Issue No. 15 Spring through Winter 1990-91

Retail price: $7

In this issue: A new and highly accredited contributor to our journal discusses in depth the current state of CD player technology, from DAC architecture and digital filters to analog stage design, and rates the various approaches. Inspired by the above, your Editor reviews a baker's dozen (count them: 13) CD players and D/A processors. The introductory firstpa rt of our promised expose of the wire/cable scene makes its delayed appearance. A short (but not last) installment of the "Seminar 1989" transcript is shoehorned between timelier matters. Plus our accustomed columns and features, including the return of "Hip Boots" and lots of CD reviews. Issue No. 15 Spring through Winter 1990-91

Editor and Publisher Peter Aczel Contributing Editor David Rich Cartoonist and Illustrator Tom Aczel Business Manager Bodil Aczel

The Audio Critic@ is an advisory service and technical review for consum­ ers of sophisticated audio equipment. The usual delays notwithstanding, it is scheduled to be published at approximately quarterly intervals by The Audio Critic, Inc. Any conclusion, rating, recommendation, criticism, or caveat pubished by The Audio Critic represents the personal findings and judgments of the Editor and the Staff, based only on the equipment avail­ able to their scrutiny and on their knowledge of the subject, and is therefore not offered to the reader as an infallible truth nor as an irreversible opinion applying to all extant and fo rthcoming samples of a particular product. Address all editorial correspondence to The Editor, The Audio Critic, P.O. Box 978, Quakertown, P A 18951. Contents of this issue copyri ght © 1990by The Audio Critic, Inc. All rights reserved under international and Pan-American copyright conventions. Reproduction in whole or in part is prohibited without the prior written permission of the Publisher. Paraphrasing of product reviews for advertising or other commercial purposes is also prohibited without prior written permission. The Audio Critic will use all available means to prevent or prosecute any such unauthorized use of its material or its name.

For subscription information and rates, see inside back cover. Contents 8 The Present State of CD Player Technology: Who Is Doing It Right? By David A. Rich, Ph.D. Senior VLSI Design Engineer, TLSI, Inc. Adjunct Assistant Professor, Polytechnic University 47 Current CD Players and D/A Processors, New and Not So New, Multibit and One-Bit By Peter Aczel, Editor and Publisher 48 Aragon D2A 49 Carver TL-3220 49 Euphonic Technology Mk II Signature 50 Harman/Kardon HD7600 50 JVC XL-Z1010TN 50 Meridian 208 51 Onkyo Integra DX-7500 51 Philips CD-80 52 Pioneer Elite PD-71 53 PS Audio "Digital Link" 53 Sony CDP-608ESD 54 Sony D-555 "Discman" 54 Theta DS Pre Basic 55 Recommendations S6 The Wire and Cable Scene: Facts, Fictions, and Frauds Part I By Peter Aczel, Editor and Publisher S8 Hip Boots Wading through the Mire of Misinformation in the Audio Press 58 John Atkinson in Stereophile 59 Neil Levenson in Fanfare 60 Seminar 1989: Exploring the Current Best Thinking on Audio Part III of the Continuing Transcript 63 Records&Recording CD's from the Golden Age of Audio (Meaning Right Now) By Peter Aczel, Editor and Publisher 63 Beethoven 66 Handel 64 Berlioz 66 Medelssohn/Schubert 64 Bernstein/Barber/Gershwin 66 Piston 64 Bruckner 67 Ravel 65 Dvorak 67 Schubert 65 Elgar 67 Verdi 65 Franck 3 Box 978: Letters to the Editor

1 From the Editor/Publisher: More of That Running Commentary

This is the first issue of The Audio Critic in which the editorial "we" has been dropped. I have decided that it's a little stiffand old-fashioned, not right for the 1990's, especially not in the context of a small, informal publication. (Paul Krasner, radical editor, chiefYippie, and weirdo humorist of the 1960's, had an anecdote about an editor of The New Yorker who used the editorial "we" in the climactic utterance of his most intimate encounters; since then I haven't been truly comfortable with that time-honored usage.) So, from now on, when the speaker is I, the pronoun shall be the same, and the authors of guest articles will also be encouraged to use the first person singular. Another new step in the same direction is the more liberal use of conversational contractions such as "isn't" and "that's." I somehow started offon the wrong foot and became too rigid a formalist on that subject, beginning with Issue No. 10. Maybe it was a delayed reaction to the studied folksiness of Madison Avenue advertising copy, which had been my bread and butter for so many years. Anyway, what I'm doing in this issue reflects my current prefer­ ences in editorial style. I reserve the right to change my mind. * * * All right, all right, you want to know what happened to the promised quarterly publishing schedule and why this issue is again so late. What happened was that I tried my best but just couldn't get the whole act together for 1990; however, I believe I do have it together for 1991. It's mostly a question of staff, contributors, and ready material well in advance. I expect to have a Spring, Summer, Fall, and Winter issue in 1991. Famous last words, you're thinking, so I'll say no more; when it happens you'll believe it. Mean­ while, look at this issue: it has at least twice the editorial content (as distinct from the paper bulk) of the typical high-end audio journal that comes out more often, and since you're paying $22 per four issues, not per year, you come out way ahead and I get paid less often. That, of course, is my greatest incentive to accelerate. As far as up-to-date audio information is concerned, don't imagine that I'm withholding from you a great deal that I already know because of the laggard publishing schedule. I always try to play catch-up at press time and cram recent findings into the issue about to be printed. * * * Let me conclude with a rhetorical question. If audiophile A has read every issue of Stereophile and The Absolute Sound ever published but doesn't know The Audio Critic, and if audiophile B has read every issue of The Audio Critic ever published but doesn't know Stereophile and The Absolute Sound, who has a more solid understanding of the realities of audio, and who is likely to make smarter purchasing decisions, A or B ?

2 Box 978 Letters to the Editor

Have you noticed that certain high-end audio magazines publish only weak, unconvincing letters of criticism and, on the other hand, always the strongest letters of praise? They obviously want you to think that you are getting a scrupulously fair, balanced sampling of reader opinion. But write them a highly articulate, authoritative put-down that makes them look incompetent or insincere, and they will suppress it. This journal is of a different persuasion. We are not impressed by groupie flattery and we welcome intelligent disagreement, no matter how strong. Tr y us. Letters printed here may or may not be excerpted at the discretion of the Editor. Ellipsis (.. .) indicates omission. Address all editorial correspondence to the Editor, The Audio Critic, P.O. Box 978, Quakertown, PA 18951.

The Audio Critic: pologies used in the output stages, as well Why have transformer-coupled pure I read with great interest your review as to the output transformers themselves. cathode fol lowers or totem poles not been of Bob Carver's t-mod "Silver Seven" Let us first consider the matter of out­ commercially produced? Probably because clones. I have seen laboratory data on one put impedance. If, for instance, a unity­ of their high drive requirements, which pair of Silver Sevens-a pair which may coupled output stage, or a pure cathode fol­ simply can't be met with the single-ended have been defective-but I have neither lower or a totem pole (series-connected voltage amplifying circuitry that has al­ measured nor auditioned any of the t-mods, output stage), is employed in conjunction ways prevailed in the vacuum-tube realm. so I am not prepared to comment on Mr. with a correctly designed output transform­ With differential circuitry, totem pole or Carver's success in duplicating the output er, the output impedance can be brought to cathode fo llower outputs are entirely practi­ of his big tube amp. I would, however, like well below one ohm. I hasten to add that cal, and amplifiers using such output stages to address your comments on the wisdom this is not mere theorizing on my part. Uni­ do yield low output impedance and good of Carver attempting to replicate the "tube ty-coupled amplifiers are currently manu­ damping. As you're undoubtedly aware, sound." factured by Nestorovic, EAR, and Vacuum both totem poles and pure cathode fol low­ You state that the output of an audio­ State Research, and were formerly made by ers have been used in commercially pro­ phile tube amplifier is characterized by McIntosh. In all cases, output impedance is duced OTL designs, but that needn't con­ high output impedance (over one ohm), a or was a fraction of an ohm. A pure cath­ cern us here becauseOTLs do not have low bit of second harmonic distortion, and ode-follower output constructed with volt­ output impedances. abrupt rises in distortion at the frequency age-regulator-type tubes such as the 6336, More important, such unorthodox out­ extremes. These observations are very gen­ 7242, or 6AS7 might practically achieve an put stages make possible extremely wide­ erally correct but are not universally true. output impedance of a quarter of an ohm or band power delivery without the rising dis­ That they should apply at all is a conse­ even a tenth of an ohm. I can cite no pro­ tortion at frequency extremes which you quence of the design deficiencies in most duction examples of this type, but a cath­ mention. Ordinary plate-coupled tube am­ currently produced tube power amps rather ode-follower amplifier is available on a plifiers require primary impedance loads in than of the inherent limitations of vacuum­ special-order basis from Vacuum State Re­ the thousands of ohms and thus critically tubecircuits themselves. search. A transformer-coupled totem pole high turns ratios to achieve good matching Most of the performance limitations in output would also yield an extremely low to conventional loudspeakers. Even at low tube amplifiers made today are attributable output impedance, though here again a power, power bandwidth is inevitably limit­ to the singl e-ended circuitry used in the commercial application appears never to ed by shunt capacitance and leakage induc­ main gain stage and the plate-coupled to- have been attempted. tance in the transformer, but the problems

3 become critical as output approaches and gain stages class A and use judicious unless you count quasi-feed forward schemes exceeds the 100-watt mark. It is worth not­ amounts of feedback, and the small amount like Stasis operation, or Quad's "current ing that until the seventies nobody manu­ of third harmonic will be just about elimi­ dumping." In any case, even with fe edfor­ factured a plate-coupled consumer tube nated also. In fact, the active circuitry of a ward, clipping will still produce lots of amplifier of over 100 watts output. The fully differential all-triode tube amp may high-order distortion, and there are other only consumer tube amps in the 200-300- be designed to produce practically no THD subtler aspects of transistor behavior which watt range were the big Macs, which were at all, and with values of global feedback resist the linearizing effects of feedforward, all unity-coupled (half in the cathode cir­ many, many dB less than would be re­ class A operation, diffe rential operation, or cuit, half in the plate). quired to achieve similar results with a any other linearizing technique I know. With an output stage having a low out­ solid-state design. Interestingly, the fully Transistors have more than a dozen distor­ put impedance before transformation, a differential all-triode Morikawa amp from tion mechanisms--far more than triodes­ transformer can be designed that has ex­ Japan has THD in the 0.01 % range with a and these include thermal debiasing, collec­ tremely low distortion and phase shift in mere 6 dB of global fe edback, and most of tor/base capacitance, etc. In toto, these are the audio band, and extremely extended that distortion is probably due to the output productive of not only high inherent values frequency response. If, for instance, your transformer. of THD, but also complex intermodulation primary impedance is 100 ohms, you can Of course, one may object-why both­ distortions which are not adequately repre­ get a secondary impedance of 0.1 ohm with er with tubes when vanishingly low THD sented by simple two-tone tests. Richard a very low turns ratio. Indeed, if the amp is values can be achieved with transistors? In Bell, currently of Carillon Technologies, designed with split supplies, an autoformer other words, what particular merit is there wrote a series of seminal papers on the sub­ can be utilized in lieu of a conventional to tube design today? ject back in the sixties, and his observations transformer for an al most total elimination I think you can gather by now that I still hold good today. of shunt capacitance and leakage induc­ still favor vacuum tubes, and that being the Vacuum tubes are also far better iso­ tance. I won't belabor the point. A tube out­ case, I wish I could make an irrefutable lated from electrical disturbances than are put stage with an inherently low output technical case for their superiority and for transistors. They are less disturbed by either impedance working into a well-designed the impossibility of a solid-state amp per­ RF or reactive loads, and can be operated transformer will give transistor amps a pret­ fectly replicating the transfer function of a in amplifying circuits without regulation, ty good run for the money in terms of properly designed tube amp. I'm afraid I something that just isn't possiblewith tran­ damping and bandwidth. can't settle the matter, but I can state a sistors. Transistors want a rigidly control led I am not implying that those who de­ number of reasons why a tube amplifier operating environment, and really superior sign plate-coupled amplifiers are fools. Plate should outperform a solid-state type in re­ solid-state amps such as the Threshold Sta­ coupling makes for simpler, cheaper de­ producing music. sis have extraordinary amounts of house­ signs and a much wider choice of output The better audio-frequency triodes keeping circuitry to keep the signal devices tubes. And it can work extremely well in produce negl igible distortion above the in linear operating modes. low-powered units. But today's audiophile third harmonic even at full clip. Bipolar I am not suggesting here that solid­ wants high power, and I am convinced that transistors produce a relatively unfavorable state amplifierdesign is an exercise in futi l­ plate coupling is not the way to go. And harmonic spectrum, especially at clipping, ity or that tube amps are the only valid I'm in pretty good company. Frank McIn­ and heavy use of global feedback exacer­ choice for the music lover. I am only sug­ tosh fel t the same way. bates the problem by eliminating low-order gesting that the ancient technology of the I might mention in passing that the distortion while leaving higher-order prod­ vacuum tube still has something to offer bandwidth of a pl ate-coupled tube amplifier ucts. (MOSFETs are a special case; they and can deliver a level of performance that can be considerably improved by using a have a somewhat tube-like distortion spec­ is very hard to emulate. pair of output transformers in parallel, each trum, but total distortion is quite high, and By the way, if Carver is indeed pro­ optimized for one half of the frequency they're hard to drive, especially in multi­ ducing a preponderance of second harmon­ range. It's an old trick, and I believe that ples.) Cl ass A operation doesli nearize tran­ ic in his t-mods, it might interest your read­ Carver uses it on the Silver Seven. Double sistors considerably in this respect but ers if you were to explain how he achieves transformers are also used on the new Cary doesn't completely solve the problem, inas­ this result. It would certainly interest me. I amps and were used on the now discontin­ much as class A transistor amps derate have seen distortion spectra from Iitera ll y ued Meitner tube amps. I don't think the when operated into low-impedance and/or scores of high-end solid-state amps, and scheme works as well as a pure cathode fo l­ highly reactive loads. (One can, of course, with one exception-a Mark Levinson lower, and it's certainly not cost-effective. provide the amp with output impedance ad­ product-I have yet to see a solid-state amp You mention distortion. Not all tubes justments, an interesting technique used in with a tube-like spectrum. I know that Car­ produce a preponderance of second har­ the class A solid-state amps made by Sony ver is using neither class A operation nor monics. Most triodes do, but pentodes tend Esprit and Sphynx, a Dutch manufacturer.) feedforward, so I am curious. to produce a more transistor-like distortion The technique of feedforward also tends to I must conclude an already overly spectrum with plenty of high-order nasties make a transistor output stage emulate the long letter. Thank you for producing a live­ (pentodes were regularly denounced as out­ performance of triodes, and may well rep­ ly publication and for discussing technical put devices in 1940's consumer publica­ resent the optimal output topology possible issues generally not covered elsewhere in tions). Now, if an all-triode power amp-­ with today's devices, but it still fa ils to the consumer press. the preferred design in my opinion-is achieve a perfect replication of tubecharac­ Very sincerely, made fu lly di fferential, most of that second teristics, and it is very expensive to imple­ Daniel Sweeney harmonic will be bucked out. Operate the ment properly and not terribly practical- Burbank, CA

4 Yes, as a letter it's overly long, but as and reason after reading some of the other sion in Issue No. 13, it was clear to me that a minitutorial on pro-tube philosophy it's high-end publications. The others seem to some individuals could get the impression about the right length, and I think our read­ be promoting a mysticism designed to ap­ that all amplifiers sound the same. As a ers will appreciate it. Although you have peal to those with a fear of technology. I consequence, they could purchase any am­ obviously been recruited by the vacuum­ am afraid this is a very expensive cult. plifier they wanted based on factors other tube lobby (not necessarily commercially, I hope to see The Audio Critic grow than sound quality. Consequently most au­ just ideologically), your arguments appear and prosper, but please do not become diophiles would tend to buy the cheapest to me to range from at least plausible to glossy and flashy like the others. amplifier within the appropriate range of basically valid. On a fe w points I and my Please find my check and subscri ption specifications, or the one that had the most associates part company with you totally, form enclosed. desirable bells and whistles. This is exactly as future reviews in this journal will un­ Rod Hickerson what a large majority of amplifier manufac­ doubtedly evidence. Th e biggest hole I can Portland, OR turers would want. Then they could con­ punch in your exegesis is that you carefully centrate on marketing, distribution and avoid mentioning the heat generated by As of the last couple of years, the po­ advertising, the things they do best and the tubes, their fa r from unlimited life span, larization along the reason/mysticism axis tactics that lead to the most consistently and their changing performance character­ is even worse than you think. I come from high sales figures. I speak with firsthand istics as they age. A tube amplifier is not an an era when audiophiles believed what the knowledge on this point. You see, I once install-it-and-forget-it type of audio compo­ professionals told them. Today their heads believed that all amplifiers were the same, a nent, whereas a properly designed solid­ are filled with garbage fe d to them by self­ notion that was reinforced by your maga­ state amplifier is-and to me that overrides appointed pundits with no credentials, who zine, among others. For example, I was all other considerations. I'm willing to con­ tell them to distrust the professionals. Th e talking to a Carver representative a month cede that a superior solid-state amp will witch doctors are elbowing aside the real ago, who told me that all amplifiers sound possibly be more complex than an equally doctors. Unless a strong countertrend the same and he could prove it using the good tube amp, from which it fo llows that a emerges, I fe ar fo r the future of audio. This ABX comparator. Your magazine's en­ semipro or dilettante designer/constructor publication alone can't hold the line. dorsement of ABX double-blind testing, will findthe tube approach more congenial. No, we won't become another slick coupled with your claim that any amplifier I think you make too much of the sec­ magazine, but we must run color ads on working within certain parameters sounds ond-harmonic issue. The Carver "Silver coated stock. It's a matter of survival in a the same, perpetuates that misconception. Seven " doesn't have such a heavily second­ highly competitive situation. Your response to my letter did little to clear harmonic-dominated signature, or any oth­ -Ed. that up. The fact remains that the real­ er peculiarity in its distortion spectrum, world audiophile (or the nonaudiophile that there should be any problem copying The Audio Critic: consumer) cannot make the Radio Shack its transfe r function into a solid-state am­ Dear Peter, amplifier sound like the Boulder 500 under plifier. Ye ars ago, Bob Carver sent me at Thank you for printing my letter in Is­ any real-world conditions. Yet there are my request a memorandum detailing the sue No. 14 of The Audio Critic. many audiophiles who will read your mag­ successive steps in his t-mod procedure. If I may comment on your response to azine and think that all amplifiers sound the Th ere are far too many steps, and too many my letter: 'The point of my letter is and was same. Thus they will be easy prey to their details within each step, to repeat them to get the word out to audiophiles who sub­ desire to save money and buy the cheaper here. Maybe one day I can prevail on him scribe to your magazine that all amplifiers amp. Moreover, they will be tricked by the to write an article. Th e thing to remem­ do not sound the same! I carry that message manufacturer or dealer who would rather ber-and what many audiophiles somehow to anyone who is interested in getting the sell a lot of cheap amps for a large profit fa il to understand�s that you can't t-mod best out of his system. I assume that the than a few expensive ones for a small prof­ a sow's ear to duplicate a silk purse; in majority of your readers have that as a it. The ABX comparison is an intellectual other words, the amplifier being modified goal . I have read every issue of your maga­ exercise with no real-world application. If has to be potentially as good as, or better zine Uournal?) since its inception. I contin­ the Silver Seven-t sounds exactly like the than, the amplifier being copied. The basi­ ually get the feel ing that your readers are Silver Seven, then the Silver Seven is a cally sound circuit topology and high volt­ being misled into thinking that all ampli­ consumer and intellectual fraud. I do not age/current capability of Bob's amplifiers fierssound the same. As a matter of fact, I know if you were aware of it when Issue give him a lot of latitude. think that anyone who has been exposed to No. 14 went to press, but the Silver Seven -Ed. the ABX debate could be misled. Let me is currently being sold as a real -world prod­ say now that I have no problem with the uct through Lyric Hi-Fi (see Bob's inter­ The Audio Critic: ABXper se. I do not have any use for dou­ view in the Fe bruary 1990 Stereophile) at I am extremely pleased with the back ble-blind testing. I know that I am intellec­ the rate of about ten per month. issues of your publication I received recent­ tually honest and have no motive to use Having said that, let me respond to ly. They are well written and cover the some clue (cue) other than the merit of the your criticism of my letter point by point. finest equipment with an absolute mini­ component to identify it. I do, however, use You state that my use of the terms mum of filler and nonsense. blind testing whenever practicable. (Yes, I "frankly gorgeous" versus "splitting head­ Your assessments regarding the rela­ can guess what component is playing by ache" hardly constitutes expert testimony. tive effect of amplifiers, CD players, loud­ identifying external cues, but I am honest Excuse me! You are the expert! That is speakers, interconnects, room and software enough to admit it.) what I pay you for! I am just a humble au­ on sound quality are a breath of fresh air After reading the roundtable discus- diophile depending on you for advice. If I

5 sound like an amateur, it is because I am. are absolutely right, Peter, I did know that ers, not fo r those in need of a remedial Try this. Hard, brittle, dry high frequencies you needed an extremely high-quality amp reading class. What I print is what I mean, and overall dry sound versus a euphonic, for the lab and, more importantly, for your exactly as you read it, nothing more and lush sound. More learned audiophiles at­ listening pleasure. (Although the Boulder nothing less. Yo ur simplistic and subjec­ tribute this to different orders of harmonic would not have been my choice.) What do I tively distorted restatement of what I wrote, distortions in tube amps than in transistor need in my home to create the illusion of or what the seminar participants said, is amps. But then you knew that, didn't you, live music? Is your need any greater than not something I should have to deal with. Peter? (Incidentally, in Issue No. 14, an ex­ mine, or for that matter than that of the That said, let me comment on some of pert like yourself describes the sound of thousands of other audiophiles who desper­ your rhetorical maneuvers, counselor. You various components using terms that I ately seek to create the illusion of live mu­ say you are intellectually honest, but I find would not consider expert terminology-I sic in the home? I think I see it now, Peter: certain indications to the contrary in the assume you meant terminology instead of you are not a hypocrite, you are a snob! above letter. For example, you say you testimony--e.g., page 13, paragraph 2, What you are really saying is that all have read every issue of The Audio Critic " ... The square-pulse response of the Plati­ things being equal (and you concede that in since its inception, but then why are you num II is gorgeous ..."; page 13, paragraph the amplifierworld they never are), you can unaware that my present position on elec­ 3, " ... Platinum II's sounded sweet, smooth, make two amps sound the same. For that tronic soundalikes is quite recent, not even open and uncolored ..."; page 15, paragraph matter, I could compete in the ring with fully evident in Issue No. 1O? As for your 2, " ...snappy or open ..."; page 16, para­ Mike Tyson if you equalized (pun intend­ out-of-context quotations of my descriptive graph 7, " ...too many flavors... "; page 55, ed) our strength, size, physical skills, adjectives and phrases, maybe a jury of paragraph 2, " ...sweeter ...") knowledge of and experience in boxing, trade-school dropouts from the District of Peter, those double standards certainly and desire to win. But it just does not hap­ Columbia would fa ll fo r such a shabby are convenient, aren't they? pen that way in the real world. courtroom trick, but not our readers. Of As I have already said, I have nothing No, you did not say the Radio Shack course I use those basic and very useful against the ABX comparator per se. I prefer amp will sound like the Boulder under all words-in conjunction with other words single-blind testing of the unaltered compo­ conditions, but you implied it. I'm glad you presenting objectively verifiable data. Com­ nents, since that is what the real-world au­ cleared that up. The only way to get the pare: "Please describe the young woman diophile takes home. "Today there exists sound of the Boulder 500 is to buy the you saw walking away from the scene of no halfway respectable opponent of ABX Boulder 500. That is why you purchased the crime. " "Quite frankly, she was gor­ testing who attacks the box itself." I think I the Boulder instead of any of the cheaper geous. " As against: "She was a Caucasian can guess who it is you are attacking. I brunette, about 24, app roximately 5 fo ot amps available to you, including the Car­ 7, guess you have your own definition of re­ vers! very shapely with long legs, wearing a tight spectable. My definition of respectable is a I intended this letter to be provocative blue dress and little makeup. Quite frankly, manufacturer of musical components that and to spark debate. Sending it to Stereo­ she was gorgeous. " Yes, counselor, I meant represent good value, and the reviewers phile and The Absolute Sound increased the and still mean testimony, credible testimo­ and dealers who lead me to it. I assure you chances that it would be printed. Indeed, it ny. The terminology can vary, depending there are plenty who question not only the was printed in The Absolute Sound. In an­ on what is being said. ABX box but the motives of those who en­ swer to your question, I often comment on You also seem to be innocent of the dorse it. While we are on the subject, I pending legislation that affects my clients commercial fa cts of life in audio. Th ere is wonder why no one talks about the fact that and, yes, I do send copies to all involved. more money, for the manufacturer and fo r AlB testing has no real scientific value. The Thank you for considering my point of the dealer, in selling a $6000 amplifier only question is, does the system as a view. than there is in ten $600 amplifiers. Th e whole do an acceptable job of recreating Reginald G. Addison entire cost/wholesale/retail structure is fa r the illusion of live music at a price you can Attorney at Law more stringent in the lower price ranges afford . AlB testing usually is just a gim­ Washington, DC than at the high end, where the figures can mick. Do I purchase component A because CC: Th e Absolute Sound be arbitrarily set at whatever level the it comes close to the real thing or because it Stereophile traffic will bear. As for the Silver Seven is better than B? More in line with" our dis­ "fraud, " Bob Carver tells the consumer cussion here is that the ABX really mea­ Yo ur point of view? You have no point right up front that the "t-mod" sounds ex­ sures sonic memory. Can you remember of view, except fo r a knee-jerk aversion to actly the same, aft er which the high-end what you just heard well enough to match moderately priced audio equipment and an tube freak is on his own, free to part with what is now being played and then com­ obvious fo ndness fo r the sound of you own $17,500 or not. Does that constitute fraud? pare it to something you are now listening voice as you keep repeating yourself I Should there be a Surgeon General's warn­ to? flatly refuse to defend what I "implied" (in ing on the chassis to the effect that this I knew you would refuse my offer be­ your opinion!), or what "some individuals product is dangerous to your pocketbook cause amplifiers do sound different, and could get the impression " of or whatever and offe rs no benefits beyond what is ob­ you know it. You proved it with your audio gave you "the feeling" you elaborate on. tainable fo r $2000? The Silver Seven was dollars. The fact is that, even assuming ar­ Th e defense would, in any event, consist of originally intended to be a tongue-in-cheek guendo that you could get the Radio Shack a literal quotation of what I wrote, and that engineering exercise; it was Harry Pearson amp to sound like the Boulder, the condi­ you already have in your possession. I said and Mike Kay who took it seriously and tions that you were able to do it under it all then and there. The Audio Critic is made it into a business. would have no real-world application. You written fo r attentive, comprehending read- Your cavalier dismissal of double-blind

6 testing has me in a quandary. The world's constant-directivity loudspeakers have been Such attacks are not in keeping with leading authorities in psychophysics con­ getting good results for only about twenty your stated mission and philosophy. Con­ sider it to be absolutely essential in order years. Most horn systems are still bad tinue to present your data and opinions in to obtain valid results, not only in audio enough to lead cri tical listeners to the con­ ways that readers can enjoy them most, and but in dozens of other areas of investiga­ clusion that the concept of constant direc­ try to ignore those aspects of your competi­ tion; the world's leading medical research­ tivity is wrong. The concept that's wrong, tion which reflect poorly on all of us (the ers insist on it fo r testing new drugs; but however, is the belief that it's enough for a audiophile community-the sensible part). Reginald the Lawyer says it has no value. loudspeaker to be flat on axis. Yours truly, Now I don't know whom to believe. Sincerely, Rob Bertrando All this is more than I really intended William J. Roberts Reno, NV to write in response to an inconsequential, Toledo, OH hassling letter. Please do not reply again. I have two, and only two, reasons fo r Be happy with your two soapbox opportu­ What you say is absolutely correct, castigating another audio publication. One nities so fa r. but what you fo rget is that a re cording that is that they attacked me, or The Audio -Ed. was balanced by the producer over cOnl'en­ Critic, first. The other is that they are tional loudspeakers will in most cases spreading major disinformation on the sub­ The Audio Critic: sound fa r too bright when played through ject of audio. Stereophile qualifies on both Your long efforts to define the perfect constant-directivity speakers. Unless some­ counts. In 1988, they tried to destroy my loudspeaker give me the feeling that you're body says, "One, two, three, go!" and the credibility via their "Letters " column and trying to prolong the suspense. Let me sug­ entire audio community changes to con­ did not publish my perfectly civil but highly gest a concept. A loudspeaker should have stant-directivity sp eakers at the same time, embarrassing corrective reply. Since then the same frequency response in all di rec­ this problem will remain with us. Further­ they have been taking potshots at me and tions. An equivalent statement is that a more, not all constant-directivity loud­ my publication at every opportunity. Are loudspeaker should have the same di rec­ speakers sound the same, either. Yeah, it's you asking me to turn the other cheek? tional pattern at all frequencies. This char­ a lot easier to design amplifiers. (The first and only persuasive advocate of acteristic is called constant directivity. -Ed. that policy ended up in a situation I would As described in Benchmark Papers in find unacceptable.) As for disinformation, Acoustics, the loudspeaker was still a re­ The Audio Critic: Stereophile is now the most influential dis­ cent invention when the best minds in the Although your publication continues seminator of audio myths, fa ntasies, and audio business were called upon to design at a generally high level of quality, your fe tishes, the chief apologist for the school systems fo r movie theaters. Obviously, flat snide remarks about other publications is of "A blows away B because my exquisite frequency response was required over the inappropriate and insulting to your readers. ears say so and I don't have to prove it. " band to be covered. The designers quickly Just because Th e Absolute Sound has re­ To me that constitutes an irresponsible realized that flat frequency response with­ sorted to such a style in the past doesn't journalistic butt just begging to be kicked. out constant directivity was inadequate, be­ mean you should emulate it; in fact, I no When accountability becomes the rule rath­ cause there would then be some direction longer read that magazine, in part because er than the exception within the high-end in which the frequency response would not of that (and I'm sure I'm not alone in this). audio community, the prevailing intramu­ be flat. Floyd Toole rediscovered this just a The comment about Stereophile's re­ ral tone will surely be less confrontational. few years ago. The microphone people views of noncone-type loudspeakers being I agree with you 100% that the Carver have always known that flat frequency re­ suspect is wrong, for two reasons. First, the "Amazing Loudspeaker" should not have sponse in all directions was the ideal char­ mere fact that Carver is able to modify his been released before the altitude sensitivity acteristic. Having lots of tricks available, ribbon to make it less altitude sensitive problem was fixed, as it is now. Bob Carver they've beenable to come pretty close for a means (to me) that this should have been a himself admits it was a serious oversight. variety of patterns. design parameter from the start; there are Th at, however, doesn't make Dick Olsher's For two reasons, it is not natural for a enough audiophiles who live at altitudes amateurish and often downright silly loud­ conventional loudspeaker to give flat fre­ high enough to incite altitude sickness that speaker reviews more respectable. The man quency response with constant directivity. they should not be ignored. Second, like is an undisciplined seat-of-the-pants dab­ First, a conventional loudspeaker becomes nearly all audiophiles, Stereophile's writers bler in a technology he only partially understands. a carefully designedloud­ less efficient at high frequencies where the prefer various noncone speaker systems, Yes, radiation resistance stops rising, so that and in general the same ones audiophiles speaker should work properly at all inhabit­ equalization is required to maintain con­ prefer at sea level. If in fact they don't like able altitudes, but no one will convince me stant power output. Second, the beam width some systems at their elevation which seem that electroacoustic measurements made at tends to increase at low frequencies where to sound better at sea level, I would not 7000 fe et above sea level are good, reliable the diaphragm becomes small relative to consider that speaker to be a reasonable practice. I never saw any qualifications or the wavelength, so that a horn is typically recommendation. Imagine the audiophile caveats in Stereophile's test reports. required to restrict the angular coverage. who, shortly after his audition and purchase John Atkinson's obfuscatory editorial During the thirties, for reasons of cost, of Carver's Amazing Loudspeaker at a sea­ of May 1990 on this altitude controversy is most designers of loudspeakers for the level dealer, gets it to his home in the addressed elsewhere in this issue (see "Hip home abandoned correct design forever. mountains and finds himself (rather) disap­ Boots"). The few people still trying to design pointed with the sound. -Ed.

7 The Present State of CD Player Technology: Who Is Doing It Right?

By David A. Rich, Ph.D. Senior VLSI Design Enginecr, TLSI, Inc. Adjunct Assistant Professor, Polytechnic University

This is an attempt to clarify at one fe ll swoop all the diffuse bits and pieces of information that keep cropping up in the audio press on the subject of CD playback and to put that whole technology in a critical engineering perspective.

