Moog 904A lowpass filter Question

[Terry Michaels]

Message text written by "Colin Fraser"
>----- Original Message -----
From: "jh." <jhaible at t-online.de>
To: "Gerry Dahl" <gerryd at ntx-tech.com>; "Synth DIY"
<synth-diy at node12b53.a2000.nl>
Sent: Sunday, January 07, 2001 10:34 AM
Subject: Re: Moog 904A lowpass filter Question

> Apparently Moog used slightly different buffer amps
> for each VCF they made, and these will most likely affect the sound
> way more than all the above points.

Some time ago AHer Michael Caloroso put together a list of most variations
on the ladder Moog built, in order of his preference, and compared the
buffer amp design, and resonance feedback method (ac or dc coupled).

You'll find his mail in the AH archives at midiwall -
http://www.midiwall.com/archives/ah/lookit.cgi?-v1997-10.121

Very useful reading if you're interested in the differences between Moog
ladders.

Certainly in my experience, the filter in my mini, Prodigy and the Source's
I've borrowed or repaired all sound noticeably different.

Colin f<


I recently got around to building a Moog ladder VCF, utilizing the 904-A
circuitry, but with some updates.  The primary change I made was the buffer
at the top of the ladder.  Instead of the AC coupled descrete design in the
904-A, I used a 3 op amp instrumentation design like that shown on page 425
of "The Art Of Electronics".  

I initially left out the 4 capacitors in the ladder to run some tests of
the overall transfer characteristics.  It is well known that the Moog
ladder filter is somewhat non-linear, which no doubt is responsible for a
part of its "charm".  Because the audio is fed into the differential pair
at the bottom of the ladder, I expected the transfer characteristic to show
a flattening out at higher signal levels, the same principle that is used
in some triangle to sine convertors.  However, it was linear all the way up
to the clipping point.  After thinking about this, it occured to me that
the signal at the top of the ladder is read out across the base emitter
junctions of the transistor pair, these junctions have a logarithmic
characteristic which cancels out the exponential characteristic of the pair
at the bottom, hence a linear overall response.  

I then installed the ladder capacitors.  The ladder was still linear for
signals at all amplitudes well below its cutoff point, but when I ran a
sine wave through and swept the filter cutoff around the sine frequency, I
could clearly see distortion of the sine wave near the cutoff point.  The
distortion is very dependent on input signal level.   I'm guessing this is
because the caps are within the non linear part of the ladder, before the
log response at the top of the ladder.

Finally I added a descrete output to the top of the ladder using the
original 904-A design, so I could compare the response of the op amp and
descrete buffers.  The descrete buffer DOES show the classic flattening out
of signals well below cutoff (or with the ladder caps removed).  This is to
be expected, because the descrete output buffer is the classic differential
pair, with resistive networks on the collectors, which saturates at higher
signal levels.

Conclusion, there are at least two non-linear components to the transfer
characteristic of the Moog ladder filter.  It would be interesting if
someone could accurately model this with a Spice simulation.  

Terry Michaels