Signals & Noise (Letters to Editor) (Jan. 1995)

"Man Bites Digital": The Final Chapters?

Dear Editor:

I am writing in regard to the apparently skewed negative responses (May 1994) to Keith Mackenzie's letter, "Man Bites Digital" (August 1993). The letters span the purely technical, mid-philosophical, and purely philosophical, so I assume these are representative of the total responses you received.

I used to be a dedicated digifanatic, and then was coaxed into the analog camp for a while--but after thorough research, I now find myself in the middle, waiting for something better. Here is why:

In digital audio, there are two corner stones that determine the quality with which the signal is recorded; the first is word length, and the second is sampling rate. The word length determines the S/N ratio, and the sampling rate determines how easily the D/A converter can recon struct the waveform. For Compact Discs, the word length is 16 bits, and the sampling rate is 44.1 kHz.

The reason so many audio buffs feel that the analog LP is superior to the CD is that the LP was developed to its maximum capability, which was really quite good (keeping in mind that the final sound emerging from our speakers is analog, whether the source is LP or CD). The CD, on the other hand, has the potential for superior sound, but the word length and sampling rate are at the lower end of the scale in terms of what is necessary for truly accurate sound reproduction.

Depending on how carefully the digital sound was originally recorded, the CD can end up with some lost bits, for a total word length of 14 bits. A noisy CD results, and I have heard them-even some CDs that have been released on top-brand labels, and supposedly DDD discs. Hiss as loud as, or louder than, that of an analog LP is present on these discs. Moving to a 20-bit word length will help this, for even with lost bits, the S/N ratio should remain at a satisfactory level.

According to the Nyquist theorem, the minimum sampling rate necessary for waveform reconstruction of sampled music is twice the highest frequency that is going to be sampled. The CD sampling rate of 44.1 kHz is just slightly more than twice the 20-kHz cutoff frequency in CDs. Thus, the minimum sampling frequency was chosen, and we now know that increased sampling rates improve the sound quality dramatically.

The bottom line is that current CDs are on the bottom line. Once digital sound is moved to a higher plateau, with 20- to 24-bit word lengths and higher sampling rates, there will be no further dispute as to the superior fidelity of CDs. But for now, there are two crowns worn in the audio world:

One is made of vinyl, and the other of polycarbonate.

-John E. Johnson, Jr., Ph.D. Redwood City, Cal.

Editor's Reply: Depending upon how one defines S/N ratio (more about which later), Dr. Johnson is correct when he states that "word length determines the S/N ratio" of a digital system, but he is incorrect in saying that "sampling rate determines how easily the D/A converter can reconstruct the waveform." Sampling rate has nothing to do with playback (signal reconstruction); it applies to the recording side of the equation.

Sampling rate sets an upper limit on the frequency of signals that can be handled by the sampler without "aliasing"-creation of "images" or "aliases" that appear in the sampled spectrum and cannot be distinguished from "true" signals at those frequencies. According to Nyquist, the theoretical limit is one-half the sampling rate, but to achieve that limit requires a non realizable brickwall filter-one that passes signals from d.c. to half the sampling rate perfectly and totally rejects signals above half the sampling rate. In practice, there fore, the frequency range that a sampler can accommodate is somewhat less than half its own sampling rate.

Note the following:

1. It is the sampling process that imposes the bandwidth limitation.

2. The sampling process is not "digital" (numeric), but an analog process that is a necessary precondition for digitization.

In other words, present-day A/D converters convert ("digitize") sequential samples of analog information.

3. It is the sampling rate of the sampler that limits the maximum frequency that can be accommodated at that point in the signal chain.

4. By sampling an analog signal at a multiple of the CD rate and using noise-shaped digital filtering, a substantial improvement in A/D linearity and dynamic range can be achieved. Modern A/D converters that use such techniques are far superior to the converters of earlier days. Thus, these modern converters are a major reason for the perceived improvement in CD sound quality despite the fact that there has been no change in the CD Standard itself.

Dr. Johnson's statements that "Depending on how carefully the digital sound was originally recorded, the CD can end up with some lost bits, for a total word length that "Moving to a 20-bit word length will help this, for even with lost bits, the S/N ratio should remain at a satisfactory level" are incorrect. Each digital word on a CD is 16 bits long-no more, no less. Whether each bit conveys useful in formation (determined, in part, by the characteristics of the A/D converter in the recorder) and whether all 16 bits are accurately converted from digital to analog (determined, in part, by the characteristics of the D/A converter in the player) may be in doubt, but CDs do not "lose bits" and end up with 14 rather than 16. Since the faults lie not in the CD (or the CD Standard) but in the limitations of the converters, in creasing word length from 16 to 20 bits will not help unless and until A/D and D/A converters with more than 20-bit capability are common-true 20-bit capability, not advertising hype. At that point, one might argue that CD word length is limiting dynamic range, but the practical significance of that is questionable given the back ground noise levels in which CDs are recorded and reproduced and the maxi mum sound pressure levels at which we can/should listen.

Despite the advertising, there's no such thing as a 20-bit CD. However, it is possible to achieve a perceived S/N ratio equivalent to that of a digital system with a word length in excess of 16 bits-from a 16-bit CD-by using oversampling and noise-shaping in the recording chain. These techniques do not reduce the total noise energy below that of a 16-bit system (in fact, the total noise energy is almost certain to in crease), but the spectral distribution of the noise can be modified so as to make it less audible; i.e., a greater percentage of the noise is shifted into regions in which the ear is less sensitive.

In summary, if Dr. Johnson has heard "Hiss as loud as, or louder than, that of an analog LP . . . present on these discs," the fault lies with the recording A/D converter.

