Audible Jitter/amirm vs Ethan Winer

[Ethan Winer]

I would argue that riding the volume up during quiet passages, or any manner of playing the material at abnormally high volume for the purpose of pushing the distortions into audibility is cooking the tests.

I agree, and I have learned to say "at normal levels" when discussing jitter and dither and the benefit of 24-bit recording. I'm pretty sure I never said "jitter is never audible in any circumstance," but if I did I meant any normal circumstance. If you have to crank the volume on soft parts such that playing other parts of the same music would blow out your speakers or your eardrums, then it's not a valid condition.

As for blind versus double blind, I think that single blind can be adequate in many cases. I've tested people blind in my own studio to my satisfaction. I knew which file was which, but the listener didn't and I'm confident my facial expressions gave nothing away. The last the I did such a test I was facing away from the listener. I also test myself blind by closing my eyes so I can't see which state the solo buttons are in.

--Ethan

 

[Phelonious Ponk]

That is too funny. I think you should "trademark" that term in audio at the very least

Perhaps, but in a hobby filled with people who believe they are impervious to expectation bias, it's probably not worth much more than an occasional laugh, a consistent sneer. This place has its share of fire-breathing audiophiles, but it has some balance as well. Other boards bristle at the mention of any kind of objectivity. On another board I visit, there has been an on-going discussion of counterfeit hi-res - SACDs on Amazon, files on HDD Tracks - that have been discovered to be nothing more than upconverted 16/44.1, complete with the 20khz ceiling. The discussion has taken up 3 or 4 threads, many pages, and has covered the ethical considerations, the need for renumeration, a debate over the responsibilities of the producers vs. the retailers and on and on.

What has hardly been touched are the implications of the fact that so many owners and users of these recordings had no idea they were not genuine until someone put a scope on them and published the information on the internet.

P

 

[Ron Party]

Amir, <snip>

I've read the description you've posted here and (in what seems like another lifetime) at AVS, and IIRC in all of the tests you've run you were listening through cans. Have you heard what you describe as jitter through loudspeakers?

No I have not tested jitter using speakers. <snip>

I was asked once whether I thought it would be as easy to hear these artifacts with speakers and I said no. Happy to elaborate once we cover some of the main points of the debate.

I agree, and I have learned to say "at normal levels" when discussing jitter and dither and the benefit of 24-bit recording. I'm pretty sure I never said "jitter is never audible in any circumstance," but if I did I meant any normal circumstance. If you have to crank the volume on soft parts such that playing other parts of the same music would blow out your speakers or your eardrums, then it's not a valid condition.

--Ethan

A few additional thoughts, all of which tie into the concept of a person actually detecting what Amir has identified as jitter in real world situations.

I would like to see this discussion to evolve in the direction of detecting jitter while listening through loudspeakers.

Amir stated he has only detected what he has identified as jitter while listening through cans. He left a piece of bait out there about the unlikelihood of detecting the same while listening through loudspeakers. Given the wide misunderstanding of the subject matter and numerous (but IMO erroneous) declarations of the audibility of jitter, including some right here at WBF, Amir, I would like to ask you to further expound on this.

I also think it is important to focus on Ethan's point about *normal* listening levels.

I suppose there is a range of what might be considered normal, e.g., I know I like to listen at much louder levels than most I know. (I have a SOTA headphone preamp and what many have described as one of the top 3 cans ever made, but I cannot imagine turning up the volume anywhere near what I do when listening through loudspeakers.)

Having stated my own preference, however, a discussion on the detection of what Amir describes as jitter at much lower listening levels through loudspeakers is essential to the discussion, particularly if we're looking for something that is -60 to -80 dB.

A couple of other passing thoughts. It is my understanding that there is an abundance of gear for which jitter has been measured at less that 250 picoseconds. If this is the case, then when trying to detect the audibility of jitter by listening through loudspeakers, are we discussing a moot point, a question made quite significant since there are many DACs on the market in the 5 figure range?

And, finally, how come there has been no discussion in this debate about jitter, or a jitter equivalent, in vinyl playback, which in a best possible case has only about 12 bits of dynamic range?

 

[DonH50]

To address only the last question Ron raised: The jitter under discussion is related to the "timing noise" on the sampling clock in a digital system. There is no sampling noise in a purely analog system. There is broadband noise that could be compared to random jitter, and phase issues in tracking (and the electronics) that could be compared to deterministic jitter.

