Why The Audio High End Is About to Climb
To The Top of the Minicomputer Junk Pile.

Francis Vale

A standalone, high end audio DAC (Digital Analog Converter) component is nothing more or less than a digital device to process data coming off your CD transport. I.e., it's a computer. Unfortunately, a DAC also happens to be a special purpose computer, or, if you prefer, a smart peripheral device.

The high end audio industry now has several dozen companies actively enaged in creating and building these monomaniacal machines. And like pungent mushrooms, more DAC vendors seem to pop up almost daily.

So what's wrong with this picture? For the clueless, let's put it this way: The whole high end digital audio industry is ridiculous. You, dear high end audio reader, have been known to blow as much as $15,000 (or more) for a DAC just to listen to the occasional CD. But for the same $15K, you can buy three Silicon Graphics UNIX workstations, network them together in parallel, and produce enough computer horsepower to create our own multimedia epic.

Ever wonder how Terminator2 and Jurassic Park got their awesome special effects? They were likely created on SGI machines. Want to create 3D, virtual reality, sound 'walkthroughs' that makes AC3/DTS/etc. look lost in space? Get an SGI. Want to edit, create, enhance, and store audio/video files that can be easily catalogued, cross referenced, and retrieved based on music/image content? Get an SGI. Want to have home televideo conferencing, with minimal cost connection to the rest of the world? Get an SGI with an Internet connection. Entry level SGI machines start at about $5k.

Even for the price of a 'low end' $2,000 DAC, you can buy a Pentium or a Mac PowerPC multimedia machine that approximates many of the SGI's video and audio processing capabilities. And now, you're going to say that your $15,000, single function DAC with its Mickey Mouse ten button remote control is a great value? Right.

Oh, we can already hear the swelling pundit chorus: DACs require unique, painstaking design, specialized components, custom chips, special power supplies, filters, manufacturing, blah, blah, blah...

The computer industry had the equivalent of its own DACs a while back. They were called mainframes and minicomputers.

Do you know why Digital and Wang went onto ER life support? It's because they were so wedded to their specialized multi-board computers and system software that they couldn't -- wouldn't -- see the all-on-a-chip microprocessor revolution bearing down on them.

One of the principal means that Wang made its now squandered fortune was through the sale of dedicated word processor machines. Then along came the general purpose, inexpensive PC with relatively cheap WP software, and splat! Goodbye Wang cash cow.

But today, you, the high end audio customer, routinely buy the equivalent of these extinct WP bovines when you purchase a specialized DAC, or other such unique digital audio processing component.

And the high end situation keeps getting more and more ridiculous. E.g., need a device to get rid of CD jitter? Well, just rush out and buy another specialized audio computer; which then requires yet another grundge-emitting power supply; and a new, maybe jitter producing, cable interconnect, etc. Would you buy a business or home computer that was as functionally constricted?

So, OK, smart guy, what's the answer? Easy! Moving the digital high end rapidly forward rests in two synergistic components:

1) Faster, cheaper, standardized microprocessors, and
2) better, smarter, software-based, audio processing algorithms.

These two technologies go together like bad coax interfaces and interconnect company profits.

In high performance computing, you always face a tradeoff: you can either build special purpose computer hardware, like the Seymour Cray's of the world used to do; or you can put the solution in software-only algorithms and use clusters of inexpensive, general purpose workstations.

Special purpose computer hardware has a fun way of hiding the basic stupidity of your software algorithms. For example: several years ago, there was a classic NASA study in which they carefully examined and rewrote some of their vector machine supercomputer software so it would run on new, multimillion dollar, massively parallel processing machines (MPP). Sure enough, the newly ported MPP codes executed much faster.

Then someone at NASA had the bright idea to put this now carefully reworked and looked over software back onto their vector machines to see what would happen. Lo and behold! Those dusty old codes suddenly got up and ran faster than a vinyl junkie after a flea market copy of Casino Royale!

The obvious moral: focusing on fancy hardware for performance can make a kludge out of your algorithms.

