F-V converter and Ring Mod. stuff

[gstopp at fibermux.com]

     On 2-16-96 Don Till (don at till.com at ccrelayout) wrote:
     
     >
     >Hmmm, I calculate maybe 10 bits for an octave, 16 for a six octave 
     >range if you're using an exponential lookup.  Maybe 13 bits if you 
     >use the exponential counter.  Add a few more bits if you value 
     >accuracy.
     >
     
     You're probably right Don - I stopped trying to figure it out before I 
     understood the solution. Sometimes ya gotta start down a road before 
     you decide it's too rocky...
     
     Regarding loop response - I sure know more about this now than I did 
     before, more from trial and error than from engineering. Designing a 
     PLL from discrete components is a real learning experience. A lot 
     depends on what you want out of the thing - as it happens I needed a 
     fairly ripple-free loop voltage because I want to use it as a CV for 
     external use.
     
     I discovered the following:
     
     1. This PLL tracks the best when there is no loop filter. That's 
     right, no frequency-dependent components at all - comparator output 
     tied to VCO input. The VCO pulse width becomes real narrow, but the 
     tracking extends as far as I can stretch it. There is no useable CV 
     from the loop filter in this case however - just a crazy-looking 
     pulse. The VCO triangle wave looks more like a hyperbola.
     
     2. When the loop filter consists of a simple R/C lowpass followed by a 
     buffer, the PLL will zoom to the lock point and then "ring" for a long 
     time around it like a spring ("booooiiiiing!"). Adding a resistor 
     between the cap and ground increases the damping and kills the 
     springey effect.
     
     3. It's better to let the loop filter output keep some ripple for the 
     drive to the VCO, then tap off of it and filter it some more to 
     provide the CV output for external use. If you squish all of the 
     ripple out of the correction voltage before it goes into the VCO, the 
     PLL becomes very sluggish and prone to locking onto subharmonics.
     
     I'm not sure but I suspect that a lot of this behaviour has to do with 
     the phase comparator design that I used. I have here a "Designer's 
     Guide" from Philips on the 74HCT4046 PLL chip, which is a whole book 
     dedicated to the 4046-type PLL, with lots of graphs and equations and 
     schematics of phase comparators. After browsing through this I went 
     ahead and ignored it and made what I thought would be a good phase 
     comparator out of 4 NAND gates (it's an edge-triggered set-reset 
     flip-flop). All I wanted was something that was:
     
     mostly low for F(ext) < F(vco)
     mostly high for F(ext) > F(vco)
     square wave for F(ext) = F(vco)
     
     The output of this flip-flop goes into an op-amp comparator that slams 
     between +12 and -12, to provide a correction voltage centered around 
     zero volts.
     
     Once again, I must admit that all of these tests were performed on the 
     bench, with a scope, and the "external signal" was a manually
     controlled oscillator. Therefore all of the tracking response findings 
     were based on the PLL tracking my hand turning a pot shaft. Real-live 
     signal tests may revise my circuit parameters. I'm not sure when I'll 
     be able to get to this...
     
     - Gene
     gstopp at fibermux.com


______________________________ Reply Separator _________________________________
Subject: Re: Re[2]: F-V converter and Ring Mod. stuff
Author:  don at till.com at ccrelayout
Date:    2/16/96 6:16 AM
     
     
I've wanted to try this out for a while, but never got to it.  It's 
cool that your giving it a go.
     
The major problem I see here is that, with an exponential VCO, the 
loop transfer function (the loop's response to a change in input 
frequency) changes drastically with the running frequency, to the 
point of ringing at some.  I don't know how much of a problem this is 
in real life.  And of course it can be compensated for too.
     
[Later... Wait a second, this effect might also be a feature.  I have 
to think about this some more.]