In my opinion, dynamic range is a red herring in much of this discussion. Not because it does not matter, but because (a) there's usually more than enough dynamic range to hear the music above the noise, (b) like many other systems, our ears can pull signals from below the noise floor via filtering and averaging, and (c) pure noise is much less objectionable than distortion, whether induced by jitter, nonlinearity in the electronics, or something else.

My 0.000001 cents - Don

 

[RBFC]

I have what is perhaps an uninformed question:

If jitter distortion is at -80dB (for example), is it possible that the combined distortions of the reproduction chain can become cumulative enough for total system distortion to reach typical audibility? Is it possible that jitter is one of the items to be summed as distortion approaches the threshold of audibility? Would it not make sense to reduce those distortions that can be effectively reduced to avoid an audible "sum"? Which distortions contribute most heavily to audible disruption?

Maybe best for another thread or debate, but this "summation" seems like an important concept to me.

Lee

 

[DonH50]

"Summation" can be coherent or not, which changes significantly how much each component "adds" in the chain. Gain, bandlimiting and such also have an impact. Random jitter, like random noise, is not "distortion" from an engineering perspective, btw.

Of course it makes sense to reduce all noise and distortion compenents as much as reasonable, where "reasonable" implies not only theoretically possible but practically realizable and at a given price point.

I am not sure what you mean by "disruption". If objectionable to the listener, then distortion is likely more disruptive than noise in practical systems, and intermodulation or non-harmonic disortion more so than harmonic distortion. Most error sources do not sum linearly due to phase differences and are RSS'd -- root-summed-squared, e.g. RSS = sqrt(a^2 + b+2 + c^2 + ...) where a, b, c... are individual error sources.

And yeah, this is getting a bit off just jitter... - Don

 

[Phelonious Ponk]

I have what is perhaps an uninformed question:

If jitter distortion is at -80dB (for example), is it possible that the combined distortions of the reproduction chain can become cumulative enough for total system distortion to reach typical audibility? Is it possible that jitter is one of the items to be summed as distortion approaches the threshold of audibility? Would it not make sense to reduce those distortions that can be effectively reduced to avoid an audible "sum"? Which distortions contribute most heavily to audible disruption?

Maybe best for another thread or debate, but this "summation" seems like an important concept to me.

Lee

I suppose that it's possible, if a number of distortions were occurring at exactly the same time and at the same frequencies, that they could "sum," but this concept is more pertinent, I suspect, to the noise floor than it is to non-linear distortion. And in the context of this conversation, Amir has identified specific jitter artifacts he can hear at high volumes. If they are "summed," how could we possibly identify them as jitter?

P

 

[amirm]

If they are "summed," how could we possibly identify them as jitter?

P

Assume you have a digital transport feeding a DAC. And assume that you make certain changes to that source and the sound changes. What would you conclude is changing?

 

[Phelonious Ponk]

Assume you have a digital transport feeding a DAC. And assume that you make certain changes to that source and the sound changes. What would you conclude is changing?

I would assume it was jitter. But you know what they say about "assume."

P

 

[amirm]

I would assume it was jitter. But you know what they say about "assume."

P

I this case, they don't say anything bad .

A digital source conveys two types of information:

1. The PCM samples. If something goes wrong with these, you get glitches and audio drop outs. These failure modes are not of interest to us so we can assume all is well here if the music keeps playing.

2. The timing of said samples. If the sound changes, and #1 is not impacted, then by induction, it is the timing of the samples which is impacted and hence jitter is the sole reason for audibility differences.

If you read my posts though, you often see me mix in the term "other distortion" to talk about downstream conversion issues which indeed wind up being the sum of jitter and DAC non-linearities. There, we can't distinguish between the two without measurements (and even then it may be challenging at times).

 

[Phelonious Ponk]

I this case, they don't say anything bad .

A digital source conveys two types of information:

1. The PCM samples. If something goes wrong with these, you get glitches and audio drop outs. These failure modes are not of interest to us so we can assume all is well here if the music keeps playing.

2. The timing of said samples. If the sound changes, and #1 is not impacted, then by induction, it is the timing of the samples which is impacted and hence jitter is the sole reason for audibility differences.

