CU Errors and Recovery Algorithms
default -- Mon, 09/21/2009 - 11:35
Hello TAS Readers & Editors:
I am posting this comment under D/A Converters because it was spurred by reading the "Further Thoughts - Benchmark DAC1 PRE" article in the October '09 issue of TAS. When Benchmark's own John Siau replies to Alan in the 2nd half, he attempts to explain why Alan found the sound quality of the Rotel CD player / Benchmark DAC1 combo to be only so-so. Insisting that it couldn't be jitter causing the problem, he asserts that it must be due to the transport or CD itself having "non-recoverable read errors (commonly referred to as CU errors)", and/or the transport's "concealment algorithms" which are invoked on-the-fly to repair them.
His assertion raises a big question about the accuracy of CD transports in general. Assuming I have a fairly well-made CD transport (and some clean CDs), HOW BIG of a problem are these CU errors (and the attempts at recovery) typically? In the millions of samples on a given CD track, roughly how many of them end up needing immediate "filling in" (i.e., being fudged) by the transport? My guess is that there shouldn't be enough of them (especially in a first-rate player) to justify explaining why the Benchmark DAC1 would perform less than perfectly with a non-reference-quality CD transport such as the Rotel.
Phil G. - TAS Subscriber
All content, design, and layout are Copyright © 1999 - 2011 NextScreen. All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
If you read even other manufacturors response, they all say thanks and are super happy if they get a positive review(Benchmark was happy too and advertised accordingly). And naturally they defend their product if they get a negative review. Anyone would after putting their heart and soul into it. And that alone makes one wonder that how come for 3 years Benchmark has been making huge sales on this and they didn't mention anything about transports all being so different. Now its all the fault of the transports. I think the new competitive products are getting ahead addressing the problems (jitter sensitivity) much better than benchmark is and its time for them to work on upgrading their product (Which they are working on according to John Siau). As for CD transports, Per TAS reviewers, Transport quality makes a difference and with the new DAC's it makes less of a difference but it still does make a difference. Also most believe that Hard disks sound better than CD transports, although hard disks can be unreliable. FYI, RH will be doing a review of the P.S. Audio transport in the near future; AT compared 2 transports with the Benchmark followup. I think it would be best if REG would start participating here and on TAS and possibly get together with the Benchmark guys(John Siau) and do a test with multiple transports to match the Benchmark. This will help all the bench mark owners to get the best out of their unit. I don't think it should be that difficult for the two to come together and arrange for a followup of this type with transports. Adding a volume control over the years is all great, but the competition is getting stiff. Others are also getting better and cheaper in this segment of the market.
Sam - Thanks for your reply. I agree that the DAC competition is heating up, and that Benchmark should be mindful to keep their products ahead of the other similarly-priced DACs. My main concern, however, is the underlying message implied by Mr. Siau regarding transport CU errors. Has anyone ever performed a "sample" comparison between a digital master file created, say, from a live performance (at the Redbook CD sample-rate & bit-depth) and the digital output of a reference CD transport playing a freshly stamped CD from that file? What percentage of the samples might we find in error here? 0.004%? 0.02%? 0.3%? What value should we say is insignificant? And how much do we have to spend on a transport to render these errors as insignificant? Finally, if transport CU errors aren't generally that big of a problem, the Mr. Siau will need to look elsewhere to explain the Rotel/DAC1 situation.
As George Carlin once said, "Someone has to think of these things".
Phil, I am not qualified to answer that question. It would be nice if R.H. can answer your question. In the past he has had direct experience working in audio labs. Also if you search 2008 RMAF digital seminars posted on AVguide.com they breifly discuss these things. You might want to contact Benchmark directly as well for a transport recommendation. My DAC company and my dealer both were very straightforward with what kind of transports worked well in their experience with the DAC that I bought. Best of Luck!
That is the issue....are CU errors a significant issue in player performance, or not. If not, then Benchmark is not being straight about their machine. If it is a problem, then why does no one ever measure or mention it. I know my DVD player sounds awful and my CD player sounds very good. Why the difference?
Phil G, you have stated very well my similar concerns upon reading John Siau’s reply to AT’s comments about the Benchmark DAC. There must be many others wondering about disc unrecoverable read errors and where it sits on the scale of one-to-ten for importance in the audio chain. I don’t recall anyone giving much attention to measuring and discussing read errors previously, leaving me with the impression it’s a non-issue. So, one less thing for an audiophile to worry about (I hope).
