Violin Signal Processing

[Blacklace]

Hi, I've built myself an electric violin. It plays fine, but the sound is dull and uninteresting, as one might reasonably expect from a minimal solid bodied instrument. The violin has piezo buzzer pickups fed into a high-Z preamp.

I've also built myself a DSP box. It has a 200Mhz SH4 processor, with an FPGA and AC97 codec. I've been using it as a reverb and eq box for the violin. It helps some, but I've run out of ideas to improve the sound.

Adding reverb just makes it sound dull and uninteresting in a church hall, and with some eq I can make it sound like a (dull and uninteresting) church organ or a clarinet or a variety of other (dull and uninteresting) textures. Adding distortion just makes it a distorted dull and uninteresting.

I initially hoped to use small time constant reverb to simulate the resonances in an acoustic violin body & bridge, but so far no amount of that has even begun to approach the niceties of even my cheap chinese acoustic. I'm missing something vital here.

I'd love some more ideas... Thank you.

[suntechnik]

Real violin sounds juicy due to higher frequencies build up based on fundamental one by Helmholtz resonance therefor electric instrument lucks appealing timber of an acoustic instrument.

Tube amplification using awkwardly biased tube is one possible approach that is in use for electric guitars.

DSP route would be using TI sound processor and wavelet algorithms to achieve some sound coloration of original signal from a pure string in a manner of real acoustic instrument. That needs some comprehensive R&D for developing proper mathematical algorithms and I believe might be out of scope for a DIY project.

WADIA implements similar approach in the decoding computers for RedBook playback BTW.

[phase_accurate]

I also think that it is not that easy to imitate the sound of an acoustic violin with a solid-body instrument but there are those "silent" violins by Yamaha etc that sound quite real.

I own an electric upright bass. That one doesn't sound like the real thing either it rather sounds like something between an upright and a fretless bass guitar. I guess all these resonances caused by the bodies of the instruments of this group are hard to imitate.

If you play it unamplified - how is the sound ? Is it closer to your likening than the amplified version ? Where is the piezo mounted ? Where is the preamp located (right in the instrument or after some hundred yards of cable ?) ?

Regards

Charles

[Blacklace]

Suntecknik, do you have any links to info/papers on this wavelet approach? I know what wavelets & wavelet transforms are, but I don't see how I'd apply them to this problem.

Phase_Accurate, the acoustic sound isn't hugely different to the raw amplified signal. a bit thinner (just less LF?). It's not terrible, but again, totally uninspiring. But perhaps it is a little warmer, the piezos do tend to accentuate scratchiness, but some eq does takes care of that. The preamp is 2m of coax cable away. The piezos (2) are currently mounted under the bridge feet - but good suggestion, perhaps elsewhere on the instrument is worth trying.

Maybe I should try electric guitar strings and coil pickups too... My goal here isn't to imitate an acoustic exactly, I realise that that's never going to happen. I just want something that plays warm and easy (and acoustically quiet, so I can practice with headphones when I got home from work late without annoying the neighbours)

[Simon B]

Interesting project, shame it isn't really doing what you'd like yet.

Phase_accurate makes a good point asking about the pickup and what it's attached to. You need a good rich output to be able to do anything useful with it.

The short reverb idea for simulating body resonances doesn't work because it's basically a time delay, so you end up with a series of harmonically related peaks and valleys.

What you need is more likely as big a set as possible of largely harmonically UN-related resonances! So each different note you play ends up with a different but still somewhat similar spectrum - the spectra will look much more different than they sound.

If you could record a set of notes from your normal violin in as dead a room as possible, with a decent flat mic quite close, and a similar set from your electric, then compare their spectra, that might be a reasonable starting point for thinking about filtering.

Do post more, interesting stuff.

[Blacklace]

OK, so I guess getting unharmonically related components means some form of smooth non-linear transfer function? (hence the awkwardly biased tube preamp suggestion) Maybe hyperbolic tangents, or just even cubic polynomials?

Anyway, thank you, this looks like a promising avenue to pursue too.

[Simon B]

Originally Posted by Blacklace

OK, so I guess getting unharmonically related components means some form of smooth non-linear transfer function? (hence the awkwardly biased tube preamp suggestion) Maybe hyperbolic tangents, or just even cubic polynomials?

Anyway, thank you, this looks like a promising avenue to pursue too.

No, I'm talking about filtering that is linear (ie, it generates no harmonic distortion), but has a non-linear frequency response (different frequencies come through with different amplitudes).

If you were to feed it a sine wave of ANY frequency, you'd get a sine wave out, though with phase and amplitude that varied according to what its frequency was.

So long as you have a rich, sawtooth-like (at least as far as its spectrum goes) signal to feed into your filter network, you don't need to distort it, at least not to start with.

Your challenge is to find out what set of filters sounds good - you need to be looking at the placing of the peaks and troughs on a logarithmic scale, and placing them at irregular spacing.

Remember you're trying to get something that affects the sound in something like the way that the complex set of resonances that a violin body has.

[pixpop]

Not saying that this is easy... but it ought to be possible to do what you want using a convolution reverb. This is at the heart of many modeling amps/instruments, such as those made by Line6.

[FrankWW]

violins are very rich harmonically. this might be of interest:

Master Studio for Violinmaking - Martin Schleske Munich, Germany Harmonic Structure

[Chris Daly]

Hi Blacklace
It would be interesting to know what style of music you play. As regards electric violin, a few artists really stand out. David Cross ex King Crimson, his setup in 2000 was: Zeta violin, Art processor, Peavey combos. Then: Barcus berry bridge pick-up on a Vincent Violin, Pete Cornish pedal board, H and H stack.

Papa John Creach, who contributed to the band Hot Tuna. also Don "Sugarcane" Harris and Jean-Luc Ponty. where electric violin stands out on Frank Zappa's Hot Rats. you might be able to follow up how they achieved their sound.

If you are recording I recommend DBX mic preamps, or build your own, DBX type 1 which are companders a combination of compression on recording and expansion on playback, as found in the 150x and 155 models provides a nice solution.

Cheers / Chris