Those who expect one CD player to "blow away" another in the current generation of 16-bit, 4-times oversampling, high-end models should read what follows here with particular attention.

Before we can meaningfully discuss individual pieces of equipment, we must clarify certain basic facts of digital audio which appear to be confusing or painful to a surprisingly large number of audiophiles. We touched on these topics in Issue No. 10, but our mail indicates a need for further elaboration.

In general, audiophile attitudes, insights and reflexes honed on analog components are often the main intellectual obstacle to the understanding of digital technology. The rules of the game are suddenly unrecognizable. The analog purist is obsessed with a multitude of subtle details because he knows that the tiny intricacies of the signal are highly vulnerable to any kind of coarseness or inaccuracy in the electrical/mechanical signal path. It becomes difficult to grasp that such concerns are irrelevant to the digital reproduction of music, a process in which the accuracy of detail is an automatic assumption if the system is working properly. That leaves no opportunity for the compulsive little gestures and rituals of perfectionism; any slob can press a button and get just as good results as the nerd with the anti static gun, stylus brush and pocket microscope. There is no poetic justice in the digital domain.

Let us take a quick look at some of the major points that keep coming up in this context.

Quantization.

Many audiophiles seem to be under the impression that a 16-bit linear PCM system is some sort of borderline compromise, as if the number of bits were a figure of merit and 16 a barely adequate number, not as good as 18 or 24 or what have you. That is a simplistic and misperceived view of the subject. The fact is that 16 bits will give you absolute perfection, as long as you are willing to live with a signal-to-noise ratio of 98.1 dB. If you want 110.1 dB, you need 18 bits. If you can be satisfied with 86.0 dB, all you need is 14 bits. It is not a question of good or bad but simply of signal-to-noise ratio. The proof of this involves probability theory and is not within the editorial purview of The Audio Critic, but you can take our word for it. (Over sampling and noise shaping in the D/A converter introduce further complexities but do not negate the validity our basic point.) Needless to say, a signal-to-noise ratio of 98 dB is incomparably better than anything ever dreamed of by the analog purist. The most sophisticated direct-to-disc recording systems, supposedly better than tape, claim 80 dB or thereabouts, and that assumes totally flawless vinyl.

Now, it is perfectly true that all 16-bit D/A converters (at least all those known to us) exhibit some degree of low level nonlinearity, so that the standard 16-bit encoding on CD and DAT may not result in true 16-bit resolution in the playback. One of the few good things ever to come out of the now defunct CBS Technology Center (birthplace of the discredited Copycode system) is the CD-1 test disc, which has among other goodies a track with a series of dithered low-level signals for measuring linearity and distortion. The smallest error we have measured so far at the -90.31 dB level was 3 dB (0.5 bit); a more typical deviation is 6 dB (1 bit); 9 to 12 dB (1.5 to 2 bits) and worse can also be found.

Is that sort of thing audible? Only in a passage recorded at a whisper level and then amplified to the point where a subsequent normal passage would have to be turned way down again, The Yamaha quasi-18-bit conversion system (probably an interim design) and the new Sony true 18-bit linear D/A converters are intended to achieve better low-level linearity, in effect full 16-bit resolution with 18-bit decoding; we hope to be able to report the specific details and exact amount of the improvement in our next issue. In any event, 16-bit A/D encoding is here to stay, and you can relax about it-there is no fly in the ointment.

Sampling frequency.

We get more flak on this subject than any other (see, for example, the last letter in the " Box 392" column in this issue). Many audiophiles have trouble understanding how a sampling frequency (or Nyquist rate) of 44,100 Hz can generate a sufficient number of samples to define signal components up to 22,050 Hz in both magnitude and phase.

At the limiting frequency that is only two samples per cycle; what if they are both taken at the zero crossing, they ask, or both at the peaks, or both somewhere in between? How big or small is the 22,050-Hz signal then, and where does it start and stop? Again, the mathematical proof of the Nyquist-Shannon sampling theorem is outside the scope of this publication, but that does not make it less valid. It may be better not to try to understand it intuitively but simply to accept it as true. We have, however, one little hint and a somewhat trivial demo to ease the pain of the doubters.

The hint: 22,050 Hz in the above case is the absolute limit, in effect the first frequency that cannot be reconstruct ed, because the sampling theorem actually requires that just a little more than two samples per cycle be taken. Let us look at 21,048 Hz, for example, which is represented by 2.095 samples-very little more than two, right?-and can be shown to be totally reconstructible within the system.

