Hydrogenaudio Forums

Hydrogenaudio Forum => General Audio => Topic started by: tigre on 04 September, 2003, 03:36:53 PM

Title: Test 24/96 vs CD resolution
Post by: tigre on 04 September, 2003, 03:36:53 PM
Hi.

At afterdawn forums there is a new high resolution audio forum. Wilkes, a guy who has much experience in this field (e.g. rips vinyl to DVD-A AFAIK) thinks he can hear differences. So I asked him to do a blind test and he said he's ready.

My idea how this should be performed:

1. Take some audio sample(s) recorded at 24/96

2. Apply fb2k's scan per-file tack gain to avoid clipping and too low volume.

3. Create "A" sample(s) by using fb2k's diskwriter, 24/96 resolution, trackgain used.

4. Convert "A" to "A_downsampled": 16/44.1 using fb2k's diskwriter, noiseshaped dither, slow resampling

5. Create "B" by "A_downsampled" to 24/96 using fb2k's diskwriter, slow resampling, dither not important.

6. Use KikeG's fileABX (http://www.hydrogenaudio.org/forums/index.php?showtopic=12387&) to create randomized files from "A" and "B"

7. Author a DVD-A using the randomized files

8. Try to ABX


This should be about the capabilities of the formats, not about quality of DACs, mastering quality of different formats etc.

I'd like to know your ideas and suggestions of you about this.

2 1/2 Questions: Will fb2k's resampler give good results? Especially I'm concerned about downsampling and the filters applied there (-> aliasing). Or would it be better to use e.g. CEP at least for downsampling, maybe even to apply a lowpass manually? (which one?)

Thanks.

Cheers tigre
Title: Test 24/96 vs CD resolution
Post by: Patsoe on 04 September, 2003, 04:21:47 PM
Wouldn't it be easier to skip the resampling, and just do lowpass filtering plus dithering 8bit deep on file A?

Edit: needless quoting removed.
Title: Test 24/96 vs CD resolution
Post by: tigre on 04 September, 2003, 04:42:12 PM
Thanks for the input. Certainly it would be easier but
- One important thing from the theory why > 44100Hz sampling rates *could* sound better is this: If a high resolution file is lowpassed before resampling you have to choose: Apply a steep lowpass, e.g. 20000Hz: 0dB -> 22050Hz -90dB; Result: No aliassing but (pre)ringing and/or phase shift. The less steep a lowpass is the less ringing and/or phase shift you'll get but OTH at some point you'll get noticable aliassing and/or audible high frequency loss. I don't know which lowpass is the best compromise, that's why I ask. If this lowpass doesn't result in complete silence above 22050Hz, not resampling would be cheating as aliassing would be avoided. Additionally resampling isn't a lossless process, so without resampling the test would be biased against high resolution IMO.
- About dither I *think* you're right, but if resampling is done anyway ...

I think changing the resolution in a way that it could be burned to CD audio 1:1 will lead to less complaints afterwards.
Title: Test 24/96 vs CD resolution
Post by: music_man_mpc on 04 September, 2003, 05:43:04 PM
Might it be easier to create two spertate files: one 24bit and one 16bit from the same analogue source?  Although I suppose this would create a whole new set of factors.
Title: Test 24/96 vs CD resolution
Post by: Pio2001 on 04 September, 2003, 07:01:53 PM
Will the 8 bits dithering remove all informaton ? Won't there be some info still audible below the dither noise ? I think it's safer to go through a real 16 bits step.
Title: Test 24/96 vs CD resolution
Post by: Continuum on 05 September, 2003, 02:19:05 AM
Quote
2. Apply fb2k's scan per-file tack gain to avoid clipping and too low volume.

3. Create "A" sample(s) by using fb2k's diskwriter, 24/96 resolution, trackgain used.

Is this step necessary? It could possibly introduce some degradation on both files created (A and B), thus maybe reduce the perceivable difference between them.

There is not much reason to listen to file A instead of the original, outside this test, anyway.
Title: Test 24/96 vs CD resolution
Post by: tigre on 05 September, 2003, 02:30:44 AM
Quote
Quote
2. Apply fb2k's scan per-file tack gain to avoid clipping and too low volume.

