Hello,
sorry for being away from my own topic. Somewhere I decided to first find my own way and then look into what you wrote. Which now opens up a new bunch of ideas.
In the attachment you can take a look at what I produced so far (pd-extended). It is not finished and I think some sliders could be thrown out. It was more an intuitive approach and not a physical model. The result seems to be noisy enough for me and has some interesting effects and uses in it. Not that it is near any speaker simulation like you discussed here. The -old version was the first draft. But the heavy use of fexpr~ made my CPU cry. My current version uses one block delays and tries to be the same. But the result is quiet different, as the 64 samples delay makes it impossible to cut the waveform like in the -old version. But I reduced CPU usage to 1/5th, which is still too much for my liking.
By the way it is kind of sad that everyone (including me) always uses tanh, while there are much much more functions out there which could be better suited, produce less high harmonics (aliasing!) or just sound better. A while ago I took a small dive into this topic but only remember some sigmoid functions, but not what their effect was: atan, 1/(1+exp), x/sqrt(x^2), [sqare-law], erf() [I couldn't test because its only in the documentation of expr but not useable] But this should be a topic of its own.
@Martin Thank you for the complex and interesting physical analyzation. Now I have a bunch of new ideas of what I could do next. No I didn't know that page but it looks interesting. I'll soon take a closer look at it.
I'd be interested what kind of speakers fallen_devil was interested in when originally posting his question to the list. Also a record of the noise would be nice for comparison...
I am interested in the whole topic and the possibility to create completely different sounds. Not the sound of one special speaker. I wanted to do a record too, maybe I get around to it soon. But I'm not sure what I should use as a test sample. I thought about increasingly loud snare + bassdrum and a bassynth. Any other suggestion?
@Mathieu Great distortion you created there. Also a topic of its own. Thanks for the lop2~ now I don't have to always use the iemlib filters.
Greetings
Am 16.10.2010 06:46, schrieb Martin Schied:
Hi!
I'm no speaker modeling expert at all, but I can try to describe what produces sounds in an overloaded speaker. There are various sources of distortion, symmetrical (mechanic suspension) and asymmetrical (magnetic field) and also time variant (temperature) and modulation (doppler effect / amplitude modulation) effects. I don't know which effects have a stronger or weaker influence, but I describe what I imagine:
High peak amplitude, positive wave:
speaker moves to the front, parts of the coil will not be in the magnetic field anymore. The field isn't zero outside the magnet gap, but I guess it decreases rapidly and is almost zero (so for example if half the coil is inside the magnet, the parts outside will not produce a force. So the force is only half as strong as it should. For simplicity you could say the field outside the magnet's gap is zero, so you have a linear function of excitation / current. Also If the coil moves out of the field its impedance will decrease which has influence on frequency response for higher frequencies a bit.
Heat
If the impedance is reduced as described above, a higher current will flow and heat the coil more than usual. The resistance of the coil will increase when it gets warmed and thus the efficiency of the speaker decreases (up to -7 dB I read somewhere, but this wasn't meant for almost dead speakers but heavy load). The heat needs some time to dissipate, so some kind of slow pumping compression effect occurs.
High amplitude, negative wave:
The speaker's coil might crash into the magnet and create different mechanical noises. Also the speakers diaphragm will be deformed by this crash and create various kinds of noise. Additionally it carries the noises the coil created - depending on the material and shape this sound different (paper, Kevlar, aluminum, etc sound different). If this crash doesn't occur (professional speakers don't have this issue usually) the negative wave will not be distorted too much and maybe distortion can be ignored.
In both directions the spider (basically a spring) starts to become nonlinear. Different manufacturers have different curves, but for small amplitudes they all pretend to be linear - so some experiments with sin~ or tanh~ might do it here.
Then generally there are happening doppler-effects on all speakers with big excursion. You could model them through a variable delay, modulated by a differentiated, low pass filtered signal (don't bite if I'm wrong, it's already very late... ). Amplitude modulation can be applied the same way (lowpass and apply it to higher frequencies).
so to sum it up:
apply symmetric distortion for the spider, split the path into positive and negative parts, for positive samples: tanh~, polynomials or other wave shapers, for negative parts let the signal untouched or add noises of a crashing coil (don't know how to achieve this), then sum both signals up, apply doppler effect, amplitude modulation and pumping compression. perhaps that sounds like your speakers then :)
I'm not sure if this works at all, but it definitely will sound very distorted in the end.
did you already discover http://www.klippel.de/pubs/papers.asp ?
cheers Martin
On 15.10.2010 21:10, - wrote:
Thank you for your answer, but as I wrote I don't want the sound of simple clipping like clip~, tanh~ or overdrive~. I want the sound of a speaker crying for mercy because you put just too much through it.
But I don't know where to start. I know there are complex distortion effects, which are able to simulate different speaker cabinets after variable amps recorded by different microphones. But they all cost big money. Also I don't need the physical simulation. I just want the sound. If you want I can try to record the sound I'm talking about.
I tried to search for information how to do this but couldn't find anything usable. Not even an analysis what happens inside the speaker when you torture it like this.
I already know the forum but don't want to doublepost. I really liked the post about the oto biscuit. Neat distortion possibility's.
Am 15.10.2010 17:12, schrieb George Ker:
Hello, I can't really understand, so , you mean something different from
[clip~] ?
I' m sure you can find really good patches in the puredata.hurleur.com
forum searching about distortion , overdrive clip etc
GeorgeKer~
On 14 October 2010 23:19, - fallen_devil@gmx.de wrote:
Hi, does anyone know how to simulate the sound of an overdrivenspeaker? You
know the crunchy sound when you torture it with a strong bass. It's nowhere near the sound of an normal overdrive with some kind ofclipping.
Greetings - _______________________________________________ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management ->http://lists.puredata.info/listinfo/pd-list
Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list