How? The demo: 1002.27 Hz is the frequency of the square wave recorded on CD-1 and reproduced in the oscilloscope photo that appears in just about every CD-player review in the slicks. You must have one lying around somewhere.

Take a look. A square wave consists of a fundamental sine wave and all of its odd harmonics. The 21st harmonic of 1002.27 Hz is 21,048 Hz, and you can easily count the 21 little blips over a full period of the square wave, all of them correctly placed. The 23rd harmonic (23,052 Hz), the 25th (25,057 Hz), etc., are not there because they are above the limiting frequency (Nyquist frequency) of 22,050 Hz. That is why the square wave ripples on top and bottom instead of being completely flat. We rest our simple little case.

Thus, a 16-bit system with a sampling frequency of 44.1 kHz is of the most heavenly and incontrovertible perfection (barring vulgar hardware and software foul-ups, of course) if one is willing to accept just two limitations: a signal-to-noise ratio of 98.1 dB and a top-end cutoff of 22.05 kHz. Those whose hearing is so exquisite as to require better figures are out of luck.

Bad vibes and Reed-Solomon.

Does a good CD player have to be as precisely built and acoustically dead as a good turntable? Must the disc itself be mechanically damped? Analog-conditioned purists fervently believe so, needless to say, and the industry caters to their compulsions without embarrassment. Look at the various steel-and-plastic sandwiches, other high-tech plastic and ceramic materials, mechanical stabilizers, isolation feet, etc., among the features touted by the manufacturers, as well as the gizmos you are supposed to put on top of, around the rim of and who knows where else on your CD's.

The underlying assumptions are that some mechanical shock and vibration will be inevitable (true), that the tracking of the laser mechanism will be affected as a result (true), that the disc itself might vibrate to some degree (true), that some of the "pits" in the disc will consequently be skipped or misread and digital errors will ensue (true), and that these errors will degrade the signal to the point where clarity, depth, imaging, etc., will suffer (false).

We recommend the article in the October 1987 issue of High Fidelity by David Ranada (who, incidentally, is the unsung hero of the successful fight against the Copycode), titled "Error-Correction Myths Exploded," which explains among other things the glories of the Reed-Solomon error correction code incorporated in the CD system. The main thrust of Reed-Solomon has little to do with the relatively trivial shock and vibration issue, defects in the disc itself being by far the most important concern; however, the same basic principles apply to digital errors due to any cause. The damping fetishists believe in effect that fewer errors result in better sound because the error-correction system is less busy. David Ranada points out very tellingly the fallacy of that belief, but we also want to report an independent experiment which yielded the same conclusion.

In collaboration with one of our consultants, a proto type CD player (brand irrelevant) was set up in such a way that the error-correction circuit could be closely monitored.

The high-quality Burr-Brown IC in this model has "flags" identified as Reed-Solomon 1, Reed-Solomon 2, Interpolation and Muting, permitting counters to provide readouts of the separate error totals. Now Reed-Solomon 1 and 2 (the distinction between them is unimportant for our purposes) are by definition 100%accurate restorations of the original data, without the slightest loss of information, thanks to the massively redundant recorded data on the disc and the sophisticated mathematics of the error-correction system. At this level of correction, the original binary numbers before the occurrence of error and the corrected numbers are identical. Several dozen CD's were played without any ritualistic precautions, and there was not a single interpolation registering on the counter, let alone muting. Nothing but Reed Solomon 1 (extremely busy) and Reed-Solomon 2 (less busy). In other words, correctable and therefore inaudible digital errors are constantly with us, under all conditions, and errors requiring concealment (i.e., interpolation) are very, very rare. Finally one heavily abused and badly gouged CD tripped the interpolation counter a few times; the effect was completely inaudible even then. Kindly spare us, there fore, the caveats about microscopic disc flutter, chassis damping, etc.; all that is grist for the Reed-Solomon mill.

To avoid causing distress to small firms selling CD damping gadgets in small quantities, we have decided not to bring up specific names, although we have tested such items on our ABX setup and found them meaningless. They are solutions in search of a problem. We also want to add, just in case someone has totally misunderstood what we mean by the shock and vibration issue, that we are very much in favor of the kind of construction in a CD player that provides immunity to minor bumps and jars during play, now fortunately the rule rather than the exception.

The sound.

