Comparing CD audio quality to higher resolution
What is the question?
- To test 44,1k / 16-bit files against different higher resolution files or
- Or to test CD as a media against higher resolution download?
It is not a straightforward question and easy to make a meaningless comparison.
- As usually the recording is made in higher than 44,1k/16-bit sampling, the 44,1k version always is downconverted. So technically, it will be one step further away from the master file, which should not sound better in any way in a correct setup reproduction chain. Technically.
2. How different should it be?
It depends.
Practically all the tests I know about is using an oversampling or upconverting D/A converter. It can hide some of the differences. So I think a proper comparison should use a so-called non-oversampling D/A converter which runs both kinds of files without processing “natively”.
3. The other question is how we play back the high-res files? What way (using PCs, digital audio workstations, streamers or some dedicated hardware platform) will show a meaningful and generally representative enough result to make a widely valuable conclusion?
4. There is an exciting test published two weeks ago by Milind N. Kunchur about cables and their effects in high-end audio:
http://boson.physics.sc.edu/~kunchur//papers/Audibility-of-cable-pathways--Kunchur.pdf
It is as scientific as it could be and shows many, many exciting conclusions. I would say it is a game-changing test. It is very well documented and thoroughly setup, and probably it will be part of the authors’ forthcoming book about high-end audio.
His notes about blind testing and short music samples, and quick switching is eye-opening:
“Overall, the brain appears to cumulatively gather and retain fragments of information to progressively synthesize an increasingly detailed picture. This cumulative process might potentially lead to another complication in listening comparisons that may especially affect the very first (virgin) trial. If the lower resolution configuration B is followed by the higher resolution configuration A, listeners may be more likely to judge that A is more detailed by noticing the extra revealed information. However, when A is played first, the brain may remember its more detailed sound and fill in the details when B is played later making it harder to discern the cable interchange (this should be especially true for musicians). Such filling in of missing information based on memory/expectation also occurs in speech recognition as the well-known phonemic restoration process [32]–[33] and occurs in tone perception as the illusory continuity effect [34]. The measurements of the present work show a suggestive hint of this hypothesized effect.
The above observations explain why a blind testing procedure that quickly switches back and forth between A and B might be less effective than one that uses extended segments as in the present work. The latter allows the listener to “see the distortion from many angles” to notice an unmistakable (or less mistakable) pattern. “
5. Another exciting question how can different kind of genres (and post-production, mixing, mastering) influence the result? Different genres require different post-production processes, and they can result in different loudness and distortions. These can mask the real difference in several ways. Does it matter if the recording is primarily acoustic and made in or out of the studio on a natural location? Studio recordings (because of the overdubs) can show very different result than minimally miked acoustic recordings with natural reverberation and not with artificial reverbs.
So as always, in high-end audio, clear, straightforward answers are not existing. Anyone to do proper testing should follow how Prof Kunchur’s test was conducted and documented.
