In the 90s physical modeling seemed like it was going to be a big thing, but then it just kind of went away. There was the Yamaha VL, the Korg Z1 MOSS stuff, and the AAS Tassman soft synth was around in an early form. What is there now that can make the full variety of sounds? It seems like there are a few plugins that specialize in one type of model each ere doesn't seem to be anything that does it all. Also the only thing that seems to be doing any kind of brass/horn models is arturia brass.

I think the reason why physical modeling may have died off is that people where wanting realistic sound out of it, which it didn't really seem to acheive. That is EXACTLY why I want it actually. I wan't it to sound sort of like a saxophone, sort of like plucked and struck stuff. My problem with what I've heard from Arturia Brass, is that it sounds pretty darn lifelike. I think when something sounds close like that it's worse than sounding abit farther away from it. When it's close listeners just say oh that's fake, vs that's an interesting synth sound that's kind of like a saxophone.

That Kaivo synth looks interesting, but only does 2 types of models. The current AAS stuff just does one type of model each it seems. I guess I'm looking for something that's where the Korg Z1 left off but in software, or maybe just an old trinity/triton w mossor a prophecy if no all in one modeling thing ever surpassed them.

There is Resonans Rack Extension for Reason:

You can load and record your own samples in it I think.

Not sure if it does what want.

https://youtu.be/rl9F6I0vI6c

Structure in Logic Pro X sounds amazing and has a lot of parameters which can be mapped for expressive playing.

AAS Chromophone 2 does multiple models, is intuitive to use and user friendly and allows you to couple 2 resonators together as well as adjust the complexity of the models. The other AAS plug-ins are specific to the modeled instrument but are still very good and get very sophisticated.

Yeah, it seemed like such promising tech, but then everyone said, “ooo, ANALOG!” and that was that. You see bits of it here and there, like in the AAS stuff (which seems largely abandoned) and Arturia stuff, but nothing that seems to be pushing things forward. I really liked what could be done in Tassman, but I started having issues with some of the PM modules and their response was, “yeah. It’s broken.” That was years ago.

I think that when every old analog is finally modeled to death we’ll see people starting to mine the past for ideas. I don’t think that will happen until we get a proper CS80 and Prophet VS emulation. Then we might see movement.

How about the STR-1 engine on the Kronos (and Oasys)?

As Korg describes it:

Based on the physical modeling know-how which KORG has developed over the course of many years, this struck/plucked-string physical modeling sound engine takes advantage of cutting-edge technology. The STR-1 begins with highly playable sounds whose tone responds to your touch in ways that are difficult for a PCM sound engine to replicate. These include acoustic guitar, electric guitar, harpsichord, clavinet, harps, bells, and ethnic instruments. Then the STR-1 goes beyond, allowing the creation of sounds that obey physical laws, but that don’t exist in reality. Play harmonics on a metal bar “plucked” by a piano or sing into a guitar string; with the STR-1, it’s all possible. You can also process samples and live audio input through the physical model—including audio feedback from any of the KRONOS X effects!

I know very little about the tech, so this may just be bald marketing hype, but there it is. It *is* a fun engine to play with...

There's something about the sound of physical modeling that's really pleasing to my ears. May be alone in this, but I wish it would have taken off instead of virtual analog, which I never have liked that much. It's long overdue for a comeback. Seems like there is an abundance of resources for developers out there... probably just a matter of time.

In my opinion, physical modeling took a hit from the following:

- decreasing cost of RAM and hard drive storage, which made it possible to store larger multisamples for more realistic emulation of acoustic instruments
- the invention of "Gigasampling" (my name for when I first saw the technology). Keeping the first short section of all relevant samples in RAM, and then streaming off of the hard drive once a sample started, allowed for sample sets to get far bigger than the amount of RAM in a system
- the playability of the physical model was often directly proportional to the playability of the acoustic instrument itself. A piano physical model is easy to play - just hit a key. A violin physical model, played by a novice, will sound like a weird electronic version of an acoustic violin that is also played by a novice. Lots of screeching.
- lots of patents around physical modeling, while other synthesis techniques were more open during the late 90s.
- analog synthesis came back into style

I worked for Staccato Systems at the turn of the millennium, which was a spinoff of some of the physical modeling work at CCRMA. At one point, we reviewed several hundred patents from Yamaha, that were related to physical modeling. There was a LOT more IP ownership and territoriality around physical modeling than there was around the early days of analog synthesis modeling.