3. Create "A" sample(s) by using fb2k's diskwriter, 24/96 resolution, trackgain used.

Is this step necessary? It could possibly introduce some degradation on both files created (A and B), thus maybe reduce the perceivable difference between them.

There is not much reason to listen to file A instead of the original, outside this test, anyway.

You're right. This could be a weak point. But:
If the level of the source material (I don't provide it unless someone here knows a source for lossless high resolution samples) is too low (like peak level = -40dB), this could lead to cranced up volume on playback and the 16bit noise floor becoming noticable which wouldn't be fair IMO. If the source material contains already clipping, this could become worse by lowpassing and 2x resampling (similar to lossy encoding introducing clipping), which wouldn't be fair either. IIRC I've read here that DAC's (or was it ADC's) can't preserve full 24bit resolution anway because of molecular movement introducing noise (or similar).
Title: Test 24/96 vs CD resolution
Post by: tigre on 05 September, 2003, 02:42:02 AM
Quote
Will the 8 bits dithering remove all informaton ? Won't there be some info still audible below the dither noise ? I think it's safer to go through a real 16 bits step.

Sounds reasonable, but I have no idea. (Wouldn't 9bits dithering be necessary BTW?) I have experienced some surprises (e.g. noise bumping if dither depth is too low) when trying to understand how dither works a while back. This could be tested by comparing A: adding noise several times to a sample vs. B: adding noise + change res. 14->16bit; change res. back to 24bit several times but for me this isn't "A" priority ATM.


EDIT:
Any ideas what to use for lowpassing (what settings) and resampling? Is there anything that could give better results than fb2k? Anyone?
Title: Test 24/96 vs CD resolution
Post by: Patsoe on 05 September, 2003, 04:14:42 AM
Title: Test 24/96 vs CD resolution
Post by: tigre on 05 September, 2003, 04:32:52 AM
Quote
  • If you have trouble finding samples I can get them easily. Just call

Call!  This would be great. It's kind of hard to get 24/96 samples if you are no audio engineer as DVD-A isn't rippable at full resolution so far (or have I missed something?). Maybe you could put some ~ 20 seconds samples to a new thread in "Uploads" forum so others who are interested in performing similar tests could use these samples?
Title: Test 24/96 vs CD resolution
Post by: KikeG on 05 September, 2003, 05:16:44 AM
I'm planning to setup a similar test, but using real-world 16/44.1 and 24/96 DACs. The analog output from those DACs would be recorded with a 24/96 ADC. The first test would be about trying to tell the original from the 16/44.1 DAC - 24/96 ADC recorded one. If nobody can, then there's nothing more to check. If somebody could, then one would have to try other tests in order to know where the "bottleneck" is: at ADC, then at DAC, then you would have to try just with DSP (no hardware DACs between the process) and test if is the type of filtering used the bottleneck.

Patsoe:
Quote
If you have trouble finding samples I can get them easily. Just call.

All I lack for preparing the test is a proper 24/96 file (well, and spare time), because I didn't know how to have access to some, so help here would be very appreciated
Title: Test 24/96 vs CD resolution
Post by: tigre on 05 September, 2003, 05:42:21 AM
Hi KikeG! I've been waiting for your appearence here. What do you think about lowpassing/resampling quality? All I can find is that for 48kHz->44.1kHz SSRC/fb2k does a very good job, but I don't know about 96kHz->44.1kHz. 2Bdecided's post in this thread (http://www.hydrogenaudio.org/forums/index.php?showtopic=210&st=0&&#entry1794) sounds like there are really good resamplers available for this task, but I'm not completely sure if SSRC/fb2k's is one of those.
Title: Test 24/96 vs CD resolution
Post by: KikeG on 05 September, 2003, 06:01:28 AM
Tigre, about you proposed test, if you want to avoid superfluous processing as much as possible, then all you have to do is first remove frequencies over 22050 Hz, then convert to 16-bit with dither, and then again to 24 bit, this last process is lossless.

I recommend to perform the filtering with CEP FFT filter, using spline curves and a 256-point (in order to keep pre-echo as short as possible) Blackmann window that starts rolling at near 20 KHz and attenuates up to 100 dB at 22050 Hz.

