New idea for a Shepard Function Generator....

[Magnus Danielson]

>>>>> "CC" == Chris Crosskey <chrisc at zetnet.co.uk> writes:

 CC> Hi Folks....
 CC> I've had a sort of idea, and I've talked it over with JH and he seems to think 
 CC> it's sane........

 CC> Basicaly the problem with Shep Functions is you've either got to build a true 
 CC> octature oscillator (pretty damn hard) or you've got to build seven 45 degree 
 CC> phase shifters that work over a decent frequency range (pretty damn hard as 
 CC> well)....the problem is the 45 degree shift.... well I can guarantee and 
 CC> absolutely locked 45 degree shift....#

 CC> Basically my idea consists of a dual pot of about a meg or so with one pot in 
 CC> the frequency control of a 555 timer circuit. This feeds a 4022 divide by 8 
 CC> counter, if the 555 is running at 8x the speed you want for the Shep function 
 CC> then you've got 8 outputs 45 degrees apart at the Shep speed......
 CC> These eight outputs act as resets on eight separate integrators being fed 
 CC> current from buffers running off the other part of the dual pot. If the pot is a 
 CC> Meg and you put 15V into one side and a 10K resistor between the other side and 
 CC> ground then the output (centre) sweeps between 0.15V and 15V in a basically 
 CC> loinear fashion. This ensures that as the clock speed of the 555 goes up 
 CC> (basically linear if all you want is 100:1) then the current being seen by each 
 CC> integrator goes up, therfore the reset voltage on the integrators stays 
 CC> constant......try to match the caps in the integrators and trim it in on the 
 CC> resistance. The voltage at output of the integrators will be pretty low so amp 
 CC> up to 8V P-P (probably best off trimming here rather than at the integrator 
 CC> iteself) .....Then you need to extract a 0-10V triangle wave from the sawteeth 
 CC> and you have a Shep generator...

 CC> OK bad points....no VC, good points no insanely matched components. I make it 
 CC> one 555 a 4022, eight TLO74's and a bunch of R's and C's.....
 CC> If you want VC then you'll need to give it an input for a VCO and one for the 
 CC> control voltage for the VCO, you'll need a clock divider for the VCO and an expo 
 CC> converter built into the integrator stage, however you'll only need one expo 
 CC> converter but with eight outputs, so that's nine matched transistors.....
 CC> Anybody willing to run with this, I'm busy ATM so don't hold your breath waiting 
 CC> for my version.....remember where you heard it first......

All this rings a familiar bell, it sounds like something similar to
the Paia Shepard kit. Just looking at the theory description of it is
quite enlightening. The Shepard kit has CV and envelope out for 8
VCO/VCA stacks.

The Paia kit will do a similar arthmetic in the digital domain but
messing around with the upper bits of the counter in order to generate
the 16 diffrent versions.

I have been thinking about some voltage tricks to achieve similar
effects using just op-amp cursuits from a common oscillator.
The benefit of such a cursuit would be CV-in/CV-out. Look at it as a
saw-wave VCO followed by a custom waveshaper. A gut feeling says that
it can be done but I also think that some minor CMOS coulp help the
sequencing and thus propably reduce the complexity.

I think that your cursuit probably could benefit from taking a good
VCO instead of the 555 to benefit the CV properties. Actually, when
you consider the full complexity of the Shepard generator in action I
think you can motivate the better oscillator.

Cheers,
Magnus