Another problem with custom chips/systems is that they can take a long product run to amortize -- Most especially so when you are a small, high end-type company. These, and other techno-go-slow effects, negatively impact the overall rate of innovation in digital audio. While the change rate may sometimes appear dizzying in the audio high end, it is a snail's pace when compared to the computer industry.

Moreover, if you think about it, most of the so called innovations in digital audio have to do with the bit medium format; e.g., DCC, DAT, minidisc, DVD, etc. But the core technology for processing the bits into audio analog waveforms has hardly changed since the inception of the CD format more than a dozen years ago. ("CD: Perfect Sound Forever" -- Who knew that meant also using the same technology forever!)

This is glacial progress. In the PC world, the rate of obsolescence is measured in calendar quarters, not years. PC users are also accustomed to great processing flexibility. You can readily buy inexpensive add-in cards to process digital sound in 8, 16, or 32 bit formats for ever higher resolution. Ditto for video. (When a DAC vendor offers plug in, plug out digital filters, it is hyped-up like the second cyber-coming.)

[This PC flexibility does have its downside, however. Remember Disney's Lion King CD ROM fiasco? Simba, et al, roared in 16 bits, forgetting that many people still had only 8 bit PC sound cards. The result: little Johnny or Mary's Xmas was ruined while their parents ground their teeth down in functional frustration. This risible scenario is now being played out once again. Millions of hapless owners of PCs with either NEC or Toshiba CD-ROM drives manufactured before August, 1995, are discovering that the much ballyhooed CD Plus format discs (which merge music with multimedia content) either crash their systems, or just can't access the data. No matter how much it tries, the entertainment industry just can't seem to figure out technologies which exceed the complexity of a toaster oven.]

So how come all the change seems to be in digital audio storage formats, and not in better audio signal processing? Easy answer: Manufacturing player widgets by the metric ton is what makes consumer electronics vendors money. Mucking around with new algorithms for better sound processing does not.

It is the classic hardware vs. software situation all over again. The fact is, Japan, Taiwan, Korea, etc., where the majority of these devices originate, still cannot hold a candle to the West when it comes to rapid software innovation.

So why, then, are we enslaving the DAC industry, and the hapless high end user, to endless arguments about digital storage formats? Just so a bunch of we-could-care-less-and really-don't-know-how-to-do-it-anyway manufacturers can make more money? Why isn't the focus instead on whaling away at new and innovative software-based algorithms for better audio?

If the consumer high end were to suddenly follow the PC industry example, we could have a standardized audio system box that:

1) utilized the computer industry standard PCI (Peripheral Component Interface) local bus;

2) into which bus could be easily added special function cards; and

3) which overall, used off the shelf shrink wrap software to enhance digital audio signals.

A DSP-based home 'Audio PC' makes the most sense in this new application. There are those among you who will now start screaming; "Using DSPs as general purpose devices is ridiculous! There aren't any good high level software development tools for them!" Wrongo. E.g., see the Impact article on DSPs in the Jan.1, '96 issue of 21st.

By doing everything via smart algorithms on powerful DSPs (maybe even running in parallel), you might, for example, entertain the notion of having 500kHz digital sampling when recording your masters; which might then be stepped down on the fly, via software-only selection, to the CD standard 44.1kHz. And why not 32, or even 64 bit audio data?

And jitter? Why must one do such technical handstands over this issue? If spurious rise/transition times are causing jitter down the digital food chain, why put all the jitter reduction in ever more boxes, and not in intelligent, self-adaptive, algorithms?

Still think this software-based approach to high end digital audio is nothing more than smoke and mirrors? Then look in the appendix below. It describes cheap 'shareware' for the Macintosh that does interesting things to audio. Please note the manifold digital sound formats supported by this software; as well as the ability of some of this shareware (e.g., SoundApp) to convert numerous types of digital sound file formats from one to another, including CD 'standard'. The other shareware product, "SoundEffects" clearly illustrates the power and ease of using modular software on a general purpose machine.

So, what do you think? Were the developers of all these computer audio formats humming "Perfect Sound Forever" when they created them?

In the next issue of 21st, we let the Impact section show the future of the audio high end. When things converge, it's either move, or die.

Copyright 1996, Francis Vale, All Rights Reserved

21st, The VXM Network,