(Other alternatives could be explored: a shorter pre-echo 128-point FFT filter that starts rolling at 19 KHz, or some of the scientific filters that have no pre-echo at all, but have phase distortion at cutoff frequency. Human ear is deaf to phase at high frequencies, so this can be a true alternative)

Save results as 32-bit fixed point, and then use FB2k to convert to 16-bit with strong noiseshaping. Then, convert again to 24-bit, either with CEP or FB2K, with no dither.

Edit:
SSRC uses a longer filter, so that pre-ringing when downsampling will be longer. I'd use fast mode, which I believe uses a shorter filter than slow mode. Still, it will be longer than what I propose with CEP. However, even so, I think this pre-ringing effect importance is quite small, and maybe inaudible in practice. Haven't checked, though.
Title: Test 24/96 vs CD resolution
Post by: 2Bdecided on 05 September, 2003, 06:19:59 AM
In that old thread, I was thinking of CEP being more than good enough for the task in hand.

You do need to resample to 44.1kHz, and then back again. The reason is because the quantisation noise is spread evening across the bandwidth of the signal. So a 16-bit 88.2kHz signal would have half the quantisation noise (at a given frequency) of a 16-bit 44.1kHz signal.


If I was doing the test, I'd do the following:

1. get a 2496 file which has no chance of clipping. In this case, it seems enough to record an LP at a sensible level.

A native 2496 digital recording might be better - I sent some to ff123 years ago. 2496 audiophile recordings on DVD-video discs (which came out before the DVD-audio standard was finalised) so allow digital output - so you can make a digital clone by recording the digital output into a 2496 capable sound card.

2. resample to 44.1kHz

3. dither to 16-bits. I'd be tempted just to use 1-bit RMS (2-bit peak to peak) triangular dither, no noise shaping.

If you want to use heavy noise shaping to maximise the SNR, do so, but it might be interesting to compare this with the above suggestion.

4. Convert the result(s) to 24-bit. That's just zero padding.

5. Resample to 96kHz.


or...

4. Convert the result(s) to 32-bit.

5. Resample to 96kHz

6. Dither to 24-bits, no noise shaping.


I'd use CEP because I'm familiar with it, but I'm sure foobar2k can do an excellent job. Someone would need to try some test files to check that everything is in order before carrying out the test.

Hope this helps. btw, count me in for listening!

Cheers,
David.

P.S. Another test could compare a 44.1 16-bit recording to a 2496 recording of same analogue source without any sample rate conversion. If no one can hear the difference here, then there's no point of any other test. If they can, then you move on to using a higher rate for A>D or D>A, as suggested by KikeG.
Title: Test 24/96 vs CD resolution
Post by: tigre on 05 September, 2003, 07:13:48 AM
KikeG & 2Bdecided, thanks for the responses.

Quote
Someone would need to try some test files to check that everything is in order before carrying out the test.

As it would be great to use free software for this (so everyone who wants can reapeat it) I'd like to test.

I have some ideas about what samples to use and what to look for:

- some sine sweep with frequencies > 22.050Hz, look for aliassing after lowpassing/resampling in frequency view + frequency analysis
- some single-sample clicks, look for (pre-) ringing in frequency analysis

any other useful samples?

BTW: Here's something that's post-ringing in my ears: In "Nyquist was wrong" thread (http://www.hydrogenaudio.org/forums/index.php?showtopic=8177&st=49&) Azeteq said
Quote
As it has been stated in this thread, it should be impossible to hear filter ringing with fc beyond human hearing. This is INCORRECT. It might be tempting to think so. The human hearing cannot hear steady-state sines over 20kHz. This has been concluded in millions of hearing tests all over the world. However, when listening to impulse responses, we have to take into account what is analyzing these sounds. The human ear has its own set of filters, analysis windows. Analyze a long enough impulse response with an auditory model and you will find that these filters are indeed triggered. This is why we can hear steepness of filters even when fs=96kHz and fc=47kHz.

As this discussion ended without real conclusion, I'm not completely sure if > 20kHz ringing is inaudible. Does anyone know facts about this (e.g. ABX-tests or similar performed)?
Title: Test 24/96 vs CD resolution
Post by: Mac on 05 September, 2003, 07:45:20 AM
Quote
(I don't provide it unless someone here knows a source for lossless high resolution samples)

I have the Cool Edit's Loopology cd, which contains 32bit samples, I don't think they're >44.1khz however.  Is that any use?
Title: Test 24/96 vs CD resolution
Post by: 2Bdecided on 05 September, 2003, 08:32:57 AM
I missed the revival of that thread (it was on my birthday!).

