| Home | Audio mag. | Stereo Review mag. | High Fidelity mag. | AE/AA mag. |
| CORRECTIONS & AMENDMENTS OUR APOLOGIES TO Detroit's 40-member audio construction club whose name as it appeared in the letter of four of its members in 4/79, p. 57 may have led some searchers to the wrong corner of Michigan. We said, in explaining the club's acronym SMWTMS (pronounced, we understand, smootyms) that the letters stood for South Western Michigan Woofer and Tweeter Marching Society. Any of you who are wandering around in the wilds of Southwestern Michigan should head for the other side of the state to locate the Society: SMWTMS = Southeastern Michigan Woofer and Tweeter Marching Society. They welcome new members too: drop a card to David Carlstrom at 10155 Lincoln Dr., Hun tington Woods, MI 48070 for details on the next meeting. -Ed. COMPEX MISS IN REG WILLIAMSON'S article on his "Compex" Expandor/Compressor system, (TAA, 3/79, p. 15), the expandor parts list has two listings for R4. The first of these is a correct value, but wrong designation. It should have been listed as RS and the value 5k6. (This means 5.6k) R4 is 1Meg. 40/40 FAULT Although R, is shown correctly on the schematic , Fig. 1, of Reg Willamson's 40/40 amp, (4/79, p. 7) it is incorrectly listed in the parts list below. R3 is 10k t 5 To or better, not 820k as shown in the parts list. The alternate transistor type for TR4 is the MPS U56 and not the BFX40 as the U06 as the parts list shows. JUNG-WHITE PAT/5-WJ-1A THE PARAGRAPHS BELOW were originally submitted for publication with the PAT-5/WJ-lA notes and updates (TAA 4/79, p. 58), but did not appear. For proper context, they should have appeared just prior to the second from last paragraph (beginning with "Allied to the above..."). -WALT JUNG AS MAY BE NOTED from the discussion in the article on the PAT-5/WJ-1A (3/79) the exact components chosen for use in the power supply are critical to the extent of the design's aural quality. One of the last few of these components to be finalized was the pair of t 15V bypass capacitors, that is, the 5uF units mounted on the PC-34 cards. Initially, it was hoped that polyester units would suffice here, and the limited number of 5uF poly propylene units on hand prevented their test for this application, until after 3/79 was wrapped up. As it turns out, the 5uF polypropylene as t 15V bypasses are much superior to the polyesters mentioned in the article, and this is reflected in superior sound (smoother highs). Accordingly the Old Colony kits will use the polypropylene units, so that the builder is availed o£ this highest potential in sound. In addition to the TRW units, Electro-Cube W950D types have also been found suitable, for coupling and bypassing. For scratch builders, these units will be next to impossible to obtain in small quantities, so othet units will no doubt be desirable for substitution. Use polycarbonate types if you can, or use the polyester unit if nothing else is available. You can operate the circuit even without this capacitor pair, (while temporarily waiting), as there is no stability problem. It is recommended however, that you eventually incorporate it, if at all possible, as it does make a difference in a top rate system. OC-2N AND OC-2H OP AMPS: WHAT AND WHY ONE OF THE SORE issues from the readers' stand point regarding the PAT-5/WJ-lA preamp kit has been the subject of the OC-2N and OC-2H op amps. Apparently some number of readers are unhappy with the approach taken with regard to controlling these devices and their specifications. For the record, it should be noted that the fundamental interest towards assigning "house numbers" for these devices was rooted in protecting the kit builder from the use of a loose tolerance IC, and the resulting uncontrolled DC offset. Obviously no one wants to fry your power amp or speakers for you, so it was decided to totally control this situation by providing defined limits for these two devices and make them available through Old Colony. This resulted in numerous letters and indignant phone calls, some bordering on the irate. From the viewpoint of these readers, we apparently have done a grave disservice in not allowing them the freedom to install other IC's. Unfortunately we don't see it this way, and we still feel our first responsibility lies more towards caution and safety. Nevertheless, we know this is a sore point and we feel compelled to make a qualified disclosure as to the "mystery" IC's. If readers then want to procure their own IC's and install them in the preamp, it is entirely up to them to verify correct DC operation. The OC-2N is the more critical of the two IC's, and must have an input offset voltage of 2mV or less. This can be satisfied by selecting from a lot of LF351N's (National, available from Digi-Key) devices, or preferably using a device such as the LF351AN, which has a factory 2mV spec. Other devices may also work for the OC-2N function, such as Texas, Fairchild, or Motorola 2mV (or less) offset FET units, but we have not evaluated their sonic characteristics. The stipulations above apply to the JW-33(A) circuit, as used with the DC servo control, which is preferable for best sound. Alternatively, if the capacitor output option is used, the DC offset is not as critical and the LF351N unit will function for IC 1. The DC offset limit for the OC-2H op amp is 1 mV or less, since it does not have benefit of the