SV: Re: [sdiy] PAIA VCO, LM13700 diode temp comp.

[Ian Fritz]

Hi Karl (inline) --

At 06:19 PM 10/18/05, karl dalen wrote:

> > I built a similar VCO a few months ago, as an aux VCO to use in my chaos
> > box.  My main trouble with this circuit involved the V/Oct 
> tracking.  After
> > much experimenting and tweaking the best tracking I got was +/- 0.5% from
> > 100 Hz to 6 kHz.  Nothing even close to the 2 cents accuracy claimed in 
> the
> > writeup.  I didn't check the temperature drift.  It shouldn't be too bad,
> > but I know from another active-compensation circuit that you need good
> > matching of the OTA pair.
>
>Wich would be a pain! I have never found any pair that have matched each
>other!!

OTA performance really seems to be a crap shot.  I have a batch of 13700N's 
that almost all have low offset change with Iabc, and some of these are 
also matched in this characteristic.  (Of course, the overall offset still 
has to be trimmed.)

You might be able to get decent temperature stability, but again my main 
problem with this circuit is the rather poor tracking.  And BTW, I ran the 
converter into a high-performance Saw oscillator core, to be sure the 
problem was in the converter circuit.

>BTW, there was previous discussion on the CV offset trim methods
>the usless Serge - volt trim, and the typical National trim pot
>app note, but i just realized i havent done the:
>"Ground one of the inputs (no 330 resistor to gnd "just to gnd")
>and apply a balancing voltage to the active input pin trough a resistor!!
>Wonder if that would improve the CV feed trough problem a tiny bit?

I'm not familiar with this Serge method, but I do know from his ENotes 
writeup that he studied the various OTA's in great depth.

As far as the best trimming method, I would stick with the (symmetrical) 
method of the manufacturers, although grounding one input directly seems to 
work fine most of the time.  The CV feedthrough problem can only be solved 
by selecting devices for low variation of offset with Iabc, at least as I 
have been able to figure out.

Best regards,

   Ian