I'd like to test the theory he's proposing. But he suggests that "Don't try the tests with Soundblasters, Midiman Deltas or anything in that range... The conversion process will most probably mask all of the effects."

Isn't the audiophile 2496 based on the Midiman Delta series? Doesn't that mean that, supposedly, no one here has a good enough sound card to test out these claims? Or am I mistaken about the model mentioned?


If I understand him correctly, the idea is that the cochlear amplifier (ie. the active process within the cochlea, which isn't fully understood I hasten to add!) does respond to HF sound that we can't actually hear when presented as a steady state tone. This response isn't to let us hear HF sound, but to trigger a change in the cochlea tuning and dynamic compression so that audible sounds are perceived differently.


For this to happen, these HF components need to

a )  get through the middle ear
b )  cause a movement of the basilar membrane
c )  be picked up by some hair cells on the basilar membrane.

These are difficult questions. For (a)... well, the middle ear isn't believed to be the limitting factor - in many mamals it passes sounds way above the audible frequency for that mamal. In humans, it's not so obvious that this is the case, but it doesn't seem to contribute greatly to the HF hearing limit.

For (b)... the sharp tuning curves we often see for the BM are due to the active process within the cochlea. In a dead cochlea, the curves aren't do sharp. It's typical that the highest resonant frequency on the cochlea is around 14-18kHz, with the skirts of the filter going higher than this. Maybe, without an active process, or with the active process not sharpening the filters, the skirts of the filter will go even higher. To 20kHz easily, but to 40kHz?

For ©... I don't know. Really.

Anyway, it's interesting to think about and ponder and test - is Azeteg still around to provide input?

If so, on what equipment do you consider this effect audible? Can you explain your idea (and where it came from, if it's not your idea) in any more detail? i.e. what is happening within the ear?

Cheers,
David.

P.S. I'm not buying this idea yet, but I want to find out where/why it's right or wrong.
EDIT: removed extra smilie
Title: Test 24/96 vs CD resolution
Post by: KikeG on 05 September, 2003, 09:50:23 AM
Quote
You do need to resample to 44.1kHz, and then back again. The reason is because the quantisation noise is spread evening across the bandwidth of the signal.

You're right. I thought about that before replying, but then I thought about other things that would be needed and I finally forgot about that important point. So I would use same process I talked, but: after the filtering, resample to 44.1 KHz and then dither to 16 bit. Then, convert to 32 or 24 bit, and upsample to 96 KHz. The resampling can be performed with FB2K or CEP high quality with pre-post filter, since the previous filtering at 20 KHz will eliminate any ringing at this resampling process.

Edit:
about dither and such: I think flat (no noiseshaped) dither won't make an audible difference in practice, and maybe it's a wise idea to use it instead of strong noiseshaped filter. Maybe I'd use intermediate triangular shaped dither. Abut clipping: just make sure that neither of the files clip.

More edit:
David, what do you think about using an IIR filter (CEP scientific filters) with no pre-echo but more post-echo and non-linear phase distortion at high frequencies?

About ultra-hf becaming audible inside the ear: it could be due to basilar membrane not being perfectly linear, and still vibrate with ultra-hf sound. I recall about JJ (James Johnston) talking about this once. However, AFAIK it has been tested that hf inaudible tones don't intermodulate inside the ear and cause audible tones, but I could be wrong. So... I don't know that much.
Title: Test 24/96 vs CD resolution
Post by: Patsoe on 05 September, 2003, 10:06:55 AM
Quote
It's kind of hard to get 24/96 samples if you are no audio engineer as DVD-A isn't rippable at full resolution so far (or have I missed something?).

DVD-A I cannot rip either. But there are some DVD-V discs having 24/96 stereo sound which should be available from my library.
It will take me a couple of days (I fetched too many CDs - over my quotum for this week...), then I'll try to get some good ones.
Title: Test 24/96 vs CD resolution
Post by: KikeG on 05 September, 2003, 11:06:08 AM
Quote
It will take me a couple of days (I fetched too many CDs - over my quotum for this week...), then I'll try to get some good ones.

Great, thank you!
Title: Test 24/96 vs CD resolution
Post by: KikeG on 05 September, 2003, 11:16:06 AM
Quote
I have some ideas about what samples to use and what to look for:
...
any other useful samples?

