What I have learned about synthesising musical instruments is to understand the shape and performance space of them and not strive for high spectral realism but for consistent acoustic viability.
Because musical instruments are part of a well recognised group of sounds within which we can easily distinguish quality variance the unreal use of synthesis makes facsimiles extremely dull, especially those based on S+S synthesis, and those made by modern physical (waveguide modelling) too.
The problem is that as we approach high spectral realism the critical faculties we use to hear are more likely to focus on features that don't feel right. You might have heard this referred to in computer graphics as the "uncanny valley" where _almost_ photo-real human models seem revoltingly unreal while crude cartoon characters seem "more human", and are able to make expressive gestures that pseudo-humans cannot get away with.
For very good spectral realism of brass try expanding upon something like "Contiguous group wavetable synthesis of the french horn" [Andrew Horner and Lydia Ayers (187 in the Csound book)] combined with a Chebyshev table shaping approach for the spectral dynamics. These two techniques combined will lead towards realistic results.
But what really works, in my opinion, isn't to go for classical realism at all. Look for something that gives you the parametric expression of brass but extends it to the hyper-real.
This is to be found in
- transient overshoot, tonguing, stopping
- very careful amp and spectral correlation (probably
using lookup tables for the xfer functions), embrochure postion 3) Non linear, cubic and intermodulation artifacts from overdriving the cavity. (careful amp controlled distortion) 4) Careful high pass design for the bell + horn end correction in tuning
Listen to the Travelogue version of "Being Boiled" on the Human League's second album. That is synthetic brass (done on a Roland system 100 I think) that pulls down it's trousers and shits all over _any_ modern digital synthesiser I have ever heard. Not because it's real, but because in it's simplicity it uses a few parameters to extend the performance space of brass into the hyper-real, an acoustically viable expanded performance space.
On Sat, 17 Jan 2009 08:13:11 +0000 errordeveloper@gmail.com wrote:
hi
i'd like to make a synth which would sound quite like the metal wind-blown instruments .. unfortunatly i haven't got Perry Cook's book handy, i had from the library before, but now ..
has anyone done any of similar stuff in pd? my project is not about making the same sound and/or simulate the instruments, but just make a synth which would have some similar sound character, i mean something that wouldn't just remind the instrument sound (like some cheap old kidies synthie could //) .... more like a sax put through some spacy effects is what i'm looking for!
cheers, happy patching! -- ilya
Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list