Circuit modeled filter, how to?

[Aleksey Vaneev]

Can anyone please write down a simple non-linear system of equations consisting of three or four equations (without differentials), with one free variable (which will be an input signal) which cannot be solved other than approximately? I'd like to try the approach I've offered, but I do need a validation afterwards - so you should know what to expect from such system.

[nollock]

Here is a page with some self oscillation harmonic plots, which I find useful only as a double check once you have modeled the circuit of the filter, but thought might be useful to others:
http://vellocet.com/dsp/analog/SelfOscHarmonics.html

So basicly these give you an idea of what kind of saturation is in the resonance feedback? Those harmonic spikes trailed down after the main resonant spikes are a result of distortion in the filter?

So the moogs actualy have less 'in filter distortion' than the others?

The SH-01 is likely to be more assemtric clipping than the Juno 106.. as the former has a balance of even and odd order harmonics, the later more odd order harmonics?

Very interesting stuff!

[nollock]

yeah yeah yeah... i think andy is as least as bitchy as aciddose, period.

I dont remember Andy calling accidose a retard, or blind ect..

[nollock]

I think the key sound to strobe is the sub oscs. There is a saw and sqr with pwm, then a tri, sin, saw and sqr with pwm set of subs each of which can be switched to -1, -2, or -3 octaves. There is self sync, which is cheating a bit, but very useful, oscillator stack. The filter has around 30 filter responses created by making sums of the different taps. I especially like the notch filters where I have a notch placed exactly an octave down or up from the resonant peak. There is also a cool one with two notches, one an octave down, one an octave up, and a resonant peak in the middle, so this is kind of 6 poles of action out of a 4 pole filter Here are some graphs of the responses:
http://vellocet.com/dsp/CascadedFilterResponses/
Andrew

So is this similar to what mystran described in his recent post about the xpander type filter?

http://www.kvraudio.com/forum/viewtopic.php?t=207647

How is the resonant peak done? With feedback across the whole filter?

Nice graphs btw!

[Urs]

So is this similar to what mystran described in his recent post about the xpander type filter?

I think it's rather similar to the one williamk posted. It's afaik a cascaded diode design.

But the principle might be the same: You tap the filter at different points and mix the taps in certain levels to harvest different filter responses.

The Xpander/Matrix12 service manual is most helpful because there's a multiplexer for different configurations of resistors and inverters from each filter pole. From these configurations one can easily derive the levels in which the poles must be mixed to achieve the 15 filter types available in these machines.

Cheers,

Urs

P.S.: Andy, I finally had a long play with Strobe last night and I think it's awesome! It's got that really creamy sound.

[mystran]

But the principle might be the same: You tap the filter at different points and mix the taps in certain levels to harvest different filter responses.

Yup, the same basic idea works with any filter with 4 one-pole lowpasses in a row with global feedback around the thing. That includes the classic Moog ladder, Xpander, and the filter posted here (IIRC as I only took a quick look at the schematic). As long as the gain characteristics don't change, you can use the same coefficients for any of the filters (if they change, same principle still works, just take that into account when calculating the coeffs), since looking at the thing purely in a linear sense, they are exactly the same filter; it's just the non-linearities that change, and as long as those have little or no effect on phase or gain characteristics, they make no difference when it comes to getting different modes out of them.

In a real circuit other considerations might apply ofcourse.

The Xpander/Matrix12 service manual is most helpful because there's a multiplexer for different configurations of resistors and inverters from each filter pole. From these configurations one can easily derive the levels in which the poles must be mixed to achieve the 15 filter types available in these machines.

I would actually argue that it's easier to just derive the stuff from scratch rather than work backwards from the information in the service manual, but either way will naturally work. The Xpander implementation actually seems slightly more limited than is necessary for a DSP implementation, since in DSP you can choose arbitary points to extract your signals from, while the Xpander has to live with the limitations of the filter chip they use.

[WilliamK]

Why don't leave it to WilliamK to ask him if he wants? I know aciddose has better things to do than continue to propose how to approach these things to you and then have you argue with him about anything he says instead of just take some of his advice even though you have about 0.1% of the knowledge as well as experience in synthesizer and filter circuits and synthesizers in general that he has.

17 pages?! its impressive on how a simple thing can get out of hand so fast...

Anyway...

Thanks guys for all the input so far. It does look like I'm trying to chew more than I could eat.

I have to stop now since I'm working on another 3 projects, but I will come back to this in a month or two and check out how things are.

Best Regards, WilliamK

[WilliamK]

Haha, this thread ftw! \o/
WK, see what you did!!!