Editor's Note:Dr. David Rich is the new 1. We start with an analysis of the most sample of a DAC are correctable by Contributing Editor on our masthead important component of a CD player, the modifying the circuit design or layout of and a very serious classical music afi cio­ digital-to-analog converter. the chip. Distortion worsens as signal nado on top of his fo rmidable electrical level decreases, and fe edback cannot be engineering credentials. His app roach to Digital-to-Analog Converters used to linearize the DAC. One research­ audio is a little more technical than what [Burr-Brown 1989], [Tex. Instr. 1989] er has fo und that very small linearity er­ our readers are accustomed to; for that The digital-to-analog converter rors in DACs can "produce audible mod­ reason some of his more esoteric digres­ (DAC) has the greatest effect on the ulation noise and extremely noticeable sions are broken out in sidebars to sepa­ sound of a CD player. The DAC accepts audio distortion" [Fielder 1989]. rate them from the main body of his arti­ digitally coded data and produces an ana­ The step height of a DAC is the dif­ cle. You may want to pass over these on log output in the fo rm of currents and ference at the DAC's output between ad­ first reading and return to them later­ voltages (see Figure 1). jacent steps in the transfer curve of the but please, do return. Linearity is the principal perfor­ DAC (see Figure 1). A perfectly linear * * * mance specfication for a DAC. Linearity DAC has steps of equal step height, as Warning: The author will not-repeat, is not a new concept in audio. In analog shown in Figure 1. Note that the step will not--take telephone calls from our systems, it is the deviation from a linear height has been normalized to 1 unit. readers at the company where he is em­ transfer function, which gives rise to har­ This is the smallest analog step at the ployed or at the university where he monic and intermodulation distortion DAC's output. An LSB step occurs when teaches. You can reach him by mai� how­ [Borbely 1989]. The deviation from lin­ the Least Significant Bit (the last bit of ever, in care of this publication. earity in analog systems is usually well an n-bit digital word) is changed while * ... * characterized. For example, bipolar de­ leaving other bits constant. The resolu­ Introduction vices have an exponential transfer char­ tion of a DAC is the number of digits This article explores the design of a acteristic. In analog amplifying devices, necessary to express the total number of modern CD player, offering insights into the distortion increases with increasing steps. For example, a 16-bit DAC has the design trade-offs of midpriced and signal amplitude. (The crossover distor­ 65,536 steps. There are many definitions high-end players. Armed with this tion in class B output stages is an excep­ of linearity error in a DAC. The most knowledge, you will be in a better posi­ tion.) The deviation from linearity of the common to characterize the performance tion to distinguish differences between amplifying device is reduced in almost of a DAC are integral linearity (also CD players. all designs by global feedback. Care called end point linearity or linearity) The best source of information must be taken in applying feedback to and differential linearity. Integral linear­ about an electronic product is its service prevent the formation of dynamic distor­ ity is defined as the difference between manual. Service manuals were consulted tion products [Otala 1974]. the actual step value and the nominal extensively in preparing this article. For A DAC's deviation from linearity step value, as shown in Figure 2. (The small American companies that do not differs from those characteristic of ana­ actual step values must be corrected for publish schematics or service manuals, log systems. The deviation, generally, is offset and gain errors where absolute DC marketing brochures and interviews with stochastic, randomly varying from one voltage levels are required to be main­ designers were the primary sources of in­ sample of the converter to the next. Sys­ tained. These errors are not important in formation. Data is summarized in Table tematic linearity errors occurring in each audio applications.) The maximum lin-

8 Analog Output Value

Ide.1 Straight line 5

4 \./"11',? Step Height I � '-'u'" 11 lSBI - 3 II': ..(: ., " 2 II' " / Step Value

Digital o Input 000 0 .. 001 0 .. 010 0 ... 01 0 .. 100 0 Code O .. ... 101 \ , Step � CONVERSION CODE -

----<--- ' r-- Digital Input Code a . 000 O . 001 a . 010 :0.011: 0.. 100 0 101 , , Analog Output Value a 2 3 4 5 1 :... _ _.J: , ...... Figure 1 Analog OUlput Value ILSBI

6

5

4 AtStepOll • -r/ I· 'I. lSBI �'-

Analog End,Point Lin./ Error Output 2 Value / ILSBI At Step 001 I· 'I. lSBI 6 0 ��r--1-�--+--+--+--4 000 001 010 011 100 101 110 111 5 Digilal Input Code

4 Figure 2 Differential linearity Error I '/J lSBI

3 ,. / , I \ I 2

Differential linearily E"ol I . 'I. LSBI Digital o Input Code a 0.. 001 0 ..011 0. 101

Figure 3

9 earity error is given in the DAC's data 100). It is not a single monolithic chip. a wide dynamic range [Dijkmans and sheet. Linearity errors are often ex­ Rather, it is a hybrid circuit incorporat­ Naus 1989]. The test uses a 400 Hz sine pressed as multiples or submultiples of 1 ing numerous state-of-the-art custom­ wave recorded at -80 dB in combination LSB. Differential linearity is defined as designed chips. The price of the with a .03 Hz sine wave at -20d B. This the difference between the actual step DAC729, a product of hybrid technolo­ test will expose differential linearity er­ height and the ideal value of 1 LSB (see gy, is 10 times greater than the maximum rors that are not found using single-tone Figure 3). If a DAC has a differential Iin­ price a DAC in a high-end CD player THO measurements. earity error of greater than 1 LSB, then would cost. If a DAC does not have a low glitch the transfer function can be nonmonoton­ The $12,000 Stax DAC-Xlt digital­ energy, or if it does not settle within a ic, i.e., the output of the DAC can de­ to-analog processor uses a similar hybrid small percentage of the sampling period, crease even when the value of the digital circuit, the DAC 020400 manufactured it must be followed by a sample-and­ code is increased. A resolution of 20 bits by UltraAnalog. The UltraAnalog circuit hold circuit. The sample-and-hold sam­ for a DAC is feasible, though the accura­ guarantees 18-bit differential linearity, ples an analog input signal value and cy of the DAC is a function of the linear­ but no specifications are supplied for in­ then holds the instantaneous input value ity errors, and the DAC may be accurate tegral linearity or settling time to a 16-bit upon the command of a digital control to only some smaller number of bits. If LSB. The diffe rential linearity specifica­ signal. Figure 5a is an idealized sample­ the 16-bit DAC has a maximum diffe ren­ tion holds only at room temperature. The and-hold circuit. In the sample mode, the tial and integral linearity error of ±2 Stax unit uses tubes in its output stage; switch opens and the capacitor stores the LSB, it is no more accurate than a 15-bit thus the DACs may not perform to specs value of the input voltage at the point the DAC with a maximum differential and because of elevated operating tempera­ switch opens. The circuit samples the integral linearity error of ±1 LSB.In oth­ tures. The distortion specifications of the output of the DAC after the converter has er words, the accuracy of a DAC, not its Stax unit show that the very low distor­ nearly settled to its final value. This val­ resolution (the figure quoted by the mar­ tion at the output of the UltraAnalog ue is then held on the capacitor when the keting departments of CD player manu­ DAC is compromised by the tube output next data word is presented to the DAC. facturers), determines how linearly the circuit's nonlinearities. [Ha-ha!-Ed.] Figure 5b shows a simplified circuit dia­ signal will be reproduced. Although typical specifications for gram of a sample-and-hold. In the sam­ Figure 4 shows the dynamic re­ differential and integral linearity are giv­ ple mode, the circuit acts as a unity-gain sponse of a DAC to a step change in the en for a consumer DAC, these ratings are inverting amplifier. In the hold mode, the digital input code. The glitch is a short, not guaranteed. Instead, a set of THO capacitor holds the value of the output at undesirable transient in the analog output values is specified. In this test, a se­ the time the switch is opened. As will be following a code change at the digital in­ quence of digital words which represent discussed, the implementation of this put. The glitch area, the time integral of sine waves of different amplitudes is particular circuit has its problems. Ulti­ the analog value of the glitch transient, transmitted to the DAC, and THO at the mately, it is not possible to build a cost­ should be as small as possible in DACs output is measured. Quoting from [Burr­ effective sample-and-hold that does not used in audio applications that do not in­ Brown 1989], '�HO is the measurement distort the input signal. corporate a deglitching circuit. The set­ of the magnitude and distribution of lin­ All of the current DACs designed tling time of a DAC (tsd in Figure 4) is earity error, diffe rential linearity error, for use in high-end CD players operate the total time required for the analog out­ noise and quantization error. [Distortion, without a sample-and-hold circuit. These put to settle within an error band around attributable to quantization error, can be include the Philips TDA1541A (16-bit its final value after a change in the digital eliminated if a dither is added to the sine resolution), the Burr-Brown PCM56P input. The error band is usually ±1 LSB wave {Lipshitz and Vanderkooy 1988}.] (16-bit resolution), the Burr-Brown wide. However, settling characteristics to There is a correlation between the THO PCM58P and PCM61P (18-bit resolu­ wider error bands are important if the and the square root of the sum of the tion), the Analog Devices AD1856 (16- DAC is to function without a sample­ squares of the linearity errors at each dig­ bit resolution) and AD I860 (18-bit reso­ and-hold circuit. The value of settling ital word of interest. However, this does lution). The PCM56P was revised so it time varies with the magnitude of the not mean that the worst-case linearity er­ can be used without a sample-and-hold. change in the digital word value. The ror of the O/A is directly correlated to Older CD players that used this chip had conversion period is the time between the THO." The OAC with the lowest a sample-and-hold circuit. (The high­ successive digital codes being applied to guaranteed THO levels is the UltraAna­ priced Burr-Brown DAC729KH, unfor­ the DAC. The conversion period should log DAC 020400 hybrid. Since the THO tunately, does require a sample-and-hold equal or exceed the settling time. The of a OAC can be difficult to measure be­ circuit. The UltraAnalog DAC 020400 number of words presented to the DAC cause of the low absolute value of the includes a sample-and-hold as part of the in one second is called the word rate. distortion products, an alternate, simpler hybrid circuit. The differential linearity The word rate is a rec iprocal of the con­ test is often used to assess DAC linearity. and THO specs for the 020400 include version period. This test is called gain linearity. Gain the sample-and-hold circuit.) The precur­ linearity is the measurement of the devia­ sor to the PCM58P, the PCM64P, was The Best DACs tion of the amplitude of the sine wave's the first 18-bit resolution DAC. It re­ Only the Burr-Brown DAC729KH fundamental component from the ideal quired a sample-and-hold circuit for digital-to-analog converter has a guaran­ value, for sine wave signals of varying proper operation. I compared the sound teed 16-bit differential linearity error of amplitude ranging from fu ll scale to be­ of the Pioneer PD-91 and Sony COP- one bit (it can be externally adjusted to Iow an LSB.This is commonly called the 707ESD, which used the PCM64, to that 1/4 LSB) and an integral linearity error linearity test, and the errors are reported of their respective successors, the Pio­ of 1/2 LSB for 16-bit resolution. The in LSBs. This test, widely used by audio neer PD-71 and Sony COP-X7ESO. The DAC729 can also settle within 1/2 of a magazine reviewers, should not be con­ latter are similar but not identical to the 16-bit LSB when it is sampled at a 4x in­ fused with the more stringent integral older models, with the principal differ­ terpolation rate. Unfortunately, the DAC- and differential linearity tests. Philips re­ ence that they incorporate the PCM58P. 729, which is not designed for consumer searchers have developed a test signal Although my listening comparison was audio, sells for $197 (in quantities of which explores differential linearity over not double-blind or even single-blind, 10 Analog Output

t�., -----""""'i

t" . --_"'i Specilied Error Band I � I I

Glitch

, 1\ I '_ Ft',

1", fDigital!,."., ","" Settling J Time S !Digitall Slew Rate IOigitafl Delay Time OMO 'tid - I �,. Anatog Settling Time

Figure 4

Sample

Sample �/ ------�---- Input Input I Output I Output Figure 5a Figure5b

11 Figure 6a Figure6b

� ---- r---- ST ------VDD- 2 I l I ¢M I (400Fs or more) I I I WCO weo. SHL SDSY SHL Bel SHR generation SHR � ___�l i\B '!: _

---- I:� r-� ------�� 16 - 18 I Timing circuit I Coefficient ROM Micro-program I Beo BCI I _.r_--, I I BCI I I I DLO I I & I VSS I Multiplier Accumulator p I Shifter I � s �Q) VDD l I > ::;0. �I I o I , ::; I o DRO I I I I I SDI Temporary RAM I I I I I l L_ ------_J �------J Figure 7

12 nor at exactly matched levels, my im­ es, and they do not match between dies. claim show the PCM61P-K to be more pression was that the sound of the new Thus, only the small systematic nonlin­ linear than the PCM58P-K; however, units is significantly more open and less earities will be canceled. This approach they are now using the ADI860-K. Some "electronic." I attribute this difference to is very expensive because each channel manufacturers use lower selection grades the elimination of the sample-and-hold is serviced by a pair of DACs. A single, of DACs, asserting that the selection pro­ stage. highly linear DAC will outperform two cess is done in-house. This is hardly con­ One potential problem in removing less linear DACs wired in a push-pull vincing, since the binning process en­ the sample-and-hold circuit is the feed­ configuration. Hence, the push-pull to­ sures that the lowest-grade DACs will through of digital clock signals to the an­ pology should be used only with the have the poorest performance; all DACs alog output of the converter. Digital sig­ highest-grade DACs ava ilable. with superior performance have already nals typically have a peak-to-peak It is not possible to fabricate the in­ been removed. (An exception to this rule amplitude of 5 V. These signals enter the ternal circuit components in a DAC to occurs if the number of top-grade DACs DAC to encode the next digital word into match closely enough to achieve 16-bit produced exceeds the demand for them. the DAC and they determine when the accuracy. Burr-Brown and Analog De­ When this happens, some of the top­ conversion process begins. Normally, vices adopt a technique called laser trim­ grade DACs are marked with a lower these signals would not be running dur­ ming. A laser adjusts the value of the grade). ing the period that the output of the DAC critical resistors on the chip during the The Burr-Brown PCM1700P and is sampled. When the DAC's output is initial testing of the silicon wafer. After Analog Devices AD1864 contain two continuously connected to the input of this test, the wafer is split into individual DACs on one silicon chip. This allows the analog circuit, a change in a digital chips, which are then placed in plastic the use of a single chip for stereo appli­ signal's value may slightly affect the an­ packages. The packaging of a die can ef­ cations. The Analog Devices part comes alog output. This is probably a third­ fect its performance through exposure to in blank and J grades, and its perfor­ order effect, though some designers of mechanical stress. Final packaged parts mance is identical to that of other Analog high-end CD players minimize it by con­ are retested. Not all of these parts per­ Devices parts of the same grade. The trolling the rate of rise and fall of the dig­ form equally. Devices with the best THD PCM1700P yields performance that is ital signals connected to the DAC. They performance are separated. These DACs slightly poorer relative to the PCM58P also align the transition time of all the are then sold at different price points, de­ for a given part grade. (The comparisons digital signals that enter the DAC. pending on their respective THD perfor­ are complicated by the fact that the THD Denon, JVC, Technics, and Yamaha mances. DACs are generally classified levels of the PCM58P are specified at a add external circuitry to increase the res­ into one of three grades. Suffixes are at­ diffe rent sampling rate than those of the olution of digital-to-analog conversion tached to the part number to indicate its PCM1700P.) two bits beyond the resolution of the quality. The architecture of a practical DAC DACs they are using. High glitch energy causes the worst differential linearity er­ is one problem with these systems. Con­ Rankingthe DACs ror to occur at the most significant bit sequently, a sample-and-hold circuit is A current ranking of DAC accuracy, (MSB). The MSB is the largest incre­ required. The Yamaha CDX-1120 elimi­ in ascending order of improved perfor­ mental output change obtained by nates the sample-and-hold with a new mance, is as fo llows. switching a single input bit. This is un­ low-glitch implementation of their Category 1: AD1856 AD 1860 fortunate because the MSB change oc­ floating-bit DAC. In addition, the prob­ PCM56P PCM61P curs when the output of the DAC passes lem of matching the components added Category 2: AD1856-J AD I860-J through zero. For small-amplitude sine to the DACs in these systems tends to PCM56P-J PCM61P-J waves, the differential linearity of the limit the accuracy of these CD players. Category 3: PCM58P MSB can have a significant effect on sig­ These systems, I believe, are incapable Category 4: AD1856-K AD1860-K nal distortion. Figure 6a shows a sine of matching the accuracy of the best PCM56P-K wave when it is reproduced by a DAC available monolithic DACs. I would Category 5: PCM58P-J with large positive differential linearity greatly hesitate to purchase a CD player Category 6: PCM61P-K error at the MSB. Figure 6b shows a with external circuitry for the purpose of Category 7: PCM58P-K DAC with a large negative linearity er­ enhancing DAC resolution. The PCM58P and PCM61P are ror. The DAC of Figure 6b is not mono­ Denon (in the DAP-2500 digital au­ guaranteed to meet their distortion per­ tonic. To reduce distortion, the Analog dio preamplifier), Kinergetics (in the formance specifications without a de­ Devices and Burr-Brown DACs allow an KCD-40), and Yamaha (in the CDX- glitcher. In my opinion, only the DACs external trim adjustment that trims the 1120) use a novel approach for reducing in the last four categories should be used differential linearity error at the MSB the nonlinearity of a DAC, in which a in a midpriced or high-end CD player. change close to zero. The PCM58P al­ pair of DACs are wired in a push-pull The PCM56P-K and AD1856-K, albeit lows adjustment of bits two through four configuration. One DAC in the pair re­ 16-bit DACs, offe r better linearity than in addition to the MSB. Designers debate ceives digital information which has some of the 18-bit resolution devices. whether these adjustments offer addition­ been modified so that the polarity of the These 16-bit DACs are cheaper than the al sonic performance improvements. The signal entering the DAC will be inverted. less accurate 18-bit models, though mar­ accurate adjustment of these potentiome­ The analog current-to-voltage converter keting considerations curtail their use. ters is difficult in a production environ­ takes the difference between the respec­ One of the leading DAC designers calls ment, and independent laboratory tests tive current outputs of the two DACs. the extra two bits "marketing bits." For a confirm that many units are shipped with Matched even-order nonlinearities that CD player with a four-figure price tag, I the adjustments incorrectly performed appear at the output of the two DACs are would consider only a PCM58P-K or a [Lipshitz and Vanderkooy 1988]. [See then cancelled. This approach is success­ PCM61P-K. Engineers at Madrigal, The­ also the CD playback equipment reviews fully adopted in analog circuits charac­ ta, and CAL have selected the PCM61P­ in this issue.-Ed.] terized by closely matched even-order K device for their machine, claiming that The Philips TDA1541 uses a differ­ nonlinearities. Most DAC nonlinearities, it offers better sonic performance. Theta ent architecture than the Burr-Brown and however, are caused by random process- has performed measurements which they the Analog Devices DACs. The architec- 13 ture applies a proprietary technique Burr- Brown and Analog Devices are tion of the new DACs is 20 bits. (The called Dynamic Element Ma tching. The continuing the development of audio higher resolution also satisfies the mar­ architecture allows the DAC to achieve DACs. For this reason, you should make keting department.) It takes approximate­ 16-bit linearity without laser trimming or sure that any very expensive CD player ly 6 to 12 months after the introduction external adjustments. The elimination of or decoder you are contemplating to pur­ of a component before it begins to ap­ an external adjustment is especially ad­ chase can be upgraded to the newer pear in a commercial product. Pioneer is vantageous since a trim pot can change DACs when they become available. The the first company to announce the use of with age. The top-of-the-line TDA1541A most recent DACs from Burr-Brown and the PCM63P. The new Pioneer CD Sl offers first-class performance, but it is Analog Devices are the PCM63P and players are the PD-73 and PD-93. These not possible to rank this DAC with the AD1862 respectively. The PCM63P uses units are expected to become available American DACs above because the guar­ a new topology which steps away from by the time this article is in print. [That's anteed specifications for the American zero in small steps in both directions to sufficient lead time.-Ed.] DACs and the Philips are different. (The reduce low-level nonlinearity. The American manufacturers guarantee dis­ AD1862 uses a digital offset technique Digital Filters and Interpolators tortion and Philips guarantees differential which shifts the zero level away from the [Lipshitz and Vanderkooy 1988] nonlinearity.) In the CD players tested MSB transition. The PCM63P-K and All quality CD players now place a for review in this issue, the best DACs AD1862N-J chips have better linearity digital filter and interpolator (the term from Burr-Brown were found to have performance than the PCM58P-K. The oversampling should be reserved for an converter sampling at a rate much lower distortion at -90 dB than the Phil­ PCM63P-K data sheet lists slightly better AID ips TDA1541A Sl. Until the new Ameri­ THD specifications at the -20 dB and fa ster than the Nyquist rate) ahead of the can DACs that did not require a sample­ -<50 dB levels in comparison with the DAC. A digital filter affords significant and-hold circuit made their appearance, AD1862N-J. (Again, the comparisons reductions in the complexity of the ana­ the TDA1541A S1 was used almost ex­ are complicated by the fact that the THD log filter. With a 4x interpolation rate, clusively in all high-end CD players. levels of the PCM63P are specified at a the analog stage could be formed with Manufacturers such as Sony, CAL, and diffe rent sampling rate than those of the only two active gain stages. With an 8x Kinergetics have now chosen alternative AD1862. In addition, the THD for the interpolation rate, a single active gain chips from Burr-Brown and Analog De­ AD1862N-J is an A-weighted measure­ stage can be used. A well-designed digi­ vices. Moreover, newer companies enter­ ment.) The lower grades of these DACs tal filter should introduce virtually no ing the field (e.g., Krell, PS Audio, Ara­ are not as linear as the PCM58P-K. The signal distortion. This is in contrast to a gon, and Proceed) use the Burr-Brown THD level of the UltraAnalog DAC high-order analog filter, an analog circuit devices. Philips remains the preference D20400 hybrid is 6 dB lower than that of which, owing to unavoidable nonlineari­ of European companies. the PCM63P-K at -90 dB. The resolu- ties in the active devices, can introduce

Inside the Digital Filter croprogram, which is stored on a ROM in­ has no advantage, provided analog dither The topology in which a digital filter ternal to the timing circuits, controls the op­ was added during the recording process. is implemented is a highly specialized mi­ eration of the filter. Adding dither is standard practice in mod­ crocontroller called a digital signal proces­ The word length of the coefficient ern recordings [Lipshitz and Vanderkooy sor (DSP). A block diagram of a digital ROM partially determines the accuracy of 1988]. fil ter, the Yamaha YM3434, is shown in the filter response. The effect on passband The bus of data connected from the Figure 7. The circuit blocks which consti­ response is not important, e.g., the NPC arithmetic unit and the temporary memory tute the DSP "engine" are all inside the SM5805, which has a short 16-bit word, is is called the data path. The word length of smaller dashed rectangle. Other blocks fiat ±0.OO025 dB. The added word size has this path is another parameter which affects within this rectangle (BCO generation, PIS, a more important effect on the stopband re­ the filter's performance. The data path output temporary buffer) and those outside jection. The Sony CXD1144BP, which has word length is usually the word length of it are specialized digi tal circui ts for format­ 293 taps and a 22-bit coefficient word the DAC, though it may be larger if noise ting data from the CD player's disc-reading length, has a stopband rejection exceeding shaping is performed. The marketing de­ circuitry. These circuits also make the data 120 dB. When two digital words are multi­ partments have recently taken notice of the available to the filter's DSP engine and plied, the resultant word length at the out­ word length of the coefficients, accumula­ send the output of the latter to the DAC. As put of the multiplier is the sum of the input tor output, and data path. They are using for the DSP engine itself, it functions as word lengths. This word is too large to use these in advertising copy, perhaps with follows: The coefficient ROM stores the and must be shortened (requantized). The hopes that readers will confuse these larger digital words which control the filter's process of shortening introduces quantiza­ numbers with the resolution of the DAC. shape. The multiplier/accumulator per­ tion distortion [Dijkmans 1989]. Some digi­ The optional noise shaper can round fo rms the arithmetic operations required for tal fil ters truncate the word. This is a less the data at the accumulator. Normally, the the filter. The accumulator stores partial desirable process than a rounding opera­ noise power is constant from DC to half the and complete computations from the multi­ tion. Lipshitz observes that, in addition to sampling frequency. A noise shaping filter plier. The shifter manipulates the digital rounding, a di ther signal must be added is an IIR filter, a filter with an infinite im­ words during the multiplication process. during the multiplication process to ensure pulse response, which redistributes the The temporary RAM block is required to that all quantization artifacts are removed. quantization noise shape. A noise shaper store the output of the accumulator because No current monolithic DSP chips use dith­ reduces the noise power in the audio band the processing of the cascaded filter blocks er. (Dither has been used in certain Theta and increases it outside the audio band. The is performed in parallel, and the data and Wadia digital decoders, but given the signal-to-noise ratio of the signal bandlim­ emerging from the accumulator is not the constantly changing filter algorithms used ited in the audio band increases. Noise data for the next computation. The ROM, by these companies, it is unclear if dither is shapers exhibit low-level instabilities called RAM, and the arithmetic unit are con­ used in current production models.) Note limit cycle oscillations. Proper rounding op­ trolled by the timing circuit block. The mi- that adding dither at the input of the DAC erations and use of dither prevent this, as

14 significant distortion and frequency­ measured by the number of delay blocks the-art product is to design with a gener­ response variations. An engineering in the filter and is called the tap length of al purpose DSP chip, external RAM, trade-off must be made between the re­ the FIR filter. (There are N + 1 taps in a ROM, and a number of smaller glue duction in the complexity of the analog filter with N delay blocks.) The compari­ chips. The glue chips, as the name im­ stage and the increase in nonlinearities at son of the tap sizes of different digital plies, form the interfaces between the the DAC output. This is because the lin­ filters is valid only if the filters have the other chips in the system. Because of the earity of the DAC can be degraded when same interpolation rate. A properly de­ large computational requirements, more the conversion period of the DAC is re­ signed filter's passband flatness is, re­ than one DSP chip may be required (one duced. A DAC operated at an 8x interpo­ markably, less than ±0.0001 dB. This is for each channel, for example). The cost lation rate will have half the conversion insignificant compared with the frequen­ of this group of chips is much larger than period of one operated at a 4x rate. cy response errors in the analog chain. that of a single monolithic device. The Digital filters often use a set of cas­ A designer of CD players who wish­ Motorola DSP56000 is proving to be the caded, finite impulse response (FIR) es either to improve upon off-the-shelf most effective chip for this application. filters. The sampling rate of each filter filter chip designs or to incorporate addi­ The high prices of the Theta ($2000 to section is increased relative to the section tional ideas of his!her own will need $45(0), the Krell ($3500 to $8950), and which precedes it by a power of 2. The plenty of money. A design team of engi­ the Wadia ($1995 to $7995) reflect the bulk of the filtering takes place in the neers would require at least a year, ow­ cost of implementing the digital filter first section, since this section operates at ing to the chip's complexity, to develop with general-purpose DSP chips. Some the slowest clock rate and is therefore and fa bricate such a chip, which may of these units run at much higher interpo­ easier to design. Finite impulse response need upwards of 100,000 transistors. The lation rates (16x to 64x) than the mono­ filters, which are difficult to design in the cost of engineering time and materials lithic filters. The manufacturers of these continuous time (analog) domain, have would be in the $200,000 to $500,000 units claim that the monolithic chips do the significant advantage of being Iinear­ range. This investment is well out of the not perform the interpolation function ac­ phase if the coefficients are chosen prop­ reach of small American audio compa­ curately. In the monolithic chips, an ideal erly. Because of the linear phase charac­ nies. Because a custom chip, fabricated brick-wall filter, which is required by the teristic, such a filter exhibits better time for a single company, would be used in sampling theorem for the exact recon­ domain response to a pulse than an ana­ much smaller quantities than a standard struction of the input data, is approximat­ log filter. product, the cost per chip would be much ed by the FIR filter. A brick-wall filter The smoothness of the passband, the higher than that of off-the-shelf digital has a sin x/x impulse response. The time slope of the transition band, and the at­ filter chips. domain form of the sampling theorem tenuation of the stopband are determined The only solution for a small Ameri­ states that when a sin x/x function is con­ by the size of the FIR filter. The size is can audio company designing a state-of- volved with samples of a bandlimited in-

apparently does dither added in the record­ test tones. The extra bits available from an relative to the CXD1144. The lower cost of ing process. Noise shaping is used on two 18-bit DAC could provide the headroom, the CXD1244 also supports this notion. 18-bit digital filter chips, the NPC SM5803 though this is not done on current mono­ The CXD1144 is often cited by circuit de­ and Sony CXD I244. The noise shaping can lithic filters. The designers of these filters signers as the best-sounding single-chip be turned on and off under software con­ prefer to use the extra bi ts to reduce the digi tal filter. It should be noted that none of trol . Therefore, a service manual may not quantization noise introduced in the trunca­ the diffe rences in passband or stopband show whether a given CD player is using tion process at the output of the accu­ characteristics given in Table 3 should be noise shaping. mulator. audible. Therefore, it is unclear how the in­ Interpolating digital filters are also Currently, there are only four manu­ creased complexity of the CXD1144 results plagued with potential overload. This over­ facturers of monolithic digital filter chips: in better sonic performance. Most Philips load arises because signal amplitude at the NPC, Sony, Yamaha, and Philips. The TDAI541A's are used in tandem with the output of the filter can be greater than that prices of the chips are dependent on the Philips SAA 7220P!B digital filter, although allowed by the word length of the filter. complexity of the DSP section. A filter some Sony digital filters have functional The amplitude increases because of the with more taps, longer coefficient ROM modes which make them compatible with Gibbs phenomenon [McGillen and Cooper words, longer data path, or a larger accu­ the TDA1541A at a 4x interpolation rate. 1974], which occurs when a signal is band­ mulator will be costlier. The CXD 1244 chip allows the repro­ limited and all its Fourier coefficients are The Sony CXD1144 18 by 8 filter is duction of very small signals without not present. An example of the Gibbs phe­ the most complex chip to date, and it is switching the MSB bit because it can apply nomenon is seen in test reports on CD priced at double the competing NPC a small DC offset to the digital code emerg­ players as an oscillation on the top and bot­ SM5813 (the similar SM5803 adds a noise ing from the filter. This feature is unique to tom portions of a square wave. The prob­ shaper and other features that do not impact the CXD I244. Very small signals are re­ lem is worsened by an increased filter inter­ on the performance of the DSP core) and produced with lower levels of distortion. polation rate. Lipshitz calculates that two Yamaha YM3414. Sony's newest design, The DC offset would cause the positive bits of headroom are required in a 4x inter­ the CXD1244, has not as yet been adopted peak of large-am'plitude signals to exceed polating filter. Attenuation of the input sig­ by any manufacturer other than Sony. The the maximum digital word size of the filter, nal to the digital filter will solve the prob­ performance of each of the 8x interpolating thereby clipping the positive peak of the lem, but attenuating the input penalizes the filters is summarized in Table 3. Sony has signal. To avoid this, all signals entering signal-to-quantization-noise ratio of the not quoted the multiplier size, coefficient the digi tal filter are slightly attenuated in filter beyond acceptable levels. Hence, word length, or filter tap length in its data amplitude. most monolithic filters detect the presence sheet of the new CXD 1244. The ripple re­ The Philips digital filters have a very of an overload and allow the filter to clip. It jection of the CXD1244 and its stopband similar feature but differ from the Sony is unlikely that the filter will clip in the rejection are slightly inferior to those of the CXD1244 inasmuch as the DC offset can­ presence of music signals as distinct from CXD1144, indicating a simplified design not be defeated.

15 • • Original Saund Regenerated - -- -X x Waveform

Figure 8

Latch ,ignal IWCKO) Latch signal from digiti' filter. a to Of A COnlo'eft8f . S.4672MH. Q-_....3--i t--.L.--i::> CK 8.4672MH.

X'tal 16.9J44MH.

Latch signa' Irom digital lilter. f8.4672MHz)

Masterclock puis•. f16.9344MHz)

Latch signa' to DIA conwrter. @ f8.4672MHz)

Figure 9

16 4. VSS2 S.XI 5.XO

13.S't'NC

15.L/R

IS.WC r------, {'lock Synch. pul5e selection ClOCk ,e:neralion I D I . A "" veo 19.DIGR I �� I L-.---r--r---r--r-T-' I loop I I I filter 3. vco 20 . DIGL I I PLL I --� : �r- 9.¢B I I 0 L ______� 12.BCO

Parity check I LlIIl,h synchroni7.a lion

2a . DIN

17.00

2t .ERR 10. TI D" .mol de,..,,;o. �

It.n

I.VDDl o-J �ub·code OUlput

o 22. SEL 24 .S2 23 .S1 16.DEF 26. SSYNC 25.SCK 27.S00 Figure 10

N x Cs .= 44 1 kHz 16 .... N x C. N x Cs Serial Djital filter .... --... 1-4 4 Optional One-bit digital ..... a linear ... Loopfilter ... Quantizer �...... PWM ,1 output data interpolator .,16 DAC �l1_4

Figure 11

17 put signal, the bandlimited input signal system rings less than a standard digital Wadia has implemented the glue cir­ will be reconstructed exactly [Papoulis brick-wall filter. The amount of ringing cuitry in a programmable gate array man­ 1980]. The impulse response of the FIR in the impulse response is directly related ufactured by Xilinx. The interconnection filter is finite andof the sin x/x function by the Fourier transform to the stopband of the circuitry is controlled by software is infinite. The coefficients in the filter rejection of the filter. The lack of ringing programmed into a programmable ROM are slightly modified to account for this in the tails of the digital impulse re­ (PROM). The circuitry for the SPDIF de­ [Papoulis 1980]. Coefficient modifica­ sponse curve for the Wadia system is a coder (see below) is also implemented in tion routines are well-known and give direct result of the poor stopband rejec­ the programmable gate array. Wadi a can good results. Wadia, Krell, and Theta tion. The sampling theorem requires that thus correct errors or upgrade the circuit claim that the method used by the mono­ the signal at the input of an analog-to­ configuration of the decoder box by lithic chip manufacturers is less than digital converter must be bandlimited to changing the code in the PROM. ideal. In the case of the Theta, the one half the sampling frequency. An ana­ coefficients are also adapted, depending log or digital brick-wall filtermust be in­ Phase Jitter and SPDIF on signal conditions. The Wadia and the cluded in the digital tape recorder to sat­ A sampled data system's perfor­ Krell perform the interpolation function isfy the sampling theorem. Thus, the mance is critically dependent on the ac­ directly in the time domain rather than impulse response of a complete digital curacy of the sampling time interval. the frequency domain. As none of these audio system (analog in to analog out) Variation in the absolute timing of suc­ companies have published their current will be that of a brick-wall filter regard­ cessive spacings is called phase jitter (or algorithms in the open literature, it is im­ less of the response of the playback sys­ jitter). A sampled sine wave signal cor­ possible to assess their methods. Wadia tem to a digitally generated unit impulse. rupted with phase jitter will, when exam­ has published information on an earlier [Touche again.-Ed.J Martin Colloms ined on a spectrum analyzer, appear as a time-domain algorithm using Lagrangian also fo und that the frequency response of narrow Gaussian-shaped band of signals. interpolation [Moses 1987]. The perfor­ the Wadia 1000 and the Krell units was The resulting signal spectrum is very mance criterion for a digital fllter that down by 3dB at 20 kHz. This indicates similar to that of an analog signal repro­ performs interpolation and smoothing is that the Wadia and Krell algorithms have duced from a turntable or tape deck with the Mean Square Estimation Error not been optimized for maximum pass­ flutter. The effect of jitter on the time­ (MSEE) [Papoulis 1984]. None of the band flatness. domain plot of a sine wave is shown in manufacturers have published any data In a recent piece of promotional lit­ Figure 8. The crystal oscillator which showing that their filters have a lower erature, Wadia implies that the results of generates the clock signal is generally jit­ MSEE than a low-cost monolithic filter. the sampling theorem cannot be applied ter-free. Time-base jitter can arise if the [Touche!-Ed.} to music signals. They argue that in the power supply signals to the crystal oscil­ Time-domain interpolation algo­ derivation of the sampling theorem the lator become noisy. Kenwood proposes rithms have the major disadvantage of Fourier series is used. They claim the that clock noise induced by the CD track­ not rejecting the out-of-band spurious Fourier series cannot be used to represent ing system gives rise to time-base jitter. signals with as great attenuation as a FIR the stochastic music signals. This state­ According to Kenwood, this explains the filter [Cezanne 1988]. Martin Colioms ment is completely fa lse. It has been claim that CD ri ngs and disc stabilizers has fo und this problem in his measure­ shown that the sampling theorem is change the sound of a CD. ments of the Wadia 1000 and the Krell equally valid for bandlimited random The obvious solution is to ensure units [Colloms 1989], [Colloms 1990]. signals [Papoulis 1984]. One hopes that that the supply to the crystal oscillator is The Wadia sales literature points out that the Wadia copywriters misunderstood well-regulated. A problem is that the the digital impulse response of the Wadia the Wadia engineers. crystal oscillator circuit is often incorpo-

Inside the SPDIF Circuitry often not possible to optimize the loop filter in the decoder box, run at slightly different A block diagram of the industry stan­ of the PLL to acquire the SPDIF signal frequencies. Resyncing is possible only if dard YM3623B SPDIF receiver chip is quickly and also produce a clock that is low the two signals have the same frequency, shown in Figure 10. The SPDIF signal en­ in ji tter. Generating a stable clock signal in but they may still have different phases. To ters the DIN pin. The various digital data the outboard box (where the quality of the equilibrate the frequencies, an elastic store and clock signals recovered from the clock signal is important) and using an ad­ which accepts data at one rate and reclocks SPDIF signal appear on the output pins of ditional signal line to send the clock to the it at another would be required. If the CD the chip. An analog circuit, a phase-locked CD player-this clock replacing the inter­ player runs faster than the decoder box, loop (highlighted in Figure 10), implements nal clock of the CD player-would be a then data accumulates in the elastic store, this clock recovery. The clock recovered by better solution for connecting an outboard since data enters at a faster rate than it the phase-locked loop (PLL) will contain decoder box to a CD player. Sony's former leaves. If the CD ' player runs slower than more jitter than the crystal-oscillator­ top-of-the-line CDP-R1 and DAS-R1 com­ the decoder box, the elastic store fills its generated clock signal in the CD player. bination adopts this approach. This is sur­ memory with data before placing data on One major source ofjit ter is the volt­ prising, given that Sony established the its output line. Data could be read out at the age-controlled oscillator (V CO) in the SPDIF fo rmat. Unfortunately (for the rest faster clock rate fr om the data in the elastic PLL. The amount of phase jitter in the of us), all other CD players generate only store. As the CD plays, the amount of data VCO output signal is dependent on the an SPDIF output and will not accept an ex­ in the elastic store's memory decreases as VCO's circuit topology. The phase jitter of ternal clock signal. data is removed fa ster than it is replaced. A the VCO can appear at the clock output of The clock generated from the VCO large, uneconomical elastic store would be the SPDIF decoder [Gardner 1979]. The cannot be resynced with a clean clock gen­ required because of the abundance of data type of phase detector (marked phase dif­ erated from a second crystal oscillator in on a CD and the variation in crystal fre­ ference detection circuit in Figure 10) used the decoder box(as was done in the NC, quencies in a CD player. (Technics has in the PLL also strongly impacts upon the Kenwood, and Sony CD players). The two very recently announced that they are phase jitter at the output [Fourre 1989]. It is crystals, one in the CD player and the other working on a practical implementation of

