Greg discovered an apparent contradiction in my article on jitter: Why did I say that DATs don’t seem to pick up jitter, but AES/EBU CDR’s seem to pick up incoming jitter.
Date: Wed, 17 Jul 1996
Tp: Greg Simmons… Jodie Sharp [email protected]
Dear Greg:
Thanks for your comments. I’ll see if I can clarify…please consider the article on jitter a work in progress. Your question did hit the nail on the head on some missing information in the article, and I thank you. In my copious (????) free time I’ll try to clarify the article.
Dear Bob,
Your essays on dither and jitter are excellent, authoritive and very informative – I’ve learnt heaps from reading them. However, in your essay on jitter, you say some things that appear contradictory and have me slightly confused.
Firstly, you say that “…Playback from a DAT recorder usually sounds better than the recording, because there is less jitter. Remember, a DAT machine on playback puts out numbers from an internal RAM buffer memory,locked to its internal crystal clock.” Shortly afterward, you say: “…a compact disc made from a DAT master usually sounds better than the DAT…because a CD usually plays back more stably than a DAT machine.”
Under the heading ‘Can Compact Discs Contain Jitter’, you say “…An AES/EBU (standalone) CD recorder produces inferior-sounding CDs compared to a SCSI-based (computer) CD recorder. This is understandable when you realize that a SCSI-based recorder uses a crystal oscillator master clock.” The text continues by discussing the differences between the PLL system used by a standalone AES/EBU recorder and the crystal oscillator system used by a SCSI recorder.
The paragraph closes with “…No matter how effective the recorder’s PLL at removing incoming jitter, it can never be as effective as a well-designed crystal clock.”
Do you see the source of my confusion? When discussing DAT and dubbing in general, the essay suggests that jitter on the incoming data is irrelevant during playback (so long as the jitter is not so high as to cause actual errors) because after being recorded, the data is re-clocked by the playback machine’s internal crystal oscillatorlock. And yet, the essay also suggests that jitter on the incoming data (as in the case with the two different types of CD recorders) does affect the final sound.
Excellent point. I will have to include this answer in the next revision of the article:
First of all, the remnant playback jitter (“intrinsic jitter”) of a DAT machine is significantly higher than the remnant jitter of a good CD transport. Bob Harley measured 1.2 nanosecond RMS jitter on the output of a Panasonic 3500! The remnant jitter on the output of a great CD transport is on the order of 10 to 100 picoseconds. A nanosecond is 1000 picoseconds, so you can see there is an order of magnitude difference.
Power supply design, grounding, can affect the quality of these cloks, and audiophile CD transport designers pay special attention to the power supply. A poor power supply can affect the remant jitter both by contaminating the crystal clock, and the AES/EBU trasmitter in the digital output stage. Until someone examines the internal mechanisms of both reproduction systems with very sophisticated measurement equipment, we can only hypothesize. But for now, it is enough to say that the measured intrinsic jitter of a DAT reproducer is greater than 100 times the jitter of a good CD transport. We all know that shouldn’t be happening… all digital reproducers should measure perfectly–right? Good thing we are able to measure those differences, or the golden ears would all be in a pickle trying to demonstrate why DAT playback just doesn’t sound as good as CD playback.
Now for your question, you recall that my tests with DAT recorders seem to confirm that jitter coming in is irrelevant to the final playback. This does seem to confirm the proper action of the phase locked loops and FIFO’s in the DAT machine. But why doesn’t it seem to hold true with CDRs? Well, the deeper we dig, the more we learn. On the surface, the science holds true, but…
One of my hypotheses is that the residual jitter level of 1200 picoseconds in a DAT reproducer has a very large masking effect. It is highly unlikely that any small remaining differences or variances (say plus or minus 100 picoseconds) due to variant jitter of sources could ever be heard or reliably measured. Especially after reduction through the recorder’s input PLL, and especially after the separate blocks on tape are reassembled and then retimed through the reproducer’s output FIFO. So the question with DAT machines becomes a moot point, as far as I’m concerned.
Another hypothesis is that the data block structure of the CDR is different from that of the DAT, and may have an effect on outgoing jitter. In both types of playback, however, data is extracted in a “jittery” manner, and always smoothed by FIFO, so the different data block structure would have to have an indirect influence on the output clock.
