems like it would be a reasonably good source.
Anywho, hope I haven't stepped on any (more) toes...
Tom
Just to reiterate, this post is not about theory. It is an inquiry about what people have experienced. Has anyone been able to do a valid ABX test? Is so, on what?
I know negative results aren't what you are looking for, but I have spent a while trying this on native 24bit material and failed. This is probably due to one of two things. First, my listening environment isn't very quiet. My PC is extremely quiet and I use good headphones, but the traffic noise from the nearby road is still audible. Second, a large amount of 'native' 24 bit material I have come across doesn't have a better noise floor than properly dithered 16bit material - that is, the noise floor is above quantization noise.
It would be pretty cool if somebody could point us towards a real 24bit recording with a noise floor of less than, say, -100dB.
I suppose it's quite easy to construct a track with very low-level musical content, with very high-level transients, so that the steady-state content is 80-100db below the transient peak. However, for that to work, you'd need to ABX quantization noise in a signal that is perhaps only 20-30db in volume. Still quite hard, you'd need IEMs with good noise isolation, but certainly possible. The transient, obviously, needs to be short enough not to cause permanent damage.
Oh oh oh - another idea. Dump all the high-amplitude energy into low frequencies. Think of the infamous cannon shot in Teldec's 1812 Overture LP. You'd need a sub for that, but it would probably avoid the more pressing issues of ear damage.
Also, how contrived can the listening environment get? Would it be reasonable to use a compressor in the playback chain?
The question isn't whether or not it is possible to distinguish something in 24 bit vs 16 bit, although that is a challenge in itself, but to distinguish some music. Can it ever make a difference with music?
For instance, using a test tone, it is fairly easy to demonstrate the difference between converting from 32 bit to 16 bit with dithering vs no dithering, but doing the same thing with real music is more problematic.
Also, in my test tone trials, I can easily enough hear the difference between dithered and undithered 16 bit, but even under these simplified conditions, not between 32 bit and dithered 16 bit. Those differences are easy to see on screen and to measure, however.
Well.. how much dynamic range can you coax out of a close-mic'd drum? Virtually any drum you hear on any recording is either heavily compresser or not recorded in the nearfield.
I'm sorry, that went over my head. What is the relevance to the search for some music recording that has audible differences between 16 bit and 24 bit?
haha nothing.. see 24 bits is purist technology. it is in theory suppose to provide more resolution and hence better sound quality but at the same time there are other factors to consider during playback, such as DACs, the speakers, the listening room/environment, etc. plus even if you COULD abx between 16 bits and 24 bits i doubt that the difference will be that big to fuss about. with today's advanced dithering systems such as pow-r, waves idr etc. and plus given how most music is mastered these days, loud and proud, you won't hear a difference on your consumer level equipment, maybe a slight difference on a very audiophile setup if your hearing ability is that good. and i mean BLIND tests.
I'd wager that in order to distinguish the difference, you'd damage your hearing.
In the spirit of ABX testing aren't you supposed to listen at a comfortable level without aggressively adjusting the volume or eq? IOW, isn't it unrealistic to crank the level while listening to quiet passages just to see if you can hear a difference in the noise floor?
OSHA's recommendations for 120db is that it should not be listened to for over 7.5 minutes, and the recommended exposure time halves for every 5db increase. So, you'd figure that for 1 minute of exposure time, you could go as far as 135db. But obviously very few headphones could even go that far safely, and the OSHA figures are only a model. But I see nothing wrong with a single 120db transient as part of an ABX testable wave file. The biggest problem might be that temporary hearing thresholds can change by 10-20 db the first time you hear it, so it might take several minutes between tests for your ears to cool down.
But are you talking about real music or, effectively, just a test signal? Even a verified test signal would be something, considering my experience with test signals, but I don't understand the loud part. It seems that only something of very low signal level is likely to be distinguishable -- if anything is. I'm think I'm missing your point completely.
My point is that you don't need a "test signal" to do this, and that if you limit the discussion of wide dynamic ranges to percussion, one can think up a plausible scenario where such dynamic range is necessary to capture something. Think of any musical instrument capable of 120db of measurable dynamic range, and there you go. Or, alternatively, you only need 100db, or 80db, or however much you need while still being able to hear the quantization noise.
OK, you need more than the dynamic range of 16 bit to faithfully capture the analogue. Easy enough to understand. Then the test can be made. Do a proper resample to 16 bit with noise shaped dither -- or not noise shaped if you have some (strange) idea that will be better. Compare the two.
Of course, the question is still open. Can anyone find or make such a recording that can be successfully ABXed? If you choose your dither carefully, especially without noise shaping, that alone could differentiate the two, but properly done (single pass) noise shaped dither isn't very audible.
With a test tone, I can hear the quantizing noise at the end of a fade when the resample to 16 bit is not dithered but I cannot hear anything at the same point when I dither. With CoolEdit's Statistics I measure Peak -94dB, RMS Average -96dB on the undithered (can hear; at high volume level, of course) vs Peak -77, RMS Average -89dB dithered (can not hear). These numbers compare with Peak -150, RMS Average -161 on the original 32 bit file. Go figure.
I think the best volume for the test would be 'normal' listening levels. Low level signals have less bits to respresent them. If you only listen to low level signals at higher volumes then you aren't really comparing 16-bit- to 24-bit. You may be comparing 10-bit to 18-bit.
Also, most analogue equipment has a noise floor significantly higher than a pure 24-bit noise floor. And, when recording, some of the additional dynamic range (gained from more bits) is used for headroom so that actual amount of bits used for the recording may be as low as 21 or 22.
I would suggest, therefore, it is impossible to realise true 24-bit performance from an actual recording.
Without cryogenic cooling the intrinsic noise of electronics puts the useable limit at about 20 bits, maybe 19. This is a great deal more resolution than 16 bits can provide directly. Properly dithered 16 bit can approach that but at a higher noise level.
I think many of us are fairly well convinced there is nothing to gain from a greater bit depth on playback. Recording, mixing and mastering is a different question not being addressed here. However, there are a fair number of people out there who are convinced that 24 bit systems sound "so much better." The challenge is to find some evidence; to produce any 24 bit recording that at least one person can objectively differentiate from a properly dithered 16 bit version of same.
If someone can tell them apart, it isn't necessarily because of the greater resolution of very low level detail. There are other measurable differences. The best evidence seems to be that those differences just are not perceptible to humans.
I think it's a good idea to mention the importance of the sampling rate when we talk about the audibility of quantization noise. For example you can use a 16/96 setup to get a similar SNR performance compared to 24/48 in the audible frequency range (by use of clever noise shaping filters) plus a flat response upto 40 kHz. In this case 16 bit would be more than enough for distribution.
Even a simple 4th order IIR filter can lower the noise floor about 20 dB below 0.4 times the nyquist frequency -- in case of 96 kHz that'd translate to a 20 dB increase of the dynamic range below 19,2 kHz (and a 20 dB decrease above 28.8 kHz but that shouldn't bother anyone)
I think it's a good idea to mention the importance of the sampling rate when we talk about the audibility of quantization noise. For example you can use a 16/96 setup to get a similar SNR performance compared to 24/48 in the audible frequency range (by use of clever noise shaping filters) plus a flat response upto 40 kHz. In this case 16 bit would be more than enough for distribution.
First you would have to demonstrate the audibility of quantization noise in 16 bit/44.1kHz stuff to find any excuse to go to a higher sampling rate. Such can be demonstrated in test situations, i.e. simple test tones at very low levels -- but only if the 16 bit file is undithered.
If dithering is used, the quantization noise is not audible. If rational noise shaping is used (it may be clever but it is quite ordinary in digital audio), even the dither is unlikely to be heard. For me, and I suspect for most people, 16 bit, dithered and noise shaped, cannot be differentiated from 32 bit under ABX test conditions (at 44.1kHz) even with simple test tones.
Maybe things would be slightly different under exceptional conditions of no external noise and
extremely low noise analogue electronics, but has that been demonstrated? Going to a higher sample rate would move the shaped noise to a still higher frequency, and therefore further frrom the audible band, but would hardly be worth the markedly increased storage cost if it in fact, as is likely, has no audible benefit.
Even a simple 4th order IIR filter can lower the noise floor about 20 dB below 0.4 times the nyquist frequency -- in case of 96 kHz that'd translate to a 20 dB increase of the dynamic range below 19,2 kHz (and a 20 dB decrease above 28.8 kHz but that shouldn't bother anyone)
I don't know what you are writing about. What are the parameters of such a filter (filter variety, cutoff, highpass/lowpass/band pass/, transition bandwidth, etc)? What is it filtering out? Is it hardware or software? Where might it be employed? If it does something useful, why is it not in normal use?
First you would have to demonstrate the audibility of quantization noise in 16 bit/44.1kHz stuff to find any excuse to go to a higher sampling rate.
I don't have to do anything. I didn't claim that higher sampling rates are needed -- just that lower sample resolutions can be used at higher sampling rates to get the same SNR performance in the audible band. I commented because everyone was
only talking about bits/sample and not about samples/second. That's all.
Even a simple 4th order IIR filter ...
I don't know what you are writing about. What are the parameters of such a filter (filter variety, cutoff, highpass/lowpass/band pass/, transition bandwidth, etc)? What is it filtering out? Is it hardware or software? Where might it be employed? If it does something useful, why is it not in normal use?
I'm talking about
noise shaping filters. These filters are usually applied in conjunction with dithering
while a signal is requantized in software (ie from 24 to 16 bits) to shape the quantization noise spectrally. See this thread (http://www.hydrogenaudio.org/forums/index.php?showtopic=47980), download noise.zip (http://homepages.uni-paderborn.de/sgeseman/noise.zip) and checkout the
sym20.sos filter. If you feel like plotting its response in Matlab: Here's what you should do:
sos = [ 1.0 -0.6160081826103387 0.7225 1.0 0.8253356149096804 0.25
1.0 -0.8253356149096804 0.25 1.0 0.6160081826103387 0.7225 ];
[b,a] = sos2tf(sos);
freqz(b,a);
This (http://homepages.uni-paderborn.de/sgeseman/sym20.png) is a screenshot of the filter design tool showing the filter's response on a logarithmic amplitude scale (Y-axis, each horizontal line = 10 dB, the bright line correspondes to 0 dB) and linear frequency axis (X-axis, 0..nyguist frequency). If you requantize 24/96 to 16/96 using this filter you can still have a SNR of above 110 dB below 19 kHz -- not just 90 dB. If you use the
sym30.sos filter instead you can go down to 12 bits/sample and have an SNR of 102 dB below 19 kHz and the PCM data rate would be equal to the rate you need for 24/48.
Cheers!
Sebastian
If you have decent equipment and a quiet environment, it should be easy for you to distinguish this test file I created (source 48khz, 24-bit) to a 16-bit conversion.
It's 5 seconds long, and under 100KB. Very low-level pure tone(sine wave) @ 3.5khz (since this is where our hearing is most sensitive).
Test Tone File(48/24) (http://www.sendspace.com/file/9w4hm1)
Convert it to 16-bit using any method you choose, any dither type of any level, I could abx it easily. I have already tried using at least a few dozen combinations of noise shaping and dither levels, along with no dithering (which obviously sounds worse). Always ABX'able. When properly dithered, the tone is still audible at 16-bit, but the noise floor is much worse and the tone becomes more difficult to distinguish from the noise (where it sounds fine at 24-bit).
You need fairly high output levels to do this, but since the file contains no peaks or anything that would resemble "loud", you wont risk damaging your hearing or equipment so long as no other "computer sounds" play during testing.
[Edit: If anyone prefers, here is a 44.1khz version of the same file, created fresh, not via SRC]
Test Tone File(44.1/24) (http://www.sendspace.com/file/9t9w9x)
Yes, it is easy to create test tones that demonstrate a difference. That was my opening statement. The question is, is there any music?
The vast majority of all music most of us have access to is in 16-bit.
The real question to me is not if the difference is audible in music during critical listening, but whether we can get a sufficient sample library of 24-bit tracks in order to appreciate such a difference. Since we do not have this (or at least, I do not), we are reduced to using synthetic test tones to demonstrate the difference in the meantime.
And as for "is there music?" I would most certainly say yes, if you consider all genres that people currently count as "music". That may just be 1 "test tone" I posted, but I could easily make it part of, say, a PC-Based techno song which people actually enjoy listening to, and at low levels they would easily be able to ABX my track between a 24 and 16 bit version.
These articles on ff123's website (http://ff123.net/) may be of interest to you:
The MAD Challenge (http://ff123.net/madchallenge.html)
Analysis of Ethan Winer's Bit-Depth Listening Test (http://ff123.net/24bit/24bitanalysis.html)
The first link compares 16-bit and 24-bit decodes of an MP3. I just glanced at the article, so I'm not sure if the MP3 is provided, but if it's not linked in the article it probably is on ff123's samples page. (Look for a link from the home page.)
The Bit-Depth Listening Test provides a dynamic 24-bit audio recording and several bit-truncated versions, and it's pretty well put-together. Click through the links in the article and you'll get to the samples themselves (http://www.pcavtech.com/test_data/).
You need fairly high output levels to do this, but since the file contains no peaks or anything that would resemble "loud", you wont risk damaging your hearing or equipment so long as no other "computer sounds" play during testing.
That may just be 1 "test tone" I posted, but I could easily make it part of, say, a PC-Based techno song which people actually enjoy listening to, and at low levels they would easily be able to ABX my track between a 24 and 16 bit version.
Well, which is it?
Greynol,
What I meant to imply by "at low levels", was parts of the song where the overall level of the track itself is very quiet. High gain playback would still be necessary, as I mentioned prior. Thank you for helping me to clarify.
You need fairly high output levels to do this, but since the file contains no peaks or anything that would resemble "loud", you wont risk damaging your hearing or equipment so long as no other "computer sounds" play during testing.
Any noise (even 24-bit dither) can be made audible with enough monitoring gain. That's not really a fair test. It can be handy though to "zoom in" to details that would normally be inaudible, but are still good to know they exist. Just like a printer checking a printout with a magnifying glass.
Quite a few mastering engineers claim audible differences between dither flavours at normal playback levels. I'm pretty sure these claims have not been confirmed by double-blind testing though.
Best thing to do is find a mastering engineer who is willing to share a sample with a clear 24/16 bit difference. I'll try to contact a few. Please do so too if you have the possibility.
The vast majority of all music most of us have access to is in 16-bit.
What about taking a DVD-Audio release (generally they're 24 bit) and rerecording the two-channel analog output digitally, at 24 bits?
I've got a couple of dozen DVD-As, several of which I've re-digitized in just that fashion via my M-Audio 2496 card so i could make flac files from them (after converting to 16-bit -- but I still have the 24-bit 'masters').
(Alternately those who have the DVD-A ripping software could use that).
As far as I can see, the ff123 page does not provide the 24 bit file, just various 16 bit, and lesser bit depth, versions. It isn't addressing the question of this thread, it is just about various way of presenting material resampled to lower bit depths.
At least one of the engineers I've asked seems prepared to provide some samples where the difference between 24 and 16 bit should be "easily" audible. I'm waiting for more samples to come.
Would it be an idea to set up a public listening test for this ? I've never done that, so any help and suggestions would be welcome.
-How long should a sample be in order to be valid ? (my guess: 30 sec. max., see HA TOS #9 (http://www.hydrogenaudio.org/forums/index.php?showtopic=3974&st=0&p=292534&#entry149482))
-What is the preferred sample rate ? (my guess: 44.1 kHz with 48 optional)
-What kind of music is preferred (my guess: any should do)
IMO it is better to keep samples much shorter: 10-15 sec. That way it will be easier for people to concentrate just on the part that is important. To conduct listening test you can simply follow Roberto's document about conducting listening test.
Maybe 48 kHz would be better to avoid issues if some listener use those soundcards that have crappy sampling rate convertors. But then on those soundcards it might also be impossible to hear the difference anyhow.
There are a few 24/96 and 24/44.1 files here:
http://www.pcabx.com/technical/reference/index.htm (http://www.pcabx.com/technical/reference/index.htm)
McDougalsMen24bit_48kHz.flac is here:
http://ff123.net/samples.html (http://ff123.net/samples.html)
Cheers,
David.
Setting aside for the moment whether this donated tone, 24-bit-ABX, might ever be considered music, my first listening based impression was that I can tell no difference between the 24 bit and a dithered 16 bit version. A run of 14 trials with WinABX said otherwise, however. I guessed correctly 12 of those times, for a guessing probability of 0.6%.
Whatever I based my guesses on was below the level of my conscious awareness. I doubt that it would ever occur to me that there might be a difference if I did not know there was supposed to be.
I turn my headphone amplifier up quite far rather often. I need to listen to samples that are down 60 to 70 dB, as part of the restoration process I do (mostly from LPs). I know from painful experience that forgetting to turn the dial back down is very unpleasant when switching to "normal" level music.
In order to even hear this test tone, I need the volume setting to be much higher than I ever use it in my daily noise sample testing. In order to hear it well enough to do the ABX test, I had the volume set on maximum. I don't know what sound pressure level is produced by this. Your equipment might provide significantly greater amplification; your hearing might be significantly more sensitive. However, I am reasonably sure that no human could listen to music at anything approaching that setting on my system unless the entire composition was at a similar input level as this test tone.
That, of course, would be too contrived to accept. To be reasonable, the file would then need to be normalized to "normal" music levels, where the differences between 24 bit and 16 bit would probably be much less. After normalizing it to -16dB (which is loud at a volume control setting below my normal listening level), and resampling a copy to 16 bits, a run of 14 trials produced 13.something% probability of guessing for me. I'm not sure how to interpret that.
Again, laying aside the question "is it music?", is there any disagreement that a piece that contained this tone, along with parts at more normal levels (thus precluding the volume control being set at an extreme level) would need to be listened to under extraordinary conditions for such low level parts to be heard at all? In real music, such low level sounds might occur at a fade out, but probably no one ever hears them.
Also, music is not always at a constant level. For this sample to be useful in the ABX tests, it needs to be part of something, as emphasized by Axon. Adding higher level music might change the ABX test results considerably -- unless one insists that really extended durations of such a low level tone are a worthwhile contribution to some composition. Who knows what musical genius might wrought? Tunes for the Sensory Deprivation Vat
So, let me restate part of the hypothesis that led to this thread. I don't remember just what was written in what other thread, but something was being discussed with one or more proponents of "24 bit is better" and that prompted me to start this thread (I've made similar challenges in other forums. This test tone is the first offering with which anyone has ever responded).
I acknowledge that there are real differences between 16 bit and 24 bit, but is there any actual music where that difference can be distinguished under real world conditions? While most of them might not hang around this neighborhood, there are more than a few people who proclaim that 24 bit recordings (as finished products) are so much better than CDs. I rather doubt it is because said recordings contain parts at -110dB. I want someone to produce even one sample of some 24 bit music that can be successfully ABXed against a properly resampled 16 bit version of same. I'm not saying it doesn't exist but I want to hear it for myself.
An addendum to the comments about higher sample rates by SebastianG, which comments I afraid didn't make much sense to me: I generated two samples of 32 bit silence, one at 48kHz (keeping with the parameters of the tone under consideration here) and one at 96kHz. I resampled both to 16 bit with the dither I normally use.
At maximum headphone volume, the 48kHz sample sounds the same as silence; I hear nothing (from the added noise shaped dither). However, the 96kHz sample is definitely audible. Perhaps this is a defect of the soundcard when running at 96kHz. CoolEdit's Statistics say the 96kHz file has both lower peak and lower RMS levels.
Repeating the 96kHz silence sample with a noise shaping preset labeled for 96kHz, the perceived loudness is decreased (but still audible), but the measured levels are higher than for even the 48kHz dithered sample.
I don't know what a "public listening test" might involve, but I want to be able to resample the 24 bits to 16 bits myself. I suspect it is possible to bias the results towards 24 bits by doing a less than splendid job of resampling.
I don't know what a "public listening test" might involve, but I want to be able to resample the 24 bits to 16 bits myself. I suspect it is possible to bias the results towards 24 bits by doing a less than splendid job of resampling.
My idea was to ask a few Majors in the audio industry for samples at 24 and 16 bit. Preferably at 44.1 kHz and with the same dither flavour they use for their cd-releases. These samples will be made publicly availble for testing purposes, allowing anyone to try to hear a difference.
That doesn't stop you from making your own 16-bit version of the 24-bit sample if you think you can do a better job.
IMO this test essentially boils down to comparing various dither versions. It seems that there are no double blind test results publicly available. This could be an opportunity for HA. It's recommended that the listening test is performed at normal (healthy) listening levels with real life music samples, preferably from commercial releases to be representative (this might be open for discussion).
An ABX test to compare 16 and 24 bit dithered audio (44.1 kHz) won't be easy (although at least one of the ppl I've asked for samples thinks it is).
It will probaby require quite some time and motivation. Would HA be a good place for such a test, since 320 kbps is already considered transparent by a majority of the members ?
Just to get an idea, who of you would like to (seriously) participate in such a test ?
I don't know what a "public listening test" might involve, but I want to be able to resample the 24 bits to 16 bits myself.
I suspect it is possible to bias the results towards 24 bits by doing a less than splendid job of resampling.
Out of curiosity, what kind of dither would you use ?
I believe the resampling is more important than the dither. I can use any dither possible with CoolEdit, and I'm open to suggestions for test purposes. Mainly I use shaped triangular to a depth of 0.5, CoolEdit's noise shaping curve C1. I've also used triangular, dither depth 0.7, C3 quite a bit.
Neither of these is audible by itself. I guess that doesn't necessarily mean they don't have audible effects on the music but nothing I've noticed. In fact, it is quite difficult to find any music where dithering or not dithering makes any audible difference when going from 24 bit (32 float) to 16 bit.
Test tones are easy but most music has enough self dither, from the microphone preamp if nothing else, that adding more, or not adding any, is irrelevant. I've also asked several forums, this one included, I believe, for any 24 bit music sample anyone thinks will exhibit a difference. I've always done this when the thread was in some manner stressing the importance of dither or some aspect thereof, at least by some posters. No one has ever offered any actual music.
Hi everyone.
A year has passed since the previous post in this thread.
Some of the posts have referred to listening for the noise floor. That to me would be a means of testing but would not go to the heart of the matter.
If the superiority of leaving the final mix at 24-bits is not evident in soft passages when listening with a normal amount of gain, one must wonder whether it is worthwhile at all for the final product to have more than a 16-bit sampling depth.
So I ask, is it still a moot point whether a 24-bit musical extract has been successfully ABX tested as sounding better than a 16-bit dithered version?
"sounding better" is a subjective evaluation. The question here is whether or not it is possible to tell any difference what-so-ever. Just as one can not prove there is no Easter Bunny, no one can supply evidence that there is no 24 bit music recording that can be differentiated from a 16 bit version of same. So we have to approach the problem from the other direction: can anyone supply a 24 bit recording that people can tell from a 16 bit version of same simply by listening to them (in an ABX presentation, of course)
I am new to ABXing but I think that for playing 24-bit material it may be wise to use ABX software where the audio driver in use is known, as some drivers ignore the extra bits. I have just installed a full version of foobar2000 and I see it offers ABX testing, and selection of the audio driver.
When I get some spare time I will compare some selections of 24-bit material (e.g. some amateur 24/96 recordings) with 16-bit derived versions. I see that foobar2000 can even be used to convert files so as to create 16-bit versions (with or without dither).
And if I find I can reliably hear a difference I will report back, and upload (or identify) the test file. (In the past I have felt quite confident I could detect a more liquid sound with greater than 16-bits, but the proof of the pudding will be in the ABX testing.)
Even it is comparing apples with oranges:
Do you expect to hear differences between 16 bit and 24 bit at normal listenig conditions, when you can't seperate 192kbit CBR-MP3 vs. Original CD-16bit/44kHz rip?!
Any lossy Encoder does much more "harm" to the sound, then 24bit audio is noiseshaped & dithered to 16bit audio.
Wow: Nov 2nd 2006 to Feb 27th, 2008… and still waiting for A/B/X-able material (!)
Cool thread nonetheless - I’m just hoping that someone will be able to source two lossless .wav files in the aforementioned bitrates. I'll offer to host em' free for a couple of years if someone can dig them up in the first place.
Andrew D.
www.cdnav.com
Since DVD-A can now be ripped, having 24/96 stereo material is not an issue.
The success of 16 vs 12 bit tests should tell us a lot about the prospect of 24 vs 16 being detactable, I cant understand why those arent properly discussed.
The line between test-signal and music is blurred when you search for killer-samples. Lets just say something like "any snip of sound less than 30s long that is normalised to 0dB FS and still comfortable to listen to?"
Caution
Is it concievable that the 24bit version might sound different but worse than dithered 16bits in some PCs if 24 bits was truncated to 16 bits by soundcard/OS?
-k
Caution
Is it concievable that the 24bit version might sound different but worse than dithered 16bits in some PCs if 24 bits was truncated to 16 bits by soundcard/OS?
Good point. Any playback chain that is not capable of processing 24 bits correctly is not suitable for this kind of testing.
I remember remarks (don't know where) that indicate that "a slow fade out" could be one of the things you might want to try. In effect bits are gradually taken away during a fade out.
The reason we see 24 bits anyway is probably the idea (as said before) that it is useful during recording/editing in the studios and so why not leave it without reduction (regardless if a difference can be heard at home). But, sorry, that has nothing to do with proof.
I think that the difference between 12 bits and 16 bits is a lot easier to prove, both due to technical and perceptual issues.
If one tries 12 vs 16 first, and cannot proove the difference, then continuing with 16 vs 24 is a waste of time (either due to your equipment, ears, room, content... ).
-k
I think that the difference between 12 bits and 16 bits is a lot easier to prove, both due to technical and perceptual issues.
If one tries 12 vs 16 first, and cannot proove the difference, then continuing with 16 vs 24 is a waste of time (either due to your equipment, ears, room, content... ).