18 rated in the digital filter IC. Noise on in­ may be reintroduced by the filter IC. I 59, which sells for 20 cents per foot. You ternal chip power supply lines cannot be think these manufacturers may be attack­ can bet that the snake-oil manufacturers removed. The solution to this problem is ing a second-order problem while leav­ will be introducing SPDIF interconnects to build the crystal oscillator as a sepa­ ing other major design flaws in their at $100 apiece. rate circuit and drive the digital filter players unresolved. For example, the One final observation. I found the with the output of this circuit. This ap­ Sony CDP-508ESD uses an unselected PS Audio "Digital Link," which uses proach is used by Stax, Sony, and CAL, Burr-Brown PCM58P DAC. It would only the Yamaha YM3623B chip, to have among others. Another source of time­ have been preferable to eliminate the an excellent sound quality when used base jitter occurs when the crystal oscil­ CXD8003 from the 508 and channel the with a high-quality transport such as the lator signal is divided down to the word cost savings towards an upgraded DAC. one in the Philips CD-80 or the Pioneer rate of the DAC in the digital filter chip. The problem of jitter is more signifi­ PD-71. This indicates that the jitter level The source of the jitter is again power cant when the SPDIF (Sony-Philips Digi­ of the YM3623B is sufficiently low so supply noise. Resyncing (realigning) the tal Interfa ce Format) signal is employed. that it does not significantly affect the edges of the clock waveform at the out­ This is the data format used in connect­ sound quality of modem decoder boxes put of the digital filter with the master ing a CD player to an external digital de­ that use it. The early Philips and Sony clock will reduce this timing jitter mech­ coder box [Rumsey 1989]. The SPDIF decoder boxes had less sophisticated anism. Sony, in its CDP-508/608/X7ESD uses a single cable to transmit data re­ SPDIF decoders and may have more jit­ players, performs the resyncing with a covered from the CD player. The data is ter. The Ya maha CX-1000U preamplifier/ custom IC, the CXD8003. Kenwood uses specially encoded so that the clock signal decoder incorporates several clever cir­ a similar chip called DPAC (Digital can be recovered from the data. (See cuits placed around their YM3623B chip Pulse Axis Control) in its DP-8010 sidebar for technical details.) to reduce jitter even further. Ya maha is player. The basic DPAC circuit is shown Some subjective reviewers have cit­ trying to keep these circuits for them­ in Figure 9. The system in Figure 9 al­ ed differences in the sound of the output selves. They are not discussed in the lows the latch signal to the DAC to of a decoder box when driven with dif­ YM3623B chip data sheet. New SPDIF change only on the rising edge of the ferent transports. If the differences are decoder chips are currently being intro­ master clock. The practical implementa­ real-and that remains to be proved with duced by Philips (SAA7274P) and Crys­ tion of the DPAC circuit is significantly properly conducted ABX tests-then jit­ tal Semiconductor (8411). I have not re­ more complex than the circuit shown in ter is the culprit. Some high-end manu­ ceived data sheets on these chips and Figure 9. facturers propose expensive transports thus cannot comment on their perfor­ JVC has found [JVC 1989] that ad­ with low jitter. This is of little conse­ mance relative to the Yamaha. ditional sources of jitter include noise quence, since the jitter problem can only coupled into the clock line from adjacent be completely eliminated in the decoder "Bitstream"D/A Conversion signal lines by mutual inductance and box. Wadia and Theta claim that the The ability to integrate highly com­ mutual capacitance. In addition, signal bandwidth of the optical cable system plex digital systems using low-cost reflections in the interconnection lines used for the SPDIF fo rmat is inadequate. CMOS IC technology is one of the im­ between ICs can cause jitter. The K2 In­ This can give rise to jitter. Coaxial cable portant advances in this decade. Early in terface is a functional block developed designed to transmit wide-bandwidth the development of fine-line CMOS tech­ by JVC to suppress jitter. The K2 Inter­ digital signals is recommended instead. nology, research focused on more eco­ face combines optocouplers and a data Unlike audio cables, these coaxial cables nomical implementations of ADCs and resyncing circuit. The K2 Interface is will transmit the SPDIF signals precisely. DACs. Architectures were chosen that placed before the digital filter; some jitter An example of such coaxial cable is RG- could take advantage of the cheap digital such a system [Willenswaard 1990]. Over uses an FLL. The RockLok will only work istics of each PLL in the system are dif­ 1.5 megabits of memory is required. If the with CD players having a clock that devi­ ferent. memory overflows, the unit switches to a ates a maximum of :1:.75 ppm from the nom­ The Krell, Theta, and Wadi a SPDIF PLL clock decoder. Technics will use the inal data rate. (The Technics system also decoders appear to be innovative designs. technology in the SH-XlOOO decoder box. requires an accurate CD player master Aragon uses a similar decoder, designed Technics of America has chosen not to im­ clock, :1:.50 ppm, to function properly.) Wa­ for Aragon by Theta, in its D2A product. port this unit. Technics of America refused dia reports RockLok provides a 2500:1 jit­ Owing to the difficulty of designing a to give me any information about the new ter reduction over a conventional PLL­ SPDIF decoder that has low jitter, these technology or to explain why the SH­ based clock recovery system. companies were less willing to talk about XlOOO is not being imported.) The amount Multiple phase-locked loops are an al­ the design techniques they used in the of memory can be reduced by slowly adapt­ ternative. The first PLL is designed to cap­ SPDIF decoder than in any other part of ing the frequency of the crystal oscillator in ture the data free from error, without gener­ their design. Only Theta was willing to the decoder box once the CD player starts ating a jitter-free clock. The second PLL is give figures for the peakji tter amplitude of sending data. This approach is often called designed to attenuate the jitter present on their SPDIF decoder. Theta reports a peak a frequency-locked loop (FLL). the clock line of the first PLL. The second jitter of less than 1 nanosecond. This Ins Designing a crystal oscillator that is PLL often uses a VCO which incorporates peak jitter is still 2.5 times larger than the both tunable and jitter-free is a difficult de­ a tunable crystal. This approach is adopted 400 ps peak jitter amplitude necessary to sign problem. Cheap CD players often use by JVC, Kenwood, Nackamichi, and San­ ensure that the full 16-bit performance of master oscillators which may not be very sui among others. the DAC is realized [Harris 1989], [Fourre accurate. The jitter performance of a crystal JVC strangely uses in the second PLL 1989]. Technics reports that their new osci llator would be compromised if it were a VCO (74LS624) which has jitter levels SPDIF decoder achieves a 500 ps jitter am­ required to have enough tuning range to op­ comparable to the VCO in the first PLL. plitude--but, as I said, Technics of Ameri­ erate with these cheap CD players. The The NC system does offer some jitter· at­ ca will not be importing this state-of-the-art Wadia "RockLok" clock recovery circuit tenuation because the loop tilter character- system into this country.

19 technology in order to minimize its ex­ 3 bits. The 3-bit PWM DAC creates sin­ filters can display low-level instabilities pensive counterpart, analog technology. gle pulses at its output at 7X the word due to the truncation function of the The most promising method involves a rate. The PWM DAC is inefficient and is quantizer. The low-level instabilities give low-order (1 to 4 bits) DAC operating at inappropriate for a system with a large rise to oscillations called limit cycles very high interpolation rates (64 to 256, number of quantization levels. Consider, [Lipshitz 1988], [Ardalan 1987]. These for example). A digital loop filter is for example, a 16-bit DAC: 65,535 one­ oscillations will change in amplitude and placed between the digital interpolation level pulses would be required. The oper­ frequency depending on the signal filter and a quantizer which truncates the ation of an oversampling DAC is often present at the input of the DAC [Dijk­ lower-order bits present at the quantiz­ incorrectly described as a PWM system. mans 1989]. These problems can be er's input. The output of the quantizer is The term bitstream describes a DAC made worse with the use of a high-order returned as a feedback signal to the loop which converts a multi bit data stream to loop filter, which is required to keep the filter. A block diagram of the system is an analog signal by using a one-bit data interpolation ratio to a reasonable value shown in Figure 11. stream. Producing a DAC that uses digi­ [Fielder 1989]. The oversampling architecture re­ tal technology almost exclusively offers Because the oscillations are signal­ duces the quantization distortion present the following two advantages: (1) cost dependent, the distortion characteristics at the output of the low-order DAC and reduction and (2) identical performance of oversampled DACs must be evaluated redistributes the noise above the audible for each properly functioning DAC. (For with multiple-tone (2-3) test signals. Tra­ range. This process is called noise shap­ a given input code, the bitstream output ditional test signals used to evaluate au­ ing. The quantization noise inband is re­ will always be the same for each func­ dio equipment are not capable of fully duced to levels equivalent to a 16-bit tioning DAC.) The second advantage im­ characterizing an oversampled DAC. DAC. If the noise-shaping circuit has plies that all bitstream DACs exhibit Stikvoort has found that some audible ef­ more than a single-bit output, another identical, good linearity performance. fects in noise-shaping DACs "could not technique, called pulse width modulation Properly designed, a bitstream DAC will be detected by just measuring them" (PWM), can be used to reduce the output have better linearity than a low-grade [Stikvoort 1988]. Stikvoort found that he signal to a single-level output. In a PWM multi bit DAC. could not resolve audible "birdies" on his DAC, the area of a pulse represents the The noise shaper in the bitstream spectrum analyzer because they were "al­ DAC output. In a 3-bit DAC, seven one­ DAC is a special form of an infinite im­ most harmonic." He found that some level pulses are required to represent the pulse response (IIR) digital filter. IIR pulse-train-like effects called rattle had a

About Bitstream DAC Architecture of digital circuits, CMOS switched-capac­ plications, although its design performance Currently, two groups have been itor circuits generate noise, coupled from (using characteristics given in the data working on bi tstream DACs, namely NIT the digital circuits into the analog secti on. sheet) falls far short of what can be in Japan [Matsuya 1989] and Philips in Eu­ This leads to 1M sidebands in the region of achieved using the Philips TDA1541A-S1 rope [Naus 1987]. The two DACs use dif­ the audio band. Furthermore, it is difficult and SAA7220P/B chip set. ferent architectures. Expounding upon the to isolate from each other the two stereo Philips suggests the use of two relative merits of each architecture is out­ channels integrated on the same chip, espe­ SAA7320's in a differential mode in con­ side the scope of this article. It should be cially at 20 kHz. junction with external op amps to improve noted, however, that the DACs are differen­ Analog CMOS circuits require addi­ performance. This appears to be an expen­ tiated by the amount of analog technology tional processing steps to manufacture the sive solution, since the digital filter section implemented on the chip. The NIT chip capacitors, thereby increasing manufactur­ of the SAA7320 is then needlessly dupli­ implements its one-bit DAC with only two ing costs. In addition, it may not be possi­ cated and some of the op amps on the transistor switches, used in conjunction ble, because of the special requirements of SAA7320 are not used. Moreover, if Phil­ with an external passive RC or LC filter analog circuits, to use processes that allow ipshad intended the chip to be used in the and external op amps. Philips adopts a the most compact digital circuits. This also differential mode at the onset of the chip's switched capacitor filter (SCF) in its bit­ raises costs. In telecommunications sys­ design, its internal circuitry would have stream implementation. SCF circuit perfor­ tems, the analog portions of the oversam­ been implemented with fully balanced op mance is dependent on: (1) the settling pled coders are often implemented as a sep­ amps and switched capacitor circuits [Lee time, bandwidth, noise, and distortion of arate chip, using di fferent processes for the 1985]. It should also be noted that the de­ the internal op amps, (2) clock feed through analog and digital chips. sign of the analog differential-to-single­ and charge leakage of the FET switches, Why would Philips use an SCF? The ended converter can be difficult. Meridian (3) linearity of the monolithic capacitors, answer, I believe, is the lure of a chip that uses a novel differential-to-single-ended (4) parasitic resistance and capacitance, requires no external op amps. In addition, converter which eliminates the common­ which are unavoidable in a monolithic im­ the use of a single 5 V supply (analog mode input signal present in standard im­ plementation. Moreover, fine-line CMOS ground is generated internally on the chip) plementations of the circuit. Ben Duncan processing limits power supply vol tages to is attractive, since such a supply is cheaper uses the SSM2016 in his digital decoder ±2.5 V. This is one-sixth the voltage used in than a ±15 V supply. Moreover, the chip do-it-yourself project [Duncan 1990]. A bipolar analog circuits. This reduced volt­ could be used in portable compact disc special input stage that has true di fferential age severely impacts on the performance of players, which have a limited voltage sup­ inputs and a very high common-mode re­ the DAC, especially on distortion and sig­ ply from the batteries. The general descrip­ jection ratio is used in the SSM2016 inte­ nal-to-noise ratios. The current state-of-the­ tion section of the data sheet for the grated circuit. Harman/Kardon uses a dis­ art designs of CMOS circuits, operating at SAA7320 bi tstream DAC confirms this crete op amp which is optimized for a high 5 V, are insufficient to allow a DAC incor­ suspIcion. "The SAA7320 (DAC3) common-mode rejection ratio. Sony uses a porating these technologies to yield perfor­ is ...designed for applications in low/mid­ single inexpensive 5532 or 5534 op amp. mance levels equivalent to what is achieva­ cost, portable compact disc systems." These op amps have been shown to have ble with bipolar ICs. There is no question that the SAA7320 is a poor distortion performance in the presence When integrated with large amounts good device to be used for its intended ap- of a common-modesignal [lung 1987].

20 repetition rate so low that they did not correct application of dither and the care­ (higher than 3) passive filter is em­ appear at frequencies in the audio band. ful design of the loop filter are reported ployed, the performance requirements for Stikvoort points out that computer simu­ by all aforementioned researchers as the active stages can be reduced in com­ lation methods which are used to design strongly reducing the undesirable effects parison with the requirements in multi bit a noise-shaping DAC cannot be run long found in oversampling coders. Va nder­ systems. A problem with this approach is enough to show these effects. Goudie re­ kooy reports that one type of dither (rect­ time-domain distortion from the higher­ ports that the tones grow sidebands and angular pdt) is not recommended for bit­ order filters. In addition, the sonic conse­ then break up into a continuous spectrum stream DACs [Vanderkooy 1989]. The quences of using inductors in the signal of noise as signal level increases [Goudie optimum type of dither and the place­ path, as would be required in a high­ 1989]. Carley states that perception of a ment of the dither signal in bitstream sys­ order passive filter, are not well docu­ system's sonic performance cannot be tem are still the subject of research. The mented. [Come on, David, some of the based solely on its signal-to-noise ratio if dither used (if any) in present monolithic finest loudspeaker sys tems use them.­ the level and spectrum of the background bitstream chips is not likely to be opti­ Ed.] The high out-of-band energy present noise vary with input signal levels [Car­ mum. at the output of the CD player may cause ley 1987]. The very high quantization noise problems for some preamplifiers if the Naus has shown that the modulator level at the output of a bitstream DAC re­ signals at the output of the DAC aren't does not respond well to small-signal in­ quires a third- to fifth-order filter, in com­ filtered sufficiently. Sony, in the new put changes below threshold level [Naus parison with an 8x interpolated multibit CDP-X55ES and X77ES, only uses a 1988]. Naus reports that with an input system, which can use a first-order filter. first-order passive filter before connect­ signal slightly higher than the threshold The first-order system reduces the com­ ing the signal to an active filter stage. level, a gain error will introduce harmon­ plexity of the analog section, and conse­ The active filter stage uses an NE5532 ic distortion and inband whistles. An ad­ quently distortion from the analog sec­ op amp, which does not have the re­ ditional problem with an oversampled tion is decreased. Modern op amps will quired settling time, small-signal band­ DAC can occur when high-level signals enter slew-rate limiting if required to width, or slew-rate specifications (see be­ are present. The signal-to-noise ratio of process a bitstream signal directly. To low) to operate in this application. Static the DAC may decrease as signal ampli­ prevent this, a passive RC or LC network THD testing has not shown any abnor­ tude increases because of the nonlineari­ is used after the bitstream DAC and be­ malities in the Sony players, but dynamic ty of the quantizer [Ardalan 1987]. The fore the active circuitry. If a high-order distortion products may occur on com-

Philips has recently introduced a in a very advanced CMOS technology. This these chips. I am aware, however, that modified version of the SAA7320. (This is allows the chip to run at a 45.1584 MHz Sony has added a muting circuit to the the second modification; the first clock rate. The faster clock allows a larger CDP-X55ES and X77ES, which shorts the modification was the SAA7321.) The new number of computations to occur in a given output signals of the analog stage to ground chip, which is called the SAA7323, is time period. This clock is not at an integer when a silent track is detected. This circuit claimed to offer improved idle pattern per­ multiple of the system clock (16.9344 increases the measured signal-to-noise ratio formance at low levels. The specifications MHz). Sony uses an expensive frequency of the CD player, since the idle pattern of for the SAA7323, in the single-ended multiplier circuit to generate the DAC the DAC is not allowed to appear at the mode, are limited and are given for typical, clock. (It seems that Sony spares no ex­ output. The circuit does not appear to per­ rather than worst-case, performance: THD pense in the design of its top-of-the-line form any useful function for the consumer, at 0 dB = -90 dB; gain linearity, ±2 dB. CD players-except in the analog section.) and it can add distortion because the non­ These performance levels are between Cat­ The Te chnics chip [Ainslie 1990] uses a linear semiconductor junction of the mut­ egory 1 and Category 2 performance. modified NITstructure which has 11 quan­ ing circuit is connected across the output of Technics (MN6471M) and Sony tization levels instead of 7 at the output of the CD player. Even so, the Philips (CXD2552) have implemented chips based the quantizer. The Te chnics chip runs at a SAA7323 is clearly inferior to the Sony on the NIT technology. Both chip sets can slightly slower 33.8688 MHz, which is an CXD2552. Philips has meanwhile an­ also be operated in the di fferential mode. integer multiple of the system clock. Be­ nounced its very latest chip, the SAA7350, However, unlike the Philips chip, the Japa­ cause of its higher clock rate, the Sony sys­ in an attempt to catch up with Sony. The nese DACs achieve this operational mode tem interpolates at a 64x rate, while the SAA7350 uses a thi rd-order noise shaper, with a single chip. In the September 1989 Technicssystem interpolates at a 32x rate. like the NITsystem, instead of the simpler issue of CD Review, the Sansui AU­ Both the Technics and Philips chips second-order noise shapero It has differen­ X911DG digital decoder/amplifier, which incorporate digital interpolating filters with tial outputs, a feature the NIT system has uses the Technics chip, was tested. [No test a smaller number of filter taps and a small­ always had, and it uses an external digital of this Sansui model is being contemplated er data path size than the digital filters ana­ filter, as does the Sony CXD2552. Philips by The Audio Critic.-Ed.J From these lyzed in Ta ble 3. The simplification of the has not yet completed development of the tests, it appears that this chip also operates digital filter is required to make room on 20-bit digital filter to be used with the at a performance level below the state-of­ the silicon for the circuitry associated with SAA7350, but the latter can be used with the-art multi bit DACs. Distortion on single the bi tstream DAC. The result is more pass­ the Japanese digital filters shown in Ta ble tones is 4 dB higher than state-of-the-art at band ripple, among other problems. The 3. The SAA7350 continues to use switched 1 kHz and, strangely, 25 dB higher at 20 Sony CXD2552 DAC does not incorporate capacitor analog circuits and a ±2.5 Vana­ Hz. From the early reviews in the hi-fi the digital interpolation filter on the same log power supply. The specifications forthe slicks of preproduction samples of the Sony silicon. Sony uses the excellent CXD1244 SAA7350 are once again limited and given CDP-X55ES and X77ES, it appears that digital filter (see Ta ble 3) in conjunction for typical, not worst-case, performance:

the Sony chi p set has better performance. with the CXD2552. I have not seen the data THD at 0 dB = -93 dB; gain linearity, ±1 Sony appears to have implemented the sheet for the CXD2552, or for the Te chnics dB. This performance is equivalent to that NIT system without modification [Mat­ MN6471M, and thus I cannot make defini­ of a chip between Category 3 and Category suya 1989] . The CXD2552 is manufactured tive statements about the performance of 4 and is far from state-of-the-art.

21 plex test signals. (For further bitstream [Fourre 19B9]. This is the most difficult free as possible. This is best accom­ information, see sidebar.) problem faced by the designer of a CD plished by placing the optocouplers after Bitstream technology certainly of­ player's power supply. Three compo­ the digital filterchip. fers superior performance in low-cost nents of the power supply are critical in A comparable technique to separate CD players that have .used inexpensive, determining its performance. These are transformers is separate windings for the nonlinear DACs until now. I expect it to the transformer, the storage and bypass analog and digital section on a single become the dominant technology in such capacitors, and the voltage regulators. transformer. This technique is used in low-cost players. The performance of the The servo circuits operate at frequencies most midpriced CD players. An undesira­ analog section in low-cost CD players is in the audio band and require large cur­ ble trend in many low- and midpriced clearly compatible with the performance rent resources to position the laser as­ CD players is an attempt to reduce trans­ of the bitstream system. In mid- and sembly accurately. Signals from the digi­ former costs by reducing the voltage and high-priced CD players, I believe that a tal sections of the CD player operate at current available at the output of the properly adjusted Burr-Brown PCM5BP­ much higher frequencies than the audio transformer. Especially disturbing is the K or PCM61P-K, when used with a prop­ band. These digital signals make their use of transformers whose voltage output erly designed analog section, will outper­ presence felt as 1M distortion compo­ is so inadequate that only %5 V regulated form the Philips and Te chnics bitstream nents which appear in the audio band analog supply rails can be generated. The systems, including the new Philips [Miller 19B9]. These distortion compo­ normal analog supplies are three times as SAA 7350. The Sony CXD2552 repre­ nents occur when the high-frequency sig­ large at ±15 V. Using ±5 V supplies re­ sents the next generation of bitstream nal is modulated with the audio signal in quires the inclusion of special low­ DACs. It will be interesting to compare an analog stage. This gives rise to sum voltage analog ICs in the analog section. its performance--especially when, and and difference sidebands. The difference These analog ICs do not have the perfor­ if, Sony implements the analog section sidebands can be in the audio range, and mance of ICs that run on ±15 volts be­ better than in their current products­ the sum sidebands can interact with other cause their design requires that they op­ with the next generation of multi bit high-frequency components to produce erate with small margins between the CD DACs such as the Burr-Brown PCM63P­ secondary 1M products, which may also player's maximum analog output swing K. Both bitstream and multi bit DAC appear in the audio band. The reduction of ±2.B V and the power supply rails. An technology will obviously continue to of digital and RF noise in an outboard example of a fa irly recent CD player improve in the future, and it is unclear decoder box, in comparison with a CD with ±5 V supplies is the Sony CDP- which technology will finally prove to be player, may improve the performance 50BESD. This player also has a single superior. For the moment, only Harman/ considerably more than the small amount transformer winding to supply both the Kardon, Meridian, Sansui, and Sony are of jitter in the clock signal will degrade analog and digital sections. The Sony using bitstream technology in their top­ it. CDP-505ESD, which was replaced by of-the-line CD players. Even Philips and The best method of isolating the an­ the 50B, had separate windings on the Technics use multibit DACs in their alog and digital sections is to use a sepa­ transformer and ±15 V analog supply flagship models. Most manufacturers are rate transformer for each section. High­ rails. The Sony CDP-508ESD has yet using bitstream in low- and midpriced end CD players such as the Ya maha another cost-cutting circuit in its power players only. CDX-1l20, Sony CDP-X7ESD, and De­ supply. The 50B uses the PCM5BP (unse­ Please note that it is in the interest non DCD-3520 apply this technique. lected) DAC. This DAC requires a +12 of the Japanese electronics industry to Even these players combine the ground V supply. Because the 50B has only a +5 push the bitstream technology. The Japa­ line of the two transformers, and conse­ V analog supply rail, a voltage multiplier nese have been strong in the production quently noise on the ground line from the circuit generates the DAC supply volt­ of digital integrated circuits like the bit­ servo and digital sections is imposed on age. The voltage multiplier produces stream chips. The Japanese have been the analog section. Careful attention to only an B V output, which is not the opti­ unable, on the other hand, to compete in the ground routing can greatly minimize mal supply voltage. In addition, the high the field ofhigh- performance analog in­ the effect. A more sophisticated method output impedance of the voltage multipli­ tegrated circuits. For that reason they is the use of optical couplers that isolate er compromises the performance of the have had to purchase multibit DACs the analog and digital sections. This ap­ DAC. from the USA or from Philips.The verti­ proach is championed by Onkyo, who of­ Many midpriced CD players use cally integrated Japanese companies can fe rs dual power supplies even in the ±12 V supplies (e.g., the Sony CDP- save a significant amount of money if $700 Integra DX-7500. Optocouplers are 60BESD and Kenwood DP-BOlO). While they can convince the consumer that the also used in the CAL Tercet Mk III and the impact on sound quality is not as se­ bitstream product is superior. the JVC K2 Interface. Both Ya maha, in vere as with ±5 V rails, it appears to me its CDX-1l20, and Denon, in its DCD- that it is more sensible to retain the better Power Supplies 3520, have abandoned the optocouplers power supply and eliminate trivial fea­ The power supply for the DAC and found in the previous generation of these tures, such as nonvolatile memory for the the analog stages can significantly im­ machines. It is unclear whether this is a programming functions or remote vol­ pact on the performance of a CD player. move to cut costs or an effort to elimi­ ume controls. Not all manufacturers have Power supply designs for CD players nate technical problems with the opto­ compromised their transformers. For ex­ vary widely. The power supplies of many couplers. The optocouplers can be placed ample, the $429.95 Philips CD-60 has of the new, low-end CD players severely either between the data recovery chip full ±15 V analog supply rails. compromise the performance of these and the digital filter chip, as in the On­ The active voltage regulator, which players, as their power supplies have kyo, or between the digital filter chip and is placed after the unregulated power been designed to such a low price point. the DAC, as in the Philips LHH1000. supply rails, creates a supply rail voltage Power supply rails which service the Since the output of the digital filters is set to a precise value. The regulator sup­ analog section must be shielded from sig­ connected to the DAC inputs, and noise presses ripple and noise fo und on the un­ nals originating from the digital circuitry on these digital signal lines can corrupt regulated supply. The regulator output and from the servo circuits which move the analog output of the DAC, it is desir­ has a low output impedance which main­ the CD and position the laser mechanism able to have the digital signals as noise- tains a constant voltage on the regulated 22 line under changing load conditions and ulators seen in Japanese equipment be­ tors, which must be used with each ac­ prevents noise from coupling into the cause they require external components. tive regulator, can cost several dollars regulator's supply line. Separate regula­ Integrated regulators of a given device each, even in high quantities. Hence, tors on the supply lines for the DAC and type are available in a variety of different their use can significantly affect the final analog sections isolate these circuits current ratings. Regulators with high cur­ price of a CD player. Unfortunately, from each other. Ideally, each power rent ratings have a lower output imped­ some high-end audio manufacturers supply terminal on the DAC (2 for the ance, but they are more expensive. A claim mystical properties for their pri­ PCM58P, 3 for the TDA1541, and 4 for well-designed discrete regulator will vate-label capacitors. There is little cor­ the PCM56P and PCM61P ) should have have lower output impedance, especially relation between the brand of capacitor a separate regulator. Since the signal at at higher frequencies, when compared used in a power supply and the power the output of the digital filter is connect­ with integrated devices. Moreover, when supply's performance. The size and type ed to the DAC, a separate regulator for compared with integrated devices, dis­ of capacitor makes the performance dif­ the digital supply to the digital filter chip crete regulators can have lower output ference. Capacitors are not a panacea. (separate from the regulator which sup­ noise, higher ripple rejection, better tem­ Va rious small modifiers of audio equip­ plies the other digital circuitry in the CD perature stability, and better regulation ment, who do not have solid engineering player) is advantageous so as to mini­ under load [Marsh 1983], [Breakall foundations, will add expensive capaci­ mize signal interference. An additional 1983]. A high-performance regulator is tors across almost any power supply regulator may be used in the crystal often formed in two stages, with a pre­ point to be fo und on a CD player. Since clock oscillator circuit to prevent supply regulator near the power supply and a set these modifiers are unable to perform noise from modulating the oscillator (this of subregulators, connected to the prereg­ careful engineering analyses to demon­ would result in clock jitter). Improve­ ulator, near the active analog circuits strate what types and values of capacitors ments in stereo separation may possibly [Didden 1987]. Integrated regulators can are best suited for each circuit, the be achieved with a dual mono sometimes be incorporated in the prereg­ changes that result are often too su­ configuration that includes separate regu­ ulator or the slave section (but not both) perficial toaffect the sound of the player. lators for the left and right channel ana­ in high-performance regulating systems Mark Brasfield of MSB Technology (a log sections. It is also possible to use without significant performance degrada­ degreed engineer working at Stanford separate regulators for the left- and right­ tion. The vast majority of American CD Research Institute) takes an interesting channel DACs, in the case of DACs designers and modifiers use all discrete minority approach which seeks to elimi­ which process only one channel of the regulators or a combination of discrete nate as many bypass capacitors as possi­ stereo pair, such as the Burr-Brown and integrated regulators in their CD ble. Mr. Brasfield designs the voltage PCM56P, PCM58P, and PCM61P, and players. regulators to present a very low imped­ the Analog Devices AD1856 and 1860. Some Japanese and European manu­ ance to the active circuitry, even at very Dual mono configuration cannot be facturers also adopt discrete primary reg­ high frequencies. achieved in the TDA1541 since a single ulators in their high-end CD models For cost control reasons, some of chip houses both the left- and right­ (e.g., Philips CD-80, Sony CDP- these midpriced players show smaller channel DACs. Players are often adver­ 608ESD, and Sony CDP-X7ESD). The storage capacitors than earlier designs. tised as dual mono or twin mono, even regulator stage in the Pioneer PD-71 uses The Sony CDP-950, for instance, has though both channels share the same a very interesting regulator circuit. The only a 1000 !IF capacitor on the unregu­ power-supply regulators. The 1989 bro­ regulator has a complementary class AB lated negative analog supply node, while chure for the Sony ES line, for example, output stage. Normally, regulators are de­ its predecessor, the CDP-91O, used a claims that the CD players are twin signed to either source or sink current 3300 !IFcapacitor. Transformers, regula­ mono, although both channels use the and consequently only an NPN or PNP tors, and filter capacitors are good sites same analog supplies. This mistake is transistor is found at the regulator's out­ for compromise from a marketing point possibly the result of changes incorporat­ put. A standard positive regulator circuit, of view, as these changes are not visible ed after the preparation of the brochure. for example, presents a low impedance to the consumer in the showroom. Only an examination of the service man­ output only when it sources current. In ual will indicate whether or not the the presence of large-amplitude, high­ Analog Stages in CD Players player has the required regulators for frequency symmetric noise sources, a Let us say you own a Sony CDP­ dual mono operation. The total number complementary regulator may reduce the X7ESD, for which you paid $2000 (mi­ of regulators can be high, usually in the amount of noise which couples to the nus whatever discount was available). low to middle teens. The CAL Te rcet Mk regulator'S output line. That bought you a player with one of the III uses 23 regulators. On the other hand, The size of the storage capacitors most advanced transports in the business. the number of regulators used in some used after the power transformer deter­ Your player also has separate transform­ Japanese CD players is being reduced. mines the ripple voltage and current re­ ers for the analog and digital sections, For example, the discontinued Sony serve of the power supply. This can vary and the power supply has discrete prereg­ CDP-505ESD had 7 regulators (includ­ from hundreds of microfarads to tens of ulators for the analog section and nine in­ ing 2 for the servo system), while the thousands of microfarads. The shunt ca­ tegrated subregulators. In addition, the CDP-508ESD that replaced it uses only 3 pacitors placed on the regulated power CDP-X7ESD has a special-selection regulators. supply rails ensure that a low source im­ grade of the Burr-Brown PCM58P, ex­ Like op amps, regulators can either pedance is present at frequencies at clusive to this unit. Unfortunately, you be purchased as integrated circuits or im­ which the active regulator stage's imped­ also have three op amps in the signal plemented with discrete resistors, diodes, ance is increasing. Many manufacturers path, with a price of much less than $1 and transistors. The 78XX, 79XX, 317, combine low-ESR electrolytic capaci­ per chip in the quantities purchased by and 339 are the most popular devices. tors, filmcapac itors, and ceramic capaci­ Sony. Do not feel too distraught; at least The 317 and 339 are frequently cited by tors in parallel to achieve low shunt im­ you did not purchase a Sony CDP-Rl/ American high-end manufacturers as of­ pedances to noise present within the DAS-R1 for $8000 or a Philips LHlOOO fering better performance than the 78XX wide spectrum of frequencies inherent in for $4000, both of which use $1 op amps and 79XX devices. Rarely are these reg- a CD player. These high-quality capaci- in their respective analog stages. 23 It is one of the great mysteries of loop distortion, bandwidth, and large­ gain elements (and tubes for those who high-end hi-fi why the large manufactur­ signal settling time for a given applica­ must have a device that glows in the ers shun modern discrete amplifier de­ tion [Cherry 1982]. Second, the return dark) may be mixed in one amplifier to­ sign techniques in their high-end prod­ loop gain, which sets the amount of feed­ pology. Fifth, passive components of any ucts. The audio performance of a CD back of the amplifying stage, is adjust­ type or value are practical. This is of par­ player could be greatly improved if the able. Refer to [Otala 1980] for a rigorous ticular importance when designing an inexpensive op amps were replaced with analysis of the detrimental effects of ex­ amplifier with low open-loop distortion, discrete amplifiers. The problems with cessive feedback. The validity of the good power supply rejection ratio, and inexpensive op amps and the proper de­ analysis when applied to analog am­ good transient performance. Sixth, there sign of discrete amplifiers is beyond the plifiers has been questioned, however is the option to use power supply rails of scope of this article. A well-designed dis­ [Cordell 1983], [Cherry 1983]. Third, de­ more than ±15 V. Seventh, there is the crete amplifier offers many advantages signing an output stage with class A op­ option to use fully complementary topol­ [Marsh 1985], [8orbely 1987], [Howe eration, no IN current limiting, and low ogy to reduce open-loop harmonic distor­ 1989], [Jensen 1980], seven of which output impedance becomes an option. tion. If the open-loop distortion is low­ will be cited here. Fourth, the designer can select from the ered, the amount of negative feedback First, the gain and transfer response thousands of discrete transistors avail­ can be reduced by the same amount with­ of the individual stages and the complete able and use transistors with different op­ out changing the closed-loop distortion. amplifier can all be adjusted, as can the erating parameters in different stages of A disadvantage of discrete circuitry, operating current of each stage. This al­ the amplifier to optimize performance. In when compared to an IC or hybrid, is in­ lows the optimal trade-offof noise, open- addition, bipolar, JFET, and MOSFET creased parasitic capacitance from the

