New idea for a Shepard Function Generator....

[Martin Czech]

Or another idea: use several ota integrators, Iabc for all otas is from
the same source (linear or exp source with resistive current splitter),
much like in ota state var filters.  One ota integrator makes a master
vco, the others are slave vco (waveform saw). The master hard syncs
all the slaves, there is a (window) comparator for each slave that
gives a sync signal.  All otas are overdriven, so no offset problem,
more or less ideal current mirrors. The slaves themselves don't need
any comparator. This way the phase shift should be very precise with
the expense of amplitude problems. But this could be minimised if you
measure the gm of every ota right in the circuit board with some jumpers
(dc circuit, recommended also for filters, gm can vary by a factor of 2),
and adapt the integration cap, this way obtaining more or less the same
time constant for each osc. OTOH one could also take otas right out of
the box and trim the amplitude for each slave. I'm pretty shure that
this will give a good saw shepard (without any artefacts).


By the way: what application do you need the shep gen for?
I only know the psychoacoustic ever rising glissando...

m.c.