Jitter vs no-Jitter

[徐中銳]

Recalling Digital Audio: The Possible and Impossible :

How Much Jitter is Too Much?

One way we measure the impact of jitter is by making the simplifying assumption of it being a sine wave (it is not many times but let’s go with it). We can then compute how much it needs to be to generate distortion equal to the voltage represented by one bit of our audio sample as described earlier. For CD audio, it would be the amount of jitter that makes 16 bit audio sample equivalent to 15 bits. The idea here is that if we are trying to play 16 bit audio samples we would ideally want our reproduction system to have sufficient transparency to rise up to at least that level of transparency.

 

Jitter's effect on music is that it modulates all the tones in it. Using mathematics which in the interest of not boring you I won’t go into, we can model sinusoildal jitter as a signal that generates two distortion products, one of which is the sum of the jitter and our source frequency and the other, the difference between the two. Using this model, and the fact that CD music has a response of roughly 20 KHz, we can compute how much jitter it takes to overwhelm a single bit of our 16-bit audio sample. Performing the math, we arrive at the unbelievable fact that jitter amplitude cannot be more than 0.5 nanoseconds! You read that right. If timing variation of the DAC is more than half a billionth of a second you generate enough distortion to swamp one bit of your audio sample. It is not opinion that says that. It is pure mathematics. And this is for the simplest case of jitter, not the more complex but typical scenarios where there are many jitter frequencies and spectrums acting on your your music.

 

To visualize this obscure explanation in practice, here is a measurement of jitter acting on a single tone as created by the late Julian Dunn which was published in the digital audio measurement handbook of Audio Precision Audio Analyzer:

The 10 KHz tone is our source frequency. The distortion sidebands (smaller peaks on each side) are purely the result of changing the timing clock of the DAC by a sine wave at a frequency of 3 KHz with an amplitude of 5 nanoseconds. This is for a system with 20 bits of resolution and hence, has a noise floor of -120 dB. Jitter distortion of -80db is hugely above that and has reduced the system performance to well under what CD can do at 16 bits (96 dB). So even though 7.6 billionths of a second sounds like a very small value, its impact in distorting our audio signal is quite significant.

 

Jitter Sources

The next question becomes what causes jitter. The answer is...

[jabbr]

Those are cortical evoked potentials, not brainstem evoked potentials. You've confused two separate tests.

 

Evoked potential/shmoked potential -- the study I was recalling must be several decades old -- makes sense it would be cortical:)

[sandyk]

Jabbr

Did you receive my PM ?

 

Alex

[bibo01]

I think it means that skilled listeners with enough exposure to live musical instruments to truly know what they sound like can hear audible differences like those we're discussing, if they exist. But there's a lot of psychoacoustics research showing that the less discriminating listener is often "fooled" by extraneous factors. For example, there are many studies showing that people routinely choose as "better" otherwise identical reproduction (equipment, source etc) that is simply 1 db louder than the comparator. Adding a bit of white noise to a source results in listeners' reporting "better high frequency response", too.

 

I'm sure some prefer the sound of the jittery DAC because it does make the flute sound fuller and smoother, even though it's fuller and smoother than the instrument sounds in person. A real flute being played next to you can even sound a bit harsh and breathy compared to the recorded flutes that serve as reference for most. So if you're used to a softer flute sound on your recordings, the sound of a real one might well be perceived as shrill. Similarly, the plucking of a harpsichord string causes a "spectral splash" of high order harmonics that could be blurred (if my interpretation is correct) by jitter, making the recording of the instrument sound less "tinkly" for lack of a better descriptive term.

 

I'd love to do this test with pizzicato strings in the scherzo (3rd movement) of Tchaikovsky's 4th. If my concept is correct, there should be a clearly audible difference as all those strings are plucked in unison - jitter should blur the individual instruments and make each note sound like it was plucked on a few huge violins plus a few huge violas plus a few huge cellos instead of a string section.

 

I think you are correct here with our perception of flute and in this case jittery flute. In my first listening I perceived that A1 had a lean sound and I assumed it was more correct. In my second listening, however, I perceived A2 fuller and smoother - as you said - and I assumed A2 was more natural but in reality I was expressing a preference rather than a perception.

[bit01]

Wow! if DAC A2 = track A2 = + jitter, then that little more jitter helps the music IME

[bibo01]

The difference between the two tracks is minimal.

This is a NTD between the two DACs. (NTD is a special null test method; some explanation here)

 

[bibo01]

I also want to follow bluesman's suggestion of another round of poll with other two tracks, to show if we understood the character of the jittered dac and we are able to recognize it.

...so there is going to be a "Part II"

[sandyk]

The main thing here is that 87.5% of the recipients were able to hear clear differences, despite the level differences being very small. Unfamiliarity with the real sound of a flute certainly would not have helped many .

[bibo01]

The main thing here is that 87.5% of the recipients were able to hear clear differences, despite the level differences being very small. Unfamiliarity with the real sound of a flute certainly would not have helped many .

The differences were laid out originally. Everybody knew before hand that the tracks were different.

If the test was to detect differences, I would have prepared it differently.

