[sdiy] more pitch shifter thoughts

[Magnus Danielson]

From: Scott Gravenhorst <music.maker at gte.net>
Subject: Re: [sdiy] more pitch shifter thoughts
Date: Mon, 10 Jan 2005 10:00:48 -0700
Message-ID: <200501101800.j0AI0mV10516 at linux6.lan>

> Grant Richter <grichter at asapnet.net> wrote:
> >You can build one with just a top octave chip and a CD4046. The TOC just
> >need a high frequency clock phase locked to a reference VCO square wave. A
> >switch can select the high or low C from the TOC and give and octave above
> >or an octave below the reference VCO. (IIRC with a 50240)
> 
> Being a 4046 PLL kind of guy, I have wondered about stability and tracking
> when doing this.  I have a PLL module that uses a divide by 12 counter in
> the loop and it works just fine.  But the MK50240 has a range from divide
> by 239 to divide by 478.  
> 
> What kind of problems, if any, will there be with such large counters?  

Larger dividing ratios will divide the loop gain by the same amount. You need
to take that in count.

> I would think that proper VCO capacitor selection is important.

Indeed it will. The loop-gain being different for different ratios (/1 being
refernce for "normal" cases) and thus you end up with basically two troubles,
that of changed steady-state stability point and that of changed Q-value which
potentially renders the responce very resonant and thus unstable, beside having
it providing phase-noise gain at the PLL bandwidth, which is not a good thing
most of the time.

> Over/undershoot?  

Yes.

> Does a large counter affect lag very much?  

Yes.

> My own circuit has adjustable damping for the loop filter and allows me to
> adjust it so that frequency changes are so immediate that I do not hear any
> over/undershoot or lag.  But then, my PLL has only a divide by 12 counter
> in the loop.

That is a good move. However, by spending a little care will an octave of range
not be as much of a trouble. The trick is to make the loop-gain change occur
over a well damped case, so it's loosest end is more or less critically damped.
It is a bit of numerics but it is fairly simple to do.

There is a bunch of different sources for the calculation, which one are you
using?

I've been doing multiplications beyond 2000 with no problems, but that was no
4046 in there. Infact, for some uses the 4046 is so darn bad, and especially
the charge-pump buissness has better reputation than it should have. Too many
use it blindly without understanding the issues. Let's say I'm not impressed
when seeing it in action in what proved to be a critical application, where as
using a much simpler setup locked and just kept going without headaching
trimings and problems.

OUPS! Now I am 4046-bashin again!

Cheers,
Magnus