The THD Wars Are Over? Letters

Letters in Response

National Op-Amps
Editor: I was quite dismayed to read "The THD Wars Are Over?," John Atkinson's recent "News Desk" item on the Stereophile website In it, he asserted that "there is no doubt in my mind . . . that amplifiers (and preamplifiers) will sound better, the less spuriae [distortion] with which they contaminate the signals passing through them." Not only do you provide no evidence whatsoever for this assertion, but it flies directly in the face of 45 years of the subjective listening tests that Stereophile pioneered. To put it simply, to date there is absolutely no demonstrable correlation in audio amplifiers between measured total harmonic distortion (THD) and subjective audible performance.

The fact that an integrated circuit has achieved a (best case) THD+noise level of 0.0003% is impressive, no doubt. But what does this really mean in the real world? It merely means that the designers have applied such a high level of feedback that nonlinearities, as measured with a static test signal (ie, one that does not mimic actual music), are minimized, along with achieving a respectably low level of noise. (These parts are specified with an input noise level of 2.7 nanovolts per root hertz—which, by the way, is about 10dB worse than an Ayre or John Curl–designed Vendetta phono stage, for example, both of which use Toshiba's discrete low-noise JFETs.)

But does this low level of measured distortion actually translate to improved sound quality when listening to music? The answer is a resounding "No!" We can examine the fallacy in this presumption from two completely different perspectives and arrive at the same conclusion:

1) From a purely logical viewpoint, it is only a matter of time before somebody develops a circuit that has 0.00003% distortion, or 10 times better than these new ICs. Does this mean that this circuit will sound ten times better than these new ICs? Of course not. It simply means that it is possible to use higher and higher levels of feedback to achieve lower and lower levels of distortion as measured with steady-state test signals. And to date, nobody has made a defensible correlation between low measured distortion levels and "better sound."

For a good example of how extremely high levels of feedback can affect the sound quality of an audio circuit, we have to look no further than the review in Stereophile a few years ago of an ultra-low-distortion power amplifier. In that review, the reviewer noted that "dry, dull, or aggressive recordings sounded much more so" through these amplifiers. The very idea that such an amplifier could make a mediocre recording sound worse than it actually is contradicts the idea that low distortion leads to "sonic accuracy" (to say nothing of musical enjoyment).

2) As a circuit designer with decades of experience, I agree with JA's remarks in that same "News Desk" item that the individual parts used in a circuit are of paramount importance. But please consider this fact: There are not only hundreds of transistors and diodes inside that ultra-low-distortion IC, there are also hundreds of resistors and many capacitors to boot.

If anyone thinks for a moment that a resistor or capacitor fabricated from a dirty piece of sand (ie, the doped silicon used to fabricate ICs) is a better-sounding part than what is available in a discrete form, I can assure you that they are flat-out wrong. If you don't believe me, try to find any company that sells discrete parts made in this fashion. It simply isn't done, as the limitations imposed by these materials make it literally impossible to make high-performance parts in this way.

The bottom line is that there are inherent sonic limitations and compromises to any piece of analog electronics that is built from ICs. To think otherwise is just fooling oneself. Only a fully discrete circuit allows the designer to use the very best-sounding parts, adjust the values of each resistor for the best sonic results, dial in the correct bias currents, etc., etc.

It is to the discredit of the specialty audio magazines that they don't expose the "dirty little secret" of many so-called "high-end" audio companies. That is, these companies don't actually have circuit designers, they instead employ packagers who simply package the "latest, greatest" ICs into a fancy box with a thick faceplate. Is it any wonder that many of those companies are facing financial problems, or that there is a lack of enthusiasm from the buying public for so many "new, improved" products? (How many times in the last 10 years have we heard that high-performance audio is a shrinking market?)

When one is building a product to a low price point, there is probably no choice but to use ICs (or else build it in China). But when a "high-end" company builds a flagship preamp for $15,000 and the circuit uses nothing but ICs, it is simply a tacit admission that their "designers" know less about making a circuit sound good than do the engineers at the IC manufacturing companies. And while National Semiconductor is to be commended for setting up a dedicated listening room to evaluate their future products, please note that these new ICs were not designed on the basis of listening tests, nor were any other ICs from any other company.

To summarize, there are really only three reasons to choose an IC over a discrete circuit when building an analog audio product: a) cost, b) size, and c) power consumption. None of which has anything to do with high-performance audio and musical satisfaction, but everything to do with selling feature-laden goods to a consumer-oriented society.—Charles Hansen, Ayre Acoustics

Come on, Charles
Editor: I read Charles Hansen's letter in the October issue (p.208) with a combination of disbelief and amusement. I must take strong issue with most of what he said.

His comment about passive components of integrated circuits being made of "dirty sand" was a colorful play on words but terribly misleading. These ICs are made in clean rooms that are probably 1000 times cleaner than the factory where Charles makes his amplifiers. Moreover, all of the transistors he uses are made of the same "dirty sand." This is like saying his polypropylene capacitors are made of dirty carbon-based organic substances. Actually, glass capacitors (silicon dioxide or silicon nitride) are very, very good. The National op-amps that John Atkinson wrote about in the same issue (p.13) use only a single, very small compensation capacitor in each op-amp, not the "many" that Charles claims.

And what about the IC-based DACs that Charles uses in his fine CD player? These are built with the same "dirty sand." Would Charles argue that these are not part of the analog signal path? Perhaps Charles would like to make these out of discrete transistors.

There are numerous ways to make resistors and small capacitors on an IC, some of which produce outstanding performance. As someone who is not an IC designer, perhaps Charles is unaware of these options and the many flavors of secret sauce available to IC designers.

Perhaps more seriously, Charles took some Stereophile review comments out of context. In trying to justify his claim that negative feedback is bad, he said, "For a good example of how extremely high levels of feedback can affect the sound quality of an audio circuit, we have to look no further than the review in Stereophile a few years ago of an ultra-low-distortion power amplifier. In that review, the reviewer noted that 'dry, dull, or aggressive recordings sounded much more so' through these amplifiers. The very idea that such an amplifier could make a mediocre recording sound worse than it actually is contradicts the idea that low distortion leads to 'sonic accuracy' (to say nothing of musical enjoyment)."

Imagine my shock to discover that he was referring to Paul Bolin's review of the Halcro dm58. Talk about taking something out of context! That was a great review of the dm58. If anything, the review was a testament to the very high sonic quality achievable through the proper use of negative feedback.

I was also sorry to see Charles attack audiophile designers who use ICs as mere "packagers." This is certainly below the belt, as there are many fine analog high-end audio designers who have made well-considered decisions to use ICs in certain places in their products. I tip my hat to any designer who can achieve 95% of the sonic quality of a Halcro or an Ayre at one quarter the cost. I don't care what kind of parts he uses to do it.

Finally, we have an IC company that is taking seriously the needs of the high-end audio community. Rather than attacking them, the high-end community should be applauding them—and listening to them (pun intended).

There are many remaining mysteries to audio, and that is why we must listen, and that is precisely why Charlie is wrong in asserting that something can't sound good for thus and such reasons. In doing so, he is guilty of the same sin as those who would assert that an amplifier cannot sound bad if it has low THD.—Bob Cordell, Holmdel, NJ

ARTICLE CONTENTS

X