Latest / Electronic Music / The Art of Physical Modelling - David Bessell

Transcript

The Art of Physical Modelling - David Bessell · Electronic Music

• Hello and welcome to the Sound On Sound podcast channel about electronic music and all things synth. I'm Rob Puricelli and in this episode I talk to David Bessel about his work in the realm of physical and acoustic modeling synthesis. For decades, synthesizer manufacturers and sound designers have strived to recreate the sounds of acoustic instruments Such as brass, woodwind, strings and more. And for many years they failed. Then in 1979, digital sampling was unleashed on an unsuspecting world and musicians were now able to sample and play back such instruments with ease. But it still wasn't perfect. Sampling recorded a single, often static performance, but lacked all the nuance of the real instruments. After decades of research in university labs around the world, physical modeling came to prominence in the 1980s and 1990s and offered a way to accurately mimic the detail and expressivity of acoustic instruments. Music David Bessel was inspired by this and sought to explore it further, something he still does to this day. I started off by asking him how he got into music in the first place and what prompted him to explore the world of physical and acoustic modelling. And, I mean, I started playing in bands like most people did, um, went through various bands in various styles with varying degrees of success. And in the process of that, I started working more and more in studios. So I ended up doing session work, um, you know, uh, in studios, mostly guitar stuff, um, but also programming. And then as I got more into the classical side of music theory and stuff, I started to do arranging things for other people, um, worked quite a lot with, um, uh, as an assistant, uh, programming and that kind of stuff with Ed Buller, who is, uh, also in Node. Um, so Node is an electronic band I play in from time to time, like about, let's see, 20 years? Um, but it has, uh, people in it like Noel Wesson, who works with Hans Zimmer, and, uh, Flood, who people will probably have come across, Depeche Mode, The Cave, and a whole Um, so in between doing all that, uh, I also went, uh, went and studied, uh, classical orchestration at the Royal College of Music, and ended up doing a PhD on the, um, writing for orchestras in a manner that imitates electronics. Um, and, uh, and then after that I, I went into a full time academic job running music courses at the University of Bedfordshire and, uh, Plymouth University. Uh, and then. In the past couple of years, I've bailed out of that academic world again, and, um, I've gone back to just making my own music for my own amusement, which I have the luxury of doing at this particular point in my career. What sparked your interest in physical modelling? What, where did, where did that all start for you? Uh, well, I've always been interested in synthesis, and understanding as much as I can about sound is obviously a way to, Becoming a more versatile synthesist, um, and, um, I guess, without really knowing it, I always used acoustic sound as a kind of model or ideal in my mind when programming even synthetic sounds. Um, I always tried to approach the, um, the emotional impact of an acoustic sound with an electronic sound. Um, And in the process of doing that, I explored a whole variety of different synthesis methods, granular and analog subtractive, and all sorts of things like that, and I came across physical modeling, which intrigued me because, firstly, it seemed very impenetrable to start with, um, you know, it works in a quite different way from an analog synthesizer, for example, or even a FM synthesizer, which are supposedly more difficult to manage. Um, and so, uh, when I went into the academic world, I had the time and the tools, um, to, to really in depth research into that. I taught myself how to do physical modeling and then started to push. The physical models which I was building myself, uh, into areas which were more kind of suited to my own musical needs, rather than off the shelf stuff. But when I first started it, there wasn't an awful lot of off the shelf stuff available. The Yamaha VL 1, which nobody could afford, and it was only about, you know, half a dozen made or whatever, you know. Uh, and so that was really the only decent commercial offering at that time. And so I was kind of forced to build my own stuff. And learn about it myself in order to be able to use it. How do you begin to create a physical model without, you know, going right into the weeds? Where does one start? The weeds are the interesting bit, but I don't think No, really they are, because actually I think probably a lot of people sort of know a little bit about physical modelling without even realising it. And So there are different approaches to physical modeling, but the most common one is based on extensions of the Karplus Strong algorithm, or Karplus Strong synthesis method. So that's something you can look up on YouTube. It's a very, very crude approximation of what happens if you pluck a string. And it's done by just, um, having a very, very short delay line and firing a little burst of noise into it, and then the delay line will ring if you have the feedback up enough, and decays in a way which is similar to a string. It's a very crude approximation, but that basic principle, a noise source, or what's called in the