http://www4.head-fi.org/forums/showthread....20&pagenumber=6 (http://www4.head-fi.org/forums/showthread.php?s=&threadid=44927&perpage=20&pagenumber=6)
This theaudiohobby guys is really bugging me.
I'm not trying to say that SACD must be superior to DVD-A (since actually they're probably both 'perfect'), but this guy's justifications for SACD are ridiculous, starting from the page before this:
Originally posted by theaudiohobby
Joe,
most of your post is simply not so, but rather than get into what is and what is not read this arcticle from optical-disc systems journal, remember that is a partially a critique of DSD and you will see that the experts give more due to DSD that you want to accept. DVDA v Bitstream article (http://www.opticaldisc-systems.com/NovDec2000/BITSTREAM/BITSTREM82.htm) also look at this article from this dcs paper (http://www.dcsltd.co.uk/papers/effects.pdf) and this paper (http://www.dcsltd.co.uk/papers/aes97ny.pdf) and see that there are valid scientific reasons why folk might prefer DSD to DVDA happy reading.
As you can see his links are irrelevant to or even against his argument. I'm especially irked by him using the dcs paper to say that they find DSD to be superior. Yes, they pointed out flaws for 24/192 (does anyone here know more about this test? Was the ADC/DAC combination faulty or something??) but they hadn't even made any observations on the corresponding performance criteria for DSD yet! For all we know it could be even worse!
And bit resolution? Time resolution? I'd say 24/192 trumps DSD in all these areas. Sampling at 2.8MHz doesn't mean anything when you have practically zero S/N ratio at 100kHz and above.
What about this (http://www.discwelder.com/pdfs/1-Bit%20SD%20is%20Unsuitable%20paper.pdf) paper?
From what I understand of the format, it is presenting the signal in a similar manner to something that might happen inside a very cheap CD player. It (partly) makes up for it with sheer bandwidth, so it could be worse... but it just seems like an such an awful waste considering that PCM can do so much more with less.
As you can see his links are irrelevant to or even against his argument. I'm especially irked by him using the dcs paper to say that they find DSD to be superior. Yes, they pointed out flaws for 24/192 (does anyone here know more about this test? Was the ADC/DAC combination faulty or something??) but they hadn't even made any observations on the corresponding performance criteria for DSD yet! For all we know it could be even worse!
And bit resolution? Time resolution? I'd say 24/192 trumps DSD in all these areas. Sampling at 2.8MHz doesn't mean anything when you have practically zero S/N ratio at 100kHz and above.
I just love this dcs paper...
96kS/s, 24-bit: "some stereo image formation"
Oh damn, all my audio cd's are mono then !
192kS/s, 24-bit: "bass can appear light and slightly out of time".. "stereo image can be strong but widened (1.5 times)"
I think these sentences are so ridiculous that they're not even worth arguing
In my opinion switching from PCM (96+ kHz) to DSD is a bit like going back from weighted number systems to Roman numbers
A bit like saying MMMMMMCCXXXXIV instead of 6244... oh well, as long as it can simplify a $10 DAC design
I don't even buy what some of these dCS papers say. Again, they rely on non-rigorous (non-blind) anecdotal evidence, and on "possible" audible effects, again not confirmed by blind testing.
They look more as hi-res marketing than as anything else.
192kS/s, 24-bit: "bass can appear light and slightly out of time".. "stereo image can be strong but widened (1.5 times)"
I think these sentences are so ridiculous that they're not even worth arguing
Before you laugh, remember one thing: DCS sell these things, and in this paper, they're basically saying that none of them sounds perfect.
I ask you to seriously consider the possibility that they are simply reporting what they hear, in the interests of scientific research.
They have no reason to lie, to invent this, or even to imagine it. I believe they sometimes (but not always?) use blind testing. Whichever this was, the placebo effect doesn't usually make the "better" thing sound worse!
btw - Joe Bloggs - you're wasting your time. The SACD vs DVD-A vs CD vs analogue thing will be scientifically sorted out, eventually. But it hasn't been yet, and for now, SACD is winning the marketing war with audiophiles. I think this is largely because SACDs are carefully mastered, and the ultra-sonic noise has an interesting effect on people's equipment. There are probably genuine audible differences between the formats, but the differences people hear in the home are probably mostly due to other things.
Cheers,
David.
remember one thing: DCS sell these things, and in this paper, they're basically saying that none of them sounds perfect.
Which is just not believable, in my opinion.
Also, from that reasoning, it's unavoidable to derive that the higher the format resolution, the better the sound. So they are clearly pushing high-res formats as better sounding, and those formats are what they are more interested in selling now.
remember one thing: DCS sell these things, and in this paper, they're basically saying that none of them sounds perfect.
Which is just not believable, in my opinion.
I thought you'd say that! I thought you'd pick me up on "There are probably genuine audible differences between the formats" too!
So they are clearly pushing high-res formats as better sounding, and those formats are what they are more interested in selling now.
Read the paper - better than CD, but not transparent. It's hardly a sale pitch! Which, to my logic, makes it all the more believable.
Cheers,
David.
There are probably genuine audible differences between the formats, but the differences people hear in the home are probably mostly due to other things.
Amen, David... such as that little button (or buttons, on some models) that let you choose between the internally preset EQs! My wife and I spent a good chunk of the weekend comparing models, and she had all sorts of fun going from "Concert Hall" to "Dolby ProLogic II," and everything in between. And there I was, standing beside her, trying to explain to her that the "2-Channel Stereo" mix sounded more like the original CD - yes, that's right, the demo disc which so fascinated her was a standard 16bit/44.1KHz CD - than any of the others. Silly me. >_< (No sympathy... I love her more than life itself, and have plenty of opportunities to listen to my own music in the original mix, so it's not as harsh a sacrifice as it sounds. No pun intended. )
- M.
Read the paper - better than CD, but not transparent. It's hardly a sale pitch!
??? It's clearly a sales pitch. According to them, 24/192 or SACD are the best, and CD the worst. They use a typical "audiophile-style" argument: there's always something better, no matter how good what you have now. So you can always buy something better. According to them, it's possible that 24/384 is not transparent either, which is just absurd.
Back to real world, if 24/192 is not transparent, then *nothing* is transparent. There's no analog audio playback format that is remotely close to 24/192. Not to say that there are no real-world mics or speakers that can cope with it adequately either.
Edit: moved addendum to next post.
Read the paper - better than CD, but not transparent. It's hardly a sale pitch!
??? It's clearly a sales pitch. According to them, 24/192 or SACD are the best, and CD the worst.
Not the 1997 aes97ny.pdf, but the 1998 effects.pdf.
Back to real world, if 24/192 is not transparent, then *nothing* is transparent.
Nothing is more likely. Things
can be perceptually transparent for many people and many signals. But in engineering, nothing is perfect. Things can, at best, be "good enough".
I know you know that, and that you meant "perceptually transparent". I was just being picky. It's just worth remembering: even at 24/192, it would be trivial to show that the signal coming out of the A>D>A process was different from the one going in. But probably not using your ears!
There's no analog audio playback format that is remotely close to 24/192. Not to say that there are no real-world mics or speakers that can cope with it adequately either.
But there's no analogue equipment with brick-wall filters, or that kind of time-domain ringing either. I'm not convinced that it's audible, but simple intuition says that it's not a good thing to have it flying around an audio system either.
We're having this discussion
again, aren't we?
And no, I couldn't hear anything in the ringing test!
Cheers,
David.
But there's no analogue equipment with brick-wall filters, or that kind of time-domain ringing either.
It depends. Some analog parts of most record/playback chains will have glentle (or not so gentle) rolloffs below the ringing frequency of a 96 KHz or 192 KHz sampling system. What importance does some subtle, occasional, ringing at 40 KHz (on a high-res system) have, if your speakers or mics roll off quite below 30 KHz?
And of course, all discussion is related to audibility. If not, we wouldn't talking about audio.
Edit:
More about dCS papers: if dCS people were truly interested in scientific research, they would at least have presented some DBTs that supported their claims. Those DBT results shouldn't be difficult to obtain, if the sonic differences were so clear, as they suggest. Also, those results would make a world of difference in the credibility of their claims. But still, they have not presented them. The same old story we've heard so many times.
AFAIK, so far nobody has been able to prove such claims by means of a DBT, not even to prove that well-implemented, plain-old cd-audio is sonically distinguishable form high-res formats, when playing regular music.
David: I may be wasting my time, but the way this guy fudges technical terms to support his POV is ridiculous and he must be eradicated
More from that guy
Joe and czilla9000,
read the papers before we discuss further at the rate you guys are going my response will probably be good enough for an AES paper.
suffice to say that the dcs paper does say this
quote:
--------------------------------------------------------------------------------
detailed comparisons not yet performed on enough systems, but well
liked by (classical) artists after sessions
no busy signal break up
very good separation of reverberation and room acoustics
no observations on bass so far
strong stereo image formation, no observations on width so far
--------------------------------------------------------------------------------
The philips response also says this
quote:
--------------------------------------------------------------------------------
10. The impulse responses of 4 different systems in a multichannel configuration are depicted: a 48 kHz system, with a bandwidth of 20 kHz (that is, 8 kHz transition bandwidth is allowed for anti-aliasing filtering), a 96 kHz system with 35 kHz bandwidth (26 kHz transition bandwidth), a 192 kHz system with 75 kHz bandwidth (42 kHz transition bandwidth) and an SACD system with 95 kHz bandwidth. Though none of the systems reproduces the input exactly, the DSD systems shows the least artifacts. Clearly, the 48 kHz system has great difficulty in reproducing the click; due to the steep filtering it starts ringing at a -30 dB level approximately 1 ms before the click, which is very audible. Also at the higher sampling frequencies, the ringing phenomenon cannot be removed, though it is reduced significantly. Only the DSD system is very effective in suppressing the ringing effect, due to very slow filtering above 95 kHz. The price to pay for this is the increase in noise floor with respect to the other systems; however, as the noise floor contains only high frequency components which are uncorrelated with the audio, they are not perceptible.
--------------------------------------------------------------------------------
and that correlates exactly with the main decenting paper I gave you that says
quote:
--------------------------------------------------------------------------------
An alternative bitstream converter has been reported that uses a combination of a linear quantizer with dither to guarantee linearity, together with 4th-order noise shaping. Conversion to a 1-bit code (typically from a 4-bit code) is then performed by an open-loop, optimal code conversion table which minimises spectral modulation. Potentially the system produces high resolution with no low-level correlated distortion. However, the bit rate is again very much greater than PCM, and although solving the problems of correlated and idle-channel distortion it is too bit inefficient.
