Digital Domain -- Beautiful Or Beastly? (By Ken Pohlmann)

(adapted from Audio magazine, Sept. 1985)

For the past year or so we've examined many of the elementary aspects of digital audio technology. Discussions of topics such as sampling and quantizing, CD-player design, Compact Disc format and lowpass filters have, I hope, illustrated their workings and taken away some of the mystique often associated with them. Digital audio is, after all, merely a kind of technology used to store the information comprising music, just as analog audio is.

Of course, an analytical look at music technology goes only so far because, when it comes to the music business, a host of other factors ultimately becomes the overriding concern. Sales of LPs peaked (along with the 8-track tape) in 1978 at $2.473 billion, declining to $1.69 billion in 1983. Meanwhile, prerecorded cassette sales rose from $250 million in 1978 to $1.812 billion in 1983, thus displacing the LP as the leading audio medium.

What is the key to a medium's success? When it comes to market share, is fidelity the primary concern? Obviously there are other factors. For example, the cassette's quick rise to preeminence could probably be traced to the consumer's desire for convenience and portability.

Now, the cassette's rapid success is being humbled by that of the CD. In less than two years the CD achieved the same U.S. market penetration which took the cassette six years to reach. In the American market, CD sales rose from 800,000 units in 1983 to 4.9 million in 1984, an increase of more than 500%. Worldwide, 350,000 CD players and 5.5 million discs were sold in 1983, and in 1984 900,000 players and 17 million discs were sold.

Clearly, the CD is a big hit; it is, in fact, the most successful new electronic product ever introduced.

Why is that? Let's avoid a lot of endless debate such as, "It's because the economy is good," or, "It's in spite of the bad economy." And let's even avoid the obvious non-sonic elements of the format's popularity: Convenience, longevity, portability, etc. Instead, let's go to the mat with one criterion: The aesthetics of the medium.

Specifically, the question we can ask is, does the Compact Disc's fidelity

justify its success? Or is it merely the fast food of the audio world? The question is difficult to answer because of the dissimilarity of analog and digital media. Most performance specifications are throwbacks to the days of analog, and thus they measure parameters relevant to the quality of analog circuitry, but not necessarily to fidelity itself. Audiophiles have always questioned the omniscience of specs; with digital, that suspicion is even more suitable. For example, in terms of flat frequency response, signal-to-noise ratio, THD, etc., the CD is quite good.

In my experience with many players, I have come to expect a frequency response from 20 Hz to 20 kHz, ±0.5 dB; S/N greater than 95 dB, and THD of less than 0.003% at 0 dB and 1 kHz-or numbers in those ballparks. In fact, the specs for CDs are so good that I routinely employ CDs as a reference standard on the test bench for all kinds of measurements. Need a 1-kHz sine wave? How about a sweep tone, or tone bursts, or SMPTE IM? No problem--a CD test disc provides a highly accurate reference for all kinds of audio testing. But already we have diverged from our stated task of comparing aesthetics. Strictly according to specifications (at least those designed to evaluate analog equipment), the digital medium is clearly superior. But what do specifications mean in relation to what we hear? Beyond specifications remains the question of sound. It is apparent that a CD sounds different from an LP; this is evident even over broadcast FM. And the difference could stem from factors beyond those measured by our specifications. Exactly what constitutes it? Apparently there is something missing from, or added to, the sound of a CD (or both). But why is this "± Factor" judged good by some and bad by others? Is it music information, or distortion, that is missing? Has new music information been added, or a new kind of distortion? Or a mixture of both? With those kinds of questions, the battle lines are clearly drawn. A digital advocate would state that additional music information has been added and distortion taken away, and would point to the excellent specifications and sound quality to prove it. An analog advocate would state that music information has been lost and distortion added, and would similarly point to the sound quality to prove it. Leaving aside the question of specifications, we are left with differing opinions of the sound quality. It appears that in lieu of more meaningful specifications only a subjective evaluation can begin to settle the issue.

So, what is different about the sound quality when comparing digital to analog? A CD has several obvious advantages, such as lack of clicks and pops, groove noise, and wow and flutter.

Other, perhaps more debatable, advantages are solid low-end response and accurate localization. This brings us to the essence of the debate. Many analog advocates might grudgingly concede the above, but would single out high-end response and ambience as deficiencies in the CD. Although the Nyquist Theorem was proved in 1926, it is still not accepted by some. Despite healthy skepticism, the facts of the theorem haven't changed in the last 59 years. Simply stated, all of the information below half the sampling frequency can be encoded, and any information above that frequency must be excluded. The frequency response of a CD is thus flat from 0 to 20 kHz, less any deficiencies of the analog output circuitry interfacing the player with an analog amplifier.

If you like the frequencies between 100 and 400 Hz, then you have to like the frequencies between 10 and 20 kHz; they are absolutely identical during the digitization process. Of course, the Nyquist Theorem is as much of a limitation as it is an opportunity. Only frequencies below the half-sampling frequency are encoded. If you want a higher frequency, then you must increase the sampling frequency.

So what does all this mean? Let's consider the case of the highest note of the highest pitched instrument in an orchestra (no, it's not the cymbals), a piano. The C note four octaves above middle C is 4,186.01 Hz. Let's round it to 4 kHz for simplicity. Since it is a complex waveform, it will have an overtone series at multiples of the fundamental: 4, 8, 12, 16, 20, 24, 28, 32, 36 kHz, etc. Although each higher partial generally decreases in amplitude, in theory the overtone series would extend infinitely. Our digitization system would encode everything below the half-sampling frequency; in the case of the CD, up to about 20 kHz. Thus the fundamental and four partials would be encoded, and all musical information above that point (24 kHz and up) would be excluded.

The question thus ultimately boils down to hearing. If your hearing response extends beyond 20 kHz (and is acute enough to perceive the low-amplitude partials present there), then the CD has indeed robbed you of information-specifically, all information above 20 kHz. The 20-kHz high end was felt to be sufficient for most listeners; that, of course, explains why 44.1 kHz was selected as the sampling frequency.

How high is 20 kHz? Well, most televisions emit a loud oscillation of about 15 kHz which drives some people nutty. That's three-quarters of the way to 20 kHz.

What about ambient information? Analog advocates might say that a CD recording "has no air," in other words, that some reverberant or spatial information is missing or has been affected.

With 16-bit quantization the noise floor of a CD is about 98 dB down; thus the reverberation is not being lost in quantization noise. And it is not the length of the reverberation time, either; timings between analog and digital recordings show an equal period of decay. And why should ambient information be at all different from any other kind of acoustical information? The pressure function identifying the acoustical characteristics of a Steinway, and that identifying the acoustical characteristics of Orchestra Hall in Chicago, follow the same laws of physics. Once again, there is no analytical means yet available to demonstrate the subjectively felt differences between digital and analog sound.

Nevertheless, if a digital recording is compared to an analog recording, some listeners might describe the digital recording as being "cold" while others might describe it as being "clean"-two subjective expressions of striking descriptive similarity, yet one is a negative judgment, the other positive. Has the digital recording failed to capture some aspect of live sound, or merely eliminated some distortion which we have grown accustomed to in analog recordings? The weight of evidence favors the latter, though it must be conceded that perception of beauty lies between the ears of the beholder.

(adapted from Audio magazine, Sept. 1985)

= = = =