But to be cynical for a moment, what if it’s easier to just assume that read errors don’t exist significantly in any transports because it’s too hard to test for read errors without expensive equipment and special skills and it requires more time and energy to test than most want to invest. That’s just a what-if. But, let’s not forget that John Siau also stated provocatively that error rates vary significantly from CD to CD and from transport to transport and he mentions a white paper that discusses it. So are bad error rates really out there in the wild?
Normal
0
MicrosoftInternetExplorer4
/* Style Definitions */
table.MsoNormalTable
{mso-style-name:"Table Normal";
mso-tstyle-rowband-size:0;
mso-tstyle-colband-size:0;
mso-style-noshow:yes;
mso-style-parent:"";
mso-padding-alt:0cm 5.4pt 0cm 5.4pt;
mso-para-margin:0cm;
mso-para-margin-bottom:.0001pt;
mso-pagination:widow-orphan;
font-size:10.0pt;
font-family:"Times New Roman";}
Dear Alan Taffel, Robert H
Re Benchmark DAC1 Further Thoughts
Benchmarks John Siau says differences heard by Alan Taffel may have been caused by “the audibility of a difficult-to-hear CD playback defect ,” non-recoverable read errors known as CU errors.
Alan Taffel says given the jitter elimination techniques of the DAC1, the two sources should have sounded the same.
They did not, they sounded distinctly different.
RH ‘Do Data Errors Really Degrade Sound Quality’ (TAS #171 p131) tells us from his “analysis it seems obvious that bit errors are not a source of audible degradation in CD playback.”
Charles Altmann / www.jitter.de tells us “the cheapest CD players are able to digitally output the exact data (with jitter, but original data).”
He also says “Unfortunately CD and other transports generate a lot of jitter below 10kHz,” which seemingly standard jitter attenuators don’t handle well, but apparently his patented Jitter Scrambling Decorrelator does.
Alan heard far more similarity between the two sources through the Bryston BDA-1 DAC.
‘Difficult-to-hear’ and ‘distinctly different’ are not compatible terms ie distinctly different means ‘Not-difficult-to-hear.’
So if the Benchmark DAC1 removes jitter, and CU errors are difficult-to-hear, what did make the same sources sound distinctly different through the DAC1 but far more similar through the BDA-1?
I spent nearly four years working in a CD mastering lab working on a wide range of projects. One of those projects was modifying Philips CD players to extract the error detection codes for output to a custom PC-based error analyzer. These modified players were used in the QC department of the small pressing plant. Another project involved performing bit-for-bit comparisons between the source data and replicated discs made from those data. Yet another related job was intentionally scratching discs and analyzing the error syndromes.
My experience suggests that uncorrected data errors on CD are extremely rare unless the disc has been abused. CD's error correction system can completely correct (not conceal or interpolate) burst errors of up to 4000 consecutive bits. Such an error would require an easily visible scratch on the surface that happened to be radial, following the spiral track rather than a scratch from the center to the outer diameter. CD's error correction is amazingly robust. One technique is interleaving, in which the data are not recorded on the disc in linear form but are scattered. Deinterleaving on playback puts the data back into the correct order. Burst errors, such as caused by a scratch or local defect, are converted into several shorter errors that are more easily corrected.
Error concealment (interpolation of missing data) begins at 4000 consecutive bits and operates up to 12,000 missing or corrupted bits, after which the player mutes.
Early CD players used several separate chips that allowed access to the various error-syndrome flags, making it possible to see the severity of corrected errors. The worst error that could be completely corrected, called an "E32" (three consecutive corrupted symbols at the C2 decoder), was extraordinarly rare. Production CDs simply don't have errors that even approach interpolation.
Thankyou Robert - very informative.
Your experience suggests uncorrected data erors on CD are "extremely rare," (and difficult to hear according to Benchmark).
The musical waveform generated by a CD player is two dimensional, amplitude vs time (vertical axis vs horizontal axis).
We have amplitude covered seemingly, leaving the big ugly jitter bug the culprit.
Is this the key difference between transports ie the best sounding transports/drives have audibly less jitter?
If the data are identical, the only factor remaining that could influence the sound is timing (jitter). There is one other potential factor: noise on the source ground getting into the D/A.