I think it would be enough just to use a RMAA 5.1 24/96 wave test file that you can generate using asychronous testing, and analyze also with RMAA. Plus an impulse signal (1 high amplitude sample, surrounded by a few seconds of zero samples (silence), and maybe a sweep tone signal. The impulse signal is good for checking filter performance, mostly hf rejection and ringing in time domain. The sweep for checking for aliasing.
Title: Test 24/96 vs CD resolution
Post by: tigre on 05 September, 2003, 12:51:48 PM
I've found 2 short 24/96 samples:
http://64.41.69.21/product/reference/keys.wav (http://64.41.69.21/product/reference/keys.wav)
http://64.41.69.21/product/reference/triangle-2.wav (http://64.41.69.21/product/reference/triangle-2.wav)
Title: Test 24/96 vs CD resolution
Post by: Patsoe on 06 September, 2003, 05:40:08 PM
Quote
I've found 2 short 24/96 samples:
http://64.41.69.21/product/reference/keys.wav (http://64.41.69.21/product/reference/keys.wav)
http://64.41.69.21/product/reference/triangle-2.wav (http://64.41.69.21/product/reference/triangle-2.wav)

This disc should be of great help: http://www.chesky.com/catalog/body_catalog...0010&CATEGORY=2 (http://www.chesky.com/catalog/body_catalogdtl.cfm?PRODUCT=100010&CATEGORY=2) - unfortunately I can't find it at the library. Heck, I'd buy it if I could find it at all in Holland...
Title: Test 24/96 vs CD resolution
Post by: 2Bdecided on 07 September, 2003, 12:40:52 PM
Have you tried mail order from somewhere?

I've just ordered two of Chesky's DVD-As from amazon.co.uk.

You could ask at the Chesky forum about buying the discs outside the US, or visit a hi-fi shop, or just email chesky directly.

Remember - IIRC those DVD-V discs with 24/96 audio on them have no copy protection at all - I think you should encourage people who release things like this by buying their products, not copying them. What's more, Chesky is a great campaigner for having digital outputs put into DVD-A players - something the majors have said shouldn't happen.

Cheers,
David.
Title: Test 24/96 vs CD resolution
Post by: Patsoe on 07 September, 2003, 04:24:36 PM
Quote
Remember - IIRC those DVD-V discs with 24/96 audio on them have no copy protection at all - I think you should encourage people who release things like this by buying their products, not copying them. What's more, Chesky is a great campaigner for having digital outputs put into DVD-A players - something the majors have said shouldn't happen.

If you're referring to my copying them from the library - I'm paying a small amount to the copyright foundation for that for every disc. However, you're right in that it wouldn't be quite right to share those copies. But I believe using short samples (less than 30secs) is legal, and wouldn't do harm?
Title: Test 24/96 vs CD resolution
Post by: 2Bdecided on 08 September, 2003, 06:46:23 AM
It would seem the general consensus is that 30-second samples are OK for our purposes (and plenty of others).

I don't know the legal situation if your country re: copying library discs, and paying a fee.

I was suggesting, as a general principle, that it's good to support companies that produce excellent products, and don't restrict the way in which we can use them. It's very bad to take advantage of such companies, because that's a sure way to put them out of business!

all IMO.

Cheers,
David.
Title: Test 24/96 vs CD resolution
Post by: tigre on 09 September, 2003, 06:38:55 AM
Most likely I'll be able to get this DVD (http://www.chesky.com/catalog/body_catalogdtl.cfm?product=100168&category=2) from a public library (probably next week). Hopefully I'll be able to create some samples from it.
Title: Test 24/96 vs CD resolution
Post by: Patsoe on 09 September, 2003, 08:08:14 AM
I've spent several hours browsing (real life, that is ) through the library this afternoon - they have 1852 music DVDs, which aren't searchable by label...

I couldn't find a single Chesky SAD or Classic Records DAD... So, since I already promised to contribute, I'll be looking for a local shop now...

