[obsidian] vault backup: 2025-01-20 19:03:04[

content/Max Mathews Full Interview by NAMM Oral History Program.md

@@ -80,7 +80,7 @@ Now where did you grow up?
 
 Max V. Mathews 01:24
 
-I grew up in Nebraska, and when I was 17, I guess I enlisted in the Navy as a radio technician trainee. Now we were called radio technicians, but we were really trained to repair radars, but the word radar was secret at that time. And so I finished school there and then went to Seattle and helped commission a destroyer, and then we shot the guns and shook the boat down and went back to Seattle, and then I was mustered out because the war had ended and VJ Day was over.I met Marjorie in San Francisco at the radar training school on Treasure Island, and we hit it off immediately. So I stayed in the West Coast, went to school at Caltech, studied electrical engineering there because I was in love with radar circuits. I wish I had studied physics there, but nevertheless it's a wonderful school.And then I went on to MIT and got introduced to computers. Those days analog computers were the useful computers, digital computers were still being developed, and I sort of loved these big complicated systems, and so we solved the kinds of problems that analog computers could solve, and that was my schooling. 
+I grew up in Nebraska, and when I was 17, I guess I enlisted in the Navy as a radio technician trainee. Now we were called radio technicians, but we were really trained to repair radars, but the word radar was secret at that time. And so I finished school there and then went to Seattle and helped commission a destroyer, and then we shot the guns and shook the boat down and went back to Seattle, and then I was mustered out because the war had ended and VJ Day was over. I met Marjorie in San Francisco at the radar training school on Treasure Island, and we hit it off immediately. So I stayed in the West Coast, went to school at Caltech, studied electrical engineering there because I was in love with radar circuits. I wish I had studied physics there, but nevertheless it's a wonderful school.And then I went on to MIT and got introduced to computers. Those days analog computers were the useful computers, digital computers were still being developed, and I sort of loved these big complicated systems, and so we solved the kinds of problems that analog computers could solve, and that was my schooling. 
 
 Interviewer 03:03
 
@@ -88,7 +88,7 @@ Very interesting. Can you give me a little background on your family? Did your p
 
 Max V. Mathews 03:11
 
-Yes, my parents were both born there and grew up there. They were both teachers.My father was the principal of the teachers' training high school in Peru. There was a little teachers' college there. But what he really enjoyed was teaching the sciences. So he taught physics and biology and chemistry. And he let me play in his laboratory as well as in his workshops. And that was another thing that set the course of my life. I still enjoy working in a workshop and I still enjoy the sciences very much. 
+Yes, my parents were both born there and grew up there. They were both teachers. My father was the principal of the teachers' training high school in Peru. There was a little teachers' college there. But what he really enjoyed was teaching the sciences. So he taught physics and biology and chemistry. And he let me play in his laboratory as well as in his workshops. And that was another thing that set the course of my life. I still enjoy working in a workshop and I still enjoy the sciences very much. 
 
 Interviewer 04:00
 
@@ -96,11 +96,11 @@ Very interesting. Well, what were the computers like when you first started gett
 
