Originally Posted by Swampfox
Great. It is a rare occasion that someone has read these reports, and even more rare that they follow it with a post that shows they have
No, the Dolby paper indicates the level of audible jitter under 'ideal' conditions.
Where? What makes their conditions ideal? An ideal condition would include worst case jitter. Sinusoidal, data-independent, jitter is not worst case jitter.
True, but misleading. Uncorrelated jitter is less audible, not more audible.
I am glad you agree they used uncorrelated jitter and hence it is less audible
. Your response however doesn't address the point that I made. In the Dolby test they are injecting only single jitter frequency. There is no evidence to back that this is the profile of jitter in real devices. Here is a random measurement from Paul Miller:
You see all of those spurs (spikes) around the main signal frequency? For a single tone in the middle, we got all of those sidebands. Not one, but all of them. Now imagine your music being made up of thousands of those main bands, with each generating all of those distortions.
Still think the Dolby test created the ideal conditions for jitter?
Actually, the Dolby experiment is statistically biased for finding audible jitter because they excluded all individuals who couldn't be trained to hear jitter induced distortion.
Not really. They may have just thrown one guy off the train. It is not like they surveyed 1000 people and out of them, they picked these 8. And the ones they picked, some are pretty tone deaf relative to the best guy. Mind you, I think it is great that they eliminate some people. Just that it doesn't rise up to the level you are mentioning. To wit, it is still entirely possible that the authors of TAS article could beat even their testers.
Totally, and absolutely unsupportable conclusion based on the design of the TAS trial.
I was specific on where the two agree and I can quote the Dolby and TAS articles if they like:
1. They both believe, as I do by the way, that you can be trained to hear these artifacts better (AJ: take note).
2. That music selection is very important as some are more revealing than others.
I don't see how you can factually disagree with either when they are written in black and white. I am not asking you to the entire report if that is what caused you to go non-linear with that comment
. Only that the TAS authors have said something that agrees with this report. Which also happens to agree with what we know about subjective audio testing in general (that content matters and expert listeners always do much better when it comes to discrimination).
Pure test tones trials totally bias jitter distortion detection toward 'hearing it' because auditory masking doesn't come into play.
There is no evidence that pure tone are best. They found it to be the case in their testing but that is because they did such a limited test. It is entirely possible that other content could be more revealing. The data is simply insufficient to draw your conclusion.
Also, keep in mind again that they used data independent jitter tests. And used continuous tone jitter.
Music is not pure test tones, and auditory masking clearly and unequivocally comes into play when listening to meaningful content.
It is true that a lot of masking is going on in music. But it is also true that you can't control what is music and masking can go away from under your feet all of a sudden, allowing the distortion to be heard. We use simultaneous masking as the cornerstone of audio compression. It works well 'till you get a sharp transient and the distortion spreads in the audio frame well past the skirts of critical bands and it becomes audible as pre-echo. Since masking effect comes and goes, distortion becomes data dependent and hence more annoying than it would be otherwise.
Well designed systems have constant quality. If your system only works when masking is there and not otherwise, then you have variable quality. And with it, lack of ability to put a number on its fidelity. We use lossless audio compression for this reason rather than lossy.
We need to also keep in mind that the Dolby test used headphones. As rightly noted in the Hawksford, when using stereo speakers masking argument becomes weaker. This is due to "unmasking" that occurs when you use more than one speaker. What worked in mono, all of a sudden doesn't work in stereo or multichannel. Both of the test systems in TAS report utilized speakers. I don't want to overemphasize this point as headphones can be wonderfully revealing too and we use them quite a bit but we should keep this in mind as another difference between the tests.
Well, I have read it. It is a well designed study to find the audibility of jitter distortion.
It is. It provides a data point. It is interesting though that you find it well designed even though it doesn't even appear to be a blind test. It has rather small number of testers (8?). Plus the other things I mentioned. I personally agree with you as I just said. It is a very good test. What I find unfair though is that you all look past the faults in its report because you like its conclusion, but can't in the TAS report. Terry said it long time back. We don't seem to be consistent.
The TAS study does not isolate jitter, measure jitter, or even test jitter in any fashion what so ever. Any, and all 'conclusions' about jitter can simply not be made based on the design of the study, and frankly the findings actually exclude jitter as a possible mechanism for their findings.
How do they exclude it?
I'm not trying to turn this into a 'prove jitter is audible' thread, but rather trying to explain why the TAS article is self contradictory, and the concept that jitter explains the finding is at best conjecture, and at worst totally inconsistent with the findings of their study.
Nobody said jitter explain their findings. I have already said their finding is fallacious. So don't know why you are drawing that conclusion by arguing with me about it.
My point is simply this: you cannot say lossless files play out the same because the bits are the same. Both jitter and electrical interference cause enough variation such that this theory cannot be used to veto their results. If we agree on this, we are done. If we don't, then you have to prove that the Dolby test duplicates their condition and that you have measurement data form TAS systems that were below their findings. I think that is impossible to do.