[sdiy] OTA performance (was SSM chip reissue)

[Mattias Rickardsson]

On 26 April 2017 at 11:51, Tom Wiltshire <tom at electricdruid.net> wrote:
> On 26 Apr 2017, at 09:06, Roman Sowa <modular at go2.pl> wrote:
>
>> At low cut off frequency, when integrator currents are really small, Zeners are damping the oscillation, because most of the current is eaten up by the zener, not the cap - at that small current their voltage is way much smaller than, say, at 1mA, the top of VCF frequency.
>
> I've got some evidence this is true. I built an 2164 SVF and swapped the 30/33K input resistors for 15K, thereby doubling the currents. This improved the situation somewhat, but it's only a octave's worth of improvement. But running the filter "hot" helps.

I just swapped the 30k resistors to 15k for another reason: to double
the signal level and get the noise down almost by half. If it also
reduces any effect of zener leakage it's a nice bonus! :-)

Also, I guess raising the gain (with lowered resistors or increased
VCA gain) would imply that the integrator capacitor values be
increased accordingly - which would also reduce the leakage problem,
right?

But note that 15k is too low input resistance for signals above 7 V.
Clipping might occur, and in the 2164 clipping isn't something you'd
want. The SSM2164 datasheet uses 30k for maxing out the signal with
+/- 15 V supplies, but with full signal swing using 15k would only be
safe with +/- 8 V supplies.

Speaking of "running the filter hot"... I usually just use the
"attenuation side" of the 2164 CV in SVFs, letting a CV of 0 V (which
gives 0 dB gain) set the maximum frequency and then the audio range is
covered by positive CVs of roughly 0-2 V (giving lower gains of 0 dB
to -60 dB). Would it be worthwhile to use some of the "gain side" of
the 2164 as well, like perhaps 15 dB to -45 dB, bringing the
well-speced optimum 0 dB point into the audio band and avoiding the
most imprecise high-attenuation regions? I tend to avoid it in filters
due to the rather uncontrolled THD that slowly grows on the "gain
side"... I get a bad feeling allowing this to happen in the highest
frequency region, but how far do you guys usually go?

On 26 April 2017 at 04:11, David G Dixon <dixon at mail.ubc.ca> wrote:
>> From:  Andrew Simper
>> positive feedback around the bandpass integrator to
>> push the damping more negative and can easily push the
>> circuit into self oscillation, which makes it just like any
>> other topology where you can increase the feedback past the
>> point of self oscillation with more gain.
>
> Yes, exactly.  A largish resistor between the BP output and the input
> ensures robust oscillation at all frequencies.  Without that resistor, just
> hitting the "theoretical" level of damping required for oscillation will not
> guarantee oscillation.

In general, how largish would this largish resistor typically be? :-)

> A couple of zeners across the BP cap ensures relatively clean sine waves.
> However, I still haven't figured out how to get the sine waves to have
> uniform amplitude at all frequencies.  One would think that the zeners would
> ensure this, but they don't.

I've been puzzled too about how different the levels can get.

/mr