Btw, if anyone is interested, here is a list of Classic Records 24/96 DVD-V discs: http://www.audionautes.com/lp_cd/classicre...dad/dad_eng.htm (http://www.audionautes.com/lp_cd/classicrecs/dad/dad_eng.htm)
Title: Test 24/96 vs CD resolution
Post by: tigre on 10 September, 2003, 06:46:56 PM
Thanks for the help, Patsoe. I'll have a look if the library (browsable online  ) has got some DVDs from the list.
Title: Test 24/96 vs CD resolution
Post by: tigre on 10 September, 2003, 06:53:10 PM
OK. Finally here's a detailed description of how the test should be performed. Probably most of the explanation isn't necessary for people here, but I just want it to be as fool-proof as possible. B)

===============================


Test CD resolution (16bit/44.1kHz) vs. high resolution (e.g. 24/96)

Test samples
2 short samples are here: http://64.41.69.21/product/reference/keys.wav (http://64.41.69.21/product/reference/keys.wav) and http://64.41.69.21/product/reference/triangle-2.wav (http://64.41.69.21/product/reference/triangle-2.wav)
Anyone who can get 24/96 material please post short samples (~20 seconds, reasonable volume = peak at -0.1 - -10dB) here: Upload thread (http://www.hydrogenaudio.org/forums/index.php?showtopic=13054&)

Hardwardware needed: (*)
Soundcard capable of 24/96 playback.

Software needed: (*)
foobar2000 http://foobar2000.org (http://foobar2000.org)
ABCHR http://ff123.net/abchr/abchr.html (http://ff123.net/abchr/abchr.html) (*)
______________________
(*)
for those who have
DVD-A authoring software and a DVD-A player OR
DVD-V authoring software that can handle 24/96 audio and a DVD player capable of DVD-V-with-24/96-audio playback,
instead of ABCHR use KikeG's fileABX (http://(http://www.hydrogenaudio.org/forums/index.php?showtopic=12387&) to create randomized files and use your authoring software to create DVD-A/V

========================


How to test

0. Install the software needed etc.

1. Take some audio sample(s) recorded at 24/96 = "A" sample(s)

2. Convert "A" to "A_downsampled": 16/44.1 using fb2k's diskwriter, dither (strong noiseshaping), slow resampling

3. Create "B" samples by converting "A_downsampled" to 24/96 using fb2k's diskwriter, slow resampling, dither.

4. Try to hear differences and prove that you hear them by ABXing using ABCHR (**)
________________
(**)
If you want to use DVD-A/V + hardware player:

4.1. Use KikeG's fileABX to create randomized files from "A" and "B"
4.2. Author a DVD-A/V using the randomized files
4.3. Try to ABX = play back with your hardware player and tell which files are "A" and which "B"; compare to fileABX's log

*AFTERWARDS*

=======================


Details/suggestions for "How to test":

Given the directory you want to use for the test is d:\test

0. Install the software needed etc.
- create subdirectories: \test\A_samples; \test\temp; \test\B_samples
- put ABXHR (or fileABX) to d:\test
- install foobar2000 using installer

1. Take some audio sample(s) recorded at 24/96 = "A" sample(s)
- put the samples you want to use for testing to A_samples directory

2. Convert "A" to "A_downsampled" ...
- run foobar2000
- enqueue (e.g. drag'n'drop) all samples from \A_samples to foobar2000's playlist
- highlight all files
- Rightclick -> convert -> Settings ...:
-- Output directory: d:\test\temp
-- Output filename formatting: %_filename%
-- Output format: WAV (PCM 16bit dithered)
--
Title: Test 24/96 vs CD resolution
Post by: KikeG on 11 September, 2003, 03:15:59 AM
For generating the files, maybe it's a better idea using triangular dither instead of strong ATH dither, because the later can work worse if further processing (such as digital attenuation or eq.) is performed when listening to the files. Also, there's no need to use dither when converting from 16 to 24 bit.

Edit: BTW, you can use WinABX too for the test 
Title: Test 24/96 vs CD resolution
Post by: tigre on 11 September, 2003, 03:55:22 AM
Quote
For generating the files, maybe it's a better idea using triangular dither instead of strong ATH dither, because the later can work worse if further processing (such as digital attenuation or eq.) is performed when listening to the files. Also, there's no need to use dither when converting from 16 to 24 bit.

About 16->24 conversion: Dither or not should be equal, so no need to change this - or...?

