Low-Jitter USB: Dan Lavry, Michael Goodman, Adaptive, Asynchronous

[regal]

Quote:

Quote:

The level of jitter in tmost modern DACs ia 100 times below the threshold of audibility. It isn't anything worth worrying about. Better to focus on things you can actually hear.

 
Hello Steve, long time no see, been in prison ?
 
Steve, you are being a little bit disengenuous here. You know full well that there are different variations of jitter and only signal-correlated jitter has been empirically tested to be detected at levels of about 20ns (Benjamin and Gannon) in music which is 100x the level of that found in a competent DAC (200ps) or 10ns as pure tones at 17K, random jitter however has been empirically tested to be undetectable at up to 250ns (Ashihara et al) which is 1250x worse than a competent DAC, please get your facts right

 

The tests are silly at best.  First there is no means to calibrate "injected" jitter,  where is the ASTM?  Is random jitter ever encountered from a transport in a level worth discussing?,  its the deterministic jitter that must be studied deeper.    So we have this one stupid deterministic jitter "study",  but its a fact that a pure 17khz sine wave doesn't exist in nature.  The only time we ever here one is a hearing test,  pretty foreign to the ears and a pretty poor test my freind,  hell most people can barely hear 17khz,  there is no music up there.   No one can unequivically say that a multimillion dollar transport industry and hundreds of thousands of customers are experiencing mass auditory jitter hallucinations.  That stance borderlines on the schizophrenic ....

 

[DaveBSC]

Quote:

 hell most people can barely hear 17khz,  there is no music up there.   No one can unequivically say that a multimillion dollar transport industry and hundreds of thousands of customers are experiencing mass auditory jitter hallucinations.  That stance borderlines on the schizophrenic ....

While its true that most adults have lost most of their hearing above 16-17kHz, it's not true that "there is no music up there". Many instruments indirectly produce sounds at extremely high frequencies, and any analyzer can show you that nearly all commercial recordings contain at least some information up to 22kHz.

 

[bigshot]

Generally speaking, the more money you throw at a technology, the higher the performance you can achieve.

I've never seen a direct correlation between cost and quality in electronics. But I listen with my ears, not spec sheets.

 

[bigshot]

While its true that most adults have lost most of their hearing above 16-17kHz, it's not true that "there is no music up there". Many instruments indirectly produce sounds at extremely high frequencies, and any analyzer can show you that nearly all commercial recordings contain at least some information up to 22kHz.

Cymbal crashes mostly. Not much else up there, even first level harmonics.

An interesting test to find out just how much music there is up there is to filter out all sound above 10 or 12 kHz. You'd be surprised how little of a difference it makes to sound quality. The audible range in practice has a handle of about an octave. Most music falls between 40 Hz and 10 kHz.

 

[DaveBSC]

Quote:

Cymbal crashes mostly. Not much else up there, even first level harmonics.An interesting test to find out just how much music there is up there is to filter out all sound above 10 or 12 kHz. You'd be surprised how little of a difference it makes to sound quality. The audible range in practice has a handle of about an octave. Most music falls between 40 Hz and 10 kHz.

UHF response seems in my experience to have a direct relationship to soundstaging, air, and just the feeling of a real musical space. I don't know why, but it does. Use a high pass filter for everything above 10kHz and I can guarantee that the result will be a sound that is utterly devoid of life, and completely two dimensional.

 

[regal]

Quote:

Quote:

 hell most people can barely hear 17khz,  there is no music up there.   No one can unequivically say that a multimillion dollar transport industry and hundreds of thousands of customers are experiencing mass auditory jitter hallucinations.  That stance borderlines on the schizophrenic ....

While its true that most adults have lost most of their hearing above 16-17kHz, it's not true that "there is no music up there". Many instruments indirectly produce sounds at extremely high frequencies, and any analyzer can show you that nearly all commercial recordings contain at least some information up to 22kHz.

I agree with Bigshot listen to your favorite album with everything below 17khz cutout,  then get back to us whether you still call whats left music.   Point being it is rediculous to base a jitterr audibility test at 17khz unless your subjects are dogs.

 

[nick_charles]

Quote:

 

  Point being it is rediculous to base a jitterr audibility test at 17khz unless your subjects are dogs.

That was a worst case scenario, at lower frequencies audibility was far harder !
 
The point remains though, that there is really no good reliable evidence for jitter audibility at the levels found in competent kit. We could take transport systems with known levels and spectra of jitter and do controlled tests but nobody does this, surely if jitter was such a problem it would be a slam dunk to take a cheap transport with high jitter and a transport with good jitter, connect to a DAC that does not remove jitter in the signal and both plot the FR and empirically test the audibility of the differences, it just is not done, why ?
 
I have plotted the FR for differentially jittered samples, a lot of added jitter really makes a very very small difference even 100ns of added jitter barely registers. For instance the average effect of 100ns of added jitter is a deviation of 2/100ths of a decibel over 20 - 20K and the worst deviation at any frequency is 9/100ths of a decibel at 19961hz. At 10K or below it never gets past 0.023db at 15K or below it never gets past 0.053db.
 
