A very frank article about digital audio.
Article By Barry Blesser
From Volume 34 No. 3, published October 2012 Boston Audio Society
[This article was sent to me by AES-DC member Jim Mastracco. It was originally published in the September 1984 db magazine (which no longer exists). While some of the details reflect the state of digital technology at that time, the concepts and thoughts from this digital pioneer are still relevant in and beyond digital audio. Reprinted with permission from the author. DJW]
There comes a time in every activity when we must ask why we bother. Each month for the last four years I have sat down at my typewriter to select an interesting topic on digital audio. And each month, many of you spend a few minutes to read what I have written. Perhaps it might be interesting to examine what I have achieved over these years. Have I been successful? Since I have not received any direct communication from readers, I can only speculate that I have been either very successful or a total failure. I would like to take a few minutes to write about what I had hoped to achieve, and we can then evaluate my performance.
Although I have had extensive design experience in digital audio, these articles in db are not intended to make the reader a design engineer. Although digital audio has its roots in higher mathematics, the db articles are not intended to provide a formal education. The articles are, however, intended to give the reader a feel and intuition for the subject. Having good intuition about the subject is not the same as having formal training. Curiously enough, there often are people with formal training who have no feel for their subject, and people without formal training who do have such a feel.
How does one get a feel for a difficult subject? Next time you are at an AES convention, take an old-timer for a free beer and ask him to talk about how he learned audio in the good (bad) old days. The stories always have a common idea: play with it and experience it through your senses. He will invariably talk in terms of senses: touch, sight, hearing, and smell (when something overloaded). The grooves of a record are visible under a microscope; the filaments of the tube glowed red hot; the vibrations of the loudspeaker cone could be felt by the touch of the hand; the harmonic distortion could be clearly heard when grossly overdriving the amplifier. Even the sophisticated equipment had a physical and sensory component: you could see and touch the reverberation plate, you could see and touch the microphone ribbon, and you could see and touch the contacts of the fader. Wires could be visualized as pipes containing the flowing music. Even the magnetic fields on recorded tape could be made visible with the aid of magnetic fluid.
Another element of the old-time stories is that the audio people never worked at their profession; they were always having fun playing. Learning and doing was always fun. They would say "What happens if I...?" Then they would go do it to see what happens. When I was a kid, I learned about the differences in vacuum tubes by breaking them open and taking them apart. I had the largest collection of grids and plates on my block.
Then came the revolution!
The transistor was so small that it no longer had a physical-sensory reality — just the mystical belief that there was something in the little blob at the ends of the wire. Have you ever seen an emitter? Maybe it exists. Maybe it doesn't. Nevertheless, engineers could gain an under- standing by playing with it, although it was more difficult and more complex than the vacuum tube. This was only the beginning of the revolution. By the 1970s, manufacturers were placing hundreds of transistors into a single blob. This actually made life easier because the new blocks could be described in much simpler terms, e.g.: an amplifier. The ease of use brought us to the last stage in the revolution. A designer could make a system with thousands of these new blocks. Then the final systems could contain extremely complex functions.
Why is the large system the last stage in the revolution? Because a human being, regardless of his IQ, could not keep the entire system in his head. It was too large. The size is what changed everything. If you have tried playing with a home computer, you might have gotten the sense that it is easy. The first simple program [for which you write the code and it works] gives a great sense of satisfaction. Now consider the problem of understanding a program made up of 50,000 instructions. One of the reverberation machines I designed has the equivalent of 30,000 actions for each sample of audio. Who can understand the behavior of such a system by playing with it? What is the sensory basis for sitting at a computer console editing in all of the commands?
If you open up a defective piece of equipment with the expectation of repairing it, where do you begin to look when you see 500 ICs, some of which have 48 pins. And each of these pins might have 300 different time-shared functions because each IC contains 40,000 transistors. The fact that the input and output wires might be related to the functionality of reverberation does not change the fact that there is no location in the machine that contains "reverberation." Not only can't you repair it, but the manufacturer also has trouble. Many of them replace defective equipment and throw away the returned unit. One computer manufacturer throws away $50,000 backplanes, which contain nothing more than connectors and wires, when they have not been able to repair it within a week.
Your anxiety in dealing with modern technology is shared by those who are the "experts." This is a hard business because the basis for the technology is not physical-sensory; it is intellectual and inferential.
1. "It's broken because it sounds bad when I play organ music. The music sounded better two years ago, if my memory is correct."
2. "With low frequency input signals, there is an additional wideband noise modulation which rises with increased input level. It gets better after the equipment warms up and it only appears on the left channel."
In neither case is the quotation from a sophisticated design engineer who has detailed knowledge of the circuitry. But the second quotation is obviously from somebody who has the concept of noise modulation in his vocabulary. He has the feel and intuition to be able to relate his sensory auditory experience to such a phenomenon as tape modulation noise.