physical modeling world an excitation, fired into a delay line, uh, Usually, the first extension you would make to a Carpenter's Strong, um, a basic Carpenter's Strong setup is to put a filter in the feedback loop of the delay, and that makes the string decay more naturally. Otherwise, you get something a bit like a sitar. It's very twangy. But the filter will tame that a little bit. It also introduces other problems, and then you go down the rabbit hole into the weeds, as you said. But it's basically that. It's a noise source, um, uh, a short delay line with a lot of feedback, and a filter in that delay line. And that approach, you know, when you extend beyond carbon strong, that's called waveguide synthesis. And that's the easiest, most approachable style of physical modeling. And is that exclusively string type sounds because of the, the Karplus Strong principle? No, it's not, because it turns out in the physics of a, of a blown tube is very similar to a string. Um, and, um, and so if you, you can do, uh, woodwinds, you can do brass, uh, you can do plucked strings. Um, strings. And then if you go further into the weeds, you can do, um, metallic objects, wooden objects, um, and then further and further down the rabbit hole. And so that, that basic technique, if you extend it enough and understand enough about it, maybe add a second delay line, have cross feedback networks between the delay lines, multiple filters in multiple places, uh, you can get some very sophisticated stuff. You can model just about any physical object. What do you use to, to create your physical models? Because I understand that you do some stuff with analog equipment and then some stuff, uh, with digital equipment. So, so how do you, how does your process manifest? That's right. Well, when I first started doing it, it was around about the time when it became, computers became fast enough to be able to do it in real time. And prior to that, um, uh, computers could make a model, but you'd have to wait half an hour for the sound to come out the other end, you know, while it, while it churned away, crunching the numbers. Um, but around about the time I started playing with it, it was just probably, I don't know, early 80s or something like that. Um, computers got fast enough so you could do it in real time. So I started off doing it with computers. I'm not a terribly proficient coder, but I use MaxMSP as my main tool because it was approachable. As a musician, which I am basically, I'm not. I'm not particularly a coder. That was my main tool. I mean, it's not perhaps the best tool for doing it, but I found ways to do it. And then I tried porting some of those things to Reactor. Reactor handles some of the things which are more difficult to do in Max. Um, I wasn't terribly happy with the sound quality of either of those. Uh, and so eventually I, once I understood enough about physical modeling, I began to think, if it was possible to reproduce the software that I've written in analog hardware with a modular synth, if you had the right modules in the right configuration, you, you can see. You know, you can get delay lines, you can get filters, you can get noise sources. And I was thinking, well, it should be possible to do it. And so then more recently, I started building a Eurorack system specifically designed to do physical modeling. Uh, I know in Eurorack, you've got things like mutable elements, which are prebuilt physical bottles and they're They're digital models, you know, all contained within a, a module already built for you and you have access to some parameters there, but you don't have the flexibility if you build the entire model from scratch yourself. And so some of the things I wanted to do musically were not accommodated by, uh, Elements, for example, and so I, I started playing around with the analog modules and I managed to, I actually managed to get it to work somewhat to my own surprise. And I also came, what was very helpful in that process was I came across a guy called Ron Berry. He's hardly known at all, but he ought to be very well known because almost single handedly in the early 80s he built a modular synthesizer. himself, you know, from scratch with the soldering iron, um, to do physical modeling. And he's just about the only person who did it that way at that time. Um, and you can find his website. Um, I mean, he's retired now. I'm not quite sure how old he was, how old he is, but, um, I did manage to contact him and I picked his brains quite extensively. And, uh, he was very helpful, and, uh, because he's sort of been there and done it himself. Um, although, although he was using custom modules with the currently available Eurorack stuff, I was able to reproduce some of his ideas and, um, um, solve some of the practical problems that you inevitably run into, you know, something doesn't work the way it's supposed to in theory, you know. And so, so my most recent interest is trying to do it in the analog world. Because, to me, it sounds, it sounds more lively. You're using modular gear. What modules would one require? Is, are, are these the sort of basic modules that, you know, the standard kind of oscillator and a regular kind of filter? Yeah, I mean, in physical modeling, you don't use any oscillators. It's the feedback from the delay line which acts as the oscillator, or the equivalent of the oscillator. It's a bit like giving a spring reverb a kick. You know, you send a burst