About noise shaping vs. flat: I thought it's recommended to use flat dither for all steps of processing and noise shaped dither only in the last step (can't remember where I've read that, maybe CEP help or some threads about running winamp with several DSP plugins...). As the source is 24bit there hasn't been (hopefully) a noise shaped dithering step to 16 bit before, so this would be the one and only.
Additionally, if someone is able to hear the difference between 16bit and higher resolution (at low volume parts of a sample, great equipment, silent environment), it should be easier to hear with flat dither used than with noiseshaped as the perceived noise level is lower with noiseshaped (I just want to explain my thoughs - I know that you know these things  ).

So ... In your statement there's a "maybe" and a "can". This means to me either you're a polite person or it's hard to tell if noise shaping should be used or not. Given someone who uses 24/96 audio for storage/playback probably uses high quality equipment - will this decrease the danger of noise shaped dither giving worse results when processing is applied?
Giving the advice not to use equalizers/further processing in the test is no option as in RL (some) people use DSPs and the test should give RL results.

So - as I have no idea how dangerous noise shaped dither+dsp is compared to flat dither noise becoming audible at low volume, please give more information about this. - Or just tell me what to use. I'm sure you know what you're talking about.
Title: Test 24/96 vs CD resolution
Post by: KikeG on 12 September, 2003, 12:41:48 PM
Quote
About 16->24 conversion: Dither or not should be equal, so no need to change this - or...?

No, in practice there will be no difference, so for sake of simplicity you can leave it on.

Quote
So ... In your statement there's a "maybe" and a "can". This means to me either you're a polite person or it's hard to tell if noise shaping should be used or not.

Well, it means that I'm not really sure, and for some things I don't like to be very categorical.

Quote
Given someone who uses 24/96 audio for storage/playback probably uses high quality equipment - will this decrease the danger of noise shaped dither giving worse results when processing is applied?


Probably not. Dither break would happen just if you further process the samples with a resolution of 16 bit, or process them at 24 bit or more and then dither/truncate to 16 bit. I don't think this will be usual when listening to 24/96 samples. Also, I guess that in practice the kind of dither won't make much of a difference, because I believe flat dither noise won't be audible under regular listening conditions. Note that I was suggesting triangular shaped dither, which is a little bit less noisy than flat dither, but more resistant to processing than strong ATH shaping.

Personally, I use triangular shaped dither when listening to music, because I don't like the idea of having such higher noise levels at high frequencies. I prefer a more uniform noise distribution, even when it would be more audible under critical conditions. As I can't hear this noise anyway, I use it. But it's more of a "fuzzy feeling" thing than anything else, because in practice there's probably no difference. But being rational now, under unrealistic critical listening conditions, strong ATH would most probably be better. (sorry for not being categorical, I can't avoid it  )

So, use what you prefer, I guess strong ATH is fine for the test.
Title: Test 24/96 vs CD resolution
Post by: Pio2001 on 16 September, 2003, 03:57:33 PM
Quote
AFAIK it has been tested that hf inaudible tones don't intermodulate inside the ear and cause audible tones, but I could be wrong. So... I don't know that much.

I ran one of these tests (playing an audible frequency in one speaker and an inaudible one in the other speaker). It was very short, I just tried 6 kHz and 18 kHz and heard nothing else than 6 kHz. That's all. The playback was loud, enough to cause intermodulation in the speaker if the two tones are played into the same one. But I would probably have heard no intermodulation either with two audible tones. Ear intermodulation must occur at high levels (>100 dB). The test just showed that the common intermodulation comes from the hifi equipment and not the ear.
Nika Aldrich also ran the same kind of test (his forum : http://recpit.prosoundweb.com/viewforum.php?f=18 (http://recpit.prosoundweb.com/viewforum.php?f=18))
Title: Test 24/96 vs CD resolution
Post by: nwn on 16 September, 2003, 05:14:11 PM
i see you look for 24/96 sound file ..
here perhaps no expensive ..
http://www.aixrecords.com/catalog/sampler2.html (http://www.aixrecords.com/catalog/sampler2.html)
Mlp and Pcm files..

i don't see such file here .. http://www.kellyindustries.com/sounds.html (http://www.kellyindustries.com/sounds.html)


and this http://www.chesky.com/catalog/body_catalog...0010&CATEGORY=2 (http://www.chesky.com/catalog/body_catalogdtl.cfm?PRODUCT=100010&CATEGORY=2) is on a d*nk** p*p network ..  with some others ..

but you know as there are dvd-audio, envy 24ht and Audigy have restriction and degrade output signal desactivating digital output  .. licence ...
http://www.dtcp.com/data/Compliance_Rules_...udio_020610.pdf (http://www.dtcp.com/data/Compliance_Rules_Audio_020610.pdf)
Title: Test 24/96 vs CD resolution
Post by: askoff on 18 September, 2003, 02:08:22 PM
Quote
It was very short, I just tried 6 kHz and 18 kHz and heard nothing else than 6 kHz. That's all.