If you read my posts though, you often see me mix in the term "other distortion" to talk about downstream conversion issues which indeed wind up being the sum of jitter and DAC non-linearities. There, we can't distinguish between the two without measurements (and even then it may be challenging at times).

In your example, it could also be electrical noise from the transport (unless it is connected via optical or well-isolated) getting into the analog signal. If there is anything I've learned in my explorations of audiophilia, it's that even when it is simple, it's not simple.

P

 

[amirm]

I agree, and I have learned to say "at normal levels" when discussing jitter and dither and the benefit of 24-bit recording. I'm pretty sure I never said "jitter is never audible in any circumstance," but if I did I meant any normal circumstance.

I just did a quick web search on your name and jitter and the second Google link said this: http://www.skeptic.com/eskeptic/10-01-06/

"Digital recorders have a unique type of timing deviation called jitter, but with all modern equipment, jitter is so much softer than the music that you’ll never hear it. "

Now, don't go searching for me and what I have said in the past or surely you will find skeletons like this and then some .

I sleep in peace if I have gotten you to move from "never" to "normal circumstances." If everyone did away with extreme views of this field on either side, we would all be happier for it!

There is no question that we are talking small distortions here. So I am happy to concede as I have repeatedly that this is hard stuff to hear although let's not confuse having to to turn up the volume to train yourself to hear it than actual listening tests post that. I also object to saying that we should all listen to music that is normalized and follows "normal circumstances." I should be free to turn up the music when it is recorded at softer levels. If I hear jitter there, then it is my business to search for ways to eliminated it.

 

[amirm]

In your example, it could also be electrical noise from the transport (unless it is connected via optical or well-isolated) getting into the analog signal. If there is anything I've learned in my explorations of audiophilia, it's that even when it is simple, it's not simple.

P

I grant you that if I didn't know what jitter sounded like as opposed to electrical noise unless you are saying that electrical noise caused jitter in the DAC in which case, the distinction is moot .

Still, good point.

 

[Phelonious Ponk]

I grant you that if I didn't know what jitter sounded like as opposed to electrical noise unless you are saying that electrical noise caused jitter in the DAC in which case, the distinction is moot .

Still, good point.

No, I'm talking about electrical noise that is distinct from jitter. I've had to answer for why I don't hear differences between digital audio players or hard drives on isolation plinths sending identical data to identical systems. Those who believe they do hear such things are quick to accuse my hearing or the quality of my equipment. I know better, so I have learned to briefly summarize that digital players move data from one place to another within a system and if the error corrected data is the same at the DAC as it is on the original file, there are only two opportunities for it to sound different - electrical noise and timing errors.

Of course timing errors are not audible under "normal" listening conditions , and my systems is very well-isolated, so I guess there's nothing left to blame but my hearing. By the way, of course you have the right to crank quietly recorded material up, but I still don't think that's going to make the jitter in competent digital devices audible. I believe (cannot prove, but believe) that to hear jitter in such devices you will need to crank the quiets passages up to a point where the crescendos would threaten your hearing, your voice coils, or both. And they would probably have to be the voice coils in your headphones unless your listening room is an anechoic chamber and your speakers are both active and remarkable. Which means, to summarize my personal position on this subject, that while I will concede that unusually bad jitter might be audible under extreme circumstances, it is, in practice, a non-issue.

P

 

[Ron Party]

Of course timing errors are not audible under "normal" listening conditions , and my systems is very well-isolated, so I guess there's nothing left to blame but my hearing. By the way, of course you have the right to crank quietly recorded material up, but I still don't think that's going to make the jitter in competent digital devices audible. I believe (cannot prove, but believe) that to hear jitter in such devices you will need to crank the quiets passages up to a point where the crescendos would threaten your hearing, your voice coils, or both. And they would probably have to be the voice coils in your headphones unless your listening room is an anechoic chamber and your speakers are both active and remarkable. Which means, to summarize my personal position on this subject, that while I will concede that unusually bad jitter might be audible under extreme circumstances, it is, in practice, a non-issue.

P

This directly ties in to my Post #63 of which (hint, hint, errr, hmm, pssst: Amir?) I'd like to see more discussion, namely: (1) the detection of jitter while listening through loudspeakers (2) at *normal* listening volumes and (3) as you stated, PP, the detection of jitter while listening to *competent* devices, which includes plenty of gear with measured jitter specs of less than 250 picoseconds.