Digital audio has its own language in order to represent new phenomena. Unfortunately, I cannot teach the phenomena by demonstration since db magazine is made out of paper. If I could show you what granulation noise was by listening, I would. I can, however, give you a feel for the origin by explaining the mechanism that creates it. This has been my goal. Learning new ideas and concepts is not expected to be easy. For many of you the digital audio series is more than a little difficult. How- ever, it is not important that you understand everything, but that you get a sense for the way in which one can look at an issue.
The notion of language was made very graphic to me a few months ago. I had the opportunity to work with a lawyer on a patent case. In the process, I began to get a feel for the language of patent law. The lawyers have three words for "proving." As a simple scientist I had only one word for proof. A lawyer has the concept of "preponderance of proof", which means that there is more evidence for than against; he has the concept of "proof beyond a reasonable doubt", which means that there is no evidence to the contrary; and he has the concept of "strongly proved", which is in between. When a lawyer says that he has proved something, he means that for the particular legal issue he has satisfied the required proof level. Legally true does not mean scientifically true. After getting a feel for the legal language, I could much more easily communicate with him; the experience did not make me a lawyer. Reading my articles will not make you a digital audio design engineer.
The process of teaching the language of digital audio is made more difficult by the fact that the profession is immature. New concepts and issues appear; old concepts are better understood and refined. In the year 1995, it will be much easier to understand digital audio. Not only will each of us be 11 years older, but the profession also will be. I can re- member the first time I was presented with a "problem" in a digital audio delay line. There were five us looking at a "distorted" sinewave that appeared very strange. None of us could figure out what created such a strange-looking problem. A year later, I understood that I had seen what we now call image-frequencies at the output D/A caused by inadequate anti-image output filtering. We have the language and that allows us to communicate and to evaluate the issue. It also means that we can fix and redesign it with ease.
In other professions, as in audio, we spend a great deal of effort making the language very sophisticated. Often this sophistication is mathematical, but ordinary words can be used with the mathematics. The math is much less ambiguous but often lacks a physical feeling. Nevertheless, the mathematical language is so powerful that large systems are de- signed on the computer. In a recent project, all of the design work was done by using a mathematical model of the process on a computer. Over $100,000 was spent just on the computer simulation without ever going into the laboratory. When the answers appeared, the designers began to play with the model to get a feel for the issue. "What happens if I change the...? What happens if I increase the X parameter...?"
When all of the results were achieved, the specifications for the equipment were sent to the shop laboratory for fabrication. The laboratory model was then a test of the computer simulation. If it checked out OK, that meant that the model was OK. Otherwise, the new phenomenon was investigated in order to fix the model. The laboratory was not the place for invention and design; it was the final examination. If you pass, you go to the next step; if you fail you go back to the start again. We thus see that the last stage of the revolution has reduced engineering to that of information manipulation. Computers are pure information. When we make an accurate model of material, we have reduced it to information. High technology is abstract because information has no physical form.
It is only in the last few years that audio engineers are beginning to realize that part of the profession is being subject to this part of the revolution. Before you argue with me, let me say that the artistic part of the profession will never be subject to pure information because it is an emotional sensory experience. Only the technical part of the profession is changing. The distance between the users and the designers is widening and a common language is required to communicate across the gap.
• Dynamic range
To illustrate the complexity of our language. I would like to take two of these terms and define them clearly.
• The first example will be dynamic range. This is a badly used term that is usually used incorrectly. Dynamic range is defined as the range of signal level from the largest to the smallest. It is expressed in a ratio of levels such as dB. Dynamic range has nothing to do with noise! Dynamic range refers to the range of usable or interesting signals. Some- times, the smallest usable signal is covered by noise. A signal can have much less dynamic range than the tape recorder. Popular music might only have 10dB of range. That does not mean that the noise level is 10dB under the peak signal. We could record the signal on a tape recorder with a signal-to-noise ratio of 90dB. At playback we would still have a signal dynamic range of 10dB. When using our language carefully, we would say that we cannot record a larger dynamic range than the signal-to-noise of the tape recorder. One term refers to the signal, the other to the recording equipment.
• The difference in concepts becomes more obvious when we describe a signal such as a flute tone. We can record such a tone and play it back successfully, even when it is several dB under the noise level. The usable dynamic range of the tape recorder is greater than its signal-to-noise ratio. Signal-to-noise ratio is related to the maximum peak signal and the wideband noise in the absence of signal. The fact that one can hear a signal that is smaller than the noise level de- scribes a property of the human auditory system, not of the signal or the tape recorder.
The above discussion is a demonstration of the power of language when used carefully. I now ask you to evaluate the success of the db series on digital audio in terms of your learning the new language. Your comments are welcome!
Enjoy the Music.com highly encourages our readers to join the Boston Audio Society by clicking here).
This article is copyrighted © by the author or the Boston Audio Society. It is posted on Enjoy the Music.com with their permission, and with all rights reserved.