At this point, a lot of the physical modeling patents are probably expired, or are close to expiring. This might open up the path for younger developers to try their hand at making cool sounds with physical modeling, without worrying about patent infringement lawsuits.

That’s really cool you worked at an off shoot of CCRMA. Did you go to school there, or UCSD, or another of the big computer music schools?

Are you able to elaborate on how the physical modelling from Korg and Yamaha relate to Sondius? Some things I've read seem to be conflicting on this topic.

It seems that Yamaha and Stanford did some joint development at some time, though both organisations have also conducted independent development.

It seems that Korg may have been convinced to sign a Sondius licensing agreement, on the basis of potential patent infringement, though didn't initially enter into a collaborative arrangement. Is that accurate?

IIRC, Sondius is mentioned in reference to both the VL and Oasys PCI platforms.

edit - just saw the book The Sound of Innovation: Stanford and the Computer Music Revolution by Andrew J. Nelson, which seems to cover the topic, and the historical context.

That's not the case. They release soundbanks and updates for their VSTs fairly regularly (granted, mainly bugfixes and stuff like VST3 support intead of new features), plus they have a few newer things, like working with Intellijel on the Plonk module, Objeq Delay, and the Objeq iOS app.

Well, I think they’ve come right out and said that Tassman is dead. I’m pretty sure I read that. No bug fixes there. It seems to me that they’re only making some cosmetic changes on their other plugins. Their updates have been super disappointing. I seem to remember that the last one was just a UI change with some effects tacked on. It’s like they lost their original developer and they just don’t have the talent to do true upgrades, or even make the plugins they have all work correctly. Objeq looks like a fun toy, but I can’t help but think that it’s an attempt to repackage their old tech to grab some iOS app cash.

Anyway, I hope that I’m wrong. I think AAS has some good products. I think I own them all... I just think of other developers like U-He that continually improve their software and also develop new products. AAS seems like a guy who tinkers with plugins from time to time on the weekends.

CCRMA and Yamaha were close for a while. I don't know the exact specifics. I remember seeing a whole bunch of new Yamaha everythings at CCRMA as an undergraduate in the late 80s/early 90s.

I'm not sure. I wasn't that high up in the hierarchy of Staccato Systems, so I wasn't party to the actual IP rights discussions. From looking at Korg's patents of the time, they certainly took things in a different direction than the CCRMA stuff.

I was going to recommend that book. VERY detailed overview of that time.

I was an undergraduate at Stanford, and took courses at CCRMA, but I was never officially in one of the CCRMA-focused majors. At the time, I would have either had to go EE (not smart / focused enough for that major) or Music (not enough classical music background to pull that off). My senior year, they introduced a CCRMA-focused undergraduate major that didn't require classical music or a graduate degree in EE, but I was pretty much done with my Anthropology degree at that point.

I went back in the late 90s and took a year-long computer music course at the University of Washington. From then on, most of my education happened on the job. Working with a bunch of CCRMA PhDs at Staccato Systems, and later at Analog Devices (who acquired Staccato Systems in 2001) taught me a ton. I also took a few workshops at CCRMA in the mid-2000s, which taught me a LOT about plugins.

I was always more interested in the electronic/experimental aspects of physical modelling. Emulative accuracy was pretty low down on the list. A twenty foot violin body made of a crystal tube with strings strummed by a slinky was more in my wheelhouse.

The Yamaha VL line, the Korg Prophecy/Z1, original Wavedrum, were all capable of delivering really interesting sound design. But the emulative aspects were overtaken by competing tech. Didn't know about the patent minefield. Hope your speculation about the possibilities of a resurgence of interest pans out.

My only experience with physical modeling is with Collision and Tension in Ableton Live but the reality doesn't live up to the promise imo. I think there are a number of issues with physical modelling, such as:

- A large number of parameters, many of which aren't always intuitive to use or predictable in their results.

- Small sweet spots of usable sounds that can quickly get out of control with a small adjustment of a parameter.

- Unpredictable noises and glitches within an otherwise nice sound (may be bugs or may be something inherent to the process)

And then when you find a sweet spot you can get some very realistic emulations of an acoustic instrument, but they're still not as realistic as a sample. You can get other sounds that are close to realistic but pushed more toward a synthetic sound, and these often to my ears sound like FM patches. Like something that's almost an electric piano, bell or plucked string, but not quite realistic. Just sounds like a DX7 patch imo.