-k
True, however 12 bits is very easy to pick simply by listening for the background noise, at a normal listening level, so perhaps a more reasonable initial challenge would be to try with 14 bits.
For example, there is a well presented set of samples at this (http://64.66.184.67/technical/bits44/index.htm) webpage of the
PC ABX Web Site. Here are my results from this evening, mainly involving 12 bits vs 16 bits:-[blockquote]Sample 2.2.6 (Triangle 44.1KHz 12 bits) is much noisier than sample 2.2.1 (Triangle 44.1KHz 16 bits).
The poor quality of 12 bits is even more stark with the gliding tone samples. Artefacts are clearly audible in sample 1.1.7 (Tone sweep 44.1KhZ, 12 bit, undithered) but sample 1.1.2 (Tone sweep 44.1KHz, 16 bit, dithered) is of good quality.
However, even that 16 bit sample can be heard to have slight deficiencies if you listen to the last section of it at high gain, and compare it with sample 1.1.1 (Tone sweep 44.1KHz, 24 bit). To satisfy hydrogenaudio protocol I confirmed the samples sounded different with ABX testing, but I had no hesitation identifying which sample was which, just through noise levels.
Turning to the piano extracts provided on the webpage, the recording is very noisy and the difference between the 44.1Khz 12-bit sample (2.1.6) and the 16-bit sample (2.1.2) is not particularly obvious [to my ears]. I find it disappointing that such a poor sample was used.[/blockquote]
If I start with an ordinary 16-bit file of quiet music, drop the volume level by 24dB, save the file, open the file and raise the volume level by 24dB I think this should get rid of about 4 least significant bits, leaving 12 bits. I performed this procedure tonight and the noise introduced was apparent, even at a normal listening level.
Sorry for interrupting your conversation, but I don't want to create another topic only for this:
I would like to know if there is any difference in playing 16bit lossless file with 24bit or 16bit output format (i use foobar). Is there also a difference (degreed/improved/same quality) for lossy formats? And while my card support 24bit output, should I use it instead of 16bit.
Almost all my files are 16bit flacs, but I need some space so I'll rip my cds with mp3 until I get a new hdd.
Caution
Is it concievable that the 24bit version might sound different but worse than dithered 16bits in some PCs if 24 bits was truncated to 16 bits by soundcard/OS?
Good point. Any playback chain that is not capable of processing 24 bits correctly is not suitable for this kind of testing.
An you should prove the playback chain is really 24 bit capable (ie not rely only on the specs of the hardware/software).
Sorry for interrupting your conversation, but I don't want to create another topic only for this:
I would like to know if there is any difference in playing 16bit lossless file with 24bit or 16bit output format (i use foobar). Is there also a difference (degreed/improved/same quality) for lossy formats? And while my card support 24bit output, should I use it instead of 16bit.
Almost all my files are 16bit flacs, but I need some space so I'll rip my cds with mp3 until I get a new hdd.
I'll try to answer this.
With the gliding tone examples at http://64.66.184.67/technical/bits44/index.htm (http://64.66.184.67/technical/bits44/index.htm) on the
PC ABX Web Site using foobar in ABX mode and set so as to play only the last few seconds of the samples (i.e. the quietest part), setting the foobar output at 24-bits seemed to give the best performance. What I found was as follows:-
[blockquote]The result with the option set to '24-bits' was good for sample 1.1.2 (16-bit dithered). The dithering algorithm used for preparing the sample must have been of good quality. Switching the output option to 16 bits gave a much noisier result (and this was much the same result whether or not dither was selected as an output option in the foobar preferences).
With sample 1.1.11 (16-bit, not dithered), artefacts were more noticeable (spurious whistles) on my setup (an Audigy 4 with external hub) than with sample 1.1.2.
Setting the foobar output to 24-bits gave the quietest output. Setting the output to 16-bits increased the noise (somewhat hiding the whistles). There was little difference between setting the foobar output dither to on rather than off.[/blockquote]
With playing of 16-bit music more generally, I don't see why it should make much difference whether the codec is lossy or lossless. If it is lossless but derived from a source digital recording at more than 16 bits the source needs to be reduced to 16-bits resolution somehow using dither, rounding, or -- conceivably -- truncation. It has to be remembered that 16-bit 'lossless' still has the limitations of 16-bits: it is a perfect copy of a 16-bit resolution approximation. The question we are trying to answer in this thread is whether that approximation is so good that no difference can be detected between 16 bits and a greater number of bits when listening to music. I think it must be assumed that the gain used when listening to the test extract cannot be unrealistically high. By unrealistically high, I mean that the loudest parts of the music must not be unbearably loud when played back with the same gain setting used to listen to the test extract.
It is trivial to prove that 24-bit music sounds different to 16-bit music if the 24-bit source has a very high signal to noise ratio, part of the performance is extremely quiet, and that part of the performance is listened to at an unrealistically high gain setting.
Caution
Is it concievable that the 24bit version might sound different but worse than dithered 16bits in some PCs if 24 bits was truncated to 16 bits by soundcard/OS?
Good point. Any playback chain that is not capable of processing 24 bits correctly is not suitable for this kind of testing.
An you should prove the playback chain is really 24 bit capable (ie not rely only on the specs of the hardware/software).
Choosing a quiet part of a high signal to noise ratio 24-bit sample should enable the noise level to be assessed. When 16 bit sound-card mode is selected, background noise should rise quite substantially compared with operating the card in 24 bit mode. Foobar conveniently makes the output bit-depth selectable and also allows playback to be confined to just a selected part of a file.
Obviously this is only a rough and ready test, but it may be sufficient to indicate that the system is operating at more than 16 bits.
AndyH, perhaps you might be kind enough to answer my two questions below.
Being a newcomer to this site, I’ve only now read some of the other threads on the 24 vs 16 bit debate. I’ve also now read some of the material on the complex subject of dithering.
Before submitting any sample of 24-bit audio that contains a section that can be demonstrably distinguished from 16 bits (testing with ABX software), I’d just like to check on the ground rules for such an exercise.
Question 1:
Given that it can be demonstrated (using low level test tones) that artefacts in a 24-bit to 16-bit conversion are more evident if dithering is not used, and given that dithering is widely in use and recommended, is the use of some form of dithering compulsory for this exercise? If it is not compulsory, then I might seek to identify a sample where artefacts were in evidence.
Question 2:
If the answer to question 1 is that dithering is compulsory, or at least acceptable, and given that dithering introduces an amount of noise, does this exercise exclude from consideration simpler forms of dithering that introduce much more noise than is necessary? More specifically, are any of the dither methods available with readily available software such as Audacity or foobar acceptable for this exercise? If acceptable dithering method(s) can be indicated, I would seek to identify a sample where the noise of the dithering is detectable when comparing a derived 16-bit version with a 24-bit original.
************************************************************
What is the theoretical signal to noise level available for 16-bit format of a conventional audio compact disc?
It is often suggested that 16-bits provide around 96.3 dB of signal to noise ratio, based on there being 2^16 instantaneous levels that can be represented with 16 bits. However, it’s an extremely complicated question. The S/N ratio can be argued to be infinite (if a series of zeroes is used between tracks of the CD), or extremely high (if a series of zeroes is sent interspersed with an occasional positive bit, or occasional negative bit [which could apply is using high frequency dithering to make a low frequency low level tone audible]).
Substituting a slightly different (and much simpler!) question, for sine waves with no direct current offset and which are at a frequency not far below half the sampling rate, the maximum [peak] amplitude that can be represented either side of zero is half of 2^16 , and the minimum [peak] amplitude that can be represented either side of zero is 1. Half of 2^16 is 32768. Voltage ratios can be expressed in decibels using the expression 20 x log (voltage ratio). That gives us the figure 90.3 dB as the ratio between the peak level of the loudest sine wave and the softest [highly imperfect] representation of that sine wave without dither. The low level wave doesn’t really have a shape to make the description ‘peak’ meaningful, being some variety of rounded square wave. And the filtering needed might alter the peak value of the low level ‘sine wave’ somewhat.
Recording at 24-bits and keeping the noise below that which applies for the 16-bit format
Without attempting to pin down a definitive answer to the S/N ratio question, I have found in practice that the required signal to noise ratio of recording equipment needed to compete with that available in the 16-bit format is quite high. What I have found is that is stretches my equipment to the limit. A Behringer analogue mixer struggles to provide a signal that has noise much less than 90dB down when set up so as not to clip in loud sections of a recording that approach 0 dB. A Creative Audigy 4 soundcard with external hub struggles to provide better than about 94 dB with some audio drivers. By locating the Audigy 4 card in the pc in a PCI slot away from interference, and by using an ASIO ALL driver (instead of the normal a creative driver), and by setting the electronic fader in the creative recording mixer to less than 100%, I was able to get over 100 dB available for use for 24-bit recording.
Then there is a question of room noise where the recording is made. A close microphone technique is needed.
When all of this is done, I find that a dithered 16-bit version of a 24-bit recording can be audibly noisier at a realistic (though rather high) listening level, during quieter passages of a recording.
However, a lot may depend of the type of dither used. I would prefer not to submit any sound recording extract until I have a better understanding of what forms of dither are considered acceptable.
Being a newcomer to this site, I’ve only now read some of the other threads on the 24 vs 16 bit debate. I’ve also now read some of the material on the complex subject of dithering.
Before submitting any sample of 24-bit audio that contains a section that can be demonstrably distinguished from 16 bits (testing with ABX software), I’d just like to check on the ground rules for such an exercise.
Question 1:
Given that it can be demonstrated (using low level test tones) that artefacts in a 24-bit to 16-bit conversion are more evident if dithering is not used, and given that dithering is widely in use and recommended, is the use of some form of dithering compulsory for this exercise? If it is not compulsory, then I might seek to identify a sample where artefacts were in evidence.
Yes, dithering should be compulsory for this kind of test. It is well understood than the quantization process works "better" if there is the "right" amount of noise added to the signal. This has to do with reducing the correlation between the quantization error (or residue) and the signal to be quantized. Where error is not correlated with the signal, the only result of quantization is the addition of noise - not the addition of distortion.
Not applying the optimal type of dither before performing such a test would seriously call into doubt the validity of the results.
Question 2:
If the answer to question 1 is that dithering is compulsory, or at least acceptable, and given that dithering introduces an amount of noise, does this exercise exclude from consideration simpler forms of dithering that introduce much more noise than is necessary? More specifically, are any of the dither methods available with readily available software such as Audacity or foobar acceptable for this exercise? If acceptable dithering method(s) can be indicated, I would seek to identify a sample where the noise of the dithering is detectable when comparing a derived 16-bit version with a 24-bit original.
If your hypothesis is "reduction of 24bit music samples to 16 bits using optimal techniques is not transparent", then you would have to use the theoretically "best" dither available. If your hypothesis is "reduction of 24bit music samples to 16 bits using X technique is not transparent", then you can use any kind of dither available. Note that confirmation of the second hypothesis is still useful.
************************************************************
Substituting a slightly different (and much simpler!) question, for sine waves with no direct current offset and which are at a frequency not far below half the sampling rate, the maximum [peak] amplitude that can be represented either side of zero is half of 2^16 , and the minimum [peak] amplitude that can be represented either side of zero is 1. Half of 2^16 is 32768. Voltage ratios can be expressed in decibels using the expression 20 x log (voltage ratio). That gives us the figure 90.3 dB as the ratio between the peak level of the loudest sine wave and the softest [highly imperfect] representation of that sine wave without dither. The low level wave doesn’t really have a shape to make the description ‘peak’ meaningful, being some variety of rounded square wave. And the filtering needed might alter the peak value of the low level ‘sine wave’ somewhat.
Without dither, some of the assumptions behind the quantization process break down. In most basic analyses, it is assumed that the quantization error for high bit counts is not correlated with the signal to be quantized - where this does not hold the analysis becomes much, much more complex. I would recommend borrowing or buying a copy of a good DSP book (Oppenheim and Schafer, for example) for answers to these questions better than those that I can offer you.
Hi!
I'd like to remind you that there's a free tool around for requantization, dithering and noise shaping that supports arbitrary bit depths and custom noise shaping filters: NOISE (http://homepages.uni-paderborn.de/sgeseman/)
Note: The built-in filters as well as the other supplied filters might not be the best ones. I encourage you to try creating your own filters.
Using the appropriate noise shaping & dither settings 12 bits/sample @ 44kHz already sounds good enough to me. So, I feel quite safe with the common 16/44 format.
Cheers,
SG
I know this isn't the only time I raised the topic in this forum, and I have done the same thing in a number of other forums. Every time I write sort of off the cuff, that is, without a preset script. I'm not in the mood to reread this entire thread so I can't say exactly what I might have already discussed in detail here, as compared with other times I addressed the subject.
However, I did include the necessary basic conditions in the first post:
a properly resample to 16 bit version of same
It may be possible, using poor quality dithering, with no noise shaping or poor noise shaping, to make an audible difference because of the added noise. That is not "properly" done. Many people here will remember that in the recently published year long ABX test of DVD-A & SACD vs resampled to 16 bit, the only differences detected by anyone were the "silence" between tracks, turned up to extremely high levels, where the unshaped dither used in their resampling was audible.
Proper resampling means, to me, what CoolEdit/Audition does with resampling, dithering, and noise shaping. I won't go into the evidence that has been presented in various places, but the program's resampling is probably as good as it gets.
It also has a fair number of dithering and noise shaping options. They might not be equivalent to the industry's best, but some of them work extremely well. I am only interested in the options that work well. The only point of using something else would be to encourage failure, not something anyone knowledgeable would do with a real music project.
So, I'm looking for a 24 bit music sample that I can resample to 16 bit in CoolEdit and tell the difference. I know there is other very good resampling software, but I also know that quite a few of the commonly used mixing/mastering programs do a relatively poor job of it. Just saying that someone has managed an ABX test isn't satisfactory without adequate evidence that the resampling wasn't done in such a way as to encourage an audible difference.
I'd like to remind you that there's a free tool around for requantization, dithering and noise shaping that supports arbitrary bit depths and custom noise shaping filters: NOISE (http://homepages.uni-paderborn.de/sgeseman/)
Cool tool, thanks for letting us know about it (and writing it )
However, I did include the necessary basic conditions in the first post:
a properly resample to 16 bit version of same
Good post, but you neglect the fact that there is an important difference between changing the sampling rate (resampling) and changing the number of bits per sample. In my opinion, the latter should never be called resampling, as this is misleading.
~ Not applying the optimal type of dither before performing such a test would seriously call into doubt the validity of the results.
~ If your hypothesis is "reduction of 24bit music samples to 16 bits using X technique is not transparent", then you can use any kind of dither available. Note that confirmation of the second hypothesis is still useful.
Thanks cabbagerat for your various remarks. I am not sure I am all that attracted to the idea of proving that last hypothesis, 'X technique is not transparent' if better dithering is apparently so readily available. As I have recently discovered, dither using Audacity appears to be much better [certainly quieter] than another method I've used (software I paid for!), and Audacity is a free download.
~ It may be possible, using poor quality dithering, with no noise shaping or poor noise shaping, to make an audible difference because of the added noise. That is not "properly" done. Many people here will remember that in the recently published year long ABX test of DVD-A & SACD vs resampled to 16 bit, the only differences detected by anyone were the "silence" between tracks, turned up to extremely high levels, where the unshaped dither used in their resampling was audible.
Proper resampling means, to me, what CoolEdit/Audition does with resampling, dithering, and noise shaping. I won't go into the evidence that has been presented in various places, but the program's resampling is probably as good as it gets. ~
Thanks, AndyH. Just as well I asked. I had assumed the dither that is built in to some software I've been using for a few years, n-track studio, would do a passable job in terms of added noise. But I find that is not the case, now that I've started comparing the results of different forms of dither.
With some recordings I have recently made at 24-bits, and processed to 16-bits, I have obtained these results:
[blockquote]1. N-track Studio 4 set to 1-bit dither - the added noise is quite noticeable.
2. N-track Studio 4 set to 1-bit dither plus shaping - the added noise not as noticeable as in 1, but nevertheless is detectable at a loud but realistic listening level (detectability verified with ABX software).
3. Audacity: dither left at default (triangular) - the noise is slightly different to that in the 24-bit original but so low in amplitude I think it unlikely to be detectable at a realistic listening level [by me anyway].
4. Audacity: dither set to shaping - same as 3, though a slightly different quality to the noise.[/blockquote]
The extracts in which I have detected differences due to the noise level have used the dithering method of 2 above, so I guess that will not be good enough for the current exercise.
If I am to pursue this further, I will probably need to concentrate on other artefacts than noise, though the more I read, the less optimisitic I am of success.
It makes current complaints, in other web forums, about home theatre personal computers not being able to access all 24 bits from some Blu-ray audio formats (partly because of unresolved issued regarding HDMI and content protection protocols) seem unnecessarily alarmist, unless there is no dither used for the reduction to 16 bits. It even has to be queried why 24 bits are being transferred in the first place to some Blu-ray discs, which I guess is what was behind the DVD-A & SACD testing you mention. In relation to surround sound, I guess where 6 or so full range channels are in place, the total noise floor may begin to be an issue, particularly if a movie is created with a very wide dynamic range. It may be a slightly different exercise to produce a two channel CD at a normalized maximum recorded level.
When I first used a 24-bit audio card, a few years ago, for capturing amateur musical performances, I could immediately detect the improvement compared with capturing with a 16-bit sound card. However, it appears that provided dither is used in the mixdown from a 24-bit source, enough of the 24-bit intensity detail should normally find its way into the 16 bits, at least at lower signal frequencies, where we might otherwise detect the lack of intensity resolution.
I haven't given up, but as I say, I am no longer as optimistic.
When I first used a 24-bit audio card, a few years ago, for capturing amateur musical performances, I could immediately detect the improvement compared with capturing with a 16-bit sound card.
Not to be pedantic - but the only objective conclusion you can draw is that you immediately noticed a difference between the quality of the two cards. Unless you performed more testing than you have indicated you have no reason to believe the quality difference you detected had anything to do with one being 16 bit and the other 24 bit.
Thanks, Soap. I should have been more precise. When I started using the new card (an Audigy 2), I compared recordings made with it with its drivers set to 16-bits and with drivers set to 24-bits; and the 24-bit setting gave a noticeably different result (not a dramatic difference of course). As I was not intending to release the recordings in red book CD format, or other uncompressed 16-bit format, I never investigated dithering techniques, though I was aware they were recommended.
I'd also mention that once I had a 24-bit card available I no longer sought to make recordings at close to clipping level, but allowed the recordings much more headroom.
If I am to pursue this further, I will probably need to concentrate on other artefacts than noise, though the more I read, the less optimisitic I am of success.
With proper dither, there will be no artifacts other than noise.
I'd also mention that once I had a 24-bit card available I no longer sought to make recordings at close to clipping level, but allowed the recordings much more headroom.
I don't think anybody really doubts whether 24 bit is better than 16 bit for recording, mixing, processing, mangling, mastering, etc. the sensible debate is only whether 16bits is enough for final delivery to the customer.
With proper dither, there will be no artifacts other than noise.
If I am not mistaken, that is the question we are trying to resolve, whether the arte[/i]facts that can be heard (even with optimal dither) when test tones are used, set at very low levels, and listened to at a high gain setting, can also be heard by the human ear at realistic listening levels, with music.
I found the following software quite useful for getting a taste of artefacts:-
[blockquote]
What is the effect of a non-white random source in dithering?
I did a small and free software to show/hear the effect of dither. The dither itself can be noise but also a sine wave. So you can try and listen to the result.
Get it here (http://www.ohl.to/about-audio/audio-softwares/ditherer/)
[/blockquote]
With proper dither, there will be no artifacts other than noise.
If I am not mistaken, that is the question we are trying to resolve, whether the arte[/i]facts that can be heard (even with optimal dither) when test tones are used, set at very low levels, and listened to at a high gain setting, can also be heard by the human ear at realistic listening levels, with music.
Well, if that's the question, then to my mind it's already been resolved for 24bit -> 16bit conversions. Consider our 24bit music samples x[n] and dither signal d[n]. Now, we define e[n], which is the difference between the x[n] and quantize_to_16bits(x[n] + d[n]). Ok? Now, optimal dither would do two things:
1) Make e[n] and x[n] completely uncorrelated, and
2) Reduce the total power of e[n], while still achieving (1), or
3) Reduce the audibility of e[n] while still achieving (1).
From what I understand, for the case of quantization from 24bits (or 32 bits) to 16 bits, the formulation of the jitter signal d[n] to achieve (1) is well understood. The formulation of d[n] to achieve (2) and (3) is a little bit less well understood, but still can be very well approximation with common processes. What (1) means is that there is no distortion at all caused by the quantization process - only the addition of noise.
I think the whole 24 bit/16 bit thing boils down to the hypothesis:
Consumer music and movie soundtracks have sufficient dynamic range, that when listened to at "normal" levels, the noise introduced by optimal dithering and quantizing to 16 bits is audible.In my opinion, this hypothesis is yet to be resolved conclusively for real world music and movie soundtracks. This has been demonstrated to be true for some classes of test signals.
~ Well, if that's the question, then to my mind it's already been resolved for 24bit -> 16bit conversions. Consider our 24bit music samples x[n] and dither signal d[n]. Now, we define e[n], which is the difference between the x[n] and quantize_to_16bits(x[n] + d[n]). Ok? ~
Er, yes, "ok" after several re-readings!
Expressed non-mathematically, I can understand how a high frequency dither can be used to supplement missing intensity level detail for lower frequency signals; and to mask the remaining low levels of quantisation error.
I've decided that my chances of being able to hear artefacts when playing back real life music in a 16-bit format derived from a higher bit depth source, and using a quiet dither, are likely to be confined to the artefact of the dither noise itself.
Using a 300Hz test tone at -60dB, my AVRs when fed a 24-bit PCM bitstream, play the tone with some background noise, and a small edge of distortion. Substituting a 16-bit PCM undithered version of the test tone, the distortion becomes quite pronounced. Adding triangular dither, the pronounced distortion disappears but dither noise is quite apparent. [It is certainly the lesser of the two evils.]
Minus 60 decibels is a very low signal compared with 0dB, a ratio of 1:1000 in fact. And the dither noise is quite a bit quieter than that.
The noise level of my microphone preamplifier/mixer competes with the dither noise if the gain is set to avoid clipping at 0dB. So to make a recording of music (e.g. a pianoforte) where the dither noise was appreciably greater than other noises, and to include in the recording a passage within a few decibels of 0dB, I would need a new preamplifier. [Fortunately the microphones have very low noise.]
If I were to succeed in this exercise [ABX tested, and dither method acceptable], what would I have proven? Simply that at very high (many would say artificially high) listening levels, dither noise can be audible when listening to a high signal to noise ratio 24-bit stereo recording of music, reduced to 16 bits.
That is not the narrow result I had assumed. I had assumed that some other artefacts might be audible. But I gather that other artefacts will almost certainly be neutralised, or sufficiently masked, by the dither noise.
I generally issue this type of challenge when someone is loudly proclaiming the obvious superiority of “high definition” audio. That generally includes both bit dept greater than 16 and sampling rate greater than 44.1kHz. I don’t recall just why I open a new thread to ask about 24 bit, maybe just to see if anyone could come up with something in a non-confrontation posting (every time I’ve asked someone who is so certain of his/her beliefs for evidence, the conversation has stopped). I don’t think there is an way to differentiate except by way of the dither noise, and rarely even then, but the question is genuine. If such music exists, I would like to experience it.
I suspect, however, that even if someone can come up with a real sample, it will only be noticeable in a close comparison of the two versions. If one is just listening to a performance, one could never figure out which it was.
I've a added a 24-bit 48KHz original, and 16-bit dithered derived version, in the uploads forum under a new topic Short samples: 24 bit vs rendered to 16 bits dithered (http://www.hydrogenaudio.org/forums/index.php?showtopic=61833). It's 9 seconds of a simple piece by Handel, played on an upright piano by an amateur pianist.
These wave files certainly show a difference in noise level as a result of the dither (if listened to at very high gain). They also seem to show a difference in the sound quality when listened to, briefly, at high listening levels. However, before making too many further comments, I await any critique of the dithering method.
I presume comments on these files may be more likely to be made in the uploads forum under the new topic, rather than here, but I'm not sure what the preferred practice is.
Hello,
In september, I did an ABX between a 96 kHz 24 bits file and a 44.1 kHz 16 bits one, made by Voxengo R8brain from a 192 kHz 24 bits file from 2L (the Mozart one) : http://www.2l.no/hires/index.html (http://www.2l.no/hires/index.html) (you can even download a 352 kHz 24 bits DXD version and a 2.8 MHz DSD version !)
Voxengo's software says that 8-, 16- and 24 bit output is dithered with a gaussian noise and a slight noise-shaping. I have no idea if it is the proper way to do it.
The original account is here : http://www.homecinema-fr.com/forum/viewtop...59&start=60 (http://www.homecinema-fr.com/forum/viewtopic.php?f=1029&t=29891959&start=60)
It's part of an epic 24/96 ABX challenge in which a forumer fell from an "obvious, huge, unmissable" difference just before the ABX test, to a complete failure !
I got 10/10 in the ABX test, with closed headphones, listening to the initial fade-in. Comparing the playback level to speakers, and using a sonometer, I think that the listening level was such that the 0dB of the digital scale was at 113 dB(A) on a white noise. Since the track has a peak level of -1 dB and an RMS level of -23 dB, the listening level of my ABX test is something like 90 dB RMS, with peaks at 112 dB.
The use of closed headphones, that isolate from the environment, especially the computer noise, is useful. I also had to "release" the internal pressure in my ears (like divers) several times in order to stay at the top of my sensitivity.
What is interesting is to compare the quantization noise to the file content. By ascending order, there is :
-The 16 bits noise. Without this ABX test, I would never have thought that it was audible
-The background noise of the microphones and of the recording location. The 16 bits noise doesn't seem to be completely masked.
