I enjoyed Robert Harley’s “What Exactly Are CD-Quality and High-Resolution Audio?” in the July/August issue and found much to agree with. I also found a few things that are at odds with my own experience.
As one of the first engineers in the world to create masters for CD, I have often spoken about how difficult it was to get original masters and how those instances where I did remain vivid in memory after all these years. However, the situation was not as bad as the “three, four, five or more generations removed from the original” the article mentions. More often than not, the sources I was given to work with were a single generation away from the mixes.
When it was an “EQ’d copy” created during the vinyl mastering, this was most often made from the original mixes (which were used to cut the lacquers for vinyl). On other occasions, it was a flat, 1:1 transfer made from originals which might have been say, in in Europe. (Of course, different engineers might create different sounding “flat” transfers.) If that 1:1 was the tape sent to the vinyl cutting house the EQ'd copy would of course then be one more generation away from the originals. That, to my knowledge, is a far as it ever went - at least for the tapes I worked with.
One slight advantage the 1630 system at Atlantic Studios had was the Apogee retrofit filters we installed. These made for an appreciable improvement in the sound of the converters. Of course, in the ensuing years, converter technology improved significantly, with devices such as the landmark Pacific Microsonics unit mentioned in the article and more recently, with converters such as Metric Halo’s ULN-8 - the unit which has so far outperformed all others in my experience and when used at 24/192, marked, for my ears, the point where digital was no longer inferior to the best analog and indeed could surpass it.
One last point, regarding sample rate conversion: I often see mention of integer conversion (e.g. converting from 176.4 or 88.2 to 44.1 as opposed to converting from 192 or 96 to 44.1) being “better” or “less sonically detrimental” than non-integer conversion. In my experience, the "even multiples" (i.e, integer math) requirement is usually the result of a given sample rate conversion algorithm's difficulty in "handling the math". That is, the "requirement" exists only for not-so-good algorithms.
What I've found is that with the better algorithms, there is no such requirement. In fact, the better algorithms don't seem to "care" and will do non-integer conversion much more transparently than many other algorithms will do the simpler integer math. In other words, it comes down to the quality of the algorithm and not the task which it is asked to perform. Lower quality algorithms will balk at the more complex math and provide (even) lower quality results than they would with the simpler math. The better algorithms don't care either way.