Head-Fi.org › Forums › Equipment Forums › Sound Science › Hi-Rez - Another Myth Exploded!
New Posts  All Forums:Forum Nav:

Hi-Rez - Another Myth Exploded! - Page 8

post #106 of 156

Finally, this seems an argument from authority, particularly in how you frame the initial post.

 

If you truly do mean "argument from authority," I'm very puzzled.  In the initial post I ask questions.  I don't claim to know the answers (i.e., I don't claim to be an authority myself), and I don't cite answers from anyone else, nor claim anyone else to be authoritative.  Without authorities or argument, how can this be an "argument from authority"?

 

Regarding the rest of your comment, that's fair discussion of the topic, which is what I hoped to promote in that thread and by referring to it here.  So thank you.

post #107 of 156

Quote:
Originally Posted by judmarc View Post

Finally, this seems an argument from authority, particularly in how you frame the initial post.

 

If you truly do mean "argument from authority," I'm very puzzled.  In the initial post I ask questions.  I don't claim to know the answers (i.e., I don't claim to be an authority myself), and I don't cite answers from anyone else, nor claim anyone else to be authoritative.  Without authorities or argument, how can this be an "argument from authority"?

 

Regarding the rest of your comment, that's fair discussion of the topic, which is what I hoped to promote in that thread and by referring to it here.  So thank you.


You seem to think posters in that thread are pretty authoritative:

 

 

Quote:
Originally Posted by judmarc View Post

Got lots of good feedback from audio engineers, university professors, authors of audiophile playback software, etc., a couple of whom are quite familiar with details of the recording and playback process.  There's at least one academic research article cited as well.

post #108 of 156
Quote:
Originally Posted by judmarc View Post

Finally, this seems an argument from authority, particularly in how you frame the initial post.

 

If you truly do mean "argument from authority," I'm very puzzled.  In the initial post I ask questions.  I don't claim to know the answers (i.e., I don't claim to be an authority myself), and I don't cite answers from anyone else, nor claim anyone else to be authoritative.  Without authorities or argument, how can this be an "argument from authority"?

 

Regarding the rest of your comment, that's fair discussion of the topic, which is what I hoped to promote in that thread and by referring to it here.  So thank you.


You're quite welcome. Perhaps I'm reading this too closely, but your post over on Ausdiophile leads off with the following:
Quote:
IMHO, Shannon/Nyquist (perfect reconstruction of a waveform is possible if one samples at a rate exceeding 2x the frequency) may be one of the most misused bits of mathematics/information theory ever. I've seen it trotted out over and over again to "prove" digital is "perfect sound forever," and that various items in the signal chain (CDs, CD players, cables, DACs) cannot make a difference in sound quality. In this thread I'd like to get a better undertanding from Those in the Know about harmonics and Shannon/Nyquist.

This suggests a slightly adversarial stance to a fairly noncontroversial piece of mathematical theory, not to mention those scare quotes. The theorem has nothing to say about the effect that different pieces of equipment can have on the signal chain. Nor has anyone (IIRC) made that claim here. What is does say is that higher sampling levels are not going to get you more bang for your audio buck.

HeadInjury has covered the argument from authority issue, nicely. You don't have to claim authority, but you can point to people who you've accorded some expertise or knowledge to make the point for you. It's a rhetorical maneuver.
post #109 of 156

While I am in no position to argue the mathmatics of this equation I do reguarly digitize vinyl and find that going from 44.1 to 96 to 192 has audible benefits on the adc front. The soundstage in noticiabley wider in higher resolution encodings. Playback is less clear as when I resample the 24/192 to dithered 16/44 the difference is less apparent.

 

I will say that Waltz for Debby mastered by Paul Stubblebine at 24/192 is the best of the digital versions that I have heard. Sadly there is no 16/44 equivilent and I have yet to process this high res sample to a lower format. I use a fairly high end adc/dac the metric halo uln - 8.

post #110 of 156
Thread Starter 
Quote:
Originally Posted by judmarc View Post

Just wanted to invite folks here to have a look at a thread on Computer Audiophile (yes, I started it) : http://www.computeraudiophile.com/content/Better-Understanding-ShannonNyquist

 

Got lots of good feedback from audio engineers, university professors, authors of audiophile playback software, etc., a couple of whom are quite familiar with details of the recording and playback process.  There's at least one academic research article cited as well.  Some of the information there does seem to run counter to a few things stated in this thread as fact.  That doesn't mean the folks there are necessarily correct and those here not so, or vice versa; but I do think it's healthy to get different perspectives from (apparently) knowledgeable people.