Designing the Analog Stage what, a first-order filter function, which back is equal to the input current. Since the Analog stage design in a CD player is forms part of the reconstruction filter, can input node is held at ground potential, the similar to preamplifier stage design. The also be incorporated in the current-to­ voltage output of the stage is the voltage functional first section of the analog stage, voltage converter. An additional operating acrossthe resistor. Hence, namely the current-to-voltage converter, is constraint requires that the input of the cur­ Vo ut = the most difficultto design. The current-to­ rent-to-voltage converter be held at its This is the ideal -1inRequation for a volt­ voltage converter takes the current output ground potential so that the DAC is termi­ age-to-current converter with conversion of the DAC as its input and generates a nated into a virtual short. If the output of impedance R. voltage signal at its output with a level Iin­ the DAC is not held at a constant ground If the designer has chosen to reduce earl y proportional to the current entering potential, the linearity of the converter is the amount of negative feedback in an au­ the stage. Design problems arise in this degraded. The current-to-voltage converter dio amplifier, the transimpedance amplifier stage because the DAC output is not a should also show good rej ection of high­ has an additional advantage. Cordell argues bandlimited analog signal, but rather a set frequency noise, which may be present on that an amplifiercan be designed with high of current steps changing at the word rate the power suppl y rails. A current-to-voltage global feedback rates and avoid exhibiting of the DAC. The stage must respond to the converter stage should have a minimum dynamic distortion [Cordell 1980]. D. C. step change in current with a step change in power supply rejection ratio of 60 dB at Wadsworth of Phototronics contends that voltage. The voltage change must occur 100 kHz. this argument is valid only if the input sig­ within a small percentage of the conversion A current-to-voltage converter can be nal is bandlimited to the audio band. The period, and the voltage output must settle to crafted from a voltage amplifier [lung input to a current-to-voltage converter has a the new signal within an accuracy of 1986], though the performance criteria are bandwidth well into the megahertz region, 0.0015%(for 16-bit accuracy) of the output more easily met using a current feedback thereby violating this condition. A voltage expected from an ideal current-to-voltage (transimpedance) amplifier [Evans 1985], amplifier, when used in a current-to-voltage converter. To achieve this level of perfor­ [Goodenough 1987]. This topology offers converter, requires high rates of feedback mance, the amplifier must have a high slew higher slew rates, greater bandwidth, and to reduce the input impedance of the am­ rate, a wide bandwidth, linear operation lower settling times than is possi ble with an plifier [Millman 1979] to a point at which with "large-signal" inputs, and large stabili­ op amp topology. This circuit is more easi­ the DAC output will not move significantly ty margins. As an example of the require­ ly understood than the op amp circuit. It, from ground potential. Only transimped­ ments in a current-to-voltage converter, therefore, will be used as a point of refer­ ance amplifiers and current amplifiers-see consider the required slew rate for an am­ ence in explaining the operation of a cur­ [Didden 1989] for an elegant treatment of a plifier to slew completely from the maxi­ rent-to-voltage converter stage. The trans­ current amplifieras part of an IN converter mum positive signal to the maximum nega­ impedance amplifier (Figure 12) works by stage-are feasible if a low rate of negative tive signal (a swing of 5.6 V typically) in placing a low impedance across the input to feedback is to be used around an amplifier one-tenth the conversion period of the act as a short. This forces the input voltage which simultaneously provides a low input DAC. For an 8x interpolating system this is to held at groondpctential as required fa' impedance. The principal disadvantage of n be 285 ns. A slew rate of at least 20 V/IAS is re­ Viproper D/A performance. Feedback, in the current-mode amplifiers is that they are not quired. form of current, is applied to the input as easy to use as op amps [Goodenough To settle within 0.0015% of a signal's through R. Current from the DAC and cur­ 1990]. Increased noise levels are also a final value, a circuit with a single-order rent from the feedback branch enter the potential problem with current-mode am­ lowpass response must settle in 11 time short-circuit input of the transimpedance plifiers. constants. This requires the current-to­ amplifier. Accuphase, Precision Audio, Barclay voltage converter to have a bandwidth 11 The gain stage, whose transfer func­ Audio (a company no longer producing times the word rate. For an 8x interpolating tion is A (with units in ohms), senses the complete CD players), and M. S. Brasfield system, this requires the current-to-voltage current in the short and transforms it into a adopt transimpedance amplifiers in their re­ converter to have a bandwidth greater than voltage. If A is large enough, the feedback spective current-to-voltage conversion stag­ 4 MHz.To ease these requirements some- loop will be satisfied when the current feed- es. Interestingly, the latter three are small

24 large PC board traces. This reduces the expensive DAC and a discrete analog verter (the circuit is holding the data for speed and bandwidth of discrete circuits stage. half a clock cycle). In addition, it must and increases settling time. Companies It is possible to use op amps and have a high input impedance and show using discrete amplification include CAL other monolithic functional blocks, such stability with capacitive loads. Often, (Tercet Mk III), Aragon, Harman/Kar­ as buffer circuits [Williams 1986], in more than one amplifier is needed to don, and Krell. conjunction with a few discrete transis­ meet all the requirements. Additional Latitude in discrete amplifierdesign tors to form an amplifying stage that will considerations in the sample-and-hold means that a designer can create an am­ approach the performance of a discrete circuit are the accuracy and speed of the plifier that will impose almost no audible amplifier. Companies adopting this ap­ transistor switches. Furthermore, the per­ distortion on the signal or, alternatively, proach are AVA, Proceed, Pioneer (PD- fo rmance of the storage capacitor, where an amplifierthat is distinctly inferior to a 71), and PS Audio. the data is held, is critical [Jung and $1 op amp. Discrete amplifiers are ex­ Performance requirements for an Marsh 1980]. None of the sample-and­ pensive to build properly. Hence, op amplifier in the sample-and-hold stage hold circuits used in CD players that I amps play an important role in low- and which fo llows the current-to-voltage con­ have examined are sophisticated enough midpriced CD players. An analog back verter are even more difficult to meet to meet these performance requirements. end cannot solve problems in the stages than for the current-to-voltage converter The most practical solution in the design that precede it. I believe that a CD player itself, if the DAC must be deglitched. of a high-performance CD player is the that incorporates an op amp analog stage The settling time is the culprit here. The elimination of this stage by using a DAC and the highest-quality DAC available amplifier in this stage must settle in half with low glitch current. will outperform a CD player with a less the time of the current-to-voltage con- The last stage in the analog section

American companies with degreed electri­ their development phases. Older IC devices data for the NE5532, NE5534, and LM833 cal engineers heading the design depart­ are priced closer to the direct cost of manu­ op amps used in most European and Japa­ ment. All three companies developed the facturing. The price of the PA630 will r�­ nese CD players. As can be seen from the solution independently. Small CD modifi­ strict its use in budget machines, but mid­ chart, these devices do not meet the 20 ers who do not have an engineering back­ priced machines should be able to V/I-IS slew rate requirement calculated ground continue to use op amps in their de­ incorporate these devices. above. Furthermore, the critical specifi­ signs. Unfortunately, no matter how many A number ofhigh-p erformance mono­ cation of s�ttling time is not disclosed for different types of op amps they try, they lithic current feedback amplifiers have been these chips. Manufacturers often state that will never overcome the fundamental de­ recently introduced to the market. The prin­ they are forced to use these op amps be­ sign problems understood by trained pro­ cipal speci fications for these op amps are cause they have lower noise than higher­ fe ssionals. given in Ta ble 4. Unlike the PA630, these speed devices. As can be seen from the A designer who has chosen to use inte­ chips are not specific to CD players. They chart, this objection is substantiated, with grated circuits in the analog stages now has offer lower THD levels, though dynamic the exception of the Burr-Brown OPA627. the option of using an integrated circuit de­ distortion products may result. Two provi­ The reduction in the signal-to-noise ratio signed specificallyfor high-end and profes­ sos: these wide-bandwidth IC devices can that would result if the other high-speed op sional applications. The chip is the Photo­ be difficult to work wi th and will oscillate amps were employed is an insignificant 9 tronies PA630 current conveyor, designed if not properly employed. For this reason, I to 12 dB-insignificant because all of them by D. C. Wadsworth [Wadsworth 1989]. do not suggest that you attempt to replace exceed the inherent signal-to-noise ratio of The current conveyor (patents pending) is a the op amps in your present CD player with a 16-bit digital encoding system, which is special form of current amplifier that re­ current feedback amplifiers. 98.1 dB and therefore the effective limiting quires no global feedback. Output buffers A few low-noise, high-slew-rate, parameter. If this reduction is not accept­ included on the chip for filtering functions wide-bandwidth monolithic operational able, a high-speed op amp can be used in (see below) also use no global feedback. amplifiers can be used with good results in conj unction with a simple discrete pream­ The chip is processed using an advanced the current-to-voltage conversion stage if a plifier to form a low-noise, high-speed am­ (and expensive) complementary bipolar designer chooses to implement the current­ plifier. As an example, the data sheet for the process. All passive components are left off to-voltage converter with op amps. Jung SSM-22lO transistors shows an amplifier the IC so that high-quality discrete passives uses the op amp's offset adjustment pins in with a 1.7 nV/YHz noise level, 40 V/I-IS can be used. No short-circuit protection, an innovative manner to linearize the first slew rate, and 63 MHz bandwidth. The which could introduce nonlinearity and in­ stage of some op amps [Jung 1986) . This noise level of this amplifier is lower than crease settling time, is included. Because raises the linear input range and lowers the that of any commercial op amp. . the entire circuit is implemented on a single global feedback rate by forming an inner Mike Moffat of Theta uses the PMI chip, the rise time of the current conveyor feedback loop around the first stage of the OP42 operational amplifier in his designs. is less than 25 ns. Since no global feedback amplifier. A relatively inexpensive Signet­ Mr. Moffat claims that the use of Te flon PC is used, THD is an order of magnitude ies NE530 op amp is used so that the cir­ boards (which are extraordinarily expen­ (0.02% at 0.5 V rms) higher than that for cuit can create a current-to-voltage convert­ sive) makes a significant difference in circuits that use even moderate levels of er with good performance in a popularly sound quality. [A hem.-Ed.] A discrete de­ global feedback. The level is still low priced CD player [Jung and Childress sign would take up too much board space enough to be almost certainly inaudible. 1988]. A popular op amp used in some to allow the use of a Te flon PC board, Mr. The chip also has fairly low power supply high-end CD players as a current-to­ Moffat argues. Mr. Moffa t states that an op­ rejection; consequently, it requires good voltage converter is the PMI OP42. The amp-based analog stage will outperform a supply regulation. These chips are priced performance of the OP42 and of some discrete stage on a cheaper PC board. Mr. higher than standard op amps because of more recently introduced op amps is given Moffat also states that a current feedback the advanced processing technology. In ad­ Ta ble 4. As can be seen, the OP42 is an ex­ amplifier, while exhibiting fast settling to dition, new monolithic devices must be cellent choice for cost-sensitive applica­ 10- or 12-bit accuracy, may not settle as priced high enough to cover the cost of tions. Ta ble 4 also gives the performance quickly to 16-bit accuracy as the OP42.

25 CURRENT FEEDBACK

R

I. III V out

V. III I � 6 (A)I �

ANALOG GROUND

Figure 12

26 of a CD player is the reconstruction filter rated in the feedback loop of the current­ cuit, and the reactive components around stage. This stage removes the high­ to-voltage converter, the effects of the the Sallen-Key filterare not in the circuit frequency spectral components inherent processing variations are canceled. Un­ at audio frequencies. The GIC implemen­ in a sampled data system. The use of dig­ fortunately, the absolute value of the tation requires two additional op amp ital interpolation reduces the required or­ feedback resistor varies between differ­ stages and fo ur additional passive com­ der of the analog filter stage consider­ ent dies, so it is incapable of forming a ponents. ably. For an interpolation rate of 8x, a filter with a precise time constant. The A preferable approach would save simple one-pole filter is adequate, al­ current-to-voltage converter and first the cost of these components by building though most Japanese players (Sony, De­ filter stage cannot be combined because a Sallen-Key stage with an improved op non, Pioneer) use a third-order filter. The of this, and an additional active gain amp. Pioneer, in their PD-71, and Ken­ Onkyo DX-7500 offers the option of a stage is needed. CAL, Madrigal, Theta, wood in the DP-801O use a filter section first-order or a third-order filter. The Ya­ and Kinergetics adopt the Burr-Brown which is formed around an op amp in the maha CDX- 1120 has this option; howev­ PCM61P by carefully matching the cur­ inverting configuration. This eliminates er, a bipolar analog switch switches the rent output of pairs of the DACs to obvi­ the common-mode distortion of an op filter in and out of the circuit instead of ate the need for an internal feedback re­ amp when it is used in a buffer separate jacks for the two options. Only sistor. This approach requires significant configuration. The disadvantage of this the passive components are bypassed in testing, and some working devices may circuit is the added inverting stage which the CDX-1120, not the additional active not be usable if they cannot be matched is required if absolute polarity at the out­ stage. It is possible that the bipolar ana­ to another DAC. All four DACs listed put is to be retained. This can be accom­ log switch adds more distortion than is above have a built-in op amp which can plished with a simple digital circuit, removed by shorting out the passive form the current-to-voltage converter. though Pioneer chose an analog circuit. components.The higher-order filter is re­ The op amp on the PCM56P was not The circuit used by Pioneer and Ken­ quired if the preamplifierstage following adopted by the Japanese in their mid­ wood also has the additional disadvan­ the CD player is not linear outside the priced and high-end products. The op tage of a low input impedance, which audio band. Most modern preamps amp on the Analog Devices may have may be difficult for the previous stage to should have no problem with the low­ better performance, since it is imple­ drive. order filter. Several designers of CD mented in a more advanced (BICMOS) The electrical requirements of the playback equipment disagree, finding technology. The Burr-Brown PCM58P active voltage gain stage in a filterstage that some otherwise excellent pream­ has a laser-trimmed current source so its are similar to those of a preamplifier plifiers and power amplifiers require current output is matched between de­ stage, except that the bandwidth of the greater attenuation than a first-order sys­ vices. stage should be wide enough to ensure tem can provide. For this reason, Aragon The Philips TDA1541 has matched proper operation of the filter in its stop­ uses a third-order Bessel filter. As stated current outputs because both channels band region. The PSRR (power supply above, the first-order filter can be incor­ are on a single die. Unfortunately, the de­ rejection ratio) of the amplifiershould be porated in the current-to-voltage convert­ vice can only work at a 4x interpolation high at frequencies outside the passband, er. Consequently, a single amplifying rate, thereby requiring a third-order filter. so that noise on the power supplies is not stage is used for the entire analog signal­ To implement a third-order filter, two ac­ coupled to the output. processing section. In addition to the On­ tive stages, including the current-to­ The Sallen-Key filter circuit does kyo DX-7500, the Krell, PS Audio, and voltage converter, are required. Mark not require a stage with voltage gain. Theta digital decoders have a single ana­ Brasfield only uses a first-order filter in Hence, a simpler unity-gain buffer can be log stage. The Wadia goes even further, his TDA1541-based design. This creates used. Economics favor the design of a incorporating only RF filtering. a significant amount of high-frequency discrete buffer when compared with a If the output stage of the IN con­ output from his player, centered around discrete voltage gain stage. If it is uneco­ verter is not robust enough, the settling 176.4 kHz. Mr. Brasfield states that the nomical to use a discrete circuit in the performance of the IN converter can be level of these out-of-band signals is filter section, a high-quality monolithic affected by the loading of the preamp sufficiently low to render them inconse­ op amp or buffer should be used. Jung and cables. This problem is most likely quential. and Childress discuss the performance of to occur with an IN converter that uses Sony, Denon, and Onkyo use GIC­ a variety of op amps and recommend only an integrated circuit. To prevent the based third-order filters in their more ex­ those that yield the best performance settling time degradation, Theta and Wa­ pensive players. The GIC (generalized [Jung and Childress 1988], [Jung 1987]; dia incorporate an additional buffer stage impedance converter) filter creates high­ they also also discuss monolithic buffers after the IN converter. order (fifth or greater) filters which have [Jung and Childress 1988]. In addition to It is not possible to combine the cur­ low sensitivity to passive component the active stage of the filter, the passive rent-to-voltage converter and the filter in variation. Accuphase used the GIC filter components of the filter can also affect one stage for all DACs. The problem in its first CD player (the player did not sound quality [Jung and Marsh 1980]. arises in DACs with mismatched fu ll­ use digital interpolation) to implement a The output from the reconstruction scale current outputs (Burr-Brown high-order filter and achieved good re­ filter has a DC offset originating from PCM56P, PCM61P, Analog Devices sults. Subsequently, the GIC circuit ap­ both the active gain stage and any ran­ AD1856 and AD1860). The mismatches peared in other CD players with low­ dom or systematic DC offset in the DAC are caused by processing variations in order filters under the assumption that and/or the digital filter. The problem is the values of the resistors at the DAC the performance advantage of the GIC not unique to CD players; preamplifiers core. To match the voltage output be­ circuit would be retained. It is argued and amplifiers also will have DC offsets tween the two stereo channels at the out­ that the GIC circuit is not in the path of at their outputs. This offset can be elimi­ put of the current-to-voltage converter, the analog signal as in the more standard nated with a coupling capacitor. Howev­ an additional monolithic resistor-which Sallen-Key topology. The argument is er, great care must be taken to ensure that tracks the value of the resistors in the flawed because a unity-gain buffer must the capacitor does not affect the sound DAC's core-is used, since it is on the follow the GIC filter. The Sallen-Key to­ quality of the signal. High-quality capac­ same die. When this resistor is incorpo- pology uses an identical unity-gain cir- itors can be expensive and may not be 27 available in large enough values to en­ quency response. This additional filtering complished with a transistor or a relay. sure a low enough cutoff frequency to reduces the quantization noise present at The relay is preferred, since the transistor provide the best possible bass response. the output of the DAC. The noise arises will present a nonlinear impedance to the As with the power supply, bypass capaci­ when the original data is quantized to 16 output when the muting circuit is not ac­ tors and multiple capacitors of different bits in the recording process. The de­ tivated. The relay and its associated cir­ types and values may be combined to emphasis filter is usually implemented cuitry (which drives the relay) are obvi­ yield the most transparent sound possible by placing a passive network across the ously more expensive than a transistor through the DC blockingstage. feedback resistor of the current-to­ switch. The relay, therefore, is found The alternative to a DC blocking ca­ voltage converter or the first filter stage. mostly in high-priced CD players. One pacitor-used by Theta, Precision Audio, This approach is problematic, since a sol­ exception is the midpriced Onkyo DX- Philips in the CD-80, and Onkyo in the id-state device is used as a switch and, 7500. Some modifiers eliminate the tran­ DX-7500, among others-is a DC servo consequently, nonlinear junction imped­ sistor switch but do not replace it with a that nulls out the DC component from ances of the device can affect the sound relay. They short out the power switch, the output by placing a compensating DC quality when the switch is off. The prob­ letting the unit run continuously. This is voltage at the input of a stage that pre­ lem can be remedied by using a relay. a risky move, since a power interruption cedes the output [Clark 1982]. Unlike a Regrettably, in this application the relay will cause pulses to appear at the CD coupling capacitor, a DC servo is, itself, must be of a high-quality design if reli­ player's output. If you are using a pas ­ not in the signal path. Hence, it does not ability problems with the relay contacts sive preamp and a power amp with a affect the sound quality, provided it is de­ are to be prevented [Duncan 1988]. large supply filter bank, these pulses will signed properly. The DC servo often is Some of the relays I have seen in CD appearat your speaker terminals and pos­ incorrectly described as allowing fre­ players are not of high enough quality to sibly destroy your speakers. quency response down to DC. The servo overcome these reliability problems. An­ For the best possible performance, it actually does not allow DC or very low­ other problem with the placement of the is desirable not to load the output of the frequency signals to pass to the output. If de-emphasis network across the feedback CD player with a remote volume control the DC servo fa ils, or a power supply rail resistor is instability arising from the re­ circuit. Eliminating the remote volume collapses, or a circuit in the forward path active components in the network. This control-as well as a headphone am­ of the servo fa ils, a 15 V or higher DC can be resolved only by reducing the plifier-helps keep PC board traces as voltage will appear at the output of the open-loop bandwidth of the amplification short as possible,improving stereo separ­ CD player. A power amplifier or loud­ stage. A passive de-emphasis network ation. Eliminating rarely used features speaker can be readily destroyed if this placed between the current-to-voltage also lowers the cost of the player. Many fa ilure mode occurs. A fail-safe protec­ converter and the first filter stage has modifiers disconnect these fu nctions dur­ tion circuit can be included, though only been employed by some designers to ing the modification process, and high­ Philips among the above manufacturers avoid compromising the amplification end CD players generally do not include has included the protection circuit. Jon stage. A single analog stage cannot ac­ them. Surprisingly, the Pioneer PD-71 Schleisner of Precision Audio argues that commodate a passive de-emphasis cir­ also does not have these features-an un­ the chances of failure of the circuit are cuit. Mark Brasfield believes the disad­ usual example of engineering consider­ small. Further, he suggests that a large vantage of placing reactive components ations winning out over marketing con­ coupling capacitor provides a time con­ in a feedback loop outweighs the advan­ siderations. Finally, four words regarding stant of sufficient length to allow a 15 V tages of removing the second amplifier the remote volume control function: do pulse of over a second's duration to ap­ stage. Brasfield'sCD players perform all not use it! These circuits are built to very pear at the output of the CD player if the filtering, including de-emphasis, passive­ low price points with cheap potentiome­ power supply or active circuitry should ly so that no reactive components are in ters and IC-based amplifiers. The line fail. This pulse could be of sufficientam­ the feedback loop. The Phototronics stage in your preamp should offer much plitude and duration to damage a power PA630 chip also uses passive filtering better performance. If this is not the case, amplifieror speaker. and de-emphasis. the purchase of a new preamp might The connection of the output of the The NPC SM5803 and Sony prove worthwhile. DC servo to the analog stages in a CD CXD I244 digital filters include an op­ player presents some difficulty when tional I1R filter to perform the de­ Recommendations compared to its use in a preamplifier or emphasis function. This eliminates the The October annual equipment issue power amplifier. Jung argues that the DC analog components required for the func­ of Audio magazine lists 80-odd different servo should not be terminated at the tion. The disadvantage of performing the manufacturers and modifiers of CD play­ junction between the output of the DAC filtering in the digital domain is that back equipment. Obviously, I have exam­ and the input of the IN converter if the quantization noise is still present at the ined only a small sampling of these ma­ TDA1541 DAC is used [Jung and Chil­ output of the filter, since the signal has chines. This is particularly true in the dress 1988]. Jung reasons that the DC not yet been converted to the analog do­ low-priced category. output drift during DAC warm-up and main. This means that the advantage of One CD player I can recommend is low-frequency noise present at the output the de-emphasis circuit is not fully real­ the Magnavox CDB-630. The CDB-630 of the DAC will cause problems. In con­ ized. The problem becomes less sig­ has a list price of $399.95 but is heavily trast, Precision Audio and Philips, among nificant for DACs with 18-bit linearity. discounted. This machine offers a mid­ others, terminate their servo at this junc­ The final design consideration is the grade (A) selection of the Philips tion and have reported no problems with muting circuit at the output of the CD TDA1541, a ±15 V power supply, and a the DC servo circuit. player. This circuit prevents large voltage well-designed CDM-4 disc transport. A CD may be encoded with digital pulses from appearing at the output of While the CDB-630 is a good value at its data either flat in frequency response or CD player as it is powered on and off. price point, the plastic chassis, single­ emphasized with a high-frequency boost. These pulses can destroy an amplifier or sided phenolic printed circuit boards, ex­ The latter encoding requires a de­ speaker if they are not suppressed. The tensive use of surface-mount compo­ emphasis filter at the filter stage so that muting circuit places a short across the nents, and an abundance of plastic com­ the disc is played back with a flat fre- output of the CD player. This can be ac- ponents in the CDM-4 are indications 28 I

MIGHT AS WELL BE SH OPPI NG .. LE FOR RAZOR BLADES . GI'- Ur.\6

You CALL THIS AN AlBUN\ LINER 7 WHER.E AI<..E: l"\-IE OKA.Y, LET'.5 TR.'{ PIC.TUR.ES? I OUT t{\Y NEW THOUGHT tD GET CD PLAYER. A POSTER ... GRUMBI-E: GRLlt<\Bt-E:

29 that the reliability of this player will not which replaces the digital filter with the ridian 203 decoder box ($990.00). The match that of the more expensive units. NPC SM5813 and the DAC with the An­ unit was introduced too late to be re­ The CDB-630 is a favorite of the CD alog Devices AD1860-K. At $300 this viewed along with the 208, but at least it modifiers. Some modifiers replace the upgrade is a good value for owners of is more realistically priced. The analog DAC with the Sl selection grade (and these older CD players, provided the section is similar to that of the 208 and some do not!), add new power transform­ transport mechanism is in good condi­ thus lacks the level of sophistication ers, and nearly rebuild the analog sec­ tion. Given the quality of performance fo und in the Aragon and PS Audio units tion. The result is often a player with a available from the PS Audio "Digital (the NE5534 op amp is used, for exam­ four-digit price tag. Given the quality of Link" (see below), I suggest that you ple). Also the 203 uses the obsolescent construction and mission of the CDB- spend no more than $400 on the modifi­ SAA7321 DACs. The only circumstance 630, I recommend you avoid these cation of a CD player. under which I would recommend this modifications and purchase a better­ CD players using the Burr-Brown unit would be the availability of an up­ constructed CD player. Another low­ PCM58P or PCM61P in its selected grade path to the SAA7350 (or at a mini­ priced CD player which can be recom­ grade should also be considered. The mum the SAA7323). To my knowledge, mended is the Rotel RCD855. This unit Onkyo DX-7500 ($700.00), Pioneer such a program has not been announced. uses the same basic parts as the Magna­ PD-71 ($850.00), and Sony CDp· At a higher price point, you might vox CDB-630, but according to the com­ 608ESD (900.00) are all excellent consider examining the new Theta DS pany it has upgraded power supply regu­ players and highly recommend. Each of Pro Basic decoder ($2000.00) and the lators, passive components, and these units has distinct advantages and Wadia DigiMaster X-32 ($1995.00) . improved op amps in the analog section. disadvantages, which are discussed in Both decoders use a digital filter imple­ The RCD855 has fewer features than the full reviews in this issue. [A gain, succes­ mented on a general-purpose DSP chip CDB-630 and lists for $349.00. Rotel de­ sor models exist in the case of the Pio­ set. Both represent good value given the clined to send the unit for review. A final neer and the Sony, but the similarities sophistication of the technology used in unit to consider in this price range is the are much greater than the differences.­ these units. The "Pre" version of the HarmaniKardon HD7500, which sells Ed.] These units, for a variety of techni­ Theta receives a full review in this issue. for $449.00. [However, the Ma rk II ver­ cal reasons, may be more difficult to Wadia declined to send the unit for re­ sion, which may be the only one avail­ modify than the Philips CD-80. If you view. The Wa dia unit uses a time-domain able by the time you read this, costs $80 are contemplating having the units algorithm. This algorithm causes a high­ more.�d.] A complete review of the modified at a later date, you might want frequency roll-off and results in a high Harman/Kardon HD7600, which is iden­ to talk to the modifier before purchasing level of out-of-band energy at the unit's tical to the HD7500 except for features, the base unit. You should also look at the output. Others' reviews of the Wad ia is included in this issue. CAL Icon at $750.00. According to the have shown very nonlinear DACs. The The Philips CD·SO ($799.95) is a manufacturer this unit is DC coupled and principal advantages of the DigiMaster much better value than any modified uses the PMI OP42 op amps. I have not X-32 are the excellent jitter performance Magnavox can offer. The best selection examined this unit and cannot make a of the RockLok clock recovery circuit grade (Sl) of the TDA1541A is com­ firmreco mmendation. and the ability to reconfigure the glue bined with an all-discrete power supply The recent release of separate digi­ logic and SPDIF decoder with PROM regulation system. The unit uses a metal tal-lo-analog converter boxes is very chips. Overall, I think the disadvantages chassis and the CDM-1 Mk II transport, good news. Complete reviews of the Ar· are greater than the advantages. which has far fewer plastic components. agon D2A and PS Audio "Digital Link" Given the steady rate of improve­ A full review of the CD-80 is in this is­ are in this issue. I can recommend both ment in CD player technology, I cannot sue. Most modifiers give nebulous ex­ units. Yo u should not choose the Aragon recommend spending more than $2000 cuses for their preferences in using the unit if you are using a passive preamp for any CD player or decoder. Magnavox CDB-630 rather than the Phil­ (see the full review). Bolh Aragon and ips CD-80 as the fo undation of their PS Audio have made provisions in their Acknowledgement modifications. Yo u might consider buy­ designs to allow updating to new DACs Thanks are extended to Dr. Stanley ing a CD-80 if $800 is the maximum you and filter chips. The ability to update Lipshitz and to Rex Nathanson for their want to spend on a CD player, and later these boxes is an advantage over a high­ careful review of the manuscript and go to a modifier service to have it up­ end all-in-one CD player. (Aragon and their perceptive commentary. graded when your cash flow position im­ PS Audio have not yet announced an up­ proves. I think this two-step purchase ap­ grade to the Burr-Brown PCM63P or An­ References proach is the best reason to consider alog Devices AD1862. I hope these com­ Ainslie, A. "What Is Noise Shap­ buying a modified player over one that panies will keep the promise to upgrade ing?" Hi-Fi News & Record Review has been designed from scratch. The their products.) The major disadvantage (February 1990): 45. modifier offers you a CD player in the of a decoder box is increased time-base Allgaier, Jr., J. "Greening Magnavox $1000 to $1500 price class even if you jitter from the SPDIF interface. 14-Bit CDs." The Audio Amateur 19.4 cannot afford it in one purchase. Yo u I could not evaluate the Proceed (October 1988): 19-23. might also consider having a modifier PDP decoder box ($1295.00) because the Ardalan, S. and J. Paulos. "An Anal­ such as Paul McGowan Designs or Preci­ manufacturer, Madrigal Audio Laborato­ ysis of Nonlinear Behavior in Delta­ sion Audio modify your current CD ries, refused to a send a schematic (or, in­ Sigma Modulators." IEEE Transactions player provided it is well-built and has a deed, a unit for review). Based on the on Circuits and Systems CAS-34 (June linear DAC. The discontinued Philips limited information I have, I cannot see a 1987): 593. CD960 and Sony CDP-910, 705ESD, justification for spending $3"00 more than Borbely, E. "A Moving Coil Pre­ 605ESD, and SQSESD are also good can· the price of the Aragon or $496 more amp." The Audio Amateur 18.1 (January didates. If you have an older Philips­ than the price of the PS Audio on this 1987): 30. based player with a 14-bit DAC (CAL unit. As for the ergonomic problems of Borbely, E. "A Multi-Tone Inter­ Te mpest I, Kinergetics KCD-20A, etc.), the Meridian 208 (see the full review in modulation Meter." Th e Audio Amateur Digital Upgrades offers a modification this issue), they do not exist in the Me· 20.2 (April 1989): 7-15; 20.3 (August 30 mRRTln-LOGRn LT D the electrostatic loudspeaker technology co. p.o. box 74 1 200 1 delaware street lawrence. kansas 66044 913.749.0 133

31 DORIAN RECORDINGS is proud to present Jean Guillou in the first recording of the new van den Heuvel organ of St. Eustache. Inaugurated at a gala concert on September 21st, 1989, the Great Organ of St. Eustache is already being hailed as one of the greatest instruments in the French symphonic organ tradition. At 147 ranks, it is also considered to be the world's largest mechanical-action pipe organ.

In The Great Organ of St. Eustache, Paris - Inaugural Recording IDOR-90134j, Jean Guillou recreates the program of the gala dedicatory concert. It is a thrilling 75-minute melange of Bach, de Grigny, Mozart, Liszt, Widor and Guillou, thoroughly exploring the virtually unlimited resources of this majestic instrument. Jean Guillou's tour de fo rce performances combine the artistry, imagination and stunning virtuosity that have made him one of the legendary figures of the organ.

Critical A.cclaimfo r Jean Guillou on Dorian

" ... his rhythmic elan and digital dexterity are, in a word, awesome ... the finest organ recording I have heard to date." -Fallfare

"Jean Guillou brings clarity, rhythmic vitality and imaginative registrations to the music ... " -Stereo Review

"electricity, virtuosity and genius, ... consistently entertaining." -Americall Record GlIide

"The most impressive organist ...that I have ever heard ...is Jean Guillou ...his virtuosity can only be described as outrageous; his temperament is fiery... " -The Alldio Critic

"I have collected organ discs and tapes for 32 years and have heard nothing that approached this recording in both the power of its pedal bass and the bright singing quality of its higher pipes." -Newhollse News Services

(1 17 State Street, Suite 2£ Troy, NY 12180 (518) 274-5475 DORIAN RECORD[NG� for hard Bach youneed a Boulder. From rolling overtures to smashing finales, our audio products will get the maestro in you moving. Ta ke the Boulder 500AE Power Amplifier. It s engineer­ ing breaks new ground in sonic clarity, so you'll hear every note on a grand scale. And because it sounds magnificent in both balanced mono and stereo, you can orchestrate your system one step at a time. Like all our components, the Boulder 500AE conducts a solid performance you can take for granite. Audition it soon at a dealer near you.

AMPLBoulIFIERS,der INC. · There's no stopping a Boulder.

4850 Sterling Drive • Boulder, Colorado 80301 • (303) 449-8220 ... is part of the Premier System One. Audition in your home.

For details, call 1-800-346-9183 or write Audio Concepts, Inc., 901 South Fourth Street, La Crosse, WI 54601. FAX 1-608-784-6367. O NNOUNCING THE BRYSTON TWE NTY

YEAR WA RRANTY

For over a quarter-century Bryston has been committed to designing and producing audio products with musical accuracy, reliability and value as our primary focus. It is widely known Bryston 12B pre-amplifier that Bryston's policy on the warranty of our products has always been extremely generous if We discovered that some parameters of ever required. To fu rther enhance our long term transistors must be controlled as much as 100 commitment Bryston is instituting a 20 year times more closely before their contribution to warranty program as of January 1st, 1990. This, audible distortion is rendered negligible. as far as we know, is a first in our industry and Each of the steps or stages in every Bryston as such will further demonstrate our continuing amplifier, from the input section to the output dedication to our products and customers. section, without exception, are designed to

Musical accuracy is reflected throughout all optimize the musical experience. Bryston takes

Bryston power amplifiers. This includes the very seriously the correct functioning and long necessity for wide-band transient accuracy, open term reliability of its products. loop linearity ahead of closed loop This new twenty year warranty is also

retroactive. It includes all audio products

previously manufactured and sold under the

Bryston name. This warranty is also fu lly

transferable from first owner to any subsequent

owners.

Bryston has always been dedicated to

designing and producing audio power

amplifiers, crossovers, and pre-amplifiers that

Bryston lOB electronic crossover deliver uncompromised performance,

outstanding reliability and exceptional value. specifications, and power supply design as an We believe our new 20 year warranty is one integral part of the overall sonic and electrical more example of our continuing commitment to performance of a power amplifier. this ideal.

