So I’ve got my vinyl back up and running after an eight-month hiatus during home renovations. And lovely it is to slide some big black discs onto a chunky platter. But I’ve also spent plenty of time this issue sampling the latest in digital sources. Not ye olde CD players, no no. We’re talking boxes that aim to combine the convenience of digital files and online sources with a bit of quality in the output stage. First up was Bryston’s BDP-1, a thrilling performer particularly with the high-res files it was built for. Then the Marantz NA7004, surely a harbinger of the future, accessing assorted online digital music sources, plus operating as an effective DAC and standing ready to receive the wonders of Apple AirPlay should Apple ever push its particular button (hopefully pushed by the time you read this).
And I also bought myself an Apple TV, predicting it would do a fine job of streaming my tunes ‘Airport Express’-like to the newly-built and datacabled areas of my home. Which it does. It took me a while to notice, however, that Apple TV’s output on optical digital is a constant 48kHz, regardless of the file it’s playing. While this may provide marvellous compatibility with anything on the back-end, my audio hackles were raised by the idea of such unnecessary resampling, which can only blur the accuracy of sample levels, particularly with frequencies close yet mathematically unrelated.
Then it occurred to me — this resampling process is, in fact, reversing a previous resampling. Recording studios nearly always master at 96kHz or 192kHz, multiples of the early standard of 48kHz. Only as a final step do they downsample to 44.1kHz for the CD. So we have been living with this perilous resampling since the early days of digital recording.
Sadly the reverse process in the Apple TV will not simply return the original bits. Both processes are destructive to the original data.
So why did we end up with such an odd mismatch of frequencies for CDs in the first place?
The rough figure of 40kHz has an obvious derivation — the top reproducible frequency is half the sampling frequency*; humans can hear up to 20kHz**, so there’s your 40kHz sampling rate. Add a bit to take account of a filter slope, and we end up around CD’s 44.1kHz. But why the difference between this and the 48kHz of studio recording?
There are some delicious rumours that the difference was deliberately introduced by record companies (Sony) paranoid about people making perfect copies of CDs to such media as DAT (digital audio tape). Quite believable, though I never heard of this at the time.
More likely is the explanation from British audio writer John Watkinson, that the 44.1kHz comes from the days when digital audio was recorded to video tape, back when hard drives didn’t have enough capacity for long recordings, even if the write times were fast enough for sufficient bandwidth. I still have one of these ‘pseudo-video’ tapes from the first time I ever experienced the terror of paying for studio time (tick tick, another $10), and if you play that old tape on a VCR, it looks like rippling snowy vertical stripes. These are visually-stored audio bits, so the achievable bit-rate depended on frame rate, number of lines per frame, etc. The sums can be found online; basically 44.1kHz was ideal to work for both 50Hz and 60Hz video systems once blanking lines were considered. It was locked in by the Red Book standard for CD mastering.
So all our CDs have been carefully blurred in the path from studio to CD. Now that we have digital downloads, we might hope this would be unnecessary. Even if we can’t yet get hold of 24/96 files for everything (and why not, I may ask?), could we not at least download 16/48? Sadly, not yet; downloads continue to be derived from the CD masters; the frequency remains 44.1kHz.
Should I panic at Apple TV’s double conversion of my tunes? I don’t think so. I attempted a blind A-B listening test between the original 44.1kHz files played from USB (through Bryston’s digital music player and DAC), and the Apple TV output (also through the Bryston DAC). I thought I could hear differences, but they didn’t correspond to changes in input so, effectively, I couldn’t pick a difference.
Still worries me, though. I’ll play some more vinyl while I’m thinking…
Jez Ford, Editor
* Nyquist’s theorem
** This may now apply only to young virgins as yet undefiled by a surfeit of popular music supplied through earphones.