 

[terryj]

I have nothing to add technically, but I would like to laud and validate the nature of the thread.

This MUST be the most polite, rational and coherent debate I have ever seen on a forum that touches on anything remotely contentious. And, as a result of this, it must also be the most educational.

Thankyou, looking forward to many more like this.

I too would like to see the emphasis now switch to 'less contrived' circumstances, real world, real rooms, real systems that we use.

 

[Phelonious Ponk]

This directly ties in to my Post #63 of which (hint, hint, errr, hmm, pssst: Amir?) I'd like to see more discussion, namely: (1) the detection of jitter while listening through loudspeakers (2) at *normal* listening volumes and (3) as you stated, PP, the detection of jitter while listening to *competent* devices, which includes plenty of gear with measured jitter specs of less than 250 picoseconds.

Yes, but be careful. We could very quickly get to a place where we'll be concluding that properly implemented DACs, integrated into very common consumer electronics should be transparent and, therefore, audibly identical to even the most esoteric products. And that's just a short step away from concluding that many are transparent, and we may find a lot of people very resistant to that notion.

P

 

[amirm]

Speakers present challenges and perhaps some opportunities for hearing digital distortion:

- One speaker can mask distortion in the other. Let's say there is a 2 Khz jitter sideband and nothing close to it in that channel but the other, has a strong 3 Khz signal in it. The latter will mask the former. Headphones eliminate this chance.

- There is less in an audio chain with headphones directly driven by a DAC or source with its internal DAC. There is no power amp. There is no crossover. There are no speaker wires.

+ Speakers do have the advantage of providing imaging. If you have crosstalk issues between channels (e.g. in low cost stereo DACs), you may be able to hear that better than with headphones.

On jitter being less than 250ps, keep in mind that the math makes some simplifications to get there. Namely, it assumes jitter is a simple sine wave. If a DSP pluses the system once every millisecond causing the jitter to have a spectrum similar to that, that math doesn't apply. Also keep in mind that the number is driven based on bandwidth of 20Khz and sample resolution of 16 bits. If you listen to high res music, then you need to follow much strictly timing spec than this number.

As to equipment today being competently designed, and jitter not being an issue, you can only say this if you have seen their measurements and not manufacturer's claims of jitter numbers. But sure, if you can interpret a chart like this, you are golden:

 

[DonH50]

Hmmm...

1. I understand the spatial imaging aspect, but am not sure why crosstalk would be easier to hear with speakers? I have always used the cans for that (actually, unless it's pretty bad, I resort to a 'scope or other test equipment to measure crosstalk). However, I usually do it "old school" by putting a full-scale signal in one side and terminating the other input, then measuring (or listening to) the "off" side to see (hear) what gets through.

2. In the "normal" math I have seen, jitter is treated as white noise, or perhaps broken into various types of colored noise depending on distance from carrier on a phase noise plot in an RF sampling system (have not seen that in audio, or very rarely). The signal, however, is usually assumed to be a single-tone sinusoid at full-scale to generate the most often-used jitter formula relating SNR degradation and jitter, e.g. this one from an ADI app note:

SNR = 20*lo(2*pi*fin*tj) where
pi = 3.141592654...
fin = input signal frequency
tj = amplitude of time jitter

For random jitter, the SNR is related only to the input signal frequency and jitter amplitude (in time). System bandwidth does not enter into it, and resolution and signal amplitude falls out of the equation. I agree that the math is invalid if these assumptions are not valid, e.g. the math does not work properly if:

a. The jitter is not (or can not be statistically treated as) white noise;
b. The input is not a pure sinusoid;
c. Even for a pure sinusoid, the equation is not valid over the entire period of the signal due to changing slew rate at different points on the sine wave, but the average is good enough;
d. There are no other noise or error sources in the system;
e. Sampling is perfect, with only time jitter added.

Plus other assumptions I've forgotten and others can point out, but I need to get to bed. IMO - Don

 

[rsbeck]

You need to trust me that I know the difference

Since you know the difference, that's all the more reason to use the correct terminology when reporting.

So let's move on and talk about other points. I think we have beat this one to death.

Agree. I see this excellent discussion has moved on and IMO, that's great to see.