[sdiy] SSM2164 state-variable filter

[Tom Wiltshire]

Hi all,

I've done a bit more work towards trying to eliminate the frequency  
effect from the resonance on the SSM2164 SVF filter I'm playing with  
(http://www.sowa.synth.net/modular/m_vcf.gif). I'm just looking at a  
single filter section, without any of the CV mixer stuff.

I drew a graph of the resonance response:

	http://www.electricdruid.com/ResonanceGraph.gif

Note the linear scale. It probably doesn't look as bad on a log scale!


Neil J also suggested that the effect might be because of limited  
slew rates of the op-amps. I haven't got a high-speed op-amp handy,  
but figuring that if that was the case, a worse slew rate should make  
the effect worse, I tried a LF444.
The slew rate of TL074 is 13V/uS, against 1V/uS for LF444. There was  
no change at all.

I've also experimented adding various capacitors over the 30K input  
capacitors. This just seems to limit the resonance overall, without  
much effect. I haven't yet started trying compensation on the first  
mixer stage.

If anyone has any further ideas, I'd appreciate it. I'm running a bit  
short myself.

Thanks,
Tom


On 10 Aug 2008, at 18:41, Neil Johnson wrote:

> Hi Tom,
>
> As I'm just finishing up the design of an SSM2164-based 3-pole  
> state variable filter I think I can offer some thoughts...
>
>> Firstly, the resonance varies with cutoff frequency, being more  
>> pronounced at higher frequencies (in fact, it'll oscillate at high  
>> frequencies when it won't when you tun the cutoff down. Is this  
>> normal? Do all SVF designs do this?
>
> I think in theory resonant fequency *should* be independent of Q.   
> However my suspicion is that various non-linearities and other  
> frequency-dependent effects (op-amp slew rate, limited GBW, etc)  
> change the behaviour sufficiently that there is a connection  
> between the two.
>
> This is one aspect of my design that I am currently experimenting  
> with.
> An early draft of the schematic is here:
>    http://www.njohnson.co.uk/pdf/MSM/msm-vcf.pdf
>
> [Note: there are some errors in that version - a revised version  
> will be appearing sometime soonish once I'm happy with the core  
> design.]
>
>> Secondly, the resonance increases as the resonance CV increases.  
>> Whilst this is logical enough, the CV serves to increase the  
>> attenuation of the VCA, so increasing the CV actually decreases  
>> the output from the resonance VCA. Does this mean this VCA is  
>> being used to cancel out the effect of the other feedback loop?  
>> (through R6 or R53 on Roman's schematic)
>
> Yes - in reality what you are changing is the 'damping'.  At 0V the  
> gain cell has 0dB attenuation, so you are at maximum damping  
> (lowest Q).  As the control voltage increases so does the  
> attenuation (at 33mV/dB), so there is less damping (higher Q) and  
> the filter starts to ring - eventually going into self-oscillation.
>
>> Thirdly, what is the function of the 5V6 zener diodes on the final  
>> integrator? I haven't put them in yet, and it seems to work.
>
> They limit the loop gain when in self-oscillation, otherwise you  
> can end up slamming against the rails.
>
> Cheers,
> Neil
> --
> http://www.njohnson.co.uk
>
>
>
>