None of that thread, as far as I can see, addresses the point I've made in this thread. There was quite a lot of talk of alias images but that was a bit of a red herring as the anti-alias filters in modern ADCs operate in the megahertz range and are really a non-issue. No, the real problem comes in the decimation filter and this was not addressed in the thread you linked to.

I agree in theory with Miska's post, it would be nice to record everything at 384kS/s and not have to think about any missing information above Nyquist. In practice though Miska is missing the point that the higher the sample rate the less accuracy we get. Operating at 192kS/s requires 4 times the number of taps (coefficients) for the decimation filter to reject aliasing but we have only a quarter the amount of time to execute those taps, it's just not possible. I have yet to see (or hear of) any AD chip achieve any better than 80dB alias image rejection when running at 192kS/s, they just can't run enough taps in the time. At 44.1kS/s a decent AD chip should easily manage 120dB (or better) of alias image rejection and therefore any aliasing would be below the noise floor.

I've tried to keep this thread simple but it's becoming almost impossible. I've mentioned mics being a problem from the simplistic point that few studio mics have any response above 48kHz. I saw someone's post (in your linked thread) about using mic arrays but again, there are some inherent problems with mic arrays: 1. There's the obvious phase coherency problem with any pair (or more) of mics, with the exception of MS pairs (but that configuration is not suitable for this application). Even a near coincident pair introduces phase cancellations in high frequencies. 2. There is the problem of colouration variations between different mics within the same array, even worse in this case, when there is nothing you can do about it because you can't hear the ultrasonic range to mix or EQ the mics properly! It's just a can of worms up there (ultrasonic freqs) and I haven't even mentioned the possibility of intermodulation distortion on playback. I just don't see the logic of having to deal with a whole bunch of complex issues created by trying to record and mix frequencies which we can't perceive. Time is always a limiting factor in today's world of music and sound creation and I would far rather spend my time improving the quality of the recording and mixing where it will make a real difference, in the hearing spectrum!

G
Edited by gregorio - 9/21/11 at 6:48pm
post #111 of 156
Thread Starter 
Quote:
Originally Posted by jp11801 View Post

While I am in no position to argue the mathmatics of this equation I do reguarly digitize vinyl and find that going from 44.1 to 96 to 192 has audible benefits on the adc front. The soundstage in noticiabley wider in higher resolution encodings. Playback is less clear as when I resample the 24/192 to dithered 16/44 the difference is less apparent.

 

I will say that Waltz for Debby mastered by Paul Stubblebine at 24/192 is the best of the digital versions that I have heard. Sadly there is no 16/44 equivilent and I have yet to process this high res sample to a lower format. I use a fairly high end adc/dac the metric halo uln - 8.


In this thread I have attempted to explain why 192kS/s is flawed. The theory and mathematics behind this claim is, as far as I am aware, uncontested (see the Lavry paper I linked to in the OP). So, all I can say with confidence is that there would have to be quite a serious design flaw for your ADC to produce more linear results at 192k than at 96k. However, the fact that 192kS/s is less linear (more distorted) than say 96kS/s does not necessarily mean that to some people it will sound worse. We can say that putting a tube in your signal chain is far less linear than not having a tube but again, there are some people who prefer the tube sound. I'm not trying to be insulting, but there can be only three explanations for your observations:

1. You prefer the additional distortions exhibited by 192kS/s recordings.
2. Your ADC has a design flaw.
3. Placebo effect.

Of course, it maybe a combination of these. For example, there is no logical reason why you should be able to hear any difference between a recorded 16/44 file and a resampled 16/44 file unless there is a design flaw in your ADC.

