[sdiy] FatMan Hacking

[jhaible at t-online.de]

> >The ladder filter normally has unity resonant frequency 
> gain. 

Not to split hairs here, but I thought the resonant peak
goes *up* as feedback is increased. (And the low end goes
down, of course, which was the main point.)


> Two ways that I, and others, have used to overcome this:
> 
> 1. Increase the overall gain of the filter as resonance 
> is increased. 
> The Roland SH-2000 does this, as did the System 101 I 
> think. And my own 
> VCF-1, of which I thought was very clever and unique 
> until I saw the 
> schems for the SH2000. This has the benefit of keeping 
> S/N high too. 

Yes. And I recently found an interesting thing about
this compensation a la Roland. Every Roland filter
is different, and they also used this trick on
their later (4pole) OTA filters, but now I'm talking
about an early diode filter. I think It was the 
System 100 filter. I'm sure it was a 5pole diode filter.

Background: The low end drops on both, transistor and diode
ladder filters when resonance is increased. But the amount
of bass drop is bigger on diode filters: This has to do with
the different phase & amplitude values for the self oscillation
point of a straight 4pole vs. a 5pole filter which approximates
the 24dB/Oct slope.
So in this circuit there is a dual ganged pot, one for feedback,
and one for the filter input level.
To compensate the bass loss ?
Yes, but (big surprise!) *only* to the level of a straight Moog
ladder! It's funny to see this in a circuit simulation. Rather than
bringing the bass end up to unity gain, they only bring it up
to the Moog filter's level. (That's -12dB compared to no resonance
if memory serves.)


> On a side note: I believe the Korg Polysix actually turns 
> up the volume 
> of the whole instrument to compensate for the drop in 
> passband gain when 
> Q is turned up.

I think I've seen it in the Mono/Poly. (Long time since I looked.)

JH.