Speaking of ABX comparisons, they are an absolute necessity when evaluating equipment as similar in design and performance as the CD players discussed below. We simply cannot understand the reviewer who removes, say, the Tandberg player from his reference system, inserts the Denon, listens to the latter at a level that comes naturally and then delivers strong opinions on the big difference in sound. That is an approach left over from the days when big differences actually existed; today it borders on the irresponsible unless the comparison is between components as divergent as loudspeakers or pickups (see also our comments on the same subject in the context of amplifiers). In this particular instance, a complete round robin structured to ABX each of the four CD players one by one against the others would have entailed six series of comparisons; plugging available digital outputs into the D/A converters of the Denon DAP-5500 preamp would have created nine other such series. You can imagine that we did not go through the entire ordeal; we did enough ABX-ing, however, to conclude--without much statistical authority, to be sure-that there exist no readily audible differences within this group of players, plus or minus the DAP-5500. We concede the possibility that many more long-suffering hours of auditions might have uncovered identifiable mini-distinctions of little importance, but we can assure you that not one of these units "blows away" any of the others.

Before you decide to cut off the head of the messenger who brings you this blasphemous news, consider the given facts of the situation. What is the mechanism or process whereby these CD players could sound different? They all have separate left- and right-channel 16-bit D/A converters with 176.4-kHz sampling and digital filtering. The analog low-pass filter at the converter output also appears to be the same in all of them (except the DAP-5500, which is not really a CD player). Three of the players are Philips-based and presumably have identical lasers. In low-level linearity as well as in mechanical construction there are some differences, granted, but of quite unlikely audibility as already explained. The main difference from model to model is in the analog circuits, which are well designed in all of these top-of-the-line units and not terribly critical because of the very low gain. Why, then, should they sound different? By the way, they all sound wonderful. Better, at least to our way of thinking, than any phono system we have heard (pace Sao Win).

Denon DCD-1500 II

Denon America, Inc., P.O. Box 5370, Parsippany, NJ 07054. DCD-1500 II compact disc player, $675.00. Tested sample on loan from manufacturer.

This model represents the top of the line only temporarily, while the DCD-3300 ($1700.00) is being phased out and until a new ultrahigh-end model is introduced. Mean while, in combination with the DAP-5500 digital preamp, it is as high-end as anyone could reasonably ask for and Denon's current recommendation to the audio perfectionist.

Even by itself, without the preamp, the DCD-1500 II leaves very little to be desired. Its front-panel features are at the very least the second best in this group, in terms of control facilities and display information; only the Philips could be argued to be ahead of it and not indisputably. For example, the Denon is the only player we have seen that indicates on the front panel whether or not the de-emphasis circuit is active-now we know which CD's have been recorded with preemphasis. The DCD-1500 II is also one of the two in the group with reassuringly sturdy disc drawers that go "clunk" instead of "click" (the other is again the Philips); this has nothing to do with playback quality but rather with the promise of durability in the same sense as a good car door. The chassis of the Denon is also one of the two most solidly built and heaviest (the other is once again the Philips); immunity to bumps, jars and knocks is good.

Our measurements indicated pretty decent low-level linearity in the D/A converters; interestingly, the small errors in the two Denon models are all positive-in other words, the decoded levels are higher than the encoded levels-whereas in all the Philips-based models the errors are negative. Square-wave reproduction is OK but not perfect; the waveform shows a slight asymmetry reminiscent of previous generations of converters, before the four times oversampling digital filters, but not as pronounced.

The analog low-pass filter may be the reason; there is also a very non-Japanese -45 dB beat tone of 24.1 kHz when a test tone of 20 kHz is being reproduced. No big deal.

On a per-dollar basis, the DCD-1500 II is very hard to beat. It even has variable volume buttons on the remote control (quite rare) and an optical as well as a coaxial digital output. Highly recommended.

Denon DAP-5500

Denon America, Inc., P.O. Box 5370, Parsippany, NJ 07054. DAP-5500 digital audio preamplifier, $1400.00. Tested sample on loan form manufacturer.

There are two entirely separate chassis within this unit, sharing only the line cord and the front panel. Divided between them, with some necessary encroachments, are two quite different audio components. One is a more or less conventional preamp/control unit minus the phono stage. It has a very fine unity-gain buffer output amp (designed without feedback), a switchable 16.5-dB gain stage which shows some transient overload problems, a choice of balanced and unbalanced inputs and outputs, and the usual tape loops.