Another possibility is error correction, and again, only through an indirect influence, common impedance coupling through the power supply. Perhaps the CD player’s design is more susceptible to that than the DAT. It seems the problems I was alluding to are only relevant to a very low jitter medium (such as CD). In a low-jitter CD player, we can examine and test for “microcosmic” influences on the stability of the player’s crystal clock and see if they are caused by “microcosmic” differences on the CD disc. It took David Smith and Sony Corporation months and months to devise some sophisticated audio tests in order to conclude that the golden ears were right!
This brave work was undertaken by design engineers working for the very company that had designed the “perfect” FIFO system for CD players which is supposed to eliminate all outgoing jitter (or at least reduce it to the residual of a crystal oscillator). It will take months to years before scientists with sophisticated measuring instruments find and eliminate the subtle internal mechanisms in a CD player that are somehow permitting jitter differences to be heard through a supposedly “perfect” system. So far, no equipment designer has succeeded in producing a jitterless playback system (everyone says it’s possible)–although great improvements have been made (listen to some of the newer audiophile digital reproduction systems and to a couple of the finest professional D/A converters).
We’re trying to find flies on an elephant, here. Unfortunately, on the audio side of this mixed metaphor, the human ear can hear the flies very well. A good scientist mustn’t assume anything, nor take anything for granted, and must recognize that all conclusions are based on some underlying hypothesis or axiom.
Surely a CD player is just like a DAT machine on playback, and uses it’s internal crystal oscillator to clock out the data, therefore reducing the problem to the jitter inherent in the internal crystal oscillator clock and eliminating any jitter caused by the disc recording process, whether the data was recieved via AES/EBU or SCSI. Or is there something I don’t understand?
I’ve heard terms such as pit jitter and land jitter, are these what you’re referring to?
Pit and land jitter on the CD may or may not be the cause of the differences we are hearing. Some other mechanism on the CD (size of pits, not necessarily the spacing of pits) may be causing the servo mechanism in the player to be more jittery. It is definitely not data errors. Research has shown that these CDs which we claim to sound different have identical data. But part of the problem may be due to error correction, with the error correction system causing problems, again by power supply coupling. Very far-fetched argument, yet to be proved. Same with the servo mechanism leaking into the power supply for the output crystal…engineers have found a 25 cent power-supply bypass capacitor in the digital section to do wonders on the audio quality, so this is pointing to the reasons.
And to complicate the matter, the analyzers which look at pit and land jitter on CDRs generally do not look at its frequency distribution. For example, 10 picoseconds of peak to peak jitter with a central peak at 3 Khz is likely more audible than 500 picoseconds of random (uncorrelated) RMS jitter.
It takes far more sophisticated equipment to make the second measurement. I’ve had a plant analyzer show the reverse result, RMS jitter was higher in the CD that played back with apparently lower jitter. That is, if pit and land jitter on the CDR is even the root cause of the sonic differences we are hearing. When we hear a CD that has a wider soundstage, greater apparent low level resolution, and other audible differences, we assume that is caused by jitter differences on the CDR itself. But this is only a very unscientific hypothesis. And no standards as yet have been developed that correlate measured jitter against listening differences.
However, advancements are being made, and a specialized test system that looks at the analog outputs of CD players (and D/A converters) has been developed.Paul Miller’s company employs Julian Dunn’s specialized test signal for this purpose.
Also remember that what I said in my article remains true: that you can copy from a CD that supposedly sounds “degraded” through a SCSI interface back to another CDR or to a hard drive, then cut another SCSI CDR, and the end result can sound better than the original if the new writer is better than the original writer! Jitter is NEVER transferred with the data to a new medium, if a clock is not involved. And SCSI does not involve a clock. Jitter is strictly an interface phenomenon, whenever a clock is involved.
Finally, I’m sorry to send you such a long winded message but I genuinely would like to understand these things. I suspect I may be interpreting your essay incorrectly. Please don’t take this message as a criticism of your essay – I have nowhere near enough knowledge or experience in this field to criticise your essay!! I’m just confused…
No problem… Jitter is a complex subject, no one knows all the answers. I ask more questions than I have answers. Someday I will try to rewrite my essay to incorporate all of these considerations. I hope this letter helped!
BK