-The breathing and movements of the musicians, far above the previous
-The music, still far above.
Playback system : Foobar2000 v0.8 in 24 bits mode (note that this version used to equalize the RMS levels of files to be ABXed).
Marian Marc 2 soundcard, analog output, MME drivers, Windows XP (Waveout, MME device in Foobar).
Superex Pro-B VI headphones.
Voxengo's software says that 8-, 16- and 24 bit output is dithered with a gaussian noise and a slight noise-shaping. I have no idea if it is the proper way to do it.
The shape of the dither is quite significant. I'm not sure a "slight" noise-shaping is sufficient. I respect the difficulty in obtaining these results. However, I am afraid that all they show is that this dithering algorithm is discernible from the high-resolution original. If I were to perform this test, I'd try it with foobar2000's dither.
Nonetheless, thank you for the results. As I've always considered 24/96 to be excessive, you've given me a point to ponder.
Voxengo's software says that 8-, 16- and 24 bit output is dithered with a gaussian noise and a slight noise-shaping. I have no idea if it is the proper way to do it.
The shape of the dither is quite significant. I'm not sure a "slight" noise-shaping is sufficient. I respect the difficulty in obtaining these results. However, I am afraid that all they show is that this dithering algorithm is discernible from the high-resolution original. If I were to perform this test, I'd try it with foobar2000's dither.
Nonetheless, thank you for the results. As I've always considered 24/96 to be excessive, you've given me a point to ponder.
I guess your listening conditions are pretty extreme and uncomfortable in terms of overall loudness, Pio, but roughly as loud as you'd need to have any chance of ABXing 16-bit from 24-bit. i.e. you need to have the programme peaks about as loud as a petrol (gasoline) chainsaw at arm's length without ear defenders (113-116 dBa is typical) to be able to come close to picking up quantization, dither noise or noise modulation at the LSB level without noise shaping while you listen to the fade-in. With optimal dither and strong ATH noise shaping, you probably need to near the pain threshold.
Gaussian pdf dither isn't optimal. Triangular dither -1 to +1 peak to peak (sum of two independent rectangular dithers of -0.5 to +0.5 peak to peak) is optimal for the first and second moments, which prevents quantization distortions and noise modulation - the only two effects (moments) of bad dither that are considered audible. Gaussian might be more audible (higher energy) while failing to completely eliminate both effects.
I'd agree that foobar2000's dither options, esp on older versions, should allow a better test of theoretically valid dither. Versions around 0.8 allowed:
- no dither (i.e. rounding/truncation quantization)
- flat dither (no noise shaping) - i.e. triangular pdf,
- dither with soft ATH noise shaping (less HF boost),
- dither with strong ATH noise shaping (recommended setting in fb2k).
I think the range of dither/shaping options has been reduced and simplified in the 0.9 versions of fb2k.
From calculations based on commonly-accepted figures for human audition (difficult for me to test in practice with my equipment and noise environment) I'd expect flat dither (triangular pdf) to be inaudible in comfortable listening but near the borderline of audibility/ABXability at extreme but bearable loudness.
I'd expect good valid dither with noise shaping (not noise shaping alone) to be practically inaudible unless you selectively turn up the fade-in to extreme levels and turn it down to avoid painfully loud peaks during the main programme material. This would probably remain the case for tracks with replay gain values as positive as +12 dB or even a little greater (i.e. 77 dB SPL calibrated, or 6 dB below the 83 dB SPL reference pink noise specified in the Replay Gain calibration spec and used in movie theatre loudness calibration). Such quietly-recorded tracks could typically be increased by about 12 to 15 dB without clipping, but their noise-shaped dither is probably about 15 dB or so below the level of spectrally-flat (triangular pdf) dither.
I'd be very interested to see disconfirming evidence (ABX) from you or anyone else and to know how the results vary with different dither and noise shaping options.
I got 10/10 in the ABX test, with closed headphones, listening to the initial fade-in.
There you go ! During the fade-in the noise level of the recording is probably too low to mask the 16-bit dither. IMHO the crucial condition for audibility is that the level of the added (dither) noise is relatively high compared to the noise level of the source. During digital silence (no modulation) or fade-ins/outs this condition is met, so with enough monitoring gain the noise difference can be easily made audible. The cumulative noise of the hall/studio, microphones and pre-amps is likely to be the most important source, usually with a gaussian spectrum. Most dithers use white noise, which has a slightly more aggressive character compared to gaussian, so predicting masking isn't very easy.
Pio, most classical recordings have "room tone" (acoustical noise) between the movements. Could you redo your test by using such a fragment without fades ? Note that there are recordings where this "silence" has been lowered somewhat in level so it doesn't necessarily have the same masking properties as the noise during the music.
Thank you for the precisions about dither. This is interesting.
I'm sorry, but I won't be able to go on with this test for the time being. The test was already very difficult with that kind of dither, and I'm currently very busy with a bunch of ABX tests on hifi components that have suddenly started to occur in France : http://www.homecinema-fr.com/forum/viewtop...&t=29898890 (http://www.homecinema-fr.com/forum/viewtopic.php?f=1034&t=29898890)
I fully recognize that, and want to thank you again for the results you've given us. Best of luck with your new tests!
There are real differences between 16 bit and 24 bit files, just as there are real differences between uncompressed and mp3. As with the differences between wav and mp3, 16 bit vs 24 bit is often difficult to identify by listening.
Test signals are one thing, but has anyone found any 24 bit music recording that can be successfully identified via ABX testing against a properly resample to 16 bit version of same?
I quickly scanned this thread and couldn't see a reference to the paper by Meyer and Moran in JAES Sept 2007 called "Audibility of CD-Standard A/DA/A Loop Inserted into High-Resolution Audio Playback" (http://www.aes.org/e-lib/browse.cfm?elib=14195). If it has been mentioned already I apologize.
They looked at the audibility of various programs played at higher bit-rates versus the same loop played through a standard CD (16 bit, 44.1 kHz) A/D, and could not find significant differences except when the noise floor of the 16-bit version was played at an elevated playback level.
This paper was so controversial when it was published it overloaded the AES editorial staff, which led to the start of an electronic forum on the site for people to air their beefs about AES papers.
Cheers
Sean
http://www.hydrogenaudio.org/forums/index....showtopic=57406 (http://www.hydrogenaudio.org/forums/index.php?showtopic=57406)
Greetings, found your forum recently, so here's my first post! Wanted to share another set of files that might be useful for listening tests. The source material is a 'raw' unprocessed master recording downsampled from 96kHz. Uploaded the files here:
Digital audio resolution test files (http://www.nordicaudiolabs.com/dart/)
Below you can see a quick ABX result I got using my laptop and a pair of Beyerdynamic DT250 headphones.
(http://www.nordicaudiolabs.com/samples/ABXreport.gif)
The second set of files are from an old master tape, comparing 24 bit versus 12 bit versus 320kbps mp3.
Martin
Below you can see a quick ABX result I got using my laptop and a pair of Beyerdynamic DT250 headphones.
That's an interesting result. Thanks for sharing and welcome to the forum!
I've quickly checked the audio files:
The music peaks at -2.6dBFS which means it's not a low-level signal that makes it easier to hear noise floor.
AFAICT the difference between the 24 and 16 bit version is just (dither) noise.
It's a bit surprising that this difference can be ABX'ed, but your test results look valid.
How would you describe the difference that you've heard ? Is it just the noise level ?
Can anyone confirm that Foobar treats 24 and 16 bit files the same ? It might have been a good idea to convert the 16 bit file to 24 by padding the last 8 bits with zeros to exclude any additional variables.
I can't ABX it. I agree it's a perfectly good 24>16 bit conversion.
The noise floor of the original recording comes quite close to the noise floor of the 16-bit conversion during the quiet points - watched on a spectral display.
I use this method to compare noise floors in Cool Edit Pro...
Open 24.wav
Open 16-bpt.wav
Channel Mixer - both channels = right (run on both files)
In 16-bpt.wav, select the left channel only, hit copy
In 24.wav, select the left channel only, click mix paste, select invert, 100.
Analyise: frequency analysis. 1024 fft. Linear view.
Now click play, and watch the dancing display. You can try any FFT size up to 4096 to get a dancing time/frequency display with varying time/frequency resolution. The lower curve is the 16-bit noise floor, the upper curve is the signal.
Note that, when playing, there's some averaging going on. Click stop, and put the cursor on a quiet bit of the music to see the exact FFT for that moment.
I tried running this through the last MATLAB version of lossyFLAC / lossyWAV. It doesn't have the refinements added by Nick, so it's not as trustworthy as his latest version, but it's an interesting indicator. I've plotted the number of bits it wanted to keep, and attached the graph.
As you can see, it often wanted to keep 17 or 18 bits during the quiet passages, if adding triangular dither without noise shaping. I.e. lossyWAV thinks it needs more than 16-bits to guarantee transparency.
If adding no dither (and remember lossyWAV aims to work below the noise floor, so no dither should be necessary - it's never added harmonic distortion IME), you can get away with 1 fewer bit than shown on that graph.
With Nick's refinements, and a shaped dither spectrum, it's quite possible that lossyWAV would decide 16-bits is sufficient.
Verifying the ABX
It would be helpful if Martin Kantola could describe the differences he heard, where in the file he heard them, and most importantly can verify that his laptop plays back 16-bit and 24-bit digital audio correctly.
The most important test is to check that 24-bit files are replayed correctly - if everything is truncated to 16-bits, then you are not comparing what you think! Replaying a -96dB 1kHz tone in a 24-bit file, examining the output using a high quality sound card or audio analyser, and looking for harmonic distortion, will verify true 24-bit capability, or a lack of it.
The next most important test is to verify that 16-bits padded to 24-bits with 8 zeros is still ABXable.
Cheers,
David.
I ‘m going to see if I can borrow a fast connection to download, those files are just too big for this dial up.
As I specified before, only the 24 bit source is of interest. I know I can convert properly; possible audible differences with other conversions to 16 bit are of no value. Tests based on them on very suspect. I understand most anyone can do a proper conversion, given the necessary software, but a fuller specification of the particular process employed is desirable.
Blue printing on a black background is extremely difficult to read. Under most circumstances that kind of presentation is enough to make me very quickly decide the site is not worth the bother. I still haven’t figured out all the headings on this one.
I know I can convert properly; possible audible differences with other conversions to 16 bit are of no value. Tests based on them on very suspect. I understand most anyone can do a proper conversion, given the necessary software, but a fuller specification of the particular process employed is desirable.
What exactly do you need to know apart from the dither used? Please let me know. While I understand that this is slightly controversial, we have to remember that most professional audio is done in 24 bit today, probably for good reason. So we could at least suspect an audible difference. But it's a good idea to download only the source file to save download time.
Blue printing on a black background is extremely difficult to read.
Oh, I'm sorry, looked fine on my screen, but fixed it now! Please reload the page. Thanks for pointing that out.
Martin
It would be helpful if Martin Kantola could describe the differences he heard, where in the file he heard them, and most importantly can verify that his laptop plays back 16-bit and 24-bit digital audio correctly.
While the hardware of my laptop is 24bit/192kHz capable, the specs are not exactly impressive:
(http://www.nordicaudiolabs.com/samples/ALC260data.gif)
Will get back to the differences I heard, have to try a 16 bit padded to 24 bits to make sure it works as it should.
Martin
While I understand that this is slightly controversial, we have to remember that most professional audio is done in 24 bit today, probably for good reason. So we could at least suspect an audible difference.
The primary reason we use 24-bit in production is due to the fact that 24-bit words allow for greater resolution of low-amplitude audio. In production, it's not rare to track or otherwise record a source at very conservative levels where the majority of the audio may only be effectively represented with 12 to 16 bits per sample so as to prevent full-scale clipping of rogue transients. If we were to record at 16 bits per sample, such audio would be only effectively represented with 6 to 8 bits per sample. This is especially true with certain types of recording, such as recording Foley, where often times there may be only one or at least very few opportunities to record a given sound, thus recording at conservative levels is critically important.
When audio is compressed, limited or normalized, we must necessarily bring up the noise floor to obtain an increase in volume (a decrease in dynamic range), so longer word lengths are extremely advantageous in this regard. Since most types of audio, music especially, end up passing through several layers of compression and limiting, working at or above 24-bit is almost mandatory (in my opinion) to achieve the highest quality result.
As an engineer, I never expect any audible difference between 16-bit and 24-bit for a compressed master. For a final delivery, where no additional processing is expected to be applied, 16-bit should almost always suffice.
That being said, I'll have a listen to the samples you've provided. The odds I'll be able to detect a difference is probably fairly slim, but I'll at least give it a go.
EDIT: Ran an ABX test. I didn't have the slightest chance, even listening at roughly 10dB higher than I would be comfortable listening to music at for any extended period of time. I focused my attention on the quietest passage, which begins at about 9.5 seconds and dives down to an RMS of about -41dBFS for about a second and a half. Absolutely nothing seems to stand out.
As an engineer, I never expect any audible difference between 16-bit and 24-bit for a compressed master.
Thanks for the post, good explanation, I understand and agree with you. Please note that this is definitely not about a compressed master, but about our ability to hear differences between 16 and 24 bits in the first place. Hopefully also about what these possibly detected differences are from a musical perspective, because noise is hardly the issue. I should have explained myself better. Thanks for checking out the files.
Martin
18/32 - there's a harshness in the left channel around the 8.5 mark, which I had originally mistaken for ground hum, which I felt I could possibly distinguish in character between 16hpt and 24. I started pretty good at 5/6 but I felt like I could not hold on to the difference.
There seems to be a tremendous amount of distortion in this recording when I listen to it critically
While the hardware of my laptop is 24bit/192kHz capable, the specs are not exactly impressive:
I think your assumption that there's a
likelihood of hearing a difference is very wide of the mark. There's a possibility, yes - but there's a stronger possibility that the sound from your laptop, like most other laptops, isn't all it's cracked up to be.
I can't find full RMAA results for what you have, but that codec on another board, tested at 16-bits, wasn't exactly spectacular...
http://www.hardwareluxx.de/Images/Alderwoo...ar%20output.htm (http://www.hardwareluxx.de/Images/Alderwood/Realtek%20HD%20Audio%20rear%20output.htm)
The first thing to demonstrate is that you're getting 16-bits out
without distortion, and 24-bits out
without distortion (at least down to ~ 18-20 bits). Having a noise floor substantially below the 18th-20th bit would help too.
That test lists 70dB SNR which is nominally (though not accurately) equal to about 12 bits.
Cheers,
David.
I think your assumption that there's a likelihood of hearing a difference is very wide of the mark. There's a possibility, yes - but there's a stronger possibility that the sound from your laptop, like most other laptops, isn't all it's cracked up to be.
Thank you for looking into this. What I'm doing is putting that 'assumption' of mine to the test with ABX. While I agree that my laptop sound is not anywhere near 24-bit performance, we can't ignore that there are several types of both distortion and noise. It's been pointed out that speaker distortion is very high in level, yet we can detect distortion of other kinds listening through one. FWIW, there are two fellow forumers at Gearslutz who scored a full 10/10 on this test, but they probably had much better equipment. You should know that I found testing with my laptop quite a challenge. It was anything but easy, and required a lot of focusing.
Martin
Found it...
http://www.gearslutz.com/board/so-much-gea...ot-audible.html (http://www.gearslutz.com/board/so-much-gear-so-little-time/358301-24-vs-16-bit-not-audible.html)
...interesting thread. Some amazing miscomprehension in there, but still interesting.
I'm not saying you can't hear it. My own algorithm (lossyWAV) says you can. (But I can't!). It's just that 90%+ of people who come to HA claiming to hear a difference between 16-bits and 24-bits, or 44.1kHz and 96kHz, are really hearing problems with their equipment.
Potential equipment "problems" aside, would you be willing to try another ABX or two?
Cheers,
David.
Here you go...
http://rapidshare.de/files/43321357/16-hpt-24-n.wav.html (http://rapidshare.de/files/43321357/16-hpt-24-n.wav.html)
Your 16 bit file, converted back to 24-bits, with a little high-pass noise added to make sure the bottom 8 bits are moving.
Is it better or worse than your 16-bit version?
(where better = closer to the 24-bit original!)
Cheers,
David.
My own algorithm (lossyWAV) says you can. (But I can't!).
Have always thought that if you can hear a difference (as long as there
is one) depends on a number of factors:
1. the source material
2. the equipment and room
3. you critical listening skill or experience
4. your ability to find and focus on the artifacts (this can change from day to day)
In other words, by having an excellent playback system and good acoustics in a relaxed environment, and possibly by discussing the differences, we can most likely maximize the number of people that detect a difference.
For me, the point it not if most people hear it or not. This is all about how sensitive our hearing is or can be, and how we can keep improving the sonic experience. To some extent questioning the quality of both our hearing mechanism and digital audio from a musical perspective. I hear a lot of claims that digital is so linear and accurate while our ears are so non-linear and inaccurate, but isn't that a simplified view and an attempt to compare two things of completely different nature?
Just because mp3s are today good enough for most people and applications, we don't have to give up our quest for better sound. Look at this way, the day we can record
any musical performance and play it back with such quality that the playback cannot be distinguished from the original performance by anyone, then we can rest... Not saying that's what music recording should be about though...
Martin
Found it...
http://www.gearslutz.com/board/so-much-gea...ot-audible.html (http://www.gearslutz.com/board/so-much-gear-so-little-time/358301-24-vs-16-bit-not-audible.html)
...interesting thread. Some amazing miscomprehension in there, but still interesting.
I'm not saying you can't hear it. My own algorithm (lossyWAV) says you can. (But I can't!). It's just that 90%+ of people who come to HA claiming to hear a difference between 16-bits and 24-bits, or 44.1kHz and 96kHz, are really hearing problems with their equipment.
Potential equipment "problems" aside, would you be willing to try another ABX or two?
Cheers,
David.
Let me get this straight... comparisons were 24bit vs dithered-to-16, and another was 24-bit .wav , converted to 16 vs 320 mp3 made directly from the 24 bit .wav ...and at least TWO people in this tiny sample of subjects managed
perfect or near-perfect ABX scores on
both -- with the mp3 trial reported being HARDER to ABX than the wordlength comparison??? (Except for the THIRD guy who claims he aced the 320 lossy vs lossless ABX, while failing to score significantly on the wordlength ABX). And another guy got 10/10 on just the wordlength ABX?
But Martin K
can't successfully ABX two files made using Chris Johnson's superspecial dither?
I'm not implying nefarious intent on Martin K.s part, but to me these data suggest an obvious control: someone here needs to replicate the sample pairs (take the original 24-bit .wav and convert to 16 bit with TPDF dither; take the 24-bit file and convert it to 320 mp3), and see how those guys 'over there' do at ABX'ing them.
But Martin K can't successfully ABX two files made using Chris Johnson's superspecial dither?
My result for the "avd" dither was 8/10, as posted in the thread. And yes, I really want to see more people trying this too! Regardless of outcome. But so far many many posts, but very few ABX results.
You are welcome scrutinize my conversions, I'd be happy to correct any mistakes I might have made, but these are very basic operations on a DAW, honestly think I can handle them.
Martin
Here you go...
http://rapidshare.de/files/43321357/16-hpt-24-n.wav.html (http://rapidshare.de/files/43321357/16-hpt-24-n.wav.html)
Your 16 bit file, converted back to 24-bits, with a little high-pass noise added to make sure the bottom 8 bits are moving.
Is it better or worse than your 16-bit version?
(where better = closer to the 24-bit original!)
Hi David,
took a different approach with these two, hopefully you will find it useful.
Not exactly sure what that 'little high-pass noise' you added would do, so I just started out listening and guessing. This was my one and only attempt at the files, and hopefully you can see how the guessing slowly gets better in the log file. Tried not to think too much, just go by feel and keep going (with some breaks in between). By the time I had reached 100 attempts, the probability of guessing was down to 1% according to the computer.
This time it was much more random as I didn't have the energy to focus properly. It's one thing to stay sharp for 10 attempts, a different story with 100. Even after I found something to pay attention to, my concentration failed and I made many mistakes. It is very hard to do this on the laptop.
(http://www.nordicaudiolabs.com/samples/ABX100.gif)
Here's the complete log:
foo_abx 1.3.3 report
foobar2000 v0.9.6.1
2009/01/23 22:20:10
File A: C:\Documents and Settings\Valued Customer\Desktop\TEST\samples\16-hpt-24-n.wav
File B: C:\Documents and Settings\Valued Customer\Desktop\TEST\samples\24.wav
22:20:10 : Test started.
22:37:01 : 01/01 50.0%
22:37:45 : 01/02 75.0%
22:38:28 : 02/03 50.0%
22:39:19 : 03/04 31.3%
22:40:15 : 04/05 18.8%
22:42:27 : 05/06 10.9%
22:47:28 : 05/07 22.7%
22:48:01 : 06/08 14.5%
22:50:24 : 06/09 25.4%
22:50:51 : 07/10 17.2%
22:51:23 : 07/11 27.4%
22:58:36 : 07/12 38.7%
22:59:43 : 07/13 50.0%
23:03:40 : 08/14 39.5%
23:03:59 : 08/15 50.0%
23:04:17 : 08/16 59.8%
23:04:25 : 08/17 68.5%
23:04:39 : 09/18 59.3%
23:04:53 : 09/19 67.6%
23:05:04 : 10/20 58.8%
23:05:28 : 10/21 66.8%
23:05:46 : 10/22 73.8%
23:06:45 : 11/23 66.1%
23:07:01 : 12/24 58.1%
23:07:13 : 13/25 50.0%
23:07:46 : 14/26 42.3%
23:07:59 : 14/27 50.0%
23:08:19 : 14/28 57.5%
23:09:20 : 14/29 64.4%
23:09:33 : 15/30 57.2%
23:11:16 : 16/31 50.0%
23:11:30 : 16/32 57.0%
23:32:28 : 17/33 50.0%
23:32:51 : 18/34 43.2%
23:33:21 : 19/35 36.8%
23:33:45 : 20/36 30.9%
23:34:27 : 21/37 25.6%
23:35:06 : 21/38 31.4%
23:35:35 : 22/39 26.1%
23:35:46 : 22/40 31.8%
23:36:04 : 22/41 37.8%
23:36:16 : 23/42 32.2%
23:36:25 : 24/43 27.1%
23:36:41 : 25/44 22.6%
23:37:59 : 26/45 18.6%
23:38:33 : 27/46 15.1%
23:38:40 : 28/47 12.1%
23:38:50 : 29/48 9.7%
23:40:29 : 30/49 7.6%
23:41:12 : 31/50 5.9%
23:42:28 : 31/51 8.0%
23:42:51 : 32/52 6.3%
23:43:03 : 33/53 4.9%
23:43:56 : 34/54 3.8%
23:46:56 : 35/55 2.9%
23:48:04 : 36/56 2.2%
23:48:59 : 36/57 3.1%
23:49:54 : 37/58 2.4%
23:50:23 : 37/59 3.4%
23:50:53 : 38/60 2.6%
23:51:51 : 38/61 3.6%
23:52:01 : 39/62 2.8%
23:53:02 : 39/63 3.8%
23:54:50 : 40/64 3.0%
23:55:01 : 40/65 4.1%
23:55:31 : 41/66 3.2%
23:55:42 : 42/67 2.5%
23:58:10 : 43/68 1.9%
00:00:02 : 43/69 2.7%
00:00:53 : 43/70 3.6%
00:01:34 : 44/71 2.8%
00:02:14 : 44/72 3.8%
00:02:32 : 45/73 3.0%
00:02:41 : 45/74 4.0%
00:02:53 : 46/75 3.2%
00:03:16 : 47/76 2.5%
00:03:32 : 48/77 2.0%
00:03:54 : 49/78 1.5%
00:05:15 : 50/79 1.2%
00:06:28 : 50/80 1.6%
00:06:38 : 51/81 1.3%
00:07:45 : 51/82 1.8%
00:08:06 : 52/83 1.4%
00:08:36 : 52/84 1.9%
00:10:12 : 52/85 2.5%
00:10:23 : 52/86 3.3%
00:10:50 : 53/87 2.7%
00:11:06 : 54/88 2.1%
00:11:25 : 54/89 2.8%
00:11:57 : 55/90 2.2%
00:12:35 : 56/91 1.8%
00:13:14 : 56/92 2.4%
00:13:40 : 57/93 1.9%
00:14:03 : 58/94 1.5%
00:14:51 : 59/95 1.2%
00:15:14 : 59/96 1.6%
00:15:51 : 60/97 1.2%
00:16:29 : 60/98 1.7%
00:16:41 : 61/99 1.3%
00:17:31 : 62/100 1.0%
00:18:38 : Test finished.
----------
Total: 62/100 (1.0%)
Martin
Thank you Martin.
Putting noise in the last 8 bits actually removes most of the potential "equipment" issues, even though the resulting file is mathematically even further from the 24-bit original than the pure 16-bit version.
So I now think it's very likely that you're hearing a genuine difference, rather than an equipment artefact (unless it's quite an exotic one).
I'm not implying nefarious intent on Martin K.s part, but to me these data suggest an obvious control: someone here needs to replicate the sample pairs (take the original 24-bit .wav and convert to 16 bit with TPDF dither; take the 24-bit file and convert it to 320 mp3), and see how those guys 'over there' do at ABX'ing them.
I haven't even looked at the 320kbps mp3, but the 16-bit from 24-bit version seems fine - it's exactly what it says it is, no more, no less.
I agree we need to do what you suggest: try another known 24>16 conversion.
However, like I said, the noise floor in that 24-bit file does dip very close to the theoretical limit of 16-bits - such that, in some blocks, lossyWAV wants to keep 18 bits of the 24-bit original (if converting with triangular dither).
Cheers,
David.