Why can't we all just get along?



Wk

[Rock Hardbuns]

I will come back to this in a month or two and check out how things are.

At which point this thread will have 300 pages of mostly semantic discussions interspersed by deliberate trolling.

It's a good time

[aciddose]

i think it was shy that had linked a clip of a saw on a sh-101;

http://xhip.cjb.net/temp/public/noisey_osc.wav

this is pretty close, just routed a little brown noise to osc fm here, res at zero. the example he posted is way more gritty, a lot more noise mod on that one. anyway certain synthesizers do a very good job of those basic sounds in some cases. you just need to patch things up in order to model effects like noise/60hz and others.

i've got this example here though which is something software really sucks at:
http://xhip.cjb.net/hardware/beefxor.mp3

just a basic pwm, no eq or anything else going on. i've never heard software come even close to that, and even my attempts have failed. (ok, i've never really tried hard, but..)

the asymmetric slewing combined with other factors give an analog pwm bass such life compared to what software can produce. same effects are extremely audible in my sh-101 self-osc example. all software synthesizers i've ever used/heard, including spore are completely unable to reproduce these effects. compared to the real deal, spore sounds completely lifeless and 'static' in situations where the real sh-101 is producing complex dynamics.

as i've mentioned, a combination of factors influence the sound, not merely one such as a saturation function in the filter. one of the reasons i have trouble understanding why people would be picky about the specific saturation function they use is that in my opinion it is not important. the general behaviour is all that is required.. if you concentrate on only one of these systems and do not model all the others (as evident in spore and other synthesizers) you are wasting your time. andy's work has produced moderate results, but none in my opinion that are worth the effort.

using per-sample modulation rather than your block processing might improve sound quality by a huge factor, andy. i can explain this in greater detail if you wish. an accurate envelope generator would also help greatly. (requires per-sample operation.)

[whyterabbyt]

So at 12.30 today Andy offers aciddose a copy of Spore, and by 5.30 aciddose is able to make claims as to what Spore is unable to do. Thats impressive all by itself.


Mandy Rice-Davies time, again, I reckon.

[aciddose]

i haven't used it yet, no time today. i'm still commenting upon the demos. i can tell he's using block processing as the envelopes are so damn slow, what is this, minimum time of 2.5ms? i can make more comments after i've used it myself.

you're free to post demos as a rebuttal, wr.

[whyterabbyt]

i haven't used it yet, no time today. i'm still commenting upon the demos.

so you're making claims of what it cant do based on some predone MP3's ???
Now that's a shocker. Again.

[aciddose]

are you trolling again? if you'd care to listen to the demos you'd see that there are the same problems. sounds like a typical cascaded lossy integrator with tanh() just like all the others. you can immediately tell that this isnt producing output like the electronics, if you have experience working with both software and hardware.

i think we can all accept that you're of the opinion that it isnt possible to judge something based upon a glance, many would share your opinion. i however do not. many years of absolute shit software claiming to be "the big one" has made me cynical at best. i hear absolutely nothing from the demos of spore which would convince me to look further at this point. wouldn't it make sense to post demos of the best quality output available? if these demos are in fact of very poor quality then i will retract my statements after i have tried the software myself.

at this point i'm assisted in my judgment by the opinions of a number of other highly skilled and experienced individuals.

[mystran]

i've got this example here though which is something software really sucks at:
http://xhip.cjb.net/hardware/beefxor.mp3

just a basic pwm, no eq or anything else going on. i've never heard software come even close to that, and even my attempts have failed. (ok, i've never really tried hard, but..)

Hmm. I wonder, what sort of PWM modulator is that? Triangle or sine or what? What sort of design? How stable? I'm wondering 'cos I'd need to kinda... well let's say I'd like hear what sort of PWM people generally find nice.

as i've mentioned, a combination of factors influence the sound, not merely one such as a saturation function in the filter. one of the reasons i have trouble understanding why people would be picky about the specific saturation function they use is that in my opinion it is not important.

I would disagree to a point, as a carefully selected and/or tuned saturation function can provide one with a trade-off between sound quality (aliasing artifacts and such), processing time (especially the oversampling required), and subjectively desirable distortion characteristics. If one would only had to consider two of those, I'd agree with you.

using per-sample modulation rather than your block processing might improve sound quality by a huge factor, andy. i can explain this in greater detail if you wish. an accurate envelope generator would also help greatly. (requires per-sample operation.)

In some cases there's even audible difference between driving an oversampled filter with oversampled-rate or base-processing-rate "cutoff CV" (with or without some sort of interpolation).