--------------------------------------------------------------------------------
and philips answer to their objection is this
quote:
--------------------------------------------------------------------------------
Other issues which often appear to be confusing, are data rates in connection to the bandwidth claimed by SACD. The SACD format comprises (apart from its red-book conforming CD layer) two different music streams: a stereo 2 channel stream, and a surround 6 channel stream. Hence, an SACD contains 8 channels of high-quality audio. Because all channels are 2.8 MHz sample rate, 1-bit signals, the total data rate equals 2.8 Mbyte/s (or 22.6 Mbit/s). On these signals, lossless coding is applied. This lossless coding scheme is specifically developed for coding 1-bit signals. From experience of over 100 recordings, the average coding gain is roughly 2.4 - 2.5 for pop recordings, and 2.6 - 2.7 for classical recordings. This corresponds to a data rate per channel of about 1.1-1.2 Mbit/s. This indicates that on average 70 minutes of a DSD signal can be recorded on an SACD in the 8-channel format. For 6 channels, this amounts to roughly 95 minutes. Also, the high sampling rate of DSD allows for the use of filters with slow roll-off. We can compare this to DVD-A. The DVD-A format that gets closest to the SACD characteristics is DVD-A at 192
kHz, 20 bit, which reaches the same dynamic range, but is either of lower bandwidth than SACD if sloppy anti-aliasing filters are used, or has the same bandwidth using steep filters. Using a compression factor of 2 the data rate amounts to 1.9 Mbit/s, which is almost twice as much as the data rate for DSD. Hence, even if only six channels are used on the optical disk (compared to 6+2 on SACD), only 55 minutes of music can be stored - much less then the 74 minutes that we are accustomed to from CD.
--------------------------------------------------------------------------------
I will return with some detailed answer to objections when the need arises. but as it is stands, it is easy to see here that SACD has a few aces up its sleeve that DVDA as it stands simply cannot match.
The responses from philips are interesting, although I have no idea where he pulled them from (!) So SACD has a higher compression rate than DVD-A now? Better impulse response?
Tell him some uncontrovertible facts:
- 24/192 DVD-A has much better dynamic range than SACD. DVD-A has 24-bit performance (144 dB SNR) up to 96 KHz, whilst SACD has the equivalent of just 20-bit performance (120 dB SNR) just up to 20 KHz, and quite worse performance over this frequency. Over 50 KHz the signal is very noisy, in fact Sony recommended to filter SACD output over 50 KHz.
- It's impossible to totally avoid quantization distortion in SACD. Due to its 1-bit nature, it's not possible to adequately dither it.
It's true that SACD may have better impulse response due to its higher sampling rate. However, instead of having higher ringing just in the proximities of high-frequency impulsive signals (as DVD-A has), it will have relatively high levels of a mixture of distortion and constant noise at these same frequencies where DVD-A ringing occasionally appears.
Is there an exact dynamic range profile for DSD?
Can't totally avoid quantization distortion? Can't you just make the dither range fullscale?
There's a sense in which they're trying to have their cake and eat it.
They state that ringing around 96kHz will be audible, but noise above 95kHz will not be. Now look, either audio signals above the generally accepted hearing limit are audible in some way, or they're not. If they're not, let's stick with CD. If they are, then we'd better have a much better understanding of what's involved before we start making such bold claims!
"Clearly, the 48 kHz system has great difficulty in reproducing the click"
Yes, and the human ear will have great difficulty in hearing most of it, too!
Be very careful with these comparisons. It's easy to do one or more of the following:
1. average the data many times. The "ringing" on DVD-A adds up, the noise on SACD cancels.
2. Use a gentle filter on the SACD signal which makes the noise just below the width of one pixel on a linear time-domain plot. This noise is about 20dB down.
But there's no point waging a holy war over this. For all we understand at the moment, either SACD or DVD-A should sound fine. The reasons quoted why SACD isn't good enough ("imperfectible") are very true, but the problems can be minimised to the point where they don't really matter. Also, this time domain smearing in DVD-A (and, more so, CD) has not been shown to be a cause of audible problems.
No matter how many times they try to write "time domain ringing/smearing" in the same sentence as the word "audible", this is just conjecture.
At the end of the day, the technical arguments will be irrelevant. They're not what will decide the format war. I too find it very annoying that there is so much false and misleading science used to sell SACD, but until more work is done, you can rebut it by saying why it's wrong (or at least, unproven/unjustified), but can't provide anything better.
Cheers,
David.
Is there an exact dynamic range profile for DSD?
Can't totally avoid quantization distortion? Can't you just make the dither range fullscale?
No, it depends on the DSD modulator. You can make it equal to almost anything you like. It depends on how aggressively you want to push the noise into the ultra-sonic region.
20-bit (i.e. 120dB) to 20kHz is possible (but not always used, I might add). But there's no useable hard fixed digital full scale in DSD either - it just goes unstable above a certain input, so you stay at least 6dB below the theoretical 0dB FS.
Cheers,
David.
No, it depends on the DSD modulator. You can make it equal to almost anything you like. It depends on how aggressively you want to push the noise into the ultra-sonic region.
How does it alter that? I'd like to learn more about noise shaping. You've also left behind a very confused crowd in my last thread about DSD. http://www.hydrogenaudio.org/forums/index....howtopic=13450& (http://www.hydrogenaudio.org/forums/index.php?showtopic=13450&). Help us.
Found this from here: http://www.hometheatermag.com/hirezaudio/ (http://www.hometheatermag.com/hirezaudio/)
All digital signals begin as 1-bit representations of the analog waveform after they're converted by 64x-oversampling Delta-Sigma modulation. Whereas the PCM system then sends the signal through a series of filters, which can cause audible problems during recording and playback, DSD records the 1-bit signal directly and eliminates several steps of the record/playback process.
True? False? If that's indeed the first recording medium I know it can't be only 64x oversampling And what is the process to transform delta-sigma into PCM? Can it be lossless?
(I guess probably not, otherwise there wouldn't be all that fuzz about mastering in DSD. Just master in PCM and then release as SACD.)
Noise shaping is like negative feedback. It can be modelled as negative feedback. You look at the output, look at what you don't want, and subtract this from the input - hopefully correcting the output "next time around" as it were.
The reason I didn't come back and explain it is because it doesn't make that much intuitive sense to me, but the result falls out of the equations nicely. I can't type the equations, and I can't scan the diagrams, so I thought you'd get on better with a web search. Unfortunately, you've found a marketing based site, rather than an engineering based site.
I've tried google, and it hasn't helped me either. I think you've been here:
http://www.cs.tut.fi/~rosti/1-bit/ (http://www.cs.tut.fi/~rosti/1-bit/)
and I don't think it's too helpful.
Assume you have a signal, and you quantise it. Quantise = round it to some value. 8=bit audio has 255 possible values. 1-bit audio has only 2: 1 or 0.
The rounding introduces an error - noise or distortion. When the signal is big, and the error is small, then the error sounds like noise, and fills the audio band equally. When the signal is small and the error is (relatively) large, then it sounds like distortion, at harmonics of the signal frequency.
If you put negative feedback (i.e. noise shaping) around the quantisation process, then it will attempt to correct it's own faults. Of course, it can't correct them - you'll always have exactly the same amount of noise or distortion, because you
have to round. However, if you put a filter in the feedback loop, then what gets through the filter is removed from the error. So, you let the audible error through, subtract it from the input, and you now have less error in the audible band, but twice as much (say) above the audible band.
In DSD, it's almost all error. The signal is much smaller than the error. But you have a tiny signal band (0-20kHz), and a huge ultrasonic band (20kHz-1.4MHz), so you can push the (huge) error down very severely in the audible band - there's plenty of room to push it into the ultrasonic region.
How does it alter that?
The filter in the feedback loop is altered. If it lets through ten times as much signal in the audio band as in the ultrasonic band, then there will be 1/10th as much nosie in the audio band as in the ultrasonic band (remember - what gets through the filter is subtracted, but the equivalent amount of energy is moved into other frequency regions). The limit is where the ultrasonic region is saturated, and/or the whole thing goes unstable. Because it's a loop, if you get it wrong, it can make a sound even when there's no input. Like when you put a microphone in front of a speaker which is playing the output of that microphone.
(but remember - In DSD, it's classic negative feedback, but similar - I'm struggling to explain this, and I'm hoping you might have looked at op-amps, negative feedback, and filters at some time in physics at school or something to help with this - if not, I'm probably not helping at all!)
All digital signals begin as 1-bit representations of the analog waveform after they're converted by 64x-oversampling Delta-Sigma modulation. Whereas the PCM system then sends the signal through a series of filters, which can cause audible problems during recording and playback, DSD records the 1-bit signal directly and eliminates several steps of the record/playback process.
Marketing BS. SACD fans want you to think that all A>D and D>A convertors are basically 1-bit convertors. Well, some are, some aren't. The best aren't. (The best
were when SACD was invented, which is why we have it, but this is no longer true!).
Interestingly, before SACD, 256x oversampling wasn't uncommon. They use 64x oversampling on SACD, because 256 gives too high a data rate. But people used 256x,
even for D>A conversion of CD quality material because it was measurably better than 64x. Draw your own conclusions. Hint: SACD is compromised, though probably still good enough. 256x 3 or 4 bits would be better, if they're guesses and theory is correct. If they're not, then it's overkill anyway.
And what is the process to transform delta-sigma into PCM?
1-bit > 16-bits. 20kHz low pass filter. Keep every 64th sample, throw away the rest. (there are more computationally efficient methods!!!).
Can it be lossless?
Of course not. You lose everything above half the destination sample rate. Whether this is audible or not...
(I guess probably not, otherwise there wouldn't be all that fuzz about mastering in DSD. Just master in PCM and then release as SACD.)
There are a lot of SACDs sourced from 24/96 masters. And some SACD fans claim they sound better than the 24/96 versions. Draw your own conclusions. Hint: SACD can't add anything useful, but there's always distortion, placebo, poor DVD-A players etc etc etc.
I hope this is some help.
I still think you're wasting your time trying to convince him!
Cheers,
David.
Re: noise shaping, I have this (http://members.chello.nl/~m.heijligers/DAChtml/Digital%20Theory/Digital%20theory.html) page. There's a section on noise shaping. I've found the rest of the article quite useful, so I might find this useful too. Unfortunately I don't know what the terms in the equation stand for. E.g. x(n): function varying with frequency? Or varying with sample number?
Edit: definitely sample number. Give me a few days, I may be able to sort this out yet... Especially with you help in giving the equations some intuitive background to place it on
Edit: u ( n ) = x ( n ) + ( y ( n ) - u ( n ) ) * h ( m ), huh? How can you have u(n) on both sides of the equation? (And I've done my share of programming-- i=i+1 but I still can't figure out exactly what this stands for )
Edit: Is it that the whole noise shaping process is being treated as an offline process here, where u(n), y(n) etc. represent the entire audio clip; so this step takes the quantized output y(n), puts in on the RHS, calculates the difference between the quantized waveform y(n) and the unquantized waveform u(n), filters the difference waveform by convolving with h(m), and then adding it back on x(n) to obtain the *new* u(n), and try quantizing again with the new waveform? And 2nd order noise shaping would be taking the quantized version of that new y(n), calculating y(n)-u(n) again, filtering, etc.?