servo control to null its error. The device type optimum for this function is a selection from the Analog Devices AD544 series, described below. Note that these devices and the LF351 types, while generally similar, are not equivalent, and cross substitution is not recommended. Each of them was selected for the optimum combination of characteristics for the specific mode of use. A new, precision FET input operational amplifier which features a typical slew rate of 13V/us, in combination with guaranteed warmed-up input bias current of 25pA, maximum, has been introduced by Analog Devices. The AD544 also features low noise of 2uV peak-to-peak over 0.1 to 10Hz, low quiescent current of 2.5mA, max i mum, low input offset voltage of 0.5mV, max i mum, and minimum open loop gain of 50,000 (RL > 2k-Ohm). The device has a typical unity gain, small-signal bandwidth of 2mHz. Input offset voltage drift is available as low as 5uV/'C, maximum. The AD544 settles to ±0.01% in 3us, making the device applicable as an output buffer for fast digital-to-analog converters. Other applications . include uses in sample-hold circuits, high speed integrators and use as a front-end buffer for digital voltmeters and other instruments, as well as any application where excellent and ac and do performance is necessary at low cost. The high common mode rejection ratio of 80dB, minimum, ensures 12-bit linearity in high impedance buffer applications, for A/D conversions. The AD544 is available in J, K, and L grades, for operational over the 0° to + 70°C temperature range. Packaged in a TO-99 metal can, the device is available from stock for $3.50 in 100's, for the K version, which has an input offset voltage and in put offset voltage drift of 1.0mV, maximum, and 10 AV/°C, maximum, respectively. We evaluated early samples of this device as part of the development cycle in our PAT-5/WJ 1A preamp modification (TAA 3/79), and found it's audio performance parameters to be quite favorable. It was designed into that project in the high level stage, in the AD544K version. Old Col ony is making this device available to the readers as the OC-2H. For Engineering Information, please contact Don Travers Analog Devices Semiconductor, 804 Woburn Street, Wilmington, MA 01887, (mentioning this article). We hope the above discussion clarifies the situation for readers, and clearly states the subject of responsibility. Readers are free to use the types mentioned as specifically recommended, which will realize the performance of the WJ-IA mod, when properly operating. However, under no circumstances will we answer correspondence concerning complaints due to substitution of other ICs and resulting malfunction of the DC operating points. -WALT JUNG UPDATES FOR THE JUNG/WHITE PAT 5/WJ-IA IT HAS BEEN FOUND that improvement in reproduction occurs with two minor modifications to the PAT-5/WJ-lA circuitry (reference TAA 3/79 issue). The first of these is a value change of R„ 9 to 10 megs (was 1 meg), which serves to further minimize loading on the volume control. The other is the addition of a pair of 220uF/50V radial lead electrolytic bypasses across the 15V zeners in the power supply (reference Fig. PS-2, p. 25). These two changes provide increased mid and high frequency naturalness, while the latter one also increases bass response and separation. There has been some confusion about the exact state of the signal path now, particularly in the HL section. This is no doubt due to the nature of the update article, and we regret the confusion. We consider the high level signal path to now consist of the modified input buffer (1/78), which can be alternately bypassed with the EPL switch (see letters 1/79). The high and low filters are still available via this signal path, but we freely admit the sound quality of this path does not approach that of the phono direct. Some readers have criticized the approach taken in presenting the concepts inherent in this design, implying they are weakened by being tied to a non-essential PAT-5 chassis. While in an ab solute sense of things this is indeed true, we'd like to express our reasons why this was done. One of them was to get the new information out as soon as possible, which was accomplished. This is so anyone, scratch builders as well, can freely use the ideas presented. The second major reason was to minimize the impact on those who had implemented the first kit-they can update what they have with no new PC cards, and their investment has been protected. A new from-the-ground-up design would obviously be much nicer and prettier, but would not appear before a year's time and certainly make few friends with the initial buyers. In the meantime, no one has had the benefit of the new ideas. Dave and I are constantly working on new ideas and ways to improve (as exemplified above), and hope in time to incorporate them into a new "ground up" preamp. In the interim we will pass along ideas like those above, when substantial sound benefit with ease of implementation is inherent to the particular idea. -WALTER JUNG HEATH'S EDUCATION: PRO or NO? THANK YOU FOR THE REVIEW on our Electronic Circuits course in your issue 1/79, p. 44. Your reviewer, Ronald Miller did an excellent job and we thank him. We do take issue, however, with his comment regarding the ability of our courses to train technicians. While no where do we actual ly claim in print as Mr. Miller