I can hear 18 kHz sine wave. With my SB Live and without resampling and tone samplerate is 44,1 kHz. 
(well maybay i don't hear that 18kHz wave but all that distortion...)
Title: Test 24/96 vs CD resolution
Post by: tigre on 23 September, 2003, 06:32:45 PM
OK. I've got a DVD-V containing 24/96 PCM audio now.
[Edit]
Blah ... I had some problems extracting audio using smart ripper but probably I've found a sollution.
[/Edit]

Another point that just came to my mind: IIRC when ripping DVD-V (e.g. to DivX + mp3 audio .avi files) sound is often low volume and needs to be amplified (or was it AC3 sound only ...?). If this is the case here too, some of the 24bits resolution could be lost. (I don't think there's a difference between 24 and e.g. 22bit but the design of this test shoud satisfy audiophiles as well ...  .) So - is there a way to get sound from DVD-V at decent volume without loosing resolution due to amplifying?
Title: Test 24/96 vs CD resolution
Post by: tigre on 23 September, 2003, 08:12:03 PM
OK. Finally I've managed to get 24/96 audio from the DVD-V "Chuck Mangione - The Feeling's Back" I've linked to some posts earlier. FYI: I've used LPCM24.exe from rarewares.

It contains music-related content > 24kHz and according to CEP's frequency analysis noise level is somewhere arround -108dB. Volume is reasonable -> from CEP statistics of 1 randomly chosen track:
Code: [Select]
Max Sample Value: 27755.65
Peak Amplitude: -1.44 dB
Minimum RMS Power: -70.73 dB
Maximum RMS Power: -8.69 dB
Average RMS Power: -22.1 dB


It's a jazz record, instruments are:
Flugelhorn
Piano
Acoustic guitar
Drums
Percussion
Acoustic bass, electric bass
Cello
Flute, ato flute
female vocals

Tomorrow I'll start to create some test samples of it. I'd appreciate any thoughts/wishes about what kind of samples to choose (as much high frequency content as possible / high dynamic range within the sample / many instruments playing at the same time / only one or few instruments / certain instruments playing / strongest available transients / ... ?).

Bedtime now. CU.

edit: typos
Title: Test 24/96 vs CD resolution
Post by: KikeG on 26 September, 2003, 03:02:11 AM
I think a good test sample could be one that would have as many variety as possible. For example, sample containing a soft part with a single instrument playing, and then a louder part with several instruments. But I don't mean editing, just getting a piece of the recording that encompasses all these.

Still downloading your 1st. sample...
Title: Test 24/96 vs CD resolution
Post by: Patsoe on 03 October, 2003, 05:34:21 PM
Quote
This disc should be of great help: http://www.chesky.com/catalog/body_catalog...0010&CATEGORY=2 (http://www.chesky.com/catalog/body_catalogdtl.cfm?PRODUCT=100010&CATEGORY=2) - unfortunately I can't find it at the library. Heck, I'd buy it if I could find it at all in Holland...

I finally got my hands on it: CHDVD171 "The Super Audio Collection and Professional Test Disc" from Chesky. Would you still need some of that, Tigre?
Title: Test 24/96 vs CD resolution
Post by: tigre on 04 October, 2003, 03:38:48 AM
Hi Patsoe.

It would be good to have more samples, because all samples that I've uploaded are similar somehow. I hope there's something completely different on your CD ... Maybe you find something with a single (a few) instrument(s) playing, something acapella-like and/or something with huge dynamic range ...?

Besides, since it's called "The Super Audio Collection and Professional Test Disc" maybe there's some information in the booklet about mastering, e.g. what equipment used etc. This could help to persuade some people that the quality of these recordings is good enough for this test.

Thanks. CU tigre