| Home | Audio mag. | Stereo Review mag. | High Fidelity mag. | AE/AA mag. |
|
Catch-Penny CD--A Guest Editorial If you have wondered why the sound of compact discs differ so markedly, ponder some intelligence provided by Keith Armstrong, one of our professional readers who lives and works in the United Kingdom. Others have observed long since that some of the digital processing equipment recording companies are using is inferior to the home playback systems we own. Armstrong, responding to Vytenis Babrauskas' article, 'CD Group Delay Corrections’ (TAA 4/87, pp. 8-11, 59), says that the latter is quite correct to conclude that the desirable place to insert recording error compensation is in the recording chain itself... ’ I take the liberty of including the remainder of Mr. Armstrong's letter as an important editorial comment worthy of our consideration. -E.T.D. As a designer of professional digital recording systems, I would strongly discourage anybody from attempting to correct recording-side group delay error by adding compensation in their playback side. The reason is the complete lack of standardization of this aspect of digital recording. John Meyer, in ‘Time Correction of Anti-Alias Filters used in Digital Systems ’ (JAES 32, 132 7, March 1984), measured digital recorders containing one stage of analog to digital conversion and one digital to analog, and found a typical group delay error of 340° lag at 20kHz. What is not generally realized is that a CD may be the result of a varying number of conversions: from one to five (three analog to digital and two digital to analog), and the group delay errors could add up to as much as 1,000° at 20kHz using some equipment. (And five conversions is by no means a maximum number!) The reason for the large number of possible A-D-A conversions is that most people still use analog mixing con soles, and every time they use a digital tape recorder they have to revert to the analog domain to process its output. Even when transferring from the final stereo digital master to CD most people use an analog transfer mixer. As a rough estimate, if the letter code on your CD is AAD it has had one conversion, if it is ADD it has had three conversions, and DDD implies five conversions, possibly more. As if this were not Dad enough, currently 'available analog to digital converters used in professional s recording equipment have a group delay error that cannot be relied upon to be 340° at 20kHz: some are as low as 5°. Clearly, the only sensible place to compensate for group errors is where they occur (and in fact the best way to design an accurate correction network is at the same time as designing the anti-alias or reconstruction filter itself). When I designed for Neve (a prestigious name in the audio mixer business) on their digital mixing console, I took this approach. The anti-alias filters that Neve uses thus have only 30° group delay error at 20kHz, an error that compares well with the performance of most analog consoles. What is more, the stereo channel tracking of this error is a maximum of + 2° in the range of 400Hz to 15kHz to preserve the stereo image as far as possible. The filters are made for Neve in England by Kemo Ltd., and they are not the cheapest. This sort of attention to detail is a characteristic of Neve desks and is why they are more costly than most others. Unfortunately, too few recording studios are sufficiently committed to sound quality to have purchased the Neve digital desk, which is expensive by any standard. Neve also makes a Digital Tape Transfer Console, which allows the final transition from stereo master tape to CD to be accomplished entirely in the digital domain. Quite a few of these have been sold, but the AAD-ADD-DDD label on your CD will not tell you whether or not one was used. An ideal digital recording studio would consist of a digital console (for mastering and mixdown), digital recorders (for multi-track and stereo masters), and a digital tape transfer console to create the CD information. Only one stage of A to D conversion would be employed, following soon after the console's input amplifiers, and if it was a Neve desk the A to D would not create group delay errors appreciably worse than a good analog system. The only CDs I know of that can boast this degree of all digital recording using the Neve digital consoles are: ‘Blue Skies,’ the Nelson Riddle Orchestra with Kiri Te Kanawa on Decca 414666-2, 1985; and ‘Fire and Ice, ’ a recording for a Christmas television show, which I believe is released by First Night Recordings, Ltd. Most digital audio recording systems are poor in their group delay correction for two main reasons. First, the A to D converters are only just fast enough, and to prevent aliasing (which would generate nonharmonic distortion products in the audio passband), very sharp cut-off filters are required at around 20kHz. The design of these filters, which can be 9th-order elliptical, is not trivial -never mind their group delay compensation. Second, many digital recording systems are designed by digital engineers-strong on digital processing and software, but weak on analog. Thus decades of experience in what makes good sound are lost because the essential analog ingredient is left out. These problems did not exist in the Neve digital con sole design, and group delay distortion in particular will be less of a problem in the future as A to D converters get faster-allowing the use of oversampling techniques similar to those that have revolutionized the quality of CD playback systems. It is likely that adding so-called Group Delay Correction circuits to your CD player will alter the sound: similar phase manipulation techniques are used in some recording studios to ‘brighten up ’ dull recordings. The results will not be repeatable from one CD to another, and they are also artificial and inaccurate. If you want low group delay errors from your CDs, lobby the recording companies and studios to encourage them to take more care and buy better equipment. ------------------ Also see: CD REVIEWS, By John Sunier
|
Prev. | Next |