We have found that a simple carbon film Bryston Marketing Ltd. resistor can contribute more static distortion to a Tel: (416) 746-0300 Fax: (416) 746-0308 signal than the entire circuitry of the amplifier. Brystonvermont Ltd. Te l: (802) 223-6159

1989): 13-29. Proceedings of the AES 7th International 1989) : 969. Breakall, J., R. Simons, and F. Me­ Conference: Audio in Digital Times McGillen, C. and G. Cooper. Con­ rat. "Measuring Power Supply Output (May 1989): 45-62. tinuous and Discrete Signal and System Impedance." Th e Audio Amateur 14.2-3 Fourre, R. D. "Testing 20 Bit Audio Analysis, 159. Holt Rinehart and Wins­ (April and June 1983). Digital-to-Analog Converters." Th e Pro­ ton, 1974. Burr-Brown IC Data Book. Vo l. 37. ceedings of the AES 7th In ternational Miller, P. "Resonance and Repercus­ Burr-Brown Research Corp., 1989. Conference: Audio in Digital Times sions." Hi-Fi News & Record Review Carley, L. "An Oversampling Ana­ (May 1989): 79-80. (June 1989): 35. log to Digital Converter Topology for Gardner, F. Phase lock Te chniques. Millman, J. Microelectronics, Chap­ High-Resolution Signal Acquisition Sys­ 2nd ed., Chapter 6. John Wiley and Sons, ter 12. McGraw-Hill, 1979. tems." IEEE Transactions on Circuits 1979. Moses, R. W. "Improved Signal Pro­ and Systems CAS-34 (January 1987): 83. Goodenough, F. "Design Innova­ cessing for Compact Disc Audio Sys­ Cezanne, J. and A Papoulis. "The tion." Electronic Design (16 April 1987): tems." MONTECH '87 Proceedings: Use of Modulated Splines for the Recon­ 59. Conference on Communications (9-11 struction of Band-Limited Signals." Goodenough, F. "New Processes, November 1987): 203-11. IEEE Tr ansactions on Acoustics, Speech, Designs Boost IC Op Amp Speeds." Naus, P. J. A, E. C. Dijkmans, E. F. and Signal Processing 36.9 (September Electronic Design (12 April 1989): 45. Stikvoort, et al. "A CMOS Stereo 16-Bit 1988): 1521. Goudie, A "Idle Tones in Oversam­ Converter for Audio." IEEE Journal of Cherry, E. M. "Feedback, Sensitivi­ pled ADCs." 87th Convention of the Solid-State Circuits 22.3 (June 1987): ty, and Stability of Audio Power Ampli­ AES, New York, NY (18-21 October 390. fiers."Journal of the Audio Engineering 1989): Preprint 2882. Naus, P. J. A and E. C. Dijkmans. Society 30 (May 1982): 282-94. Harris, S. "The Effects of Sampling "Low Signal-Level Distortion in Sigma­ Cherry, E. M. "Amplitude and Phase Clock Jitter on Nyquist Sampling Ana­ Delta Modulators." 84th Convention of of Intermodulation Distortion." Journal log-to-Digital Converters and Oversam­ the AES, Paris, France (1-4 March of the Audio Engineering Society 31 pled Delta Sigma ADCs." 87th Conven­ 1988): Preprint 2584. (May 1983): 298-304. tion of the AES, New York, NY (18-21 Nethisinghe, S. S. Introduction to Clark, B. "DC Servo Loop Design October 1989): Preprint 2844. Bit Stream AID, D/A Conversion. Philips for Audio Amplifiers." The Audio Ama­ Howe, "Stretching Marsh's Pre­ Components, 1988. T. teur 13 (August 1982): 14. amp." Th e Audio Amateur 20.1 (Febru­ Otala, M. and R. Ensomaa. "Tran­ Colloms, M. "Wad ia 2000 and 1000 ary 1989): 29-37. sient Intermodulation Distortion in Com­ Processors." Hi-Fi News & Record Re­ Jensen, D. "1E-990 Discrete Opera­ mercial Audio Amplifiers." Journal of view (December 1989): 49-53. tional Amplifier." Journal of the Audio the Audio Engineering Society 22 (May Colloms, M. "Krell Digital MD-l/ Engineering Society 28 (January/Febru­ 1974): 244-46. SBP-64X/SBP-16X." Hi-Fi News & ary 1980): 26-34. Otala, M. "Feedback-Generated Record Review (June 1990): 58-63. Jung, W. and R. Marsh. "Picking Phase Nonlinearity in Audio Amplifiers." Cordell, R. R. "Another View of Capacitors." Audio 64.2-3 (February and 64th Convention of the AES, London, TIM." Audio 64.3 (March 1980): 39. March 1980). England (25-28 February 1980): Preprint Cordell, R. R. "Phase Intermodula­ Jung, W. "Op Amp Meets CD." The 1576. tion Distortion Instrumentation and Mea­ AudioAmateur 17.3 (August 1986): 7. Papoulis, A Circuits and Systems: A surements." Journal of the Audio Engi­ Jung, W. Audio IC Op Amp Applica­ Modern Approach, Chapters 6.3, 7.4, and neering Society 31 (March 1983): 114- tions. 3rd ed. Howard W. Sams, 1987. 8.3. Holt Rinehart and Winston, 1980. 23. Jung, W. and H. Childress. "POOGE- Papoulis, A Probability, Random Didden, J. M. "Wideband Power 4: Philips!Magnavox CD Player Mods." Va riables and Stochastic Processes, Supply." The Audio Amateur 18.1 (Janu­ The Audio Amateur 19.1 (February Chapters 11.2 and 13. McGraw-Hill, ary 1987): 22. 1988): 7-17; 19.2 (May 1988): 19. 1984. Didden, J. M. "A Simple, High­ JVC. "K2 Interface ." JVC Company Pryce, D. "Audio DACs Push CD Quality CD Output Amp." The Audio of America, 1989. Players to Higher Performance." EDN (7 Amateur 20.2 (April 1989): 26-31. Lee, K. and R. Meyer. "Low­ December 1989): 112. Dijkmans, E. C. and P. J. A. Naus. Distortion Switched-Capacitor Filter De­ Rumsey, F. "AES/EBU Interface "The Next Step Towards Ideal and IEEE Journal of Solid­ Studio Sound NO sign Te chniques." Comes of Age?" (Decem­ D/A Converters." Th e Proceedings of the State Circuits 20.6 (December 1985): ber 1989): 29 AES 7th International Conference: Audio 1103. Stikvoort, E. F. "Higher-Order One­ in Digital Times (May 1989): 97-103. Lipshitz, S. P. and J. Va nderkooy. Bit Coder for Audio Applications." 84th Duncan, B. "Supertuning CD." Hi­ "Are D/A converters Getting Worse?" Convention of the AES, Paris, France (1- Fi News & Record Review (November 84th Convention of the AES, Paris, 4 March 1988): Preprint 2583. and December 1987, January 1988, and France (1-4 March 1988): Preprint 2586. Te xas Instruments. Linear Circuits: March 1989). Marsh, R. N. "Power Up: An Over­ Data Acquisition and Conversion. Data Duncan, B. "Evaluating Audio Op view of Power Supply Considerations." Book, Vo lume 2. Te xas Instruments In­ Amps." Studio Sound (July, August, and The Audio Amateur 14.4 (September corporated, 1989. September 1990). 1983): 16. Va nderkooy, J. and S. P. Lipshitz. Duncan, B. "Building Bitstream." Marsh, R. N. "Low-Distortion, "Dither in Digital Audio." Journal of the Hi-Fi News & Record Review (Septem­ Low-Feedback Power Amplifiers." The Audio Engineering Society 35 (Decem­ ber and October 1990). Audio Amateur 16.3 (July 1985): 24. ber 1987): 966-75. Evans, S. "Design Entry." Electron­ Matsuya, Y., K. Uchimura, A Iwata, Va nderkooy, J. and S. P. Lipshitz. ic Design (28 November 1985): 125. and T. Kaneko. "A 17-bit Oversampling "Digital Dither: Signal Processing with Fielder, L. D. "Human Auditory Ca­ D-to-A Conversion Te chnology Using Resolution Far Below the Least Signifi­ pabilities and Their Consequences in Multistage Noise Shaping." IEEE Jour­ cant Bit." The Proceedings of the AES Digital Audio Converter Design." Th e nal of Solid-State Circuits 24.4 (August (continued on page 40) 37 Ta ble 1: Design Features of CD Players and Separate D/A Converters

SPDIF Digital D/A Interpolation Sample and Analog I to V Decoder [1] Filter Converter Rate Hold Circuit Stages Converter

Aragon D2A YM3623B + Sony Burr-Brown 8x No 2 (discrete) Voltage 2addnl PLLs CXDl144 PCM58P-J feedback CAL Icon NA NPC Burr-Brown 8x No 2 (i ntegrated) Voltage SM5813 PCM61P feedback CAL Te rcet NA NPC Burr-Brown 8x No 2 (discrete) Mk lll SM5813 PCM61P Carver NA Yamaha Burr-Brown 8x Yes 5 (integrated) Voltage TL-3220 YM3414 PCM61P feedback Denon Yamaha Sony Burr-Brown 4x Yes 2 (integrated) Voltage DAP-2500 YM3623B CXD1162 PCM56P-J [3] fe edback Denon NA NPC Burr-Brown 8x Yes 3 (i ntegrated) Vo ltage DCD-1560 SM5813 PCM1701KP feedback Harman/Krdn NA Included in NIT NA NA 3 (discrete) NA HD7500L7 600 Bitstream DAC MN6471M JVC NA Yamaha/JVC Burr-Brown 8x Yes 2 (integrated) Voltage XL-ZIOIOTN YM3414/K2 PCM56P fe edback Kenwood NA NPC/Kenwood Burr-Brown 8x No 3 (i ntegrated) Voltage DP-8010 SM5813LQroQr PCM58P feedback Kinergetics NA Sony Analog Devices 8x No 2 (composite) Voltage KCD-40 CXDl144 ADI860N-K fe edback Krell SBP-64X Proprietary 4 Motorola Burr-Brown 64x Yes 1 (discrete) Voltage 56001 [4] PCM64 fe edback Magnavox NA Philips Philips 4x No 2 (integrated) Voltage CDB630 SAA7220PLB TDA1541A feedback Meridian 208 NA Included in Philips NA NA 2 (i ntegrated) NA Bitstream DAC SAA7321 [5] Onkyo Integra NA Yamaha Burr-Brown 8x No Choice of 1 or Voltage DX-7500 YM3414 PCM58P-K 2 (integrated) feedback Philips CD-60 NA Philips Philips 4x No 2 (integrated) Voltage SAA7220PLB TDA1541A feedback Philips CD-80 NA Philips Philips 4x No 2 (integrated) Voltage SAA7220PLB TDAI541A-Sl feedback Philips Sony Philips Philips 4x No 2 (integrated) Voltage DAC960 CXDI076 [6] SAA7220PLB TDAI541A-Sl feedback Pioneer NA NPC Burr-Brown 8x No 3 (2 composite Voltage Elite PD-71 SM5803 PCM58P-K & 1 integr ated) feedback Proceed PCD NA NPC Burr-Brown 8x No 2 (composite) Voltage SM5813 PCM61P feedback PS Audio Ya maha Yamaha Burr-Brown 8x No 1 (composite) Proprietary "Digi tal Link" YM3623B YM3434 PCM61P-K Sansui YM3623B + Included in NIT NA NA 4 (integrated) NA AU-X911DG 2nd PLL Bitstream DAC MN6471M Sony DAS-Rl Sony Sony CXD1144 Philips [8] 8x No 2 (1 integrated Voltage CXD1076 [6, 7] & CXD 1329 TDAI541A-Sl & 1 comQQsite) fe edback Sony NA Sony CXD1244 Burr-Brown 8x No 2 (integrated) Voltage CDP-X7ESD & CXD8003 PCM58P-S feedback Sony NA Sony CXDI244 Burr-Brown 8x No 2 (integrated) Voltage CDP-608ESD & CXD8003 PCM58P-J feedback Sony NA Sony CXD1244 Burr-Brown 8x No 2 (i ntegrated) Voltage CDP-508ESD & CXD8003 PCM58P fe edback Theta DS Pro YM3623B + 2 Motorola Analog Devices 8x No 2 (integrated) Voltage 2addnl PLLs 56001 ADI860N-K feedback Theta YM3623B + 2 Motorola Analog Devices 8x No 2 (integrated) Voltage DS Pro Basic 2addnl PLLs 56001 AD1860N-K feedback Wadia 2000 Proprietary FLL 4 AT&T Proprietary 64x No 2 (hybrid) Voltage (V COLc rys tal) DSP-16 feedback Wadia Digi- Proprietary FLL 4 AT&T Proprietary 64x No 2 (hybrid) Voltage Master X-64 (V COLc rys tal) DSP-16 feedback Wadia Digi- Proprietary FLL 2 AT&T Proprietary 32x No 2 (integrated) Voltage Master X-32 (YCOLc rys tal) DSP-16 fe edback Ya maha Ya maha Ya maha Burr-Brown 8x Yes 3 (integrated) Voltage CX-loOO YM3623B [9] YM3414 PCM56P-K [1 0] feedback Ya maha NA NPC Burr-Brown [3] 8x No 5 (integrated) Voltage CDX-1120Ti SM5813 PCM58P-J [10] [1] Applicable only to outboard con- smaller number than complete CD players [4] General-purpose DSP (digital sig- verter boxes and digital preamplifiers. or digital preamplifiers). nal processor) with custom software. [2] To tal number of regulators in unit [3] Balanced operation with 2 DACs [5] A proprietary configuration allows (outboard converters naturally require a per channel . one SAA7321 to be used in each channel .

38 (Blank space signifies unavailable data. See review section for successor models, if any. )

Filter Output Numberof Optical Regulator Dual Analog Sect Price Order Coupling Transformers Couplers Stages [2] Mono Pwr Sup V $

3rd order Capacitor 1 No 11 Yes :t18 995.00 Bessel 3rd order Capacitor 1 No 5 No 750.00

3rd order DC servo 2 Yes 23 1295.00

3rd order Capacitor 1 No 5 No :t 8 529.00

3rd order DC servo 1 No 8 No :t15 1000.00 Butterworth Passive (order Capacitor 1 No 7 No :t12 650.00 unknown) 3rd ord Btrw + Capacitor 1 No 7 No :t12(fil ter) 449.00 3rd ord Qu lse int 599.00 3rd order Capacitor 1 No 10 No :t10 700.00 Butterworth 3rd order Capacitor 1 No 7 No :t15 650.00 Butterworth 2nd order DC servo 2 No 15 No :t18 2295.00 Butterworth 1st order DC servo 3 No 10 (not sure) 8950.00

3rd order Capacitor 1 No 4 No :t15 399.95 Bessel 2nd ord Btrw + DC servo No 5 No + 5 (pulseint) 2950.00 3rd ord Qu lse int + 15 (fi lter) Choice of 1st DC servo 2 Yes 7 No :t15 700.00 or 3rd order 3rd order Capacitor 1 No 4 No :t15 429.95 Bessel 3rd order DC servo 1 No 12 No :t15 799.95 Bessel 3rd order Capacitor 3 Yes 7 No :t15 999.00 Bessel 3rd order Capacitor 1 No 7 No :t15 850.00 Butterworth DC servo 1 No 11 1650.00

1st order Direct (with DC 1 No 6 Yes :t15 799.00 offsetcanc Qot) 1st order + 3rd Capacitor 1 No 7 No :t15 1100.00 order Qu lse int 3rd order Capacitor 2 No 7 No :t15 8000.00 Butterworth (w ith CDP-R1) 3rd order Capacitor 2 No 10 No :t15 2000.00

3rd order Capacitor 1 No 5 No :t12 900.00 Butterworth 3rd order Capacitor 1 No 3 No :t 5 550.00 Butterworth 1st order DC servo 4 No 13 Yes :t15 3500.00

1st order DC servo 2 No 7 No :t15 2000.00

RF filtering Direct (with DC 2 No Yes :t15 7995.00 only offset canc Qot) RF filtering Di rect (with DC 2 No Yes :t15 4995.00 only offsetcanc Qot) RF filtering Direct (with DC 1 No Yes :t 8 1995.00 only offset canc Qot) 1st order DC servo + 2 No 18 Yes :t12 1199.00 caQacitor Choice of 1st or Capacitor 2 No 5 No :t12 1199.00 3rd order Btrw [6] Only the PLL phasedetector is in- [8] Staggered configuration with 2 used. Crystal oscillator is stopped when tegrated on this chi p. DACs per channel provides 2x increase in phase lock is achieved. [7] Clock signal is internally generat- conversion rate. [10] Bit shifting circuit is used after ed when used with Sony CDP-R1 transport. [9] External adaptive PLL filter is the DACs.

39 Ta ble 2: Design Features of Modified CD Players

Models D/A Converter Analog I to V Filter Modified Replaced Stages Converter Order

Euphonic Magnavox CDB-650 Yes, with Philips 2 (integrated) Voltage feedback 3rd order Bessel Te chnology & most other TDA1541A-S1 + Mk II Signature Philips-based players separate transformer Arpeggio in DAC pow er supply MSB Technology Magnavox Silver 630 CDB-630 No 2 (composite) Transimpedance 1st order Gold CDB-582 Yes(T DA1541A-S1) 2 (composite) Transimpe dance 1st order Paul McGowan Any player with No 1 (composite) Proprietary 1st order Design s >±12 V supplyrails Precision Audio Any player with No 1 (discrete) Transimpedance 2nd order DVIC PCM58 or TDA1541 = 0.5) (0 DACs and >±12 V supply rails, as long as mod fitschassis

Ta ble 3: Comparison of 8 Times Interpolating Digital Filters

NPC SM5803 NPCSM5813 Sony CXDl144 Sony CXD1244 Ya maha YM3414 YM3434

Passband Ripple ±0.00005 dB :0.00005dB :0.000005dB :0.00001 dB :0.00005 dB

Passband Frequency :0.00005 dB :0.00005 dB +0, -0.003 dB :0.00001 dB :0.00005 dB Response

Stopband Attenuation 24.1-150 kHz -115 dB -115 dB -120 dB -100 dB -100 dB 150 - 180 kHz -115 dB -115 dB - 70 dB -100 dB -100 dB

Noise Shaper Selectable No No Selectable No

Digital De-emphasis Yes No No Yes No

DAC DC Offset No No No Selectable No

N umber of DSP 1 1 2 1 1 Processors

FIR Thp Length 199 199 293 Not available 287 from manufacturer

Coefficient 22 bits 22 bits 28 bits Not available 19 bits Word Length from manufacturer

Data Path Length 20 bits 20 bits 22 bits Not available 18 bits from manufacturer

Accumulator Size 25 bits 25 bits 32 bits 45 bits 18 bits

References (continued from page 37) 7th International Conference: Audio in tion of the AES, New York, NY (18-21 13.10 (October 1990): 59. Digital Times (May 1989): 87-96. October 1989): Preprint 283 1. Williams, J. "Designer's Guide to Wadsworth, D. C. "A Professional Willenswaard, P. van. "Industry Up­ Op-Amp Booster Stages." EDN (29 May Audio Integrated Circuit." 87th Conven- date: The Netherlands." Stereophile 1986): 131. 40 (Blank space signifies unavailable data.)

Output Replace Filter Dual Output Muting Price of New Price to Modify Coupling & De-emphasis Mono Function Modified Player Yo ur Player Components $ $

Capacitor (film) Yes, with Philips No Varies SAA7220P/B 1595.00 1200.00

DC servo Yes Removed 1695.00 DC servo Yes Removed 2350.00 Capacitor Yes No Removed 300.00

DC servo Yes No Relay 1300.00 450.00 to (Philips CD-80) 550.00 (depending on player)

Ta ble 4: Comparison of ICs Suitable for Use in CD Players

VOLTA GE FEEDBACK

Settling (ns) Noise Slew Rate Small Signal Output Cur Input Bias Price 0.1% 0.01 % (nVWIIZ (V /flS) Bandwidth for ±10 V Current ($ per chip @lkHz) (MHz) (rnA) (!ANnA/pA) in lots of 100) Analog Devices AD841 90 110 13 300 40 50 3.5 11 6.00 AD845 250 310 25 100 12.8 20 500 3.25 P Burr-Brown OPA627 450" 550" 5.6 135'" 16 45 20 p 7.50 OPA602 600 1000 13 35 6.5 20 lO p 4.50 OPA606 1000 2100 14 35 13 10 lO p 3.85

National LM6361 120 Not specified 15t 300 35 7 3 11 1.91 LF400 200 365 23 27 16 20 200 p 3.95 LM833 Not specified Not speci fied 4.5 7 15 Not specified 500n 1.05 (dual)

PMI OP42 450 1000 13 50 10 10 BOp 2.25

Signetics NE5532 Not specified Not specified 5 9 10 20 200 n 1.44 (dual) NE5534 Not specified Not specified 4 13t/6 10 20 200 n 1.00 "' @AYCL = -1 t@ lO kHz t@AyCL = 3

TRANSIMPEDANCE (CURRENT FEEDBACK)

Settling (ns) Slew Rate Small Signal Output Cur Price 0.1% 0.01 % (V/IlS) Bandwidth for ±10 V ($ per chip (MHz) (rnA) in lots of 100) Analog Devices AD846 80 110 450 46 20 6.25

Burr-Brown OPA603 50 Not specified 1000 55 80 4.95

Elantec EL2020 90 Not specified 500 50 50 3.30

Harris HA5004 50 Not specified 1200 100 95 5.87

PMI OP160 120 155 1000 90 20 4.50

41 Snake Oil and Gimmicks. Quite the contrary. I and other members of the scientific audio community know that just four factors deter­ mine the sonic characteristics of an amplifier:

1. Current output 2. Vo ltage output j,Power output 4. Transfe rfu nction asevidenced �)' the interrelationship offr equency response and Olltput impedance.

These factors transcend the usual trivial debates over tubes vs. solid state, MOS-FETs vs. bi-polar, Class A vs. AB, silver Leitz wiring vs. Thumb through Audio 's Annual Equipment copper, gold-plated front panels, Wo nderCaps Directory and you'll see vivid proof that all and my favorite: hand-ground-open transistors power amplifiers are neither created equal nor filled with a proprietary crystalline substance priced equally. that stops ringing (honest, I'm not kidding!). An Two hundred watts per channel can amp can have any combination of these enter­ cost you as much as or as little as $8,400 taining variables (plus special bricks stacked on 5599. You can own an amp from a multi­ top) and yes, sound wonderful...provided it national mega-manufacturer who also ALSO has high current, voltage and power out; makes TV's, microwaves and cellular put and the correct output impedance. phones. Or an amp from a company Thus the Four Factors explain why expen­ so small that the designer is sive amplifiers generally sound better than also the assembler and cheap amplifiers. But also why that doesn't shipping clerk. necessarily have to be the case. Can it be that ampli­ fiers are sonically FAOORS 1·3: tHE POWERSUPPLY equal' Some seem to BEHIND tHE SOUND have muscular power An amplifier's power supply produces cur­ reserves far beyond rent and voltage. A preponderance of one their FTC-rated out- without the other is meaningless .2 To maxi­ put. Others sound great mize SIMULTANEOUS current and voltage out­ until they're challenged by a dynamic passage put using traditional design approaches costs and then sound like a Buick hitting a row of serious garbage cans. Some are (to indulge in audio­ money. For phile jargon) so "fluid" that you practically need example, we a drop cloth under them. Others seem to sound recently tested a harsh, "metallic" and brittle at any output level. competitor's A casual comparison of perceived sound $2,000 amplifier quality versus price tags may lead to an erro­ that was rated at neous conclusion: that an amplifier must be 20 watts/channel. expensive to sound good. Believe me, from The !nllh is a bit more complicated: a parts and mate­ Cosmetic glitz aside, an amplifier's cost is pri­ rials standpoint, it marily determined by its power supply. 1 In was wonh other words, within reason, you generally do $2,000, with most get what you pay for when you buy a conven­ of that money tional amp design. But the key word here is being spent "conventional. ., on an amaz­ My deCidedly un-conventional Magnetic ingly rugged Field Power Supply is capable of outperforming power sup- conventional power supplies of the same size. ply. Another more Absolute "/(iXilllllm output Result: A Significantly better power amplifier extreme example is value for you. my own ultra-con­ CURRENTI!tlcb challnel Let me explain. ventional Silver Seven Tube amplifier design. Its "money-is-no­ NO MAGIC. JUST FOUR CRITICAl object" power supply helps set the price of a QUANTITATIVE FACTORS. pair of S-7's at around $20,000.00. When I fervently state that " the sound of an Now, since it is universally agreed among amplifier need not be related to its price, " you amplifier deSigners that current/voltage/power might think we're veering off into the land of output directly affects the sound of an amplifier,

42 and since good current and voltage another amplifier design to within 99 parts out traditional levels previously of 100 (a null of 40dB). For example, we've power supplies only found in used Transfer Function Calibration to closely are costly, price extremely ex pen- emulate the sonic characteristics of my refer- and sonic qual­ sive "esoteric" ence Silver Seven in our TFM-45 and TFM-42 ity ARE often designs. Or to solid state designs. In other cases we've used closely related. look at it another the process to simply adjust the sound of an But what if way, in a given amplifier to have pleasant but unique sonic there was a price range (say characteristics: in general, a warm "tube" sound way around the $900-$1,000), with rich, rolling bass and soft yet detailed tre- economic con­ Carver simply gives ble (such as our TFM-22/25 , S-7t and TFM-1S). straints of con- TFM-45; 375 watts RMS/ch. into 8Q you far more for Either way, we use painstaking measurement ventional, inefficient power 20-20kHz with no more than 0.5% THO your money. and adjustment processes to finetune output ($949 sugg. retail.}and TFM- 15; 100 watts . suppIies? What if there was RMS/ch. int0 8Q 20-20kHz with no more Impe d ance/f requency response. Not magic. than 0. 1% THO ($399 suggested retail). ' ,' 1" 1/' 'I .IJI[. '[ )11 a power supply that could It,t�l,l.t!i'I en... 1. '1I • And, needless to say, we start with highly Consider two hypothetical ampli- deliver awesome simultaneous current and volt- . .. capable power amplifier designs before the flers wlth IdentICal power supplies. Same age into real-world speaker impedances with- . Transfer Function Modification process. Powel. . ra t mg; same gam, etc. Y et t h ey Stl out shocking your pocketbook? II sound different when powering identical That's just what my patented Magnetic Field .',;1 1'£.1111h 1;1 (31)31 8.];1 1: 1;1 }.\j 3 1 1319 speaker through IdentICal cables. My Transfer Function Calibrated power Power Supply does. Without gimmicks, mysti- � . Why. A fourth quantifiable factor IS at amplifiers have suggested retail prices of from cism or loss of bass response. Simply put, a work. One that, unltke power supply output, $399 to $1,000. That I even dare to suggest Magnetic Field Power Supply uses progressively IS totally mdependent of economIC constramts. they can sound as good as designs in the more of each line voltage swing as amplifier I've left Factor 4 (transfer function/frequen- $2,000 to $6,000 price range has not endeared power demand increases. It's just plain more response/damping) until last intentionally. me with some audiophiles or underground efficient. How and why this works is explained cy Because until an amplifier can deliver sufficient magazine writers. in our new White Paper called "TheMag netic power wah simultaneous current and voltage That's a real shame, because I have abso- Field Story Parts 1, /J &Ill " which you can get lutely nothing but respect for free by calling 1-800-443-CAVR. (Factors 1-3), transfer func- tion is immaterial. well-made, high-ticket con- Right now, let's consider the tangible bene- + + + ventional amplifiers. Like Frankly, I'm guilty of + fits. The series of comparison charts in this ad + + + + + + : not makmg thiS fully clear . Rolexes and Lamborghini's, shows how my Magnetic Field Power Supply Amplifier with resistor test load m the past. Some readers 1+ I they are a joy to own if you s'JCcessfully challenges the previously hard- may have gotten the can afford them. But just as a and-fast rule that high-performance power sup- ImpreSSion Ihat by magi- - Rolex doesn't tell time any plies must be expensive. Amp X is a highly- � cally adJustmg some + + + + + better than the inexpensive respected solid state ...... + + + arcane parameter called Same amplifier connected to cables and watch I'm wearing right now, design rated at 200 t++ 1+ transfer function, one loudspeaker. good sound does not neces- watts into 8 ohms. It could cause a sarily have to be costly. cost $S,500. My somehow cheap amp to sound ltke an expensive one. Nothing could be fur- If this concept intrigues you, please visit a TFM-45 is rated at ther from the truth. If there's no guts (power Carver dealer soon. Bring demo material 375 watts per supply), there's no glory (optimized transfer you're familiar with and be willing to do some channel both function). critical listening. Compare my designs to com- channels driven By transfer fu nction, I mean the effect an petition costing about the same amount as well into 8 ohms 20- amplifier's output impedance has on real world as to more expensive models. 20KHz with less frequency response. I don't mean the flat, "DC Yo ur ears alone should be the final arbiter. I than 0_1% THD. It to light" Rated Full Power Bandwidth found in feel confident that you will join the tens of has a suggested retail column 11 of Equipment Directory, thousands of audiophiles who have gotten the of $949. Audio 's which is a resistor best possible value by owning Carver. Even more measured using as a load. Rather, I'm referring to the frequency response impressive is this curve that occurs when an amplifier and same sort of speaker cables interact with a specific speaker. comparison chart with the As distinctive as a fingerprint, this curve Bob Carver, President RNSoulplll TFM-45 vs. other determines the "sound" of each amplifier amplifiers in its own design. Its warmth or harshness. The quality VOLTAGEetlchcJxmllei price range. In defer- of the bass. The definition of its upper regis- ence to how utterly ters. Even the configuration of the stereo we trounce Similarly-priced, conventional com- "sound stage" it can create. My engineering department and I are capa- CARVER petition, we've confined those charts to our CARVER CORP., LY NNWOOD, WA. U.S.A. 1-800-443-CAVR new White Paper. ble of making one amplifier design sound like Distributed in Canada by Evolution Audio Inc. 1-(416) 847-8888

To summarize: Magnetic Field Power 1 My definition of cosmetic glitz i� any part of an ?m,?lifier whose sofe audio contribution is to cause one's friends to go, "Ooooo.l!" when they see o;e new purchase. My own Silver Seven amplifiers hand-rubbed plano lacquer and salrd. granite surfaces meet this definition. Supply technology allows reasonably-priced � Smce power (watts) equals voltag� times curren t, the same wattage can represent Significantly different combinations of voltage and current - power amplifier designs to deliver simultaneous and thus verydiffer ent performance Into the same load

43 BREAKING ALL BA RRIERS...

THE REVOL UTIONARY NE W FE T-10 PHONO CA RTRIDGE

At Win Research Group, we've found that brilliant engineering often begins with creativity and fundamental research. Ye ars ago, when we got interested in th e art of phono transducers , we figured very innocently, that the way to play a phonograph record was to measure the ripples in the groove by assigning a bias current to the cantilever position - while everyone else was derivi ng a voltage signal from the canti lever velocity. Why this complexity, we asked . Why not do it simply?

Now, after six years of research, we can abandon the ex­ hausted technology of magnetic generators . We changed everything, the operati ng principle, support electronics, the stylus shape - everyth ing.

Fundamental research can make a difference: while others play end-games with jewelled to p plates, signature models etc., we went ahead and created a precision measuring instrument. The Win FET-1 0 - the last phono cartridge that you will ever need to buy.

WIN RESEARCH GROUp, INC.

7320 HOLLISTER AVENUE, GOLETA, CA LIFORNIA 93 117 FA X: 805-685-278 1 TEL : 805-968-52 13 44 co·da \'kodd\ n -s [It, lit., tail, fro L coda , cauda ] : a final or concluding portion of a musical or dramatic work ;

usu : a portion or scene that rounds off or integrates preceding themes or ideas

Coda Technologies is a new expression of art and science at a level of refi nement and quality that is rare in the audio industry. The company is the creation of a small group of dedicated engineering audiophiles whose goal is to uti lize the best techniques and materialsthat are available and appropriate to a given design. Providing a foundation for these efforts are over two decades of experience in pursuing a sonic and aesthetic ideal. Coda Technologies products are limited production items, which are available at carefully chosen audio dealers, allowing the company to interface more completely with the audiophile user than would be possible for a company with a more "Mass Market" orientation.

As with any audiophile product the proof is in the listening experience. We encourage you to audition the Coda Te chnologies FET PREAMPLIFIER 01 and realize its superior quality for yourself. For more information about Coda Technologies or its products please call or write us.

E$DI1 CODA TECHNOLOGIES INC. 9233 Wausau Way Sacramento. CA 95826 Business: (9 16) 366-6420

45 WE'VE CREATED A monumEnT TO GOOD sounD.

To a true music lover, nothing is more important than sound quality. For 17 years that's been our specialty We help you select a PETER Mc G RATH'S system that is perfect for you and ideal for your budget. Now with our new super store open­ ing this summer, we're even more dedicated COSOUMPONENTSND to the proposition that good sound is the High performance audio and video. most important thing on the face of the earth. 11927 South Dixie Highway, Miami, FL 33156 AllMajor Credit Cards Accepted. - We Ship Anywhere. 305-232-8848

ACCUPHASE . ARCICI . AUDIO QUEST . BRYSTON • B&W . CLASSE . CHICAGO STANDS . C.w.O. CABINETRY . DUNTECH • ENTECH . GOLDMUND • GRADO . HARMONIA MUNDI . HITACHI

• KINERGETICS • KEF . KOETSU . LEXICON . LINN SONDEK . MADRIGAL . MAGNEPAN . MARK LEVINSON . MERIDIAN . M.I.T. CABLES . MONSTER CABLE . NAD . NAKAMICHI . NILES

• NITTY GRITTY . PROCEED . PROTON . PYGMY . QUAD . REGA . REVOX . ROCKUSTICS • S.M.E.• SONANCE . SOUND ANCHORS . SPECTRAL . SPICA . STAX . STUART SCREENS . SUMIKO

46 • TALISMAN . TARGET . TERA lV.'s . TERK • THIEL . THORENS • TDL . VIDIKRON PROJECTION T.V .•WADIA . WILSON AUDIO . WELL TEMPERED . VAN DEN HULL Current CD Players and D/A Proc essors, New and Not So New, Multibit and One-Bit

By Peter Aczel Editor and Publisher

Inspired by David Rich's formidable treatise, your Editor tests a bumper crop of hardware on the lab bench, in the listening room, and some through the ABX Comparator (if you'll pardon the expression).

My favorite cliche, "the plot thickens," is quite safely What he needs to know, rather, is whether or not the applicable here, since just about every manufacturer capable unit is as well engineered-with regard to basic circuit con­ of mounting a printed circuit board on a chassis has come cept, quality of parts, mechanical construction, control fa cil­ up with CD playback hardware of some sort, more often of ities, and ergonomics-as he has the right to expect for the several sorts, and rival design philosophies are proliferating money he is paying. Significantdifferences exist in those re­ at all price levels, not just the high end. Armed with the spects. What about the sound? I'll be coming to that shortly. technical insights provided by David Rich, I managed to keep my head above the water in the lab, but boy, there are Low-level linearity: not so simple. more difficult choices in this sector of audio tOday than on When the deviation from theoretically perfect gain the menu of a good Chinese restaurant. Let us, therefore, try linearity with a dithered test signal at -70 dB, -SO dB, and to sort out the relevant issues and criteria. -90 dB is small enough to be approaching the accuracy and resolution limits of my instrumentation (see above), I call How they differ, how they don't. the gain linearity essentially perfect, although it can happen All CD playback equipment currently or recently that the spectrum of the signal still shows some harmonic manufactured can be expected to have almost dead flat fre­ blips above the "grass" of the noise floor. Usually, when the quency response, negligible de-emphasis error, stupendous amplitude deviation is a fraction of a dB, no such blips are stereo channel separation, virtually zero phase difference visible. Quite a few units turned out to be that good, even at between channels, extremely low THD at reference level, -90 dB, but a number of those that showed some nonlineari­ insignificant wow and flutter, and nearly always noninvert­ ty led me to a totally unexpected conclusion. ing impulse response. The elaborately presented curves for I now believe, very firmly, that trimmer potentiome­ these measurements in the typical magazine review are ters should be left out of designs using the J-grade and K­ mere gingerbread in my opinion (look, Ma, I have an Audio grade DACs from Burr-Brown and Analog Devices. Those Precision!); my trusty Hewlett-Packard 3580Ahas a guaran­ who use these DACs as is, exactly as they come from the teed amplitude accuracy of ±0.3 dB and a visual resolution fa ctory, get essentially perfect linearity (e.g., Sony and PS of about 0.1 dB, leaving me little or nothing to say about Audio), whereas those who add the trim pots (e.g., Pioneer even smaller differences that are meaningless in any event. Elite and Theta) end up with far from negligible errors in At the same time, there are very real though rarely impor­ actual production samples, correctable of course on a suit­ tant differences in noise floor, dynamic range, 0 dB square ably equipped lab bench but not by the consumer in his wave clipping, RF output, possibly 1M distortion and, most home. In fact, an untouched J-grade appears to be more lin­ significantly, low-level gain linearity and monotonicity. ear than a typical trimmed-in K-grade. The lowest grades, These are worth discussing, but here again I consider a 0.3 dB on the other hand, should always be used with trim pots to deviation from theoretical perfection, or a 0.3 dB superiori­ bring their linearity up to a respectable par, but production ty in device A versus device B, to be quite insignificant and tolerances are even sloppier in the lower-priced CD players, not really what the enlightened audiophile needs to know. so that the end result is generally unpredictable.