That sounds like a plausible enough way to shape the noise. Now to figure out how to do this in realtime. ->
Can't totally avoid quantization distortion? Can't you just make the dither range fullscale?
No. In fact, the amount of dither applicable before unstability is quite sub-optimal. It's explained at the link at the 2nd. post of the thread.
quote:
--------------------------------------------------------------------------------
10. The impulse responses of 4 different systems in a multichannel configuration are depicted: a 48 kHz system, with a bandwidth of 20 kHz (that is, 8 kHz transition bandwidth is allowed for anti-aliasing filtering), a 96 kHz system with 35 kHz bandwidth (26 kHz transition bandwidth), a 192 kHz system with 75 kHz bandwidth (42 kHz transition bandwidth) and an SACD system with 95 kHz bandwidth. Though none of the systems reproduces the input exactly, the DSD systems shows the least artifacts.
They mean this ? http://forum.cdfreaks.com/showthread.php?s...1864#post376461 (http://forum.cdfreaks.com/showthread.php?s=&threadid=61864#post376461)
2BDecided: so how is my understanding of noise shaping going?
To everybody else: what is linearity error and what has it got to do with jitter and choosing between SACD and DVD-A, if anything? There's this dCS Elgar SACD DAC thingo (http://www.stereophile.com/showarchives.cgi?814:6) and it has this great measurement in the linearity error category and theaudiohobby is using this to push the superiority of SACD.
To everybody else: what is linearity error and what has it got to do with jitter
It has nothing or very little to do with jitter.
and choosing between SACD and DVD-A, if anything?
Nothing either. there's no reason why a good DVD-A DAC can't achieve good linearity too.
There's this dCS Elgar SACD DAC thingo (http://www.stereophile.com/showarchives.cgi?814:6) and it has this great measurement in the linearity error category and theaudiohobby is using this to push the superiority of SACD.
It doesn't say at anywhere that a good 24/192 DAC can't achieve same or better performance.
2BDecided: so how is my understanding of noise shaping going?
To everybody else: what is linearity error and what has it got to do with jitter and choosing between SACD and DVD-A, if anything? There's this dCS Elgar SACD DAC thingo (http://www.stereophile.com/showarchives.cgi?814:6) and it has this great measurement in the linearity error category and theaudiohobby is using this to push the superiority of SACD.
The dCS Elgar plus is a multibit DAC!
5-bits, 64x oversampling, to be precise.
http://www.dcsltd.co.uk/Elgar.htm (http://www.dcsltd.co.uk/Elgar.htm)
Anyone who uses the (excellent) performance of the DCS Elgar as some kind of justification for the superiority of SACD doesn't know what they are talking about!
DVD-A, correctly dithered, in infinitely linear. The dCS Elgar manages linearity down to the equivalent of the 27-th bit level with a DVD-A source.
SACD, as a format, is not linear, as proven in the Lipshitz and Vanderkooy papers at the start of this (or the other?) thread. I haven't seen results which conclusively show the level of the non linearities of a real SACD disc, played through the elgar Plus. I suspect they'd be worse than DVD-A. The Stereophile result looks like it is (linear down to the equivalent of the ~22nd bit), but it's swamped by noise, so this isn't a fair comparison.
Did I mention that you're wasting your time? :-) When you've dismissed all "technical" arguments, he'll just say "it sounds better to me", and you can't argue with that!
btw - noise shaping is an active (feedback!) process. You can't do it in the way you suggest - that's not feedback, that's subtraction. In a feedback loop, the output affects the next input, which affects the next output, which affects the next input, which... ! You don't get that with a single subtraction because the result of the subtraction on this sample can't affect the next sample.
Cheers,
David.
btw - noise shaping is an active (feedback!) process. You can't do it in the way you suggest - that's not feedback, that's subtraction. In a feedback loop, the output affects the next input, which affects the next output, which affects the next input, which...
Then I'm back to this question
Edit: u ( n ) = x ( n ) + ( y ( n ) - u ( n ) ) * h ( m ), huh? How can you have u(n) on both sides of the equation? (And I've done my share of programming-- i=i+1 but I still can't figure out exactly what this stands for)
Especially as u(n), x(n) etc. really seem to stand for a whole waveform instead of just one sample?
And my interpretation does have 'feedback'--the output waveform affects the next input waveform, which affects the next output waveform, which affects the next input, which...
How can you convolve a single sample with a filter?
I know my interpretation won't work as a real-time process as it is, but is it workable for an offline process? :x
Did I mention that you're wasting your time? :-) When you've dismissed all "technical" arguments, he'll just say "it sounds better to me", and you can't argue with that!
True , but I feel like there is more at stake than what his final beliefs are; this thread is also going to sway the beliefs of the rest of the population in the forum. >_<
this thread is also going to sway the beliefs of the rest of the population in the forum. >_<
You poor misguided fool.
!
As for noise shaping and SDMs, I'll use these pictures for a simple explanation:
http://metrology.hut.fi/courses/s108-180/L.../Luento9/sd.pdf (http://metrology.hut.fi/courses/s108-180/Luento9/sd.pdf)
Look at the first and second order SDM ADCs. Ignore the last box in each - that converts the 1-bit output to multi bit.
The first order is the simplest to understand: Figure 8.20, the SDM is inside the box. Note that the system is clocked, so it can only send things round the loop once per clock cycle.
Let's assume the output can only be -1, or +1. It's really 0 or 1, but changing the range makes the maths easier. Also, let's assume that the comparator has a very slight positive bias - otherwise, in this simple system, with absolutely no noise, we're going to get stuck.
Now, consider silence. That's an input equivalent to zero. The input to the integrator is zero, the 1-bit comparator has to jump one way or the other, it's slightly biassed, so jumps to +1. Output = +1.
Next clock cycle. (B) = +1, Vin=0, the output of the integrator = -1, the output of the comparator =-1. Output = -1.
Next clock cycle. (B) = -1, Vin=0, the output of the integrator = 0, the output of the comparator = +1. Output = +1.
We're back where we started.
So, a first order undithered DSM encodes pure silence as a square wave at half the sampling frequency. That's true. That's why we use higher orders, and a little dither to break the pattern up. But even in this simple example using a simple system, you've got to admit that the noise shaping works: there's none in the audio band - it's all at fs/2, which is 1.4MHz on SACD.
Now, consider your idea. Calculate the output throughout the duration of the signal, and
then do the subtraction. The output would be +1 throughout, and then the subctraction would make it -1 throughout. So silence gives a negative digital full scale signal? Hmm - that doesn't work at all!
Any good?
If not, my fees as a private tutor are £50 per hour. ;)
Cheers,
David.
this thread is also going to sway the beliefs of the rest of the population in the forum. >_<
You poor misguided fool.
!
LOL I guess you're right... anyway, I asked a mod to close that thread and it actually got closed with a somewhat positive note for me, so I won't be wasting any more time on this either way... er... congratulate me! CONGRATULATE ME! (w00t)
Hmm, I understand some of what you are saying, but then we are back to the basic definition of delta-sigma as 'sum the difference...'. I have people at my university I could ask about this, I suppose that will be faster Thanks
LOL.
You are to be... congratulated.
But you better be ready if he starts a new thread.
Please don't engage in personal attacks, but be VERY subtle if you want to. By calling him names people will have a lowered impression of you no matter your technical superiority in the issue that was at hand.
I don't know if you should be congratulated - if you used that kind of insulting language (and unattributed quotes!) on HA you'd be warned, then banned. But you probably already know that.
Next time, if I have time, I'll join the discussion if it helps.
Cheers,
David.
Well, there are differing standards on different forums--head-fi.com was a very polite place when I first posted there but apparently it has gotten a lot more rough lately. bbs.stardestroyer.net, the forum I've been frequenting more these days, is a madhouse, where you're just not treating yourself fairly if you don't throw some insults back at all the people that have piled curses and mockery on you, and people have this elaborate theory that insults are just used to emphasize points and do not detract from the quality of your argument if the argument itself is sound. (!) I guess some of that has rubbed off on my posting style. :x
But the people at bbs.stardestroyer.net were driven into their current style of debating by a certain very annoying debater who uses very polite language, whose arguments are all wrong but all take an extraordinary amount of time to pick apart, and jumps from point to point, never addressing the main issue and never admitting defeat. Quite similar to the person I was debating
But enough excuses I'll be sure to conduct myself properly here.
Quite similar to the person I was debating
No, he thought it was right, and was doing his best to express it. I think you were unfair on him.
Anyway, you can kill his argument with some facts:
Prerequisites:
The psychoacoustic theory in this field is almost non existent. (So you can't say, for certain, "the signal is different in x, y, or z dimension, hence it
will sound better". You can say the signal is measurably better/different, but you can't draw concrete conclusions about what this will sound like).
As a consumer, you will
never be in a position to make a fair comparison.
DSD fundamentals:
1-bit coding was proposed because the best DACs
at that time were 1-bit DACs. That is no longer true.
1-bit coding is neither optimal, not perfectible. There are inherent non-linearities, which it is impossible to remove. (They can be reduced to a level at which they are probably inconsequential).
1-bit coding adds large amounts of ultrasonic noise. The effect of large amounts of ultrasonic noise upon audio systems is well known, and the effect is audible. Whether it is a good or bad effect is debatable (though it may explain the preference for the "sound" of SACD), but incorporating it within the audio format itself is a dubious decision.
The use of 8-bit "DSD Wide" doesn't remove any of these problems. It just makes it possible to edit the stuff without cascading and multiplying the problems. The fact that DSD wide was created proves that the problems are real, and that they become significant when 1-bit DSD coding is used several times in series. That should ring alarm bells.
The delivery format (SACD) is still 1-bit. There is no magic that says you can avoid all the above 1-bit problems by originating in 8-bit and then converting. (You
can reduce the problems, but you don't have these same
fundamental problems with DVD-A
at all).
DSD Advantage? :
Material mastered on Analogue tape has a limited high frequency response (bandwidth). You can say that the tape deck is experiencing frequencies which are simply so high (so fast) that the equipment cannot respond. You could also call this effect a "filter".
This high frequency response (bandwidth limit) of most analogue master tapes exceeds the limit of CD. It does not exceed that of DVD-A. A "brick wall" (or more gentle) filter, as used at the high frequency limit of PCM systems, will only ring if it experiences spectral content within its transition band. An analogue master tape does not contain any frequency content this high, or any transition this rapid, so the filter in a PCM system is inconsequential.
For such content, the anti-alias filter has no effect in the time or frequency domain (apart from removing any very high frequency equipment noise). There is no signal present, temporal or spectral, that it can possibly impact. Hence, the shorter impulse response of SACD (which is only really shorter under certain conditions and constraints) is immaterial when transferring material from analogue masters.