If anything, your observations support my argument against 192k sample rates. Baring in mind there is nothing on your vinyl which falls outside the limits of 16/44 and certainly no frequencies above 48kHz, the only thing which could possibly make any audible difference with your 192kS/s vinyl rips is the distortion!

I obviously can't comment on the quality of mastering on a particular recording. As I said earlier in the thread, the quality of recording, production and mastering is going to have massively more impact on perception of quality than which format it's on. Even though 192kS/s is flawed in comparison to other digital formats, we are still talking about differences which should be near the limits of audibility (depending on implementation).

G
Edited by gregorio - 9/21/11 at 6:21pm
post #112 of 156

Jud, respectfully, you keep re-framing the same argument over and over to attempt to get a response that you'd like, even on the other forum. You just *know* that the whole digital sound thing is hooey, so if you can latch it down to one flagship topic, like Shannon Nyquist, and find holes in it, then you can sink the whole fleet. The problem is that you don't seem to really understand the mathematics involved and so your rhetorical arguments come off as straw men. I mean, yes, you certainly appear to understand the basic concepts and such, but in your comments I still don't see a good grasp of the Fourier series or how taking advantage of it allows for a perfect transcription of a piece of audio.

 

In determining real-world performance of sampled and quantized audio, aliasing and jitter are two notable hurdles. I don't think you'll find any disagreement that they need to be addressed. And they have been, to the extent that current designs render them inaudible.

 

The big joke is that the demands for audio transparency really aren't that high in a piece of kit. You need 20-20,000hz, 80db SNR, -40db crosstalk & less than 1% THD (-40db) and presto! Most modern hi-fi DACs beat 100db SNR & .01% THD without breaking a sweat (and a good amp will match those figures). Good news, we can stop worrying so much about the source and focus on what makes magnitudes more of a difference - the speaker/headphone.

post #113 of 156


 

Quote:
Originally Posted by gregorio View Post

Quote:
Originally Posted by jp11801 View Post

While I am in no position to argue the mathmatics of this equation I do reguarly digitize vinyl and find that going from 44.1 to 96 to 192 has audible benefits on the adc front. The soundstage in noticiabley wider in higher resolution encodings. Playback is less clear as when I resample the 24/192 to dithered 16/44 the difference is less apparent.

 

I will say that Waltz for Debby mastered by Paul Stubblebine at 24/192 is the best of the digital versions that I have heard. Sadly there is no 16/44 equivilent and I have yet to process this high res sample to a lower format. I use a fairly high end adc/dac the metric halo uln - 8.


In this thread I have attempted to explain why 192kS/s is flawed. The theory and mathematics behind this claim is, as far as I am aware, uncontested (see the Lavry paper I linked to in the OP). So, all I can say with confidence is that there would have to be quite a serious design flaw for your ADC to produce more linear results at 192k than at 96k. However, the fact that 192kS/s is less linear (more distorted) than say 96kS/s does not necessarily mean that to some people it will sound worse. We can say that putting a tube in your signal chain is far less linear than not having a tube but again, there are some people who prefer the tube sound. I'm not trying to be insulting, but there can be only three explanations for your observations:

1. You prefer the additional distortions exhibited by 192kS/s recordings.
2. Your ADC has a design flaw.
3. Placebo effect.

Of course, it maybe a combination of these. For example, there is no logical reason why you should be able to hear any difference between a recorded 16/44 file and a resampled 16/44 file unless there is a design flaw in your ADC.

If anything, your observations support my argument against 192k sample rates. Baring in mind there is nothing on your vinyl which falls outside the limits of 16/44 and certainly no frequencies above 48kHz, the only thing which could possibly make any audible difference with your 192kS/s vinyl rips is the distortion!

I obviously can't comment on the quality of mastering on a particular recording. As I said earlier in the thread, the quality of recording, production and mastering is going to have massively more impact on perception of quality than which format it's on. Even though 192kS/s is flawed in comparison to other digital formats, we are still talking about differences which should be near the limits of audibility (depending on implementation).

G


Odd after reading your post I checked your profile. If you are actually in pro audio how is it that you are not aware of the Metric Halo ULN - 8.