47 The irony of it all is that the beautifully plotted low­ margin. And I must add that a Golden Ear who tells me that level linearity curves in all those highly detailed test reports such and such "blows away" thus and so-"night and day" are merely an indication of how carefully Norma tickled and "not in the same league" are in a similar idiom-had and sealed the trim pots on the production line, and how better get 16 right out of 16 or I consider his credibility gently Jose handled the carton in the warehouse, but not of blown. For 12 out of 16, the words should be "maybe a little the inherent linearity of the design itself. They are QC better" or "I can just barely hear it." curves, not engineering test curves. I have learned to take Now for for the qualifications and reservations. It is them cum grana salis. I am aware that Stanley Lipshitz and possible that there exists a piece of music that would have John Va nderkooy already flagged this trimmer pitfall in revealed small differences better than anything used in these their March 1988 paper, but at the time they did their re­ tests. It is also possible that, currently used systems of AID search the alternative of almost perfectly linear DACs right encoding being less than theoretically perfect, some fu ture out of the package did not yet exist. encoder will reveal such small differences regardless of the music used. Remember, some of these differences were Double-blind, matched-level listening tests. measurable. Finally, it is possible that there are some amaz­ The point of any or all of the above is of course the ing Golden Ears out there who would have scored higher in obtainment of the best possible sound. I carefully listened at the tests than any of us. But, please, don't talk to me about some length to every piece of equipment reviewed here, and tremendous differences in sound quality between CD a number of my associates also listened to most (but not all) players of current or recent manufacture because I'll laugh of them. In a good many instances we had two of them lev­ in your face. (And don't hassle me about the ABX method el-matched within 0.1 to 0. 15 dB of each other (meaning all because I'll instantly agree to, say, switching by hand­ four channels) and connected to the ABX Double-Blind plugging and unplugging without relays, or any kind of time Comparator. It would have been impossible, of course, to frame-such as one trial a day for 16 days--or any other ABX each unit against each of the others; the permutations method, as long as you match the levels and are listening to and combinations would have been staggering. We did an unidentified sound when you give your answer. What I enough double-blind listening, however, to come to a tenta­ want the hasslers to explain to me is why every Golden Ear tive conclusion that will not please the tweaks and cultists, I or my associates have tested suddenly turns to tin when or the major-brand lobbyists for that matter. There are sim­ the levels are matched and the brand names are withheld. If ply no reliably identifiable differences in sound between any you don't have an answer to that, please don 't bore me with two units in this group, regardless of price, reputation or ecclesiastical arguments about side issues.) measured performance. Those who can handle that without freaking out and Aragon D2A showing me the door (as in "Get lost, Julian!") will want to know the details. The listening setup consisted of a pair of Mondial Designs Limited, 2 Elm Street, Ardsley, NY 10502. D2A Quad ESL-63 USA Monitors driven by a Boulder 500AE D/A converter, $995. 00. Tested sample on loan /rom owner. power amplifier, which was fed from the balanced outputs of a Boulder MS preamplifier. In some of the listening tests This is a gorgeous piece of equipment, somewhat a Ve lodyne ULD-15 Series II subwoofer was added to the handicapped by a wrongheaded decision regarding the ana­ Quads. The room was the one analyzed in Bill Rasnake's ar­ log output interface. The output stage is of the common ticle in Issue No. 13 All of the participants in the tests were emitter configuration, with a very high output impedance, experienced audiophiles with no hearing impairments. The unsuitable for driving a low-impedance volume control (as typical test between a given A and given B consisted of 16 in a passive preamp or input-attenuated power amp) and trials, i.e., 16 successive randomized X's, with unlimited also a poor match for high-capacitance interconnect cables. time available for each X and backtracking/rechecking per­ A properly designed active preamp will wash out these in­ mitted at all times. Some of the tests were stopped at 12 compatibilities-but what if you have, or want, one of those trials. The participants listened one by one, not in groups; trendy passive preamps, or no preamp? The fu lly discrete the typical time they needed for 16 trials was about 60 to 75 analog section (by itself a noble idea) has the fu rther small minutes, some of which went into synchronizing the two boo-boo that the differential pairs are biased by a resistor identical discs (except when only one disc was needed to which causes the bias current to vary in the presence of a test a CD player straight through versus an outboard D/A common-mode signal, resulting in increased distortion. A processor). The music was quite varied-symphony orches­ minor engineering simplism. tra, piano, soprano voice, string quartet, jazz, rock-but of Other than that, the D2Ais a most attractive buy. You course there was no way to include each listener's favorite take off the cover and you can see where the money went; test CD. Statistically, 12 right answers out of 16 and 10 out the parts and construction are of near-military quality, and of 12 are needed for 95 % confidence that no lucky guessing the digital circuitry is highly sophisticated, with no chintzy is involved. Nobody came even close to that; in fact nobody solutions apparent in the details. The digital filter is the very exceeded a fifty-fifty score by a statististically meaningful expensive and complex Sony CXD1144BP. I don't even

48 quite understand how Mondial does it for $995. The DAC is Carver. I do wish, however, that the DTL circuitry were by­ the J grade of the 18-bit Burr-Brown PCM58P, which is passed instead of merely deactivated when the button is in good enough in my book, but yes, there are trim pots, and the out position. On the other hand, I rather like Bob's high­ no, they weren't perfectly adj usted in my sample. I mea­ ly personal ergonomic layout of the front-panel control sured a small amount of low-level nonlinearity that buttons, quite differentfrom standard Akihabara issue. wouldn't have been there in my opinion with just the naked, My severest criticism of the TL-3220 is that it lacks a factory-trimmed J grades (see above). On the other hand, digital output, either coaxial or optical, and that neither the the D2A is loaded with quality features, such as two coax­ front panel nor the remote control has index search buttons. ial and one optical digital inputs (no tweaky "optophobia" A CD like the Richard Strauss Eine Alpensinfonie (Telarc here!), a digital absolute-phase inversion switch (wow !), a CD-80211) has just one track but 22 indexes. The TL-3220 pre-emphasis indicator light (an almost extinct but still rele- is incapable of selecting one of the latter; it can only be fa st­ . vant convenience), and more. forwarded to the approximate location, and even that takes Summa summarum, this could have been the perfect forever. Ye ah, I know, Bob-that old Buddy Holly CD D/A processor for the audiophile who knows value when he transfer you take with you everywhere doesn 't have that sees it, if only a few little engineering decisions had gone problem. the other way. Even so, any owner of the Aragon D2A who is unaffected by those decisions can be justly proud of it. Euphonic Te chnology It's a good machine. Mk II Signature

Carver TL-3220 Euphonic Te chnology, 19 Danbury Road, Ridgefield, CT 06877. Mark II Signature compact disc player with rem ote contro� Carver Corporation, P.O. Box 1237, Lynnwood, WA 98046. TL- $1595.00. Tes ted sample updated from the ET650PX reviewed in 3220 compact disc player with remote control, $529.00. Te sted Issue No. 11. sample on loan frommanufacturer. Michael Goldfield is one of the most highly endorsed Here we are at the bottom of the price range covered Philips modifiers in the delirious little world of high-end in this survey, meaning parts and construction typical of audio. He does a beautiful job and then charges far too midpriced made-in-Japan audio components, a few nice lit­ much, for the simple reason that you can't make a living tle Bob Carver touches, some inevitable engineering trade­ charging a fair price for Philips mods. Therefore he abso­ offs, the absence of certain features-you get the picture. lutely needs a cult following-and he gets it. His mods just Nonetheless, amigos, the sound of the Carver TL-3220 was sound more "musical" than anything else you can buy, the statistically indistinguishable in double-blind, matched-level true believers will tell you. (To paraphrase Samuel Johnson listening from that of the $850 Pioneer Elite PD-71 (to outrageously, musicality is the last refuge of a tweak.) name just one example), which is a thoroughly audiophile­ Actually, the Mk II Signature differs from the 1987- oriented product. Sorry about that. vintage ET650PX reviewed in Issue No. 11 in only five This is also an 18-bit, 8 times oversampling design, ways that I can discern: the DAC is now the TDA1541A Sl but the DAC is the lowest-grade Burr-Brown PCM61P, "Golden Crown" chip, not the plain-vanilla TDA1541; the which works best with a little trimming. The trim pots ap­ digital filter is similarly upgraded to the SAA7220P/B; the peared to be reasonably well aligned, since the low-level DAC power supplies are new; the headphones output is gain linearity and harmonic distortion were quite acceptable "disabled for sonic benefit" (tweak tweak hooray !); and the though not outstanding. For reasons unknown, a sample­ price is up $600. And for your $1595 you still get the ante­ and-hold circuit is included. Other eyebrow raisers are the diluvian and relatively insubstantial Magnavox CDB650 very low ±8 V supply rails in the analog stage, low-voltage chassis (reinforced with the Euphonic Technology DPS-1 op amps of unfamiliar designation, electrolytic capacitors in stabilizer, to be sure) and the Motorola MC34082 analog the output signal path-I could go on, but then I have no output chip, wholesale cost $1.00. (Mike is very secretive solid proof that the avoidance of such audiophile hang-ups about these chips, having obliterated all identification on has any audible effect. them and refusing to reveal their provenance, but circum­ One of the unique Bob Carver touches is the-you stantial evidence and educated guessing pointinsistently to guessed it-Digital Time Lens, the flamboyantly named sig­ the Motorola. "If this be error and upon me prov'd," blame nal processor circuit designed to make early CDs sound it on the secrecy.) more like LPs. It does that by softening the upper midrange Now consider the Philips CD-80 reviewed below. and increasing front-to-back depth with some L - R tweak­ Same DAC. Same digital filter. Just as good, or better, pow­ ing. One thing is certain: when you press the DTL button, er supply regulation. The greatly improved NE5534 decom ­ the Carver TL-3220 no longer sounds like the Pioneer or pensated op amp in the analog stage (rather than the Magna­ any other CD player. I have very little use for it, but it may vox's NE5532 which Mike disliked so much that he had to conceivably be somebody else's main reason for buying the replace it with the Motorola). Plus a heavy-duty cast-alloy

49 chassis with superior moving parts and better ergonomics. very serious bargain if you can live with its austerities. The All that for $799.95, barely half the price of the Mk II Sig­ Mark II versions have been announced at $699 and $529, nature. See what I mean about the Philips mod business? To respectively; at those price points they are somewhat less tip the scales even more cruelly, the gain linearity at -90 dB attractive and begin to have more competition, unless the in one channel of my Signature sample was off by more improvements are greater than I'm currently aware of. than 3/4 LSB, which is out of spec for the Golden Crown DAC, and the noise floor of the Signature in both channels was about 5 dB higher than that of the CD-80. The DAC JVC XL-ZIOIOTN glitch is probably just a QC slipup, but the noise is not; the Motorola chip can be assumed to be the culprit there. lVC Company of America, division of US lVC Corp., 41 Slater Drive, Elmwood Park, Nl 07407. XL-Z1010TN compact disc None of this makes Michael Goldfield's handiwork player with remote control, $700. 00. Tes ted sample on loan from less handsome, or the sound of his CD player less good than manufacturer. that of any other unit reviewed here. It's just that the reali­ ties of the industry are not on his side. JVC is also beginning to make the I-bit DAC scene, so I don't predict a long life for this somewhat aging multi­ HarmanlKardon HD7600 bit unit. What they call its "quadruple fu ll-time linear 18-bit combination D/A converter" is actually the lowest-grade Harman/Kardon Incorporated, a Harman International Company, Burr-Brown 16-bit DAC, the PCM56P, with external com­ 240 Crossways Park Wes t, Wo odbury, Long Island, NY 11797. ponents added to handle the lowest 2 bits, bringing the HD7600 compact disc player with remote control, $599. 00. Tested count to 18. Something of a kluge. The low-level gain lin­ sample on loan from manufacturer. earity I measured was essentially perfect in one channel and

I should have reviewed this excellent CD player in the about 1/2 LSB off in the other channel (at -90 dB), so I last issue-that's how old it is, with an updated Mark II ver­ can't complain too much. Asample-and-hold circuit is need­ sion already in the pipeline-but then I decided to consoli­ ed with this DAC configuration, and there are other little date all CD playback-related material in this issue. ([he not-quite-audiophilic touches throughout, such as an elec­ ' same design also exists in a somewhat leaner $449.00 econ­ trolytic capacitor in series with the output signal and the omy version, called the HD7500, which is identical except plebeian NE5532 op amps in the analog stage. I also saw a that it lacks digital outputs, index search buttons, and a few bit more RF in the output than I liked. On the other hand, other minor control conveniences--almost forgivable at a the so-called K2 Interface, designed to remove jitter by re­ list price $80 lower than that of the Carver above.) clocking the digital data before D/A conversion, is a state­ This "old" HD7600 represented my first encounter of-the art fe ature, made less impressive only by the fact that with any of the new I-bit DAC architectures-in this case Sony does the same thing without making a fuss about it. the Japanese MASH system--and I was duly impressed Don't misconstrue these critical observations as a gen­ with the low-level linearity. Since then I have looked at other eral disrecommendation of the JVc. If no other CD player "bitstream" implementations, and in every one of them the were available to me, I would find it eminently satisfactory. nonlinearity was within the accuracy limits of my test setup. It has everything; it does everything; and it sounds as good as any of the others reviewed here. But in a circuit-for­ Better than trim pots, right? As for the Harman/Kardon ana­ log stages, they are the most elaborate to be seen in any CD circuit, chip-for-chip, fe ature-for-feature shoot-out with the player made in Japan, regardless of price, with fully discrete competition in its price range-not neglecting mechanical circuitry and total observance of the guidelines laid down construction and controls/ergonomics-it is clearly out­ by Matti Otala when he was with the company years ago. gunned. JVC has considerable sophistication in this area of Unfortunately, that approach is inherently more costly than electronics, and I fu lly expect them to come out with some­ the conventional one using inexpensive chips, necessitating thing more exciting in the very near fu ture. They have al­ inevitable compromises in the quality of parts and construc­ ready announced a "pulse edge modulation" I-bit DAC that tion to keep the price of the HD7600 down. One wonders looks promising in their lower-priced line. whether high-performance integrated op amps wouldn't have been a more appropriate choice in this price range. Of Meridian 208 course, some of that Harman/Kardon personality would have been gone. In any case, I measured no anomalies Meridian America Inc., 14120-K Sullyfield Circle, Chantilly, VA worth mentioning in either the digital or the analog circuits. 22021. Meridian 208 compact disc player and preamplifier; with Ergonomically I found the HD7600 to be excellent; 209 remote control, $2950.00. Te sted sample on loan from USA the buttons are well-located and have a nice, positive feel, distributor. as does the disc drawer; the display is first-rate. The sound? Again, no different from that of any other unit in this group, Ye s, I know, the English gave us Shakespeare, New­ as far as any of us could tell. At $599, however, this is an ton, the RAF fighter pilots of 1940, and the Beatles. But outstanding buy, and the $150 cheaperHD7500 version is a they drive on the left side of the road, serve lukewarm beer 50 and hang their plumbing on the outside of the house so the Onkyo Integra DX-7S00 water freezes in the winter. They're too damn impractical, and so is their Meridian 208. I still can't believe it costs al­ Onkyo U.S.A. Corporation, 200 Williams Drive, Ramsey, NJ most $3000, even with the built-in rudimentary line-level 07446. Integra DX- 7500 compact disc player with remote contro� preamp and Boothroyd Stuart pin-striped styling. (For pin­ $700. 00. Te sted sample on loan from manufa cturer. stripe in four figures,I'll take Giorgio Armani.) Seriously, though, I had an off-putting experience as This is a very impressive player for the money, with a soon as I tried to put the battery into the remote control unit. surprising number of quality features, although one of my Everybody else's battery compartment has a sliding or snap­ routine tests tripped it up in a mysterious way. More about on cover, but not Meridian's, chappies. Four PhillipS-head that in a moment. What I like about the DX-7500 is that it is screws keep it in place, torqued down by some Kentish lout very solidly built and generously engineered, with a high­ with large hands, and the screw heads are butter soft so that quality disc transport employing linear tracking (but, inex­ the resistance to any counterclockwise turning will strip plicably, without a velocity-sensing fe edback coil), control them. Unless you enjoy the sight of metal shavings shed­ buttons with a positive fe el and excellent ergonomic layout, ding from your new $2950 toy, your heart will be filled with dual transformers and optocoupling to isolate the digital and hatred by the time you have that silly, im practical cover off analog blocks, coaxial and optical digital outputs, DC­ the battery compartment. A little intelligent planning could coupled analog output with servo, and a choice of several have saved a lot of aggravation here. And that's not all. The line-level output options, one of which (labeled "direct") al­ display of the 208 is of the earliest, minimalist Philips de­ lows the signal to pass through only a single analog signal­ sign, the same that Ta ndberg used to have, and the control processing stage. Quite a serious piece of equipment. buttons are an ergonomic fiasco. For example, moving to an The DAC is the J grade of the 18-bit Burr-Brown index point is a three-step operation. Of course, inconve­ PCM58P, with all fou r of the optional linearity adjustments nience is the proof of high-end performance in some circles. included. The trim pots appeared to be more or less correct­ We ll, what about that performance? Quite impeccable, ly adjusted in my sample, with low-level errors ranging I must admit. The Philips Bitstream I-bit DAC architecture from 1/4 to 1/2 LSB, including my measurement errors. I used results in just about perfect low-level linearity, prob­ have seen slightly better J-grade performance, but only ably more perfect than my measurement setup. Although the without the trim pots! An interesting though probably not Philips SAA7321 chip is a stereo DAC, the 208 has two of very significant measurement showed the usual clipping of them connected in a balanced configuration, a good idea a 0 dB square wave by the digital filter to be symmetrical made somewhat less remarkable by the fa ct that the Har­ through the "direct" output but asymmetrical through the man/Kardon HD7500 economy CD player at less than one­ "fixed" output. Decisions, decisions. sixth the price uses a similar arrangement. What's more, the The one thing I didn't like was the response to the latter gives you discrete analog circuitry whereas the 208 high-frequency two-tone test (19 kHz at --6 dB plus 20 kHz uses the NE5534 integrated op amp, the current Philips at --6 dB, equivalent to a single tone at OdB). All players standard. On the other hand, the output of the 208 doesn't passed this test without any intermodulation products within pass through an electrolytic capacitor like that of the the audio band, but the Onkyo showed intermittent distor­ HD7500 but is directly coupled with a DC servo. (Be thank­ tion blips of the order of -40dB (1 %) at constantly shifting ful for small favors.) Anyway, regardless of the strange mix­ frequencies (3.6 kHz was typical) and of short duration, im­ ture of chintzy and quality touches on the circuit board, I possible to pin down with any degree of accuracy. I have no could measure no imperfections worth mentioning in the idea what this was, certainly not garden-variety steady state electrical performance of the 208. As for mechanical perfor­ 1M. I was ready to make the statement that nobody needs a mance, the disc transport is basically the same as in the top­ better CD player than the Onkyo Integra DX-7500, but this of-the-line Philips LHHlOOO-no complaint there-but the anomaly stopped me. That doesn't mean that any other latter is a far more deluxe and better engineered package. player beat it in the ABX comparisons; it held its own That leaves us with the $2950 question: What does against all comers, but so did all comers against it. the Meridian 208 have going for it at its price? The answer is-Martin CoJloms! In the January 1990 issue of England's Hi-Fi News & Record Review, a few months before the Philips CD-SO 208's American debut, Mr. Colloms canonized it as the Philips Consumer Electronics Company, One Philips Drive, P.O. overall best-sounding CD player known to him. What can I Box 14810, Knoxville, TN 37914-1810. CD -80 compact disc say? It certainly sounds as good as any of the others re­ player with remote control, $799.95. Tested sample on loan from viewed here, but better? Not to my or my associates' ears. manufa cturer. Of course, Martin didn't use the ABX Double-Blind Com­ parator, and I did. So the best I can say for the 208 is that From where I'm sitting, Philips appears to be headed I'll gladly use it in my system as long as (1) somebody else for across-the-board I-bit DAC architecture in their CD pays for it and (2) somebody else operates the buttons. players, even though they originally developed the concept

51 with portable and low-end models in minds. Thus the 16-bit PCM5SP-K, which is the top grade of the top 1S-bit DAC in CD-SO, representing the highest evolution of the line in the line. You can assume that the PD-73 will perform simi­ which the CDSSO used to be the corresponding model, may larly, or possibly even better if the jump from 1S to 20 bits very well be the last of the Mohicans; there seem to be no is indeed meaningful, but only if the MSB trim pot in each newer muItibit models coming out of Hasselt, Belgium. In channel is accurately adj usted for maximum DAC linearity. any case, I find the CD-SO to be an outstanding piece of Of the two samples of the PD-71 I had a chance to look at, equipment, with virtuaIIy no negative attributes. one appeared to be trimmed in perfectly, the other not. Un­ The unusuaIIy heavy chassis is very similar in con­ fortunately, it was the latter that became my long-term lab struction to that of the $4000 top-of-the-line LHHlOOO, and sample; the first I had checked out on the fly while its owner many of the electronic components are of unexpectedly waited. It would be meaningless to specify the errors in the high quality. The disc transport is not quite the same as in poorly adjusted sample, as it would provide no information the LHHlOOO but is still excellent. Special attention has about any other sample, so let me just say this: been paid to routing the digital and analog signals away The Burr-Brown PCM58P-K, with the optional exter­ from each other. FuIIy discrete double-regulated power sup­ nal. bit-linearity trimming correctly fine-tuned, is the most plies are another feature. The front-panel control layout is nearly perfect 18-bit DAC on God's green earth, so the low­ somewhat "creative" and takes a little getting used to, but level performance of your PD-71 (or, by extension, the up­ you soon learn to like it. The remote control is conventional. graded 20-bit PD-73) has nothing to do with engineering The chip complement includes, among others, the and is strictly dependent on the individual history of your TDA1541A S1 "Golden Crown" DAC (which is designed unit, starting at the QC stage. That doesn't really help you, I without external trimming options), the SAA7220P/B digi­ know. It's fa irly easy to make those adj ustments on a prop­ tal filter, and the NE5534 decompensated op amp, a combi­ erly equipped lab bench, but I'd be surprised to finda dealer nation also used in the LHHlOOO. The analog outputs are di­ able and willing to do it. The answer is, of course, to design rect-coupled with a DC servo and protected against fa ilure the equipment without the trim pots. of the servo. I found the worst-case low-level gain-linearity The PD-71 has many other quality fe atures in addition error (in the less good channel at -90 dB) to be of the order to the potentially best multibit DAC there is; in fa ct, it's of 1/2 LSB, which is not quite as good as can be obtained more or less the Japanese opposite number of the Philips with the best Burr-Browns but certainly good enough. The CD-SO, each offering minor advantages over the other. The same can be said of the low-level harmonic distortion spec­ Pioneer is more deluxe in appearance, with one of the most tra; there are some tiny blips peeking out of the noise floor elegant cabinets in the business; the Philips is much heavier, where there are none with the best Burr-Browns, but again more rugged-looking, more technocratic/industrial in style. nothing worth complaining about. Va rious high-frequency Ergonomically the Pioneer is excellent and perhaps easier to test signals produce the usual predictable small-amplitude like; I happen to prefer the Sony CDP-60SESD to either beat tones above the audio range; that's the way Philips one. The Pioneer has an excellent linear-tracking disc trans­ chooses to do the filtering and there's nothing really wrong port made with Sony components; there is some tenuous ev­ with that, either. I would have liked, however, to see less RF idence that this transport tracks certain defective discs that coming out of the back end of my sample, although it didn't the Philips has trouble with. The Pioneer appears to have create any problems in my electronic environment. slightly better RF suppression; it uses discrete push-pull The bottom line is that the CD-SO has made all those power supply regulators (good);it operates all active op amp audiophile Philips mods overpriced and irrelevant. Philips stages in the inverting mode to eliminate common-mode took a good look and included in their package pretty much input signal distortion (good); unlike the Philips, it has no all of the goodies the modifiers had come up with, only with provision for headphones nor a volume-controlled output slicker execution and at a better price. I could live with this (high-end etiquette); it passes the analog signal through player, even if the fu ture doesn't belong to 16-bit DACs. three gain stages and a nonpolar electrolytic output capaci­ tor (not so good-the Philips does it better); it also uses Pioneer Elite PD-71 some op amps that aren't exactly state-of-the-art. Overall, the PD-71 is very conscientiously built and worth its price. Pioneer Electronics (USA) In c., P.O. Box 1720, Long Beach, CA As I indicated in the Carver review above, the PD-71 90801 . Elite PD -71 compact disc player with remote control, didn't sound verifiably better or worse than the much lower­ $850.00. Tested sample on loan from manufa cturer. priced Carver TL-3220, or any other CD player reviewed here, but that's an old refrain of this survey by now. Even This is another CD player I'm catching at the tail end so, I'd rather own the Pioneer as long as I get a chance to of its natural life; by the time you read this it may have been adjust thosetrim pots. phased out and replaced by the almost identical PD-73. The One more thing. The PD-71 is Precision Audio's cur­ only difference is that the latter has the new 20-bit Burr­ rent fa vorite for their excellent D1 Analog mod (see Issue Brown DAC, on which Pioneer has a temporary exclusive No. 12). The debut of the PD-73 presages heavy discounts (at least as fa r as I know). The PD-71 uses the Burr-Brown on leftover PD-71 's if you're interested in taking that route.

52 PS Audio "Digital Link" the presence of a signal it is just about the same, rendering the issue academic. The maximum signal-to-noise ratio PS Audio, 302 South 13th Street, Grover City, CA 93433. "Digital achievable on a CD with 16-bit AID encoding is theoretical­ Link" outboard D/A converter, $799.00. Tested sample on loan ly 9S.1 dB, and the Digital Link accommodates that with a from manufacturer. very healthy margin to spare. I must therefore approve of the audio-quality-oriented noise-Boor trade-off. I was less Paul McGowan and Bob Odell of PS Audio are decid­ happy about the amount of RF I saw pouring out of the back edly in the tweako camp. They believe in a lot of the audio­ end of the unit, but I can 't report any interference problems phile voodoo I ridicule in these pages (the stuffthat I can't in my particular electronic environment (for whatever such hear and nobody can prove is hearable). At the same time, a limited criterion is worth). their products are superbly engineered and utterly sensible Bottom line: I left the Digital Link in my reference setup, driven by whatever transport happened to be handy. in terms of cost-effectiveness.As long as they make equip­ ment like the Digital Link, they can go ahead and believe in No, not because it sounded better than any other unit re­ the tooth fa iry for all I care. viewed here-as you already know, it didn't, and vice ver­ As a matter of fa ct, if an enlightened audio perfection­ sa-but because it ought to have sounded better on the basis ist made a wish list of D/A processor design elements and of its analog engineering and intelligent trade-offs. Let no fe atures requiring no extravagant expenditure, he would be one say I've stoppedbeing a hi-fi nut. describing something reasonably close to the Digital Link. It's basically what the demanding but sane aficionado needs if he decides to take the outboard processor route. Sony CDP-608ESD The DAC around which the processor is built is the Burr-Brown PCM61P, which according to some maverick Sony Corporation of America, Sony Drive, Park Ridge, NJ 07656. practitioners (but not Burr-Brown) is more linear than the CDP-608ESD compact disc player with remote control, $900.00. Tested sample on loan from manufacturer. PCM5SP and "sounds better." The J grade of the PCM61P was used in my sample; later production, according Paul When I received this outstanding unit for review, I McGowan, uses the K grade. Regardless of these distinc­ made the mistake of saving it fo r this survey instead of tions, I measured essentially perfect low-level linearity in squeezing it into Issue No. 14. Between Sony's propensity both channels of the Digital Link. The analog section con­ for constant model Changes and this journal's propensity for sists of a single stage employing a state-of-the-art, high­ delayed publication, that turned out to be a bad decision. speed, complementary bipolar IC (Analog Devices ADS47) The IS-bit CDP-608ESD has now been replaced in the in a rather unusual low-feedback configuration. A high­ Sony line by the I-bit CDP-X55ES, at the same price; in­ current output buffer circuit is used in conjunction with the deed, their entire premium ES line now uses I-bit DACs, of IC, and the output is direct-coupled with only an offset pot a newer generation than either Bitstream or MASH. Even for DC cancellation. (No copycat, this Paul McGowan.) The so, some dealers probably still have a CDP-60SESD or two current-to-voltage conversion circuit is passive, resulting in left in their stockroom, and if the closeout discount is deep the relaxation of settling-time requirements for the op amp enough, grab it-it would be a very good value. (good), greater voltage swing on the DAC output line (dan­ Frankly, I don't see how the new Sony I-bit DAC can gerous living distortionwise), and a significantly decreased improve on the performance of the Burr-Brown PCM5SP-J signal-to-noise ratio (not so good, but more about that in a used in this superseded model-without MSB adjustment, I moment). The power transformer is a bit on the chintzy side might add. I measured almost unbelievably perfect gain lin­ but is separated from the tiny main chassis by an "umbilical earity all the way down to the lowest levels and saw no har­ cord" in order to reduce hum levels; the Digital Link and monic distortion blips peeking out of the noise Boor at any the Aragon D2A were the only units in these tests with that level. The other star performer of the act is the Sony feature. No jitter attenuation circuit is used in the SPDIF de­ CXD1244 digital filter chip; here the design is one up on coder, an economy I deem acceptable considering where the others by combining S times interpolation with noise shap­ money was spent instead. On the other hand, I wish there ing and digital de-emphasis. (The advantage of digital de­ were an optical input in addition to the coaxial one provid­ emphasis is that it removes several passive components and ed; fear of optical data transmission is a high-end neurosis I a FET switch from the op amp summing junction; the disad­ can 't relate to. Physical construction is unimpressive; the vantage in comparison with analog de-emphasis is possibly crowded PC board Boats inside the generic modem chassis increased noise and distortion at the player's output when and is kept from shorting to the chassis on the fo il side by a playing a disc with emphasis.) Another advanced feature is thin piece of plastic. that the digital data is reclocked before entering the DAC. My laboratory tests revealed no objectionable weak­ (JVC built its entire "K2 Interface" promotion around the ness anywhere. Ye s, the noise Boor in the absence of a digi­ same jitter-correcting teChnology.) The disc transport assem­ tized signal is a whole order of magnitude (approximately bly is Sony's high-quality G chassis, which combines a 20 dB) higher than in the best Japanese CD players, but in lightweight electromagnetic linear drive with the KSS151

53 three-beam optical block on a marble-like resin base. The Of course, shoehorning even such rudimentary processing decoder chip is the CXD1165, which was the dernier cri capability into that itty-bitty box is an achievement in itself, when the player was introduced. It's all pretty high-tech for and that neat little LCD readout is a nice touch. a model going out of style and makes you wonder whether On the lab bench a number of little weaknesses be­ the switch to 1-bit DAC architecture is an engineering or a came apparent. The noise floor is only 90 dB below the 0 marketing decision. dB reference level, whether or not a digitized signal is The analog section, on the other hand, is not nearly as present. The de-emphasis is not very accurate; it's off by as perfectionistic. The so-so NE5532 op amps are used as the much as 1 dB at some frequencies and isn 't right on the but­ active elements; the supply rails are ±12 volts instead of the ton at any frequency. The low-level linearity is simply not ±15 volts specified by Signetics; the output signal passes in the same league with the standard-sized players; in the through an electrolytic capacitor (albeit bypassed with a less good channel I measured a full 1 LSB error at -80 dB, small film capacitor)--'--in other words, Japanese upper-mid­ and at -90 dB there was total chaos in both channels. The fi audio circuitry. Not that it made an audible difference in resolution of the Discman is in effect somewhere between double-blind comparisons at matched levels against various 14 and 15 bits-not a ful1 15 bits for sure-although the other players in this group; my objections are therefore DAC is specced at 16 bits and paired with an 8 times inter­ somewhat theoretical. polating digital filter. The digital circuitry even provides an Having said my piece on that, I must commend the optical output, but no coaxial. It should also be noted that CDP-608ESD for two things: the front-panel control layout the D-555 inverts the polarity of the input signal, a rare and display, which were my ergonomic fa vorites among all quirk these days. As for the disc drive and optical block, the the units reviewed here, and the total absence of RF at the quality seems good enough, but I was unable to play track output. Sony is the king of digital audio-if only they lav­ 56 of the 99-track Philips "Audio Signals Disc 1" ished equal care and talent on the analog part! (SBC429); the damn thing kept muting for no discernible reason. Another possible cause for concern is the question Sony D-SSS "Discman" of reliability and servicing; the owner of the D-555 I tested had had a terrible time with its predecessor, the D-TlO; in Sony Corporation of America, Sony Drive, Park Ridge, NJ 07656. fa ct, the D-555 was Sony's please-don't-bother-us-anymore D-555 "Discman " portable compact disc player; $450.00. Tested free replacement for the hard-to-repair D-TlO. Hm . sample on loan from owner. All in all, however, I must come out in fa vor of the Sony Discman. It's small, it's cute, it's not overpriced for Does this yuppie toy, conceived for headphone listen­ what it offers, and it works. Its measurable shortcomings are ing on the way to the brokerage office or the sushi bar, almost certainly below the audible threshold, and its impact belong in the company of the aUdiophile-caliber CD players as audiophile costume jewelry is undeniable. reviewed here? I definitely think so. Even though its mea­ sured performance turned out to be somewhat inferior to Theta DS Pre Basic that of any other unit in this survey, a quick ABX shoot-out between it and the $2950 Meridian 208 failed to reveal any Theta Digital Corporation, 5330 Derry Avenue, Suite R, Agoura reliably identifiable differences in sound. Admittedly, this Hills, CA 91301. DS Pre Basic digital signal-processing pream­ test didn't go to the usual 16 successive double-blind trial plifier; $2400.00. Te sted sample on loan from manufa cturer. runs but stopped about halfway. But even a sloppy nonblind comparison of the Discman with the Onkyo Integra DX- Last in this survey, because of a vagary of the Roman 7500 was quite inconclusive. The Sony sounded maybe a alphabet, is the unit with the most elaborate digital circuitry. little grittier and the Onkyo smoother to the listeners, who The DS Pre Basic isn't actually a full-fledged preamplifier; were quite happy that they didn't have to do any blind iden­ it differs from a monolithic D/A processor only in that it has tifying. The visiting owner had to take the player back home a digital tape loop, inputs for a single line-level analog with him, so I can't swear that a month's worth of compari­ source, fixed analog tape outputs in addition to the variable sons wouldn't have established some clearer differences. At analog line outputs, a volume control, and a balance control. any rate, no other CD player, regardless of price, "blows It could be the control center for a simple CD/tuner/tape away" the D-555 in sound quality. That much is certain. system, but that's it. The volume/balance controls involve The highly touted Digital Signal Processing (DSP) no additional active stage. And yes, there's no optical input; features of the Discman I found less impressive. The bass it would spoilthe high-end image, don't you see? (I don't.) boost below 80 Hz (useful on cheap headphones), dynamic Mike Moffat, the designer of Theta equipment, is yet range compression (for car listening), "surround sound," another of those strange technologists who do everything and graphic equalization at five frequency positions-al1 of knowledgeably and scientifically but talk voodoo to wide­ it digitally implemented-seem quite rudimentary; for ex­ eyed aUdiophiles. (Do they think it's good for business, or ample, the so-called surround-sound processor appears to do they with great sincerity manage to work themselves into equate adjustable phasiness with an actual surround effect. a schizzy Jekyll and Hyde mind-set?) Mike Moffat's soft-

54 ware-driven digital filter for the DS series, in effect a single­ they run the unit through QC again to bring it up to their purpose computer, is possibly the best-designed and most highest standard MSB-wise. When it came back, everything advanced in the business, at least as judged from his conver­ looked textbook-perfect on the lab bench, with no anoma­ sations with David Rich (no schematics and no computer lies or deviations worth mentioning. All right, one peculiari­ code being available). At the same time, Mike claims to ty, not a complaint: square waves at the 0 dB level are hear differences between printed circuit boards made of dif­ clipped more radically by the digital filter than in any other ferent materials and between DAC chips whose pins are D/A processor I've looked at. No big deal. Also-and this made of different metals. Are these claims based on double­ has nothing to do with D/A conversion accuracy or audio blind comparisons at matched levels, Mike? Huh? (On the fidelity-there's a lot of RF in the output, although a little other hand, nobody held a gun to my head to make me go less than before that latest mod. Still, it interfered to a slight into high-end audio and take all this guff.) degree with the performance of an FM tuner about three Other than the digital filter, which is implemented on feet from the Theta. Little gurgles, hash, and birdies. a general-purpose DSP chip, the chip complement of the DS As for the sound of the DS Pre Basic-great! The fact Pre Basic isn't proprietary but a little different nonetheless. that it's indistinguishable in ABX tests from the sound of The I8-bit DAC is the ADI860N-K from Analog Devices, any of the other units reviewed here doesn't make it less the K suffix indicating the highest grade. A trim pot in each great. Indeed, on the basis of general engineering sophisti­ channel adjusts the MSB. The op amp for the analog signal cation and palpable quality on the circuit board, this is argu­ is the PMI OP42, a high-quality but no longer state-of-the­ ably the unit of choice in this survey, but those pesky MSB art IC (a generation behind the AD847 that PS Audio uses, trim pots and the RFI give pause. My impression is that for example). The output is direct-coupled with a DC servo. Mike Moffat has the ability to engineer a flawless piece of Construction and parts quality is very high (except perhaps equipment; however, Theta as a manufacturing/marketing for the use of that op amp), making the price tag palatable to operation needs to get its act together a little more reassur­ the critical purchaser. ingly before I can be entirely satisfied. For openers, they One design parameter that obviously matters a great could include an instruction booklet. deal to Mike is jitter. The SPDIF decoder includes three phase-locked loops to reduce jitter. The PLL circuits used are inexpensive CMOS devices (CD4046) rather than the Recommendations more expensive crystal-controlled PLL circuits used in many Japanese decoders, which theoretically yield even die Wah l ist eine As the Germans say it rhymingly, lower peak jitter. Even so, Theta is the only company other Qual. The choice is agony. Trade-offs, trade-offs, nothing than Mondial/Aragon to publish a peak jitter spec for the but trade-offs. My ideal CD player, or transport/processor, SPDIF decoder, namely 1 ns. Nice. would have the nice fe el, ergonomics, and quality control of My lab-bench and in-use experience with the DS Pre the Sony CDP-608ESD ... the digital filter of the Theta DS Basic had its ups and downs. In the first sample they sent series ... the 20-bit Burr-Brown DAC of the new Pioneers me, the left and right outputs were reversed and the MSB (in the K grade but without MSB adjustments) ... the con­ trim pots misadjusted. I proved to myself, however, that the struction-quality-to-price ratio of the Aragon D2A and may­ latter could be adjusted for virtually perfect low-level lin­ be even its discrete analog circuitry (but with the addition of earity. A second sample, which looked more like a finalized an output buffer)... I could go on but what's the use? It will production unit, had the left/right errorstraightened out, but never happen. Read David Rich's article, read my reviews, the MSB adjustments were still way off, and the low-level and decide what suits you best. When the unequivocally linearity was poor until I again tried and succeeded tickling best choice arrives, I'll be less equivocal. Meanwhile it into perfection. A few weeks later I sent the second unit (chuckle, chuckle) "they all sound the same"-and if you back to Theta for a "basic capacitor upgrade" they had just disagree with that, don't just vituperate but prove the con­ then put into production. At the same time I insisted that trary in a double-blind comparison at matched levels. 0

55 The Wire and Cable Scene: Facts, Fictions, and Frauds Part I

By Peter Aczel Editor and Publisher

This is just a curtain raiser to get you in the mood: an almost two­ year old background story, told here for the first time "like it is." The technical/critical examination of the subject is coming next.