Consider that last point. Some SACD fans believe SACD sounds better than DVD-A. They claim that this due to the fundamental difference(s) between the formats. The only valid technological "advantage" is the shorter impulse response (less ringing) of SACD vs DVD-A. However, this difference (whether we can hear it, or not) is irrelevant for all those archive recordings released on SACD. (It's also irrelevant for the large number of 24/96 recordings upsampled for SACD ).
So, if there is an "audible" difference, it's due to mastering, ultrasonic noise, ADC quality, DAC quality, placebo, and/or (just possibly) non-linearity.
In other words, any "audible" difference, if it really exists, has nothing to do with the formats supposed superiority, but due to intentionally introduced (and format irrelevant) differences.
i.e. You can do a good or bad mastering job with either format. You can add ultrasonic noise in any DVD-A player if you want. You can use the same ADCs and DACs for both processes (remember: the best converters are no longer 1-bit devices). We can all imagine things. And if you want more non-linearity, well, use a distortion box!
This still leaves the possibility that a new all-digital “pure” DSD recording might sound better than a DVD-A recording, if the shorter impulse response is an audible advantage. Maybe. Who knows. What is certain is that, for the vast majority of SACD content out there, any superiority over DVD-A content (real, or imagined!) has nothing to do with the format or digital coding parameters.
To be fair to SACD, there are two other issues. Assuming that the ADCs and DACs are good enough assumes that there is a good filter in use when converting DVD-A. For high sampling frequency PCM, the requirements can be similar to SACD, but they're usually more strict. A bad filter may cause a bad sound. This is more of an issue with DVD-A than with SACD. Though it should be possible to do an excellent job quite easily when sampling at 192kHz!!!
Also, if you do need a 192kHz sampling rate, that's too high for multi-channel audio on DVD-A.
Neither of these impacts my "analogue master tape" example, which must be answered before we can take anyone seriously who claims that SACD sounds superior to DVD-A with such material.
A final thought:
Both formats should be more than good enough. Surround sound is a much bigger issue.
If the basic sampling on DVD-A and/or SACD isn't good enough (and I can imagine that! After all, we’ll have to be sold something new in another 30 years) then you need a high sample rate multi-bit system which is half way between both formats. Before going down that route (and, tbh, before deciding between SACD and DVD-A) we really need to understand the psychoacoustics and engineering that's at work here. The format war means this is in the interests of neither side, so marketing has taken over from engineering.
Cheers,
David.
EDIT: added "To be fair..." paragraph.
Hi Folks,
Courtesy of Joe Bloggs, I have come to learn of your august forum and some of the posts are rather very enligthning as well as the fact that I noticed that 2Bdecided knows Prof. Hawksford personally, whose paper in the Optical Disc Systems Journal, I refer to as the dissenting paper on head-fi.org. Some of 2BDecided and KikeG answers are very helpful and I would hope that you folks could probably help me with some questions that I have on DCS Elgar/Elgar Plus/ Mark Levinson No.30.6 linearity plots. After Joe's unwarranted outburst I went back to check out the original linearity plots of the an earlier Elgar implementation
(http://www.stereophile.com/archivesart/elgfig3.jpg)
The updated linearity plot after the implementation of DSD is very different
(http://www.stereophile.com/archivesart/DCSverFIG3.jpg)
And now the Mark Levinson 30.6
(http://www.stereophile.com/archivesart/306fig3.jpg)
The key differences to a casual observer (of which I am) is that the Mark Levinson is dancing along y-axis beyond -100 dBFS, and this becomes more aggressive as it approaches -120dBFS The original Elgar seems to have a cut-off in this region and Elgar Plus has the cleanest plot. What would account for this differences? Previous comments on the Elgar Plus are noted, thanks in advance for your helpful comments.
I thought I should ask another question with respect to some info that I read recently Auditory Display for Deep Brain Activation: Hypersonic Effect (http://www.icad.org/websiteV2.0/Conferences/ICAD2002/proceedings/Oohashi.pdf)
Could you guys give the paper a professional look over? thanks in advance.
regards, theaudiohobby.
While I'm not an acoustician or an electrical engineer, there are a number of problems with the paper linked that are not insignificant to this discussion.
For one thing, the article is not testing whether one particular method of audio data storage is superior to another. All tests use a heavily modified signal capture, master and reproduction system based on the SACD.
What the paper appears to claim is that certain (which among those used, they dont appear to say) recordings made with a certain recording system, mastered with a certain technique, and played back with a proprietary speaker system may change (how much is unclear) in certain individuals' (they dont say who, how many tested, what background) levels of blood flow and electrical activity in the brain in a statistically significant fashion when a high cut (>22KHz) filter is not on. This apparently correlated to subjective measures of sound quality ("more pleasant") in the same individuals (who were queried in some unmentionable method of "psychological evaluation", the results of which they dont bother to mention).
This has rather little relevance to whether SACD is a "better sounding" recording media than DVD-A (24/96 or 24/192) since DVD-A media was not used in any part of the testing here, and hence no comparisons between the same recording played back on DVD-A and SACD systems were made
In addition, the signal path in every stage of the experiment was manipulated in a non-standard fashion using extremely esoteric and/or proprietary equipment.
In particular, the primary DAC used was a higher frequency unit (3.072MHz) than is used in SACD mastering, largely because a lower noise floor than found at hypersonic frequencies (in traditional SACDs) was desirable. The system also featured esoteric "pre-emphasis" and "de-emphasis controllers" to furthur minimize noise at hypersonic frequencies (or something).
Most troubling about the system they used was that they did not appear to compare the low-pass signal plus ultra and hypersonic noise versus the full range signal at any point.
They also did not bother to introduce any other high-cut filters outside of the 22kHz one, which is also quite troubling, since AFAIK high-frequency perception does vary to a not-inconsiderable degree. Also, the relationship of the supertweeter to the rest of the speaker setup is questionable - the supertweeter (mimimum output frequency: 3KHz) is driven directly by the dedicated hypersonic amplifier and turned off completely in the case when the high-cut is employed.
Also, one of the two data sets they do bother to graph appears to show subjects adjusting the volume of the full range signal to be higher than the high-cut signal. The amount of the adjustment done relative to the high-cut appears to be <1dB (max).
There are a large number of questions about the methodology, and most importantly about the raw data, that are simply not addressed at all in this article.
I would wager the previous article in the Journal of Neurophysiology would have, at least, more data presented.
Strangely enough, the article appears to be intent on selling equipment (and the hypersonic test disc) manufactured by Action Research.
edit: fixed italics
The Journal of Neurophysiology paper (http://www.yamashirogumi.gr.jp/kumigashira/jnp-hse.pdf)
thanks audible!, are there other credible papers on this topic because most of those that I have come across are from Japanese universities.
thanks in advance
EDIT: added a link
I'm not supporting or refuting anyone's position concerning advantages or disadvantages of SACD vs. DVD-A, but in regards to this article...
The Journal of Neurophysiology paper (http://www.yamashirogumi.gr.jp/kumigashira/jnp-hse.pdf)
...and concerning psychoacoustic audio compression in a broader capacity, does this tend to refute the concept that lowpass does not change the effect of sound on the listener? Or has the contention always been that lowpass makes no
significant change in the effect of sound on a listener?
Could these results (and similar results in other studies, if they exist) stand as evidence that the use of lowpass in psychoacoustic audio compression is, to some measurable extent, detrimental to sound quality?
Excerpt from article (final paragraph)...
"In conclusion, our findings that showed an increase in alpha-EEG potentials, activation of deep-seated brain structures, a correlation of alpha-EEG and rCBF in the thalamus, and a subjective preference toward FRS, give strong evidence supporting the existence of a previously unrecognized response to high-frequency sound beyond the audible range that might be distinct from more usual auditory phenomena. Additional support for this hypothesis could come from future noninvasive measurements of biochemical markers in the brain such as monoamines or opioid peptides."Note: I am not making a statement or "taking a position" here, only asking questions in the pursuit of better knowledge of the subject.
Welcome theaudiohobby!
I can't explain the linearity graphs until I check how the Audio Precision determines these values. I've seen those kind of ripples before somewhere - IIRC it was due to a non-linearity, or possibly incorrect dither, or even no dither at all. That's just an "off the top of my head" guess - I'll try and find out more about what the AP linearity test involves.
IIRC The Hypersonic effect had a time constant of (typically) around 2 minutes, but I'll have to go back and re-read the paper (not this week - far too busy!). I thought we'd discussed it here before, but a search doesn't find it. Still, it's probably in one of those SACD/DVD-A/CD threads in the FAQ. IIRC, despite claims in the paper, there were doubts that high level ultra sonic signals could be reproduced without generating in-band (i.e. audible) distortion components.
ScorLibran, be careful drawing comparisons between the Hypersonic effect (if it exists - IIRC it's the only research group to have shown this, and the experiment hasn't been repeated or confirmed by others) and (selective or static) low pass filtering in psychoacoustic based codecs. 96kHz sampling adds another octave compared to CD quality, 192kHz sampling adds yet another octave. Low pass filtering 44.k1Hz sampled material down to 19kHz removes approximately three semitones.
So, even if this missing inaudible information has an effect, it's desperately tiny compared with the amount added by high resolution formats. What’s more, very few other people are claiming that these extra high frequencies have some effect on human listeners directly. What they do suggest is that the filtering in CD quality audio has some small effect. mp3 and MusePack etc use filtering by the bucket load - if it is a source of audible problems in CD, then it should be 10x more of a problem in psychoacuostic based codecs.
If proven, the Hypersonic effect raises other questions. If the main benefit of including ultrasonic frequencies is to introduce a subconscious change over several minutes, cannot the same effect be caused by simulating this ultra sonic content.
If, as a consumer, you expect to be able to answer any of these questions using information provided by any interested party, or your own testing, then you're sadly mistaken. That is, until someone issues identical material on DVD-A and SACD.
Cheers,
David.
ScorLibran, be careful drawing comparisons between the Hypersonic effect (if it exists - IIRC it's the only research group to have shown this, and the experiment hasn't been repeated or confirmed by others) and (selective or static) low pass filtering in psychoacoustic based codecs. 96kHz sampling adds another octave compared to CD quality, 192kHz sampling adds yet another octave. Low pass filtering 44.k1Hz sampled material down to 19kHz removes approximately three semitones.
So, even if this missing inaudible information has an effect, it's desperately tiny compared with the amount added by high resolution formats. What’s more, very few other people are claiming that these extra high frequencies have some effect on human listeners directly. What they do suggest is that the filtering in CD quality audio has some small effect. mp3 and MusePack etc use filtering by the bucket load - if it is a source of audible problems in CD, then it should be 10x more of a problem in psychoacuostic based codecs.
If proven, the Hypersonic effect raises other questions. If the main benefit of including ultrasonic frequencies is to introduce a subconscious change over several minutes, cannot the same effect be caused by simulating this ultra sonic content.