of noise into it, and if the feedback's up high enough but not quite feeding back on its own, it'll ring for a certain amount of time and then die away. And if you put a filter in the delay loop, The high frequencies die quicker than the low frequencies, so you get a twang at the start of the string and then it becomes more mellow as it dies away, you know, if you think about the way a piano decays. Well, I use different noise sources. I tend to use pink noise. I find white noise a little bit too bright and aggressive for what I want. And sometimes I use the noise source off my, uh, Macbeth modulo. Uh, so much of those sometimes I use a Eurorack module. I've got a couple of Eurorack modules which could generate noise. It's not entirely critical. what the noise source is. I mean a pink noise source is a pink noise source more or less. I do tend to find analog pink noise a little bit more sympathetic to the ears than digital noise, but there's not, you can use either. And then, uh, I would run that into a delay line, uh, which is that analog bucket brigade, uh, The one I use is CG 1022, made by Christian Gunther. It's a specialist module that specializes in very short delay lines. And it has an input that will allow you to control the delay line with volts per octave. So it's kind of tuned to respond in pitch. The volts per octave has to be inverted, scaled, and things like that to make it in tune. And so that's already built into the module, which saves you a lot of time. And then I Sometimes it's the bandpass filter or sometimes the lowpass filter on my Macbeth semi modular, but you can use any decent bandpass or any decent lowpass. Bandpass and lowpass have different effects. Uh, you can use them for different parts of the sound. So, for example, if I was trying to imitate whistle tones in a flute, which is the tones that you get from the very edge of the flute. blowing on the very edge of the embouchure, um, I would use a bandpass filter for that. But if I was trying to do, um, overblowing in a brass instrument, I'd use a bandpass filter for that. Um, but if I was trying to do, uh, a bowed violin or something like that, I'd use a lowpass. It has different effects, but that's, yeah, that's basically it. In the history of synthesizers, particularly in the early days, there was always this drive to make synthesizers and their patches sound like acoustic instruments. So in the early synthesizers there was always, you know, whenever they had presets they were, you know, Violin, string, oboe, trumpet, and they never sounded anything like it. And then here we have this technology, which is able to really accurately recreate the, the sonic complexity of an acoustic, uh, instrument. But is that really what. You want to use physical modeling for or are you looking to create something that has the essence of Something that's rooted in authenticity, but also maybe it's something that could not happen in the physical world for example mixing, you know mixing a string with a resonance chamber or vice versa or a reed with a you know, the resonance kind of chamber of a violin, maybe Yeah, exactly that. Um, so, um, when I'm building a model, the first thing I do is I try and recreate as accurately as I can a particular physical configuration like a flute or a trumpet or a, you know, whatever it is, uh, somebody bashing on a metal thing. Trashcan or whatever it happens to be, because that's a test of whether your model works, if you see what I mean. If you can get it to sound close to what it's supposed to be, you know you've got the, you know, the relationship between the modules you're using and the physics of the instrument that you're trying to reproduce right. So the first thing I do is I try and get it to be as naturalistic as possible. And the second thing I do. is to mess that up as much as I can, because, because what interests me most about physical models is that it has the expressiveness and some of the liveliness of an acoustic sound, but it offers the possibility of doing things which are not possible in the real world. So when I was studying orchestration, I did a lot of work with what we call extended techniques on instruments to playing instruments in wacky ways to get unusual sounds out of them, multifonic wind instruments, things like that. So I was always drawn to pushing instruments beyond their comfort zone. And that's exactly what physical modeling allows me to do. So I can take, there's an example on my sound cloud where I've got an instrument called a Clary Flupet. And what that is, is, um, if I remember all the bits correctly, it's, um, Uh, it's a clarinet mouthpiece with the body of a flute and the bell of a trumpet. Um, and so what I did was I had a trumpet model, a flute model, and a clarinet model, and I stole bits from each of them and kind of bolted them together in a Frankenstein model. Because there's something about physical modeling which tweaks your brain and it falls into the category in your brain which your brain labels as acoustic. And somehow there's something in sound which does that. Also, if you create an instrument which your brain doesn't recognize, then you get a kind of a cognitive dissonance where your brain says, this is acoustic, but it's not. And that's what, that's really what I'm after. Um, it's, it's sometimes referred to as the uncanny valley. You know, this is something that comes from robotics where you recognize something as being