But Martin K can't successfully ABX two files made using Chris Johnson's superspecial dither?
My result for the "avd" dither was 8/10, as posted in the thread. And yes, I really want to see more people trying this too! Regardless of outcome. But so far many many posts, but very few ABX results.
An 8/10 result does not support the hypothesis of 'audible difference' at a p<0.05. You guys should keep the 'bino_dist' chart on this page handy:
http://www.kikeg.arrakis.es/winabx/ (http://www.kikeg.arrakis.es/winabx/)
An 8/10 result does not support the hypothesis of 'audible difference' at a p<0.05.
No, it means a 5.5% probability I'm only guessing as the software already tells me. While I could try again to see if the number goes up or down, I'd prefer to see somebody else take the test. Not only because it's such hard work but to get a better idea of how audible it is or isn't.
Martin
Remember those stats are only true for the situation where you decide on (and fix) the number of trials before hand.
Sitting there, testing, and stopping when you feel like it (e.g when you get a good result), breaks the statistics.
Taking Multiple ABX tests, and only reporting one, or making false starts and re-starting, both also break the statistics.
This isn't what you've done here, but these things have been hinted at by others. They're complete no-no's.
Cheers,
David.
Thanks for the info David, I'm completely new to ABX but thanks to the available software it was easy to try it out. As for the procedure, what I did with the other samples was to take plenty of time to study the material first, and then take a test with exactly ten trials. The recent one I did differently, plunged straight into it and did 100 trials. Don't know if these methods are correct, but the importance of being able to find and focus on the difference is essential in audio if you ask me.
Anyway, the OT was about samples that work for testing 16 vs 24 bits. Might add some more material to the same little webpage (http://www.nordicaudiolabs.com/dart/) at some point. Would be helpful to hear comments on how well the samples work for some of you guys.
Martin
My result for the "avd" dither was 8/10, as posted in the thread. And yes, I really want to see more people trying this too! Regardless of outcome. But so far many many posts, but very few ABX results.
You are welcome scrutinize my conversions, I'd be happy to correct any mistakes I might have made, but these are very basic operations on a DAW, honestly think I can handle them.
Martin
Martin, congrats for posting the test files, and for the persistence in doing the ABX trials; particularly the 100 trials of the version with 'a little high pass noise'!
I have not tried to ABX the 24bit original against the 16 bit dithered versions (with or without extra noise) because in the past I have found that 24 bit music dithered to 16 bits (with a noise shaped dither) sounds the same to me as the original 24 bits, at normal recording and playback levels.
The effort required for ABX testing can be considerable. In my own case, my ears/brain are most responsive the first time I am presented an AB comparison. If the passage is repeated, my hearing discrimination rapidly diminishes. So to commit myself to an ABX comparison of very similar files I have to set aside a fair stretch of time, to allow my hearing to regain its sensitivity to fine differences.
Before attempting a serious ABX I would like to be assured that the 16 bit dithered version has been derived using methodolgy to the satisfaction of HA Forum technical gurus, e.g. a dither conversion they have run themselves, and preferably that at least one person has successfully ABXd such an 'approved' dithered version. Perhaps that is laziness on my part!
No disrespect intended to yourself, but I can foresee reluctance of the HA community to accept any ABX test result unless all possible perceived doubt has been removed as to the reliability of the methodolgy for performing the dither.
[Out of interest, I did ABX tests of just 4 trials of the truncated version, and 4 trials of the 320kbps mp3 version, relative to the respective 24 bit originals. I found the truncated version
relatively easy to distinguish. But the 320kpbs version needed a lot of concentration and 'refresher breaks'. The 4 correct answers I obtained in both cases merely corresponded to a 6.2% likelihood of not guessing. But all I was seeking was broad assurance I could distinguish the files. They were not what I'd call 'serious' ABX tests.]
Cheers
The effort required for ABX testing can be considerable. In my own case, my ears/brain are most responsive the first time I am presented an AB comparison. If the passage is repeated, my hearing discrimination rapidly diminishes. So to commit myself to an ABX comparison of very similar files I have to set aside a fair stretch of time, to allow my hearing to regain its sensitivity to fine differences.
Is it the repetition of A/B comparison itself, or the repetition of the same music in a series of A/B comparisons, that you find fatigueing?
If it's the latter, I wonder if it might make discrimination easier, to prepare a batch of 16 different samples -- and use a different one for each of 16 trials. Then at least each A/B is 'fresh' to the ears.
For ease of preparation, the samples could be 16 different parts of the same track. If that's not a concern, they could be from different tracks (all prepared and vetted to HA standards as you have described)
Is it the repetition of A/B comparison itself, or the repetition of the same music in a series of A/B comparisons, that you find fatigueing?
Having not tried the approach you suggest of setting up multiple different parts of the same track in advance I am not absolutely sure.
However I do find that once I have become desensitized by listening to one segment of a track too many times (this may be just 4 times), the use of a different segment only offers a temporary period of being able to discriminate. After that I may find I cannot discriminate any segments of any tracks! (unless of course a segment contains a much more marked difference).
Another factor is that I have found with experience that I cannot just rely on my memory of A and B when listening to X and Y. For example, at the beginning of a listening session I may perceive a lossy Dolby Digital version as somewhat thin -- not as full sounding as an uncompressed version. But later in the session the Dolby Digital may start to sound clearer than the uncompressed version (Dolby can have a "neat and tidy" sound), and the uncompressed version may sound thin in comparison.
So before commiting to a proposed answer I find now I have to take the precaution of replaying the A and B to make sure my hearing has not altered. This adds to the listening load per entered ABX response.
In the past, I found that on some occasions I would enter up to 5 incorrect answers in a row. Quite frustrating! This I eventually realized was because I was relying on my memory of the difference in sound of A and B (e.g. A is the "brighter" sound). During the course of the ABX trial, the fatigue had caused a juxtaposition in the relative perceived sound of A and B.
In the case of 'brightness", I could conjecture that what initially sounded "brighter" may have been brighter because of the presence of high frequency content towards the upper limit of my hearing. As time wore on, and with my ears being exposed to a great deal of high frequency content, my hearing in that range may have become desensitized, so that the character of "brightness" was lost.
These sorts of extraneous effects might normally go completely unnoticed. It's just the samples we try to ABX are sometimes very close to being indistinguishable, so that even minor changes in our hearing as a session progresses have the potential to lead us into error.
This was asked more than once but there was no response. Usually, when there is some small, generally hard to detect difference between audio samples, it comes down to something specific, some nameable artifact or distortion, or whatever, at one or several points. For instance, a particular mp3 encoding may sound identical to the original except when a triangle is struck (or whatever) -- the rest of the track is not differentiable.
Is Martin’s claim for a difference here based solely on ABX results? Is there nothing that is consciously identifiable? My fairly casual 24 vs 16 comparisons have noticed nothing so far. As Martin observes, doing much ABX testing is a lot of work. Having a target might make it more tolerable. I don’t see that as invalidating the test, since it does not allow any way around the fact that I must still chose based only on the sound, regardless that I may be waiting for a particular sound, no?
I note that Axon identified some “harshness” at “the 8.5 mark.” does this mean at 8.5 seconds from the beginning? What does the “18/32 -“ at the beginning of that line signify?
~26,000 views... and still a mystery. ;>
18/32 - there's a harshness in the left channel around the 8.5 mark, which I had originally mistaken for ground hum, which I felt I could possibly distinguish in character between 16hpt and 24. I started pretty good at 5/6 but I felt like I could not hold on to the difference.
AndyH, I took the above to mean that Axon initially used an apparent difference 8.5 seconds into the track as a basis for entering his ABX responses; and that he got 5 out of 6 correct for the first 6 entered responses relying on that technique. However his concentration/hearing then began to fail him and he ended up with only 18 out of 32 correct after giving the 32nd response.
This is reminiscent of my hearing fatigue issue. But, in my case, if I find I can no longer hear a difference I stop and wait for my hearing/concentration to recover.
I note we are still waiting for Martin to tell us the subjective difference he heard when he was doing his 100 listening trials, with 62 correct answers. That's assuming the difference had a definite "quality" he perceived.
I myself find music at 24 bits sound more "liquid" than the same music file
truncated to 16 bits. But if it is properly dithered down to 16 bits, I cannot perceive a difference when 24-bit music is changed to 16 bits.
Has anyone else heard a difference at the 8.5 second mark? Does this difference disappear if another form of dither is used? (I don't have access to playback equipment at the moment to check out the sound at the 8.5 second mark myself.)
~26,000 views... and still a mystery. ;>
It seems no-one can actually supply a music sample (recorded at a normal rather than very quiet level) that demonstrably sounds different when dithered down to 16 bits. Martin has a music sample that might possibly qualify, if his own dithering and ABXing are to be accepted at face value, but he has not responded to AndyH's query regarding what the difference sounded like. And there is a dearth of ABX submissions from other HydrogenAudio members, though Axon as made a verbal report. I can only presume a few people listened to Martin's samples and heard no difference, but have not bothered to report this. I personally regard the technical issue as important, as many Bluray discs are being released with lossless tracks at a 16-bit depth (dithered), rather than a 24-bit version of the master. Some people are prepared to pay a premium for a Blu-ray disk that provides 24-bit sound. They may actually shun a Bluray that only has 16-bit depth audio. Are such people suffering under a misapprehension?
Hi guys,
sorry for the late reply, been very busy for a while with a Critical Listening class...
Anyway, you wanted to know what I heard in David's dithered 16-bit version. Once again I found myself focusing on the tone of the brass, the 16 bit was thinner, had less body or less punch if you like. Remember that I was listening on my crappy laptop setup with very low level on the phones, so had to listen to changes in timbre and not in low level detail or similar. As a matter of fact, I tried turning down the level even more and it possibly made it even easier to hear which one was which. If you're interested, with the other "special" dither I didn't hear this effect, but another one instead which seemed to affect the stereo perspective.
Have no explanation to why I hear this. It does not make any sense, but that's exactly why I find it an interesting subject worth studying further. If lower bit depths have an effect on the musical tone itself (and not some tiny technical loss of quality) we should perhaps be worried. Especially if we're stuck with dithered 16 bit as the best delivery format ever.
Mentioned earlier that several factors are important when listening. To me, these samples work well but the equipment was far from optimal. If you can't hear a difference easily, it might also have something to do with my ears, I work with audio full time and have done so for the last twenty years, and I believe that our hearing can improve with training. If you try listening, a straight signal path without any signal processing is essential, my noise canceling headphones were useless.
Martin
Anyway, you wanted to know what I heard in David's dithered 16-bit version...
Just to clarify, "my" version, here...
http://www.hydrogenaudio.org/forums/index....mp;#entry610883 (http://www.hydrogenaudio.org/forums/index.php?showtopic=49843&st=75&p=610883&#entry610883)
...is "Your 16 bit file, converted back to 24-bits, with a little high-pass noise added to make sure the bottom 8 bits are moving".
This isn't "my" attempt to dither it (IMO there's nothing wrong with your dithering, but no one has provided an alternative rendition here) - it's simply my attempt to remove 16bit vs 24bit sound card peculiarities by filling the bottom 8 bits with something relatively benign.
What I provided is further away from the original than even your 16-bit version - as such, it should be easier to hear a difference, unless the difference is down to the sound card's handling of 16-bit vs 24-bit material.
I have to admit, I've even had a little Google this morning to check there's no organised hoax against HA - we've got Martin Kantola ABXing the inaudible in this thread, and BORK ABXing the inaudible in the lossyWAV thread.
If other people can confirm, it's quite exciting, because it takes ABXing to another level.
Cheers,
David.
P.S. I ran the track 24.wav through the latest version of lossyWAV (1.1.2d).
In insane mode, it quite often wanted to keep 19 bits.
In standard mode, it quite often wanted to keep 17-18 bits.
(I'm judging this by the amount of noise it adds, not by reading the bits_to_remove values directly, so it's not 100% reliable).
Cheers,
David.
David’s first remarks about his lossyWAV results made me curious about what I could find by a different route. The results were interesting, in an academic sort of way, but I’m not certain how to interpret them. To report seemed like a lot of writing, maybe with no purpose, so I moved on to other projects. Then David’s second post on the subject decide me to try to write up my experiment and seek comment.
Since this concept is important to understand in order to follow my process and results, I first describe the following, just in case there are any readers not already familiar with it.
If one takes any on-computer audio file, it is easy to make an exact copy. If one then inverts one of the two identical files (easy done in most any audio editor) and mixes it with the other (not inverted), aligning sample for sample, the two completely cancel each other. The result is silence, digital zeros.
If one takes any two different audio files and mixes them together, one has a combination, such as a vocal and backing track now in one file. If one then takes one of the two, inverts it, and mixes it with the combination, as in the above paragraph, one is left with the original of the other file -- almost.
In the first mixing, each resulting sample is the sum of the corresponding samples from the two inputs. A quantization error occurs in every case where the sum of the two exceeds the precision the format can carry. This will occur even when working in 32 bit float.
So, if one then inverts and mixes in the original of the second file to cancel it, one does not get digital zeros, one gets the quantization error values. In a sample I tried, working in 32 bit float, the result had a peak amplitude of -144dB and an average RMS amplitude of -155dB. Silence in any practical sense, but not complete.
To start the following, I opened the provided 24 bit, 48kHz file into 32 bit float, the normal format for CoolEdit. This changes no values, it only provides greater working precision. In all cases, unless otherwise specified, a 16 bit result was converted back to 32 bit in order to compare it against any 32 bit originals. Again, re-converting to 32 bit changes no values that were in the 16 bit version, everything below 16 bits is just zeros.
A word about dither. Comments are welcome. CoolEdit provides three parameters for dither. I don’t recall that any others are possible. One is Dither Depth, the amount of dither. Another is p.d.f. or dither type (such as rectangular, triangular, etc.). The third is noise shaping. My program version has 12 choices for noise shaping.
Dither depth is the most important of these parameters. Too little and quantization distortion still exists. Too much and unnecessary noise is present in the final result. The dither type effects the results, such as the potential for modulation, and noise shaping is mainly important for distributing the added noise for least audible impact.
As long as the dither depth is good, one can leave out noise shaping, and use any p.d.f., to get very good, if not necessarily optimum, results. While there are many favorite dithering variations, they all, I believe, come down to very difficult to notice differences that are unimportant in normal listening.
dithered silence
If one generates a file of silence, digital zeros, in float format, one can then convert that to 16 bit. The result will still be silence; there are no values except zero, so no quantization errors occur.
One can apply dither to the 32 bit silence and then convert to 16 bits. The resultant will contain only the noise that gets added to any music file one converts in the same way. This consists of the dither noise and the quantization noise from converting it to 16 bit.
Once upon a time, by experimenting with various combinations, I selected the dither settings that I use with all my music. It results in the least amount of added noise (that works) and is completely inaudible to me on my equipment. The “that works” part means that less dither does not eliminate all the quantization distortion. This is very hard to determine with music, but fairly easy using pure test tones.
Dither depth 0.5 bits, p.d.f. Shaped Triangular, noise shaping C1
Now finally, the real gist of the report. First I converted the provide sample of 24/48 (now 32/48) music to 16 bit using my normal dither, then converted it back to 32 bit. This, you recall, leaves everything below 16 bits as zeros. I then inverted and mixed it with the original to provide a difference.
I realize that this process is considered useless with perceptual encodings. The good reason for this view is that the important aspects of perceptual encodings are based on human hearing, not on any data differences. Data differences might be large but irrelevant. In this case, while I am not sure about how to interpret the results (thus this lengthy report, for critique and comment), I am not sure it is so reasonable to lump them with the perceptual encoding viewpoint.
dithered conversion
The result of these three steps is essentially dithered silence, as described above. With the tools CoolEdit provides, I cannot tell any difference between this and the dithered silence results in Spectral View, with the Frequency Analysis graph, or by listening. There is some difference in Statistics (peak amplitude), but I can’t say if it is due to anything other than quantization noise. The columns are left and right channels.
dithered silence music
Peak Amplitude: -78.27 dB -78.27 dB -75.42 dB -75.55 dB
Minimum RMS : -88.42 dB -88.45 dB -89.13 dB -89.14 dB
Maximum RMS : -87.24 dB -87.29 dB -87.36 dB -87.45 dB
Average RMS : -87.87 dB -87.86 dB -88.22 dB -88.22 dB
Total RMS : -87.87 dB -87.85 dB -88.22 dB -88.21 dB
non-dithered conversion
Repeating the procedure, but without dithering, gives these lower statistics. Spectral View show a very even noise distribution, except a slightly greater intensity at very low frequencies. Frequency Analysis show a linear rise of 20dB below 140Hz to the lower limit of frequency resolution, otherwise a flat line over the audio spectrum. Although at a considerably lower level than the dithered version final result, this is quite audible-- at the highest output of my headphone amplifier.
Peak Amplitude: -96.36 dB -96.36 dB
Minimum RMS : -101.49 dB -101.42 dB
Maximum RMS : -100.84 dB -100.83 dB
Average RMS : -101.11 dB -101.11 dB
Total RMS : -101.11 dB -101.11 dB
Since CoolEdit provides no control point between applying dither and converting to 16 bits, I created my own dither file by generating a file of low intensity white noise in 32 bit float. Its statistics are
Peak Amplitude: -80.14 dB -80.14 dB
Minimum RMS : -89.88 dB -89.88 dB
Maximum RMS : -89.09 dB -89.09 dB
Average RMS : -89.49 dB -89.49 dB
Total RMS : -89.49 dB -89.49 dB
Its Frequency Analysis graph is a flat line across the audio spectrum.
As proof of concept, I tested with a created file consisting of 10 seconds of a 3kHz tone followed by 10 seconds of digital silence. For the 3kHz part:
Mono
Peak Amplitude: -78 dB
Minimum RMS : -81.02 dB
Maximum RMS : -81.02 dB
Average RMS : -81.02 dB
Total RMS : -81.02 dB
The 3kHz tone was chosen to be easy to distinguish the result of dithering vs non-dithering. Non-dithered conversion to 16 bit produces ample distortion, easily viewed and measured. Then, starting over,
(1) Mix same 32 bit audio file with 32 bit white noise file, producing dithered audio.
(2) Convert mixed file to 16 bit, no dithering. Produces non-distorted 16 bit.
(3) Convert back to 32 bit. Nothing below 16 bit but zeros.
(4) Convert white noise file 16 bit, no dithering.
(5) Convert 16 bit white noise back to 32 bit. As always, this leave zeros below 16 bits.
(6) Invert (5) and mix with (3).
The second half of the resulting file, which was originally silence, is again complete silence. The first half of the file, the tone, is now the tone plus quantization noise.
(7) Invert the original file and mix with (6). The second half of the file is still zeros, the first half is only the quantization noise, and whatever was below 16 bits in the original tone. This is seen to be essentially flat across the audio spectrum, with a slight rise at the lowest frequencies.
Peak Amplitude: -90.31 dB
Minimum RMS : -98.97 dB
Maximum RMS : -98.23 dB
Average RMS : -98.58 dB
Total RMS : -98.58 dB
Having demonstrated the feasibility, I did the same thing with the provided audio sample (now in 32 bit format for convenience). Call the original audio A and the generated white noise B.
(1) Mix A and B
(2) Convert mixture (1) to 16 bit, no dithering.
(3) Convert (2) to 32 bit.
(4) Convert B to 16 bit, no dithering.
(5) Convert (4) to 32 bit.
(6) Invert (5).
(7) Mix (5) with (3), eliminating dither of mixed A + B.
(8) Invert (7).
(9) Mix (7) with A, leaving only the quantization noise and whatever was below 16 bits in A.
The final result:
Peak Amplitude: -90.35 dB -90.35 dB
Minimum RMS: -98.49 dB -98.53 dB
Maximum RMS: -97.71 dB -97.67 dB
Average RMS: -98.1 dB -98.1 dB
Total RMS: -98.1 dB -98.1 dB
This is whatever was below 16 bit in the original A plus quantization error noise. I believe the levels are higher than the non-dithered conversion reported above only because it includes additional quantization noise from the white noise dither. The spectra is flat, the sound is everywhere a light hiss, audible only at a very high volume setting. If there is any real difference between the 24 bit and 16 bit versions, aside from the dither and quantization noise, it is below the resolution of my tools. I suggest this is indeed, as David commented, inaudible. Any ideas or suggestions?
Unfortunately, if looks like my attempt to format so that the numbers display in aligned columns is not working. Also, this new software, with the entry window so much wider than the screen, requiring scrolling right and left, is a real pain.
Nulling does unfortunately not reveal everything, since we actually don't know what we are looking at or listening to. What is noise? Don't want to repeat myself so please have a look at my post in this thread in Gearslutz (http://www.gearslutz.com/board/so-much-gear-so-little-time/358301-24-vs-16-bit-not-audible-8.html)
Martin
I'm posting in this thread because I find it fascinating and want to learn, not to criticise anyone. I wanted to say that straight away, in case it sounds like I do criticise anyone!
AndyH-ha, I followed all of that. I concur with almost all of it. I know we've both been using CEP professionally for a long time now - though for quite different uses. I only use it as a "toy" in the field where you use it professionally. I have, but don't use, Adobe Audition. That's probably irrelevant to the current discussion.
I'm not sure about this bit...
If one takes any two different audio files and mixes them together, one has a combination, such as a vocal and backing track now in one file. If one then takes one of the two, inverts it, and mixes it with the combination, as in the above paragraph, one is left with the original of the other file -- almost.
In the first mixing, each resulting sample is the sum of the corresponding samples from the two inputs. A quantization error occurs in every case where the sum of the two exceeds the precision the format can carry. This will occur even when working in 32 bit float.
If you don't scale and don't clip - and the originals do not clip - then you should still get digital silence. It certainly works here (CEP 1.2a!). It won't work if you reduce each file to 50% during the addition (because this implies scaling and re-quantisation), but it will work if neither file peaks above 50% to start with, and you add them as-is (no scaling, no clipping, no actual re-quantisation because you're just adding two integers). It works at 16-bit or 32-bit.
If the original 16-bit audio hits negative full scale, then you can't invert it properly, because negative full scale is one sample larger than positive full scale (-32768 vs +32767). CEP reduces these clipped negative samples by 1 LSB when transforming them to full scale positive samples using the function "invert", but keeps them intact if you tick the "invert" tick-box in the mix-paste dialogue.
I think the increase at low frequencies you've seen in some experiments
could be a quirk of CEP's rounding - it certainly rounds upwards in some operations, thus introducing a maximum of one LSB (average of 0.5 LSB) DC shift. I think the FFT is showing this. The exact appearance depends on the FFT window, but what you'd expect to simply fall into the "DC" bin (and so not be shown on that graph, if it's plotted correctly) ends up being distributed across many of the low frequency bins. In transform>amplitude>amplify you can add an intentional DC shift - add it to digital silence, and look at the frequency analysis - strange, isn't it?
Then we get to the crux of the problem.
We all know the mathematical difference between the 16-bit and 24-bit files is almost nothing. It's a struggle to hear it
without the music in top, never mind with!
Martin suggests this isn't the point, because hearing doesn't work like that, and then gives two (IMO unconvincing) examples...
First, I truncated (no dither) the file to only 8 bits. Then I performed a null test, comparing it to the 24 bit version. As Ethan and others have observed, mostly noise and definitely no music there.
BUT, when I took this 'noise' (the result of the null test) and combined it with the truncated 8 bit version, it really helped to 'restore' the music. So, the conclusion would be that we can't simply listen to a null test difference and say it's only noise and not music related information.
I don't know how your ears work, but to my ears, the 24-bit file sounds fine, the 8-bit file sound horribly noisy, and the "difference" / "null" file sounds exactly like the noise in the 8-bit file, but without the music. I don't hear any magic - there's "music", "music + noise", and "noise". In my mind, if you remove the "noise" from the "music + noise" I'd expect it to sound like "music". It doesn't surprise me at all.
Also, elsewhere in that thread, there's theexample of a low pass filter - again suggesting that if you listen to the "dull" low pass filtered version, and then listen to only what you've removed (a hissy, tinny thing), you wouldn't dream that putting the two together would give you the nice bright full sounding original. Well I would - that's exactly what my ears tell me to expect!
So, I think they're really bad examples. They don't work for my ears at all.
Also, FWIW, and IMO, the logic of some members on the other board, and the understanding of "resolution" especially, is really poor. They kind of got there in the end - but what's being discussed is fundamental and important, while the understanding is really lacking.
However - there
are situations where the mathematical difference between two signals doesn't related to the audible difference between those two signals. As anyone who has ever done .mp3 minus .wav can tell you, the mathematical difference is huge, the audible difference can be tiny. Listen to the mathematical difference ("null") all you like - you'll never hear it fully in the mp3 itself as the difference between the mp3 and the wav. Here, my
ears tell me that "music" (original .wav), "music + noise" (.mp3), and "noise" (.mp3 minus .wav) cannot possibly be related in the way that I
know (mathematically) they are.
So, it's
possible that the audible difference between a 16-bit file and a 24-file is different from the mathematical difference. Surprising, but possible.
The classic paper(s) from Lipshitz and Vanderkooy, e.g. http://www.aes.org/e-lib/browse.cfm?elib=11586 (http://www.aes.org/e-lib/browse.cfm?elib=11586) (I think - my copy is on paper somewhere!) point out that while the mathematical difference (due to correctly dithered truncation) "
sounds" like pure uncorrelated noise, if you examine it properly there is still third order and higher correlation present - it's just the first two orders of correlation (which give rise to harmonic distortion, and noise modulation) have been removed. This is
much better than not dithering, but not perfect.
So, can we hear those third order and higher errors? Rather than simply hear the added noise by cranking the volume control up ridiculously high? As far as I know, this is the first time anyone has ABXed this, so it's well worth looking at.
If the equipment is working properly, at a reasonable listening level,
and digital silence vs dithered silence cannot be ABXed,
and 24-bit vs dithered 16-bit can be ABXed, then that is exciting.