The protagonist of this story is David L. Clark, noted against Dave, and another reason for his frustration. Detroit audio consultant under the name of DLC Design, All I want to do here is to present the givens and the guiding spirit of the Southeastern Michigan Woofer and results of the speaker cable tests fa ctually (this has not been TweeterMarching Society, and notorious designer (with oth­ done anywhere so fa r) and to comment very briefly on the ers) of the diabolical ABX Double-Blind Comparator. He is meaning of the outcome. of course also known to our readers as that wry but rational voice on our Seminar 1989 panel. Dave should really be the Cheap industrial wire vs. Monster Cable's best. one to be reporting here on his own work, but he is letting The high-end speaker cable chosen for the double­ me do the talking because he is, as he says, a total burnout blind listening comparisons was Monster Cable Ml, which on the subject as a result of the apathy, insincerity, misrepre­ at the time was that highly promoted company's top-of-the­ sentations, and downright hostility he had encountered. line model, at nine dollars per foo t. (Since then they have I must point out up front that Dave's story is not only come out with the "Sigma" insanity at almost five times that about wires and cables but also (perhaps even primarily) price.) According to Dave Clark, the Monster Cable people about power amplifiers, but the latter are not our concern were then asked what they considered to be an absolutely here for the moment. What happened was that Dave had unlistenable cheap cable of roughly the same gauge. Their approached the Audio Engineering Society to suggest using answer was lO-gauge THHN industrial wire, so that was their 85th Convention in Los Angeles in November 1988, chosen as the low-priced foil for the Ml. To give each cable with its many knowledgeable and highly motivated partici­ the chance to assert its sonic personality, if any, 30-foot pants, as an opportunity to run some serious double-blind lengths were compared, switched by hand (no ABX box!) listening tests in a workshop to be called "Esoteric Audio­ between a Perreaux amplifier and a pair of Ta nnoy "Dual Can Yo u Hear It?" The way I understand it, the AES con­ Concentric Studio Reference Monitors." sented enthusiastically and promised three listening rooms, Now it should be noted that 434 trials by 86 persons which they delivered, as well as nine assistants (three per (the very few persons who came back to try again are count­ room, to work in shifts), which they did not-not even a ed as two) constitute a respectable statistical base permitting single one. Strike one against Dave; he was left to run the reasonably accurate conclusions. The score: 207 correct show more or less by himself, sink or swim. identifications out of 434 same/different trials; in other LUCkily, he didn't give up, enlisted a number of inter­ words, the participants were "batting" .477 in their attempt ested helpers, and managed to complete 434 speaker-cable to identify the cable they were listening to as being the same comparison trials and 659 power-amp comparison trials as, or different than, the one before. Sheer guessing would over a two-day period. (Each listener opted for anywhere be expected statistically to result in .500, virtually the same between 2 and 7 trials in cables and either 2 or 3 trials in figure. So, to put it bluntly, no difference in sound was heard amplifiers.)On the climactic third day there was a panel dis­ between the audiophile Monster Cable and the lO-gauge cussion of the results, of which a cassette recording is avail­ cheapo wire, not even with 30-foot lengths. able, documenting voices as thoughtful and informed as The reaction to this outcome by the high-end religion­ those of Richard Greiner, Floyd Toole, and John Va nder­ ists was of course predictable. Sure, they said, the unwashed kooy, and as tweaky and undisciplined as that of Michael masses can't hear the difference, but we golden ears can­ Fremer. What emerged quite clearly from the discussion look, here's somebody who got 6 right out of 6, here's 6 out was that the flat-earthers and cargo cultists were not about of 7, here's a 4 out of 4, and so fo rth. All right, class, what's to allow the facts to overturn their belief system. Strike two wrong with that argument? That's right, we're dealing with

56 the beIJ-shaped (Gaussian, or normal) distribution curve. In met with denial by the emotionally vested interests of high­ any large sample of a totalJy random process, a very few in­ end audio. Fortunately there was no "grungy" ABX box on dividual readings will be all the way to the left under the which to blame the obscuration of the alleged sonic differ­ flare of the bell (lowest scores), and a very few will be aIJ ences. That would have been too pat. The denials had to the way to the right under the flare (highest scores). For ex­ take a more convoluted, whiny, philosophically petulant ample, if you toss a hatful of coins 1000 times, you could course-witness John Atkinson's remarks from the floor get all heads once or all tails once, but most of your tosses side of the panel discussion and his report on the workshop will come out half heads and half tails, or not far from that. in the "Industry Update" column of the January 1989 Ste­ There were several 1-right-out-of-7 scores at the AES, in reophile. What I don't understand is why these subjectivist Gaussian balance with the high scorers. In one or two in­ diehards never ask the obvious question of whether there stances, the same person had a high score in the morning exists any kind of mechanism whereby A and B could differ and a low score in the afternoon. Thus 86deaf-mutes guess­ in sound. We re the two cables sufficiently different in resis­ ing wildly about the cables might also have come up with a tance and/or capacitance and/or inductance to interact quite 6-right-out-of-6 score sheet among them. No, amigos, a differently with the source impedance and terminating im­ convincing performance by a golden ear would have been, pedance, and thereby generate significantly different trans­ say, 4/4 in the morning, fo llowed by 6/6 in the afternoon, fer functions? But no-that's not what they ask. They look followed by 7!7 the next morning. Either there is a differ­ at the brand names, they look at the prices, and they just ence in sound, in which case a golden ear will reliably hear know which one sounds better. Part II of this series will ad­ it, or there isn't. Let's face it, there wasn't. dress in depth the electrical network characteristics of I can't blame Dave Clark for fe eling disgusted after speaker cables and the resulting transfer functions at the seeing his quite unexceptionable methods and conclusions amplifier/speaker interface. 0

Music & Video Svsrems For Th e NOI'ice & Connoisseur

�auant J\uoio &. Wioeo

A New Oasis ... Where every client is given the individual attention they desire and deserve ...

Where for every taste and budget you will draw the most pleasure from what music and video have to offer. ..

Where the sight, sound and feel of being there is beautiful and "magically" created for you , ..

Where you are transported , elevated, "liberated " through art and science ...

Realize your most cherished illusions. Come to us .. ,

Consulrancy - Cusrom Sysrems - Acousric Trearmenr - !nslallarion - Relail

Apogee . Arcic'i • Audio Prism . AudioQlIesr • Basis . Bene

Cardas • Chesky • Chicago Sp eaker Sland • Clearaudio

Cogan Hall · Creek . Delos . Disrech • Dorian . Elecrron Kinerics

Eminenr Te chn ology · Firsr Soulld • Carrh • Harmonia Mundi

Klyne • Lasr • Lecrron • Magnan . Merrill . Mod Squad . Mogami

Morch • Nesrorovic • Neurrik • Opus3 • Pro Ac • QED . Rega

Reference Recordings . Rorel • Sequerra • Sheffield Lab . Sims

Slimiko • Superphon • Ta ra Labs . Ta rxer • Tice Audio

Ve nderra Research . Wa dia • Wa rer Lil\' • W B T • and More

800-628-062 7 Princeron JUllcrioll , N.J, 08550 57 Hip Boots ====;;.I___ire of Misinform _ _ lW3(fulgthIO()U�, atiOllil1tlleAlla.=-=== \ � L"""�W;%Wjr-�----�--=-· " . =-��- �,---\" --�.��-----��-;;

Editor's Note: This column was absent from Issue No. 14, and in Issue No. 13 it was still using the editorial "we. " Even though signed articles and the first person singular have now been introduced as the standard convention of this publication, I feel no need to sign my name here, since no one is likely to attribute this sort of thing to another author. Rest assured, the curmudgeonly "I" is I.

Willie Sutton robbed banks "because that's where the mentionable Carver "Amazing Loudspeaker," in the early money is," as he explained, and I keep coming back in this Platinum. Edition reviewed by Dick Olsher in the February column to Stereophile because that's where the audio misin­ 1990 Stereophile (as distinct from later versions), was in ef­ formation is. That's not the only place, I must admit, but fe ct two totally different speakers at sea level and at 7000 then it's not the only publication I take to task here, either. feet-or so I am told by its designer, who was very unhappy about the difference and began to take steps to fixit as soon John Atkinson in Stereophile as he became aware of it. The review glossed over this In the "As We See It" leadoff editorial column of the problem and treated the consequent deficiencies of the May 1990 issue of Stereophile, John Atkinson defends at speaker as engineering ineptitudes, amusing eccentricities, great length the Santa Fe magazine's practice of measuring and poor quality control. According to Bob Carver, he loudspeakers at the altitude of 7000 feet above sea level, in begged DO and JA to audition and measure the speaker at a response to the doubts I expressed about the validity of such less extreme altitude, but they flatly refused. That puts a measurements in Issue No. 14 of The Audio Critic. He very different complexion on the subject than JA's hairsplit­ writes that I feel the need to defend my reputation by attack­ ting little apologia. I'm now firmly convinced that before ing Stereophile's, and after several pages of circuitous argu­ the Carver experience no one at the magazine had the slight­ ments and frequency response curves (the thrust of which is est awareness of the altitude problem and afterwards the that, yes, altitude makes a small difference, but so what) he problem had to be declared insignificant to prevent all sorts concludes that "Mr. Aczel's hypothesis"-which he mis­ of skeletons from tumbling out of the closet. Somehow, "the states so outrageously that I refuse to quote him for fe ar of truth, the whole truth, and nothing but the truth" is consid­ giving the misstatement permanence-"is incorrect." ered bad for business in Larry Archibald territory, as I have Now, in my Madison Avenue days, I used to have a pointed out before. sign in my office that said, "I may have my faults but being To return to the Waveform speaker system, here are wrong isn't one of them ." So you can imagine how much it the plain fa cts: The National Research Council laboratory in goes against my grain to refrain from punching holes in Canada measured one sample of the system to be very flat. I JA's technical arguments (and how disappointed he will measured another sample to be very flat. In Santa Fe, at be), but a larger issue than that needs to be addressed here, 7000 fe et, they measured the system (I don't know which namely intellectual honesty in audio journalism. You see, sample-not mine) to have a big 16 kHz peak. Then JA had JA is hypocritically responding only to the lesser of the two an old, spare sample (not taken from any of these systems!) examples of altitude-skewed speaker testing I cited (viz. of the suspected Philips ribbon-type supertweeter measured Waveform) and carefully avoids any reference to the big, in Los Angeles at sea level and in Santa Fe at 7000 fe et. In embarrassing one (viz. Carver). It's the same as if, for ex­ the tiny, hard-to-read graphs accompanying his May editori­ ample, Nixon were making eloquent excuses for the Agnew al, I discern approximately 2 dB more output at 16 kHz in scandal and pretending that Watergate never happened. Santa Fe than in Los Angeles, exactly as one would expect. I'll come back briefly to the Waveform supertweeter's He calls it 0.4 dB. (Sure, John, if you say so ...) Anyway, alleged peak. First, however, I must remind JA that the un- how does this cockamamy experiment put a Q.E.D. on the 58 general unimportance of the altitude issue, and how does it acoustics, and psychophysics-forget it. It's an embarrass­ prove me "incorrect"? The mind reels. ment to an otherwise outstanding journal. One more thing. In a nasty footnote to the July 1990 Actually, Neil Levenson's bimonthly column in Fan­ "Industry Update" in his magazine, John Atkinson has this fa re, "New for Audiophiles," could just as well be appear­ to say about my participation in the 8th International Con­ ing in The Absolute Sound, of which he is an alumnus. Like ference of the Audio Engineering Society, which took place so many of the equipment reviewers from Harry Pearson's stable, he blithely inserts a new piece of equipment into his in May in Washington, DC: "Characteristically, Mr. Aczel avoided discussion of his role [as audio reviewer], choosing system, plays an old fa miliar recording (say, a CD transfe r instead to attack the other high-end magazines." He is refer­ of some 78's from the 1930's), forms an instant opinion of ring to a special session on the reviewing of audio products, the sound, and attributes the qualities he hears (or claims to where I was alphabetized to be the first panelist to make an hear) to the performance of the new equipment. Just like introductory statement. that, I kid you not. He makes absolutely no attempt to stan­ Again, being "characteristically" ill at ease with the dardize his setup and his methods in order to achieve any whole truth, JA omits that all of us panelists, including me, kind of consistency or repeatability. It never occurs to him discussed our roles as audio reviewers all evening, only my that maybe he is hearing something other than the effect of opening remarks were not aboutmyself but about the gener­ the device under test. Double-blind comparisons at matched al subject of accountability in equipment reviewing-and levels? He probably thinks that's an event at the Special that made JA squirm because it set the theme for the next Olympics. hour or so. He ran into a bit of trouble trying to explain to a What irks me in particular is his smug confidence in large roomful of. the top academics and professionals in the validity of his impressionistic evaluations and solipsistic audio why he and his staff can't prove what they claim to aper<;us. For example, in the May/June 1990 issue, he talks hear. I suppose it did seem like an attack to him, just as my about the Harman/Kardon HD7S00 CD player's "rhythmic comments on speaker measurements at 7000 feet above sea deadness which made the bass line seem remote in time." level seemed like an attack on his magazine'S "reputation." And "perhaps because the HK cannot keep the bass in time Anyone sitting there in that Washington conference room with rest of the music, the mood of the performance was could see, however, that some of the best brains in audio cauterized," he writes. "After about five minutes I was had serious doubts aboutthat reputation to begin with. bored." Joel Flegler, how can you tolerate such untutored trash in the pages of your fine magazine? There exists no Neil Levenson in Fanfare mechanism known to physicists that could make the bass go Fanfare calls itself"The Magazine for Serious Record out of sync with the rest of the music in the HD7S00. Neil Collectors," and in my opinion it lives up to that tag line. Levenson appears to be so in love with the first little conceit Published six times a year in the form of a fat paperback that pops into his mind while he listens that he never gives book, sometimes exceeding 500 pages, it is unquestionably it a second thought before putting it in his column. a quality publication, unapologetically highbrow and strictly I could go on with example after example of Levenso­ of a "scholarly" typographic format (i.e., every page looks nian howlers, but here's one that neatly demonstrates why the same). If a labeled classical recording exists at all, you he shouldn't be writing about electronics. In his review of can be fa irly certain that it will be reviewed in Fanfare, not the Sansui AU-X911DG integrated amplifier (July/August only when first released but also whenever repackaged in 1990), he writes: "I did not test or audition the built-in digi­ some other form or combination. Furthermore, the review is tal-to-analog converter. There are four digital inputs, one more likely to be intelligent and musically enlightened than optical. I noted that the non-optical digital inputs are via the those in the newsstand magazines, since Editor/publisher usual RCA-style jacks. This struck me as possibly not apt, Joel Flegler has a large number of mostly excellent free­ because the impedance of RCA-style jacks is around a lance specialists at his beck and call and lets them write as couple of hundred ohms whereas the owner's manual says they please. They're a pretty sophisticated bunch. The to employ '7S-ohm digital connection cable.' It seems to me more's the pity, then, that Mr. Flegler chose as his Audio that an RCA-style plug does not properly terminate a sup­ Editor a self-indulgent pseudo expert like Neil Levenson. posed '7S-ohm' cable." Not that I expect Mr. Flegler to be an expert on audio I couldn't believe my eyes when I read that. Here's a experts. He is obviously a music man (and a good one), per­ man who accepts money to write a semitechnical column haps without any close acquaintances in the inner circles of for audiophiles, and he doesn't know (1) that those coax audio. But it's possible for a resourceful nonexpert to push digital inputs are terminated with 7S-ohm resistors and (2) some buttons, beat some bushes, make some waves, and that it's utterly meaningless to talk about the characteristic find a genuine expert with genuine credentials. Mr. Flegler im pedance of a half-inch long conduit such as an RCA jack got snookered, I think, by a fellow music devotee--because unless you're well into the gigahertz band. Some expert! Neil Levenson does know quite a bit about music, and that Maybe the solution is to enroll Neil Levenson in one makes him half qualified for the job, even in my jaundiced of those correspondence courses in electronics and ground opinion. But his knowledge of audio electronics, electro- him at Fanfare until he gets his mail-order diploma. 0

59 Seminar 1989: Exploring the Current Best Thinkingon Audio III the (Part of Continuing Transcript) The complete transcript turned out to be much more space-consuming than we had anticipated, so we are reluctantly parceling out the rest of it in short installments to make room for more up-to-date matters .

The first two installments of our seminar have created always the possibility of increased knowledge, regardless of a veritable fa n club among our readers; this group of audio­ the editorial framework, but we concede that after two issues philes would like the transcript to go on forever, and they our mainstream concerns should no longer take a backseat. apparently go to it first, before anything else in our pages. To understand and enjoy to the fullest these shorter Others have told us that they want equipment reviews, not Parts III, IV, etc. of the sem inar, make sure you have read all this loosey-goosey rapping. To us, the decisive fa ctor is Parts I and II, including the capsule bios of the participants.

(Here we continue exactly where we left off, EARGLE: And I don't know of anybody like thal. .. in the middle of a discussion of binaural today who bandlimits an amplifier to 20 CARVER: But they'll go away thinking sensitivity to very small time offsets, on kHz or anything near it. that perhaps you've tricked them. page 51 of Issue No. 14.) LIPSHITZ: No, and that's why digital sys­ LIPSHITZ: ... but they don't do that experi­ LIPSHITZ: So there's no contradiction in tems have come in for so much criticism ment! And when you ask about these rise that statement. And I think the point is, because the transient signals, the square times and so on-put the square wave once the rise time of the system--or, said waves, don't look the way people are used through. Use a very good system for your another way, once the bandwidth of the to seeing them when they go through. transduction. If you could make one with a system--sufficiently exceeds the rise time EARGLE: They look like sine waves. They bandwidth of 100 kHz, you can have a rise or bandwidth of the hearing system, there is are sine waves. time down in the microsecond range. So no point in having it any faster because it LIPSHITZ: But, you know, an experiment that means you 're letting the first umpteen doesn't substantially or significantly alter which we frequently do with students as a harmonics of the square wave into the ear what gets through. It's rather like, once the demonstration-I think that's on that ASA -you still won't hear the difference, un­ signal-to-noise ratio of your recording sys­ disc. Ta ke a square wave of a fundamental less you make the level so gross that you tem is 10 dB better than the signal you want frequency above 7 kHz so that-we believe actually are creating significant intermodu­ to record, it degrades the signal by such a that the true square wave is only odd har­ lation nonlinearities in the ear. smidgen of a decibel that another 30 dB of monics-the third harmonic will be above CARVER: And that's the only way you can signal-to-noise ratio will not get you a de­ 20 kHz; you can 't hear it or any of the hear 20 kHz. Because for somebody my tectable change in the signal-to-noise of the higher ones. So there should be no differ­ age to hear 20 kHz requires a sound pres­ program . It's a comparable sort of situa­ ence between that square wave and a sine sure power about three trillion times greater tion. There's no pointin it. wave of the same fundamental amplitude. than to hear 1 kHz, and at 15.750 kHz, the ClARK: I think these arguments come Now the fundamental of a square wave is television osci llator, it requires about a mil­ from people who are trying to defend a not the same in peak-to-peak amplitude as lion ti mes more sound power to hear than belief that they have an emotional stake in, the square wave is; it's bigger. I think the at 1 kHz. and they search around for anything that ratio is 4/n. That's 4 over 3-point-some­ LIPSHITZ: There is a benefit in not hear­ looks like technical supportand just advance thing, so it's a bit more than l. ing the 15.750, mind you, isn't there? It it. I don't think there's any logic behind it. EDITOR: The amplitude 4 representing the hel ps not to hear that. EARGLE: Dave, if you look at the am­ sine wave and the amplitude 3.14 rep­ CARVER: Wel l, I can still hear that. plifier business for the last 20 or 30 years, resenting the square wave? LIPSHITZ: Oh, you can? I can't. I can't everybody has been advertising that they LIPSHITZ: Yes. So, if the square wave was hear that, either. could pass a square wave with a fundamen­ 1 volt peak to peak, the sine wave would CARVER: Even a perfect ear-it's just tal frequency of 20 kHz, and they were have to be 4/n volts peak to peak. A bit marginal. Yo u're only about 10 dB above damned and determined to show you that more than one, if I remember the number the threshold. the thing retained a square wave shape all correctly. I may be wrong; I think it's that. EARGLE: Yo u remember the London con­ the way up, without any overshoot or with And you've got to watch most fu nction vention--oh, about seven or eight years minimal rounding or something like that. generators--when you switch square to ago-when KEF put on that demonstration And they were the first ones who said that sine, it keeps the peak amplitude the same, of the audi bili ty of various things? it's got to be this good in order to work in not the fundmental. So you will hear a dif­ LIPSHITZ: Clipping and that sort of thing? the passband of the ear. ference solely on the basis of the level EARGLE: Ye s. And one test was the audi­ LIPSHITZ: Which was fairly easy to do change. And if you do this experiment prop­ bility of a noise signal extending beyond 20 with amplifiers; you just make their band­ erly, people don't believe it. They look at it kHz. It was a double-blind test. widths up in the hundreds of kHz range. on the oscilloscope: "I can't hear the bloody LIPSHITZ: Oh yes. Ye s. However, with digital systems to get that difference!" There's nothing that will con­ EA RGLE: And Laurie Fincham had come bandwidth is a little bit excessive! vince people more rapidly than something up with some sort of a ribbon tweeter, a 60 Japanese ribbon tweeter, that he mounted synthesizers ... spendquite a few minutes trying to identify on these KEF speakers, so that the speakers LIPSHITZ: Wel l, if you know a composer it. Because I wanted to write to CBS say­ actually went up that high-because his who is composing for that one person ... ing, "Why is Tc haikovsky's 5th playing in own speakers wouldn't do it. Anyhow, only EARGLE: Look, We ndy Carlos can make the background?" But we could not identify one person out of the entire population sta­ an arrangement of "Switched-On Bach" what it was. tistically beat the odds, and it was an 18- with an osci llator going all the way up to CLARK: "And did you pay the royalties?" year old guy from Denmark. He was the 25 kHz. Should it be heard or should it not EARGLE: Is that the early recording or the only one. be heard? new one? LIPSHITZ: Where was the cutoff of the LIPSHITZ: Well, is she doing that because LIPSHITZ: No, no, no. The new one. The filter for comparing? there is somebody who can hear up there one recorded at CBS studios in New York EA RGLE: Something like 23 or 24 kHz. and she wants him or her to hear it? City. LIPSHITZ: Ve rsus? EARGLE: Wel l, let's assume so. McGRATH: The last Goldberg? EA RGLE: Twenty. LIPSHITZ: Or just because she didn't have LIPSHITZ: If you want to try it, every­ LIPSHITZ: Ve rsus 20 kHz? a suitable filter to put at the output of the body's got it. Yes, that one. Go to around EARGLE: Ye s. And he is the only one who oscillator? 24 minutes. could really hear it. Now, when you go EARGLE: Let's assume that she is saying, EDITOR: I own that CD. from 20 down to 15, lots of people heard it. if it can be heard, I'd like it to be heard. It's LIPSHITZ: Everybody owns it, man. CBS LIPSHITZ: Oh yes, absolutely. there to be heard. must be making a very good return on that. EARGLE: The idea being-we all have to LIPSHITZ: All right, I won't answer that. EA RGLE: Incredible. ask ourselves this question-if one person I'll answer it with a question. EDITOR: I never noticed what you did. out of a population of maybe, oh, let's say EARGLE: Okay. That's fair. LIPSHITZ: Listen around 24 minutes in the 250 members of the area who might have LIPSHITZ: Suppose We ndy Carlos was not thi ng-that happens to be a good place to volunteered for these tests, if one person being recorded and released on records. listen because the music's very quiet there heard bandwidth in excess of 20 kHz, does Would you feel there's an obligation to -and you will hear this orchestral thing that make it worth pursuing, intellectually, record and release her work so that the one going on in the background. Now it's as engineers, or commercially? Probably no. person who might be interested in hearing throughout the disc; it comes and goes. It's EDITOR: Don't say that too quickly. it can hear it? almost certainly RFI; you will hear that be­ EARGLE: Okay, I won't. But I think it's a EARGLE: Well, on the other hand, let me tween variations it fades down and comes point worth elaborating for the sake of our ask you a question. Let 's say this is only on back up, so you know they digitally edited discussion here. What are our obligations? one piece and it's recorded as Band 1 on the thing. Fade it down, bring it back up ... How do you feel about it? Side 1 of an LP ... McGRATH: If it had been done in analog, I CARVER: You're right. If it were possible LIPSHITZ: And that's not intended in any would have thought maybe they didn't to hear 22 or 23 kHz ... quite degauss the tape; I've heard that on EARGLE: Ye s, if those people could hear other things. it, should we make records for them? " ...has it been an arbitrary LIPSHITZ: It's on the master tape; almost CARVER: ... if it were possi ble, and if they decision ...that 20 or 21 kHz certainly it must be coming in on their mike did hear it, I mean the SPL's must be just lines; it's RFI... incredible at those freq uencies, and it can't ought to be the upper EDITOR: The next thing would have been do anything but be hurting something, I limit-what we're going to "Breaker two! Breaker two!" would think. be recording from now until McGRATH: "Good buddy Glenn, how ya EARGLE: No, it isn't that... doin'? Come back, good buddy Glenn." CARVER: I mean, you're going to be the end of time?" (Laughter.) around at least 100 dB if you're going to EDITOR: That has happened to me, on my hear 22 kHz ... way as a reflection on We ndy Carlos. This system, in the middle of the music. LIPSHITZ: No, not necessarily. is just composer X. LIPSHITZ: ...but if those mike preamps EARGLE: No, at that age not at all. EARGLE: Okay. Let's say this tone is re­ and the digital system had been processed LIPSHITZ: It depends. Some people do go corded on Band 1 of one side of this LP, Iinearly ... that I can hear it. ..demodulate beyond 20 kHz without enormous troubles. and you know it can be handled by that di­ the radio station WQXR or whatever it EDITOR: Wouldn't you say that a perfec­ ameter. You can record to 35 kHz on the was, then... [everybody talking at once, tionist sound system should be designed for outer bands of an LP. making this undecipherable-Ed.J the most perfect ear among us? LIPSHITZ: Should you do so? EARGLE: Okay. I see the point you're LIPSHITZ: How do we find that person? EARGLE: Well, you can play it back. It's aiming at here. And it isn't quite what I'm We first have to find that person and mea­ there. talking about. sure him or her before we can design the LIPSHITZ: Suppose you could record the LIPSHITZ: I know, I know. (L aughs.) sound system then, Peter. So we have a bit RFI that your microphone line is picking EARGLE: The thing is that we have car­ of a problem. How many-3 billion people up, without demodulating it, so that when tridges and LP's and tape machines, at 30 in the world now? you played your record back you could also ips, that will handle well beyond 20 kHz. EDITOR: I could live with the most perfect get the PM broadcast that was going on in It's been shown that some people can hear one found so far. the background ... some distance beyond 20 kHz. The ques­ LIPSHITZ: We better check them all quick­ EARGLE: That would be a very wide- tion then becomes, basically, has it been an ly before we go any further. range system .. . arbitrary decision, or why have we decided, EDITOR: "Found so far" I think is good LIPSHITZ: ...as a choice, instead of being that 20 or 21 kHz ought to be the upper enough. forced to listen to the demodulated version. limit-what we're going to be recording CARVER: I've checked a lot of hearing (Everybody finds this hilarious.) For exam­ from now until the end of time? and I've only found one person who can ple, take Glenn Gould's Goldberg Va ria­ CARVER: Wait a minute. Wait, wait. It's hear even 21 kHz, and again it was a very tions, on CBS ... been shown that some people can hear be­ young person and everything was perfect. EARGLE: You hear a lot of musical exter­ yond 20 kHz but at such horrendous SPL's And it required a tremendous sound pres­ nals going on. that it will never show up at that level. And sure before he would even hear it. I think I LIPSHITZ: Ye s, you hear plenty of Glenn if there's a 21 or 22 kHz musical bit on our was overloading something. Gould. But please tell me what the orches­ recording, even that person is not goi ng to LIPSHITZ: And in music, there is no musi­ tra is that's playing in the background. hear it; it will be below his threshold at 22 cal content up there. Even if at a point you CLARK: Is there one? kHz; he just won't hear it. detected it. .. LIPSHITZ: Oh yes. LIPSHITZ: But I appreciate John's point. I EARGLE: Well, in normal music ... I don't EARGLE: Really? think my answer to you would be this. If know. I don't know. When you talk about LIPSHITZ: Ye s. And I've had friends there were no good engineering reasons for

61 wanting to restrict the bandwidth to some­ EARGLE: That's right. imate dynamic range of the ear. And that thing round 20 kHz, I'm sure we would EDITOR: It's like the fear of rising taxes--­ means that I need more than 16 bits. Where have it wider. you have a fear of lowered sampling rates. does that take us up to ... ? 18, 20 ... We CLARK: Just like amplifiers. (L aughter.) need about 20 bits to do that. And then you LIPSHITZ: Not because there are some LIPSHITZ: But it's interesting ... The fun­ get very dicey, if you need more than that. people but just to give us a bit of extra lee­ ny thing is, John, you're happy to accept 16 Because then you're designing for crea­ way. Not because I happen to know one in­ bits, but you're reluctant to have too Iow a tures we haven't fo und yet, who maybe dividual who could hear it. But the penalty, sampling rate, too close to the Nyquist lim­ will visit us from outer space one day-and the engineering penalties ... it ... you wouldn't want to limit their dynamic EARGLE: It's an economic penalty. EARGLE: Ye ah. Yeah. range ... (Laughter.) LIPSHITZ: ...for increasing the bandwidth LIPSHITZ: ... yet, of those two trade-offs CLARK: Wel l, we should have some music from 20 to 25 kHz in digital are extremely or engineering decisions, the former is the for them ... high. And given the choice that that might questionable one, and the latter is the iron­ EARGLE: I suggest then that we further mean giving you 25% less playing time on clad one, in principle. state that the decision to use 16 bits is in it­ your CD, for that one individual, I doubt CLARK: Yes, that's right. That's right. self an unfortunate compromise and say that many people would make that decision. LIPSHITZ: If you can't hear above 20 kHz, that we around this table all wish that these EARGLE: Okay. I think you 're probably you don't need a sampling rate more than limits could beextended. absolutely right and I would say that it be­ two times that, plus whatever guard band LIPSHITZ: No, I'm not sure I wish that comes a matter of the cost of real estate in you need for reasonable filterdesi gn. one. the medium. And I would expect to see this EARGLE: Of course. CLARK: I don't wish to pay for it. particular point made in the transcript of LIPSHITZ: Whereas the decision to use EARGLE: Well, okay. It sound to me as this meeting we're having today. [Come on, any finite number of bits, in principle, is a though everybody here is in agreement that John, every word is being transcribed; degradation, a loss. We can change what Philips or Sony-and I don't trust those that's why the damn thing is so long.-Ed.) would be a distortion into a noise, by dith­ bastards any fu rther than I could throw LIPSHITZ: It's an engineering trade-off. ering or doing other things, but that's what them (spoken with a fa cetious inflection)­ EARGLE: It is a trade-off. We would not fixesyour signal-to-noise ratio; it is a trade­ made the right compromise for mankind choose to do it for any other reason. off. The noise can be heard. So you've in­ forevermore. Now, I'm sure you don't McGRJXfH: But is that the reason it was troduced a noise, which in an infinite-bit think that. chosen in the first place? I mean, why did system would not have been there. LIPSHITZ: Now wait. We 've already dis­ Philips settle on the bandwidth that they EARGLE: Well, it's a noise which you can cussed the sampling rate, and I think we did? argue about, in terms of the dynamic range said, all else being equal the engineers LIPSHITZ: Oh, there's a very good reason of music. The thing is that if you measure would probably have chosen a higher one if for that. they'd been able to (inaudible word) for it. CLARK: That's an even number, and twice CARVER: For comfort reasons ... that is 44.1, which they already have ... "My point is that, EARGLE: And they probably would have LIPSHITZ: No. No. The reason is the fol­ had consciences not prevailed chosen more than 16 bits if they had the lowing. The only feasible way of recording leave to ... digital was on video recorders. If they here, we could have ended CARVER: ...for just sort of comfort rea­ hadn't established a format for recording up with a medium sons, but not for any real demonstrable rea­ on video recorders, there would have been that was IS-kHz limited sons. no digital material available at the time of LIPSHITZ: We ll, look. The original Philips the CD release. and I4-bit limited." conception was 14 bits. It was as a result of EARGLE: Soundstream was generating all Sony's prodding that it was raised to 16- the 50 kHz material, and there are a lot in any concert hall the maximum level of because it didn't require unacceptable loss tapes around which have been transcoded an orchestra and then look at the noise level of playback time on the medi um, thanks to over to 44.1. into which it sinks when the music stops, clever channel encoding and so on. LIPSHITZ: Right. But the question was, you're nowhere near the inherent dynamic EARGLE: I must say something here. I why was the CD 44.1. My answer is that range limitation of a 16-bit system. wanted to talk about unfortunate compro­ the only practical standard that they could LIPSHITZ: True, but you're nowhere close mises. come up with-we're talking now ...going to what the ear is capable of. LIPSHITZ: Yes, let him modify his state­ on to a decade ago-was based on the vid­ EARGLE: That's right. ment, then I'll comment on it. eo format, and that means you had to have CARVER: So when we combine what you EARGLE: I won't modify it; I'll elaborate. an integral number of samples per horizon­ said with what you said, we find that the ex­ I'm going to say that if we had had Philips tal line, and that led to 44.056 kHz in isting digital system very comfortably fits succeeding with their 14-bit suggestion NTSC 59.94 fields-per-second countries-it totally inside the envelope of the human here, and somebody else were saying that, would be 44.1 for 60--but anyhow, that's hearing mechanism, without being troubled well, FM, which our best medium in many what led to that number. So it's just de facto; at all. ways, is limited to 15 kHz, so why not go if there had been other ways of storing it, I'm EARGLE: Yes, it does. to 32 kHz sampling-and where do you sure a much more attractive-looking num­ McGRATH: No, no, no-the music replica­ think that unfortunate number came from? ber than 44,100 would have been chosen. tion mechanism, not the human ear. LIPSHITZ: No, that I call unfortunate. EARGLE: Yes, and you know, when you LIPSHITZ: The music replication, yes ... EARGLE: Okay. look at the entire standard, I've never been McGRATH: Theear hears much lower than LIPSHITZ: 14-bit I would call unfortunate put offby the 16-bit limitation. That's nev­ the 16 bit ... because it's not significantly better than an­ er bothered me at all. But I have sort of CARVER: Oh, yes, yes. alog tape. 16-bit I don't call unfortunate. wondered from time to time about the sam­ McGRATH: ...but the real world does not EARGLE: Okay. My point is that, had con­ pling rate because in our time, in this busi­ demand more than 16 bits. sciences not prevailed here, we could have ness, we've seen it: begin at 50, with LIPSHITZ: ...what I'm saying is, the 16- ended up with a medium that was 15-kHz Soundstream; we've seen it go to 48; and bit decision has introduced a noise floor limited and 14-bit limited. then we've seen it drop preci pitously down which in realistic music, as John is saying, LIPSHITZ: Ye s. to 44.1. One can only look at that and say, is not a limitation. He can't find a hall CARVER: That's true. We came close to well, you better watch those bastards or that's quieter to record in. However, if I that. they're liable to lower it again. It isn't go­ wanted to have a recording system that LIPSHITZ: We came close. ing to happen, but for a while ... would never add its own noise to any re­ CARVER: We came close. LIPSHITZ: There's a theorem that will pre­ cording I made, I need a dynamic range of LIPHITZ: But it's like the slew-rate ques­ (continued on page vent them from lowering it anymore. at least 120 dB-because that's the approx- 68) 62 Records&Recording

In the immortal words of Julius, jacta alea es4 the die is cast. (No, not Julius Futterman, audio freaks. Julius Caesar.) This will be strictly a CD and DAT column. There will be no LP 's reviewed here, even though that possibility was brought up three issues ago. LP 's are no longer a factor in purchasing decisions involving new releases, and our precious vinyl heritage is currently being archived on CD with considerably greater skill than was the case initially. A widely traveled industry figurerecently observed that various audio journalists known for their antidigital protestations seem to be playing nothing but CD 's these days; their turntables are always being serviced, updated, exchanged, borrowed by a friend, etc., and therefore not available at the moment. Ve ry interesting�nd not surprising.