If, as a consumer, you expect to be able to answer any of these questions using information provided by any interested party, or your own testing, then you're sadly mistaken. That is, until someone issues identical material on DVD-A and SACD.
Cheers,
David.
Thanks David. That puts it into perspective. Here (http://www.hydrogenaudio.org/forums/index.php?showtopic=7516&view=findpost&p=74075) is the section of the FAQ (the "High Definition Digital Audio" portion) I had been studying on the subject, which I had read before as well. I know the effects of lowpass were thought through and discussed many times by the codec developers, but I wasn't fully clear on the
extent of such effects.
I didn't notice the statistical analysis (table 2) in the html version (it is just linked, not displayed).
In fact, this is interesting : 26 listeners, an ABBA double blind test, 10 statistical analysis.
In 4 of them the full range playback was recognized from the lowpassed one (22 kHz, 80db/octave) with a probability that they were guessing <1 %, in one of them the probability was <5 %, and in the 5 other, the probability was >5%.
If 1 % is taken as threshold, the probability of failure in case of guessing is 99 %. Now could someone calculate the probability of getting at least 4 successes (p<1%) in 10 trials ?
I get p<0.0002 % can someone comfirm this ?
prob=0.00000200127615694084 or 1 / 500 000.
(n:=10;
p:=1/100;
sum(binomial(n,k)*p^k*(1-p)^(n-k),k=4..n);)
Thanks. So we can say that in this experiment, the 22 kHz lowpass was perfectly ABXed.
If, as a consumer, you expect to be able to answer any of these questions using information provided by any interested party, or your own testing, then you're sadly mistaken. That is, until someone issues identical material on DVD-A and SACD.
Cheers,
David.
Agreed and that is the reason that the constant dross that DVDA is technically superior all the more annoying. This thread has highlighted strengths and weaknesses of both formats and more to the point it is the commercial success or not of either format that will determine their eventual survival
Well if there was any 'constant drossing' about the 'superiority' of DVD-A I certainly didn't have anything to do with it
Re: ABXing the 22kHz cutoff:
the supertweeter (mimimum output frequency: 3KHz) is driven directly by the dedicated hypersonic amplifier and turned off completely in the case when the high-cut is employed.
Thus it seems that the 'lowpassed' and 'non-lowpassed' versions are not at all identical below 22kHz, in fact they are different starting at 3kHz?
No, because the people couldn't distinguish the super tweeter playing alone at full volume from complete silence.
at what frequencies
My understanding is that this tweeter has a frequency response starting at 3 kHz, and that it was fed with highpassed signals at 22 and 26 kHz, according to the experiments.
My understanding is that this tweeter has a frequency response starting at 3 kHz, and that it was fed with highpassed signals at 22 and 26 kHz, according to the experiments.
That's what the diagram in the first article appears to show - the supertweeters were being driven directly by the highpassed amplifiers with no contact with the "audible range" amps.
The comment about 3KHz was regarding the starting frequency of the driver - Pioneer claims flat response from 3-100KHz for that particular beryllium ribbon driver.
22KHz is likely going to be audible to some few people, but those people were apparently not involved in the test.
As I said before, my primary concern with the method involved here was that there was an assumption made that <22KHz + hypersonic (musical components) is going to be superior to <22KHz + hypersonic (noise).
This assumption did not appear to be tested in either article.
Agreed and that is the reason that the constant dross that DVDA is technically superior all the more annoying.
Given that the frequency response of DVD-A can reach 96KHz and the SACD (when heavily tweaked to reduce noise) can go slightly over that, I don't see how either article shows the superiority of one medium over another in any way (especially since they were not compared at all).
Certainly I don't see any 'dross' (not sure if that works grammatically (http://dictionary.reference.com/search?q=dross) here) about the superiority of one medium over another here, but I myself would be tempted to suggest the DVD-A is probably slightly superior for reasons of equipment price if not SNR.
High Quality 24/96 recording is readily availible to just about anyone who has a few hundred dollars in their pocket, and resolutions that high IMHO, are significant overkill to begin with. The overhead protects you from drunken mastering engineers, maybe.
The problem is, both SACD and DVD-A mediums are so high resolution to begin with that the mastering process is very likely going to be the limitation on the sound quality from any particular recording.
The problem is, both SACD and DVD-A mediums are so high resolution to begin with that the mastering process is very likely going to be the limitation on the sound quality from any particular recording.
So far nothing new under the sun
Phew, gee, I was wondering what happened to this thread and the whole General forum
Certainly I don't see any 'dross' (not sure if that works grammatically (http://dictionary.reference.com/search?q=dross) here) about the superiority of one medium over another here, but I myself would be tempted to suggest the DVD-A is probably slightly superior for reasons of equipment price if not SNR.
I have to disagree, technical superiority is more than just a better SNR. I do not understand the equipment price angle.
NB: dross - worthless commentary
Equipment price angle? Perhaps it's the bit where I can buy a 24/96 recording sound card for <$1000 (probably way less) but a DSD recording and mixing setup will bankrupt me and my family?
Equipment price angle? Perhaps it's the bit where I can buy a 24/96 recording sound card for <$1000 (probably way less) but a DSD recording and mixing setup will bankrupt me and my family?
okay, but when did the price of a technology determine it's technological superiority, for an extreme example, think of fuel cell engines and internal combustion engines, you cannot get a fuel cell for less <£1/2 m at present does that make it inferior to the good ol' internal combustion engine.
I have to disagree, technical superiority is more than just a better SNR.
That's all fine and good, but where do you see solid technical superiority in the SACD medium that outweighs the superior SNR of the DVD-A medium?
Anything anyone can actually hear?
Hypersonic frequency response of a standard SACD playback chain is quite noisy, and desiring frequency response above 96KHz is more than excessive, particularly since precious few people on any given continent possess transduction equipment capable of reproducing frequencies that high, and likely even fewer albums of music have been recorded on equipment that has viable response characteristics up that high.
Perhaps it's the bit where I can buy a 24/96 recording sound card for <$1000 (probably way less) but a DSD recording and mixing setup will bankrupt me and my family?
Exactly. You can record to 24/96 with an Audigy 2 or an M-Audio Revolution and pay less than $100.
You'd want to spend several times that or more to really get a decent piece of gear to master a DVD-A with, of course.
The LynxTWO (http://www.ramelectronics.net/html/lynxtwo.htm) features sampling rates of up to
200KHz, and costs less than $1000.
So, if you can't hear the difference between the two mediums, and the entry-level cost of creating content for one medium is orders of magnitude less than the other, which one do you choose?
If you're Sony, you choose the one that you hold the most patents for
If you're Sony, you choose the one that you hold the most patents for
it is always fun to see folks defending DVDA on the basis of money. DVD-A is not free some fees for DVD technologies (http://www.dvd6cla.com/news_20021119.html) these guys want to make money from licensing fees just as much as Sony/Philips. whatever you cut it, somebody will get your money.
it is always fun to see folks defending DVDA on the basis of money. DVD-A is not free some fees for DVD technologies these guys want to make money from licensing fees just as much as Sony/Philips. whatever you cut it, somebody will get your money.
While I see that you have a strong emotional investment in the SACD medium, this response is not particularly relevant to what was being discussed.
What
was being discussed was the
entry level costs associated with recording or potentially mastering to either medium.
This cost or recording via DSD is exponentially higher that for DVD-A, meaning the ability to
create content in native SACD is limited to only extremely wealthy individuals, or that SACD content is recorded at DVD-A resolutions and
then mastered to DSD at a mastering lab.
Which is pretty funny.
I'm sure someone will find an
excellent reason why DVD-A resolution content captured to DSD will sound better after being captured than the source itself did.
DVD-A is simple 24/192 or 24/96, which can be recorded at readily with fairly inexpensive equipment, almost none of which pays license fees to the DVD-A consortium, because it only involves recording at DVD-A
resolutions.
The issuing label or maybe even the mastering lab may have to pay the DVD-A consortium money, but that is a different situation altogether, akin to the fact that the labels also have to pay to press the medium.
Surely you wouldn't try to claim because SACD and DVD-A use the same medium and the cost of pressing is the same that this somehow relates to the fact that one cannot record directly to DSD without big money.
edit: fixed italics, clarification
I still think that both SACD and DVD-A are both pointless when you can have multichannel 24/96 Linear PCM on regular DVD Video, and achieve a much higher compatibility (ie. works with all existing DVD video players).
DVD Video supports:
Samplerates: 48khz, 96khz.
Bitdepths: 16, 20, 24.
Channels: 1-8
Bitrate must not exceed 6144mbps.
Audio in DVD (video) (http://www.mpeg.org/MPEG/DVD/Book_B/Audio.html)
At 48 Khz Nyquist gives you 24khz dynamic range. Few ppl can hear up to 20khz, so you have plenty of headroom for imperfect filters. I think using 48khz instead of 96 is a good space saving option with no loss in perceivable quality.
Problem: resampling is needed from legacy 44.1khz cda.
Advantage: None of the new "copy control" annoyances of both DVD-A and SACD.
I propose people produce their discs at 24/48, 1 ch for mono, 2ch for stereo, and any more channels for anything else (reducing bitdepth when exceeding bitrate as appropiate). Say, Dolby Surround Pro-Logic could be decoded into 3ch, and Pro-Logic II into 4ch.
DVD Video audio has everything we need, and there is no need to waste our time with either DVD-A or SACD with frequencies no one records or hear. Sure we miss the handy lossless compression of DVD-A, but with 4.38gb disc space, how much raw audio can we store at 24/48?
I think some people could even have fun recording multitrack 8ch and burn into DVD-Video (audio only) discs. 8ch at 16/48 or 5ch at 24/48.
While I see that you have a strong emotional investment in the SACD medium, this response is not particularly relevant to what was being discussed.
What was being discussed was the entry level costs associated with recording or potentially mastering to either medium.
This cost or recording via DSD is exponentially higher that for DVD-A, meaning the ability to create content in native SACD is limited to only extremely wealthy individuals, or that SACD content is recorded at DVD-A resolutions and then mastered to DSD at a mastering lab.
Which is pretty funny.
I'm sure someone will find an excellent reason why DVD-A resolution content captured to DSD will sound better after being captured than the source itself did.
DVD-A is simple 24/192 or 24/96, which can be recorded at readily with fairly inexpensive equipment, almost none of which pays license fees to the DVD-A consortium, because it only involves recording at DVD-A resolutions.
The issuing label or maybe even the mastering lab may have to pay the DVD-A consortium money, but that is a different situation altogether, akin to the fact that the labels also have to pay to press the medium.