I don't know if this is what we have here. lossyWAV thinks the dither might be audible anyway, so how loud is Martin playing these files?
Martin: can you ABX the "null" file vs digital silence? At the same listening loudness that you
can ABX 24-bits vs 16-bits?
Cheers,
David.
Greetings, found your forum recently, so here's my first post! Wanted to share another set of files that might be useful for listening tests. The source material is a 'raw' unprocessed master recording downsampled from 96kHz. Uploaded the files here:
Digital audio resolution test files (http://www.nordicaudiolabs.com/dart/)
Below you can see a quick ABX result I got using my laptop and a pair of Beyerdynamic DT250 headphones.
(http://www.nordicaudiolabs.com/samples/ABXreport.gif)
The second set of files are from an old master tape, comparing 24 bit versus 12 bit versus 320kbps mp3.
I would sincerily hope that *nobody* would try to do *anything* definitiive with these files. Based on my evaluaton of the 24/96 version, here's not enough energy > 20 Khz (or even 5 Khz!) to be meaningful, and the dynamic range is only about 60 dB. If you want to convince yourself that 12 KHz sampling and 12 bits are pretty good, be my guest!
I know I can convert properly; possible audible differences with other conversions to 16 bit are of no value. Tests based on them on very suspect. I understand most anyone can do a proper conversion, given the necessary software, but a fuller specification of the particular process employed is desirable.
What exactly do you need to know apart from the dither used? Please let me know. While I understand that this is slightly controversial, we have to remember that most professional audio is done in 24 bit today, probably for good reason. So we could at least suspect an audible difference. But it's a good idea to download only the source file to save download time.
Blue printing on a black background is extremely difficult to read.
Oh, I'm sorry, looked fine on my screen, but fixed it now! Please reload the page. Thanks for pointing that out.
Martin
It would be helpful if Martin Kantola could describe the differences he heard, where in the file he heard them, and most importantly can verify that his laptop plays back 16-bit and 24-bit digital audio correctly.
While the hardware of my laptop is 24bit/192kHz capable, the specs are not exactly impressive:
(http://www.nordicaudiolabs.com/samples/ALC260data.gif)
Will get back to the differences I heard, have to try a 16 bit padded to 24 bits to make sure it works as it should.
Martin
IME, The RealTek audio interfaces aren't really 24/192 capable. FR testing shows that they downsample everything to a much lower sample rate, and much lower resolution.
I would sincerily hope that *nobody* would try to do *anything* definitiive with these files. Based on my evaluaton of the 24/96 version...
There are two completely different recordings on that page. I don't think anyone in this thread has shown any interest in the second one.
Cheers,
David.
Martin suggests this isn't the point, because hearing doesn't work like that, and then gives two (IMO unconvincing) examples...
I don't know how your ears work, but to my ears, the 24-bit file sounds fine, the 8-bit file sound horribly noisy, and the "difference" / "null" file sounds exactly like the noise in the 8-bit file, but without the music. I don't hear any magic - there's "music", "music + noise", and "noise". In my mind, if you remove the "noise" from the "music + noise" I'd expect it to sound like "music". It doesn't surprise me at all.
Let me try to explain myself better, the "magic" would be in the last,
third file, the "nullresult-plus-truncated.mp3". By adding only what
sounds like noise to the truncated 8-bit file, two things happen, at least to my ears:
1. The noise goes away.
2. The distortion goes away.
What I tried to show was that we clearly removed more than "just noise" when truncating, and so
the removed bits must contain signal-related information. Adding any other uncorrelated noise (with the same spectrum) to the truncated file would not do the trick. Yes, as such it's just an obvious subtraction-addition operation and shouldn't surprise anybody. But the point is that we can't necessarily tell by analyzing or listening to the null "noise" if there's useful information or not in there, and certainly not how much.
I don't know if this is what we have here. lossyWAV thinks the dither might be audible anyway, so how loud is Martin playing these files?
Not too loud at all, mainly because of the high impedance (250 Ohms) of the headphones. Can't measure SPL from headphones.
Martin: can you ABX the "null" file vs digital silence? At the same listening loudness that you can ABX 24-bits vs 16-bits?
Of course not, but that's a good point. As explained above, the null file in itself is meaningless to us. The LSB information only makes sense when combined with the rest of the data.
Martin
I would sincerily hope that *nobody* would try to do *anything* definitiive with these files. Based on my evaluaton of the 24/96 version, here's not enough energy > 20 Khz (or even 5 Khz!) to be meaningful, and the dynamic range is only about 60 dB. If you want to convince yourself that 12 KHz sampling and 12 bits are pretty good, be my guest!
To begin with, the old analog tape is a fantastic recording, and regardless of the tape noise I personally find it to be of very high quality. But thanks for bringing up the 12 kHz 12 bit suggestion. Am I correct to assume that you're saying this particular recording could be truthfully captured with said digital resolution based on the dynamic range and frequency range you estimated? If so, that's where I find the sampling theorem not properly applied. Why? Because it's not trivial to determine exactly where the music ends and the noise begins. We can guess of course, but in order to make sure we must capture the tape noise as part of the signal too. And so an interesting question forms, what is the noise floor of the tape noise?
Martin
What I tried to show was that we clearly removed more than "just noise" when truncating, and so the removed bits must contain signal-related information. Adding any other uncorrelated noise (with the same spectrum) to the truncated file would not do the trick.
I don't think you've
proven the point (in the scientific sense).
Where you're making it look more clever than it is is in saying "removed". Forget that. Say "added" instead. And if you don't like that, remember than you can add a negative, so it means the same thing.
Then the situation becomes....
1 = signal
2 = signal + noise_A
3 = 2 - noise_A
= (signal + noise_A) - noise_A
= signal
= 1
...where
1 = original 24-bit signal
2 = dithered 16-bit signal (or 8-bit signal in this case)
3 = difference ("null")
You believe there's something magical (or, specifically, signal related) about noise_A because, when you subtract it, you get the signal back - whereas if you subtracted "any other uncorrelated noise (with the same spectrum)" (let's call that noise_B) you would not. That's not magic at all - that's very simple maths.
Subtracting noise_B doesn't give you the original, because you didn't add noise_B in the first place. If you had done, it would. It wouldn't prove that noise_B was related to the signal either.
I don't dispute the overall point you're trying to make - noise_A might be related to the signal. In fact, it demonstrably is (if you know what to look for, you can create waveforms where this fact is clearly visible). But what you've done doesn't prove it. We need to look deeper than that.
Here's an interesting question (for the 16-bit version at least): Does
any similar (but completely uncorrelated) noise make it sound worse, or is it only dithering+truncation that makes it sound worse? In other words, is noise_A a special "bad-case", or not?
Cheers,
David.
what is the noise floor of the tape noise?
5-bits.
(someone told me once - I have no better reference than that!).
i.e. if you have truly random white noise, and you want to sample it, you can do so transparently with 5-bits.
If it's not white or not truly random, then you might need more. (I could go on about those two for a page or more, but won't).
I can't remember how many bits lossyWAV allocates, but it's about right IIRC.
Cheers,
David.
On the dither_noise_test sample you supplied lossyWAV removes 10 bits [edit] at --portable; 9.6821 at --standard [/edit] - which is at [edit] (or very close to) [/edit] the hard internal limit of 5 bits remaining + sign bit.
BTW...
foo_abx 1.3.1 report
foobar2000 v0.9.4.2
2009/02/09 16:29:47
File A: D:\audio\digital audio resolution test files\96k24b.wav
File B: D:\audio\digital audio resolution test files\mp3-48k320kbps.wav
16:29:47 : Test started.
16:31:04 : 01/01 50.0%
16:31:25 : 02/02 25.0%
16:31:39 : 03/03 12.5%
16:31:50 : 04/04 6.3%
16:32:12 : 05/05 3.1%
16:32:20 : 06/06 1.6%
16:32:30 : 07/07 0.8%
16:32:43 : 08/08 0.4%
16:32:44 : Test finished.
----------
Total: 8/8 (0.4%)
Easy.
The "T" of "I'm
travelling light" is changed by the mp3 encoding.
This is the first time I listened to it. I agree it's a nice recording in many ways. It would have to be butchered by noise reduction to go onto a CD though. Play that as-is on the radio today and the hiss would be dragged up to be louder than the music.
I don't know what mp3 encoder + settings you used (and hissy recordings are a challenge to mp3 in general) but I bet it can be encoded better than that.
Cheers,
David.
P.S. EDIT: That "24/96" recording looks like it was copied in the analogue domain from a CD. There's a null at 22kHz, a little crap (completely unrelated to the music) a little above it, and then nothing but noise, tens of dB down, above that. I hope Arny was joking somewhat with his figures though - there's clearly real audio information in the speech transients up to 20kHz, and a 6kHz LPF is easily detected. While I'm sure you could take the bitdepth right down (especially as it's a 96k recording), lossyWAV wants to keep 16 bits at a few points. The recording, as given, only peaks at -13dB, so the top 2 bits aren't being used either. Maybe it gets louder later on.
Total: 8/8 (0.4%)
Easy. The "T" of "I'm travelling light" is changed by the mp3 encoding.
Yep, that's how it worked for me too, find a little piece in the music that changes and focus on it. Thanks for listening!
This is the first time I listened to it. I agree it's a nice recording in many ways. It would have to be butchered by noise reduction to go onto a CD though. Play that as-is on the radio today and the hiss would be dragged up to be louder than the music.
I should tell you that this is only a fragment of the whole piece of music, it gets
much louder as you guessed, so the overall dynamic range is better. To de-noise or not is an interesting question.
I don't know what mp3 encoder + settings you used (and hissy recordings are a challenge to mp3 in general) but I bet it can be encoded better than that.
The source is there, so if you want to give it a shot go ahead. I'm sure there are differences in encoders.
P.S. EDIT: That "24/96" recording looks like it was copied in the analogue domain from a CD. There's a null at 22kHz, a little crap (completely unrelated to the music) a little above it, and then nothing but noise, tens of dB down, above that.
Interesting, don't know the exact signal path that recording has been through. Wasn't born when it was made. :-)
Martin
Could this whole situation be simply from Martin's poor-quality audio card causing some kind of non-linear distortion that makes these results actually audible when normally they are not?
You believe there's something magical...
Oh, not at all, that's just a word you used and I liked it ;-)
I don't dispute the overall point you're trying to make - noise_A might be related to the signal. In fact, it demonstrably is (if you know what to look for, you can create waveforms where this fact is clearly visible).
Good, that's all I wanted to say.
But what you've done doesn't prove it. We need to look deeper than that.
Well, even if I wasn't out to prove anything scientifically, I do feel that my simple test shows that the null noise is more related to the music than not. Just try adding any uncorrelated noise instead. But I agree, we need to look deeper.
Here's an interesting question (for the 16-bit version at least): Does any similar (but completely uncorrelated) noise make it sound worse, or is it only dithering+truncation that makes it sound worse? In other words, is noise_A a special "bad-case", or not?
Not sure I follow you 100% here, please help me understand, but you make the assumption that noise_A is only... noise?
Martin
Could this whole situation be simply from Martin's poor-quality audio card causing some kind of non-linear distortion that makes these results actually audible when normally they are not?
The poor quality of the equipment is actually what makes this interesting. But we've discussed using 24 bit files padded with zeroes instead of 16 bit ones to rule out some possibilities of contamination.
As far as audible on a good system, I would go as far as claiming that any full time mastering engineer not able to hear the difference between the 24 bit source and the 16 bit on his own setup should probably get another job... But when I posted the test on the GS mastering forum, nobody seemed interested in either taking the test or posting their results.
In any case, we'd need more honest trials from more people.
Martin
Could this whole situation be simply from Martin's poor-quality audio card causing some kind of non-linear distortion that makes these results actually audible when normally they are not?
Could be. I tried to remove the most obvious possibility by taking the 16-bit version, converting it back to 24-bits, and filling up the bottom 8 zeros with noise. Martin could still ABX this against the 24-bit original (it's a couple of pages back - not sure the ABX stats are sound though).
Any other ideas how we could remove this possibility?
Cheers,
David.
Based on that statement, I've come to the conclusion that you're pushing some weird subjectivist agenda. As no one with decent equipment can validate your results (and those who try and fail are not liable to post and say they did) and your results seem to be primarily on a consumer-grade sound device, the conclusion I come to is not that you're hearing a difference between the samples, but rather hearing a difference in your hardware's rendering of the samples.
I'm not gonna argue this any more, I'm just out.
Here's an interesting question (for the 16-bit version at least): Does any similar (but completely uncorrelated) noise make it sound worse, or is it only dithering+truncation that makes it sound worse? In other words, is noise_A a special "bad-case", or not?
Not sure I follow you 100% here, please help me understand, but you make the assumption that noise_A is only... noise?
I mean testing what you have been saying...
Dither+truncation adds noise of a certain character (e.g. white, -90dB RMS), but it is correlated to the original signal in some way (not a way that most people believe is audible, but that's the point!).
You could have noise with exactly the same character (white, -90dB RMS), which has nothing to do with the original signal.
It would be interesting to know if both can be ABXed against the 24-bit original, or if the latter cannot. That would determine whether what you can detect here is any old noise, or a specific type of noise.
Cheers,
David.
As far as audible on a good system, I would go as far as claiming that any full time mastering engineer not able to hear the difference between the 24 bit source and the 16 bit on his own setup should probably get another job...
Sighted tests don't count around here. There are thousands of examples of "obvious" differences that vanish in a double blind test. Thousands of examples of things that "everyone knows" which, when it comes to it, no one can actually prove.
Sadly, the audio world is full of them.
Also sadly, many people who claim to hear such things refuse point-blank to do ABX tests.
As far as I know, you're the first person to ABX 24>16-bit conversion at a low level. That's kind of extraordinary.
Cheers,
David.
You could have noise with exactly the same character (white, -90dB RMS), which has nothing to do with the original signal. It would be interesting to know if both can be ABXed against the 24-bit original, or if the latter cannot. That would determine whether what you can detect here is any old noise, or a specific type of noise.
Oh I see now, excellent idea! Thanks for explaining. I can try ABX:ing with this shitty setup, but I'd much rather get deeper into the testing when I'm back in my studio with proper gear, where I'll be able to measure the SPL used too.
Martin
Based on that statement, I've come to the conclusion that you're pushing some weird subjectivist agenda.
Uh, is that a scientific conclusion? But seriously, my interest in this is genuine. While working in the studio the audible difference has always been obvious enough on certain material with enough dynamics, it's actually a surprise to me that it's not an accepted scientific fact. The people with good brains and good ears need to be able to talk to each other.
As no one with decent equipment can validate your results
Not 100% true. Please read the GS thread. But as I said, we need more results to either invalidate or validate.
I'm not gonna argue this any more, I'm just out.
OK, bye... But again, I'm not here to argue!!!
Martin
Yes, I begin to see this... I guess I'm more irritable today than I thought at first. Please excuse my skepticism. I've had this argument with so many people that it's not even funny. It's even worse when I'm right and I crush their self-confidence.
Have you tried ABXing this sample with other forms of dither? I see one 5.5% result in the GS thread. That is not particularly statistically significant. If you'll look at this thread before you posted, Pio also perceived this difference. Yet without the appropriate dither, there's still no proof that 16-bit is not sufficient, just that this HPT dither is. Your result for the AVD dither is 8/10, and your results seem noisy like there's no real perceived difference. Personally, given my own preferences, I'd love to see the foobar2000 dither tested.
There are thousands of examples of "obvious" differences that vanish in a double blind test. Thousands of examples of things that "everyone knows" which, when it comes to it, no one can actually prove. Sadly, the audio world is full of them.
Very true. And that's why I got so excited with ABX! But let's not forget all the work that's been done on different dithering and noise shaping methods, and how we've arrived at some that do a great job. Do we really think it's all inaudible at normal listening levels?
Also sadly, many people who claim to hear such things refuse point-blank to do ABX tests.
Shame, personally I firmly believe that anything real can eventually be both heard
and measured (or ABX:ed). But not easily in all cases.
As far as I know, you're the first person to ABX 24>16-bit conversion at a low level. That's kind of extraordinary.
You think? Don't think the 24 bit standard is only a result of clever marketing. There must have been a lot of testing at low (or normal) level among people who developed all these converters and software. My ears are probably pretty good because of many ears of training, but not extraordinary.
Martin
Have you tried ABXing this sample with other forms of dither? I see one 5.5% result in the GS thread. If you'll look at this thread before you posted, Pio also perceived this difference. Yet without the appropriate dither, there's still no proof that 16-bit is not sufficient, just that this HPT dither is.
Thanks for staying! I received a file with this "ditherbox avd" dither and the specific difference I heard with HPT was definitely gone. Instead I heard something in the stereo image change between A and B, but it was still an improvement in quality no doubt. Maybe I should test that one again with more trials for a more reliable result?
Martin
Yes. If you truly heard a difference, increasing the number of tests should not pose a huge difficulty, though I acknowledge the crucial role of listener fatigue. The most interesting ABX result to me would be foobar2000's dither algorithm, which is novel and could be superior to these other solutions.
I would sincerily hope that *nobody* would try to do *anything* definitiive with these files. Based on my evaluaton of the 24/96 version...
There are two completely different recordings on that page. I don't think anyone in this thread has shown any interest in the second one.
The first file, a 24/48 file named 24.wav is even worse!
It stands to reason that if you want to hear the benefits of 24 bits, the sample itself has to have some dynamic contrast.
The first file, a 24/48 file named 24.wav is even worse! It stands to reason that if you want to hear the benefits of 24 bits, the sample itself has to have some dynamic contrast.
Thanks for listening, but would you care to explain yourself a bit more in detail? How did you evaluate this recording and in which way did you find it so bad? You are hinting that the sample does not have dynamic contrast, which is a bit surprising considering that the dynamic range was not altered in any way, i.e. no compression or limiting was used.
Martin
I use CoolEdit 2000. This idea that quantization errors occur and thus can’t be nulled out seems logical to me. Where am I going wrong?
I created two files with Generate/Tones presets, one the Chord, the other the Bell. I used the Mix-Paste function to add them together. The peaks go way over 0dB, but since this is in 32 bit float, that should not matter.
Next I Mix-Paste-Inverted one of the originals to the sum file. Then I Mix Paste Inverted the other. Statistics on the results:
Mono
Min Sample Value: 0
Max Sample Value: 0
Peak Amplitude: -144.5 dB
Possibly Clipped: 0
DC Offset: 0
Minimum RMS Power: -167.07 dB
Maximum RMS Power: -153.7 dB
Average RMS Power: -158.69 dB
Total RMS Power: -158.15 dB
Then I recreated both starting with the presets but setting the level sliders at -24dB. This kept the sum far below 0dB. The same processing as above gave
Mono
Min Sample Value: 0
Max Sample Value: 0
Peak Amplitude: -162.56 dB
Possibly Clipped: 0
DC Offset: 0
Minimum RMS Power: -188.58 dB
Maximum RMS Power: -174.01 dB
Average RMS Power: -179.1 dB
Total RMS Power: -178.45 dB
Both remainders were made visible in Spectral View by increasing the range to 200dB. Likewise, both show up on the Frequency Analysis graph. I find it interesting that the Min and Max sample values are all zero. I am not sure what that means, but I think it is simply that the values are less than 1 and these statistics don’t resolve that far down. If I zoom into a peak sample, for instance, and look at its 24 bit range, 0.0625 is displayed.
An inverse mix of either file with only itself does produce digital silence.
I use CoolEdit 2000. This idea that quantization errors occur and thus can’t be nulled out seems logical to me. Where am I going wrong?
Am not sure I understand where you are going with this analysis, and I may be missing the point, so correct me if I am wrong. However I'll mention that intermediate calculations in digital signal processing software are commonly carried out at a higher precision than the nominal format in use. Hence, the quantization rounding to 32 bit float may only ocur when saving the final result to disc. Intermediate steps in the processing may benefit from much more precise quantization.
In any event, I would have thought that quantization error at the least significant bit of 32 bit float would be inaudible; only the cumulative effect of multiple stages of processing at 32 bit float might possibly be audible.
I think this is all done at 32 bit float. However, the post is in response to David's comment that I errored in my conclusion on this operation. This has nothing to do with audibility, only with whether or not quantization errors prevent a null final result.
Andy,
I think I know what's gone wrong there.
The 32-bit float format in Cool Edit is something like* 1 sign bit, 23 data (mantissa) bits, and 8 exponent bits. That holds 24-bit audio perfectly, since 24-bits can be thought of (or actually can be) 1 sign bit and 23 data bits. 24-bit > 32-bit = lossless.
However, if you take two 24-bit files with values over 50%, and add them, you get a 25-bit result - an overflow.
In 32-bit floating point, this overflow isn't clipped - it's preserved because the exponent simply increases by 1. However, you should have 25-bits of resolution - but you have nowhere to store the least significant bit. So Cool Edit simply throws it away.
That's the re-quantisation you're seeing. It happens when working with 24-bit files in 32-float and exceeding 0dB FS.
It will not happen if you don't exceed 0dB FS. It will not happen if the original files are 16-bit (unless you add enough of them together to generate more than 24-bits).
If that's not what you're seeing, then I don't know what the problem is. CEP certainly works properly - even when you enabled "dither transform results" it doesn't dither non-scaled addition -because it doesn't have to - they're just integers.
Cheers,
David.
P.S. * = the 32-bit float format changed at some point in CEP, and I can't remember exactly which one is in CE2k. There are a lot of different flavours of 32-bit audio! Several are listed in the CE2k help file, but not in enough detail for me to be sure I've interpreted it correctly.
Sorry, I've lost track a bit -- did anyone ever supply 16 vs 24 bit recordings for testing this, that were 'HA approved'? Or is all this still working off those files supplied on another forum?
The latter.
Would you be happier if I ran the 24-bit file through Cool Edit Pro to generate a 16-bit file and posted the result?
The latter.
Would you be happier if I ran the 24-bit file through Cool Edit Pro to generate a 16-bit file and posted the result?
Couldn't hurt. What dither will you use? CE/Audition standard (triangular, no noise shaping, depth = 1?)
But as there is now some question whether the 24 bit file is sufficiently dynamic in the first place, better still might be a
new set of files.
I've done three...
- No dither at all, no noise shaping
- 1-bit dither TPDF, no noise shaping
- 1-bit dither TPDF, 48k noise shaping
...all created using CEP 1.2a. All files 16-bit, from the same 24-bit original. It's not "the best" conversion that's available, but it's what many people use, and supposedly far more than good enough.
They're here::
http://www.mediafire.com/?sharekey=605ee2b...621d66e282a0ee8 (http://www.mediafire.com/?sharekey=605ee2b5f85656f5e62ea590dc5e5dbbb02137efc9358eb45621d66e282a0ee8)
If you subtract these files from the original, and perform an FFT analysis averaged over the whole file (using the "SCAN" button), you get this...
[attachment=4884:diff.jpg]
...which is what you'd expect.
I think it's precisely because Martin can ABX it despite it being a "normal" file that it's interesting. (how many audio tracks really use more than 16-bit dynamic range anyway?)
FWIW all the tracks in the tiny collection of 24-bit audio I have are similar - i.e. the noise floor is above that of a correctly dithered 16-bit file.
Aside:
It occurs to me that this can be a misleading way of thinking - it implies that if you have noise at x dB, and some noise at y dB, then as long as y is less than x, then adding signal y to signal x will make no audible difference. This is nonsense! y has to be "much less"* than x for the addition to be inaudible. If it's just "a little bit less", then it's clearly audible - that's basic noise addition. So, if you have a 24-bit recording with noise at the 16-bit level, you can't convert it to 16-bits without adding more noise, which may be audible.
* "much less" isn't trivial to quantify. If the addition raises the noise power across the board and in each spectral band by less than a small fraction of 1dB when added, it's probably "much less", i.e. inaudible. There must be psychoacoustic data available where this has been quantified in listening tests, but I don't have it on my desk.
Cheers,
David.
I think I understand your response, a simple addition shouldn’t overflow, and thus there would be no quantization errors, depending upon the values added.. However, as I stated, the second example of my last post used lower level components. The mixture read
Min Sample Value: -3835.67
Max Sample Value: 3778.33
Peak Amplitude: -18.63 dB
Possibly Clipped: 0
DC Offset: 0
Minimum RMS Power: -35.4 dB
Maximum RMS Power: -23.83 dB
Average RMS Power: -28.41 dB
Total RMS Power: -27.94 dB
which is way below 0dB, yet there was still leftover after invert mixing.
The file format is 32 bit 16.8 float (type 1 - 32 bit), which is, I believe, also CE2K’s working format.
...there was still leftover after invert mixing.
The file format is 32 bit 16.8 float (type 1 - 32 bit), which is, I believe, also CE2K’s working format.
This is delving really deep, but there's an explanation even for that. Yes, I can reproduce it here. I can fix it too...
btw, interestingly, the "peak amplitude" values in the analysis window in CEP 1.2a report the correct values, i.e. +/-.00012207, rather than simply 0.
What's happening is that when you generate the artificial signals, you're doing so in 32-bit float = exponent and mantissa.
All samples have a 23-bit mantissa, even the low level ones.
So a sample which happens to have a value of about -48dB FS (which, if we were working with integers, would have 8 leading/MSB zeros) still has 23 bits of accuracy - and, if we were working with integers, it would need 8 leading zeros and then another 24 bits - i.e. it's a 32 bit integer.
A sample which happens to have a value of about -96dB FS (which, if we were working with integers, would have 16 leading zeros!) still has 23 bits of accuracy - and, if we were working with integers, it would need 16 leading zeros and then another 24 bits - i.e. it's a 40 bit integer!
When you add these various samples together, all floats, all with different levels of equivalent integer bitdepth, the result is still a 32-bit float. It still only have 23-bits of actual data/resolution. So you're throwing away lots of genuine bits from at least some samples of the original. That's where your quantisation error comes in.