familiar, but there's also something uncomfortable about it. Um, so it's close to the familiar, but not quite. That aspect of, of, uh, strangeness, I guess you would call it, um, is, is an area I like to explore musically. You say you, you took the kind of the three elements of a clarinet, a flute, and a trumpet. And so you created, you know, these three models and then stole bits from each. When you say you stole bits from each, what are you actually doing? Are you still creating this, this model? On a single setup and that you're just adjusting some of the parameters, uh, on some of the elements within that, that set up to match those that you used on the flute, those that you used on the clarinet and those that you used on the trumpet. Uh, not quite. No, it's more like, it's more like a modular approach. So, so what you would do in the software, if I was trying to make a model of a, a clarinet, for example, I would model the reed behavior separately from the, from the body of the instrument. And then I would, I modeled the tone holes. And, uh, the bell shape, uh, and, uh, clarinet has, uh, is a sealed tube at one end, which does weird things to the harmonics. So, you know, there are various things that I would model separately, so, and then, and then I fit, take the output of the reed and feed it into the input of the tube, et cetera, et cetera. And so you can think of it as a modular approach to programming. So once I have that, then I've got all these components, which I can then rearrange, so I can steal the reed out of the clarinet. You know, copy and paste it into the, onto the flute. You know, body instead of the flute, I'm pretty sure, which I can ditch that bit. And that's how you combine the various things. One further extension to that is, you know, if you're trying to imitate a resonant body, like the body of a guitar, or the body of a violin, or a piano, or something like that, Uh, I, I, I sort of extended the physical modeling technique with, with, um, uh, with delay lines and I used convolution, um, to, to model the resonant body part because that seemed to be the easiest and most realistic way. That's a bit of a non-standard addition to a traditional physical model. But once I worked out how to do. convolution without any delay, without any latency, then, uh, then I started to use that as a method to do that. So if you, if you have a, uh, a convolution element to your models, um, you can take a different impulse and change the body. So you can have, you can have a carpenter's strong string and feed that into a resonator, which is you know, a sample or an impulse response of somebody kicking a trash can. And then you end up with something which is like a string stretched across the top of a trash can. And so one of my hybrid models is a pizzicato violin, um, and with string stretched across a gong, so you get a hybrid of a gong and so instead of having a violin body it has a gong as a body for the violin physiognomy, and that produces a very interesting sound. You said that, um, the, the, the type of physical modeling that you are practicing there, um, is, uh, correct me if I'm wrong, is it wave guide? Yeah, that's right. Yeah. So, um, There are other methods. What are they and do you use them? And if not, why not? Um, well, there's an extension of WaveGuide called Banded WaveGuide. Uh, which was, uh, that term was coined by a guy called Perry Cook. Uh, and I've used that quite a bit. Um, and that is a halfway house between WaveGuide. which is a classical physical modeling technique and modal synthesis or modal modeling, which is another approach to physical modeling. Um, uh, so I've dipped my toe in the water of modal modeling, but that requires a lot of, a lot of components to do it. And so it's only still really possible to do that in the computer. Um, So, I use that sometimes, my software uses that. Um, and then there is a third kind, which I call brute force, the brute force approach, which is really, um, finite difference equations. It's an entirely mathematical approach. It tries to describe the physics of the components of the instrument in terms of equations. And it feeds a value into the equation, and out the other end of the equation comes a value which represents the sound. Um, that, is, um, a serious white coat, um, um, scientist's job, and, uh, I don't usually bother with that. Firstly, because until fairly recently, it wasn't possible to do it in real time. The number crunching was just too much. But if people are interested in that aspect, you can get very good results with that. There's a guy up in Edinburgh called Stefan Bilbao. Uh, he's got some interesting stuff up on his website there. He's, uh, probably the expert in the UK on that kind of stuff. Um, but, but it's, it's slow and painstaking. It can produce very high quality results, but it's, you know, you need several lifetimes to, to really do more than a couple of modules of that. Physical modeling seemed to emerge certainly in the commercial world, um, in the early mid nineties, as you said before, with, with Yamaha and their VL and VP instruments, and then slightly later with Korg and their multi oscillator synthesis system, which manifested in the Korg Prophecy and the Z1 a little bit later on. And then all of a sudden, Physical modeling seemed to kind of drop off everyone's radar and that all went to analog modeling, modeling virtually the electronics of an analog synthesizer, which