Digital Domain (May 1984)

Home | Audio Magazine | Stereo Review magazine | Good Sound | Troubleshooting


Departments | Features | ADs | Equipment | Music/Recordings | History




TAKE A NUMBER


I was surprised when the editor telephoned me and said he wanted a second installment in this digital department. I had figured that my first number had pretty much iced the cake; my comments concerning analog recordings had been pretty abusive and seemed to be especially delinquent in a magazine which has so diligently followed the evolution of analog audio for so many years. When I picked up the phone, I figured I was being slapped with a lawsuit or something. But the editor said he was willing to go again. Although he didn't exactly sound enthusiastic, I suspected that, like so many others, he was willing to pay whatever dues were required to get to the bottom of this digital thing, before the fad was over.

Digital audio is foremost a numerical technology; to properly understand it, we must first lay some groundwork with a discussion of number systems. So let's talk numbers--and I don't mean things like S/N ratio, frequency response, and wow and flutter. As everyone knows, the specifications of digital audio products are incredibly good and thus incredibly boring: Flat, flat, none, none, etc. Besides, those specifications were mainly devised to gauge the quality of analog devices and perhaps don't even apply to digital ones.

When reviewers listen to two different Compact Disc players having identical specifications and yet hear a subtle difference, it's pretty clear that it's time to devise some new and more meaningful measures of quality. No, the numbers I have in mind connote an entirely different idea.

That's one of the most interesting things about numbers--the fact that a few numerals can represent so many different kinds of information. Let's take an example. I ride two motorcycles (rarely simultaneously); one of them is a classic BMW R50/2 built in 1962 with a 500-cc engine, the other is a Triumph 750 built in 1977 with a 750-cc engine.

The Triumph has license plate number 680314, while the BMW is licensed with 029907. Obviously, there are a lot of numbers here. Not so obvious is the important context of each of them. For example, 1977 is the year the Triumph was built, while 680,314 represents a special kind of coded information so that the State of Florida can properly monitor my moving violations. These various numbers are useful only by virtue of their arbitrarily assigned contexts. If that context is confused or missing, then the numbers quickly go awry; for example, I could have 680,314 motorcycles, 750 of them with engines with displacements of 1,962 cubic years. Similarly, the numerical operations we perform upon numbers are a question of interpretation; the sum of my moving violations determines when my license will be suspended, but the sum of my engine's displacements probably won't inconvenience me so much. Information, you see, and how it is coded, processed, and decoded, can make a lot of difference in our lives. Numbers, if properly defined, provide a good method to keep our data straight. As we've seen, the negative implication is that the numbers and their meanings have to be accurately used. Such is the price of civilization.

Speaking of civilization, ours is lucky because it has pretty good number systems. Early societies were stuck with the unitary system, which used a series of marks to answer the essential question "how many?" That's a great system, but only if the numbers are small, or your clay tablets are particularly large and slow drying. The Mesopotamians, who considered themselves to be a fairly advanced bunch, rejected the unitary system and devised a number system which used 60 symbols instead of just one. It was a little cumbersome, but even today, 3,700 years later, we still use their technique to divide an hour into 60 minutes, a minute into 60 seconds, and a circle into 360 degrees.

Frankly, I think a system with 60 symbols is a little pretentious, but the point is that any number system, in essence, is a question of how many different symbols you think you need.

Most of us feel partial to our base-10 system which we owe to some ninth century Arabs who rather cleverly conceived of the 0 numeral to represent nothing, and appended it to nine other symbols already in use. Our base-10 system uses positional notation: Each next place is multiplied by a higher power of 10; 10 is the radix of the system. Not coincidentally, a base-10 system is convenient for 10-fingered organisms such as humans (for really big numbers I take off my shoes), but any number base can be the favorite, depending on the application. Of course, you have to know the radix; the numerals "10" in base 10 is the total number of fingers we have, but "10" in base eight is the number of fingers minus the thumbs. Similarly, would you rather have 1,000 dollars in base seven, or 500 in base 60? Currently, radix 2 is hot, thanks to the digital computer. Its base-two number system is highly efficient; the idea of a voltage/no-voltage, on/off binary system is terrific for an electrical machine. Ask yourself-would you rather deal with 10 voltage levels, a hundred, an infinitely analog number of them--or two? A binary system is ruthlessly efficient-and fast. Imagine how quickly you can turn a switch on and off or watch a square wave go by (which means a machine is operating the switch for you). The result is a high-powered information handler which simultaneously processes information to yield yet more information, which in turn necessitates the need for still more information. It enables us to more dynamically direct information back into itself. A plugged-in civilization can never unplug; the information system breeds complexity and facilitates it.

Okay, this is an audio magazine, there are supposed to be discussions about recordings. What does a discussion about numbers have to do with recordings? Plenty. From your conceptual seat in your conceptual concert hall, you lean forward in anticipation as the conceptual orchestra begins its conceptual performance of the Tchaikovsky Seventh Symphony (no, I'm not conceptualizing, there really are seven). You job is to write down all the information you hear. Ready? Begin! Whew! Only after a few bars, you give up. Even writing down the score real-time in musical notation was too much, let alone documenting all the timbres, acoustics, aesthetic considerations, etc. We have to conclude that music is a surprisingly complex phenomenon; it is filled with information. To store it, we require a system which can deal with incredible amounts of information. Not surprisingly, the latest and highest technology, historically, was always utilized to make recordings, because only the best technology satisfied our current expectations of what a good recording should sound like. And as higher technology was devised and pressed into service, our expectations were redefined, often times much to the chagrin of the people with a lot of money tied up in the older technology.

Our conclusion is very clear. Changing technology has again changed our expectations of what good reproduction should sound like. Digital technology has redefined the science of information coding, processing, and decoding; thus, the art of sound reproduction must be redefined. All those who thought that direct-to-disc half speed recordings, or something along those lines, were the ultimate were right at the time, but now they are wrong. The sound will get even better.

All those who think that it can't get any better, that their ears simply won't detect any more fidelity, are wrong. It will sound better. And I can say this, too: It will sound drastically better because it will sound more real.

Also see: Philips Oversampling System for Compact Disc Decoding (April 1984)

(adapted from Audio magazine, May 1984; KEN POHLMANN )

= = = =

Prev. | Next

Top of Page    Home

Updated: Saturday, 2020-01-25 10:02 PST