Now, if you take your two signals, your -24dB tone and your -24dB bell, and
before you do anything else with them you convert them to 24-bits without dither - and then work with those 24-bit versions (still expressed in 32-bit floats, but now only with the resolution of 24-bit integers - any extra LSBs of the mantissa will have been zeroed) - you'll find that adding them and subtracting them will give samples with a value of exactly 0, and RMS measures displayed as -Inf dB (which is exactly what you get if you generate pure digital silence).
So, with real 24-bit data, 32-bit floating processing works properly, and completely avoids quantisation errors within the top 24-bits.
It's only by
starting with artificially generated 32-bit floats (with a maximum resolution which is equivalent 48-bit integers!) that you get rounding errors.
I must be quite sad, because I find all this interesting - mainly because it shows that a lot of thought went into Cool Edit to design it properly.
Cheers,
David.
I think David has cleared up the issue AndyH raised concerning very small discrepancies that can occur when adding and subracting 32-bit float files. Now, returning to the main subject matter of this thread ...
Everyone (especially Martin!),
the 24 bit original recently discussed can be found in Martin's link. It is the first file, labelled "24-bit 48KHz source", and it has the filename
24.wav :-
Greetings, found your forum recently, so here's my first post! Wanted to share another set of files that might be useful for listening tests. The source material is a 'raw' unprocessed master recording downsampled from 96kHz. Uploaded the files here:
Digital audio resolution test files (http://www.nordicaudiolabs.com/dart/)
A foobar dithered version can be downloaded using David's link in post #145, just below. It has the filename
24_16_fb2k_dither.wavCan ABXs please be done on
24.wav in comparison with
24_16_fb2k_dither.wav ?
If someone does find they can successfully ABX, can they explain how the sound of the dithered version differs from the original 24-bit sound? If the difference is most noticeable at a particular time interval from the start of the file, can that time interval be quoted?
I didn't include foobar2k before.
I have now - it's much better than Cool Edit at reducing noise in the main audible range...
[attachment=4886:diff2.jpg]
(sorry it's gone off the scale at the bottom - I'm not going to re-do all the others to fix the scale)
Here's the direct download link:
http://www.mediafire.com/file/zaiiuyx4izd/...fb2k_dither.wav (http://www.mediafire.com/file/zaiiuyx4izd/24_16_fb2k_dither.wav)
That's the one to try to ABX against 24.wav from Martin.
The "null" file (24.wav minus 24_16_fb2k_dither.wav) is much quieter - though if you boost the volume and listen very loudly (i.e. about 60dB more loudly than normal!) all that high frequency noise is quite painful. This isn't relevant to listening to the actual 24_16_fb2k_dither.wav file itself, since boosting that by 60dB would be more than deafening (it would peak at 140-160dB SPL - pretty much impossible.).
Cheers,
David.
That's the one to try to ABX against 24.wav from Martin.
OK, thanks, will try this in during the weekend sometime. How many trials should I do? Would 25 be enough? Don't want to go through 100 again if this is really difficult...
Martin
Since you need to decide on the number before the test, if you are happy doing 25, that's how many you should do.
Cheers,
David.
Since you need to decide on the number before the test, if you are happy doing 25, that's how many you should do.
Yep, that's why I asked. 25 it is then. BTW, might have another recording of a completely different nature to use for further testing.
Martin
To make a contribution, of sorts, I report that I managed a score of 8/14, using the dither I normally use. Before listening, I converted the 16 bit file back to 24 bit, although I don’t think that my system has any difficulty playing 16 vs 24 bit differently. I had intended to do 15 trials but somehow got confused. I don’t think that invalidates the idea that I do not appear to be able to hear any difference.
This is more a technical point than a practical one, but is it not the case that if I first saved my white noise dither at 24 bit, before using it in the procedure I described (i.e. as David wrote a few post back), I would remove all traces of dither from the 16 bit version, thus completely eliminating any possibility of dither noise itself being responsible for any perceived difference?
The 3kHz tone was chosen to be easy to distinguish the result of dithering vs non-dithering. Non-dithered conversion to 16 bit produces ample distortion, easily viewed and measured. Then, starting over,
(1) Mix same 32 bit audio file with 32 bit white noise file, producing dithered audio.
(2) Convert mixed file to 16 bit, no dithering. Produces non-distorted 16 bit.
(3) Convert back to 32 bit. Nothing below 16 bit but zeros.
(4) Convert white noise file 16 bit, no dithering.
(5) Convert 16 bit white noise back to 32 bit. As always, this leave zeros below 16 bits.
(6) Invert (5) and mix with (3).
The second half of the resulting file, which was originally silence, is again complete silence. The first half of the file, the tone, is now the tone plus quantization noise.
Naturally the second half of the resulting audio file remains as silence, as the white noise was mixed at step 1 with nil, i.e in the absence of a non-zero signal to disturb the white noise. The conversion of the white noise to 16 bits at step 2 created exactly the same result, as at step 4. These resuts were subtracted from each other at step 6, yielding zero.
As for the first half of the file, I do not understand the purpose of subtracting low resolution (converted to 16 bit) white noise from a dithered file created using high resolution (32 bit float) white noise. The result likely to be obtained strikes me as somewhat arbitrary, and not simply 'the tone plus quantization noise'.
I note that bits in the high resolution dither will interact with bits 17 to 24 of the original (if it were 24 bit integer) audio such that the audio will modulate the random dither, and the result will be rounded to 16 bits (integer). In contrast, rounding of 32 bit float dither to 16 bit integer without another audio signal present is a straightforward "unmodulated" conversion.
If the dither were of a maximum amplitude of slightly less than 0.5 bit at 16 bit integer resolution, the converted (to 16 bit integer) value of the dither would at all times be zero; and yet even that low amplitude of dither at original resolution would have an effect when combined with a 32 bit float or 24-bit audio signal destined for rounding to 16 bits integer. But I do not understand dither at a detailed mathematical level, and perhaps I am missing some key explanation!
The resulting silence was expected. Remember, the tone + silence demonstrations was just my proof-of-concept trial.
In regard to subtracting the 16 bit dither from the converted to 16 bit file, consider two possible variations.
First, instead of converting the dither file to 16 bit to invert-mix with the audio file, convert the audio back to 32 bit, then mix-invert the dither into that (i.e. subtract the original dither). This gives a different result because the audio file is still really only 16 bit of data while the dither has 24 bits. The second half of the tone + silence does not come out silent, although it will have a lower level noise than the dither that was added.
Next repeat the experiment that originally gave silence in the second half of the file but instead of the original dither file, use ay other with noise file of identical specifications. Convert it to 16 bit and subtract. The result is not silence because the “dither” subtracted is not the same data.
Silence only results if the subtraction is of identical data. Staying withing the 16 bit realm, from beginning to end, will produce silence, working with 24 bit data, without converting to 16 bit, will produce silence. Not only the part that started as silence, but also the part of the file that has a signal can be zeroed out completely if all operations stay within the bounds of 16 bit, or 24 bit, where ever we are operating.
Therefore, subtracting the converted-to-16-bit-dither from the dithered-and-converted-to-16-bit audio should completely remove the dither -- except for those tiny values that are the quantization errors of converting the mix of audio and dither to 16 bit. I don’t see why one would call it anything other than the conversion quantization error, or why one would expect it to be any more arbitrary than any other quantization error.
Therefore, subtracting the converted-to-16-bit-dither from the dithered-and-converted-to-16-bit audio should completely remove the dither -- except for those tiny values that are the quantization errors of converting the mix of audio and dither to 16 bit. I don’t see why one would call it anything other than the conversion quantization error, or why one would expect it to be any more arbitrary than any other quantization error.
I confess I still do not see the purpose of the exercise. If someone wanted to know the extent to which an original file differed from a 16-bit rounded version of it, they could simply subtract the rounded version from the original. Similarly, if they wanted to know the extent to which a dithered 16-bits file differed from the original higher resolution file, they could again simply do a subtraction of those two files.
What you have dubbed the "conversion quantization error' is something else again and I do not understand the significance if it. It will be different for different 32-bit float random dither files employed. It will be a function of the actual dither file, despite some part of that dither file being attempted to be eliminated. It is a creature of calculation and the steps to be followed to create it have been clearly explained, but what
in principle it represents, is I'm afraid, still not clear to me.
___________
I hope that Martin (and others) will be giving us the benefit of their ABX reports (+ comments) in the near future on Martin's 24 bit original version versus the dithered 16 bit version at post #145, even if some of those those reports are to the effect that the two versions appear to sound the same.
Hello!
Am one of those others on gearslutz that did 10/10 ABX on the 24 vs HPTPDF 16 bit dither.
The chain was foobar->RME AES32 with asio driver->Lavry DA10. A bit transparent path. The problem with such a setup is that the digital stream can only take one signal at a time. (obviously, being bit transparent). This makes for a disruptive tick sound and a slight interrupt when changing streams between A and B samples. Subtle differences are easier to hear when there's continuos playback. I think it would be easier with two DAC's and a physical ABX unit doing analogue switching. Or, drop the bit transparent path idea and use a software mixer instead. If foobar could send each sound to different output ports, the 40 bit mixer in the RME card could be used and the tick sound/interrupt would hopefully be gone. Providing for that may have an influence on the results.
ABX'ing the 16 bit HPTPDF was not that amazingly hard. I can't say exactly what it is that I'm listening for. It usually changes a few times as the test progresses. When I find a difference to hang on to, it's only good for a few trials until I'm desensitized. Then I have to find another difference, hang on to that until I loose it, and so on. Perhaps the best way to listen is to squint with the ears and simply try to catch the quality of the material. I know this all sounds very vague, but that is exactly the way I listen too, when trying to hear such things. Trying to hear an obvious conscious difference is a sure way to fail. It's a sort of lateral listening.
Could not do the same with either Chris J's special dither or the 24_16_fb2k_dither.wav provided in this thread. So I take it that this is borderline territory. If HPTPDF is audible on this file, better dither may be audible on another file. My favorite for testing such things is to use deep bass sounds, only.. Probably a big no no around here. With such contrived test files, just about any dither should be audible, at least at elevated levels. But I guess that doesn't carry much weight in itself. What I'm trying to get at is that with other pieces of music, better ears/gear/room etc, other 16 bit dithers may be audible too.
As for the differences, just about all 16 bit dithered files can be seen to have more noise than a 24 bit file, when viewed in a spectrogram(like in Izotope RX). Unless the 24 bit source is particularly dirty.
Best regards,
Andreas Nordenstam
(mastering)
When simply comparing the original with the dithered 16 bit version, the dither is so much greater than any data lost or altered when converting to 16 bit that it is hardly possible to find anything except the dither.
If your other offered possibility, the rounded version, means comparing the original with an un-dithered 16 bit conversion, there will be a great deal of distortion in that 16 bit version that hides any differences in which we may be interested.
The purpose of removing the dither after converting to 16 bit is to be able to see and analyze the difference between 24 bit and 16 bit without dither or distortion confounding the results. Does that offer any insight? I don’t know if there is anything beyond the demonstration that the difference is so very small that it is hard to imagine hearing it, or any effect of it, at normal listening volume settings.
The suggestion about using this approach to produce listening samples is to prevent dither noise itself from being the cause of an audible difference, as the most recent ABX success post suggests.
I’m not absolutely certain there isn’t something else that I’m missing. That question was the purpose of my posts on the subject.
Hi AndyH, perhaps another member will be able to provide confirmation/insight into what the "conversion quantization error" file represents.
Could not do the same with either Chris J's special dither or the 24_16_fb2k_dither.wav provided in this thread. So I take it that this is borderline territory. If HPTPDF is audible on this file, better dither may be audible on another file.
Andreas, thank you very much for this. If we find that Martin cannot successfully ABX the 24_16_fb2k_dither version either, then we appear to be back to square one. That is, with a good dither, the 16 bit dithered version sounds the same as the higher resolution original, for normal recording and listening levels.
As you say, it is possible that another high resolution audio file
will lead to an audible difference when well dithered; but at this point in time, it appears no-one has been able to supply such a file.
Thanks again,
MLXXX
Here's two runs I did today, might try again tomorrow:
foo_abx 1.3.3 report
foobar2000 v0.9.6.1
2009/02/14 20:20:45
File A: C:\Documents and Settings\Valued Customer\Desktop\TEST\samples\24.wav
File B: C:\Documents and Settings\Valued Customer\Desktop\TEST\samples\24_16_fb2k_dither.wav
20:20:45 : Test started.
20:39:06 : 01/01 50.0%
20:43:06 : 01/02 75.0%
20:49:13 : 02/03 50.0%
20:49:37 : 02/04 68.8%
20:49:46 : 03/05 50.0%
20:50:08 : 04/06 34.4%
20:50:32 : 05/07 22.7%
20:50:57 : 06/08 14.5%
20:51:25 : 07/09 9.0%
20:51:48 : 08/10 5.5%
20:53:15 : 09/11 3.3%
20:54:08 : 09/12 7.3%
20:54:56 : 10/13 4.6%
20:55:27 : 10/14 9.0%
20:56:20 : 10/15 15.1%
20:56:44 : 10/16 22.7%
20:56:57 : 11/17 16.6%
20:57:09 : 11/18 24.0%
20:57:43 : 12/19 18.0%
20:57:54 : 13/20 13.2%
20:58:57 : 13/21 19.2%
20:59:08 : 14/22 14.3%
20:59:32 : 15/23 10.5%
21:00:47 : 15/24 15.4%
21:00:59 : 16/25 11.5%
21:01:05 : Test finished.
----------
Total: 16/25 (11.5%)
***
foo_abx 1.3.3 report
foobar2000 v0.9.6.1
2009/02/14 21:14:20
File A: C:\Documents and Settings\Valued Customer\Desktop\TEST\samples\24.wav
File B: C:\Documents and Settings\Valued Customer\Desktop\TEST\samples\24_16_fb2k_dither.wav
21:14:20 : Test started.
21:15:36 : 01/01 50.0%
21:16:42 : 02/02 25.0%
21:17:50 : 03/03 12.5%
21:18:55 : 03/04 31.3%
21:20:18 : 04/05 18.8%
21:21:38 : 05/06 10.9%
21:22:26 : 06/07 6.3%
21:23:40 : 07/08 3.5%
21:24:44 : 07/09 9.0%
21:25:34 : 08/10 5.5%
21:26:57 : 08/11 11.3%
21:27:08 : 09/12 7.3%
21:27:15 : 10/13 4.6%
21:27:26 : 11/14 2.9%
21:27:42 : 11/15 5.9%
21:28:26 : 11/16 10.5%
21:28:52 : 12/17 7.2%
21:29:53 : 13/18 4.8%
21:30:52 : 13/19 8.4%
21:31:37 : 14/20 5.8%
21:33:35 : 14/21 9.5%
21:34:06 : 14/22 14.3%
21:34:39 : 14/23 20.2%
21:35:20 : 15/24 15.4%
21:35:55 : 15/25 21.2%
21:35:58 : Test finished.
----------
Total: 15/25 (21.2%)
Again so easy to get distracted and sidetracked. And my ears get very tired. More about what I think I'm hearing later...
Martin
If you add those together, you get 31/50, which is 5.9% i.e. there's a 1-in-17 chance of getting that purely "by chance". That's pretty good!
The table of these percentages is here:
http://www.kikeg.arrakis.es/winabx/bino_dist.zip (http://www.kikeg.arrakis.es/winabx/bino_dist.zip)
Cheers,
David.
Nordenstam and Martin,
Are you concentrating on the quiet part of the file? Or, to put it another way, can you hear the difference on the loud part of the file?
Common sense suggests it must be the quiet part, but lots of people claim (typically without ABX) that it's equally "obvious" during loud parts.
Cheers,
David.
Are you concentrating on the quiet part of the file? Or, to put it another way, can you hear the difference on the loud part of the file?
Not listening to the quiet parts. The artifacts seem to be present within the audible part of the music, or actually part of it. It's very interesting, even if the differences are tiny. What it seems to suggest is either that properly dithered music is not 100% perfect even above the 16-bit noise floor, or that our hearing does not work exactly as we first thought...
Another question is, would a very good D to A converter produce better 16 bit audio too (and not only better 24 bit) so that the differences would in fact be smaller on a good system?
Martin
Another question is, would a very good D to A converter produce better 16 bit audio too (and not only better 24 bit) so that the differences would in fact be smaller on a good system?
Most definitely. The better the DAC, the better the signal. You appear to be hearing a difference, though there is still a reasonable chance that you are not. Even if you are, that just means that the two signals are differentiable on your hardware. Improving (or worsening) your hardware may remove the differentiability.
Are you still doing this test on that laptop?
Most definitely. The better the DAC, the better the signal.
Thought so. Will do some testing with my favourite DACs when I'm back in my studio. Of course, the rest of the system will have better resolution too.
You appear to be hearing a difference, though there is still a reasonable chance that you are not.
Oh c'mon... Look at the results, the probability of guessing was down to less than 3% during the second run. I'm not that lucky :-) But it sure was difficult with this dither.
Even if you are, that just means that the two signals are differentiable on your hardware. Improving (or worsening) your hardware may remove the differentiability.
Obviously I can't try all brands and models of hardware... But how likely is it that it's only on this specific Sony laptop with these headphones that one can hear a difference? Remember that I tried truncated + three different flavors of dither so far.
Martin
Oh c'mon... Look at the results, the probability of guessing was down to less than 3% during the second run. I'm not that lucky :-)
It doesn't work like that. Those percentages are the stats for that point in the test
if that was the point you pre-decided you were going to stop before the test commenced.
The chance of hitting "5%" at
some point during the run are far higher than 5%. Similarly, if you start testing, with the intention of stopping when you reach "5%", the chance of hitting "5%" is higher than 5%!
Calculating the actual stats for that (it's called a "sequential ABX") are difficult, and depend on the conditions. There's an excel spreadsheet about it here...
http://www.hydrogenaudio.org/forums/index....st&p=151958 (http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=15192&view=findpost&p=151958)
...so if you intend to do 25 runs, and you'll count <5% (the "0.95" column in that table) at any point as a "pass", the chances of passing somewhere in 25 attempts are 11.2%, or about 1-in-9.
If you want to run sequential ABX tests, you need to aim for <1% to be sure the actual chance is <5%.
I'm posting simply to clarify that particular part, not to say you didn't pass. I've already said that I think you "almost" passed - just
not because the displayed value dipped below 5% at some point.
Cheers,
David.
P.S. original (non-sequential) ABX stats thread:
http://www.hydrogenaudio.org/forums/index....showtopic=14679 (http://www.hydrogenaudio.org/forums/index.php?showtopic=14679)
Calculator that doesn't do any stats, it just simulates random ABX and says how many passes there are:
http://ff123.net/abx/abx.php (http://ff123.net/abx/abx.php)
(obviously it matches the stats, but it was useful in proving the stats are correct)
Remember that I was listening on my crappy laptop setup with very low level on the phones, so had to listen to changes in timbre and not in low level detail or similar.
I can try ABX:ing with this shitty setup, but I'd much rather get deeper into the testing when I'm back in my studio with proper gear, where I'll be able to measure the SPL used too.
Will do some testing with my favourite DACs when I'm back in my studio. Of course, the rest of the system will have better resolution too. ... Obviously I can't try all brands and models of hardware... But how likely is it that it's only on this specific Sony laptop with these headphones that one can hear a difference?
Hi Martin, at this point in time you are the only person to report an ABX verified difference between 24.wav and 24_16_fb2k_dither.wav
It will be quite an important result if you can also report an ABX verified result using equipment other than the "crappy laptop setup". It would be useful too if you could comment on where in the file the difference is most noticeable, and what the difference sounds like.
Don't think the 24 bit standard is only a result of clever marketing. There must have been a lot of testing at low (or normal) level among people who developed all these converters and software.
There is plenty of evidence 24-bits are useful for recording and mixing. That is not in dispute. However, there is remarkably little, if any, evidence that 24 bits are actually needed for distribution of the final mix to the consumer, compared with a noise shaped dither. (I might mention that a related and contentious question, whether 96KHz sampling might be audibly superior to 48KHz sampling, is not part of the subject matter for the current thread. It's a separate minefield!)
Cheers
Oh c'mon... Look at the results, the probability of guessing was down to less than 3% during the second run. I'm not that lucky :-)
It doesn't work like that. Those percentages are the stats for that point in the test if that was the point you pre-decided you were going to stop before the test commenced.
Don't worry, I've understood how it works ;-) But that temporary low number shows that I was making the right choice several times in a row, which was when I felt I had 'tuned in' on the difference. It has no statistical value or significance, but hopefully illustrates how important the focusing part is of ABX listening. If I could manage to stay focus all the time my score would improve.
Martin
It will be quite an important result if you can also report an ABX verified result using equipment other than the "crappy laptop setup".
Will certainly do that as soon as I can. Would also like to involve others in the testing, to see if I can help them spot the difference by discussing it before the test.
It would be useful too if you could comment on where in the file the difference is most noticeable, and what the difference sounds like.
What I listened for was the upper, 'air' region of the frequency response. The 16 bit seems to be slightly rounded off and lack detail. The highest frequencies produced by the strings is a good thing to focus on. In general I thought the 16 bit sounded a bit dampened and maybe even grainier, but not enough to help me with the ABX.
There is plenty of evidence 24-bits are useful for recording and mixing. That is not in dispute.
As a sound engineer, I understand these benefits. But remembering the first time I had a chance to try 24-bit recording, it was not hard to appreciate the difference even on a single file without any processing and such. But of course, there was an upgrade in converters involved too.
The reason I think 16 bit format works so well for consumers is that the mastering engineers make them sound as good as possible, and are actually able to compensate for the tiny losses we discuss here, even with simple EQ. After all, music is to be enjoyed, not analyzed.
Like I might have mentioned before, the reason I'm still interested in the subject is that if we can show that 24 vs 16 bit is detectable at normal volumes, how does the lowered resolution affect us on a long-term and subconscious level? Is the CD sound less emotionally exciting than it could be? To me, well recorded music does transfer some 'energy' contained in the information and not only sound waves.
Martin
Oh c'mon... Look at the results, the probability of guessing was down to less than 3% during the second run. I'm not that lucky :-) But it sure was difficult with this dither.
It doesn't work that way, sorry.
As things stand, you haven't yet breached the p<0.05 threshold, much less the p<0.01 threshold that some would argue is more appropriate for this phenomenon.
It doesn't work that way, sorry.
Already said I knew that :-)
As things stand, you haven't yet breached the p<0.05 threshold, much less the p<0.01 threshold that some would argue is more appropriate for this phenomenon.
While I understand the importance of a probability threshold, don't see the point of me personally trying to satisfy those numbers. Instead I'd like to see more people getting similar results, to learn more if there's indeed something to this. Not trying to provide final scientific proof at this point, only make a suggestion which should be followed by more and better controlled testing.
It's really hard to hear the difference, that's simply why I can't breach the threshold. It's not necessarily because it's inaudible. Only a machine can deliver the same result time after time. Don't get me wrong, I've recently found ABX and I'm very excited about the possibilities, it seems to be a very effective way to rule out one's own beliefs. But I sincerely hope you're not saying that only p<0.01 would make this worthwhile discussing further?
Martin
What I listened for was the upper, 'air' region of the frequency response. The 16 bit seems to be slightly rounded off and lack detail. The highest frequencies produced by the strings is a good thing to focus on. In general I thought the 16 bit sounded a bit dampened and maybe even grainier, but not enough to help me with the ABX.
A few comments:-
1. Where perceived differences are very small, suspicion develops that the playback equipment may be solely responsible. For example, the very slightly higher total high frequency energy in the dithered version could be trigerring a small power drain in the laptop audio system: unlikely but possible. It will be more convincing if/when the results can be replicated by Martin on other equipment.
2. If the higher frequencies of the strings is where the difference is most noticeable, perhaps another recording of strings can be found to show the effect more obviously, permitting others to hear it.
3. Those well versed in the mathematics of dither may cringe at this, but I have always thought it unlikely that dither would be found ineffective for the lower signal frequencies. Intuitively, high frequency dither provides an ample number of random variations per cycle of a low frequency sine wave signal for a clearly defined instantaneous "moving average value" to emerge that defines a value between adjacent discrete steps in the 16 bit quantisation; for the purpose of driving a woofer, or a mid-range speaker in its lower frequency range [and independent of any sophisticated technique of waveform reconstruction].
4. The effort required to hear the difference apears to be considerable. Even if people other than Martin end up hearing differences, if that is only through intense concentration then the differences may not be "material". It might be that money would be better spent on things such as replacing recording microphones with more advanced microphones, or using better loudspeakers, than releasing recordings at 24-bits rather than 16-bits dithered. [However, we are really still at point number 1, above, i.e. waiting for tests on other equipment or by other people.]
5. Occasionally when I listen to a section of 24.wav it seems to sound a little crisper than the corresponding section of 24_16_fb2k_dither.wav, but my ABX results are hopeless. I cannot differentiate the two files with any degree of reliability. The two files are for practical purposes, for myself, identical.
Re: point 3...
yes, but, Martin first ABXed it with bog standard 16-bit TPDF dither, no noise shaping. I then added another 8 bits of noise to get it back to 24-bits. He still "nearly" ABXed it. I can't think of an equipment issue that would cause an audible difference between real 24-bit and 24-bit with the bottom 8 bits noise.
FWIW the other comments I've heard about 16-bits vs 24-bits is that it sounds "grungy".
Cheers,
David.
I've just posted something in another thread but I think it's relevant here too.
I'd love to see some ABX results. A lot of the listening tests that people refer to here are done on headphones - because it's easier to spot artifacts that way - but most music is not mastered with headphone listening in mind. I'm more interested in finding out if somebody can spot differences over speakers that aren't evident over headphones. Martin, did you say you were going to use studio gear for your comarison? Would that mean speakers?I look forward to seeing some results
Martin, did you say you were going to use studio gear for your comarison? Would that mean speakers?