indicates that our courses are designed for training technicians, our courses are nevertheless very widely used for specifically this purpose. Since introducing our courses in 1975 several hundred thousand have been sold. The greatest percentage have gone to individuals for self improvement and hobby purposes. However another large percentage has gone to industry and government for employee training at the technician level. In addition, the courses are widely used by community colleges and universities for technician training. As an example, the State of Missouri recently adopted the Heath courses for state-wide standardization of technician training programs in community colleges and voc/tech schools. Our industry buyers list looks like the Fortune 500. One of our best government customers is the tough and fussy FAA. The IEEE recently adopted our courses and approved them for continuing education for their membership. While we appreciate Mr. Miller's excellent and objective review, we would most certainly be interested in hearing his opinion as to why he does not believe this course is qualified for technician training. - Louis E. FRENZEL; Director, Education & Publishing Heath Company; Benton Harbor, MI 49022 Miller replies: It sounds as though Mr. Frenzel did not read my entire review. 1 said "Heath's claim (which has been toned down in recent catalogs) that these courses will train an electronic engineer or technician is overstated. They lack a pattern of setting design problems and giving examples of correct solutions after the student has tried to solve them. In the same way, insufficient effort is made to deal with the kind of low distortion taken for granted in the current state of audio. " (TAA 1179 p. 46) 1 made grades of 88, 100, 98, 92 respectively on the unit exams of the four courses and 1 still don't know why the phono preamp I have built from a design and circuit board layout from Wireless World oscillates at 1 megahertz. If I had been trained as a technician I should have no problem. For a good grasp of theory, the Heath courses are commendable; for a technician there needs to be a different level of practical experience. It may be that the State of Missouri, the FAA, and the IEEE expect such practical experience to be part of the overall training. THE JC-2 SAGA CONTINUES IN ISSUE 3/77 you presented under "Classic Circuitry" the JC-2 schematic. John Curl warned that matched FET's were required for proper operation. The matching information appeared under the phono preamp module; I assume it also applies for the output amplifier module? I was interested since I had read excellent reviews of the JC-2 but never had the bucks to go out and buy one. As I had no way of matching the FET's, lack i ng access to a FET curve tracer, I forgot about trying to build the circuit. Along came Issue 2/78 with a corrected schematic and an offer from Messrs. Garrington and Dewey to aid constructors and supply matched FET's. I did obtain several FET's from them but did not order all I needed at the time. They also provided much information for which I thank them. However, when I tried several months ago to order the parts for the output amplifier module, my check was returned and I was advised they could no longer supply parts of answer questions on the JC-2. Take note, any readers still interested in building the JC-2. I would therefore like to solicit help from any other readers who have built the JC-2 and might have the means of matching transistors. I need to obtain enough matched FET's for two output amplifier modules; or, better still, if you have a method of matching them without a curve tracer, I would appreciate hearing from you. Incidentally, ADVA Electronics ( Box 4181A, Woodside, CA 94062) have the FET's in their catalog. The E110's are $2.10 each, the E175's are three for $1.00, and the E212's are 75¢ each. -R. D. DARROCH; 1807 Elm Crest Dr. Arlington, TX 76012 PAS: AMENDMENTS I RECENTLY COMPLETED David Vorhis' re-modification of the Dynaco PAS preamp (TAA 2/76 p. 21). Although the improvement was significant, the bass response has not been satisfactory: it seems to me to roll off below 80Hz. Replacing the 47K positive feedback resistor restored bass to a satisfying and also realistic level by comparison with other preamps such as the Rappaport or the Hafler. I suggest this change to other readers who have experienced a similar bass decrease, and would be interested to hear the results. I believe several other modifications further improve the PAS sound. They combine concepts originated by Frank Van Alstine, Audio Dimensions, and Audio Research. 1. Increase decoupling of the B + power supply by adding 100-200uF electrolytics just before the high and low level boards. 2. Parallel all capacitors with silver micas rated at 400-1000uF. 3. Follow Audio Dimensions' circuitry for both boards. 4. Split the circuit paths on both boards between the feedback circuits and the coupling capacitors, using 400V (or higher) capacitors. I'm using 1uF values with excellent results, but this requires some surgery. 5. Increase the value of the filament supply electrolytics. I am currently using two 36,000uF capacitors. To virtually eliminate DC ripple probably requires a total minimum value of about 20,000uF. I also use silicon diodes. 6. Rewire the input selector switch to provide these phono inputs with various capacitive loads. 