AW: Another VCF Finished

[gstopp at fibermux.com]

     I'm using 100K/220 dividers to keep the 3080 input structures in 
     linear mode. I haven't honestly done noise floor tests but I think 
     that since the filter seems to handle 10v p-p signals nicely that this 
     will be the best way to keep the S/N high.
     
     Of course over the weekend I've had much time to think about this 
     filter (like in my sleep, or during dinner, etc. you know how that is) 
     and I have some additional thoughts. First I went back and read 
     Electronotes #41, because there are many pages of theory about the 
     filter before the schematic was presented, including S-plane graphs in 
     3-D. So here was a case of "build it first, understand it later".
     
     Much of the discussion was about poles, and ideal filter 
     characteristics, which led to the 4th order Butterworth idea 
     (different pole positions for steepest cutoff), but then they decided 
     to use the good old all-four-poles-identical design because this was 
     easiest and all commercial synthesizers on the market used it anyway.
     
     I was confused before about the difference between the different 
     current-controlled integrator designs:
     
     * 3080 to cap to buffer, (-) input grounded, no feedback
     * 3080 to cap to buffer, negative feedback from buffer into 3080
     * 3080 to (-) input of op-amp integrator
     
     My understanding now is as follows: the first type is the same as the 
     third type, in that they are both "open-loop" integrators. They are 
     used in state-variable configurations because in that filter type the 
     integrators are tied together with an overall feedback path to 
     establish stability. In other words, the individual integrators have 
     no feedback because the way they are connected together at a higher 
     level provides the stabilization.
     
     The second type however differs in that it is a unity gain integrator. 
     This means that it can be used by itself to provide a 6 dB/octave 
     rolloff lowpass response, with no additional feedback. It is a 
     "stand-alone" integrator. Cascading several of these in a row will 
     simply increase the slope of the rolloff, without changing the 
     stability of the overall network at all. In fact until some kind of 
     feedback is added, the stability is at a maximum at all times. Note 
     that this differs from the state-variable which is at maximum 
     *instability* when there is no feedback.
     
     Since each unity-gain integrator in a multi-pole filter is 
     non-inverting, resonance is added by bringing the output back in to 
     the input of the first stage, inverted (this design does it by summing 
     it into the (-) input of the first stage). For self-oscillation in a 
     4-pole lowpass the feedback must have a gain of 4, since each stage 
     attenuates a little at the cutoff frequency. An interesting effect of 
     this feedback is that DC levels and subsequent frequencies up to but 
     not including the cutoff frequency of the filter are attenuated by a 
     factor that is directly proprotional to the level of the resonance 
     feedback. This is the cause of the apparent drop in bass volume as 
     resonance is turned up.
     
     Adding inverted feedback to an amplifer will always cause oscillation 
     if the gain is high enough. The frequency of oscillation will be at 
     the fastest rate that the amplifier can respond, and in a lowpass 
     filter this is the cutoff frequency. This is also the reason that a 
     4-pole highpass filter has no resonance control - because the 
     oscillation will be at the fastest rate that the amplifier can respond 
     at, and that will probably be in the hundreds of kilohertz and *not* 
     at the cutoff frequency of the filter!!!
     
     Also the article in EN states that the filter's range is 1 Hz to 22 
     Khz, and I suspect that the reason that I found a 3.5 Hz to 13 Khz 
     range in my filter is that I was running it off of +/- 12 volts rather 
     than +/- 15 volts - more experiments to follow I suppose.
     
     - Gene
     gstopp at fibermux.com
     
     


______________________________ Reply Separator _________________________________
Subject: AW: Another VCF Finished
Author:  Haible_Juergen#Tel2743 <HJ2743 at denbgm3xm.scnn1.msmgate.m30x.nbg.scn.de>
at ccrelayout
Date:    4/15/96 7:30 PM


Gene,
     
great to hear about your filter!
     
What resistor deviders have you used? 100k/200R to keep the
gm stages from overdrive, or some smaller ratio for a limiting 
effect?
     
JH.