Definitely. Not sure when I can get to that proper testing, could be as late as April, but I'd be happy to report back as soon as I've done some. One factor that might make it more difficult with speakers is small head movements. Speaker distortion is another of course. But the general resolution of the system will be much better, so it'll be interesting to see how it works out.
Martin
I've just posted something in another thread but I think it's relevant here too.
I'd love to see some ABX results. A lot of the listening tests that people refer to here are done on headphones - because it's easier to spot artifacts that way - but most music is not mastered with headphone listening in mind. I'm more interested in finding out if somebody can spot differences over speakers that aren't evident over headphones. Martin, did you say you were going to use studio gear for your comarison? Would that mean speakers?I look forward to seeing some results
If anything, I'd expect loudspeaker listening to be less conducive to discriminating differences than headphone listening...not more. What differences might be evident over them that wouldn't be evidence over loudspeakers ? (Assuming loudspeakers and headphones of comparable FR.)
If anything, I'd expect loudspeaker listening to be less conducive to discriminating differences than headphone listening...not more. What differences might be evident over them that wouldn't be evidence over loudspeakers ? (Assuming loudspeakers and headphones of comparable FR.)
I'm not expecting any specific differences but I don't listen via phones and since most music is mastered to be listened via speakers - as far as I'm aware - it seems a more relevant comparison to me
The listening experience of headphones is quite a bit different than loudspeakers. Of necessity, I do my LP transfers using headphones, and in some ways that might be easier, but various small percussion instruments, and some other transients, can be hard to distinguish from vinyl noise at times on the headphones. It isn’t that these sounds are less audible, they are just distinctly different.
Every once in a while I have to write my work to CD so I can use the living room system. Sounds that I was very uncertain about keeping are suddenly revealed very clearly to be, for instance, two wood sticks being struck together. Sounds of some other instruments sound as though they are really being produced by those instruments but on headphones it is nowhere near so clear.
Another thing: background noise, such as hiss or vinyl wear noise, will seem fully acceptable on headphones but very intrusive on the speakers. That I decide more NR is necessary after listening on speakers is obviously a personal preference, but the differences in sound, headphones to speakers, is pronounced. Then there is the fact that the sound stage and room ambience are not really there on headphones.
Some might believe this all indicates defective equipment but I don’t think so. The soundcard and headphone amplifier are good. I’ve used Sennheiser HD600s, Grado SR125s, Sony MDR-V900s and V6s, and occasionally others. All sound different from each other, but none are poor quality. I’ve discussed this with audio and acoustical engineers; it seems to be well know, accepted, and considered more or less explained. Mixing and mastering people rarely consider headphones to be of any value for their work.
None of this says that differences would be undetectable on one and detectable on the other, but the fact hardly seems obvious. Much music really sounds “better” to me on the speakers, but I still enjoy headphone listening; for some kinds of music I prefer the headphone experience.
I can’t think of how it would be possible to ABX test headphones against speakers, there are too many obvious clues as to which one is using. Unfortunately I also can’t do audio file ABX testing in the living room, and I have no intention of spending whatever it would take to allow me to operate a computer from my listening chair. These difficulties might lead some to maintain that differences are unproven, but it hardly seems reasonable to deny all experience that one cannot substantiate by tests.
1. Where perceived differences are very small, suspicion develops that the playback equipment may be solely responsible.
Always possible, but not very likely, I'd like to think that the playback equipment makes it even more difficult in this case. Why do you think these artifacts would not be tiny? After all, we're messing with a few LSBs here...
2. If the higher frequencies of the strings is where the difference is most noticeable, perhaps another recording of strings can be found to show the effect more obviously, permitting others to hear it.
Great idea! Hoping to record another clip specifically for testing.
4. The effort required to hear the difference apears to be considerable. Even if people other than Martin end up hearing differences, if that is only through intense concentration then the differences may not be "material". It might be that money would be better spent on things such as replacing recording microphones with more advanced microphones, or using better loudspeakers, than releasing recordings at 24-bits rather than 16-bits dithered.
Microphones are my main field, so I agree that they are very important and where we should spend most of the money :-) As for why this digital issue matters, it's because of two reasons IMHO:
a) what's assumed about audibility and human perception must constantly be updated, especially when it comes to anything nonlinear or discrete.
b) 16 bit PCM is hopefully not the final word in music or audio, and we need to learn more about what a better format should look like.
The two files are for practical purposes, for myself, identical.
Thanks for listening. It's
very difficult to hear, wasn't joking... But we
know for a fact the files are not identical, so the question is can we learn how to focus on and detect the difference well enough to get reliable ABX results. There are so many other factors at play than the files themselves, our listening environment to begin with.
Martin
Microphones are my main field, so I agree that they are very important and where we should spend most of the money :-) As for why this digital issue matters, it's because of two reasons IMHO:
a) what's assumed about audibility and human perception must constantly be updated, especially when it comes to anything nonlinear or discrete.
b) 16 bit PCM is hopefully not the final word in music or audio, and we need to learn more about what a better format should look like.
16-bit hasn't been the final word in music or audio for years now; virtually all pro recording and production is done in higher bit domains than that, for reasons
the we all should know by now (it's a *preventative* measure against distortion, not an outgrowth of an intrinsically better audible quality versus 16)
16-bit has been the most common final word*length* for home audio listening, and so far, the evidence that it's audibly inadequate to *that* task, in any
practical way,
is seriously lacking (this thread included). I don't buy the 'over the long term it may fatigue the listener' argument on current evidence -- and I've seen that claim for
mp3s too.
However, I would strongly argue for 24 bit nowadays
only because in many modern AVRs and processors, the CD signal will be reformatted to
24-bit anyway for DSP.
Appreciate your skeptical attitude.
16-bit hasn't been the final word in music or audio for years now
Of course it has, as a delivery format it has ruled for decades. Compressed audio is taking over now. And what I'm questioning is the next standard. What we do in the studio is a different story. CD did a good job, but is it the peak of audio quality delivered to the public? The bit rate chosen fit the disc technology available at the time, just as mp3 is perfect for internet connections.
16-bit has been the most common final word*length* for home audio listening, and so far, the evidence that it's audibly inadequate to *that* task, in any practical way,
is seriously lacking (this thread included).
Never said it's 'inadequate'. That's not the point as I see it. Let me rephrase, do you think there's any room or need for improvement over 16/44, or should we sit back, relax and be content with that audio quality?
I don't buy the 'over the long term it may fatigue the listener' argument on current evidence
Don't think I used the word 'fatigue'. What I miss is some of that "whoa, this sounds so great" type of primitive reaction and feeling in the music experience with CD. Don't know about you, but I react very strongly to good music, and even more so if the playback quality is very high. CD has never given me any of those top moments as I can remember. Still, it's the only playback format hooked up in my house at the moment. No vinyl, no 1/4 inch tape and mp3 is only for travel.
Martin
Don't know about you, but I react very strongly to good music, and even more so if the playback quality is very high. CD has never given me any of those top moments as I can remember.
What formats have?
Don't know about you, but I react very strongly to good music, and even more so if the playback quality is very high. CD has never given me any of those top moments as I can remember.
What formats have?
24b/96k mixes in the studio (or the sound straight off the console), some 1/4 inch master tapes and even some good vinyl (ouch!) would probably cover the recorded moments. But actually sitting in front of an artist or orchestra definitely gives those truly spine-tingling moments (even if I close my eyes.) Have heard gorgeous sound quality from a CD, but the physical reactions never seem as strong. This is HA, so please forgive me all this subjective stuff...
Before you start the flames, let me also say that I suspect an additional effect happening especially with vinyl. We are listening not only to a
reproduction. i.e. the recording, but also to a new sound
produced by the pickup and player. It almost becomes an electro acoustic instrument, which in a good system just adds to the musical excitement (and to the flood of information feeding our brain). Maybe.
Martin
It's ok. I will never get the idea out of my head that Y&T's Black Tiger will never sound better than it did on vinyl, with the slight crackle and all. I want to believe the memory this way and that's all there is to it. However, I also know that a 44.1/16 recording of the very same experience would not sound any different to me so long as I didn't already know which format I was hearing.
However, I also know that a 44.1/16 recording of the very same experience would not sound any different to me so long as I didn't already know which format I was hearing.
Sorry for asking but how do you know this? Have you actually blind tested it?
But as I said, if you capture the sound the record player
makes you would probably get pretty close. Have done some attempts to record the 'sound of vinyl' but so far have not really succeeded 100% with the studio quality converters I've tried. A bit surprising perhaps, maybe needs more testing. Please note that I'm talking top quality analog equipment here, not just any old scratchy LP and tired turntable.
Martin
Well, let's just say I'm seriously doubtful that I would be able. I cannot hear past 17-18kHz, so that half of the equation is already decided. This leaves the resolution in question, but keep in mind I'm talking about hard rock here. It's not exactly the most dynamic music.
There has already been a lot of discussion about vinyl here, and from what I recall reading, it isn't capable of providing better SNR than what 16-bit LPCM is capable of providing; regardless of the equipment used. Perhaps you know of some research that shows otherwise.
The listening experience of headphones is quite a bit different than loudspeakers. Of necessity, I do my LP transfers using headphones, and in some ways that might be easier, but various small percussion instruments, and some other transients, can be hard to distinguish from vinyl noise at times on the headphones. It isn’t that these sounds are less audible, they are just distinctly different.
Every once in a while I have to write my work to CD so I can use the living room system. Sounds that I was very uncertain about keeping are suddenly revealed very clearly to be, for instance, two wood sticks being struck together. Sounds of some other instruments sound as though they are really being produced by those instruments but on headphones it is nowhere near so clear.
Another thing: background noise, such as hiss or vinyl wear noise, will seem fully acceptable on headphones but very intrusive on the speakers. That I decide more NR is necessary after listening on speakers is obviously a personal preference, but the differences in sound, headphones to speakers, is pronounced. Then there is the fact that the sound stage and room ambience are not really there on headphones.
Some might believe this all indicates defective equipment but I don’t think so. The soundcard and headphone amplifier are good. I’ve used Sennheiser HD600s, Grado SR125s, Sony MDR-V900s and V6s, and occasionally others. All sound different from each other, but none are poor quality. I’ve discussed this with audio and acoustical engineers; it seems to be well know, accepted, and considered more or less explained. Mixing and mastering people rarely consider headphones to be of any value for their work.
None of this says that differences would be undetectable on one and detectable on the other, but the fact hardly seems obvious. Much music really sounds “better” to me on the speakers, but I still enjoy headphone listening; for some kinds of music I prefer the headphone experience.
I can’t think of how it would be possible to ABX test headphones against speakers, there are too many obvious clues as to which one is using. Unfortunately I also can’t do audio file ABX testing in the living room, and I have no intention of spending whatever it would take to allow me to operate a computer from my listening chair. These difficulties might lead some to maintain that differences are unproven, but it hardly seems reasonable to deny all experience that one cannot substantiate by tests.
I agree with almost everything you say. The listening experience IS totally different via headphones. I too do my LP transfers using headphones and I also don't have my PC hooked up to to my main listening system. The only place my experience differs from yours is background noise. I find noise much more noticeable on headphones and am have often been tempted to try to eliminate it only to find that it's not at all noticeable if I write the audio in question to a CD and play it back on my hi-fi. I'd always assumed that was beacuse I'm about 10 feet away from my speakers so maybe the noise doesn't "travel" as well as the music whereas on headphones the noise is being squrted straight into my ear
Mixing and mastering people rarely consider headphones to be of any value for their work.
Let me make an off topic comment here:
I don't know how mixing and mastering engineers really work, but I assume they sit somewhere around the sweet spot in front of two good quality stereo loudspeakers (or in a nice large room with 5 or more speakers!).
I
do know how consumers really "listen to" or (more often) "hear" music:
- They have car stereos where they're not sat in the middle of the stereo sound stage, and the lows, mids, and highs come from easily differentiable locations.
- They have mp3s players, most with terrible headphones, some with reasonable headphones, a few with great headphones.
- They have home stereos with the speakers placed 12" apart.
- They have home stereos with the speakers placed 6ft apart, but so close to the walls that they sound terrible.
- A very small number have speakers placed in a way that can give good stereo in a single sweet spot.
- ...and[/b[] however their speakers are set up, most people don't sit in the sweet spot to listen to music very often. If they're sat down at all, it's probably at some random location off centre, probably not even facing the speakers.
I don't know the proportion of each of these, though I see a
lot of people walking around with in-ear headphones on these days.
My (off topic!) point is that if people mixing music don't listen through headphones, they don't know what a huge part of their audience will hear. It amazes me that someone hasn't spotted the commercial opportunity here: mixes designed especially for headphone listening. They love selling the same music several times over - I can't believe they've missed out on this golden opportunity.
Cheers,
David.
Sorry for asking but how do you know this? Have you actually blind tested it?
There's quite a few people here who have copied many records onto CD (or at least, digital). AndyH-Ha and CliveB come to mind. I've done a few myself to (though I just play at it - my interest is more in 78s anyway).
The usual answer when someone asks is that it makes no difference.
Maybe we're all deaf, or don't know what to listen for - or maybe the magnitude of change between one record and another, or one cartridge and another, and certainly one declicker and another, is so great that any 44.1k vs 96k or 16-bit vx 24-bit differences are insignificant.
I can't speak for the others, but for myself, I really can't hear any difference between the record and the CD - unless I've processed it (declicking, denoising) to sound better.
Often I've thought I heard a difference when I stopped the LP and started playing the digital recording, and FWIW I've also thought I heard a difference between DACs too - but when I've level matched things, and done a simple live vs digital loop-through comparison, I've decided that I
can't hear a difference (and that was in a sighted comparison!). I think the difference I
thought I heard was either down to the audible noise coming straight off the stylus itself (which is present for both live and digital loop-through, but not when you play the digital file back!) and the physical experience of putting on a record and seeing it spin around (which again, is present for both the live and digital loop-through in my sighted test, but is not present when you click play on the digital file afterwards).
I did find a CD recorder that audibly altered the sound of the vinyl background noise (silent groove) when I fed it a signal via an external ADC - but then I found that the level matching for this comparison, done via the CD recorder's own level control, had introduced really nasty undithered re-quantisation. Setting that control back to unity, and adjusting the input level on the ADC instead, made the live vs digital loop-through difference inaudible to me.
Cheers,
David.
None of this says that differences would be undetectable on one and detectable on the other, but the fact hardly seems obvious. Much music really sounds “better” to me on the speakers, but I still enjoy headphone listening; for some kinds of music I prefer the headphone experience.
Agreed.
1. Where perceived differences are very small, suspicion develops that the playback equipment may be solely responsible.
Always possible, but not very likely, I'd like to think that the playback equipment makes it even more difficult in this case. Why do you think these artifacts would not be tiny? After all, we're messing with a few LSBs here...
I agree that the artifacts of the dithered 16-bit version (a) failing in some way to capture the nuances of detail of the 24-bit original or (b) introducing noticeable noise, can be expected to be tiny at ordinary listening levels. They are so tiny they will be below the threshhold of the human physiology of hearing, or perhaps only just perceptible. When the human ear is listening for such extremely tiny differences, any other differences even small ones, can contaminate the test.
One of the differences between 24.wav and 24_16_fb2k_dither.wav is that the second file is in 16 bit format. Some PC audio drivers permit a choice between 16 bit and 24 bit reproduction. Obviously the pc will need to be set for a mode that does not limit the sound to 16 bits.
A late model laptop could be expected to have a high definition capable sound chip and to produce
exactly the same analogue sound when playing a 16 bit file, as compared with playing the 16 bit file converted to 24 bits by adding zeros. But to ensure a good scientific method it would desirable to run some ABX tests to ensure the laptop does not play a 16 bit file even slightly differently to the same 16 bit file padded out with zeros to make a 24 bit file.
At this point in time for many conservative readers there will be doubt as to whether the equipment in use has not contributed to the perceived differences in sound in some subtle way. I recall that for an earlier test file in this thread 2Bdecided modified a 16 bit file by converting it to 24 bits through the addition of noise at bit positions 17 to 24. That technique encourages a pc to play back a 16 bit dithered version with
exactly the same "operating parameters" for the software/hardware as it uses to play a 24 bit original file.
____
A little off-topic:
Some months back I reported that I could ABX an 8,333Hz sine wave from an 8,333Hz sine wave plus third harmonic (24,999Hz). This met with considerable scepticism. To try to counter the sceptisim related to the experimental method, I resorted to using a portable cd player to play a continuous tone of 8,333Hz and used my main hi-fi system to play either a 24,999Hz tone, or silence, controlled by the ABX functionality in foobar. [The tweeters in my loudspeaker system could produce significant sound pressure at 25KHz, as verified with a microphone and oscilloscope.]
With the portable CD turned off, I could not hear any difference between silence and the 24,999Hz tone. (My hearing cuts out below 20KHz). But with the portable cd player turned on and playing the 8,333Hz tone, I could tell a difference. But only just! The ABX was extremely hard to do.
The presence of the 24,999Hz tone seemed to slighty reduce the amplitude of the 8,333Hz tone.
I was not directly hearing the 24,999Hz tone, but I was hearing an effect that tone was having on a frequency I could hear!
But at the end of the day, the effect was extremely subtle and I was not sure there were any practical implications.
I have a completely open mind on the 24-bit original vs dithered (noise-shaped) to 16 bit question, and look forward to seeing more ABX reports.
Cheers.
The usual question at this point would be, was this in fact an effect on the reproduction chain? For example, does the tweeter behave slightly differently in the presence of the ultrasonic signal in a way that affects the audible signal? Or perhaps due to a non-linearity there was a beat frequency just within your hearing range, but low enough that you were not conscious of it directly?
There's quite a few people here who have copied many records onto CD (or at least, digital). AndyH-Ha and CliveB come to mind. I've done a few myself to (though I just play at it - my interest is more in 78s anyway).
The usual answer when someone asks is that it makes no difference. Maybe we're all deaf, or don't know what to listen for - or maybe the magnitude of change between one record and another, or one cartridge and another, and certainly one declicker and another, is so great that any 44.1k vs 96k or 16-bit vx 24-bit differences are insignificant.
Only asked because I was interested in your experiences, and this is a good time to say how much I enjoy the high standard of this forum and talking to you guys!
In my limited vinyl experience the setup was probably as close to optimal as we can get, but I didn't do any blind testing, so it might be I only thought I heard a difference. Remember trying up to 192kHz sampling with no real improvement though.
Double blind testing puts everything in a new light, and that's why it's so exciting to me. Now I have to go back and investigate more than one thing properly. Speaking of which, I will need to build myself a good ABX rig (in hardware). Any suggested reading or links?
Sorry for taking this OT, but please let me say a couple more words about why I'm so interested in exploring the limits of our hearing and equipment, even if it doesn't seems to matter practically speaking. For some time now, I've used a bit of simple blind testing while working with audio electronics. It's interesting to see how the method can be used in reverse to make two circuits sound the same, or even to produce an audible, and confirmed, improvement. And in the end, all those small but real improvements add up to something that truly pleases the ear. The blind testing allows me to slowly but steadily work towards a better sounding solution.
I think the difference I thought I heard was either down to the audible noise coming straight off the stylus itself (which is present for both live and digital loop-through, but not when you play the digital file back!) and the physical experience of putting on a record and seeing it spin around (which again, is present for both the live and digital loop-through in my sighted test, but is not present when you click play on the digital file afterwards).
Yes! With a reel-to-reel you also have another visual 'signal' but not the stylus noise.
Martin
Never said it's 'inadequate'. That's not the point as I see it. Let me rephrase, do you think there's any room or need for improvement over 16/44, or should we sit back, relax and be content with that audio quality?
The latter. As delivery format SR and wordlength, 44.1 and 16 are good enough. They are emphatically NOT the 'problem' with home audio repro. The real improvements in audio quality are going to come from multichannel recording, mixing and delivery technologies, combined with digital 'room correction' -- not higher SR and wordlengths (though both of those are *already* involved in much playback, since virtually all CDPs oversample, and virtually all modern AVRs perfrom DSP, at 24 or "32'" bits)
Don't think I used the word 'fatigue'. What I miss is some of that "whoa, this sounds so great" type of primitive reaction and feeling in the music experience with CD. Don't know about you, but I react very strongly to good music, and even more so if the playback quality is very high. CD has never given me any of those top moments as I can remember. Still, it's the only playback format hooked up in my house at the moment. No vinyl, no 1/4 inch tape and mp3 is only for travel.
I have music on CD that brings tears to my eyes, it moves me that much. And I grew up listening to vinyl (and still have a fairly decent vinyl playback rig). i got a kick out of it, too. As well as the Beatles from a tiny transistor radio in '65-68. As well as live performances I've attended, and performed.
David,
You are probably aware of, if not personally conversant with, binaural recordings, intended particularly for headphone listening. This depends more upon recording technique that mastering. It seems to be a pretty small market.
Proper studio monitor setup is near field, very specifically set-up to avoid room interactions as much as possible. This involves dampening the room properly, as well as positioning everything correctly.
I believe the general consensus is that anything mixed and mastered on a proper set-up will sound as good as possible in the maximum number of listening environments. One of the most common complaints of beginners who didn’t pay attention to established wisdom is “after so much time and effort getting my mix to sound perfect on my computer speakers, it sounds terrible in the living room/car, wherever.” Such mixes don’t travel well.
I think part of the loudness war approach is a compensation (however poor from some viewpoints) to deal with some of today’s listening practices.
Andy,
Thanks for that explanation.
btw, I made a binaural simulation in DSP "many" years ago now - it was my second audio DSP project (the first was a dithered volume control). It's surprisingly difficult to get right, and to make it sound _better_ than "normal" stereo - even when listened to via headphones. However, I'm still surprised that someone doesn't do slightly different mixes for, say, iTunes. As an extreme example, you can imagine lots of the Beatles catalogue could be made to sound far nicer over headphones than it currently does (though there are plenty of reasons why getting the Beatles "remixed" is "unthinkable", even though it has happened several times).
Cheers,
David.
I was going to put this in a PM, but decided to post it here instead...
krabapple, I think that's a healthy dose of reality. No one has to even think about ABXing stereo vs surround, or good vs terrible speakers, or even different microphone / mixing techniques. The audible difference is really obvious. In the case of "good" surround of "good" speakers, the audible advantage is obvious, even without a "bad" reference for comparison in the same room.
I, like you, have music that can move me to smile or cry on various formats - and in some cases it's nothing to do with "association" or lyrics or anything - it's just the sound of the music, more than adequately conveyed by various media. Sometimes it's even the recording itself with all its limitations that is so pleasing (e.g. the specific "sound" of 1930s or 1960s "pop" recordings).
That shouldn't stop us tracking down any possible issues with 16-bits or 44.1kHz - some people who believe they hear differences describe them as significant once identified - and now someone appears to be able to ABX a difference. I keep an open mind.
But for me, 44.1kHz 16-bits doesn't stop me having an emotional connection with the music on various levels.
Last night I was listening to Rachmaninoff's 3rd piano concerto via 192kbps discrete stereo mp3. I don't think the power of the music was diminished at all, even though the sound of the piano wasn't adequately conveyed by my little ear buds.
Cheers,
David.
Is part of listening to music which is both familiar to us and moves us not based to some extent on expectation - a bit like re-reading a book?
The usual question at this point would be, was this in fact an effect on the reproduction chain? For example, does the tweeter behave slightly differently in the presence of the ultrasonic signal in a way that affects the audible signal? Or perhaps due to a non-linearity there was a beat frequency just within your hearing range, but low enough that you were not conscious of it directly?
pdq, you may recall my providing details in May 2008. This litle experiment of mine was described beginning at post #100 (http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=63123&view=findpost&p=567108) of the thread
"Resampling down to 44.1KHz, Is there a method that will not colour the sound?". I think these facts are relevant to your query above:[blockquote]
* For initial tests the left and right channels of the hi-fi stereo were fed phase locked 8333Hz and 24999Hz and I could hear the effect I have described (change in apparent volume, and also a slight change in apparent tonal quality). A later test with an older AVR and a portable CD player creraed the effect, and the third harmonic was phase locked.
* For some other tests the portable CD player playing 8333Hz was several metres away from the hi-fi speaker playing 24999Hz. There was therefore no acoustic coupling between the transducers (speakers) that were creating the different source frequencies. Totally independent reproduction chains. [And not phase locked.]
* Attempts to generate audible beats by playing 22KHz and 23Khz from separate speakers to create a 1KHz beat, or closer together ultrasonic frequences to create a lower frequncy audible beat failed. However some much lower frequencies succeeded. For example, playing 2500Hz and 3000Hz through separate speakers produced an audible beat of 500Hz. This phenomenon has been described in the literature and has been explained as due to non-linearity of human hearing. The effect was slight; the 500Hz was barely audible and I think I would have missed it, had I not been specifically listening for it.[/blockquote]This discussion is now seriously off-topic! I guess you could post to the old thread if you wished to continue it. Cheers
Is part of listening to music which is both familiar to us and moves us not based to some extent on expectation - a bit like re-reading a book?
Your remark/question reminds me of the fact that if on a particular occasion we listen wth poor equipment to a recording we are familiar with, our minds may fill in a lot of the gaps, based on our knowledge of the sounds of the recording as we
previously heard it, on better equipment.
If I listen to a necessarily poor quality recording of Enrico Caruso made in the period 1910 to 1920, my mind inserts some of the missing detail because I have developed a mental construct of Caruso's voice after listening to various recordings available, but also based in part on my knowledge of how operatic tenors typically sound today. I extrapolate. This is a type of listening ability we would all have to some degree. We use our imaginations when listening to music.