 

I wasn't arguing your point or against just making an observation. I find that people that are interested in argueing a point often miss the obvious because their ego is connected to their argument :-)

 

Check them out they are used by a slew of top mastering engineers. I doubt 24/192 from this unit is exhibiting any distortion other than capturing distortion from tracking error off the cart.

 

post #114 of 156

Quote:
Originally Posted by jp11801 View Post

I wasn't arguing your point or against just making an observation. I find that people that are interested in argueing a point often miss the obvious because their ego is connected to their argument :-)


What "obvious"?

post #115 of 156
Quote:
Originally Posted by jp11801 View Post


 



Odd after reading your post I checked your profile. If you are actually in pro audio how is it that you are not aware of the Metric Halo ULN - 8.

 

I wasn't arguing your point or against just making an observation. I find that people that are interested in argueing a point often miss the obvious because their ego is connected to their argument :-)

 

Check them out they are used by a slew of top mastering engineers. I doubt 24/192 from this unit is exhibiting any distortion other than capturing distortion from tracking error off the cart.

 



The Metric Halo ULN-8 is mainly used by mastering engineers...especially those who prefer using Macs. I believe Gregorio isn't a mastering engineer but a mixing engineer. Regardless, I would trust what Gregorio has to say.

post #116 of 156
Thread Starter 
Quote:
Originally Posted by jp11801 View Post

Odd after reading your post I checked your profile. If you are actually in pro audio how is it that you are not aware of the Metric Halo ULN - 8.

 

I wasn't arguing your point or against just making an observation. I find that people that are interested in argueing a point often miss the obvious because their ego is connected to their argument :-)

 

Check them out they are used by a slew of top mastering engineers. I doubt 24/192 from this unit is exhibiting any distortion other than capturing distortion from tracking error off the cart.


I started using products from Metric Halo over a dozen years ago and although I know of the ULN-8, I haven't heard one. But, I think you may have misunderstood both my reply and this thread. I am not saying that your ADC is a bad ADC in comparison with other ADCs. I am saying that all ADCs on the planet exhibit higher distortion at sample rates of 176.4k and above than at 96kS/s. The theoretical reasons why this is so have been known for many years, measurements bear it out and a number of manufacturers confirm this is true (as stated earlier in this thread).

Also, simple logic should be telling you this assertion is correct. Vinyl does not contain any frequencies above 48kHz, in fact, vinyl is usually rolled off above 16kHz. So, with this in mind, there are a couple of questions you should ask yourself:

1. If there are no freqs above let's say 20kHz on your vinyl, what is it that you are recording between 48kHz and 96kHz which makes your 192kS/s sample rate sound different?
2. What is the frequency response of your cans/speakers?

Obvious points: Whatever differences you are hearing must be in the pass band (or slightly into the transition band) of your cans/speakers. If you are hearing differences, then by definition the differences you are hearing must be within the frequency spectrum of human hearing (20Hz-20kHz).

So what can you be hearing? If you are listening at very high playback levels and you are not using a decent noise-shaped dither algorithm, it is possible that when listening to 16/44 you are hearing dither noise. Although, that potential problem should be eliminated by using 96kS/s. Another possibility is that you are hearing inter-modulation or other distortion in the hearing band caused by the poor alias image rejection inherent with 192kS/s. The final option, as mentioned, is aural illusion (placebo effect). Without some fairly detailed measurements of your recording and playback chain it is impossible to know which of these options is responsible for what you are perceiving.

I would agree with this statement of yours completely: "I find that people that are interested in argueing a point often miss the obvious because their ego is connected to their argument :-)"

G
Edited by gregorio - 9/22/11 at 3:31am
post #117 of 156

I did some A/B tests  with the 96/24 version of Ron Carter / Rosa Passos 'Entre Amigos' vs the 44/16 bit version.  My subject was a pro jazz musician. Playback was with Audio-GD  NFB-10ES as the DAC and headphone amp, headphones were LCD-2 and HD-800. Also used was a Soderburg modified Forte 4a and a pair of Quad ESL-57's refurbed by Wayne Piquet.  We also used a pair of Mackie powered monitors that the musician was well familiar with at times. (I didn't like these speakers but he was very familiar with their sound, having used them in pro recording sessions, movie soundtrack creation and so on.)