Surely you wouldn't try to claim because SACD and DVD-A use the same medium and the cost of pressing is the same that this somehow relates to the fact that one cannot record directly to DSD without big money.
edit: fixed italics, clarification
And you don't have any emotional investment in DVDA , the introduction of most new technology into the market place comes at a cost, the fact that many studios have chosen to invest in DSD/SACD despite the higher entry costs suggests that the DSD/SACD model is more aligned to their business model than the DVDA model. If the business case of the DVDA is an open and shut one as you try to portray, it will already be a runaway success, however it is not. The reason for the lower cost of recording at higher resolution is because PCM technology is a more mature technology, that is being around longer, not because it is superior. I am sure that point should be clear since a couple of years ago the cost of the selfsame equipment was much higher.
I think you need to read up on the number of patents held by Toshiba, Warner, JVC, Pioneer, Meridian etc wrt to the DVD related technologies to appreciate why they are desperate for DVDA to succeed. They are certainly not promoting DVDA because they love music and the same applies to Sony/Philips wrt to DSD/SACD
And you don't have any emotional investment in DVDA
Not at all, I think DVD-A is huge overkill.
I'm not waving my hands trying to rationalize why its superior characteristics are more compelling than SACD's superior characteristics - but that's because I just dont see any superior (technical or audible) characteristics for SACD. Quite frankly, even with my NHT SB-3's I dont think I could tell the media apart (let alone from a CD) provided they were mastered to the same level with the same amount of care.
Perhaps you'd care, once again, to find some differences in the format you so love to proselytize for?
the fact that many studios have chosen to invest in DSD/SACD despite the higher entry costs suggests that the DSD/SACD model is more aligned to their business model than the DVDA model.
Many eh? Perhaps you'd care to find market penetration numbers rather than making vague statements.
What I (and YOU) do know for damn sure, is that anyone who can record to 24/96 or better can make a DVD-A resolution recording with little difficulty. And a very very very
large number of studios have this capability.
The number of "studios" (home, semi-pro, commercial and otherwise) with 24/96 or 24/192 recording capability is
huge, since this is the standard.
SACD is a proprietary system that appears to have been sold (and well sold, especially in your case) as being "superior" in some way to traditional high sampling rate 24 bit word recording.
Of course the fact that you are still here with no more proof of the superiority of the medium rather than "a bunch of studios adopted it", is
very telling indeed.
If the business case of the DVDA is an open and shut one as you try to portray, it will already be a runaway success, however it is not. The reason for the lower cost of recording at higher resolution is because PCM technology is a more mature technology, that is being around longer, not because it is superior. I am sure that point should be clear since a couple of years ago the cost of the selfsame equipment was much higher.
The DVD-A and SACD mediums have not "taken off" because of the diminishing returns found in investing in them.
Red book CD audio is
very high fidelity if mastered properly, and the vast majority of consumers do not have speakers of the quality necessary to determine the difference between it and other higher resoultion media. This is why higher resolution formats will not take off unless they are hybridized with red book audio CDs.
Diminishing returns.
Of course, your point here it totally moot. The number of studios which possess DSD equipment is tiny in comparison to those that possess 24/96 recording gear, primarily because of the entry level cost associated with DSD, and the lack of audible quality improvements inherent in the medium.
Which of course, was what I was talking about, so please do not attempt to change the subject.
Again, you are mistaking the success of the format at the consumer level (which in both cases is essentially negligible) with the success of recording equipment able to record at the bit rate and resolutions supported by the format.
Once again, any fool with several hundred dollars can pick up a Delta 44 or comparable product and record direct to 24/96 or 24/192 and get a really brilliant fidelity product. And many many many do.
Not so with DSD. Not So at
all. The cost associated with recording
direct to DSD is
huge in comparison (which again, was what was being discussed, not the royalites associated with pressing one or another), which is why all but the wealthiest studios simply do not bother.
Any recording made with 24/96 or 24/192 will not see
any improvements whatsoever by being mastered to DSD. Why?
Because mastering a 24/192 recording to DSD,
once again will not result in a higher quality recording than the source.
Any improvements in quality inherent in the SACD medium can only be found (if they exist, which I see no evidence whatsoever of, and which you have not bothered to present) in direct to DSD recordings or super high resolution analog to DSD recordings.
Direct to DSD recordings are few and far between, largely as a result of 24/192 gear being exponentially cheaper and sounding for all the world, just as good or better.
edited for clarity
I have to disagree, technical superiority is more than just a better SNR. I do not understand the equipment price angle.
Technical SuperiorityOK - let's just re-wind, and remember that all the "technical superiority" in the world, measured in apples on this side, and oranges on the other, is meaningless unless it matches human hearing.
(I'm in the minority in this forum in believing that there are audible differences between CD-quality audio, DVD-A quality audio, and SACD quality audio.)
You can say with certainty that one format is better in dimension X, and the other is better in dimension Y. But you can't say that the advantage in dimension X leads to a better sound - well, you can say it, but it's pure speculation. It won't be fact until the relevant experiments have been carried out, and the psychoacoustics are understood. At the present time, the psychoacoustic knowledge to make these claims
does not exist.
All these facilities world-wide for recording mastering and authoring SACD - how many have had no financial contribution or incentive from Sony? Some, sure, but how many exactly?
DVD-V for audioDVD-V isn't enough for audio. The navigation isn't properly defined (625/525-line TV displays are not appropriate for selecting tracks for in-car use!), and it does not have lossless packing. 96kHz sampling of material which has little high frequency content comes "for free" (in terms of data rate) with lossless packing. The inability to store 24/96 6-channel on DVD-V is seen as a problem by some record producers. They don't want to compromise their content anymore - they want the listener at home to hear what they hear - they want to deliver the master tape right into your Hi-Fi.
There are also issues with two channel / multi channel mixes - DVD-A allows you to define down-mix coefficients, or to do a separate 2-chanel 24/96 mix - which the lossless encoder then encodes at the same time as the 6-channel mix. The redundancy means that the 94/26 2-channel mix takes up less space than LPCM 44.1/16 material!
Basically, DVD-V is seen as too restrictive for audio. It doesn't allow the data rate or capacity that is needed.
The reason that most people have DVD-V only players is because the DVD forum didn't even think to define an audio standard for DVD. If they had done, and had consulted with the high-end audio community earlier, then everyone would have universal players, and DVD-A would be everywhere by default. As it is, all DVD-A discs can be authored for DVD-V compatibility if required, so DVD-V owners needn't miss out.
There are doubtless marketing, profit, and DRM issues which attract the majors to DVD-A. But the reason it exists is because the most concerned and enthusiastic members of the industry went to the DVD forum and said "we can do much better than DVD-V for audio - let us define DVD-A".
Cheers,
David.
I fell into the topic in CD Freaks. I posted nothing new about SACD vs DVD-A, but my post features a lot of links, for those who might be interested :
http://forum.cdfreaks.com/showthread.php?s...7207#post473974 (http://forum.cdfreaks.com/showthread.php?s=&threadid=77207#post473974)
If someone had come up with a well-standardized form of high-rez multichannel PCM with no DRM, that can be played by just pushing the play button, it'd trounce both DVD-A and SACD
If someone had come up with a well-standardized form of high-rez multichannel PCM with no DRM, that can be played by just pushing the play button, it'd trounce both DVD-A and SACD
There's nothing to stop you burning 24/96 FLAC or APE or even WAV files to a DVD-R, and programming your PC to auto play them when the disc is inserted.
However, I don't see that idea trouncing DVD-A or SACD.
Cheers,
Davd.
The number of "studios" (home, semi-pro, commercial and otherwise) with 24/96 or 24/192 recording capability is huge, since this is the standard.
Red book CD audio is very high fidelity if mastered properly, and the vast majority of consumers do not have speakers of the quality necessary to determine the difference between it and other higher resoultion media. This is why higher resolution formats will not take off unless they are hybridized with red book audio CDs...
... Not so with DSD. Not So at all. The cost associated with recording direct to DSD is huge in comparison (which again, was what was being discussed, not the royalites associated with pressing one or another), which is why all but the wealthiest studios simply do not bother.
Audible, we are in danger of getting sidetracked here, the issue at hand is technical superiority not market acceptance/business model etc. DSD commands a premium in the marketplace because it is new technology vis a vis the more established PCM technology and I still fail to see how that translates into technical superiority. Lest we loose sight of the issue at hand, you are the one who suggested that you felt DVDA was more superior because of it's superior SNR. I disputed that, because whilst it is superior in that respect, it is inferior to SACD in some other respects. I think the various aspects of that position are already well covered here by other participants of the thread.
I just dont see any superior (technical or audible) characteristics for SACD. Quite frankly, even with my NHT SB-3's I dont think I could tell the media apart (let alone from a CD) provided they were mastered to the same level with the same amount of care.
On your point of audibility, you are entitled to your position, however, it is simply not valid for the recording/mastering studios, if it were that clearcut why don't the studios just revert to 16/44.1 and forget about the higher resolutions or DSD altogether afterall the benefits are inaudible , the fact that they are doing otherwise wholesale suggests that the reverse is the case.
For the record, I have never suggested that there are benefits to mastering 24/192 to DSD.
On your point of audibility, you are entitled to your position, however, it is simply not valid for the recording/mastering studios, if it were that clearcut why don't the studios just revert to 16/44.1 and forget about the higher resolutions or DSD altogether afterall the benefits are inaudible , the fact that they are doing otherwise wholesale suggests that the reverse is the case.
Or not. I think many studios have changed to hi-res formats simply because it's what the market asks for, whether it sounds better to the people at studios or not.
Or not. I think many studios have changed to hi-res formats simply because it's what the market asks for, whether it sounds better to the people at studios or not.
I disagree here, some studios were recording & mastering in higher bit rates and samplingfrequencies as far back as the early 1980s, I do not have the exact dates to hand
if it were that clearcut why don't the studios just revert to 16/44.1 and forget about the higher resolutions or DSD altogether afterall the benefits are inaudible
The higher resolution / bitrate formats are very useful in a studio, because they allow for smaller rounding errors while applying filters, but the benefits for end users on average audio equipment are far fewer.
audiohobby, if high sample rates sound better for so many people, why there's no available scientific literature that proves its audible superiority, of course by means of blind tests? Why everytime people have been subjected to these kind of tests they have resulted in negative results?
I'm discarding Oohashi hypersonic effect paper, among other things because he used very special speakers and program material, and his results have not been yet duplicated by anyone. I'm talking about tests using commercial equipment and regular music.
Note that some very reputed mastering engineers such as Bob Katz say that 44.1 KHz well implemented is undistinguishable from higher sample rates.
Oh, and it would be interesting indeed to know what percentaje of studios used high sample rate PCM at the early 80's.
BTW, if one assumes that freq. over 20 KHz are not audible, it has no sense use high sample rates even at the studio. Bitdepths over 16 do have a sense at the studio, but high sample rates don't.
Or not. I think many studios have changed to hi-res formats simply because it's what the market asks for, whether it sounds better to the people at studios or not.
I disagree here, some studios were recording & mastering in higher bit rates and samplingfrequencies as far back as the early 1980s, I do not have the exact dates to hand
That's just not true. There were a variety of sampling rates and bit depths in use in the mid-late 1970s, before 48 and 44.1 were standard. These were proprietary systems used by specific companies and studios.
