I have thought for a while about claimed imaging differences between A/D converters. This is because I'm a bit obsessed with imaging myself - I have spent much time perfecting a main microphone setup that allows me to get both maximally sharp imaging and good channel separation at once while recording the L, C and R channels in surround. My microphones are great (all Schoeps), but my preamp & converter is not on the Schoeps level but still probably good enough for now (Focusrite ISA 428 with A/D card fitted).
Since the signal path up to the preamps is already optimized as regards imaging, the next thing to consider is the preamps and the A/D. Now I'm going to concentrate on the A/D.
1. Real or imagined?
The human brain is very powerful. It's powerful to the point that it can invent really obvious - but still imagined! - audible "differences" if there is something visual (or anything else that alters preconceptions) around. A price tag, for example. Also many people try to judge such subtle differences by testing gear A on Monday morning and then borrowing gear B on Thursday afternoon, maybe playing the same instrument as a test but having no way of putting up the microphone(s) in exactly the same relative spot (it should be within a fraction of an inch!), not to mention the effect of the varying air humidity and temperature even before the sound enters the mic. The ONLY way of making really objective and reliable gear comparisons, in my opinion, is to make everything (and I mean everything) in the signal path exactly identical except the gear under test. For an A/D test this means a single signal from the same microphone(s) split into both A/D's at once - not playing and recording first using converter A and then going back to switch cables and recording the same guitar solo (or whatever) using converter B.
My first practical experience of this fact was some 15 years ago when I tweaked my first really good headphones to achieve a more open imaging. I made repeated A-B tests adding and removing damping material in a few strategic places. Once I suddenly forgot if I had or didn't have the additional modification inside, but expected that it WAS there, and once again "heard" this expanded stereo image. It turned out that I was wrong and I had already restored the construction to the original one. But I heard the expanded stereo image very clearly, since I had expected it.
Another typical case is when I was auditioning the claimed imaging differences between two CD players. I A/B'd them for some time, and since it was not a blind test I knew which was which. And I also heard a subtle difference - since I expected it. Because after A/B'ing for so long that I had already forgot which one was "A" and which one was "B", the previous subtle difference disappeared completely.
How many people make real blind tests of their gear? Very few, I would guess. Yet this is the only way to ensure objectivity, and objectivity in comparisons is a good thing even if actual usage of the gear is personal and thus more oriented towards "subjective".
2. Why are there differences in imaging?
So here are my thoughts so far:
- Jitter. It's interesting to note that the inter-sample jitter corresponding to the 20 ps limit stated by Bob Katz in his Mastering Audio book only corresponds to a totally negligible part of the minimum timing difference that can be sensed by the human auditory system (about 1 degree of azimuth, meaning about 10 microseconds). Thus if there were only this kind of jitter, it could have no effect whatsoever on the imaging. However, there's jitter on a larger scale, so if measured between 100 samples or 1000 samples the number could be clearly bigger than expected from the inter-sample jitter. The problem is, as far as I know, that this jitter isn't measured or specified anywhere. How is it visible in a typical jitter measurement?
I have no experience of jitter measurements myself so I haven't made such tests on converters. I would be very interested in seeing such objective measured jitter comparisons of various converters (including my own converter card, which in the ISA 428 is located just above the internal power source, which kind of makes me worried about what the jitter spectrum would look like at 50 Hz!).
Another reason why I have trouble accepting the effect of jitter on imaging is that the imaging is really smeared only if there are inter-channel timing differences or fluctuations. Since all channels are usually driven by the same clock, how can there be large enough such differences to really have an effect on the imaging? (OK, uniform jitter on all channels also creates distortion and you could argue that this distortion slightly masks inner details that are normally sensed as parts of the "image", but I think in a 24-bit system this should be negligibly low compared to other distortions from microphones and preamps. Also see the next paragraph...!)
- What really creates the depth. "Depth" in a recording is conveyed mainly by good reproduction of early reflections (especially from the sides), not by smooth decay of the last parts of the reverberation "tail"! The latter is of course somewhat influenced by jitter and other low-level distortions, but the former cannot be nearly as much affected since they appear at a much higher dBFS level! Often the relevant early reflections are about the same level as the direct sound, of course, so how could a problem at the least significant bit level (e.g. bad dither, small jitter-induced distortion) affect depth? Furthermore, both reverberation and early reflections tend to have quite low high-frequency content, so the differences of various converters close to the Nyquist frequency (due to good or bad anti-aliasing filter design) would have practically no effect on the reproduction of early reflections and reverberation.
- What is imagined and what is not. How many people testing A/D converters in their studios really make sure that the levels are aligned to within a fraction of a dB? If they aren't, the one which has a slightly higher level will no doubt sound more spacious than the other one!
- Other design flaws. As we know, high-end A/D's in general have tighter component tolerances and thus the phase difference between channels (from bypass capacitors, filters etc.) would be lower, and this could account for very much of the difference.
So the bottom line is, has anybody really investigated in depth to what extent these imaging differences are due to all of these various reasons (including "human imagination")? My guess is that if done, many of the claimed differences would turn out to be products of imagination.
I don't want to spoil people's enjoyment here, but I'm the kind who doesn't want to spend excessive money on something until I know the difference is real.