 Max V. Mathews 04:10
 
-Well, the one computer that we used most, and this was to develop counter missiles to protect mostly against air attacks at that time. And this was a combination of an electromechanical system. So the integrator on the computer was a mechanical integrator, but the other components, the adders and more simple operations were done electronically. Then operational amplifiers were designed and came along at that time. And so then most of the simple integrations were taken over by the operational amplifier feedback circuit that still does that job. And only complex integrations of fairly nonlinear processes had to be done with the mechanical components.So the computer itself filled a large room full of relay racks that held both the analog components and the mechanical components. Now, there were a lot of interconnecting that had to be done at a patch field. The question would be, had you done it correctly, would it give the right solution to the original problem? And so we needed check solutions, and you could integrate the solution on a Marchant mechanical multiplying calculator machine. If you had a group of five or ten, I think in those days it was entirely women, and they worked for about a month to calculate one solution, whereas the analog computer, of course, would turn out a solution in a few seconds. So we would get these digital integrations and compare it with the analog result, and then figure out what mistakes we'd made and corrected, and then go ahead and study a lot of different conditions.When I came to Bell Labs in 1955, I started working and always worked in acoustic research there, and our main job was developing new telephone, well, new speech coders that really would compress the amount of channel that was needed to transmit the speech over expensive things like the transatlantic cable. And in the beginning, people had a number of ideas on how the encoding might work. Pitch period repeating was one of them. Channel vocoder processing was another of them. Format vocoders was yet a third, and in order to try these things, one had to build a sample model of them, and this was very complicated. The vacuum tubes were the things that we had to design and work with in those days. The transistor had not yet become practical. So it might take several years to design a trial equipment, and usually it didn't work.So then you would go back and do it again. And I thought that, well, I should say that this was just the time that computers were becoming powerful enough to do a digital simulation of many things. And in the case of speech, the essential thing was a way of getting speech into the computer and then getting it back out after you had processed it to see what it sounded like. And the key element that made that possible was not the computer, digital computer itself. You could run the computer for a few days to make a few minutes of speech. But the crucial thing was the digital tape recorder, which could take the output of an analog to digital converter at speech rates. 
+Well, the one computer that we used most, and this was to develop counter missiles to protect mostly against air attacks at that time. And this was a combination of an electromechanical system. So the integrator on the computer was a mechanical integrator, but the other components, the adders and more simple operations were done electronically. Then operational amplifiers were designed and came along at that time. And so then most of the simple integrations were taken over by the operational amplifier feedback circuit that still does that job. And only complex integrations of fairly nonlinear processes had to be done with the mechanical components.So the computer itself filled a large room full of relay racks that held both the analog components and the mechanical components. Now, there were a lot of interconnecting that had to be done at a patch field. The question would be, had you done it correctly, would it give the right solution to the original problem? And so we needed check solutions, and you could integrate the solution on a Marchant mechanical multiplying calculator machine. If you had a group of five or ten, I think in those days it was entirely women, and they worked for about a month to calculate one solution, whereas the analog computer, of course, would turn out a solution in a few seconds. So we would get these digital integrations and compare it with the analog result, and then figure out what mistakes we'd made and corrected, and then go ahead and study a lot of different conditions.When I came to Bell Labs in 1955, I started working and always worked in acoustic research there, and our main job was developing new telephone, well, new speech coders that really would compress the amount of channel that was needed to transmit the speech over expensive things like the transatlantic cable. And in the beginning, people had a number of ideas on how the encoding might work. Pitch period repeating was one of them. Channel vocoder processing was another of them. Format vocoders was yet a third, and in order to try these things, one had to build a sample model of them, and this was very complicated. The vacuum tubes were the things that we had to design and work with in those days. The transistor had not yet become practical. So it might take several years to design a trial equipment, and usually it didn't work. So then you would go back and do it again. And I thought that, well, I should say that this was just the time that computers were becoming powerful enough to do a digital simulation of many things. And in the case of speech, the essential thing was a way of getting speech into the computer and then getting it back out after you had processed it to see what it sounded like. And the key element that made that possible was not the computer, digital computer itself. You could run the computer for a few days to make a few minutes of speech. But the crucial thing was the digital tape recorder, which could take the output of an analog to digital converter at speech rates. 
 
 Max V. Mathews 09:00
 
-In those days, it was 10,000 samples per second. Today it's 44,000 samples a second for CD music and more for other things.Anyhow, take these rapid flow of samples coming out and record them on a digital tape that then could be taken to the computer to be the input, slow input. And the computer would write a digital tape and you could take this back and play it back again at the 10,000 samples per second so you could hear the thing at speech frequencies. And this digital tape-based A to D computer input and output was the equipment that we built at Bell Labs that made this possible and was a completely successful device for speech research. And most of the modern coders came from this. And now, of course, as you know, it's not only digitized speech is not only used for research, it's the way that almost all information is transmitted. The reason being that digital transmissions are very rugged and number is a number and you can hand it on from one medium to another and from one company to another. And as long as you use the proper error correcting codes why if it goes to Mars and back you'll still get the correct numbers. So that's how the world works today.
+In those days, it was 10,000 samples per second. Today it's 44,000 samples a second for CD music and more for other things. Anyhow, take these rapid flow of samples coming out and record them on a digital tape that then could be taken to the computer to be the input, slow input. And the computer would write a digital tape and you could take this back and play it back again at the 10,000 samples per second so you could hear the thing at speech frequencies. And this digital tape-based A to D computer input and output was the equipment that we built at Bell Labs that made this possible and was a completely successful device for speech research. And most of the modern coders came from this. And now, of course, as you know, it's not only digitized speech is not only used for research, it's the way that almost all information is transmitted. The reason being that digital transmissions are very rugged and number is a number and you can hand it on from one medium to another and from one company to another. And as long as you use the proper error correcting codes why if it goes to Mars and back you'll still get the correct numbers. So that's how the world works today.
 