7. Leave the rectifier tube in place. I base this recommendation on Audio Dimensions' comment on the potential increase in noise level caused by high initial voltage. My early attempt to drive a pair of Dynaco Stereo 70's and large Advents directly from the low level board provided inadequate volume, although the sound was very clean. In listening tests I compared this modified PAS with the Hafler and the Rappaport. The PAS was clearly superior to the Hafler in dynamic range and in both bass and treble clarity. The Rappaport showed tighter bass and perhaps a faster response (slew rate?), but nearly equivalent high ends. I conducted the tests with Kenwood L07M's and the Dahlquist DQ10's with improved crossovers. These modifications have used up all available space inside the PAS cabinet. I plan to increase the filament supply voltage to 12.6V (per ARC and Audio Dimensions), and to increase to about 400V the B + voltage available to both boards; to do this I shall probably have to move the entire power supply into a separate box. I also plan to build a super buffer (TAA 4/77, p. 18) into the tape outputs, as I notice a substantial difference when I disconnect my tape deck. Using a buffer for the primary amplifier outputs would also make the impedance matching easier (changing the load resistors does become inconvenient). Although perhaps not state-of-the-art, I believe these modifications push the PAS to a new order of performance at an extremely reasonable cost. Shame on Dynaco for discontinuing this classic amplifier! ALEX R. WEIBEL, III; Coraopolis, PA 15108 USA LIPSHITZ/JUNG FORUM WE WOULD LIKE TO REPLY to the letters which TAA has published in 3/79 and 4/79 concerning our tests: We never intended to withdraw our letter, but simply take into account Moncrieff's objections, by replacing the word "conclusion" by "hypothesis" in two places. Regarding our null result and its significance, we point out that initially we did hear differences, which were excised by reducing the linear error. We are not meter readers. Reread our original letter to appreciate the sensitivity of our tests. The details of our tests were provided to enable others to duplicate our test. You must provide evidence and not speculation, if you believe that audible differences existed which we failed to hear. We believe that subjective evaluations are worthless to others unless they are scientifically con trolled. We know of no alternative to double-blind testing. We note with regret that our critics have to our knowledge not yet reported any scientific evidence to overthrow our hypothesis, or to support their own. The only way out of the impasse that exists is for controlled double-blind testing to become widespread. We suggest a good randomizing A/B box is the most convenient, and its lack of audibility is also testable. The SMWTMS letter, TAA 4/79, p. 57, is a source of further details. A growing number of other careful investigators are independently reaching similar conclusions. We no longer feel alone. Can you afford not to read the following? 1. Audio Scene Canada, Vol. 14 #5 (May 1977), pp. 44-50. 2. Hi-Fi News &Record Review Vol. 23 #6 (June 1978), pp. 73-77.; Vol. 23 #7 (July 1978), pp. 49-51; Vol. 23 #11 (Nov. 1978), pp. 110-117; Vol. 24 #2 (Feb. 1979), p. 67; Vol. 24 #4 (April 1979), p. 81; Vol. 24 #5 ( May 1979), pp. 81-83. 3. Wireless World, Vol. 84 #1511 (July 1978), pp. 55-58. 4. Boston Audio Society Speaker, Vol. 7 #9 (June 1979). 5. Hi-Fi for Pleasure, Vol. 7 #7 (July 1979), pp. 45-73. We all want progress in audio. We believe the above represents the only rational approach. STANLEY P. LIPSHITZ; JOHN VANDERKOOY; PAUL YOUNG---University of Waterloo Waterloo, Ontario, Canada METHODOLOGICAL EXCEPTION AS I UNDERSTAND the ongoing exchange in recent TAA Letters columns between Walt Jung and Stanley P. Lipshitz et a(., the latter claim all preamps should sound the same if their RIAA curves are exact: they A/B tested a stock PAT-5 against a JW33 through identical electronics into Quad electrostatics, and voila-no difference! Q.E.D. I would like to take methodological exception to that. The Quad owner's manual states, on. p. 11, "Maximum Output: 6 ft. on axis in free space, 93dB referred to .0002 dynes/cm^2 i n frequency range 50Hz-10kHz." As the owner of two pairs of Quads I can attest to the fact that a Quad does not reproduce much below 50Hz. One of my torture tests for a preamp is how it handles heavy bass transients (I also have Hartley sub-woofers) like those found in a digital recording's bass drum shots which are mostly 36Hz. Clearly the Lipshitz et al. tests could not differentiate between the preamps at such low frequencies. More significantly, most preamps don't misbehave until driven to near peak levels. Usually aberrations occur when the preamp is swinging large amounts of current. A full orchestra crescendo will drive a mis-designed preamp into ringing, smearing, constricting, blurring, image drifting, or gross forms of less subtle distortion like break up or inter modulation. I doubt that Lipshitz et al. could drive any preamp to peak levels using only two "polite" Quads. At 93dB most preamps would be loafing and any sonic differences would inappreciable. In other words, driving Quads is not a torture test designed to reveal the flaws of a preamp. Driving inefficient but accurate speakers to 110dB or better would be more like it. So I question the