2

Background

I am working on synthesized patches using a combination of additive and subtractive methods to create a digital orchestra. The goal is not to recreate orchestral sounds, but rather to create an instrumentation for polyphonic works that builds on principles from orchestral instrumentation of a classical orchestra - roughly the size of what is required for Beethoven's work.

I am currently building a string-like section from a single patch with a long sustain. I would then like to process this patch into easy distinguishable variations that mimic the relationship between the string instruments: violin, viola, cello, and double bass. The overall waveform for these instruments is to my understanding similar. However, the individual characteristics seem to be heavily dependent on the resonating body of each instrument. A viola is basically a larger violin. For a cello, however, I think the bow is broader and the strings have a different characteristic.

I am familiar with Karplus-Strong synthesis, but my goal is not a realistic string instrument. Instead, I would like to apply the principles that differentiate a family of instruments to create individual patches that belong to the same "section". To achieve this for strings, I think manipulating the characteristics of the resonating body - size and shape in particular - could be useful.

I have tried to use short reverbs, long unmodulated chorus and other variations on copying and delaying a signal to simulate different resonance characteristics. I also was somewhat successful in simulating the bow using enveloped noise and very short granular delays. Ideally I am looking for a VST solution or using mix buses and filters.

Question

Are there any common techniques for simulating resonating bodies?

2

I don't know if this is exactly what you need, but in my experience I've obtained interesting results using IRs of several bodies, e.g. Altiverb's IRs have a really awesome "Design" section with IRs from a djembe, a dustbin, a flower pot and more.. This worked very well when I wanted to simulate something like a resonating body.

Cheers, d

  • 1
    You could also check this link :) scitation.aip.org/content/asa/journal/jasa/77/S1/10.1121/… – fragmentsinabox Jul 6 '15 at 16:25
  • I also reached the conclusion that convolution would be a solution. By simply tapping the bridge (in an anechoic chamber) you could potentially record the IR of a Stradivarius (Sleator, 1985, Curtin, 2015). However, I have not been able to find any database for instrument spaces. Curtin, J. (2015) in Schulman, Finding the Elusive Digital Stradivarius transistor.prx.org/2015/05/digital-stradavarius – noumenal Jul 6 '15 at 17:41
  • 1
    iZotope Trash also contains a bunch of nice weird IRs, including ones that don't belong to acoustic objects at all. — @noumenal: tapping the bridge doesn't really create a very dirac-like transient though, you'll definitely need some correction. – leftaroundabout Jul 6 '15 at 20:51
  • 1
    Csound also has some interesting opcodes for physical modeling, you could try to manipulate those opcode's parameters to obtain what you need, more or less :) – fragmentsinabox Jul 7 '15 at 12:02
  • @leftaroundabout I am new to convolution and deconvolution, so I am not sure I understand. By Dirac-like function do you mean that the amplitude of the isolated source has to be proportional to the raw signal (source + IR)? What kind of correction would be needed? If the bridge can be isolated in an anechoic chamber, would it not be possible to acquire the IR of the violin-space by subtracting the bridge? – noumenal Jul 7 '15 at 18:44
1

Are there any common techniques for simulating resonating bodies?

Look up physical modelling. Be prepared - it's a hugely academic subject and involves more equations and physics than you can shake a wooden bow at. There are so many variations depending on an almost infinite amount of factors to the point you realise it's practically impossible to manage mathematically.

A starting point however would be:

  • 2D wave equation
  • 3D wave equation
  • The Physics of the Violin (Cremer) a thorougly comprehensive book

You'll need to sit down with Matlab or similar to really get your hands dirty!

  • I mentioned a physical modelling method in the OP (Karplus-Strong). I have access to Matlab, but to me this sounds more like a PhD project. I am looking for an oscillator-free transformation that could be applied to any waveform, so I am not entirely sure if physical modelling would help me. By "2D wave equation", do you mean 2 pi f? +1 for the book tip! – noumenal Jul 7 '15 at 18:50
  • Physical modelling does involve at-least degree level maths. If you're not prepared to go down that route, then physical modelling really won't be for you! If you are looking for a transformation that could be applied to any waveform create a filter that emulates the body resonances, though this will give nowhere near the same results. Be prepared to alter that filter by the frequency of the input. – user3791372 Jul 8 '15 at 17:43
  • I have sound design as a hobby only. I could potentially brush up on my mathematics, but it would still take me ages to implement. I have a decent understanding of wave mechanics, FFT, and wavelets, because I use signal processing in my daily work on 2D data, but alas, I have no background in acoustics or physics. – noumenal Jul 9 '15 at 8:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.