Whatever numbers were used, their performance was something different! They couldn't make a 16-bit accurate DAC for the launch of CD. What’s more, for ADCs, 50kHz+ sample rates made sense when they didn't oversample, and carried out all the filtering in the analogue domain.
IIRC people were thinking about 24/96 in the early-mid 1990s. Even in the mid 1990s, audiophile record companies were using 24/48 ADCs and running the analogue master tapes at half speed (and that's not easy, because the replay EQ changes!) because they believed the existing 24/96 converters just weren't good enough.
FWIW I think there is an audible difference, but even people who can't hear a difference will probably use it if the market demands, because it can't sound worse than CD.
Cheers,
David.
On your point of audibility, you are entitled to your position, however, it is simply not valid for the recording/mastering studios, if it were that clearcut why don't the studios just revert to 16/44.1 and forget about the higher resolutions or DSD altogether afterall the benefits are inaudible
A lot of people are more than happy with 48kHz 24-bit. Mike Oldfield (Tubular Bells) for one.
KikeG: Where did Bob Katz say CD quality was enough? From what I've read, that's not his position at all. He believes it can be much much better than it usually is, but he finds 24/96 better IIRC.
Cheers,
David.
KikeG: Where did Bob Katz say CD quality was enough? From what I've read, that's not his position at all. He believes it can be much much better than it usually is, but he finds 24/96 better IIRC.
Yes, it seems Bob Katz didn't say exactly this, my memories were not totally accurate. What he said was (http://www.audiomedia.com/archive/features...steningtest.htm (http://www.audiomedia.com/archive/features/uk-0400/uk-0400-listeningtest/uk-0400-listeningtest.htm)) :
"A properly-designed 20kHz digital filter can be sonically invisible in a 96kHz sampled environment.
2. Experience and this experiment suggests that 44.1kHz sampling digital systems can sound much better simply by use of better digital filters. This includes all the filters in compact disc players, A/Ds, etc. The effects of cumulative filters must also be considered —
a situation similar to the familiar effects of group delay in successive bandpass limited analogue circuits.
3. 96kHz sampling systems do not sound better because of increased bandwidth. The ear does not use information above 20kHz to evaluate sound."
He doesn't give all the details about the filter characteristics. But if we assume it cuts everything over 21 or 22 KHz, then:
a ) the only difference between a well implemented 44.1 KHz system, and a 96 KHz system with this lowpass at 20 KHz, is quantization noise level below 20 KHz.
b ) this quantization noise is not the issue at discussion here, and is a non-issue at all in case of 24-bit systems.
Then, it could be re-interpreted like that 44.1 KHz can be sonically transparent if properly implemented. However, he really doesn't say explicitly such thing.
Maybe the article is outdated, I don't know. Oh, and I don't consider Bob Katz the Bible of Audio, either. Many of his claims over better or worse sound, jitter audibility, etc, are not backed up by any kind of blind testing, just his impressions.
That's better - because that is the quote that I had read.
It begs one question (and I think Bob sees that it begs one question, but didn't answer it at that time, and didn't come back to it afterwards)...
If I take that 96kHz sampled 20kHz low pass filtered signal, and convert it to 44.1kHz, and then convert it back to 96kHz, does it still sound the same?
Nyquist says it can be the same, but Nyquist doesn't include the effects of quantisation and less-than-infinite-length filters. In the real world, the result can never be numerically identical, but can it be close enough to sound identical for all signals?
You would think so, wouldn’t you? But stranger things have happened. Using this wonderful filter, I can’t realistically see how 96k>filter>filter>96k (i.e. just running the filter another time) could sound any better than 96k>filter>48k>96k>filter. You can theoretically get the same result when going via 44.1kHz, but you would have to use a different filter in the resampling algorithm.
btw I don't know how the 20kHz filter in his experiment was designed, but I know it was over one second in duration. That's a seriously accurate (or steep, or both) filter! If it was linear phase, that's a serious amount of pre-ringing too, but it may have been minimum phase or similar.
Did you see the part where he asked the person (who created the original accurate filter for him) to go away and create a "typical bad" filter. He thought this shorter, typical DAC filter sounded like a cheap CD player.
EDIT: These filters are in a different article - I'm sure it was on his website, but he's re-arranged his site and my bookmark doesn't work. Maybe I'm making this part up?
I don't take what he says as gospel either, but he's a clever bloke (and a great bloke too - you should see the time he spent helping with Replay Gain), and he gains credibility by doing a test where he fully expects to hear a difference, and then publicly says that he hears no difference at all.
Cheers,
David.
If I take that 96kHz sampled 20kHz low pass filtered signal, and convert it to 44.1kHz, and then convert it back to 96kHz, does it still sound the same?
Nyquist says it can be the same, but Nyquist doesn't include the effects of quantisation and less-than-infinite-length filters. In the real world, the result can never be numerically identical, but can it be close enough to sound identical for all signals?
Well, I think that you can make a single 20 KHz filter that gives exactly same response both at 96 KHz and at 44.1 KHz sampling rates. If it works at 96, KHz, why shouldn't it work at 44.1 KHz?
You would think so, wouldn’t you? But stranger things have happened. Using this wonderful filter, I can’t realistically see how 96k>filter>filter>96k (i.e. just running the filter another time) could sound any better than 96k>filter>48k>96k>filter. You can theoretically get the same result when going via 44.1kHz, but you would have to use a different filter in the resampling algorithm.
Yes, that what I mean. Filter characteristic at 44.1 KHz could be made to be exactly the same than at 96 KHz, given that the stopband starts below 22 KHz (fs/2).
btw I don't know how the 20kHz filter in his experiment was designed, but I know it was over one second in duration. That's a seriously accurate (or steep, or both) filter! If it was linear phase, that's a serious amount of pre-ringing too, but it may have been minimum phase or similar.
...
EDIT: These filters are in a different article - I'm sure it was on his website, but he's re-arranged his site and my bookmark doesn't work. Maybe I'm making this part up?
Well, according to the linked article, the filter was just 255-tap. This is quite short and I think a quite usual length in FIR filters at hardware oversampling DACs.
he gains credibility by doing a test where he fully expects to hear a difference, and then publicly says that he hears no difference at all.
Yes, that's something that speaks good of himself. He has a big reputation too.
That's just not true. There were a variety of sampling rates and bit depths in use in the mid-late 1970s, before 48 and 44.1 were standard. These were proprietary systems used by specific companies and studios.
As far back as the dawn of CD, there were companies recording at > 16 bits. Even today in the era of 24 bit DACs, the true resolution of many DACs is not 24 bits. why should it have been any different 20 years ago. Thinking about it, the studio was recording at 20/48 and that was at the dawn of CD. The whole idea of recording with more bits is surely old hat.
Well, I think that you can make a single 20 KHz filter that gives exactly same response both at 96 KHz and at 44.1 KHz sampling rates. If it works at 96, KHz, why shouldn't it work at 44.1 KHz?
...
Filter characteristic at 44.1 KHz could be made to be exactly the same than at 96 KHz, given that the stopband starts below 22 KHz (fs/2).
Oops, what I missed is the fact that at 44.1 KHz you need two 20 KHz lowpass filters, the antialiasing one at the ADC and the anti-imaging one at the DAC. However, it's not very difficult to design them so that the combination of both gives the desired filter behaviour. For example, it possible to build one of the filters using a lot of taps, getting a behaviour very similar to a ideal brickwall filter at 22 KHz (similar to the 16383-tap filter at SSRC "slow" resampling), and relax frequency requirements of the other filter (a 255-tap filter starting at 20 KHz), so that this last one sets in practice the overall filter behaviour.
As far back as the dawn of CD, there were companies recording at > 16 bits.
If that's true, I think it's quite probably it was just because it was the only way they had to achieve closer to true 16-bit performance.
Even today in the era of 24 bit DACs, the true resolution of many DACs is not 24 bits.
Well, AFAIK no DAC has true 24-bit performance. Real-world physics constraints make it impossible.
Thinking about it, the studio was recording at 20/48 and that was at the dawn of CD. The whole idea of recording with more bits is surely old hat.
Are you sure of that? First cd's I knew of having been recorded using 20 bits appeared many years after launch of cd format in 1982.
As far back as the dawn of CD, there were companies recording at > 16 bits. Even today in the era of 24 bit DACs, the true resolution of many DACs is not 24 bits. why should it have been any different 20 years ago. Thinking about it, the studio was recording at 20/48 and that was at the dawn of CD. The whole idea of recording with more bits is surely old hat.
For a DAC to count as x-bits, you have to be able to resolve 2^x different output levels. Noise may swamp the output, but you can average out the noise to reveal whether there's really anything there, or not.
For a DAC to be a
good x-bits DAC, then each of those 2^x output levels needs to be in the numerically correct order, and each needs to be equally spaced from its neighbour.
The best current DACs are linear down to the 27th bit level.
"Good" 16-bit DACs from the dawn of the CD era couldn't even reproduce the 2^16 levels in the correct order, let alone make them equally spaced!
There may or may not have been "20-bit" devices in studios at this time - can you provide a reference?
Cheers,
David.
The best current DACs are linear down to the 27th bit level.
However, that doesn't mean they are true 24-bit performance. Thanks to dither, narrowband measurements can give figures such as those, but those narrowband measurements are not representative of actual dynamic range performance. You know, using those narrowband linearity measurement procedures, you can achieve linearity levels up to the 18th or even 20th bit level using regular 16-bit DACs.
For me, true 24-bit performance means that the DAC can achieve a wideband dynamic range of around 6.03*24=144 dB. No existing DAC that I know can achieve those, simply due to electronics thermal noise floor. An ideal, true 24-bit DAC would.
For a DAC to count as x-bits, you have to be able to resolve 2^x different output levels. Noise may swamp the output, but you can average out the noise to reveal whether there's really anything there, or not.
For a DAC to be a good x-bits DAC, then each of those 2^x output levels needs to be in the numerically correct order, and each needs to be equally spaced from its neighbour.
Sorry, if this has been answered in this thread before (it's a very long thread and i havn't read everything), but isn't this the reason, for this 1bit-stream-technology. You don't have to mention, how accurate this bit is relative to your reference? And that it is easyer to spped up the sampling rate, than increasing the number of DAC-bits
The best current DACs are linear down to the 27th bit level.
However, that doesn't mean they are true 24-bit performance. Thanks to dither, narrowband measurements can give figures such as those, but those narrowband measurements are not representative of actual dynamic range performance. You know, using those narrowband linearity measurement procedures, you can achieve linearity levels up to the 18th or even 20th bit level using regular 16-bit DACs.
For me, true 24-bit performance means that the DAC can achieve a wideband dynamic range of around 6.03*24=144 dB. No existing DAC that I know can achieve those, simply due to electronics thermal noise floor. An ideal, true 24-bit DAC would.