validity of their method: I don't think it tests what they think it tests. As to the concern for linearity: if RIAA ac curacy were the primary consideration in preamp design, what would happen if one built a preamp out of the lowest quality component parts, or mis-designed it on purpose, yet calibrated the RIAA section as accurately as Lipshitz et al. suggest? One could have considerable distortion originating in the preamp, though it would be RIAA-correct. Moreover, not all forms of audio distortion are measurable. Marine biologists report the shark can discriminate blood in parts per million of salt water, and we marvel at it. Yet man--each of us--is pretty marvelous. The human eye can detect light better than any device and the human nose can discriminate between odors better than any process yet devised. This is staggering considering the high level of technology in optics and television, or in the chemical techniques involved i n quantitative measures in modern biochemistry. The human ear, linked to that fabulous bio-computer called the brain, can discriminate with equal facility. I'm told by a designer of quality audio components that one part per million audio signal in the B+ circuit of a preamp is audible. Indeed, the designers of high-quality amplifiers and preamps can differentiate between the "sound" of mylar, polycarbonate, polystyrene, and tantalum capacitors though they are all rated at identical values. Dick Marsh and Nelson Pass have commented on the ability of various capacitors to introduce distortion. Similarly, carbon, wirewound, ceramic, and metal film resistors seem to have characteristically different "sound" though rated at the same value. Metal films, for example, dramatically reduce noise. Indeed, noise can be defined as a form of distortion. A given transistor may vary from one manufacturer to another, and even from lot to lot of the same manufacturer. FET's, MOSFET's, and standard transistors each behave differently and "sound" differently though they have the same rating. Audio designers know this. Power supplies are an often overlooked element because preamps don't draw as much current as power amps. Yet if the power supply is poorly designed various problems may arise; solo instruments may drift from left to right giving poor imaging, or bass may become muddy giving evidence of overshoot or ringing on tone bursts. Yet the power supply could be theoretically correct, well-regulated, with enough capacitance. Feedback (and here I'm admittedly getting beyond my depth) may be improperly employed and constrict signals something like a compressor or a peak-limiter, making potentially exciting digital recordings sound as though their extraordinary dynamic range had been broadcast by a compressed FM station. I don't know exactly how it works, but I know it when I hear it in a preamp. +++++++++++++++++++++ Capacitors, resistors, transistors, power sup plies, and feedback may degrade the sound quality of a preamp, each in a characteristic way. Im properly used they could introduce gross distortion as well as the more subtle and hard-to measure things. Yet such a mis-designed preamp might cruise along at 93dB and be unobjectionable. It would noticeably misbehave at peak moments, those moments when the audiophile wants a preamp under control. And when it did misbehave, it would do so with scrupulously correct RIAA values. Perfumers put themselves through the exercise of changing formulas, ever so slightly, controlling for one variable at a time, in attempts to develop a marketable scent. They have trained themselves to discriminate among infinitesimal amounts of aromatic oils. The serious audiophile, it seems to me, might build a pair of PAT-5's, with a series of phono boards each using only one type of resistor, and teach himself the characteristic "sound" of the component in question. Having done that, he could systematically replace the capacitors: all mylar in Board A, polystyrene in B, etc. In this way, he could teach himself to discriminate among the various forms of coloration, or distortion, or "sound" introduced by each. Then driving his preamp to peak levels, he could listen for ringing, smearing, blurring, imaging, noise, constricting, edginess, TIM, THD, and SID. Having been a party to such an exercise on two or three occasions, I am aware it is an invaluable learning experience. I also know that RIAA accuracy is not the primary factor in the design or performance of a preamp. It is nearly a cliche to say a preamp should be nothing more than a wire with gain. Some achieve this elusive goal with obviously greater success than others. Clearly, there must be more to it than RIAA. I await the prototype Lipshitz et al. preamp to prove me wrong. If they are right, they could vastly improve the performance/cost ratio in preamps and the entire audio community would be overjoyed. I wish them the best. CHARLES HOLLANDER; Baltimore, MD 21218 ON HEARING & DYNA TUBES I WOULD LIKE TO ADD MY comments to the great Lipshitz/Jung debate and, in particular, to a remarkable pair of letters from Charles Repka ( TAA 4/79, pp. 55 and 59-60). Regarding modifications to Dyna power amps, perhaps a bit of history is in order. The design originated in the very early 1950's as a dramatic simplification of the then-new (D.T.N.) Williamson style amps. By reducing the complexity of the output transformer constructions and utilizing