For 10ns jitter the deviation never exceeds 0.0112db at any frequency

 

[bigshot]

UHF response seems in my experience to have a direct relationship to soundstaging, air, and just the feeling of a real musical space. I don't know why, but it does. Use a high pass filter for everything above 10kHz and I can guarantee that the result will be a sound that is utterly devoid of life, and completely two dimensional.

Phase is most important for soundstage. Frequencies at the top are sporadic at best, and tend to be bunched up in cymbal hits. They just occur in spikes, while soundstage is created by continuous sounds in the middle. You might be mistaking super high frequencies for the treble frequencies, which are much lower in the spectrum. Those are important cues for the directionality of sound.

It's fine to have frequencies up at the edge of hearing, but the further it goes from the core frequencies of music, the less important it is. By the time you get up to third level harmonics, the volume is getting so low you can't really hear it. Inaudible frequencies are inaudible. There's no indication that they contribute anything to the reproduction of music.

 

[bigshot]

For 10ns jitter the deviation never exceeds 0.0112db at any frequency

Which is harder for them to hear... Jitter... Or the facts?

 

[Bmac]

Quote:

Originally Posted by CHansen /img/forum/go_quote.gif
 
I strongly agree with the first and third paragraphs that Mr. Goodman wrote.
 
However the second paragraph is simply incorrect (somewhat ironic given the content of the third paragraph). The truth is that, if all else is equal, it is absolutely impossible to build a variable-frequency clock (as used in an adaptive mode USB DAC) with as low a jitter level as a fixed-frequency clock (as used in an asynchronous mode USB DAC).
 
Generally speaking, the more money you throw at a technology, the higher the performance you can achieve. And it is possible to get high levels of performance from variable-frequency clocks (generally using a Phase-Locked-Loop or PLL). But not as high as you can get with fixed frequency clocks. That is a fact. Mr. Goodman is peddling deception.
 
Sure, it is possible to make a really good variable-frequency clock that has lower jitter than a fixed-frequency clock. But only by spending a great deal more money or if the fixed-frequency clock is an extremely poor design. Engineers have worked for decades to reduce the jitter of variable-frequency clocks. High-performance ones tend to be complex and expensive. In the meantime, fixed-frequency clocks are simple to design and build. Even the simplest, most inexpensive fixed frequency clock will have lower jitter than an excellent variable-frequency clock as long as you feed it with a nice quiet power supply.
 
As Mr. Goodman pointed out in the first paragraph, the incoming USB signal has high levels of jitter. A typical adaptive clocking scheme is to use a PLL, which will filter out some of the jitter. More sophisticated (and complex and expensive) systems will use two PLLs to achieve higher levels of jitter filtering. But with asynchronous, the incoming jitter on the USB line is completely immaterial. The only thing that matters is the quality of the fixed-frequency clock, which as noted previously is practically a trivial thing to design.
 
I would be happy to compare phase-noise plots of our fixed-frequency clock with Mr. Goodman's variable-frequency clock any time. (A phase noise plot is the most sensitive way to measure the jitter of a high-frequency clock.)

I would take exception to some of these statements. While in theory asynchronous USB is better than synchronous USB, there are synchronous solutions out there that are so good (arguably as good or better) that it really doesn't matter anymore.
 
The Centrance DACport costs only $400 and I believe had lower jitter than your own $2,500 QB-9 or Gordon Rankin's $3,500 Cosecant.V. 3 DAC's as measured by Stereophile. The DACport only had an estimated 91ps of peak-peak jitter. You can't really do much better than that regardless of price. That DAC is only $400 so the theory of having to spend a great deal more money to achieve good performance with a synchronous solution just doesn't hold water. There are other synchronous DAC's as well that also have very low jitter. The original Benchmark DAC-1 USB for example measured 128ps jitter by Stereophile; again very low.
 
Even if you could get lower than those, can anyone tell the difference between 50ps of jitter and 121ps of jitter? No way. It's all just academic at that point.
 
I think that's the problem that at least some people have with the way the asynch camp presents itself. The claim to superior jitter performance makes for great marketing. In the real world, if there is an audible difference between the two methods it can't be attributed to jitter, which has been the only claim I have seen from the asynch camp with regard to subjective sound superiority. Still there are lots of people out there that still believe the hype that the only way to get good USB audio is through asynch, and that just isn't the case.

 

[CHansen]

Quote:

I would take exception to some of these statements. While in theory asynchronous USB is better than synchronous USB, there are synchronous solutions out there that are so good (arguably as good or better) that it really doesn't matter anymore.  
The Centrance DACport costs only $400 and I believe had lower jitter than your own $2,500 QB-9 or Gordon Rankin's $3,500 Cosecant.V. 3 DAC's as measured by Stereophile. The DACport only had an estimated 91ps of peak-peak jitter. You can't really do much better than that regardless of price. That DAC is only $400 so the theory of having to spend a great deal more money to achieve good performance with a synchronous solution just doesn't hold water. There are other synchronous DAC's as well that also have very low jitter. The original Benchmark DAC-1 USB for example measured 128ps jitter by Stereophile; again very low.
 