 Interviewer 10:38
 
@@ -108,14 +108,12 @@ Very interesting. Max, when did it first come into your mind that computers and
 
 Max V. Mathews 10:48
 
-I've forgotten the exact date, but it was in 1957, and my boss, or really my boss's boss, John Pierce, the famous engineer who invented satellite communication, and I were going to a concert. We both liked music as an art. And the concert was at local pianist who played some compositions by Schnabel and by Schoenberg. And at the intermission, we thought about these, and we thought that Schoenberg was very nice and that Schnabel was very bad, and John said to me, Max, I bet the computer could do better than this, and why don't you either take a little time off from writing programs for speech compression or maybe work in the midnight oil and make a music program. And as I said at the beginning, I love to play the violin, but I'm just not very good at it, and so I was delighted at the prospect of making an instrument that would be easier to play, at least in a mechanical sense, and I thought the computer would be that. So I went off and wrote my Music 1 program, which actually made sound, but horrible sound, so that you couldn't really claim it was music. But that led to Music 2 and eventually Music 5, which did make good music. And gradually, I'm not a musician, well, in any sense. I consider myself a creator and an inventor of new musical instruments, computer-based instruments. But my ideas did make an impact on musicians and composers and I think started, or it was one of the startings of the fields of computer music. 
-
+I've forgotten the exact date, but it was in 1957, and my boss, or really my boss's boss, John Pierce, the famous engineer who invented satellite communication, and I were going to a concert. We both liked music as an art. And the concert was at local pianist who played some compositions by Schnabel and by Schoenberg. And at the intermission, we thought about these, and we thought that Schoenberg was very nice and that Schnabel was very bad, and John said to me, "Max, I bet the computer could do better than this", and "why don't you either take a little time off from writing programs for speech compression or maybe work in the midnight oil and make a music program". And as I said at the beginning, I love to play the violin, but I'm just not very good at it, and so I was delighted at the prospect of making an instrument that would be easier to play, at least in a mechanical sense, and I thought the computer would be that. So I went off and wrote my Music 1 program, which actually made sound, but horrible sound, so that you couldn't really claim it was music. But that led to Music 2 and eventually Music 5, which did make good music. And gradually, I'm not a musician, well, in any sense. I consider myself a creator and an inventor of new musical instruments, computer-based instruments. But my ideas did make an impact on musicians and composers and I think started, or it was one of the startings of the fields of computer music. 
 
 Interviewer 13:05
 
 Absolutely. Tell me about music too. I'm sort of curious about that. 
 
-
 Max V. Mathews 13:10
 
 Well, Music 1 had only one voice and only one wave shape, a triangular wave, an equal slope up and equal slope down. And the reason was that the fastest computer at the time, the [[IBM 704]], was still very slow. And the only thing it could do a tall fast was addition. And if you think about it, each sample could be computed from the last sample by simply adding a number to it. So the time was one addition per sample. Well, the only thing the composer had at his disposal was the steepness of the slope, how big the number was. So that would determine how loud the waveform was, and the pitch that you were going to make, and the duration of the note. And so that wasn't very much, and you didn't have any polyphony there.So they asked for making a program that could have more voices. And I made one with four voices. And I made one where you could have a controlled wave shape so that you could get different timbers as much as the wave shape contributes to the timbre. Now, in a computer, calculating a sine wave, or a damp sine wave, or a complicated wave is pretty slow, especially in those days. So I invented the wavetable oscillator where you would calculate one pitch period of the wave and store it in the computer memory, and then read this out at various pitches so that this then could be done basically by looking up one location in the computer memory, which is fast. And I also put a amplitude control on the thing by multiplying the wave shape by number. So this cost a multiplication and a couple of additions. So it was more expensive. By that time, computers had gotten maybe 10 or 100 times as fast as the first computer. So it really was practical. So that was music too. And some thing that most listeners would call music came out of that. And some professional composers used it. But they always wanted more. In particular, they didn't have any things like a controlled attack and decay, or vibrato, or filtering, or noise, for that matter. So it was a perfectly reasonable request.But I was unwilling to contemplate even adding these kind of code, one device at a time, to my music program. So what I consider my really important contribution, that still is important, came in music three. And this was what I call a block diagram compiler. And so I would make a block, which was this waveform oscillator. And it would have two inputs. One was the amplitude of the output. And the other was the frequency of the output. And it would have one output. And I would make a mixer block, which could add two things together and mix them. And I made a multiplier block in case you wanted to do simple ring modulation. And I made a noise generator. And essentially, I made a toolkit of these blocks that I gave to the musician, the composer. And he could interconnect them in any way he wanted to make as complex a sound as he wanted. And this was also a note-based system so that you would tell the computer to play a note. 
@@ -130,7 +128,7 @@ Hmm, very interesting.
 