a high-performance television tube of the day, David Hafler designed a quality amplifier that was both inexpensive and easy to construct. Dynaco was never bashful about modifying this amp-I can name seven different tubes used in various models-but they never changed the circuit topology or (except for the MK VI) the transformer winding ratios. The only three parameters that distinguish these models are transformer weight, supply voltage/current, and tube selection. The reason today's MK III's sound inferior to the Stereo 70 and the original 1958 design is primarily tube selection. I therefore cannot see any objection in principle to experimenting with different tube types and, although the circuit can certainly be improved, I would be reluctant to make any major change in the efficient lines of the circuit design unless I thought I had a lot to gain. Shortly after the publication of my audio aid ( TAA 2/79, pp. 37-38), I received a letter from John McConnell which caused me to investigate using the output tube of the Stereo 70 (EL34/6CA7) in the MK III. While not as robust as the other types, the EL34 gives exceptional ser vice if you reduce the cathode current to about 100mA and jumper pins 1 & 8. I do not recommend the modification suggested by Mr. Repka. What he has done by moving his 16 ohm speakers to the 4 ohm tap is to quadruple the effective primary impedance of the transformer. This moves the operating point of the tubes from maximum power to well beyond the minimum distortion point. It will also considerably reduce power output, stability, and the ability of the amp to control a practical speaker load. Mr. Repka's second letter intrigues me more than the first. I think he is inconsistent when, in one breath, he appeals for tighter control over the methodology and interpretation of equipment tests while, in the next, he casually dismisses the value of the "current buzz words" (depth, ap parent dynamic range, transparency...). I have no difficulty demonstrating these properties, either from records or CBC Radio transmissions. As for a recording engineer minimizing distortion induced by microphone placement (principally aberrations in phase and polar response), I have two questions: First, can he eliminate the audible effects of these distortions or if, like reverberation, a little is desirable, what determines how much he leaves in? Second, if the amount of these distortions is greater than zero, shouldn't a properly functioning home system reproduce them as accurately as the program? I can also make similarly large changes in the above qualities by moving my Quads an inch or so--but I prefer to call it maintaining phase coherency and establishing proper polar coincidence. since Mr. Repka and I have similar power amps and speakers, I believe the difference in what we hear occurs somewhere between stylus and line out. (I reject out-of-hand Mr. Repka's argument that the golden ear is rare-most people can hear but lack the patience, training, or inclination to do so. The tin ears were eaten by dinosaurs.) All current audio sources, even studio master tapes, have gross amounts of distortion; it does not necessarily follow that the addition of small amounts of distortion will be negligible. First of all, distortion is not added, but multiplied. It is compounded like money as it passes from stage to stage. (See Neil Muncy's guest editorial TAA 1/79 p. 4), which implies that the distortion increases faster than the number of stages and that the distortion introduced by the last link is somewhat less critical than that introduced by the first. This is both a powerful inducement for simplicity and an unfortunate turn of events, since we can control distortion in the source only by selective purchasing. Nevertheless, enough good recordings-even rock albums-make it through to make it all worthwhile. How, then, can the ear make it through the jungle of distortions that characterize our efforts to preserve sound? The answer, I think, lies in the versatility of the ear itself. Let us first consider sensitivity; the ear is sufficiently sensitive to detect vibrations as minute as those of a molecule passing through its own diameter; if our hearing was much more sensitive, we would have to endure the noise of the air cur rents in our rooms. Yet the ear will adapt to sound pressure levels billions of times more intense, thanks to the logarithmic nature of its sensitivity. Next, let us consider selectivity: the same ability of the ear to pick out a single conversation in a crowded room by using the available pitch, amplitude, phase, and modulation clues is brought to bear during critical listening. I have not seen any good figures regarding the lowest level of audible distortion and do not expect to see any, since the ability of a signal to drown out its own distortion depends upon amplitude, frequency, and phase relationships between the signal and its distortion. For example, the distortion produced while the fundamental and all odd harmonics are at max i mum (clipping) is much easier to measure but less objectionable than a very much smaller amount of high order distortion occurring during each polarity reversal (crossover distortion). At these moments, the sound pressure level-and therefore the masking ability of the signal-approaches zero, leaving only a nasty spike of pure distortion. The ear cannot miss this and other audio "flags," even