AW: About VCF's

Haible_Juergen#Tel2743 HJ2743 at
Wed Apr 10 21:36:00 CEST 1996

Ok, I must respond to this ...

>     I'd like to know if anybody on the list has ever built a 4-pole
>     lowpass VCF using four identical 3080 OTA integrator stages, like the
>     4-pole from the Electronotes ENS-73 series (right around issue #41).
>  * How did it sound?

I have not built it myself, but a 4pole filter based on 3080's is part of 
elektor FORMANT synthesizer. It sounds very good. A  friend of mine
has one, but I have not done an a/b check with a 2040 filter.
It surely sounds fatter than a CEM.

>     * How did it handle distortion?

I think that the FORMANT is designed in a quite conservative way,
i.e. the resistor deviders for the OTAs are much on the save side.
*if* there is some overload, it will be symmetric, and only in the
first filter stage.

>     Also, how about the same filter type, except using better OTA's like
>     the 13600 or 3280? Anything sound like an SSM2040?

I use 13600's (and 13700's) all the way, being in the impression that
the mere OTA stage is identical to the 3080. But I am not sure about
this anymore. Maybe the 3080 has different current mirrors and stuff,
who knows?
I have built a 6-pole filter with 13700's. (Filter/Phaser with caps 
from gnd to input of each stage). It does sound good, but not at all "warm"
like a 2040. (But then again, I only have heard 4-pole VCFs and
4-stage phasers with the 2040)

>     Sadly the SSM2040 is a scarce part. When I built one of those I was
>     quite surprised at the warmth of it. I've been reading the thread on
>     the "Secrets of the 2040", and I wonder if rather than trying to bias
>     a CEM3320, if a discrete design has ever been found that sounds as
>     good (except maybe an original Moog ladder).

I am quite sure about the internal circuit of the 2040 now. The OTA
is just a differential pair and one current mirror. Compared to the
3080, they save two current mirrors. The OTA output is just the
collector of one transistor of the differential pair. (Similar as in
the 2024 quad VCA).
Maybe some of its "warmth" has to do with the fact that there are very
few pn junctions in the signal path, compared with a 3080.
And most important, the stages clip very assymmetrivally. The voltage
swing at the OTA outputs is limited to about -200mV ... +14V with a
15V supply. The buffers are some kind of npn darlington circuits that
shift the OTA outputs up by approx. 1.2V in a feedback circuit like
a stage in a 4pole LP VCF. This way you get approx. -1.4 V ... 12.8V
at the buffer outputs.
If you want to hear the effects of warmness without the clipping, try
a rev. 2 Prophet 5. There the input signal is limited by a resistor that
prevents the filter from clipping at all. (Stages are 10k/200R feedback,
input resistor is 200k)
In my modular system, I have a passive input adder with 10k pots
and 18k summing resistors (also with a 10k/200R feedback in the
filter). This allows heavy overload of the filter stages themselves
and adds a different colour to the warm 2040 sound.

>     Also, I've sent an email to Analog Devices about the availability of
>     the SSM 2044/2045/2047, no response yet. Their Designer's Reference
>     databook says that these are not in normal production but are still
>     available, and I've never tried them. Are they as wonderful as the
>     2040?

The 2044 is an integrated ladder (!) filter. Main difference to the
(semi-) integrated (CA3046) Moog  and ARP designs is that the
2044 uses a "folded" cascade. Instead of the Moog-type transistor
pairs a current mirror is used. The behaviour should be quite
similar: Input resistance is that of a pn junction, output resistance
is very high. IMMO (in my momentary opinion - but I am willing to
learn!) there should not be any advantage over the Moog cascade;
it looks like a very clever means to work around the Moog patent.
(The patent for the folded cascade is held by Dave Rossum).
There are special input and output stages in the 2044 as well,
but I don't recall it at the moment.
The sound of the 2044 was best described by someone on the AH
list as "woody". That's exactly what it is, IMO. Very smooth. If
you have heard a Xylophone sound from a Polysix you know what
I mean.
I am not sure about the 2045 and 2047, but I *think* they have some
additional gm stages to compensate for bass loss at high resonance.
(Increasing signal gain with increasing resonance.)
Haven't heard one of these so far.

Speaking of emulating the 2040 by a level shift in the CEM3320,
I don't think I can hold on this any longer. This was an idea of mine
when I thought that the assymmetrical clipping was the main factor
in the 2040. But this cannot be, because old Prophets have the warmth
without the clipping. I think the best means to emulate the 2040
would be building it from discrete transistors. (Joachim's idea.)
Joachim might also be able to tell us a lot about the Curtis ICs
in the near future ...


PS.: As I said, I have given up the tought of doing level shifts on the
CEM 3320's in my Prophet. Now I think about something different:
Build some replacement modules that plug into the 3320 sockets
and contain real SSM2040's (yes!), 2020's for the resonance and
resistor networks that kind of "override" the ones in the Prophet.
I have not checked in detail if this is possible, but chances are good,
as the resistors for the 2040 are all in a lower range than the ones for
the 3320.
Well, You must think I am mad. (;->)
And no, I am *not* thinking about a 3340 vs. 2030 swap (;->) (;->)

PPS.: I have spice data available for a 2040-like filter built of discrete
transistors (I used BC550's for the npns). If anybody is interested, tell 

More information about the Synth-diy mailing list