I think, you know, everybody jumped on is still very popular today. Why do you think physical modeling? Disappeared so kind of rapidly from the commercial market, and it's now only just starting to see a little resurgence in terms of there are, you know, there are some software instruments now that do some really excellent, uh, physical modeling of stringed instruments and percussive instruments, but it seems to have kind of gone away for a very long time. I just wondered if you had any thoughts on on why it didn't really gain any traction like all other methods of synthesis seem to do. I think there were two main reasons. Um, firstly, it was such an unusual way of doing things for musicians, even musicians who are familiar with analog synthesis and FM synthesis. It's, you know, you can probably remember yourself that the kind of shock of the new way when the DX7 came along and people just, I mean, it basically, I mean, I used to end up living for a while programming DX7s because I was the only person who'd read John Channing's paper on it. And, uh, yeah. And, and, and could understand how the thing worked. Everyone else was just reduced to pushing the buttons for the presets and twiddling knobs at random. But it had a sort of novelty factor which, you know, people liked at the time. But I think physical modeling suffered from that. It was, it was a completely alien way of doing things, you know, no oscillators. You know, even the DX7 had oscillators. How do you make a sound if you've got no oscillators? You know, from the conventional synthesis point of view. And also, you know, the information on how it works and things like that was buried in papers full of electrical engineering equations. So it was even more impenetrable than John Chowney's stuff on FM. Um, you know, for the average musician. One of the first people who really extended waveguide synthesis beyond Carver Strong was a guy called Julius Smith, and he's got enormous amounts of information up online, which is incredibly detailed, and it's got amazing stuff in it. But you need an electrical engineering degree because he basically doesn't use words, he only speaks in equations. I mean, almost literally. You'll get from three pages of equations and one paragraph explaining something about it, you know. So when, you know, when I first looked at that, I was lost, and it took me a long time to really understand that. And I think a lot of people just didn't have the patience or the interest. And then the second thing I think which stopped it catching on was the fact that Julius Smith then licensed all of that stuff to Yamaha. So Yamaha wrote pretty much own the rights to physical modeling entirely. Um, nobody else could use it because Yamaha had the rights. So they launched this incredibly expensive VR 1, um, which nobody could afford thinking that they got the next DX7. They overpriced it by a long way, and, uh, the unfamiliarity and the ridiculous price just put people off. And so they just sat on the patent for years and years and years. I mean, Corp tried to find a way around it. They did to a certain extent, but they can only go so far because otherwise they didn't finish Yamaha's patent. So it's only recently that that patent has expired. Um, and people are able to do it legally. Do you think sampling had anything to do with it? Because around the time that physical modeling started to kind of disappear from people's view, sampling was becoming much more proficient. We had more storage space, more processing power, um, and multi sampling of, you know, acoustic instruments was becoming infinitely more. complex with round robins and all that kind of stuff and we could fit that into a decent sized instrument a well priced instrument and so for authenticity instead of going for A physically modeled string or physically modeled woodwind we had samples Do you think that might have had something to do with it possibly for some people? Um, I mean personally i'm not looking to You know, primarily for realism from physical modeling. I'm looking for strangeness, but um, For some people I think perhaps it might have been marketed like that. Um, uh, but I think, you know Anybody who put a VL1 side by side with whatever Akai sampler was available at that time, in terms of expressive performance, the VL would, would absolutely wipe the floor with it. And, and, but in terms of realism, if you just press one note, the sample is more real. If you keep pressing that note, it immediately becomes less real, as you, as you well know. And so the solutions to that, like you say, round robin, multi, multi, velocity sampling and all that kind of stuff came a bit later when there was faster processes and more memory. So if anybody put the things side by side at the time, I think they would have realized the physical modeling offered in terms of performance, a completely different world. It's not just about the realism of the sound when you press one key, it's about the realism of the performance. And And physical modeling gets you much closer to the performance. A physical modeled instrument is not a single sound. It's a set of related behaviors which produce sounds that belong to a recognizable family. And a sample is not that. A sample is just a sound. That's the best description I've heard. And I've always kind of tried to figure what that, that description is. And you've just kind of nailed it there with that. Do you see now that the