24 db/oct filter questions

[tomg]

Hi Chris,

> I realize that using unequal
> cap values in the filter ruins it's correctness (gives unequal corner
> frequencies for the stages, correct?).  

> My question is, am I just
> fooling myself here or is there a logical reason why I might like this
> sound better than the "proper" way?  Perhaps the different cutoff
> frequencies "fatten" or widen the resonant peak?

It makes it lumpy..you can hear the stages sweeping a different
frequencies. I made a BP filter once that worked this way. As one
stage reached it's max freq the next stage started from it's lowest. 
This action reversed as the cutoff was lowered. It made one huge
sweep out of two smaller ones.

This "lumpy" response can be fun! Maybe a couple of switched caps
to induce the range you like best.

> Second question.  The stages of the LM13600 datasheet state variable
> filter design look  like simple passive RC filters to me, with the OTA
> replacing the fixed resistor.  Is that an even approximately accurate
> way to look at the design?  So the design in the datasheet is two
> first-order filters in series, with some clever feedback paths to
> produce band pass and low pass outputs?

Yes. For a good example of this look at the Arp 3900 filter on the
cookbook page. You can see the RC passive without having to look
very hard at all.

> Lastly, like many others, I too had to abandon the idea of using the
> buffers included on the LM13700.  I found they caused a nasty clipping
> of the output signal at very low cutoff frequencies (even at fairly low
> amplitudes) and caused a very large DC offset as the cutoff is closed.

Lower the gain of the otas and see if this helps. Maybe 330K input 220-
ohms to gnd

> Replacing them with op amps did wonders!  I guess FETs work here as
> well? 

They do work but not as well as op-amps.

> I wonder why they didn't opt for a better quality buffer in the
> chip.  Maybe cost or technology restrictions at the time the chip was
> designed.  Oh well.

It's a general purpose chip and not really optimized for quality audio.
But it's way cheap and easy to use. IMO sound wise the 3280 
sounds best then the 3080 and finally the 13600 over the 13700. 
Funny that they go down in cost in that order isn't it?

Opamp buffers sound/work better than fets and fets are better than 
the on-chip buffer.

The 3280 and TLO72/74s make a very nice "clear" filter.

> I will post my schematic as soon as I have a chance to digitize it in
> case anyone wants to have a look.

Thanks Chris. You might want to look a John's new 9700 filter. It uses
13600/324s.  There is (was?) a link to the schematic. BTW I think there 
are some reject dual-filter pcbs for like $35. 

-tg