Phasers - if you don't have enough already

[jhaible]

Ok - phasers once more.

With all this talk about the SSM 2024 and the SSM2164 im coming back
to an all pass stage that I've proposed a long time ago on synth-diy.

You're all familiar with the OTA all pass stage as found in the EH Small
Stone,
Moog 12 Stage, and probably many other phase shifters. All these circuits
work
fine with 3080-type OTAs which allow a large output voltage swing, but they
fail to work with the above mentioned SSM VCAs that need a virtual GND node
at their outputs.

The problem can be solved with a little more components: One OTA + 2 OpAmps
for each all pass stage. The idea is to split the all pass function into a
low pass (LP)
function and a second, auxiliary function. The maths go like this:

What we want is:  AP(s) = (1-s) / (1+s).

What we can easily build with a 2024-type ota abd an opamp integrator
is an inverting low pass: LP(s) = -1 / (1+s)

Now we do some calculations:
 
AP(s) = ( -2 + 1 + s) / (1+s) 
         = -( 2 * ( -1 / (1+s) ) + 1 ) 
         = - ( 2 * LP(s) + 1)

This is a function we can easily implement: Build a LP from ota and opamp
1,
and build an inverting adder from opamp 2, with 2 inputs weighted 2:1.

I have not really used this circuit so far, because of the additional opamp
that is required.
But now I have found two new arguments for this method:

(1) You probably know the switchable all pass / low pass circuit from
Electronotes.
     It allows mixed configurations of low pass and all pass stages, which
result in
     some very "musical" transfer functions. The method there (and like
I've implemented
     it in my filter /phaser, too) is switching a capacitor from input (all
pass) to gnd
     (low pass). This requires shielded cables, and if you want to swich
electronically,
     the resistance of CMOS switches might cause non ideal behaviour.
     With my new method, you have the low pass function separate to the all
pass
     function, and you can switch between two buffered opamp outputs. I
think the
     extra opamps are better than shielded cables for sensitive signals.

(2) I am even thinking about an all CMOS phaser solution. Don Tillman
showed us how
     to use a 4069 as a hex VC resistor, and built a phaser from that and
some opamps. (?)
     But what if we want to replace the opamps with 4069's, too ? Similar
problem as
     with the SSM chips. You can build nice inverting linear amplifiers
from the unbuffered
     gates, but the second (positive) input of a real opamp is missing.
     But a similar trick as shown above will work. Build a passive CR high
pass this time,
     with the R going to GND and being part of a 4069 without VCC
connected. Then 
     invert the HP output with one normally powered 4069 gate, and build a
weighted
     adder with another 4069 gate.

I have tested this stuff with PSpice simulation, but haven't built anything
yet. (Sequencer
has priority. (;->) ) IMO, the method could be highly useful for (1).
Number (2) might be
more of a fun project only, but somehow I like the idea of an "all digital
chip purely
analogue" FX device.

What do you think ?

JH.

PS.: I hope the description was fairly understandable. I've also drawn a
little picture, but
I won't publish it on my web page in its current, untested state. Mail me
if you want a
copy.