patents have expired and more and more people are exploring it? What do you see now as the future for physical modelling? Is it going to carry on being something that is done by the likes of yourself, or do you see it being adopted more widely in the near future? I think it's, it's, it's going off in two different directions. I think there are people like me who are interested in the experimental aspect of it and pushing it beyond just reproducing an instrument. And then there are some people who have been working away on one or two very specific models, trying to push the realism, you know, absolutely as far as it can get. So, you know, there are now a couple of very good, physical models that approach, you know, the limits of realism. So I think it's the SWAM instruments, which, which are really very, very good. Um, and, uh, Expressive E do some very good string, physical models, solo strings. Um, and, uh, there's the ModArt Piano Tech, which is a very good piano. You know, so these are people who've been working, you know, the ModArt people have been working on that piano model for like, 20 years or something like that, gradually refining it bit by bit using the extra computing power as it goes along. And so those people are, you know, uh, particularly the SWAM instruments have got to a level now where it's pretty much indistinguishable from the real thing, unless you really listen carefully. You know, the plain techniques are great, but from my point of view, they're not quite what I want because it's a commercial piece of software and the parameters that you are allowed to tweak are limited. They only give you certain parameters, which I think might be useful, and they limit the range of those parameters to realistic behaviors, you know, and actually I want to push it beyond where it starts to fall apart. Um, and so the thing with a physical model is actually, a physical model is quite an unstable thing, as, and the, acoustic instrument is, you know, the physical shape of a musical instrument has been refined over 300 years or whatever, um, to produce a nice clean musical tone. But you don't have to change the physical instrument much and it just makes a squeaky noise or whatever, you know. I mean, anyone who's taken a saxophone mouthpiece out and just blown through that, you know what I mean. So, uh, physical models are like that. So the people who make the commercial stuff, they don't like you to have too much control because it's easy to get into a situation where it doesn't make any sound at all, or it just croaks. Um, so they limit what you can do. And so some of the most interesting sounds right on the edge of Not doing anything at all. You can't get to. And then there are other people who have tried to make systems like Tasman and things like that which are sort of physical modeling playgrounds where you can mess around with modules which give you a bit more access to the outer limits, shall we say. So I think it's going in two different directions. Realism, prepackaged, and more expressive, more expressive, semi modular approaches for people who know a bit more about it. I mean, there's some nice models available from Reactor. You can learn a lot about modeling by taking apart other people's Reactor models. Sorry, I should just mention while I remember it. There's a Dutch guy called Han Visser. He's done about 30 physical models for Reactor. You can learn an awful lot about more advanced modeling by taking apart his Reactor models. It's model symbiotic, as you can see exactly what's connected to one, what is used. There's a lot to be said for reverse engineering of other people's work to find out exactly that. It's certainly easier to do that than it is to read Juvia Smith's electrical equations. The other thing you mentioned about, um, your love of, of physical modelled instruments is their expression. And therefore, The tools to allow you to convey that expression, uh, go beyond the traditional, you know, piano style Western keyboard and start going into, you know, ribbon controllers or wind controllers or, you know, other things. But do you use anything specific in that regard and do you have a particular favorite method or is it, um, how do you control your physical models? Um, mainly for the keyboard. Um, I mean, what I try and do is, there are lots of parameters to control in a physical model. Um, but quite often you can group those parameters together so that they, you know, when it gets louder, several different things increase or, or one thing increases and the other doesn't. Thing decreases at the same time or whatever, you know, you can group an awful lot. So if you've got a keyboard with aftertouch, velocity, a couple of, um, uh, modulation wheels, um, you can get a pretty expressive performance and, you know, because I'm working mostly with the hardware these days, um, you know, it's just some obscure parameters that I want to tweak for a particular expressive purpose at a particular point in a particular piece of music. I can always reach up and grab the control on the module, you know. So, so, um, between those things on the, on the keyboard and, um, you know, grabbing the old control on the module, I can do most of what I want. Um, but I do like the, uh, the look of the new Osmos keyboard from, from Expressive E, um, which has physical modeling capability because that gives you a bit more control from the keyboard. That was going to be my next question. The advent of the MPE format, you know, the extended version of MIDI and MIDI 2. 