5196106216

scratches and fingerprints on the disc surface.

Over and above that, according to the specifications, CIRC is capabie of reconstituting morę extensive losses of up to 12,300 bits, by adjacent sample interpolation, equivalent to 7.7mm of track length. Beyond that again, the system mutes the channel rather than et through a burst of noises.

For those incjined to tackle it, the theory bcjiind such claims is well docu-mented in academic literaturę. Alterna-tively, in his helpful book “Principles of Digital Audio** (Howard W. Sams & Co. Inc) Ken Pohlmann explains the basie principles of error correction in morę readable form but, even so, don*t expect to take it in over a casual cup of morning coffee!

In their early CD literaturę, Technics claimed to have developed an original Super Decoding Algorithm “so ad-vanced that the chances of being able to retrieve the correct data are an inered-' ible once in 5000 years”.

w

About the same time, Sony published a graph showing the performance of their own “Super Strategy’* restoration method. The “Probability of Detection and Compensation Error (data word)” axis was calibrated all the way from 1 Word/Minute to 1 Word/10,000 years!

It is intriguing to speculate how the respcctive companies could have arrived at such odds and what they really amount to in practice. But the 4(XK)-plus year time-frame certainly madę for eye-catching headiines w hen the CD system was first announced!

Apart from anything else, it is one thing to restore or conceal aberrations in a data stream. It could be — and is — quite another to maintain proper tracking when segments of the micro-scopic data spiral are obscured by scratches or fingerprints. An entirely different circuit function is fnvolved.

The decoding circuitry?

To unscramble the digital data and recover the discrete left and right audio channel information, all CD players must obviously incorporate processing complementary to the basie architecture of the encoding system. And, because it includes CIRC error correction, it is often assumed that all CD players incorporate esscntially similar circuitry to de-tect, correct and/or conceal errors.

But here we get into a distinctly grey . area, with everyone listing CIRC error correction in their specification, some invoking fancy terminology but nonę admitting to — or being accused of — cutting cost in this area.

Kcn Pohmann says, however: “Not all CD players are alike in terms of error protection. Any CD player*s error protection ability is limited to the suc-ccss of the strategy chosen to decode the CIRC data on the disc and perform conceal ment**.

According to Technics: “Many error correction Systems are employed, from the very simple to the extremely com^: plex”. Sony*s graph, mentioned earlier, contrasted their own “Super Strategy” with a “standard error correction system” and a “simplified error corTection

system”.

0

But who uses which, in currently available players, and to what effect? I’d hate to be holding my breath, await-ing the answer!

Other interesting figures:

A couple of readers' letters, referring to earlier discussion of 4-channel compact dises, included photostats which served to confirm that the 4-channel modę did receive carly mention. But the photostats happened also to carry information on “BER” — a term that is subject to some confusion but which can usually be taken to mean “Bit Error Ratę’*.

The following figures, listed in the Philips Technical Review Vol 40/6, ap-pear to agree with those in an early Sony publication credited to Nakajima, Fukuor and Iga. (For the references, I am indebted respectively to A.J. of City Beach, W A and to R.P. of Point Fred-erick, NSW).    •    •

•    Maximum completely correctable burst length: 4000 bits (2.5mm track length)

•    Maximum interpolatable burst length in the worst case: 12,300 bits (7.7mm)

•    Sample interpolation ratę:

i per 10 hours at BER = 10"*

1000 per minutę at BER = lfr³

•    Undetected error samples (clicks):

Less than 1 every 750 hours at BER = 10³

Negligible at BER = 10"* or greater

Based on what has been said to this point, I would discount the proposition as originally put forward by R.Webb of London, on the grounds that isolated data errorc appearing in individual words are (a) unlikely and (b) completely correctable, therefore non-exist-ent at the D/A converter input.

His proposition, like minę in the April issue of EA, would morę likely relate to what Sony describe as an “un-favourable” situation w here, as listed above, the “ raw” bit error ratę (i.e. as read off disc) averages lfr³ (one incor-rectly recorded bit in every 10³ or

1000), and where the sample interpolation ratę may get up to around 1000/-minute.

The figures would suggest extensive scratches or finger marks on the surface of the disc, sufficient to defy digital redundancy correction and to “flag” at least a couple of major burst errors per revolution, requiring interpolation.

Provided tracking was maintained, ' they would still not create a significant click problem (one dick every 750 hours?) although “bush” logie might well insist that all those interpolations -- all that Computer guesswork — must somehow prejudice quality.

But before proceeding to that conclu-sion, let’s get certain things into per-spcctive.

BIT ERROR RATĘ: As already

0

mentioned, BER = 10³ represents very “unfavourable” playing conditions, con-ceivable with people who are careless with their dises but scarcejy applicable to those concerned with a quality result. But even 1000 interpolated (computer guessed) samples is a very smali number compared with the 5,000,000-odd audio samples pouring down the data stream at 2 x 44.1 thousand samples per scc_: ond. BER figures are admittedly ambig-uous but anyway you care to interpret them, the interpolated samples add up to a tiny fraction of 1% — a far ery from R.Webb’s 50%.

BER = 10"* is morę typical, with the interpolation ratę falling to virtually zero.

Ken Pohlmann, quoted earlier, has this to say on the same subject:

“Theoretically, the raw bit error ratę on a CD is between 10-⁵ and 10*, that is there is one incorrectly recorded bit for every 10⁵ (100,000) to 10⁵ (1 million) bits on a disc ... In practice, because of the data density, even a mildly defec-tive disc can exhibit a much higher bit error ratę”.

What do customers say?

While the above figures tend to discount the “interpolation distortion” concept, they do not necessarily reflect everyday expcrience with current dises and current modę! players. So I sought a few reactions.

In a large department storę, I ex-plained to the lady behind the record sales counter that I was writing an arti-cle on the generał subject of compact dises and had been wondering whether they reccived many complaints about quality. She was obviously puz/.led as to why such a question should arise. Customers seemed very happy with the

27

ELECTRONICS Austrajia. September 1987