The dream is, like @ Arglebargle says, something like "A twenty foot violin body made of a crystal tube with strings strummed by a slinky." But I find that once you get into those extreme settings you find a somewhat samey palette of harsh noise through comb filters. Maybe it's the dream that's the problem because realistically, a 20' crystal violin might just sound like a really thin violin with bad sustain.

I dunno, I prefer Pianoteq to sampled pianos. I think it does deliver.

I haven't heard those before, but they do sound good. To me that reinforces the idea that the technology is best left to experts who can carefully tune the parameters to make realistic presets. Kind of like FM.

I've owned and used many PM synths from the 90's, I still have a Korg Prophecy and Korg's Oasys PCI card in an old PC that still gets used from time to time.

While I do miss my Yamaha VL synths, I have found IMHO a much beter solution in software:

1) Audio Modeling's VSTis
2) Sample Modelings's VSTis
3) Wallander Instruments WIVI

I use all these with a breath controller (TEC BBC2) and they are mind blowing! (No pun intended)...

I also use Pianoteq's various instruments and they too are simply amazing.

To illustrate this, I'll leave you with one of my short pieces using these products:

LINK: YouTube

If Yamaha were to release a new reface VL model, I think I'd be interested though...

I think you nailed it. The "harsh noise" often had some interesting chaotic qualities, but maybe not as interesting or as musical as you can get with various feedback patches on, say, an MS20 Mini.

The coolest sounding physical non-models I ever heard were done by Tim Stilson. Lots of phsyical-ish model systems, but with some sort of compressor/limiter in the feedback loop to manage the feedback levels. Tim was really into control systems, and would come up with some really interesting musical sounds that were based around control system theory.

Sean, seeing as you've sifted through a vast amount on the subject, could you possibly point us to some of the more interesting resources to learn more? Who knows, maybe some potential developers might be lurking or run across it later with a search, and the more we end users know, the better patches we can create with what we have now.

I've read this a couple times of course:

PHYSICAL AUDIO SIGNAL PROCESSING
FOR VIRTUAL MUSICAL INSTRUMENTS AND AUDIO EFFECTS

I still have hope for physical modelling, as something that can sound interesting and be quite musical.

My primary experience with it is this:


....which sounds amazing, but I was wishing at the time that it was polyphonic.

Its unpredictable, and sometimes difficult to reign in. When you do though, it can sound quite nice. Its hard to play tonally, but it can be done if really careful with a couple of settings. Some of that unpredictability can be good. Sometimes not. I mean, I know whats going to happen pretty much always whenever I turn a knob on a subtractive synthesizer. Unless I’m on mod-slot 6, and modulating modulators, you kinda just know whats about to happen.

I guess I have Sculpture in Logic also. Its really quite complex, but if somebody is really patient, you can dial in tons of expressivity via velocity and aftertouch. Its a little too much for me, being honest. Still, for a plucked string sound, its pretty great. Also no tonality issues with the software.

PM also great for odd percussive sounds.

Isn’t that big raw woody looking poly with no MIDI all physical modelling? Where’d that go?

Julius Smith's work is the best starting place for the type of physical modeling I am familiar with. Here's another good summary of his work:

https://ccrma.stanford.edu/~jos/pdf/DAFx06KeynoteII.pdf

I'd recommend searching the DAFX proceedings for physical modeling - some of the recent work has ended up there.

One of my favorite books on the subject is by Perry Cook:

Real Sound Synthesis for Interactive Applications: Perry R. Cook: 9781568811680: Amazon.com: Books

Perry Cook was one of the early CCRMA grad students that focused on physical modeling, and his web site is a useful place to go:

https://www.cs.princeton.edu/~prc/

I'm not sure if the STK is still there, but that had a lot of physical modeling building blocks in it.

I have been looking for a list of Sondiux-XG patents, but haven't been able to find this. The initiative started around 1996 or 1997, so any of those patents in the original initiative have expired by now.

I don't have any recommendations for modal physical modeling - my work was pretty much all in the waveguide / CCRMA space, while the modal stuff was more IRCAM.

Thank you!

