Originally Posted by arnyk
So Amir what you're saying here is that you now repudiate all of your former comments about the importance of jitter that has never been reliably detected in a listening test that conforms to international standards or what we know about psychoacoustics?
The research I have provided to you such as Hawksford/Dunn paper, start with perceptual theresholds of hearing as was shown in the Fletcher Munson graph. They then compute mathematically what level of jitter can exist to be below that. Since the threshold of hearing is based on listening tests, they absolutely follow that. Same was done in research into dynamic range of music I have cited here, by Bob Stuart on what the resolution of digital audio should be, and countless other research into audibility of distortion.
So no, we don't throw out good science but rather, we use them in our conclusions even though you think if I don't say "ABX" the data must be invalid.
The nice thing about what they are doing is that they prove at what mathematical level jitter must be inaudible (at least with high confidence). A listing test doesn't do that by itself. When the parameters of the problem at hand creates an infinite set, any jitter listening test will have severe limit relative to applicablity to full universe of content, equipment and listeners. Not so with mathematical proofs relative to ear's sensitivity. This is why it has been so easy to shoot down existing jitter listening tests. Your recall our recent review of Dolby test where it was not even blind yet you had believed it.
Any time we can substitute objective means vs subjective listening tests, we should do that. You can't argue with the math. But we can argue whether some listening test being valid or not.
Importantly, as I have noted repeatedly, the goals for a discussion for what the audio playback standard needs to be, must be a conservative one. There needs to be a "safety margin" as to give us high confidence that we have built what we say we have built. To the extent the mathematical proof does that, it is a good thing in my book.
In yours, you assume it is "too good." That's cool. People don't have the same standards in food, wine, or cars. Some think frozen fish is just fine to make your dinner from. Same with audio. It is perfectly fine for you to lead your audio life which in this case, relies on jitter tests that do not exist, to arrive at it not even be material enough to look up its level. I am an engineer, and I care about true performance of a device, and not just its marketing moniker of "24 bit DAC." So I take totality of the data in mind from measurements to objective analysis of inaudibility to base my audio purchases. I hate getting sand in my eyes when I stick my head into it otherwise
It must not matter because "a microphone and a meter are not the same as two ears and a brain"?
Since you don't read the references I provide to you, it is natural for you to assume that they did not take psychoacoustics into account when they very much have. So no, we don't get to throw out a rigorous analysis by experts that live and breath signal processing, are heavily admired in the industry and provide us with their wisdom across this complex topic.
Personally I have been very consistent in applying human perception in our discussions. If you recall, one of my biggest arguments with you on this forum was you insisting that low frequency jitter is more audible than high. You were trying that line because you wanted to tie it to wow and flutter in analog world.
In that discussion, the roles were completely reversed with me taking the position that very low frequency jitter was totally unimportant and that audibility went down as the frequencies got that low. You went on and on that I was wrong and that it got more audible, not less (again to defend your multi-year campaign on that analogy to tape). I used your own reference in psychoacoustics of low frequency timing modulation to prove that you were wrong on this front and that we should *not* worry about jitter of that spectrum. In a stunning turn of events, you conceded the point when it became clear that you had not read your own reference and of course, were wrong.
My entire point in that discussion was based on understanding of psychoacoustics which in this case, involved masking effect and critical bands. The fact that it went against "jitter is always bad" was not material to me. What is scientifically right, is right. And yes, in that manner, two ears and a brain was precisely why I was right and you were wrong.
The above is why you need to look at the spectrum of jitter and not just its "number." Because its spectrum determines its audibility. So simplified tests that use random or even sine waves, are not necessarily representative of the problem at hand.
All of this requires lots of knowledge about the problem at hand. Just as the discussion of the "quantization myth" I think you are led by poor understanding of digital audio and so quickly go by paucity of tests and run off with negative logic that jitter must not matter. I can't blame you. Discussions of non-linear, data-dependent of distortion is not a familiar concept to people. The notion that something doesn't show up on a static frequency response measurement is just foreign to people. But as I always say, science makes no claim of fairness
. If it is hard, it is hard.
What's funny is that you and the article in question in this thread advocate 16 bits yet in presence of jitter typical in an HDMI connection, you are lucky to be achieving 14 bits. I have always taken the stance of having a minimum performance bar of 16 bits. In the age of "24 bit converters," this should be a very defensive performance target. If an engineer can't deliver 16 out of those 24, then they don't deserve my money. I realize you are hungry to do otherwise. So be it. I explain how the science works. It is your call to ignore or follow it.