But if you will only accept that a DAC has true 24-bit performance if it can acheive both 144dB SNR and linearity beyond 24-bits, then you'll be waiting a long time!
Think of it this way... Consider a 4-bit DAC. Actually, compare 2 of them.
The first has a dynamic range of 90dB (so it can reproduce digital silence as, basically, silence), but it isn't linear at all - all the numbers are in the wrong order. So, it will make a perfect job of reproducing digital silence, but a useless job of reproducing anything else.
The second has a dynamic range of less than 24dB - so basically the last bit is always lost in noise (or full of noise, if we consider an ADC instead). However, the DAC is linear down to the equivalent of the 16th bit. So each level, when the noise is averaged away, is the corrent distance from each other level, at least to 1 part in 65535. This DAC can't reproduce digital silence, and adds noise to everything else. But beyond that noise, the reproduction is perfect.
The first DAC would sound like a bad guitar effects box. The second would sound like a very hissy radio.
So I think that (a) because we can't possibly make a DAC with 144dB dynamic range, and (b) because it couldn't possibly bring any benefit, and © we know that making a DAC linaer beyond the number of bits it's designed for will improve the perceived sound quality (see above extreme example!)... for these three reasons, I think that the linearity figure is much more important that the SNR figure.
It's better to improve both figures, but linearity is more important than SNR - at least at the levels we're talking about with modern 16 and 24-bit DACs.
Another thought: if you have a correctly dithered 16-bit recording, you can find singals within it which correspond to the 18th or 19th "bit". But these signals will be ruined by a 16bit DAC which isn't linear down to the 18th or 19th bit level.
Final thought: it's unlikely that 24-bit sounds better than 16-bit because of increased dynamic range. So, if we're to believe people who think that it does (and I know you don't believe them), we should look for another explanation. I think increased linearity in a 24-bit system is probably the reason. Even though, in theory, you can make just as linear a 16-bit system.
Cheers,
David.
Sorry, if this has been answered in this thread before (it's a very long thread and i havn't read everything), but isn't this the reason, for this 1bit-stream-technology. You don't have to mention, how accurate this bit is relative to your reference? And that it is easyer to spped up the sampling rate, than increasing the number of DAC-bits
Yes, that's the idea. But nothing is for free - if you have chance to read through the thread, you'll discover all the problems with 1-bit technology, and why state of the art converters are typically 3-6 bits, oversampled.
Cheers,
David.
Are you sure of that? First cd's I knew of having been recorded using 20 bits appeared many years after launch of cd format in 1982.
There may or may not have been "20-bit" devices in studios at this time - can you provide a reference?
I can provide the reference, however it is only on paper so I will have to grovel to get to the information but I will get to it soon enough. I think it is documented by a reliable source, a loudspeaker manufacturer, but remember these were the days before the internet.
KikeG you are right Tom Jung did use 20-bit recording to try and achieve 16-bit performance, His first commercial 20-bit release was in 1991, so I am assuming here that from this period 20-bit recording equipment was available commercially. However I will get to that the relevant paper info soon. Do these dates ring any bells?
Guys, help me here you guys are discussing linearity again without looking at those linearity graphs that require interpretation earlier on the thread. thanks in advance.
Guys, help me here you guys are discussing linearity again without looking at those linearity graphs that require interpretation earlier on the thread. thanks in advance.
That's because I haven't found an explanation of how the audio precision linearity test is carried out. We had an AP at uni, with all the manuals, but I don't have access to that now. I think they can be downloaded from their website, but you can do that as easily as me
I suspect they're using a decreasing amplitude sine wave, and looking at the amplitude of the output. I believe the plots you linked to show input level on the x axis, and output-input level on the y axis. If the output drops below the level of the input, then the system is undithered, and the input has dropped below the LSB and been lost. If it rises above the level of the input (as shown at the left hand side of the plots you linked to), then either it's being becoming swamped by noise (if the noise isn't averaged out - that's what I want to check), or there are extra harmonic distortion components present. As I haven't tried this kind of test (just looking at the amplitude of the output) I can't say what the two results you linked to actually mean. I usually examine the noise and harmonic distortion directly.
When I talk about linearity, I mean "does the ADC or DAC have a linear transfer function?". A 17-bit system has twice as many levels as a 16 bit system. An 18-bit system has twice as many levels as a 17-bit system. And so on... The extra levels in a 17 bits system will lie exactly half way between the levels on a 16-bit system. An 18-bit system will have 3 extra levels between each level of a 16-bit system.
A diagram would help. I've taken the region between two adjacent levels in a 16-bit system, and shown where the levels in a 17-bit, 18-bit, 19-bit and 20-bit system would fall:
16 17 18 19 20
__ __ __ __ __
__
__ __
__
__ __ __
__
__ __
__
__ __ __ __
__
__ __
__
__ __ __
__
__ __
__
__ __ __ __ __
These "levels" are the analogue voltage levels above which the digital output will be the next highest value.
The above diagram assumes we have a perfect system. Let's assume the 16-bit system isn't perfect, and the levels don't quite fall where they should. Imagine that the top level shown is pushed about 1/3rd of a bit too low - i.e. the top left-hand level on that diagram is moved about 1/3rd of the way down. It would still fall within the same level on a 17-bit system, but it's actually jumped into a different (wrong!) level on an 18-bit system. so it's linear to 17-bits, but no further.
Conceptually, this explains linearity. In reality, most DACs aren't made using 2^16 or 2^20 discrete levels (that would be pure multi-bit technology with no noise shaping), and are swamped by noise anyway. However, in a typical oversampled DAC, it's still possible to average away the noise, measure the distortion, equate this to a bit-level, and conclude that it is linear to so many bits.
Cheers,
David.
But if you will only accept that a DAC has true 24-bit performance if it can acheive both 144dB SNR and linearity beyond 24-bits, then you'll be waiting a long time!
I know, and I understand your explanations. But a 16-bit DAC with true 16-bit performance is supposed to have near 96 dB (94 dB with dither) dynamic range performance, and linearity levels quite above 16-bit. And there are such 16-bit DACs readily available. The fact that real 24-bit DACs can't perform accord to all these same requirements, is, in my opinion, not a reason for reducing requirements for saying a DAC has true 24-bit performance.
I mean, would you call a 16-bit DAC that has a poor 80 dB SNR or dynamic range performance, but linearity levels up to 100 dB, a DAC with true 16-bit performance? A good 14-bit DAC could outperform it in every sense, and would still be only 14-bit.
BTW, I did some empiric tests with Spectralab, and it resulted that, with an ideal DAC, using flat dither, a 65 K FFT and some seconds of averaging, it's possible to resolve signals up to 6 bit over the actual bitdepth used. With a 16-bit system, I could resolve signals whose level is just over -135 dB (22 bit level). With a 24-bit system, signals over -184 dB (30 bit level). As it has been said many times, ideally, using a infinite FFT length (infinitely narrowband analysis), infinitely small signals could be resolved with any resolution.
A quick & dirty law that I figured out according to these measurements would be:
R=-10 + 6.02 * nbits + 10 log NPointFFT
Where R is the max. resolution achievable, or the limit at which signals can't be resolved, or the limiting narrowband noise floor. I'm sure there is a more accurate mathematical law that explains this, this is just some empiric law that seems to work well for the range and kind of measurements I performed.
Um, going off topic, but I think it's better than starting a new thread:
I just got some specs for the ADC and DAC of a processor I'm working on for a project:
Analog Input/Output
Analog Input:
24bit quantization, 96KHz sampling frequency。
SNR 90dB
Dynamic range: 90dB
Harmonic distortion: -80dB
Analog Output:
24bit quantization，96KHz sampling frequency
SNR 100dB
Dynamic range: 90dB
Harmonic distortion: -88dB
Headphone amplifier output in power: 40mW
Digital Input/Output
S/PDIF input and output
96KHz sampling frequency
USB Interface
Adopting USB1.1 standard
Standard Audio USB interface
The end product will be a DSP board (to be integrated into amplifiers) with an expected unit price of ~$100. Do you think the specs (SNR, dynamic range, THD) are good? Bad? Since we are on the topic of these 3 specs here... Thanks ^^"
It's not bad, but it's not spectacular either. Output performance is a little bit below 24/96 performance of cards such as the Audiophile or Revolution. Input performance is somewhat worse, compared with these cards.
I think the target for this card is to be integrated in standalone hifi components...
BTW, I did some empiric tests with Spectralab, and it resulted that, with an ideal DAC, using flat dither, a 65 K FFT and some seconds of averaging, it's possible to resolve signals up to 6 bit over the actual bitdepth used. With a 16-bit system, I could resolve signals whose level is just over -135 dB (22 bit level). With a 24-bit system, signals over -184 dB (30 bit level). As it has been said many times, ideally, using a infinite FFT length (infinitely narrowband analysis), infinitely small signals could be resolved with any resolution.
Sorry to drag this thread up - I missed your reply KikeG.
What you say is true, but I think you're suggesting it makes the whole thing bogus, whereas it doesn't. Of course an "ideal" DAC will give infinite resolution (given an infinitely narrowband analysis). That's the reason that measuring the
actual resolution of a
real DAC is a useful indicator.
However, you're also right that it seems silly to do this in the presence of comparatively large amounts of noise.
Of course, we can quote the noise figure
and the linearity figure, but I think it would be useful to know which is the limiting factor...
I'd suggest a useful approach would be to introduce some psychoacoustics. (For once in this thread, we have some psychoacoustic knowledge which we can apply!).
Here is my idea: Perform an analysis using critical band filters (i.e. filters having the same selectivity as those in the human ear). These are quite wide, so will give much poorer frequency resolution than a stupidly long FFT. However, the output can still be averaged to remove noise, because the human ear seems to do this. You would need a limit on the amplitude accuracy of each spectral bin, but this could nominally be 1dB, subject to further tuning. This means any signal which is less than 1dB above the noise in this analysis is judged to be lost in the noise.
It should be possible to perform a "linearity" analysis of a DAC using this method. Quite simply, any noise will put a limit on
this "linearity" measurement, because it will swamp the signal. So you could say that, at 1kHz, the DAC was linear to 27-bits, the noise was at the 20th bit level, and a perceptual analysis revealed the DAC was useful for human listeners up to the equivalent of 23-bits (these are made up figures).
There's one obvious flaw in this idea: the human ear has a much higher noise floor than some of these DACs (depending on the replay level), so this kind of analysis could be perceptually meaningless.
At the end of the day, I can't argue with your assertion that the ideal 24-bit DAC does not (and will not!) exist. However, the industry needs a useful way of talking about them (and expressing performance in a single, impressive number!) - maybe this perceptual analysis is the fairest and most useful approach.
Cheers,
David.
If i am right the first 20Bit recordings appeared in 1987. Sting-Nothing Like The Sun was one of these recordings.
I have no reference found on the net but can remember back.
Wombat