I get by with Structure and AAS Tassman (which does all sorts of modelling I might add) but isn't supported anymore. Tassman 4 still runs well on any 64 bit enabled modern DAW.
I find between the two I have my needs covered.
I used to use a Z1 and WSA1r prior to this combination around 7 years ago.
If you can get hold of a copy of Tassman 4 it would still be worth the investment.

Thanks Sean. Will pick up a copy of that book.

Do you happen to know what is/was ADs interest in Staccato Systems? I have used many AD products over the years, but haven't been aware of the common ground with Staccato.

Sounds like you have had a very interesting career so far, and apparently a diverse education too. What do you have in mind for the future?

Honestly, I'm not entirely sure I know the answer to that question. My best guess is that the software wavetable engine developed by Staccato Systems allowed for Analog Devices to offer a MUCH cheaper alternative to the OEM Creative Labs cards. Instead of purchasing a Soundblaster for $10, which was a lot by OEM standards, a computer manufacturer could embed a very low cost AD codec on the motherboard, and the software (also supplied by AD) would take care of replicating the Sound Blaster functionality.

It is weird to remember how much video games used to rely on hardware wavetables and effects engines. Creative Labs OWNED that market during the 1990s. Flash forward to some point in the early/mid-2000s (I forget the chronology), and Microsoft announced that they were dropping support for hardware wavetables in their upcoming revision of Direct X. CPUs got fast enough, and the PCI bus stayed slow enough, that it became faster to resample audio on the fly in the CPU, rather than trying to stream sample data back and forth over a bus.

I didn't include all the temp jobs in the 90s in there. The interesting part of my career started somewhat later in life.

As far as the future, I'm working on another Valhalla DSP plugin in Xcode as I type this. Well, not while typing this. You know what I mean. Anyway, Valhalla seems to be working out, so more of that.

One thing is that while it's not too hard to figure out how something like an FM/wavetable/pcm synth works inside from using it + the manual + poking bits around here and there. Compared to that, nobody really knows how something like the VL-1 works inside, aside that it's supposed to be a "blown string" saxophone sim.

I've seen people suggesting it was this patent or that patent, but that's nothing compared to, say, the DX7 patent (which not only tells you how it works, but tells you even down to the level of how it's pipelined).


(incidentally, this talk of Analog Devices buying Staccato to use its software wavetable to cut out Creative out of the OEM business reminds me of how, in turn, creative bought out EMU for its wavetables and having EMU roll out an OPL3 clone so that Creative could Yamaha out of the gravy train before that)

The biggest issue with physical modeling is that it's still a huge CPU suck. I demoed Chromaphone and practically every patch brought my CPU to its knees. AAS is not dead, but Tassman is. Oddly, they found it easier to write a 32->64-bit bridge than to just fix the code for true 64-bit. But they parted it out into separate plugs to cover membranes, strings, analog, etc. (Sadly, Tassman is probably going to die with the next OSX update.)

One fun plug is Modelonia, which covers two different resonators--tubes and I think strings? It's not too bad on CPU usage. I can't remember which types of escorts they use either, but there's a demo available that's worth taking for a spin.

And two thumbs up for the WSA1. I still have no clue who/where that one came from...

It's not a synth but it's relatively new (2015) and a lot of fun. You can basically re-synthesize any sound as you please and then mix it with dry/wet which gives it a lot more flexibility than a synth for sound design: Moodal - Spectral Resonator - Tritik

I think the fact that Moodal generated very little buzz despite the fact that it's sound design capabilities are fairly novel and mind boggling shows the music tech market at large just isn't interested in stuff like that at the moment. Or maybe it would've made more of an impact as an eurorack module, seems like that's where all the innovation has gone? Unfortunately the format is fairly limiting and prohibitely expensive, I wish some of the tech would make it over to the computer.

Anyways, Physical Modelling as a means to create realistic acoustic sounds with a synth/computer obviously got killed by Kontakt multisampling. It's potential lies in bizarre hyper-real acoustic and electroacoustic sounds but I don't know how many people are actually interested in that.

I made more or less the same thing in reaktor couple of years ago ..only using 32 /64 partial filters .
I justused noise or impulse as an exciter
BUt if you wish ..the modal bank can use up to 10'000 partial filters ( steep bandpass filters ) .
Reaktor's modal bank it's where it's at for modal synthesis
modal 32 .1 | Entry | Reaktor User Library

I think they must have seen my ensemble , the overall layout is almost the same ...