Even if you could get lower than those, can anyone tell the difference between 50ps of jitter and 121ps of jitter? No way. It's all just academic at that point.
 
I think that's the problem that at least some people have with the way the asynch camp presents itself. The claim to superior jitter performance makes for great marketing. In the real world, if there is an audible difference between the two methods it can't be attributed to jitter, which has been the only claim I have seen from the asynch camp with regard to subjective sound superiority. Still there are lots of people out there that still believe the hype that the only way to get good USB audio is through asynch, and that just isn't the case.

Three things:
 
1) The measurement system that Stereophile uses was designed specifically to measure the jitter added by the S/PDIF interface. When used in other applications it can only show gross defects. It has a measurement limit of 120 psec for 16 bit data (which is how it is normally used). Furthermore there have been several changes to the hardware over the years. Quoting numbers from Stereophile measurements as a way to "prove" the audible superiority of a product is just plain wrong.
 
2) If "good enough" is good enough for you, that's great. You can save yourself a lot of money by purchasing low-cost products that are "good enough". We are interested in the highest quality products and making things that are better than anyone has ever made before. They're not for everybody, and they may not be for you.
 
3) You haven't even tried (perhaps wisely) to dispute my point that (if all else is equal) that an asynchronous DAC will have lower jitter than an adaptive DAC, simply because a fixed oscillator clock will have lower jitter than a variable oscillator (eg, PLL). So if you are happy with higher jitter, more power to you. I am not. Furthermore, Mr. Goodman's post contained falsehoods that demanded correction.

 

[USAudio]

Thanks for posting Charles, glad to have you here on Head-Fi!

 

[estreeter]

End of the day, our ears are the only true arbiters : I'm all for objective, instrument-based testing but it seems to raise more questions than it 'answers' : we need controlled, independently verified DBT to establish anything resembling 'facts' in relation to the impact of jitter on sound quality, and the veracity of different manufacturers claims in relation to jitter reduction. Regardless of the size and composition of the group assembled for the tests, I have no doubt that there would be folk who would refuse to accept the findings - given that we don't have universal acceptance of Darwin's theory of evolution, science means
 
I'm more interested in the fact that the people mentioned at the beginning of this thread have spent a lot of late nights grappling with the problem - if you compare advances in computer audio with advances in *video* technology over the last decade, the question is simple : why cant I order a desktop computer with the audio equivalent of the NVIDIA GTX 460 on board ? Consider the frame-rates of top-shelf cards two years ago, past issues with drivers etc and compare that experience to November, 2010 - I just dont see computer audio moving at the same exponential pace. I know we dont measure progress by the number of transistors on a card, but it would be interesting to see some 'trickle down' from the high end - you dont need identical specs to the guy who designed the levels for the latest 'Call of Duty' to run it, and you shouldnt need a Berkeley Alpha DAC to get good results from computer-based audio. Apples and oranges ? Perhaps.

 

[DaveBSC]

Quote:

if you compare advances in computer audio with advances in *video* technology over the last decade, the question is simple : why cant I order a desktop computer with the audio equivalent of the NVIDIA GTX 460 on board ? Consider the frame-rates of top-shelf cards two years ago, past issues with drivers etc and compare that experience to November, 2010 - I just dont see computer audio moving at the same exponential pace. I know we dont measure progress by the number of transistors on a card, but it would be interesting to see some 'trickle down' from the high end - you dont need identical specs to the guy who designed the levels for the latest 'Call of Duty' to run it, and you shouldnt need a Berkeley Alpha DAC to get good results from computer-based audio. Apples and oranges ? Perhaps.

This is a rather strange post. AMD and nVidia processors have different architecture and do things differently than x86 processors like an Intel Core i7, but ultimately they do the same basic thing, and they advance in a similar fashion. They process code, and with every new generation they run faster (or more intelligently) and they get smaller.
 
Audio is completely different than that. How long has vinyl been around? How about tube amps and horn speakers? Companies like Clearaudio, Basis, VTL, Audio Research, and Acapella Audio Arts are slowly and continually pushing these mediums, which are as old as the computer that cracked the Enigma code, into the state of the art. Compared to vinyl and vacuum tubes, computer audio has made huge advancements in a fairly short amount of time. Remember the Sound Blaster 16? How far has a Mykerinos card come from that? CD playback has advanced at nowhere near that rate in the same amount of time.
 
You don't need an Alpha DAC to get good results from computer based audio. Even a modest M-Audio or RME sound card is likely to sound better than most CD players at similar prices. What's new and exciting is now you can get SOTA audio from a computer, via the Mykerinos, or a Pace-car, Wavelink, QB-9, DAC8, DAC202, etc.

 

[bigshot]

If "good enough" is good enough for you, that's great. You can save yourself a lot of money by purchasing low-cost products that are "good enough". We are interested in the highest quality products and making things that are better than anyone has ever made before. They're not for everybody, and they may not be for you.

It doesn't have anything to do with sound anymore, does it?