 Max V. Mathews 20:06
 
-But to finish up this series, that got me to Music 3. Along came the best computer that IBM ever produced, the IBM 704, 1794, excuse me. It was a transistorized computer, it was much faster, and it had quite a long life. They finally stopped supporting it in the mid-1960s, I guess.I had to write Music 4, simply reprogramming all the stuff I had done for the previous computer, for this new computer, which was a big and not very interesting job. So, when the 1794 was retired, and I had to consider another computer, I rewrote Music 5, which is essentially just a rewrite of Music 3 or Music 4, but in a compiler language. FORTRAN was the compiler that was powerful and existed in those days. And so that when the next generation beyond the Music 5 computers, the PDP-10 was a good example of a computer that ran well with music, I didn't have to rewrite anything. I could simply recompile the FORTRAN program, and that's true today. Now the sort of most direct descendant of Music 5 is a program written by [[Barry Vercoe]], who's at the Media Lab at MIT, and it's called Csound, and the reason the C in [[CSound]] stands for the C compiler. Now you're asking about Bell Labs, and many wonderful things came out of Bell Labs, including Unix, and of course Linux, and now the OSX operating system for Macintoshes all started at Bell Labs. And the most powerful compiler, and I think the most widely used compiler, was also created at Bell Labs. It was called the C compiler, A and B were its predecessors, and C was so good that people stopped there, and now that's it for the world. Every computer has to have a C compiler now, whether it's a big computer or a little tiny DSP chip. So that's where that came from. 
+But to finish up this series, that got me to Music 3. Along came the best computer that IBM ever produced, the IBM 704, 1794, excuse me. It was a transistorized computer, it was much faster, and it had quite a long life. They finally stopped supporting it in the mid-1960s, I guess.I had to write Music 4, simply reprogramming all the stuff I had done for the previous computer, for this new computer, which was a big and not very interesting job. So, when the 1794 was retired, and I had to consider another computer, I rewrote Music 5, which is essentially just a rewrite of Music 3 or Music 4, but in a compiler language. FORTRAN was the compiler that was powerful and existed in those days. And so that when the next generation beyond the Music 5 computers, the PDP-10 was a good example of a computer that ran well with music, I didn't have to rewrite anything. I could simply recompile the FORTRAN program, and that's true today. Now the sort of most direct descendant of Music 5 is a program written by [[Barry Vercoe]], who's at the Media Lab at MIT, and it's called Csound, and the reason the C in [[CSound]] stands for the C compiler. Now you're asking about Bell Labs, and many wonderful things came out of Bell Labs, including Unix, and of course Linux, and now the OSX operating system for Macintosh all started at Bell Labs. And the most powerful compiler, and I think the most widely used compiler, was also created at Bell Labs. It was called the C compiler, A and B were its predecessors, and C was so good that people stopped there, and now that's it for the world. Every computer has to have a C compiler now, whether it's a big computer or a little tiny DSP chip. So that's where that came from. 
 
 Interviewer 23:03
 
@@ -160,8 +158,7 @@ I can give you information about the most important players there, which were th
 