0. is giving musicians, electronic musicians, this, you know, hugely expanded resolution, which must really favor physical modeling and the things that you're doing. For certain things. Yeah. Um, it, it's surprising what you can do with what we've already got, you know, um, video and stuff like that. You know, it's people say it's low resolution and things like that. It is, but, um, you can do a lot, you can do a lot with it. And although You know, to suddenly expand the resolution of a control by a factor of a thousand or whatever MPE does. Um, it's nice, but it's not absolutely essential, um, to get, you know, a good expressive performance behind, uh, from it. I mean, uh, As far as physical models and expression are concerned, I mean, my viewpoint on that is that the reason why physical modeling is so expressive is that you have a sense of effort behind the sound. Um, so, so a lot of synthetic sounds, they're kind of abstract and you don't have a model in your mind for how that sound was made. Um, whereas if you hear a wind instrument, you know, somebody's blowing into that and you know, you know, When they're blowing hard into it, because the wind instrument does certain things if you blow hard and, you know, and the same with the string. If you hit it hard, it does certain things. It goes slightly out of tune. If you hit it harder and then it bounces back into tune and all that kind of stuff. So, the sense that there's an effort behind something is incredibly important for a sense of expression. And because of the way physical modeling works is that effort is provided by the volume or the brightness of the excitation source. And because that then pushes the system that it goes into, the reverberating delay line, then you get a sense of one thing pushing it either harder or less hard. And that gives a sense of effort. And that has a lot to do with expression, because it gives a sense of, You know, somebody had to push that that hard to get it to do that kind of thing. So if somebody wants to get into physical modeling now, say they've listened to this podcast, they think, right, I really want to do this. Um, they could go out and buy a range of, um, Eurorack modules if they wanted, or they could use, um, a piece of software. For example, you know, could you do Both of those things together, um, by employing, say, one of these, um, virtual, uh, modular units such as VCV rack, where you can pretty much pick up any kind of module, noise generator, filter, uh, delay line, and could, could you start experimenting in that realm just to kind of get the feel of it and see if it's something that, that works for you? Ah, you could. That would be a slightly perverse way of doing it though, because if you're going to do the modular approach, the real reason for doing that is to get the analog sound quality. Otherwise, you're just making it more difficult for yourself. So, if you're going to do it in software and you're a beginner, I would probably look at, um, somebody else's Carpenters strong patch in Reactor, because Reactor is a fairly easy patch. It takes care of some of the software housekeeping for you in, like, polyphony and things like that, which you don't need to worry about. So I would probably, if you were starting off and you wanted to do it in software, I'd probably start with Reactor. If you wanted to do it in hardware, there's a very good tutorial on YouTube by Nicole from Befaco, and she shows you how to set up a Carplus Strong. thing with a set of modules, and she shows you how to use filters with it as well. So that's the first step of getting beyond a kind of a twangy sort of metallic y string thing, which Carper Strong tends to produce. But, but it's, she explains it very well, and it's a good way in for beginners. If you don't know what Carper Strong is, and you want to know how to do that with modules, she even tells you which modules work and which ones don't. You know, which kind of delay lines are set up to do that, because you need particular kinds of inputs on the delay line, and it has to go very, very short. So she tells you which ones work, and that she's tried out and things like that. So that would be a good starting place to do it, with modules, and To actually just dip your toe in it, you only need about four or five bunches. So not, not an extensive investment, although I think the, the, the trend with Eurorack is once you've bought a few, you, you end up buying a few more and then a few more. David's been absolutely fascinating, uh, discussing this with you. It's a subject that's, um, I've always had a, Uh, an interesting, but never really a full understanding. And I think even in this short period of time, you've helped me understand that. And hopefully we've helped our listeners too, and maybe encouraged a few to dip their toe into the waters. David, thank you ever so much for your time today. It's been an absolute pleasure. Thank you for listening and be sure to check out the show notes page for this episode, where you will find further information along with web links. Details of all the other episodes before you go make sure you visit the sound on sound podcast page at sound on sound. com Podcasts where you can explore all the other great content playing across the other channels I'm Rob Puricelli, and this has been a failed muso production for sound on sound