[sdiy] chaos update

[Ian Fritz]

At 05:41 PM 4/9/05, you wrote:

>I've often wondered about sensitivity to CV feedthrough/offset when making 
>these things VC'able.
>Some parameters in the xfer function I imagine are sensitive enough...
>
>Oren

Oren --

Ah, you are very astute.  The offset problem is a bear!  Some of the orbits 
are very sensitive to offset.  Here is what I ended up doing about the problem:

1) I selected through a dozen or so CA3280s to find the pairs with the 
lowest offset.  Some of these devices were very good, indeed.

2) I made a partial low-end compensation by feeding a voltage from the expo 
converter for the frequency control through a series diode-resistor 
combination into the offset-control node.  This extended the acceptable 
operating range by a factor of 2-3.

3) I put in a range switch to allow two capacitor values to be used (180 p 
and .047 u).  This allows a good frequency range while avoiding extreme 
Iabc currents.

With these methods, simple, robust patterns are quite stable over the whole 
frequency range.  But patterns such as narrow periodic windows will be lost 
with changes of a few octaves.  It will never be possible to totally solve 
this problem, as some patterns are extremely sensitive to loop offset and gain.

Another problem I had to attack was the common problem of Q enhancement at 
high frequencies.  Because of all the amps in the loop, this is a more 
serious problem than for most VCF designs.  To alleviate this problem I 
used high-performance op amps (OPA2134 or similar) throughout and added 
small compensating caps (47 p) across all the OTA input resistors (8 
caps).  Fortunately, with good op amps this is a significant problem only 
at very high frequencies, so this approximate compensation scheme is good 
enough.

   Ian