View Single Post
Old 10th July 2016 | Show parent
Registered User
Originally Posted by pichenettes ➡️
Roughly speaking, there are 3 approaches to physical modelling.

* Finite element methods. You solve continuum mechanics equations on a 3D model of the instrument. As if the instrument existed in a virtual world, and you simulated every single force, vibration, deformation happening in this virtual world. Very computationally expensive and completely out of reach of analog methods - unless you use the analog components to build a room-sized digital computer. Sculpture and Kaivo use a very rough version of this, limited to a 2D (or 1D?) simplification of the main resonating structure of the instrument.

* Modal synthesis. A few approximations and a mathematical trick allow you to describe the way your instrument responds to the player's actions by a network of bandpass filters and gains "tuned" to the properties of the instrument's shape and material (if the material rings a lot the filters are very resonant, if the material sounds "bright" the filters with the highest cutoff will have a higher gain, if the instrument is large the filters will have lower frequencies...). That's what Elements does and this needs many bandpass filters (about 60 in Elements), the more filters, the highest the quality. So if you want to do that with an analog system, you'll need many, many VCFs, and any single change of the material properties or note will require you to adjust the cutoff/resonance/gain of each of the filters. The filters would need extremely accurate tracking for the sound to be in tune. That would be doable in analog, but would be very expensive and totally impractical. This synthesis method is used in all the Applied Acoustics Systems products, and in the Wavedrum.

* Waveguides / Karplus Strong. The phenomena of propagation, absorption and reflection of waves in a string (or tube) are modelled by a delay, a filter, and feedback element. Cheap to do with digital systems. You can recreate that with analog building blocks (BBD delays, VCFs - people routinely patch that with modular systems) with the caveat that it'll be almost impossible to make it play in tune: not only you need a delay with a precise control of short delay times, but the VCF introduces a frequency dependent delay that needs to be compensated. This is the "mid-late 90s thing" Yoozer refers to.
CCRMA lists six unless you are combining Ruiz-Strings with Cardis-Anima and Mosiac processes