CD 's from -the Golden Age of Audio (Meaning Right Now)

By Peter Aczel Editor and Publisher

We have reached the point in the evolution of the CD Op. 57 ("Appassionata"). Vo lume II: Sonata No. 21 in C Major, where the least good new release sounds pretty nice and the Op . 53 ("Waldstein "); No. 17 in D Minor Op. 31, No. 2 best sounds awesome--and the same is true of the new CD ("Tempest"); No. 26 in E-flat Major, Op. 81a ("Les Adieux"). Vo l­ players. I call that the golden age of audio because LP's and ume III: Sonata No. 15 in D Major, Op. 28 ("Pastoral"); No. 16 in phono cartridges never came close to giving that kind of G Major, Op. 31, No. 1; No. 18 in E-flat Major, Op. 31, No. 3 ("Hunt"). Vo lume IV: Sonata No. 1 in F Minor, Op . 2; No. 2 in A consistent satisfaction, even if the absolute best of them Major, Op. 2; No. 3 in C Major, Op. 2. Te larc CD-80ll8, CD- were quite wonderful. Tod ay the medium is on an entirely 80160, CD -80185, CD -80214, respectively (all DDD, recorded new level of sonic reliability. 1985/1989 by Jack Renner, released 1986/1990). For that reason I want to get away from grouping my reviews label by label, as if that were the major determinant Prof. Charles Rosen of the University of Chicago, one of audiophile interest. Although a few labels still offe r more of the outstanding intellects of the music world and some­ predictably excellent sound than others (especially the three thing of a cult figure as a pianist, points out in a recent arti­ D's and the double R-DeJos, Denon, Dorian, and Refer­ cle in The New Yo rk Review of Books that the piano sonatas ence Recordings), surprisingly good engineering is current­ of Beethoven and Schubert were not composed for concert­ ly encountered with some regularity on many labels, major hall performance but, in most cases, for a semiprivate soiree and minor. The reasonable thing to do now is to alphabetize audience of 20 or 30 Viennese music lovers. John O'Conor by composer, like the catalogs, and treat the sound as just plays Beethoven as if that were his principal guideline, and one more attribute of the recording. I find his intimate style, with its total eschewal of declama­ tory distensions, quite refreshing. My feeling as I listen to Beethoven him is that he is playing for me, not for an adoring gallery, Ludwig van Beethoven: Piano Sonatas. John O'Conor, piano. Vo l­ and that he wants me to understand the structure and flow of ume I: Sonata No. 8 in C Minor, Op. 13 ("PatMtique"); No. 14 in the music. Not that he is lacking in virtuosity; he is fleet­ C-sharp Minor, Op . 27, No. 2 ("Moonlight "); No. 23 in F Minor fingeredand accurate enough to satisfy any nitpicker of key- 63 board technique; but his aim is to make his audience say the Alte Oper in Frankfurt is very successful; the violins are "aha" or "hm" rather than "wow." One could, of course, smooth as silk; the timpani, bass drum, and lower strings cynically speculate that a pianist who possesses neither the are audible in almost frightening detail; but what are those emotional profundity of a Schnabel nor the brilliance of a RF birdies doing on Track 3, beginning at 4:41? This is too Horowitz is probably reduced, willy-nilly, to such a "good­ good a production for something like that to be allowed to taste," musicianly approach. I feel, however, that a complete slip through. Even so, if you believe in owning more than set of the Beethoven sonatas, which this Telarc series will just one or two Fantastiques, get this disc. eventually become, actually gains a certain reference value by avoiding extremes and normalizing its performance rhet­ Bernstein/Barber/Gershwin oric to the middle of the spectrum. That helps to bring hotter and colder performances, when they occur, into proper per­ Samuel Barber: Overture to "The School fo r Scandal. " George spective. Gershwin: An American in Paris. Leonard Bernstein: Arias and Since the highest opus number in these four volumes Barcarolles. Seattle Symphony, Gerard Schwarz, conductor; Jane Bunnell, mezzo-soprano; Dale Duesing, baritone. Delos DE 3078 is 81a, it remains to be seen, or rather heard, how the (DDD, recorded 1989/1990 byJohn Eargle). O'Conor treatment suits the biggies like Op. 106 ("Ham­ merklavier") or Op. 111. His "Appassionata" is already a bit Along with the Wa lter Piston album reviewed below, tame (Le., unimpassioned) for my taste, although quite love­ this CD represents a new standard of sonic excellence in the ly in many respects. Hey, maybe his taste is more refined Schwarz/Seattle series engineered by John Eargle, which I than mine. have already glorified more than sufficiently. The sound is As fo r the audio quality of these four CD 's, Vo lumes cleaner, crisper, more transparent than ever, without even a III and IV were recorded in the same hall in Worcester, suggestion of strain or hardness. Definitely 1990 demo qual­ Massachusetts, with identical equipment, whereas Vo lumes ity. Could it be an improved method of digital encoding? I and II represent slightly earlier implementations of the The credits cryptically list "Digital Recording: Sony" where same technique in two different halls in England, so that the there used to be model numbers of the digital processing sound is similar throughout but perhaps slightly superior in equipment. Ve ry interesting ... the later recordings. A spaced pair of Bruel & Kjrer 4006 A little over 50% of the music is a Bernstein joke, a omnidirectional condenser mikes is the common denomina­ cutesy fa rrago of fluff and nonsense composed (concocted?) tor of them all; like the playing of John O'Conor, that con­ in 1988, which Lenny must have considered very important stitutes another "normative" fa ctor here--even a recording because he let one Bright Sheng do the orchestration. I am engineer as pragmatic and as different from Jack Renner as decidedly underwhelmed. The best reason to buy this disc is

John Eargle uses the very similar 4004's fo r solo piano­ a highly idiomatic performance of Gershwin's "An Ameri­ resulting in a solid, well-delineated, dynamic, completely can in Paris" in-get this!-the uncut, original version that unproblematic piano sound with a fa irly close-up perspec­ apparently has never been performed until now. About three tive but without any blow-your-socks-off ambitions. I can minutes of music already fully orchestrated by the compos­ recommend this continuing series by the rising young Irish er were excised from the premiere performance, for reasons artist to all those who like their Beethoven straight up. no longer known, and can be heard here for the first time. All that and a great bass drum, too, in the fa miliar passages. Berlioz Samuel Barber's eight-minute graduation thesis from the Curtis Institute (he was 22) is also a class act, a very nicely Hector Berlioz: Symphonie Fantastique, Op. 14. Frankfurt Radio crafted piece of 1930's eclecticism and very nicely played. Symphony Orchestra, Eliahu Inbal, conductor. Denon CO-73208 Although I wouldn't include this album in "All the (DDD, recorded 1987 by Detlev Kittler, released 1989). Classical Music Your Family Will Ever Need" (remember The most fa ntastic thing about Berlioz's fantastic "ep­ that incredible TV commercial?), I can recommend it for isode from the life of an artist" is that it was composed in half of the music and all of the sound. 1830, only six years after Beethoven's Ninth. The stylistic light-year Berlioz was able to put between himself and all Bruckner who had preceded him still doesn't fa il to astonish. This a Anton Bruckner: Symphony No. 4 in E-flat Major ("Romantic "), very fine performance of this amazing work, in the deliber­ ed. Nowak. Royal Concertgebouw Orchestra, Riccardo Chailly, ate, meticulously constructive manner of Inbal, who obvi­ conductor. London 425 613-2 (DDD, re corded 1988 by Colin ously believes that too many climaxes would be equivalent Moorfoot, released 1990). to no climax at all. He graduates the tension from move­ Anton Bruckner: Symphony No. 4 in E-flat Major ("Romantic "), ment to movement, phrasing each measure with exactly the original 1874 version. Cincinnati Symphony Orchestra, JesUs Lopez-Cobos, conductor. Te larc CD-80244 (DDD, recorded 1990 emphasis it organically needs, and at the end the work esca­ by Jack Renner). lates to an overwhelming blaze of glorious sound. There is something to be said for taking your time. I happen to be one of those who find much of Bruck­ The recording with Denon's usual B&K technique in ner's music disjointed, naively grandiose, and lacking the

64 natural fluency of other grandiose 19th century composers on a Dorian sampler, remains the high point of the album. like Wagner, Liszt, or Tc haikovsky. Nevertheless, I concur Overall, an honorable near miss. with the musicologist Alfred Einstein that Bruckner "pro­ duced his most harmonious work in his Fourth Symphony, Elgar which depends almost entirely on beauty of sound." For beauty of sound you would expect the Royal Con­ Sir Edward Elgar: Cockaigne (In London To wn), Concert Over­ certgebouw to be an easy choice over the Cincinnati Sym­ ture, Op. 40; Va riations on an Original Th eme (Enigma Va ria­ phony, but such is not the case. LOpez-Cobos appears to tions), Op. 36; Serenade in E Minor fo r Strings, Op. 20; Salut d'amour (Liebesgrup), Op. 12. Baltimore Symphony Orchestra, have much more of a can amare approach to the symphony David Zinman, conductor. Te larc CD-80192 (DDD, recorded 1989 than Chailly, or perhaps more of a sense of occasion, since by Jack Renner, released1989/1990 ). he is conducting the very rarely performed 1874 Urtext of the Fourth, published only in 1975, the first of five different The firstthing that struck me when I beganto listen to versions by the perpetually self-doubting composer. The this was how well the Baltimore orchestra was playing. No Cincinnatians respond to their conductor with beautiful so­ provincial symphony, this one. David Zinman is obviously norities and great discipline, and Jack Renner's recorded doing a good job. The recording, too, shows Jack Renner in sound is more beautiful than Colin Moorfoot's,al though the very good form ; he seems to have figured out the Joseph latter is very respectable in the Decca/London multimiked Meyerhoff Symphony Hall in Baltimore to a T, and here idiom. Jack Renner is back to his trusty old Schoeps omnis again he is back to his nice Schoeps omnis after his flirta­ and getting much sweeter upper midrange and lower treble tion with those overly aggressive Sennheisers. This is the than in his recent Sennheiser recordings. kind of sound I expect from Telarc (see Issue No. 14, p. 39, I unhesitatingly choose the Chailly, however, as the third column). musically preferable disc of the two. The original 1874 ver­ The music is of course fa miliar; I don't claim to be an sion has a Scherzo totally different from and not nearly as Elgar maven but I find his anglicized late-19th-century idiom brilliant as the "Hunting Scherzo" of the later versions, without sentimentality quite exhilarating. The "Enigma Va ri­ ations" are definitely a minor masterpiece-maybe not even which also have a considerably altered Finale. As a non­ Brucknerian, I always thought it was those amazing brass so minor-and the "Cockaigne" overture is rousing good passages in the later Scherzo that made the Fourth worth the fun. Zinman conducts all of it with authority and sensitivity; price of admission. In this case Bruckner wasn't just being short of a DDD recording of a resurrected Toscanini or Sir insecure to have listened to criticism; he did improve on the Adrian Boult, this will do. symphony, at least in my not very authoritative opinion. Therefore, historical/puristic considerations aside, I opt for Franck the standard Nowak edition-but then Chailly has lots of competition from greater conductors. Cesar Franck: Symphony in D Minor; Va riations symphoniques. Royal Philharmonic Orchestra, Claus Peter Flor, conductor; Dvorak Rudolf Firkusny, piano. RCA Vi ctor Red Seal 60146-2-RC (DDD, recorded 1988/1989 by Mike Hatch & Mark Vigars, released 1990). Antonin Dvorak: Piano Quartet in D Major, Op. 23; Piano Quartet in E-flat Major, Op. 87. The Ames Piano Quartet: Mahlon Dar­ To me, the important part of this release is the Va ria­ lington, violin; Laurence Burkhalter, viola; George Wo r!.; cello; tions for piano and orchestra, not the Symphony. I like only Wi lliam David, piano. Dorian DOR-90125 (DDD, recorded 1989 the second movement of the latter; the rest strikes me as a by Craig Dory, released 1990). bunch of overstated platitudes. Thus I am disinclined to For once I have to criticize, albeit mildly, Craig Dory's judge the finer points of what is an obviously well-played, work at his fa vorite recording site, the Troy Savings Bank musicianly performance. The Va riations, on the other hand, Music Hall. This recording of three stringed instruments are a lovely, beautifully crafted, classically restrained work

and a piano is too reverberant; it should have been miked a of lasting appeal-and an excellent vehicle for one of my little more dryly and intimately. I think Craig is so much in longtime heroes, Rudolf Firkusny. He is now 78 years old, love with the unique acoustics of the TSBMH that he wants but when he was 50 he absolutely owned the Beethoven to make sure you don't miss the slightest nuance of it, but in Third Piano Concerto, for example (in my opinion, at least), this case the frame, so to speak, threatens to overwhelm the and his Mozart was equally marvelous, not to mention the picture. Don't misunderstand me; it's still a very nice, musi­ Czech masters, which of course he played in the best native cal sound, but from Craig I expect a "ten" every time. idiom. Here he is still the aristocratic, singing-toned, lyrical The music here is one early work of Dvorak and one player of old, and the 15-minute piece becomes an intense mature spellbinder; the players are excellent but not world­ musical experience under his hands. The 37-year old Ger­ class; even so they play the exceptionally lovely Lento of man conductor, Flor, is new to me, but he is as good as is Op. 87 with sufficient flair to make me almost like the not­ needed for this effort and so is the Royal Philharmonic. quite-right recording. That movement, excerpted in advance The RCA recording is interesting because it is so

65 Coenraad Bloemendal, cello. Dorian DOR-90130 (DDD, recorded smooth, sweet, rounded, and trouble-free. 1 suspect a com­ mercially calculated restriction of dynamic range, making 1989 by Craig Dory, released 1990). the album sound good on just about any system but not This, too, is a Troy Savings Bank Music Hall chamber extraordinarily good on systems such as mine. Anyway, it's music recording, just like the Dvorak album above, pro­ a better philosophy than boosting everything to the point of duced by the same team only two months later. The results, unpleasantness. however, are distinctly superior in my opinion; the sound is a little less reverberant, with still plenty of live ness and Handel sufficient TSBMH signature, and the three instrumentalists are superb. George Frideric Ha ndel: Arias fo r Montagnana. David Thomas, bass; Phi/harmonia Baroque Orchestra, Ni cholas McGegan, con­ These are wonderful works, needless to say, by two of ductor. HarmoniaMundi HM U 907016 (no SPARS code, recorded the greatest masters of the piano trio form, so there is very 1989 by Peter McGrath, released 1990). little else to say except perhaps that this is what today's golden age of audio is all about. There may be greater per­ formances than these in the catalog (bristling with names When 1 heard excerpts from this for the first time at like Casals, Heifetz, Rubinstein, etc.), but this combination the 1990 Summer CES, 1 went to the Harmonia Mundi booth, identified myself to the man in charge, and said, "I of lovely playing and you-are-there sound is a pleasure can't live without this CD." He took pity on me and gave unique to our era. me a copy for review. What 1 didn't know then was that there's a certain sameness to all these florid, virtuoso bass Piston arias (if you've heard three, you've heard all 17), but taken a few at a time, they remain absolutely dazzling even after Wa lter Piston: Symphony No. 2; Sinfonietta; Symphony No. 6. repeated exposure to them. Seattle Symphony (in the Symphonies); New Yo rk Chamber Sy m­ Antonio Montagnana was the number one bass of phony (in the Sinfonietta); Gerard Schwarz, conductor. Delos DE Handel's oratorio company in 1732 and 1733; he was fa med 3074 (DDD, recorded 1988/1989 by John Eargle, released 1990). for his range, intonation, and superaccurate leaps. Handel composed Montagnana's showstoppers specifically to dis­ This is the CD John Eargle trots out these days when play his vocal equipment; the tendency toward repetitious­ he wants to show his stuff to the fa ns. Enough said. The ness is therefore not surprising. A good thing bears some Symphony No. 6, especially, is state-of-the-art orchestral re­ repetition, however, and this music is definitely a good cording, in transparency, spatial perspective, dynamics, and thing. freedom from even momentary nasties (it was the last to be The surprise here is the English bass, David Thomas, taped). To use John's own terminology, both "texture" and widely known as a highly competent and thoroughly musi­ "structure" are close to perfection. Again, the credits say cal baroque specialist, but not as a superstar. Here he is one. "Digital Recording: Sony" without specifying the actual en­ His stylistic confidence, vocal security, and general panache coding equipment (see also the Bernstein/Barber/Gershwin remind me of Ezio Pinza, although he may not quite match review above), so 1 don't know whether that had anything to the latter in sheer beauty of voice. His bottom notes are do with the new sonic heights scaled here. John claims it astonishing; his energy is inexhaustible. You shouldn't deny was simply the right orchestration matching the right hall on yourself the experience of hearing this. What a singer! And, the right day; he didn't do anything special. Modestman. for once, find myself in total agreement with McGegan's 1 1 am quite thrilled by the Seattle/Schwarz revival of handling of the often brilliant orchestral parts-gorgeous mid-2Oth-century American symphonists (Walter Piston, playing, wonderful articulation. Howard Hanson, David Diamond, etc.) because it has made The recording is something of a departure for Peter me more keenly aware of the ridiculous musical snobberies McGrath; the venue is a big sound studio at Lucasfilm in prevailing during my formative years. My generation was California, very different from his usual school chapels. 1 expected to delight in Schonberg, Webern, and Elliott Cart­ find the recorded sound to be a bit too reverberant for the er, not to mention Stravinsky's dodecaphonic dotage, while music, but every note is crystal-clear and sweet, and the bal­ smiling condescendingly at the highly accessible works of ances are perfect, so it really comes down to a matter of in­ those neoclassical and neoromantic eclectics. What non­ dividual taste. 1 assume this is another Schoeps omni job, sense! Wa lter Piston is more enjoyable than the twelve-tone but why no SPARS code, Peter? Have you gone under­ contortionists and at least as good a craftsman. If you like ground as an AAD diehard? Shostakovich (and even the old snobs allowed that, since Stalin had forbidden him to compose like Alban Berg, Mendelssohn/Schubert right?), then there's no reason why you shouldn't find Pis­ ton's leaner, more classically restrained, but just as "popu­ Felix Mendelssohn-Bartholdy : Piano Trio No. 1 in D Minor, Op. lar" idiom equally appealing. The orchestration alone is 49. Franz Schubert: Piano Trio No. 1 in B-ftat Major, D898. The worth your careful attention. And the Seattle Symphony is Rembrandt Trio: Va lerie Tryon, piano; Gerard Kantarjian, violin; getting better all the time. Keep 'em coming, Delos. 66 Ravel phrey, tenor; Glenn Siebert, tenor; Myron Myers, baritone. Te larc CD -80212 (DDD, recorded 1988/1989 by Jack Renner, released "Nojima Plays Ravel. " Maurice Ravel: Miroirs, Gaspard de la 1990). nuit. Minoru Nojima, piano. Reference Recordings RR-35CD (DDD, recorded 1989 by Keith Johnson, released 1990). Go to Track 9, 2:28, on this CD. Listen to the Et in­ carnatus est in the Credo of the E-flat mass. Isn't this the When I raved about Nojima's Liszt in Issue No. 12, I most beautiful music in the world? And hardly anyone ever was reacting to his astonishing combination of technique, talks about it! I must admit, of course, that I became addict­ drama, and lyricism. The man has the "chops" to play any­ ed to it on the basis of the Erich Leinsdorf recording from thing accurately at any speed and any volume, but he is the 1960's (with the Berlin Philharmonic and St. Hedwig's equally capable of expressive, singing phrasing when the Choir on EMI), not because of this performance. Leins­ music calls for it and doesn't feel the need to showcase his dorf's firsttenor in that magical , sunlit nativity passage was strength when that's not the point. Liszt gives him the op­ the one and only Fritz Wunderlich; Shaw's soloists are not portunity for such contrasts of power and gentleness; Ravel in that league-but who is? Shaw's choral work, on the oth­ is another matter. Ravel is cool, in the jazz sense. Rhythm, er hand, is outstanding, maybe better than Leinsdorf's, and color, dynamic nuances, precision are the name of the game the many great choral passages come off splendidly (for ex­ in Ravel, and it's a game that Nojima plays equally well. ample the brooding Crucifixus that alternates with the Et in­ These are awesome performances. carnatus). This is late Schubert, composed in the last year This is supposed to be some of the hardest-to-play pi­ of his life, when he produced nothing but towering master­ Gaspard La nuit), ano music in the world (especially the de pieces. The early Mass is a simple, tuneful, and unimportant but you wouldn't know it listening to NOjima. He is experi­ work. encing about as much difficulty as if he were playing Fur As a footnote, I'd like to observe that Bach, in his B ELise. The characteristic sonorities and coloristic effects of Minor Mass, treats the incarnation as a dark mystery al­ Ravel come through in his playing with greater conviction ready foreboding the pain and tragedy of the crucifixion, than I have heard from anyone since Gieseking, and the which is then followed by the unbridled joy of the resurrec­ slower passages have continuity where lesser players tend tion-whereas to Schubert the incarnation is a pastoral to fragment them. And when flying fingers are needed­ idyll, the crucifixion a fa teful disaster, and the resurrection wow! At moments the piano whistles like the wind. Of almost a matter of course. The Protestant versus the Catho­ course, Ravel is Ravel, not Beethoven. He doesn't transport lic perception, one could argue. you to Elysium (not me, anyway), and I don't expect the As for the sound, you can always assume that a Jack performer to make him sound transcendental. His music is a Renner choral recording will be excellent, and that is defi­ feast for the ear, not a probing of our emotions, and Nojima nitely the case here. Transparency, inner detail, spatial quali­ does it full justice. ties, dynamics, transients are all what they're supposed to The recording by Keith Johnson makes the upper be at the summit of the art. Surprisingly, both Sennheiser strings of the piano, so important in this music, sound more and Schoeps microphones were used, so my theories about vivid than I ever expected to hear through loudspeakers­ that are shaken, but then a choral/orchestral recording with absolutely stunning-but the hiss from the microphone elec­ solo singers has its own special rules. I love Schubert more tronics obtrudes even more than in the Nojima/Liszt album, than I love my theories, so I say- nice disc. probably because there are more pianissimo passages. I but­ tonholed Keith at the CES and asked him about this; the Verdi problem seems to be the very low output from the Coles figure-eight ribbon mike combined with less than ideal Giuseppe Ve rdi : Messa da Requiem. Oberlin Musical Un ion & Or­ transformer matching to the preamp input. Keith has various chestra, Daniel Moe, conductor; Carolyn James, soprano; Susan fixes in the back of his mind but hasn't gotten around to To th Sh afe r, mezzo-soprano; Franco Farina, tenor; Gerald Craw­ them yet. To archive such a rare performance, I personally fo rd, bass-baritone. Bainbridge Records BCD2103, Discs 1 and 2 would have opted for a hiss-free B&K recording with possi­ (DDD, recorded 1988 by Brad Miller, released 1988). bly a smidgen less upper-string realism, but that's a value judgment which may not be Keith's or Tam Henderson's. This not a very new recording, but I received it only Hiss or no-get this CD. recently and find it rather interesting. It documents a live, large-scale musical event-in celebration of the lOOth anni­ Schubert versary of the Ve rdi Requiem's first performance at the Oberlin College Conservatory of Music-strictly from the Franz Schubert: Mass No. 2 in G Major, D167; Mass No. 6 in E­ perspective of the conductor. To quote Brad Miller's note : "Our intent was to archive an 'event' as it occurred; or to be ftatMa jor, D950. Atlanta Symphony Orchestra & Choruses, Rob­ ert Shaw, conductor. In Mass No. 2: Dawn Upshaw, soprano; Da­ more succinct, digitally record this 'live' performance from vid Gordon, tenor; William Stone, baritone. In Mass No. 6: Benita the perspective of Daniel Moe. A single microphone posi­ Va lente, soprano; Marietta Simpson, mezzo-soprano; Jon Hum- tion was chosen, center stage and elevated over the podium.

67 The Colossus, a state-of-the-art 4-channel digital recording I must attribute this success to the hardware-in which case system, employed a direct feed from the producer's own other recording companies should take notice. MS-4 microphone. No outboard mixers or amplifiers were It has become something of a cliche that the Manzoni used. The MS-4 is a discrete 4-channel surround (quadra­ Requiem is operatic rather than religious in spirit, but what phonic) device, with a frequency response of 2 Hz to 20 does that really mean? I think it means that Ve rdi always kHz, ±2 dB. The microphone is DC servoed, which main­ sounds like Verdi-but so does Beethoven, in which case tains low-frequency imaging very precisely, and will handle the Missa Solemn is is symphonic, right? And isn't Parsifa l sound pressure levels (SPL) of 146 dB before clipping or religious in spirit even though it's an opera? I admit that the distorting. The front and back channels were mixed together Recordare in the Dies [rae of the Requiem could be straight for this stereo compact disc ..." out of Aida, but that's something pretty good to be straight What do I think of the sound? Stupendous-but. .. out of, as long as the singing is good. In this performance The absence of any other microphone than the MS-4 makes the singers are competent but not great, except the mezzo­ the spatial perspective and overall balance utterly natural, soprano Susan To th Shafer, who is superb. Her lower range, uncomplicated, and plausible--the best I've heard. On the especially, is quite thrilling. The playing of the conservatory other hand, the words of the choir would have been much orchestra is on a respectable professional level without pos­ more distinctly audible with helper mikes, and I think any ing a threat to the ranking of their colleagues in nearby large commercial recording company would have opted for Cleveland. Daniel Moe paces and phrases the music effec­ them, purism be damned. I'm not sorry, however; the disc tively, but then anything by Ve rdi more or less plays itself; as it is makes a wonderful imaging test tool. In every other the "interpretation" is built in. (The man was a pro.) way-dynamics, low-level decay, timbral accuracy, lack of Overall, I'd have to say that this recording makes distortion-the recording is fantastic. (No "ouch!" on the la­ more of an audio statement than a musical one, but it is far dies' fo rtissimo top notes, not even a trace.) Since no subtle from inferior musically. <) artfulness, as such, was possible in microphone deployment,

Seminar 1989 music at, something corresponding to full CARVER: I suppose, I imagine that they (continued from page 62) modulation of the medium, and then play somehow look through the thing, find the tion-isn't it?-with amplifiers. How close them a -90 dB tone-and they'll say, you loudest passage, and park it right up at the can you come to the slew rate before it's haven't played anything; they'll say there's top-and he has his 98 dB. unacceptable ... nothing on. That's what they'll say. Then LIPSHlTZ: Not just that, but multitrack CARVER: It's like falling off a cliff. you say, go to the loudspeaker, put your ear digital. LIPSHITZ: ...how much margin do you there. Oh yes, ah! You know, 90 decibels is CARVER: But when you're making the re­ need? really ... cording you don't have that luxury; you CARVER: As long as you don't fall offthat EARGLE: It's a lot. need the extra 10, 12, 15, maybe 20 dB, to cliff, you're perfectly okay. You can come LIPSHITZ: People don't quite appreciate make sure you never run out of headroom. as close as you please. it. Now, yes, the noise floor in a quiet room LIPSHlTZ: You want some at the top, and LIPSHITZ: There are people who believe like my listening room is just at the level of it's also good to have some at the bottom, you want a slew-rate margin of a factor of the subliminal. You can perceive the noise so when you're mixing twelve channels to­ 10 so that you never come close to the lim­ floor of the digital system; it's not below gether, you don 't get twelve 16-bit noise it. There are other people who believe that my threshold. It would be nice if it were be­ floors added together but twelve 18- or 20- you can come up to 0.9 of the slew limit low; a few dB more might be nice. But, bi t noise floors. and not be able to tell that you're there. goodness, for a medium designed for hun­ CARVER: So that's where your 20-bit EARGLE: And there are people who will dreds of millions of people out there at an would be, yes. tell you that a bridge ought to be built with affordable price, it's not bad. It's pretty CLARK: Essentially, there should be a dif­ 10 times safety fa ctor in every area-and damn good. Now, the professional I think ference between pro and consumer ma­ it's a prohibitively expensive bridge but it needs more than 16 bits because he needs chines then. will never crash. to be able to do some manipulations and CARVER: Yes. LIPSHITZ: But they never get the contract mixing and headroom and other things, and EDITOR: What is the widest analog dy­ because their price is too high. be able to get 16-bit signal-to-noise in the namic range that you 're aware of? Or the EARGLE: That's right. end. lowest analog noise floor-whichever way LIPSHITZ: But anyhow, on this number­ EARGLE: Are you going to go on record you want to put it. Isn't it around 80? of-bits question, I would say the following. wi th that statement? EA RGLE: No. More than that. I'm allow­ A 16-bit system gives a 98 dB signal-to­ LIPSHlTZ: Ye s. ing the use of very artful noise reduction. I noise ratio in principle, undithered. EARGLE: Good. would say that normal lS-ips tape, with EDITOR: 98.1, right? CARVER: Now I understand what your ob­ Dolby SR, is going to crack 100. EARGLE: An even wider dynamic range. jection was. And you just brought it out, EDITOR: So that's 100 versus 98.1. LIPSHITZ: The dynamic range is wider be­ Stanley. Because you need that headroom. CLARK: We ll, then there's dbx, too, which cause you can hear things below the noise. You'd love to know that you're 12 dB is ... But that's true of any system. In practice, in away from clipping, or something like that, EARGLE: There's dbx, which is much a properly di thered system, you're talking when you're operating your machines or more audible, more of the time. more than 90 dB. Now, for most people, it editing your recordings. And the person CLARK: But still, 120 dB. is real ly quite eye-opening to play some­ who makes the final copy that we as con­ EARGLE: I've never heard a peep out of SR. thing very loudly, perhaps significantly be­ sumers will listen to, he ... (This is where we must pause. To be contin­ yond the level they would ever want to play McGRATH: Will get the full 16. ued in these smaller installments.)

68 Subscription Information and Rates First of all, you don't absolutely need one of our regular subscription blanks. If you wish, simply write your name and address as legibly as possi­ ble on any piece of paper. Preferably print or type. Enclose with payment. That's all. Secondly, we have only two subscription rates. If you live in the U.S., Canada, or Mexico, you pay $22 for one year's subscription (four issues mailed at approximately quarterly intervals, barring unscheduled delays). If you live in any other country, you pay $32 for a four-issue subscription by airmail. All payments from abroad, including Canada, must be in U.S. fu nds, collectable in the U.S. without a service charge. You may start your subscription with any issue, although we feel you should have your own copy of Issues No. 11, No. 12, No. 13, No. 14, and No. 15 (the one you're reading now). That way you'll have a record of where we stood on various subjects when we resumed publishing after a hiatus of almost seven years and gain a better understanding of what The Audio Critic is all about. Please specify which issue you want to start with. One more thing. We don't sell single issues by mail. You'll find those at somewhat higher cost in selected audio stores.

Address all subscriptions to The Audio Critic, P.O. Box 978, Quakertown, PA 18951. In the next issue: We return to our favorite subject-yes, loudspeaker design-with reviews of the Velodyne ULD-15 Series II, Snell Type C/IV, JBL XPL160A, Audio Concepts "Sapphire II," Cambridge SoundWorks Model Eleven, Carver "Amazing" Platinum Mark III, and others. Deep bass from small speaker boxes: a new contributor discusses in great detail all the available alternatives. Reviews of high-end electronics: Coda Technologies 01 FET preamp, Esoteric P-2 CD drive unit and D-2 multi­ D/A converter, Philips LHH500 CD player, and more. A remedial course for speaker cable tweaks and cultists. Surround-sound processors and a bit of high-end video .