Dithering about: truth, truthiness and quantization error

[camtwister]

Christopher Montgomery, who enrooted the Vorbis codec and the Ogg container, has written a seducing essay on the futility of 24 bit/192 kHz recordings.

Please read the complete argument before responding to the stimulus.

http://people.xiph.org/~xiphmont/demo/neil-young.html

It's not all 'read it and weep'. You can scoot over here for an alternative conclusion:

http://www.meridian.co.uk/ara/coding2.pdf

[StanleyB]

Christopher is a bit late. I am sure we covered that subject ages ago on AoS.

[morris_minor]

http://theartofsound.net/forum/showthread.php?t=16828

:rolleyes:

[camtwister]

Oh. Sorry, folks.
I missed the original thread.

[morris_minor]

Oh. Sorry, folks.
I missed the original thread.

Easily done! :eyebrows:

[DSJR]

Doesn't hurt to remind though ;)

[drrd]

I thought this was a good presentation too:

http://www.youtube.com/watch?v=1CkyrDIGzOE

[bobbasrah]

A good find that Russell. Very interesting indeed...

[f1eng]

I must say I had tried the PCM-F1 in tape loop test myself decades ago and, like Ivor, could not detect its presence.

I am content that the ADC/DAC process can be completely transparent to me (doesn't mean it always is) and since the Sony device is pretty well from the birth of digital recording, has been, potentially, from the beginning of the red book standard.

As an amateur recordist I have found the only recorders i have used where I can hear no difference whether the recorder is in the monitor loop or not are all digital. I have used fairly high end analogue recorders too.

-Maybe- some people can hear the effect of an ADC/DAC here but I can't.

There are clear differences between all the analogue recorders I have used, all audible. They are all well known, tape hiss, bass uneveness, treble saturation and limited frequency response.

The hiss did not bother me much. I was always disappointed that one did not get a level FR above -20dB on even the best cassettes. It was also an embuggrance that to get HF extension you needed higher tape speed but that this dragged the bass humps further up the frequency band. You can get it not too bad if you work hard, but it is always a compromise and never undetectable in the monitor loop.

[DSJR]

This is where top engineering in the recording stage came in back then, since I'm sure it was commonplace to record KNOWING what analogue tape would do and use it to advantage.

Digital has opened the window hugely wide, but IMO the amount of software now available to destroy or at least modify the sound, has ruined what could have been the very best of times for the recording industry. Instead, the SA-CD releases get the best version of the recorded mix and the sound files seem to be deliberately dumbed down for the red book CD issues, allowing the better ones to be released later on as "re-masters..." making the record companies more money on repeat sales :(

[jkeny]

I thought this was a good presentation too:

http://www.youtube.com/watch?v=1CkyrDIGzOE
Excellent presentation & great find, Russel. Martin Mallison makes some great points & reveals some of the workings of the ESS DAC.

What is interesting to me is the shortcomings of standard Sigma Delta DACs which don't show up in the standard FFT measurements, namely noise modulation & phase shifting.

I also like the fact that, according to him, these engineers actually listen to the DACs :)

EDIT: Mind you I looked at the next video in the series & have to disagree with his evaluation of ASRC as implemented on the ESS DAC chips. The ESS DACs sound better when the ASRC is bypassed so I wonder how much listening they actually do although it is probably expecting too much for them to admit this about the ASRC.

[camtwister]

A good lecture. I would be interested to see some of the equations...to confirm how little I know. Are the synthesis algorithms multi-dimensional FFTs?

[jkeny]

A good lecture. I would be interested to see some of the equations...to confirm how little I know. Are the synthesis algorithms multi-dimensional FFTs?

But I think this misses the point camtwister. On all forums we hear incessant debate/argument about measurements Vs listening - here a chief engineer admits that measurements just don't cut it especially the venerable FFTs. It exposes the lie that "we have all the measurements we need, we just don't know how to interpret them". From this lecture it is clear that they can't measure everything.

[camtwister]

I don't think I'm missing the point because I agree with you, John. Indeed, Russell's link to the video was an excellent response to my original post. My interest in the mathematics is an academic one, not a subjectivist versus objectivist proposition. I’m perverse enough to be interested in the maths for the sake of the mental exercise. I use Fourier analysis in applications outside of audio and was simply curious about the algorithms used by the ESS engineers, in that micro-context only.

[Welder]

I was under the impression that the problems of measuring changes during the transition from one steady state to another were well understood in mathematics at least.
Any change of state involves the application of energy and the law of the conservation of energy would seem to suggest that in this case the conversion may produce noise.
I was also under the impression that the ability of the brain to disregard constant frequency noise in relation to changes in frequency was also pretty well accepted (the waterfall and tiger analogy)
I would have liked to see some evidence that people can actually hear the difference between delta sigma Dacs and other types. Stating his boss and a couple of others can doesn’t really hack it imo.
Anyway, quite interesting

[jkeny]

I was under the impression that the problems of measuring changes during the transition from one steady state to another were well understood in mathematics at least.Yes, well understood in mathematics but why not shown in audio measurements? He stated that FFTs simply did not show this noise modulation but a simpler test did reveal something of what was going on. Mathematics define idealised states which when dealing with the real world is a false premise! Moreover, the great reliance on FFTs in audio is misleading & has brought us to the situation where many argue that it is impossible to hear this as it is so far down in the noise without actually considering if there are any weaknesses to the FFT tool or to how it is being used.

Any change of state involves the application of energy and the law of the conservation of energy would seem to suggest that in this case the conversion may produce noise.
I was also under the impression that the ability of the brain to disregard constant frequency noise in relation to changes in frequency was also pretty well accepted (the waterfall and tiger analogy)You missed the bit about it not being a constant frequency but in fact changes with frequency & with level - so a modulating noise which psychoacoustics predict will not be disregarded.

I would have liked to see some evidence that people can actually hear the difference between delta sigma Dacs and other types. Stating his boss and a couple of others can doesn’t really hack it imo.
Anyway, quite interestingWell, a lot of what he says is at the cutting edge & I guess a body of evidence needs to be accumulated for some or personal experience will suffice for others. For me his presentation rings true & agrees with my thinking & others who are following the same logic. I know people who can hear these differences & say that S-D in general, lacks the "jump factor", the instant dynamics that can shock as well as other elements of the sound field that are less natural than they are in real life. I know of one person who already came to the same conclusions as the CTO of ESS - it is the noise modulation that is at the heart of the issue & this i snot shown up on FFTs. SO you can have two DACs that show exemplary FFT measurements but sound very different!!

[jkeny]

Yes, Camtwister, I can understand you being interested in the beauty of the mathematics. I was just using your post to point up what I thought were some real important points in that video about measurements & also just the refreshing vista of a chief engineer admitting that listening shows his measurements are missing something. I thought that it should be lauded as a spectacle of someone deeply rooted in science & analysis, actually saying hands up, our measurements aren't revealing enough & then doing something about it - finding other measurements that did reveal the differences in that particular instance. It was a good example of open-mindedness & all that is praiseworthy in the scientific approach!!