Rob Watts
Member of the Trade: Chord Electronics
- Joined
- Apr 1, 2014
- Posts
- 3,144
- Likes
- 12,411
@Rob Watts
While following the PSAudio DirectStream threads regarding their regular DAC firmware updates, i was surprised that several final release candidates were auditioned prior to each release - and the code was mostly functionally identical .. with only changes in signal flow, timing, parallel -vs- serial, compilation switches, etc. Yet, each variant sounded 'different' - and the final release determination was based on a compromise. What this is telling me is that we are so near the bleeding edge of silicon switching noise and ground plane disturbances affecting even a 'perfect' DAC algorithm.
I know you have a handle on the small details ... but some of the claims of M-Scaler taps and ultra-low noise floor affecting perceived sonics seem so incredulous. Is there an end to all this? Its starting to sound silly. Are humans blessed with a quantum computer in our heads that can do math at 80-bits precision in real time? Or, like the PSAudio firmware variants ...are you just shuffling around the electrons and seeing 'real' differences where there are none?
- No offence or disrespect intended, please ... i am just curious of your comments.
Thanks
Dan
I agree with you; talking of small signal errors at -350 dB upsetting depth performance is just plane crazy. Objectively, it makes no sense that the ear/brain can be so sensitive.
But... Listen to real sounds. I am typing this at dawn in the Welsh countryside. My window is ajar, and I can hear sheep 200 feet away - and it sounds 200 feet away. From the farm 2 miles away a dog barks; it clearly sounds 2 miles away. Do this on reproduced audio and things sounds comparatively flat; no-body would claim that a recording of a dog barking 2 miles away actually sounds anything like two miles distant.
Now when I listen to noise shapers, and change the performance of noise shapers, I can easily and reliably hear changes in depth perception. And when I measure the digital domain performance of the noise shaper, there is a correlation between the small signal performance of the noise shaper and depth perception; so a 200 dB noise shaper (this means it can't resolve small signals below -200 dB accurately) has poorer depth than a 220 dB noise shaper - and continuing this process on I ended up with 350 dB performance (and I could still hear improvements from 330 to 350).
Now these numbers really do not make any kind of rational sense. And when I observed this, I was convinced that the most likely explanation was that the noise shaper performance was merely a proxy for something else going on - for example an analogue aberration that was magnifying the effect of the noise shapers, for example folding down from RF frequencies (where we do not get anything like 350 dB performance).
But I soon found that idea was wrong; I use noise shapers in digital only truncators, where you are going from say 56 bits at 768 kHz down to 24 bits. So it's all digital; and again, listening to these noise shapers gave exactly the same result; 350 dB performance had better depth than 300 dB digital only noise shapers.
But these performance numbers still make me uncomfortable; but I have repeated it with different pulse array noise shapers, and truncator noise shapers about half a dozen times with different designs - and I get exactly the same result. And I can hear you saying it is confirmation bias, because the listening tests are not double blind, and I know which noise shaper I am listening too - and I can't rule that out as a possibility. But I do plenty of listening tests where I get zero difference when I am actually expecting a sound quality change as I conduct my listening tests to be as accurate as possible. Because of these multiple tests, I am convinced that these results are real.
But if indeed noise shapers are required to such accuracy, then the ADC noise shaper itself must also need this performance. So the Davina ADC project is an 11th order 104 MHz noise shaper, and is capable of better than 350 dB performance. So to test this aspect I have the ability to turn off different integrators and hence adjust the noise shaper performance. I plan to make two recordings using two Davina's from the same mic feed, with different noise shaper performances; I will then publish the files with random file names and ask people to say which file has the better depth perception. This will be a blind listening test and will give more credence to whether this effect is real or not. Of course, it may be possible that ADC noise shapers do not show this problem - then I will have a serious headache to explain why DAC noise shapers are different!
I note actually with interest your comment "Are humans blessed with a quantum computer in our heads that can do math at 80-bits precision in real time?" And that recalled an article in New Scientist about neurons. Apparently, neuron scientists do not know whether a neuron is a simple 1 bit processor (it's output dependent solely on other inputs) or whether it has internal memory and quantum processing internally, so is actually an N bit quantum processor with a 1 bit output. That's a fairly fundamental knowledge hole... And we have no idea how the brain calculates depth perception to the amazing resolution it actually has. And I am not aware of any published psycho-acoustic papers on the technical requirements for depth perception...
Rob