Instead the assumption was: within these differences can you detect the sound with no-jitter?

It was a learning process for each one of us to understand how we express a preference.

[mmerrill99]

I think this shows a number of things:

- using a variety of music is important for evaluation of audio, especially sounds & instruments that we are familiar with

- long term listening is likely to be more useful for teasing out differences & preferences

- NTD test is not a very revealing measurement of this 200pS of jitter

- I might be wrong but what I think is responsible for the audible difference between the DACs is not the jitter spikes seen in the graph on page 1 (the highest in the audible range down @ -105dB or so) but rather the close-in phase noise or jitter - shown as a widening at the base of the 24KHz tone in this graph. This close-in phase noise causes a smearing of the fundamental & is probably audible in music as less sharpness in the fundamentals?

[bibo01]

I think this shows a number of things:

- using a variety of music is important for evaluation of audio, especially sounds & instruments that we are familiar with

- long term listening is likely to be more useful for teasing out differences & preferences

- NTD test is not a very revealing measurement of this 200pS of jitter

- I might be wrong but what I think is responsible for the audible difference between the DACs is not the jitter spikes seen in the graph on page 1 (the highest in the audible range down @ -105dB or so) but rather the close-in phase noise or jitter - shown as a widening at the base of the 24KHz tone in this graph. This close-in phase noise causes a smearing of the fundamental & is probably audible in music as less sharpness in the fundamentals?

NTD does not reveal jitter. It shows the analog output of your DAC with, in this case, music.

It shows that the differences between the two tracks coming out of a dac are not at all big as much a jitter test shows.

[mmerrill99]

NTD does not reveal jitter. It shows the analog output of your DAC with, in this case, music.

It shows that the differences between the two tracks coming out of a dac are not at all big as much a jitter test shows.

Sure, I know it's not a measurement that is directed towards revealing jitter - it's a difference measurement. That's why I would liked to have seen this listening test as a test of difference rather than preference. With such a predictably low sample size there is nothing the test results tell us, apart from those who described what they heard.

 

If you did an ABX difference test we may have had something to discuss - like the fact that a difference was/wasn't identified & the NTD measurements

 

If I published 2 tracks which actually were exactly the same & said that there were differences & gathered poll results, I would also gather no information of value.

 

Sorry, bibo01 but I always considered the test to be of little use

[bibo01]

Sure, I know it's not a measurement that is directed towards revealing jitter - it's a difference measurement. That's why I would liked to have seen this listening test as a test of difference rather than preference. With such a predictably low sample size there is nothing the test results tell us, apart from those who described what they heard.

 

If you did an ABX difference test we may have had something to discuss - like the fact that a difference was/wasn't identified & the NTD measurements

 

If I published 2 tracks which actually were exactly the same & said that there were differences & gathered poll results, I would also gather no information of value.

 

Sorry, bibo01 but I always considered the test to be of little use

 

I respect your view, except on the value of this test.

We have already done an ABX on these files and for most people it was difficult to detect difference.

[mmerrill99]

I respect your view, except on the value of this test.

We have already done an ABX on these files and for most people it was difficult to detect difference.

OK, I'm not surprised that ABX of the tracks was null, not because there's no audible difference but because of the nature of auditory perception which home run ABX testing fails to recognise. We see these issues here - only a few people know what to listen for & can retain focus on it for 16 or more trials (needed for statistical significance) without beginning to second guessing themselves.

[sandyk]

OK, I'm not surprised that ABX of the tracks was null, not because there's no audible difference but because of the nature of auditory perception which home run ABX testing fails to recognise. We see these issues here - only a few people know what to listen for & can retain focus on it for 16 or more trials (needed for statistical significance) without beginning to second guessing themselves.

 

+1

Fatigue sets in well before then, and they all start to sound the same.

[esldude]

OK, I'm not surprised that ABX of the tracks was null, not because there's no audible difference but because of the nature of auditory perception which home run ABX testing fails to recognise. We see these issues here - only a few people know what to listen for & can retain focus on it for 16 or more trials (needed for statistical significance) without beginning to second guessing themselves.

 

So people need some more ABX experience. A bit of training. To stop at 4 trials. Then go back and do 4 more etc. etc. until 16 or however many are finished.

 

Nevertheless, I actually agree such testing if of limited value. Too many variables. Too much variability in playback equipment and listening conditions. Hearing acuity (like from age) and other things. Yet I also agree with bibo01 that whatever difference was in his test files was not easily discerned. Easy being relative. One example being a test done here where one file of several (the only one readily picked out at levels beyond chance) was different only in being 1 db louder. While no one noticed it was louder it was picked out at high percentage levels.

[mmerrill99]

So people need some more ABX experience. A bit of training. To stop at 4 trials. Then go back and do 4 more etc. etc. until 16 or however many are finished.

 

Nevertheless, I actually agree such testing if of limited value. Too many variables. Too much variability in playback equipment and listening conditions. Hearing acuity (like from age) and other things.

