| Author | Topic: comb filters and sound synthesis | |
| Super Chu-Hi | Posted: 26th May 2003 23:31 | |
A week or so ago, i was playing around in absynth with a flute patch. I noticed that turning the comb filter on and off made a big difference to how real the flute sounded. After doing a bit of searching, i found this on a webpage about an analogue comb filter
( http://www.analogue.org/network/rs120.htm ). Does anyone know where there is some more information on what these natural resonances are? | ||
| pHz | Posted: 26th May 2003 23:37 | |
xoxos - you there ???
i think this ones for you slainte | ||
| CreepJoint | Posted: 26th May 2003 23:42 | |
Hi, dont know if this is any good for you, probably too basic?
http://www.glenbrook.k12.il.us/gbssci/phys/Class/sound/u11l5d.html Matt | ||
| Krakatau | Posted: 26th May 2003 23:46 | |
Just be aware of this :
A "comb filter" is in fact a very short delay line ...controlled by a LFO it becomes a FLANGER Nothing to do basically with a band-pass filter ...hope this helps ! | ||
| Super Chu-Hi | Posted: 27th May 2003 00:08 | |
Thanks for the link creepjoint, the stuff about harmonics looks interesting.
Krakatau - I was under the impression that a comb filter was just a bank of band pass filters or something. i did read something about comb filters using delay lines, but it mustn't have sunk in. | ||
| whyterabbyt | Posted: 27th May 2003 01:20 | |
Krakatau quoth Just be aware of this :
A "comb filter" is in fact a very short delay line Thats one way of implementing it, yes. There are other ways to do it, though. ...controlled by a LFO it becomes a FLANGER Not necessarily. That's dependant on the length of the delay line, and the path of the rest of the signal. Nothing to do basically with a band-pass filter Ummm, it is, sort of. The result of flanging -is- effectively that of multiple band-pass filters. Almost all common digital filter models are in fact based around very short (single-sample, in fact) delay lines. Whether you use the term 'comb-filter', 'flanger', or 'band-pass filter', what happens mathematically is very similar. An attenuated delayed version of the input signal is added to the original... | ||
| Jeez | Posted: 27th May 2003 06:33 | |
This week: Comb Filters and Physical Modelling!
Comb filters are often in integral part of physical modelling. This is because they're a cheap (CPU-friendly) way to simulate the effect of a sound resonating within a regularly-shaped cavity (like a metal pipe - flute). This is based on the concept that a sound is injected into the cavity at one end, travels to the end of the cavity, bounces back, and reaches the point of origin (before bouncing back and forth again and again). You could also think of it as a kind of reverb - but the space is incredibly small, and incredibly simple. The incredibly small[/] part has two (main) effects - the 'verb time is so short that there is no percievable "reverb tail". The other effect of reverb an incredibly small space is that the early reflections are so close that they interfere with the actual signal... which brings me to the second part - the effect of the reverb in an [i]incredibly simple cavity. This means that the interference (as described above) is predictable and controllable. This property is what allows the player to create different notes. With a "normal" reverb, a simple space would result in noticable ringing (what makes "bad" reverb) - but because there is no percievable reverb tail, we never hear the ringing on its own. We only ever hear it as it interferes with the source sound. Coming up next week: Formants!! Forever, Kim. | ||
| Jeez | Posted: 27th May 2003 06:40 | |
Could you explain a little more please? I'm intregued... Forever, Kim. | ||
| Urs | Posted: 27th May 2003 08:50 | |
The result of flanging -is- effectively that of multiple band-pass filters.
whyterabbyt, it's rather a set of notches in series. Bandpasses in parallel would (in the digital domain, at least) result in additional combs (= loss of bands), due to phase shifts. Well, and when it comes to formant filters vs. bandpasses, from what we commonly understand, a formant filter is merely like a (positive) peaking eq. - It doesn't damp the outside bands as much as a bandpass does, it mostly emphasizes a certain frequency. For those who want to play with formant filters, I generally do some self promotion and recommend Zoyd (Mac only currently, but PC users can have a listen to the dumb demo song), follow link to my website Cheers, | ||
| Super Chu-Hi | Posted: 27th May 2003 09:22 | |
Thanks for your explanations Jeez and Heckmann, I'm starting to understand a little more. Zoyd sounds good, nice demo songs. But Zoyd does sound like a monster to program. its a shame i'm on a PC. So a comb/bandpass filter could be used to make some simple synth patches sound a little more realistic. ie. a flute patch in Albino. What about for something that sounds like a plucked string? Could a comb filter be used to help simulate a guitar body, or is the guitar body shape too complex? | ||
| ik | Posted: 27th May 2003 09:29 | |
Bingo! SYMPHO has two of them. ik. | ||
| Urs | Posted: 27th May 2003 09:47 | |
In fact, all those Physical Modelling Plucked String synths are based upon a special comb filter. - The feedback of the delay is full-on and damped by a lowpass filter, that's why the sound starts very brilliant and fades out becoming smoother. Very simple. It's also known as Karplus Strong synthesis. More or less, cause it's patentet, so everybody has it's own interpretation of the technique... Cheers, | ||
| ew | Posted: 27th May 2003 09:54 | |
ConcreteFX's BrushFX and BrushSyn are comb filter based-in fact,it's four comb filters and a routing matrix.The filters' cutoffs can track the keyboard,and they're lots of fun!BrushFX has audio ins and BrushSyn is just a VSTi that uses comb filters as the only sound source.Highly recommended www.concretefx.com ew | ||
| Scott @ AAS | Posted: 27th May 2003 11:42 | |
For those of you who would like to explore physical modeling in more depth, I highly recommend hopping over to our site and grabbing the Tassman demo.
Scott | ||
| ew | Posted: 27th May 2003 12:48 | |
Truth!Tassman's a wonderful tool for that.
You can do it in Reaktor as well-Steampipe's a great example of what you can accomplish with physical models. ew | ||
| Super Chu-Hi | Posted: 29th May 2003 17:58 | |
thanks for your replies everyone. i'm going to go play with some bandpass filters now. |