though conventional THD measurements are very low. My experimentation, which is basically limited to the so-called "linear portion" of the transfer curve, suggests that a maximum level of .001 would be a good goal to aim for in high-quality equipment. Since this level is beyond currently practical measurements, we must call upon the in strument in question, the ear, to perform the integration. Mitch Cotter, I believe, calls a compo nent essentially distortion-free if a signal can be passed through five such devices without noticeable deterioration; I would suggest that passing a signal through consecutive left and right channels might be a more economically feasible test for the average experimenter. Now that we have reviewed the dilemma, let me suggest the final property of the ear that makes convincing reproduction possible. After interception by the ear, information is sent to the brain (digitally!) where it is interpreted. We know that the eye mechanism constructs an image by making an initially very simple set of observations which rapidly become more detailed with the passage of time. At some point while considerable information is missing, the brain identifies the object, at tempts to integrate it into the environment, and eventually, attempts to indentify incongruous elements. I submit that the ear undergoes a similar pro cess. The normal process of hearing reproduced material involves a reaching-back technique that searches for prototypes on which to build a perception, that the very acuity of the ear in detecting distortion is what enables it to identify the signal amid all that distortion, and that the ability of the ear to ignore distractions makes it possible to suppress the distortions that accompany the signal. Eventually, the ear applies its resolving ability to error detection and we are in trouble. Certain elements, such as crosstalk, cross modulation, nonlinear phase shift, and feed back-induced "echo" cause the image to be rejected out-of-hand. Certain others, such as excessive lower-order distortion produce the real but-not-present "veiling" typical of the present state-of-the-art. In short, little things do make a difference. For example, today's experiments can obtain a given value of capacitor in a wide variety of dielectrics, each with slightly different transmission characteristics. Each should therefore sound slightly different. I am surprised that no one has yet taken the inferential step and pointed out that if this difference is significant, placing a capacitor i n a feedback loop should result in an inversion of its sonic characteristics. A mica capacitor, which has poor performance below 15kHz, should sound bright when used as a coupling element and harsh and wooly when used in a feedback loop., I have noticed this effect and have found I can largely cancel capacitor-induced colorations by proper manipulation of dielectric characteristics. Such diddling around with very subtle aspects of performance naturally leads to the subject of verification and its relationship to the audio amateur. While it is not the only, or even always the best one, the A-B test with one variable has become the mainstay of the audio establishment, probably because it naturally tends to force a decision: A wins and B loses. This test (the best of which is probably the 2 1/2 -blind variation propos ed by the Southeastern Michigan group) has won a lot of support from reviewers and salesmen. It has two important drawbacks for the experimenter however: First it implies that the amateur must maintain a standard device against which he measures changes in performance-a very expensive propositions and second and most seriously, it tends to obscure rather than magnify some differences if used improperly. The com parison is not a contest, but is rather more like a debate in which each side gains and loses points. Whether or not your modification is successful is less important than that you realize what you are trading off, and why. In fact, I usually learn more from a failed experiment than a successful one because my desire for success remains un satisfied and I am forced to push on. Form a clear i dea of what a modification has gained-and lost-and you will automatically know what your next step will be. I sometimes use a tape recorder, not to preserve the sound of a particular design, but to give me a perspective on the evolution of a design. Finally, I would like to make a brief', inadequate reply to the Sulzer/Arecibo group [4/79, p. 60]. Passive equalization involves a tradeoff of distor tion and signal-to-noise ratio in return for greater stability and dynamic range. Active equalization also suffers from the problem of supplying con siderable current to a low-impedance sum point-emitter of the first stage. Since the pro blem gets progressively worse with increasing fre quency, it plays hell with the slew rate. Transistor versions are also prone to emitter base diode nonlinearity ef'f'ects in the first stage and propagation-delay effects in the feedback. For these reasons, I would use passive designs with transistors and active designs with tubes. You may also wish to try a combination of passive high frequency rolloff before the first stage and active low-frequency compensation. This eases the slew-rate requirement immensely, but noise could be a problem. BOB MCINTYRE; Toledo, OH 43612 USA THIRD WORLD: TSS IN ADDITION