Yes, that's why ABX tests are only really of some value when professionally run by those very knowledgeable in the field of perceptual testing

Yet I also agree with bibo01 that whatever difference was in his test files was not easily discerned. Easy being relative. One example being a test done here where one file of several (the only one readily picked out at levels beyond chance) was different only in being 1 db louder. While no one noticed it was louder it was picked out at high percentage levels.

Some change which affects ALL parts of the soundstage is a different prospect to differentiating something like jitter where audible differences are more specific & may only be perceptually audible when specific conditions occur in the sound field. It then becomes a matter of how likely people are to notice such a small change that is irregularly perceptible in the sound. Inattentional blindness (or deafness) becomes a big part of the test - a big difference between your cited 1dB test & this test.

[esldude]

Yes, that's why ABX tests are only really of some value when professionally run by those very knowledgeable in the field of perceptual testing Some change which affects ALL parts of the soundstage is a different prospect to differentiating something like jitter where audible differences are more specific & may only be perceptually audible when specific conditions occur in the sound field. It then becomes a matter of how likely people are to notice such a small change that is irregularly perceptible in the sound. Inattentional blindness (or deafness) becomes a big part of the test - a big difference between your cited 1dB test & this test.

 

Yes, the 1 db difference was an example of an easy test. Glad you understood that.

[barrows]

Wow! if DAC A2 = track A2 = + jitter, then that little more jitter helps the music IME

 

A different way of looking at it would be that the fact that the sample track with more jitter sounds better to you, on your system, suggests that your system has a slight problem, and could be better tuned. The track with less jitter is more accurate, if you prefer a smeared sound, then some problem in your system is resulting in that preference.

 

I have experienced this situation quite often in dealing with customers during my time at PS Audio: a customer would make an upgrade to their system (say perhaps adding a really good power conditioner), but then they would not like the sound of their system. In most cases, the resulting increase in resolution revealed a problem. Once we addressed the problem, the customer could then appreciate the increase in resolution.

Many systems have an inherent brightness, which can be revealed as one increases the resolution of the system as a whole. Even something as simple as changing speaker positioning, or a cable choice, can eliminate the inherent brightness and enable the enjoyment and appreciation of the upgrade.

[esldude]

A different way of looking at it would be that the fact that the sample track with more jitter sounds better to you, on your system, suggests that your system has a slight problem, and could be better tuned. The track with less jitter is more accurate, if you prefer a smeared sound, then some problem in your system is resulting in that preference.

 

 

I disagree. You too are making the assumption that more accurate sound is 'naturally' going to be preferred. I think that assumption is an incorrect one. Or at a minimum one that is on very shaky ground.

[bit01]

A different way of looking at it would be that the fact that the sample track with more jitter sounds better to you, on your system, suggests that your system has a slight problem, and could be better tuned. The track with less jitter is more accurate, if you prefer a smeared sound, then some problem in your system is resulting in that preference.

 

I have experienced this situation quite often in dealing with customers during my time at PS Audio: a customer would make an upgrade to their system (say perhaps adding a really good power conditioner), but then they would not like the sound of their system. In most cases, the resulting increase in resolution revealed a problem. Once we addressed the problem, the customer could then appreciate the increase in resolution.

Many systems have an inherent brightness, which can be revealed as one increases the resolution of the system as a whole. Even something as simple as changing speaker positioning, or a cable choice, can eliminate the inherent brightness and enable the enjoyment and appreciation of the upgrade.

 

I know what you are trying to say (and appreciate it), BUT what does it matter? I prefer what I do. I like my system.

[bit01]

I disagree. You too are making the assumption that more accurate sound is 'naturally' going to be preferred. I think that assumption is an incorrect one. Or at a minimum one that is on very shaky ground.

 

 

double post

[bit01]

I disagree. You too are making the assumption that more accurate sound is 'naturally' going to be preferred. I think that assumption is an incorrect one. Or at a minimum one that is on very shaky ground.

 

 

+1

If accuracy is reproducing what was recorded then why do we need EQ mastering moves - in any case accuracy needs to be defined before it can even be talked about in this context - how many parameters are you looking at from ADC to DAC & how do they interact? As a consumer (and not a designer) I just sit, listen and pick what I like best!

[bluesman]

A different way of looking at it would be that the fact that the sample track with more jitter sounds better to you, on your system, suggests that your system has a slight problem, and could be better tuned. The track with less jitter is more accurate, if you prefer a smeared sound, then some problem in your system is resulting in that preference.

Why would you conclude that? The more accurate the system, the more clearly the effect(s) of jitter will be heard (assuming you believe that they're audible, which I do). If the track with high jitter sounds better to you, it's because you like some characteristic of the sound that's being affected by jitter.

[Jud]

Why would you conclude that? The more accurate the system, the more clearly the effect(s) of jitter will be heard (assuming you believe that they're audible, which I do). If the track with high jitter sounds better to you, it's because you like some characteristic of the sound that's being affected by jitter.

 

At least for the time being. I find over time distortions will become bothersome, simply because a distortion from the system is something that is always, always the same about the sound. Thus even if the customer might have liked the sound at the time, I think over the long term eliminating distortion does the listener a service.