If I had ever heard Caruso live, my brain would be able to insert much more of the missing detail when listening to the poor quality recordings. But lacking that, it is nevertheless not hard to imagine he had the greatest operatic tenor voice of the 20th century, a claim often made; despite the very restricted frequency response, and high distortion, of the surviving recordings.
Is part of listening to music which is both familiar to us and moves us not based to some extent on expectation - a bit like re-reading a book?
No doubt...but that doesn't explain why sometimes a brand new (to me) piece of music catches my ear and gets me 'involved' -- a phenomenon that absolutely does not
require 'high fidelity' playback conditions (the last time it happened was in a restaurant -- the staff were playing tracks from their iPods over the house system).
I've been thinking of a similar framework about audio to what krab has been saying for a while, but I'd take what he's saying a step further. (Me being more militant than krab? Yikes, what is this world coming to?)
I would go so far as to say that we've been a little too accepting and tolerant about this 16/44 issue for too long. A lot of people now - not even audiophiles - regularly denigrate Red Book as being a fundamentally inferior to vinyl/high res. That it is a less emotional experience, that a generation has been cheated out of good music, that (insert fallacious technical argument here), etc. Steven Van Zandt just claimed on an interview (http://www.bowers-wilkins.com/display.aspx?infid=4063&utm_source=bowerswilkins&utm_medium=email&utm_content=us&utm_campaign=sosnews090212) that CDs are "the biggest scam perpetrated on the public". Such people make Bob Dylan look like a beacon of reason.
It's really hard to fathom exactly how elitist and ignorant these positions really are. What these people are doing, it seems, is performing a value judgement on my musical experiences. That even though I've had plenty of great experiences with CD music, those experiences are automatically less meaningful because it's on a disfavored format (for reasons that no person has ever technically, competently articulated). That had I not listened to vinyl, I would always be missing out on the "real" way to listen to a mass-marketed, commoditized, overpriced, eternally predistorted and precompressed product that never had any authenticity to begin with.
No. F*ck them. We as music listeners shouldn't stand for that.
Almost any format is acceptable listening for music, including AM for some genres, and I've had wonderful experiences with all of them. (Raise your hand if you listened to Freebird for the first time on the car radio and almost ran off the road.) CDs dominate, but not merely because of market inertia and PR - they happen to be relatively small, cheap to manufacture, be cheap to produce, and have great-but-not-theoretically-ideal performance. Those are all good things for the distribution of commoditized mass media. But it seems that many people do not like to be reminded how sheep-like they are by trying to avoid the pack, so they abandon one commodity medium for another. We call these people "hipsters". Another name for them is "stupid lemmings".
---
On a similar note regarding downloads. I've downloaded gobs and gobs of free classical music online over the last few months. Podcasts from Deutsche Welle, the Royal Concertgebouw 120th downloads, Schubert lieder, Peabody Symphony Orchestra, .... The list goes on and on. If I'm lucky, it's 320k, otherwise it's like 192k CBR or something like that.
I will say it now and I will try to state it in equally insulting words in the future: Anybody who claims that is a step backwards from 5, 10, 20, 50, or 100 years ago is a f*cking douche who does not know what he/she is talking about, and whose personal experiences with music are dead to me.
That said, a surprising number of audiophiles do get that. Remember Todd Krieger? He hearts the YouTube. (http://db.audioasylum.com/cgi/search.mpl?searchtext=youtube&b=AND&topic=&topics_only=N&author=Todd+Krieger&date1=&date2=&slowmessage=&sort=score&sortOrder=DESC&forum=music) He's still a douche, though.
It's partly snobbery.
How can anything that costs 10pence to make possibly be good enough for those $50k audio systems?!
There's an emotional need in some users, and a financial need in some parts of the industry, for CD to be "not good enough".
A word of warning though: very few people appreciate mp3 artefacts the first time they hear them, but many people subsequently learn to recognise them, be annoyed by them, and try to eliminate them. This could be just another type of artefact that people may either learn to hate, or learn to put up with, depending on their attitude and needs.
Cheers,
David.
99% of audiophiles wouldn't know a 128k artifact if it shot them in the ass. "They" (if I may continue to entertain myself by stereotyping) talk a lot about a loss of detail or "air" or whatever in MP3s that is supposedly easy to recognize. Those sorts of statements are pretty easy to dismiss out of hand for those of us who have actually heard high bitrate artifacts.
That shouldn't stop us tracking down any possible issues with 16-bits or 44.1kHz - some people who believe they hear differences describe them as significant once identified - and now someone appears to be able to ABX a difference. I keep an open mind.
Will happily admit that my interest in the possible limitations of 44/16 is 'academic' at this point. Keep producing and consuming the very format while I explore further.
But for me, 44.1kHz 16-bits doesn't stop me having an emotional connection with the music on various levels.
Me neither, what I said was that the few 'top' moments never seemed to be associated with CD playback. Just an observation. And the other observation is that the home listening experience has become less and less interesting to me over the years. But it could be for a number of reasons, I might need another room, amplifier or speakers or just another type of daily schedule to once again enjoy music as I used to.
Last night I was listening to Rachmaninoff's 3rd piano concerto via 192kbps discrete stereo mp3. I don't think the power of the music was diminished at all, even though the sound of the piano wasn't adequately conveyed by my little ear buds.
Yesterday I took a long walk on the beach, listening to my ipod. In my case. the experience was greatly enhanced by the fact that I had new Monster ear buds to try out. Yeah, the music experience was good, but there's no doubt in my mind that it could have been better too. That's the dangerous audiophile disease in a nutshell, always looking for a better experience, while the music itself stays the same...
But as you say, the mp3 format was probably not the limiting factor at this level.
Martin
...What I miss is some of that "whoa, this sounds so great" type of primitive reaction and feeling in the music experience with CD. Don't know about you, but I react very strongly to good music, and even more so if the playback quality is very high. CD has never given me any of those top moments as I can remember...
There are a lot of people interested in preserving old Beatles records out there right now (with a particular emphasis on Stereo vs. Mon, Solid-State vs. Tube, etc.). I've listened to a friends drop of Sgt. Pepper's -1/-1 original Yellow & Black Parlophone Tube Cut (borrowed copy mind you). I believe it was taken in 24/96 and dithered down to 16/44.1 with some nice software. The sound of that CD blew me away.
In fact, investing in a cheap amp and a decent set of 'phones has caused most of my CD collection to really rock (most from the pre-loudness wars).
I've been thinking of a similar framework about audio to what krab has been saying for a while, but I'd take what he's saying a step further. (Me being more militant than krab? Yikes, what is this world coming to?)
I would go so far as to say that we've been a little too accepting and tolerant about this 16/44 issue for too long. A lot of people now - not even audiophiles - regularly denigrate Red Book as being a fundamentally inferior to vinyl/high res. That it is a less emotional experience, that a generation has been cheated out of good music, that (insert fallacious technical argument here), etc. Steven Van Zandt just claimed on an interview (http://www.bowers-wilkins.com/display.aspx?infid=4063&utm_source=bowerswilkins&utm_medium=email&utm_content=us&utm_campaign=sosnews090212) that CDs are "the biggest scam perpetrated on the public". Such people make Bob Dylan look like a beacon of reason.
I just love it (which is to say hate) when the Little Stevens and the Tom Scholzs and the Neil Youngs of rock, who between them probably no longer hear much past 12 kHz, expect anyone to take them seriously as 'golden ears'.
(Dylan seems to have been talking about dynamic range overcompression, not something evil about digital per se, so he gets a pass.)
Was it ever really about hearing? Their words are taken at face value because listeners generally respect the emotional honesty of musicians. People have favorite guitars or pianos or mics, sometimes beyond all rational explanation, and most people have absolutely no problem with that kind of subjectivism. I mean, they're freaking artists, they're expected to be weird.
I guess this is one manifestation of a larger issue of artists (especially actors) being unusually authoritative in their wacked-out philosophies. People who speak to us through mass media are very persuasive. If you want to take that sort of thing on, dealing with the actual technical truth of it all is a tar pit, because that's usually not what resonates with people, and even when it is, there will always be some other reason to justify the belief. The reality is that the importance of their emotional experiences are suspect to begin with. Few sane people should take Van Zandt any more seriously - on any topic - than those of any other random joe. This format arguments are about emotions - not psychoacoustics - but the facts are still on our side in both cases.
Mods: this is probably worth a thread split.
Is part of listening to music which is both familiar to us and moves us not based to some extent on expectation - a bit like re-reading a book?
Good point.
When we are listning to music we can make two different kinds of comparisons.
(1) Our brains maintain a short (ca. 2 second) memory of the actual sound as we heard it. Within a few seconds of hearing another sound, we have some abiltiy to compare the immediate sound to the sound we remember from the very recent (few seconds ago) past.
(2) Over any longer periods of time, our memory is not of the sound. Instead we remember our perceptions of the sound. This is highly distilled information that generally does not include small changes to things like frequency response, noise, or nonlinear distortion. It is things like chords, shapes of tunes, key, instruments used, associations with otther muisic like tit, etc.
When we casually hear music, we are probably primarily conscious of our perceptions of the music, and not any subtle details. The more familiar we are with the music, the less likely we are to pay much attention to the sibtle details.
This business of our perceptions being of only the highlights or a summary of the experience is probably more clear to us when we think about acts we repeat frequently, like driving home from work.
Is part of listening to music which is both familiar to us and moves us not based to some extent on expectation - a bit like re-reading a book?
Good point.
<SNIP>
(2) Over any longer periods of time, our memory is not of the sound. Instead we remember <SNIP> things like chords, shapes of tunes, key, instruments used, associations with otther muisic like tit, etc.
It is childish to delight in typos, but I'm in my second childhood, and anyway it points to a very important point about response to music, which is that it is linked to lots of non-musical things. Which is why, I think, the desire to make reproduced audio "sound like a live performance" is not the best target, since it's not really attainable, since the live performance experience is a melange of a lot of things, many not audible.
It also explains why the popular music of my youth means more to me than the popular music of today; not because I think the Rolling Stones, say, were better than many good bands today, but because of associations with things "like tit, etc."
Martin, could you try to abx the original 24.wav file vs this one?
http://www.kikeg.arrakis.es/dither16bit/24_nf_dither.flac (http://www.kikeg.arrakis.es/dither16bit/24_nf_dither.flac)
I have employed a custom noise shaping dither that tries to have same spectral contour as the noise floor of the original 24-bit file.
Also, just in case, make sure that in foobar2000 Playback/Output format is set to 24 bit and that in ABX component DSP and replaygain are disabled.
Thanks.
Steven Van Zandt just claimed on an interview (http://www.bowers-wilkins.com/display.aspx?infid=4063&utm_source=bowerswilkins&utm_medium=email&utm_content=us&utm_campaign=sosnews090212) that CDs are "the biggest scam perpetrated on the public".
Thanks for the link. I especially liked the part where he says that "they" aren't making analog tape "very much anymore". Because, you know, I can't walk down the street (here in L.A., anyway) and buy as much 2" tape as I could ever need. Quantegy may be out of the game for now, but RMGI and ATR are still pumping out more reels on a daily basis than anyone knows what the hell to do with.
Amazing that this guy's radio show has some two million listeners
Martin, could you try to abx the original 24.wav file vs this one?
Just saw your post and downloaded the file, will check it out as soon as I can, thanks.
Martin
I just tested KikeG's dithered file against the original. There's a slight volume difference in some sound at the background (I don't know what makes the sound) in range 0:03.8 - 0:05.1. I used X-Fi in audio creation mode using bit-perfect output settings with WASAPI output and Sennheiser HD 650 headphones. Not completely clean results but enough to convince me that I hear a difference.
foo_abx 1.3.3 report
foobar2000 v0.9.6.5 alpha 2009-04-02
2009/04/04 15:53:48
File A: G:\Download\24_nf_dither.flac
File B: G:\Download\24.wav
15:53:48 : Test started.
15:58:41 : 01/01 50.0%
15:58:49 : 01/02 75.0%
15:59:37 : 02/03 50.0%
16:01:53 : 03/04 31.3%
16:03:10 : 04/05 18.8%
16:03:37 : 05/06 10.9%
16:06:14 : 05/07 22.7%
16:07:10 : 06/08 14.5%
16:07:50 : 06/09 25.4%
16:08:45 : 07/10 17.2%
16:09:11 : 08/11 11.3%
16:10:03 : 08/12 19.4%
16:11:05 : 09/13 13.3%
16:12:55 : 10/14 9.0%
16:14:14 : 11/15 5.9%
16:14:22 : Test finished.
----------
Total: 11/15 (5.9%)
Not completely clean results but enough to convince me that I hear a difference.
Well, those results are still completely within the domain of self-deception.
If you could hear a difference, you would be able to nail it down and steadily lock in at <1% levels. I have carried out very many ABX tests. Sometimes everything seems to be quite obvious and you head straightly into <5% territory without trouble. But when I have learned one thing about ABX testing: Go the extra mile.
Try to nail it down below 1%. If you can't keep your results steadily below that, the probability is very high (much higher than the displayed percentage), that you are just fooling yourself.
For people only now joining this thread, we are still awaiting ABX reports of comparing
24.wav with a
dithered version of 24.wav :-1. 24.wavThis is the first file, labelled "24-bit 48KHz source", in Martin's link in his post #76, and it has the filename
24.wav :-
Greetings, found your forum recently, so here's my first post! Wanted to share another set of files that might be useful for listening tests. The source material is a 'raw' unprocessed master recording downsampled from 96kHz. Uploaded the files here:
Digital audio resolution test files (http://www.nordicaudiolabs.com/dart/)
2a. Dithered version 24_16_fb2k_dither.wav This foobar dithered version can be downloaded using David's link in his post #145:
...
Here's the direct download link:
http://www.mediafire.com/file/zaiiuyx4izd/...fb2k_dither.wav (http://www.mediafire.com/file/zaiiuyx4izd/24_16_fb2k_dither.wav)
That's the one to try to ABX against 24.wav from Martin.
...
2b. Alternative (custom noise shaped) dithered version 24_nf_dither.flac Or, if preferred, use the custom dithered file supplied by KikeG at his post #205 just above.
Try to compare 1 with 2a; or 1 with 2b. Any difference should be
extremely difficult to hear, if audible at all. Please explain in what way the dithered file sounds different. Nil reports are also welcome, i.e. not able to ABX because the files sound too much alike.
so 16/44.1 libraries can safely be used with 24/192 libraries along with a lot of processing (reverb etc...)?
No earthquakes, fires, or floods reported so far, nor anyone able to tell any difference in the music.
No earthquakes, fires, or floods reported so far, nor anyone able to tell any difference in the music.
The 96k24b music file's spectrum shows minimal content > 24 KHz. That raises serious quesitons about whether or not it has been previously been low pass filtered by 48 KHz digital recording or legacy analog tape. In iether case it might have also been recorded with 16 bits or less, and/or had its dynamic range degraded by the analog recording technology that was possibly used.
In short, I am very uncertain as to whether there is any signfiicant difference to hear when downsampled to 16/44.
No earthquakes, fires, or floods reported so far, nor anyone able to tell any difference in the music.
The 96k24b music file's spectrum shows minimal content > 24 KHz. That raises serious quesitons about whether or not it has been previously been low pass filtered by 48 KHz digital recording or legacy analog tape. In iether case it might have also been recorded with 16 bits or less, and/or had its dynamic range degraded by the analog recording technology that was possibly used.
In short, I am very uncertain as to whether there is any signfiicant difference to hear when downsampled to 16/44.
Can you explain this in simpler ways? Are you saying that that 92/24 recording is a bad sample and does not represent what the standard could otherwise achieve?
While that is what I get as his point, the real question is whether it could possibly make any difference anyway. There is the issue of dishonesty and misrepresentation, of course, but the main concern for this thread is: can one make a "higher resolution" recording that loses something audible if it is (properly) down sampled to 16/44.1? All the multitude of recordings that are audibly indistinguishable are irrelevant, regardless of why they are indistinguishable.
While that is what I get as his point, the real question is whether it could possibly make any difference anyway. There is the issue of dishonesty and misrepresentation, of course, but the main concern for this thread is: can one make a "higher resolution" recording that loses something audible if it is (properly) down sampled to 16/44.1? All the multitude of recordings that are audibly indistinguishable are irrelevant, regardless of why they are indistinguishable.
What about samples recorded at 44.1/16? A lot of the older libraries (sampletekk) were recorded at that resolution.
Since the question seems to be directed at my last post, I will say that I don't understand it. This thread is a request for any 24 bit music that can be distinguished (ABX test) from a 16 bit version of same, which is to say, by reducing the 24 bit resolution to 16 bits, something audible is lost. Anything recorded at 16 bit is a non-starter to begin with.
the main concern for this thread is: can one make a "higher resolution" recording that loses something audible if it is (properly) down sampled to 16/44.1? All the multitude of recordings that are audibly indistinguishable are irrelevant, regardless of why they are indistinguishable.
If the the theory (that one can make a "higher resolution" recording...) is to be proven correct, it's easy: we just need one distinguishable case. If it is to be proven incorrect, then (given that a logical proof seems out of the question) we need to evaluate recordings having characteristics that render them (based on things that we do know) possibly distinguishable, and if enough of them prove indistinguishable then we can deem, with a reasonable degree of certainty, the theory to be incorrect.
So negative results are relevant, if made on suitable test samples—Arnie's point is that this is not a suitable test sample.
If you have decent equipment and a quiet environment, it should be easy for you to distinguish this test file I created (source 48khz, 24-bit) to a 16-bit conversion.
It's 5 seconds long, and under 100KB. Very low-level pure tone(sine wave) @ 3.5khz (since this is where our hearing is most sensitive).
Test Tone File(48/24) (http://www.sendspace.com/file/9w4hm1)
Convert it to 16-bit using any method you choose, any dither type of any level, I could abx it easily. I have already tried using at least a few dozen combinations of noise shaping and dither levels, along with no dithering (which obviously sounds worse). Always ABX'able. When properly dithered, the tone is still audible at 16-bit, but the noise floor is much worse and the tone becomes more difficult to distinguish from the noise (where it sounds fine at 24-bit).
You need fairly high output levels to do this, but since the file contains no peaks or anything that would resemble "loud", you wont risk damaging your hearing or equipment so long as no other "computer sounds" play during testing.
[Edit: If anyone prefers, here is a 44.1khz version of the same file, created fresh, not via SRC]
Test Tone File(44.1/24) (http://www.sendspace.com/file/9t9w9x)
Both links seem to be broken for me.
Guys, I took the 24.wav file from above and compared it to the 16-bit file with fb2k dither (with noise shaping it seems):
(http://img17.imageshack.us/img17/6695/dart24vs16fb2k.th.png) (http://img17.imageshack.us/img17/6695/dart24vs16fb2k.png)
Red is the spectrum of the track, green the difference between both files.
Also see my post (http://www.hydrogenaudio.org/forums/index.php?showtopic=88788&st=50&p=757236&#entry757236) in the Ripping Vinyl thread.
(I used simple rectangular dithering without noise shaping there..)
If the the theory (that one can make a "higher resolution" recording...) is to be proven correct, it's easy:
I still don't understand your point. It is easy to make higher resolution recordings. Their data can be "distinguished" in a number of ways. But, is it possible to hear something in real music, recorded at 24 bits, when listened to at a useable volume level, that sounds different after conversion to 16 bits (doing a quality conversion)? Some part that is too low level to be heard without extreme amplification is irrelevant via a vis the 16/24 bit controversy. We are not here listening for evidence of technology emissions around other stars.
Perhaps this thread has been used for various purposes other that my question. As far as the bit Arnold just quoted goes: I can generate low level sine waves in floating point, then convert to 16 bit in various ways with different listening results, but so what?
For instance, a 3.5kHz tone that measures
Min Sample Value: -.06
Max Sample Value: .06
Peak Amplitude: -115 dB
Possibly Clipped: 0
DC Offset: 0
Minimum RMS Power: -118.02 dB
Maximum RMS Power: -118.02 dB
Average RMS Power: -118.02 dB
Total RMS Power: -118.02 dB
I can't tell any (audible) difference when converted to 16 bit in the way I normally do it (with good noise shaped dither) If, however, I convert with rectangular dither, no noise shaping, identification is easy. For any version, the amplifier must be turned up to what would be ear damaging levels on any real music, so this kind of signal is not relevant when talking about music.
The same listening test results are true with the same tone just below the 16 bit resolution
Min Sample Value: -.33
Max Sample Value: .33
Peak Amplitude: -100 dB
Possibly Clipped: 0
DC Offset: 0
Minimum RMS Power: -103.02 dB
Maximum RMS Power: -103.02 dB
Average RMS Power: -103.02 dB
Total RMS Power: -103.02 dB
except that I can hear it at a volume setting that just might be useable for some, if very little, music. But even if it can be heard, the different sound of a poorly done conversion isn't evidence of anything useful. Of course I cannot say what other people experience, but it seems unlikely to me that the properly converted versions could not be readily identified by how they sound, even if something like such a tone occurred in some music.
×wires I think: I wasn't trying to redefine the theory, only to avoid quoting it in full (hence the ellipsis!). Given that we're almost 3 years down the line, it seems that music that benefits from >16-bit is either non-existent or so rare and difficult-to-discern as to be inconsequential.
×wires I think: I wasn't trying to redefine the theory, only to avoid quoting it in full (hence the ellipsis!). Given that we're almost 3 years down the line, it seems that music that benefits from >16-bit is either non-existent or so rare and difficult-to-discern as to be inconsequential.
In practical, pragmatic terms, yes.
If you want to apply scientific rigour, I dont think that conclusion is supported. We know little (?) about the selection of listeners, playback systems etc in this test. Listening tests allows us to "prove" (within some level of significance) that something is audible, but not prove that it is inaudible. Further, listening tests can be used for proving conscious differences, I dont think that typical listening tests are good at proving potential subconscious long-term effects (commonly used argument for those clutching at straws in various fields).
-k
I think there is much more variability in the listener than in the object. When I am in the proper mood, music can be extremely pleasant. When I am not in said mood, the same music can be tedious, or even an irritation of considerable magnitude. I like different types of music at different times, and of those types, some fairly often, some rather infrequently.
Often, poor equipment (many automobile sound systems, boom boxes, and cheap department store sets) make me just want to get away from them as quickly as possible, but if the stars are just right, aligning my life force towards the proper constellation, perhaps, I don't care much about the quality of reproduction.
I've worked intensely on quite a few albums ("restoring" LPs) many hours, listening for the smallest defects, without having any idea what the album is about. It isn't until I just listen to the final product that I have can tell if I like the music, or, if it has vocals, what the lyrics may be about. On the other hand, when I listen for enjoyment, I may occasionally become involved in a halfway poor condition LP without noticing much about its deterioration.
When I opened this thread I was interested to find something that was somehow better for being in 24 bit, there being many people, if one reads in the right places, who seemed convinced of it. Now I am convinced that it probably doesn't exist, but if it does, it is irrelevant. Straining to pickup up such find details as might be audible only in 24 bit isn't listening to music, it is just a hearing test. For me, music is about the complete experience. This is no doubt made up of the details, but it isn't about the details, and many of those details don't reside in the recording anyway.
foo_abx 1.3.4 report
foobar2000 v1.1.11
2012/04/07 12:52:17
File A: C:\Work\Render\ABX-16 bit.wav
File B: C:\Work\Render\ABX-24 bit.wav
12:52:17 : Test started.
12:54:54 : 00/01 100.0%
12:56:38 : 00/02 100.0%
12:58:46 : 01/03 87.5%
12:59:25 : 01/04 93.8%
13:01:28 : 02/05 81.3%
13:01:58 : 02/06 89.1%
13:02:52 : 03/07 77.3%
13:04:28 : 04/08 63.7%
13:06:48 : 05/09 50.0%
13:07:35 : 06/10 37.7%
13:08:05 : 06/11 50.0%
13:09:34 : 07/12 38.7%
13:10:30 : 08/13 29.1%
13:12:23 : 09/14 21.2%
13:13:54 : 10/15 15.1%
13:15:40 : 11/16 10.5%
13:18:36 : 11/17 16.6%
13:19:21 : 12/18 11.9%
13:20:31 : 12/19 18.0%
13:21:54 : 13/20 13.2%
13:21:58 : Test finished.
----------
Total: 13/20 (13.2%)
Very hard to distinguish - but probably just possible - as long as ear fatigue does not set in. Test music was 2 minutes from 24 bit live recording of 1st movement of Havergal Brian Gothic Symphony. Recording went to -3 dB - this section never went above -35 dB. ATH-M50 headphones - listening level boosted slightly. ASIO driver for EDIROL UA-101.
Comments:
1 Pure guesswork at beginning of test - I thought this was going to be easy ...
2 When I started concentrating - things got better - in this test run
3 Level to hear differences was way above normal conditions. The end of this movement would be simply unbearable at this volume.
4 After test 16, I removed the headphones for 20 seconds - this broke the concentration on the differences.
5 I got bored listening to the same bit of music over and over again ...
Conclusion:
May be possible - effect probably enhanced at ridiculous listenings levels. At normal levels, does it matter?
24 bits is distinguishable from 16 bits when only ~12 or less bits of the 16 are being used and the volume is boosted, generally to levels where a full 16 bits would likely surpass the threshold of pain. Nothing new here.
On its own, however, the log does not rule out complete guesswork throughout the entire test as it is far from impossible to guess a coin toss correctly in six consecutive trials.
So in general as I understand CD good enough even for excellent stereo audio systems.
The only point that CD mastering must competent.
For the most part 16/44 fully covers end-product audio in the human-audible sense. A properly converted recording to 16/44 will most likely be indistinguishable from an original master recording at 24/96 stereo given there is no other processing involved and the content has typical volume (i.e. RG value around 0 to +10 dB is my conservative guess, maybe higher).
If I understood the hint about competent mastering, I agree that making the overall volume louder (i.e. RG values around -8 to -10dB) does not fully exploit the benefits of 24/96... or even 16/44 for that matter (/rant).