The other component is a digital processor/interface, which includes 4-times oversampling digital filters, 16-bit "Super Linear" D/A converters--two in push-pull per channel!--followed by analog 7th-order low-pass filters in each channel (yclept "Computer-Analyzed Linear Phase" --CALP that is), plus digital inputs for CD, DAT, DBS (satellite) tuner or any other digital creature that may come down the pike.

The sampling frequency of 32, 44.1 or 48 kHz is automatically selected according to the source. Our favorite deployment of the DAP-5500 is to feed one of its digital inputs from the digital output of a CD player and then take the signal from its DAC Out jack (not the Pre Out) into our line-level amplifier stage of choice. That is the simplest possible signal path, leaving the quality of analog amplification entirely under our control.

In terms of digital quality, the DAP-5500 is outstanding. The Super Linear D/A converters are actually no more linear at the lowest measurable levels than the ones in the DCD-1500 II, maybe even a smidgen less so, but that is linear enough (1-bit error, maximum). The square-wave reconstruction, on the other hand, is of the utmost perfection, and 20 kHz is reproduced without any 24.1-kHz beat tone. That CALP filter really works. This is a solid piece of equipment that creates confidence even if the audible benefits are more elusive than the sheer hardware power.

Euphonic Technology ET650PX

Euphonic Technology, 207 Mountain Road, Wilton, CT 06897. ET650PX compact disc player, $995.00. Tested sample owned by The Audio Critic.

Long before we tested any CD players for review purposes, we selected this one as our interim reference because of its obviously outstanding qualities. We have not regretted our choice. The Euphonic Technology does everything well, even by our latest yardstick. It is a completely reworked version of the Magnavox CDB650, until recently the flagship of the Philips fleet. Michael Goldfield, a dedicated audio perfectionist of the old school, has made major changes in, and additons to, the power supply, the analog circuitry and even the mechanical construction of the stock Philips unit. The innards bristle with new boards, premium-quality capacitors and resistors, toroidal transformer, etc., and the workman ship is beautiful. The only thing that bothers us, just a little, is the flimsy disc drawer common to all the CDB type chassis; the control/program/display facilities, on the other hand, are close behind the latest by Philips or Denon in versatility except that there is no optical output.

In our measurements, the ET650PX beat both Denons and the Tandberg in low-level linearity, equaled the Philips and the Tandberg in square-wave reconstruction (very slight positive-negative asymmetry), and reproduced 20 kHz with the typical 24.1-kHz beat tone. Everything else was perfect.

Philips CD960

N.A.P. Consumer Electronics Corp., I-40 & Straw Plains Pike, Knoxville, TN 37914-1810. CD960 compact disc player, $949.00. Tested sample on loan from manufacturer.

After stamping out excellent but somewhat flimsy Magnavoxes, Sylvanias and Philcos with a cookie cutter, Philips finally decided to market a super deluxe CD player under their own name. It is the most solidly built, most elaborately control- and display-equipped, ergonomically best-designed and generally most reassuring of the models we tested (no Mickey Mouse disc drawer here!); perhaps also the handsomest. Its low-level linearity was the best of them all (0.5 bit maximum error); square waves showed a tiny positive-negative asymmetry (exactly as in the ET and Tandberg); 20 kHz still produced the 24.1-kHz beat tone.

The only worm in the apple is that what we tested was not the current version, as we were suddenly informed at press time. The "improved" CD960 will be sent to us for a follow-up, but we wonder-if it ain't broke, why fix it?

Tandberg TCP 3015A

Ortofon Inc., 122 Dupont Street, Plainview, NY 11803. Tandberg TCP 3015A compact disc player, $1895.00. Tested sample on loan from manufacturer.

At almost twice the price of the Euphonic Technology, the Tandberg (1) is built on the same CDB-type Philips chassis with the dinky disc drawer, (2) has considerably more Spartan controls with a slightly better feel but no decimal keys, (3) gives less display information at a glance, (4) measures exactly the same except for much poorer low level linearity with 2-bit errors, (5) has no digital output of any kind, (6) looks less "commercial" and (7) has different analog circuitry with discrete devices and no feedback. That was enough for the underground reviewers to canonize it, but we think it represents questionable value. Yes, it sounds fantastic but so do the others.

--------

[adapted from TAC, Issue No. 11 Winter/Spring 1988]

---------

Also see:

Cartridge, Arm and Turntable vs. the Groove: Who's Winning? [1977]

Records Recording: At the Leading Edge of the Symphonic Recording Art: the Delos CD's; Engineered by John Eargle