 

We also used SACD and Red Book tracks from several hybrid SACDs using a well-regarded Sony SACD player (model number escapes me right now.)  

 

Neither my pro musician pal nor I could hear a difference.  We listened very carefully, for quite a while.

 

I'm not saying that there is NO audible difference.  I'm saying WE couldn't hear any difference.  I suspect that if there were MAJOR SONIC BENEFITS from these hi-res tracks, we should have easily heard them.  Since we listened very carefully to quite a number of things, my conclusion is that if there really were clearly audible sonic benefits we would have heard them, and since we heard no such thing, I suspect there are no clear sonic benefits from these high-res formats. Maybe some really subtle benefits are present, and they escaped our attention- but if the benefits are so subtle that a pro musician and a trained 'audiophile' listener couldn't hear them, then I conclude that any such benefits are not worth all the trouble and expense of obtaining these high-res tracks. Your mileage may vary- and you are entitled to your own opinion, even though it's wrong wink.gif

 

That said, there certainly IS a benefit to be had in the studio when recording music- various studio manipulations and other elements of digital production yield much better sounding results when you start with higher resolution tracks to begin with, even though in the end it's mastered to 44/16..   It's just like editing a digital photo- the results of your cropping and photoshopping the image will be much better if you start with a 12 megapixel image shot on a prosumer D-SLR than if you began with a 4 megapixel image from a cellphone.

 

As to Red Book CDs being 'compressed' as some have claimed- well, there is something over 100 dB dynamic range available from a 44/16 CD.  I doubt there is a symphony hall anywhere in the world where the range between background noise and maximum orchestral crescendo is more than 75 dB. A rock concert is usually between 80 dB in the quiet parts and 125 dB at it's loudest, a range of 45 dB!  A really good vinyl playback system struggles to reach 70 dB. So, the 100 dB dynamic range of 16 bit CDs is not compressed in any meaningful sense.  Now, what the music producers do to the signal is another matter- we all know that many CDs, even classical ones, are made LOUD, which is to say, compressed.  But this is NOT a problem with the format.  Maybe in producing 96/24 tracks from original masters, the producers don't feel the pressure to win a loudness war, so in that sense high-rez tracks may have less tampering with the dynamics than the commercial CD release.  But again, this is NOT a problem with the 44/16 format, but a problem with industry practice.

 

It's also interesting to note that there is an (in)famous Audio Engineering Society white paper from a few years back which presents listening results from a large sample double-blind study of 44/16 and 96/24, and this study showed that none of the listeners tested could hear the difference.  (Actually, there are a few papers on this topic-  see: https://secure.aes.org/forum/pubs/journal/?ID=2 and  http://www.aes.org/e-lib/browse.cfm?elib=3839  and  also see material from these studies quoted in the Wikipedia article see http://en.wikipedia.org/wiki/Super_Audio_CD#Audible_differences_compared_to_PCM.2FCD )

post #118 of 156
Quote:
Originally Posted by gregorio View Post



None of that thread, as far as I can see, addresses the point I've made in this thread. There was quite a lot of talk of alias images but that was a bit of a red herring as the anti-alias filters in modern ADCs operate in the megahertz range and are really a non-issue. No, the real problem comes in the decimation filter and this was not addressed in the thread you linked to.

I agree in theory with Miska's post, it would be nice to record everything at 384kS/s and not have to think about any missing information above Nyquist. In practice though Miska is missing the point that the higher the sample rate the less accuracy we get. Operating at 192kS/s requires 4 times the number of taps (coefficients) for the decimation filter to reject aliasing but we have only a quarter the amount of time to execute those taps, it's just not possible. I have yet to see (or hear of) any AD chip achieve any better than 80dB alias image rejection when running at 192kS/s, they just can't run enough taps in the time. At 44.1kS/s a decent AD chip should easily manage 120dB (or better) of alias image rejection and therefore any aliasing would be below the noise floor.