 Max V. Mathews 32:10
 
-When I left in 1987, there were still about 1,000 people in the research department. The rest of the Bell Labs had about 30,000 people, so he insisted that everyone use their resources wisely and not try to grow. This lasted until the Consent Decree in about 1980, which broke up the Bell System into seven operating areas, separate companies, and a company called AT&T, which would contain the Bell Labs, the research part, and also the Western Electric, which was the manufacturing arm that would provide telephone equipment to the operating companies, as it always had. But it opened the whole thing to competition, and also by that time digital transmission was coming in. In contrast to analog transmission of sound, which is very fragile, and if you want to send a conversation from San Francisco to New York or to Paris by analog, that means you really have to send it over carefully controlled analog equipment that really means all the equipment needs to be run by one company. But when digital things came along, then you could pass the digits on from between many, many companies in many, many ways. So essentially, the Telephone Research Lab no longer had the support that it did with this controlled monopoly, and so it was no longer possible really to support this group. It's expensive even to run a thousand people. The budget was something like $200 million a year. So that's my view of research in the part of Bell Labs.It was a wonderful time. It was a time when there was, of course, in the Second World War and afterwards, a strong military research group at Bell Labs and development group and things like the Nike anti-aircraft missile were developed there and many other things. Underwater sound was also another branch of the military research. I think the military research actually still goes on. Bell Labs eventually split up and became Lucent, which is the name you probably know it by. And now it's amalgamated with the French company Alcatel, so it's Alcatel-Lucent. And it's no longer limited to working in the field of communications as the original AT&T was. As a monopoly, it could not work in any field. It was allowed to work in the movie field, though, and developed sound techniques for movie film in the 1920s. 
-
+When I left in 1987, there were still about 1,000 people in the research department. The rest of the Bell Labs had about 30,000 people, so he insisted that everyone use their resources wisely and not try to grow. This lasted until the Consent Decree in about 1980, which broke up the Bell System into seven operating areas, separate companies, and a company called AT&T, which would contain the Bell Labs, the research part, and also the Western Electric, which was the manufacturing arm that would provide telephone equipment to the operating companies, as it always had. But it opened the whole thing to competition, and also by that time digital transmission was coming in. In contrast to analog transmission of sound, which is very fragile, and if you want to send a conversation from San Francisco to New York or to Paris by analog, that means you really have to send it over carefully controlled analog equipment that really means all the equipment needs to be run by one company. But when digital things came along, then you could pass the digits on from between many, many companies in many, many ways. So essentially, the Telephone Research Lab no longer had the support that it did with this controlled monopoly, and so it was no longer possible really to support this group. It's expensive even to run a thousand people. The budget was something like $200 million a year. So that's my view of research in the part of Bell Labs. It was a wonderful time. It was a time when there was, of course, in the Second World War and afterwards, a strong military research group at Bell Labs and development group and things like the Nike anti-aircraft missile were developed there and many other things. Underwater sound was also another branch of the military research. I think the military research actually still goes on. Bell Labs eventually split up and became Lucent, which is the name you probably know it by. And now it's amalgamated with the French company Alcatel, so it's Alcatel-Lucent. And it's no longer limited to working in the field of communications as the original AT&T was. As a monopoly, it could not work in any field. It was allowed to work in the movie field, though, and developed sound techniques for movie film in the 1920s. 
 
 Interviewer 36:26
 
@@ -203,7 +200,7 @@ Yeah, absolutely. What role, if any, did music concrete play in the evolution of
 
 Max V. Mathews 42:16
 
-Um... Oh, music concrete started before all this came along, and the technology used was the tape recorder technology, and changing the speed of tapes and making tape loops, which play something repetitiously, and being able to splice snippets of tape with various sounds on them, so you could make a composition, for example, by splicing the tapes of various pitches, and that was a very successful and a very tedious operation, and one of the things that I tried to do was to make the computer do the tedious part of it, which it does very well, and make the composer think more about the expressive part. Now people argue a lot about music concrete, and what was Stockhausen's alternate thing where he generated all sounds, not by recording real sources, but by using oscillators, I think.I've forgotten the name for that, but anyway, that now, I think, is an absolutely meaningless argument, because digitized sound is so universal that the sources of the sound can either come from nature, from recordings of instruments, sampled things, or they can be synthesized, and you can use FM techniques, or additive synthesis, or a myriad of other ways of making your sound. So I don't really think it's worth hashing over this very old conflict, and I guess Pierre Schaffer is died a number of years ago. 
+Um... Oh, music concrete started before all this came along, and the technology used was the tape recorder technology, and changing the speed of tapes and making tape loops, which play something repetitiously, and being able to splice snippets of tape with various sounds on them, so you could make a composition, for example, by splicing the tapes of various pitches, and that was a very successful and a very tedious operation, and one of the things that I tried to do was to make the computer do the tedious part of it, which it does very well, and make the composer think more about the expressive part. Now people argue a lot about music concrete, and what was Stockhausen's alternate thing where he generated all sounds, not by recording real sources, but by using oscillators, I think. I've forgotten the name for that, but anyway, that now, I think, is an absolutely meaningless argument, because digitized sound is so universal that the sources of the sound can either come from nature, from recordings of instruments, sampled things, or they can be synthesized, and you can use FM techniques, or additive synthesis, or a myriad of other ways of making your sound. So I don't really think it's worth hashing over this very old conflict, and I guess [[Pierre Schaffer]] is died a number of years ago. 
 
 Interviewer 44:50
 

content/Pierre Schaffer.md

@@ -0,0 +1,4 @@
+---
+date: 2025-01-20 18:08
+---
+#stub

content/private

@@ -1 +1 @@
-Subproject commit e3feb6466de376b073445bcf93121b3fce241505
+Subproject commit dd4a5c8f7bbc91ce88ed96db806af3e16af38ae0