(I'm assuming that all his CDs aren't hissy; if they are, he'd be well advised to buy a new player!) But here's the flipside of this coin. If Dr. Johnson has heard any CD even a single one-that sounds better than the finest LP he's heard, the case for the CD (and the CD Standard) is proved-for it's impossible for a CD to overcome the theoretical limits of the Standard, but it's mighty easy for a lot of CDs not to come close to reaching their potential.

These comments may help explain why some audiophiles are dissatisfied with the sound of some CDs and have come to the conclusion (as Dr. Johnson has) that "the word length and sampling rate are at the lower end of the scale in terms of what is necessary for truly accurate sound reproduction." That statement cannot be justified either in fact or in theory, and no prior analog music recording system for the consumer market has even approached the uniformity of response, absence of distortion, and wide dynamic range that are possible within the theoretical limits of a 16-bit, 44.1-kilosample/second digital system.

-E.J.F.

Dear Editor:

As a physicist who has yet to buy a CD player, let me say a few words on behalf of Keith Mackenzie and his letter, "Man Bites Digital." Some of us have clear preferences as to the type of equipment we have in our audio ensembles. For instance, I prefer amplifiers with MOS-FET power transistors to those with bipolar power transistors, and I believe there are good physical reasons for doing so.

The sound from vinyl LPs can be very good with appropriate playback equipment. The majority of consumers never had good analog disc playback, which is one of the reasons for the quick demise of the LP. Even many of the so-called high fidelity turntables and phonograph cartridges were never that in actuality. I know this from experience. The majority of equipment sold, as well as the majority of music heard (classical, jazz, rock, pop, country, jazz-rock, New Age, techno-pop, ad infinitum), can only be described as ...

Well, let me tell you an elucidating story.

Years ago, at a science fiction convention, an irate fan walked up to Theodore Sturgeon and said, "You know, 90% of science fiction is crap." Sturgeon, perhaps one of the great writers in that genre, replied simply: "Yes, 90% of everything is crap." I feel that the late Mr. Sturgeon's remark truly "containeth all."

-William Mendoza; Tallahassee, Fla.

Digital Film Sound: When? Where? How Much?

Dear Editor:

I'm writing in regard to E. Brad Meyer's article, "Digital Film Sound: Rated S for Sound" (June 1994). I enjoyed the article because I love when a movie has a fantastic soundtrack and can be played back to its full extent at the theater. Hopefully, this can be reproduced for home theater with the same quality and power. I am an owner of home theater, and this sounds like an other great advancement. But how long will I have to wait until I can get my hands on this technology, and will it be affordable? The systems from Digital Theater Systems, Dolby, and Sony seem fantastic, and with the Klipsch speakers, it makes me want to go see a movie in a theater equipped with this technology. I wonder, what powers those speakers, and how much power does their source put out? Also, how much money would any of these systems cost to install, and how many theaters are or will be equipped with these systems?

-Richard Konarski

Author's Reply:

Mr. Konarski seems to be asking about four things-namely, digital sources and HPS playback in both theaters and homes. First, the sources: The three digital encoding schemes are slugging it out, both publicly in theaters and behind the scenes in the home market. DTS and Dolby have shown LaserDisc prototypes carrying their digital signals, but in the home market one standard must prevail or all will die, and it's too soon to tell who's going to win.

The HPS-4000 house nearest to Mr. Konarski (the only one in Michigan) is the Bay Theater in Sutton's Bay, a couple of hours away. HPS theater systems use professional BGW power amplifiers; a big one has a total output of 5,800 electrical watts, which translates to around 700 acoustic watts-approximately the sound power of 10 symphony orchestras.

There is a home HPS system that uses Klipschorns as the main left and right channels, and it costs about $20,000. The theater systems, whose speakers use the same drivers and horns in larger cabinets (minus the expensive finish), range from around $12,000 to $30,000.

-E.B.M.

Praise for Carver Loudspeakers

Dear Editor:

I agree with Edward Tatnall Canby that bass has been overemphasized in the world of audio ("Audio ETC," September 1994). I have found a solution, a speaker that re produces sound flatly from 20 Hz to 20 kHz (almost). It isn't a single speaker; it is the Carver Amazing Loudspeaker Platinum, actually a "full-range" 60-inch ribbon and four 12-inch woofers. This speaker, when driven with the Carver TFM-75 amplifier or an equivalent, can accurately reproduce any music you throw at it.

The great thing is the bass. It isn't the boomy, muting type heard in home theaters and on lesser speakers, but the smooth, precise sound that only four long throw woofers can produce. Detail and accuracy are what sets this speaker apart from the average cone, and realism is what separates it from electrostatic and quasi-ribbon speakers. I also have a pair of Carver AL-IIIs with 48-inch ribbons; they are great, and I am currently using them in my Incidentally, maybe you can explain the effect I heard the other day when I was watching the LaserDisc of Star Trek VI: The Undiscovered Country. I forgot to engage Dolby Pro Logic; I didn't notice until I happened to look down at the receiver a few minutes into the film. I then tested several times to see why the LaserDisc sounded as good in stereo as it did in Dolby Pro Logic. The Carver image is so precise that there is little need for a center-channel speaker, and somehow even the surround information seemed to be coming from the right places! I can't explain it, but I assume this has to do with the bipolar design of the Carver AL-III and the lesser speakers I have in my setup (Boston Acoustics, not bad for book-shelfers). Another problem could be with the decoding of my Technics receiver.

Can you imagine a home theater with three Carver Platinums up front and the AL-IIIs as surrounds?

-Steve T Seitz; Portland, Ore.

Editor's Reply: Without hopping over to Portland and playing with your system, [...]

( Audio magazine, 01/1995)

= = = =