TO THE "GOLDEN EARS, " and " meter readers" there is a third small, but growing group "The Scientific Subjectivists." I, and most of my friends belong to this group. Several months ago, I read with some pleasure the comments of Lipshitz, Vanderkooy and Young (TAA 3/79) but was skeptical of their claim to distinguish absolute phase in musical material. A group of friends and I (SMWTMS) invited them to prove to us in true scientific fashion using double blind AB comparisons that absolute phase could be heard. Being true scientists and gentlemen, they accepted. After an evening of listening to a variety of musical material under our double blind conditions (TAA 4/79, p. 56) the results of our guesses were analyzed statistically and found to be highly significant in favor of hearing phase. We are now firmly convinced. A "golden ear" may say "so what, I've known that all along," but therein lies the importance of controlled double blind testing-it can convince even the most determined "meter reader' that dill ferences exist. Other "golden ear" advocates should submit their claims to controlled statistical analysis as Lipshitz, Vanderkooy and Young have done. They have little to lose, and much in the way of credibility to gain. BERNHARD F. MULLER; Milan, MI 48160 A REES/SHAFFER EXPERIMENT A recent experience of an experiment which I was a part lends much credit to Mr. Rees and Mr. Shaffer (Auditory Perception: Part I TAA 4/79). 1 would like to explore their subject in the light of our experiment. The experiment was a double-blind to see whether we could consistently hear the difference between two preamplifiers. We assumed we would be able to do so, consistently and easily. We also assumed we could distinguish through listening average (not "high end" gear) components. Peter Moncrieff made the set-up at his house and invited John Curl, Karen Richardson, and myself to be part of his test. We all believed that we had each been able to hear subtle changes that some were reporting couldn't be heard when put to a double-blind test. Initially, no one of us could hear a difference. Had we really been fooling ourselves all these years? For me, it was very difficult to ignore my feelings, senses, and experiences. One problem for me was an unfamiliar room, equipment and loudspeaker system. My experience had been that the longer I was exposed to a system, the more I was able to detect about that system. After hours of listening, I was tired and frustrated. I began to question the assumptions we had made, and began to rearrange things to satisfy my own biases. I didn't like the diffuse sound field of a complicated splayed group of LS3/5A's. I couldn't hear the things I listen for. After I had a sound field more to my liking, I still couldn't hear any difference between preamplifiers. Then I suggested that the assumption that we could hear differences on average gear was perhaps wrong. There were things I had heard with a first rate power amplifier that were masked on records I was familiar with. The power amplifier was the weak link in the test set-up. So we looked over a dozen or so other amps and came up with a fairly good one as a substitute. I was then able to identify correctly 90 percent of the ti me one of the two particular preamps. So, the double-blind test can work but with reservations: The better the gear, the easier it is to hear differences. A long listening exposure to and familiarization with a system will allow differences to be heard more easily, And to be tuned-in to your own perceptual biases. Perhaps audiophiles and so-called Golden Ears optimize their systems to their particular biases which might further enhance their ability to detect sonic differences. RICHARD MARSH; Livermore, CA 94550 NEW SUPPLIER MANY BIG PARTS SUPPLY HOUSES have moved their minimum order up to $50 or $75 and often won't even attempt to order small quantities of specialized components. I have had considerable luck with an outfit in New Jersey by the name of Wilshire Electronics. They stock, or will order, the following manufacturers' devices: RCA, TI, Fairchild, Hewlett Packard, Siliconix, Delco, TRW, Xylog, Adv. Micro Devices. The person to contact is: Mr. Alex Allen, Wilshire Electronics, 1111 Paulison Ave, Clifton, NJ 07015 (201-340-1900). R. L. OMEGNA; Berkeley Heights, NJ 07922 FILTERS: DISTORTION AGGRAVATION TAA READERS WILL FIND immediate interest in Peter Billam's article, "Harmonic Distortion in a Class of Linear Active Filter Networks" (Journal of the Audio Engineering Society 26 June, 1978, p. 426). Briefly, Billam points out that since most active filters rely on positive feedback for their function, they multiply the inherent distortion characteristic of the amplifier being used, especially in the corner frequency region. He cites one example in which this "distortion aggravation factor" (his term) is as high as 81. The consequences for typical op amp filters could definitely be audible, especially at higher corner frequencies. I would like to learn more about this characteristic of active filters and its effect on common audio applications. Perhaps Walt Jung would like to comment on the article and relate it to his work in this area. MARK C. SHULTS; Madison, WI 53711 USA ---- Also see: Recordings Review, by editors Williamson, Gloeckler and Boak Test Report: Listening tests of the PAT-5/WJ-1A, by Laurence L. Greenhill, M.D. |
Prev. | Next |