I've tried to keep this thread simple but it's becoming almost impossible. I've mentioned mics being a problem from the simplistic point that few studio mics have any response above 48kHz. I saw someone's post (in your linked thread) about using mic arrays but again, there are some inherent problems with mic arrays: 1. There's the obvious phase coherency problem with any pair (or more) of mics, with the exception of MS pairs (but that configuration is not suitable for this application). Even a near coincident pair introduces phase cancellations in high frequencies. 2. There is the problem of colouration variations between different mics within the same array, even worse in this case, when there is nothing you can do about it because you can't hear the ultrasonic range to mix or EQ the mics properly! It's just a can of worms up there (ultrasonic freqs) and I haven't even mentioned the possibility of intermodulation distortion on playback. I just don't see the logic of having to deal with a whole bunch of complex issues created by trying to record and mix frequencies which we can't perceive. Time is always a limiting factor in today's world of music and sound creation and I would far rather spend my time improving the quality of the recording and mixing where it will make a real difference, in the hearing spectrum!

G


Just wanted to say I very much appreciate this comment, as (like the CA thread I referenced earlier) it helps me begin to think about concepts and information I was unaware of previously, for example your discussion of the time problem at higher sampling rates.  (I actually believe Miska is aware of the time problem, based on some of the statements he's made in this and other threads at CA.)  Re mic arrays, I did wonder what Miska was trying to accomplish by having recording capability up to or beyond 250kHz!

 

Jud, respectfully, you keep re-framing the same argument over and over to attempt to get a response that you'd like, even on the other forum. 

 

anetode, get with the program!  (Just joking.)  What I'm trying to do is really two things: (1) Learn more - always the first priority.  I can't do that by accepting statements, even those I agree with, at face value, so I continue to ask questions.  Yes, I'm sometimes asking those questions from a skeptical viewpoint, but not nearly always.  (2) Somewhat as a parallel to the first, trying to look beyond "received wisdom."  I'm old enough to've had a bellyful of people who obviously don't know the science and engineering making statements like "cables can't possibly affect the sound" - as if wires in an audio system were somehow immune to common electrical interference, ground interconnection noise, RFI, etc., and audio engineers could not possibly vary in the success of their designs to cope with these problems.  One of the topics that really seems to attract this sort of ultra-reductionist thinking is digital audio - after all, "it's all just 1s and 0s."  Thus a topic like this one - hi-res a "myth" - attracts my attention.

 

But what I think gregorio in particular and the rest of you as well have managed to do in this thread is to provide backing for your statements, which as I said above I appreciate, because it really helps with priority #1.

 

post #119 of 156

So.... may I ask why we are discussing about the topic, which is already repeatedly beaten to death in hydrogen audio forum..... since as early as 2003?

 

It was very clearly answered multiple times, yet after 8 years later, we are still debating already answered topic with such passion. popcorn.gif


Edited by wnmnkh - 9/22/11 at 5:45am
post #120 of 156
Thread Starter 
Quote:
Originally Posted by wnmnkh View Post

So.... may I ask why we are discussing about the topic, which is already repeatedly beaten to death in hydrogen audio forum..... since as early as 2003?

 

It was very clearly answered multiple times, yet after 8 years later, we are still debating already answered topic with such passion. popcorn.gif


Because this is predominantly a consumer forum and unfortunately many (if not most) of the manufacturers of consumer digital audio equipment (and some music sales companies) either haven't read the information in the HAF (or other published documentation) or deliberately ignore and avoid it. This allows them to continue making and selling higher and higher sample rate products by creating the false belief that bigger numbers means higher quality.

So while you shouldn't find too many experienced, knowledgeable professionals who don't know the issues of high sample rates, you can't expect this level of understanding of quite complex issues from the vast majority of consumers. It would certainly be pointless starting this thread on the HAF or many of the other professional oriented forums but here, on head-fi, I thought it would be useful to some. The reason for all the discussion is that digital audio theory is counter-intuitive in many ways and that usually means a lot more time and effort is required to arrive at a good understanding, because logical but incorrect assumptions have to be battled along the way. That was certainly true for me when I first started looking more in depth into digital audio.

G
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Hi-Rez - Another Myth Exploded!