AW: tapped 4pole filter

[jh]

>The OBsx just had 2-pole filters 

Oh, I didn't know that !

>Interestingly, the extra diodes, resistors, and extra trimpot that they put in
>their 2-pole feedback loop look a lot like what they had in the feedback loop
>of their old SEM state-variable filter (built around two 3080's).  

I'm speaking of the 2-pole mode of the OB-8 here, but as
you said, it's very similar to the OB-Xa VCF, so ...
What Oberheim did was a trick to use one of the OTA / Buffer
pairs of the 3320 as an opamp. Remember the SEM filter
is basically built around one opamp adder and two OTA / Buffer
pairs. Now with a feedback loop around one of the 4 OTA / Buffer
pairs in the 3320 it becomes an opamp. Only problem is that
this OTA is also connected to the filter CV, so the opamp
has an open loop gain that widely varies with cutoff frequency
of the filter. With a rather tight feedback around this "opamp"
the closed loop gain doesn't vary that much. I suspect it has
*some* effect, nevertheless, because in my multimode VCF modification for the OB-8 (see my homepage for details) the
High Pass (i.e.the "opamp" output of the filter) is not really
a perfect high pass. But it's close to a real HP - close enough
to consider this modification for any OB-8, IMO !

The OB-8 also had 2-pole/4-pole switching, but accomplished it with a
byzantine maze of analog switches built around its 3320 (seven separate
switching elements!), which basically allowed them to reconfigure it into
their OBxa 2-pole mode. 

"Byzantine Maze" hits the point. Seven switches look a bit
high from memory, but I haven't looked at the schemos for some time. 

I don't remember whether the two-pole mode allowed the filter to go
all the way into oscillation, but I do remember that it didn't sound so great
to me, except on string patches.

The OB-8 and OB-Xa have a "problem" around their filters where you would not expect it. It has to do with such a simple thing as filter input level of the VCO waveforms.
The OB-8 has the 2-pole filter slightly harder driven than the OB-Xa. This is fine for *most* patches, providing the maximum of SNR. But You will get a bit of unpleasant distortion with triangle waves and a high resonance that hits the fundamental of the triangle, for example. Another side effect is that if you switch from 2-pole to 4-pole, the 4-pole is not as loud as the 2-pole (especially with high
amount of resonance), so you get the *impression* that the 4-pole filter is "weak". Which it is not really, of course.

On the OB-Xa the levels are different, so you might get the impression of a "weak" 2-pole filter. Which it is not really, either.
If there is a design flaw in these VCFs (other than the lack of VCO
level VCAs in general), then it's the chosen input level of the OB-8's filter in 2-pole mode.
I have changed a few resistor values in my OB-8, and the effect is
stunning: The 2-pole becomes softer (but a little bit more noisy, i.e. more in the SEM direction ...), and switching from 2-pole to 4-pole gives that "growing fat" effect again.

JH.