[sdiy] Temperature Compensated VCO attempt - help?

[Magnus Danielson]

From: Don Tillman <don at till.com>
Subject: Re: [sdiy] Temperature Compensated VCO attempt - help?
Date: Sun, 2 Feb 2003 02:04:00 -0800

Don,

>    > I've made an attempt at an electronically temperature compensated
>    > exponential generator and VCO. I built it into a thru-zero FM VCO based
>    > on an Electronotes design. 
> 
> I'm really curious how the thru-zero VCO works?  (Perhaps I may have
> to break down and finally buy a copy of Electronotes.)

Look at Jürgen Haibles frequency shifter. Handdrawn schematics, but they
convey the construction well ;)

Basiclly, start of with a triangle oscillator core and when you go though zero
you reverse direction by chaning load current direction. This can be done in
several different ways, but for a triangle oscillator core you allready have an
OTA to do the current direction anyway, so that part comes for free.

A method that I haven't seen, but which should come very natural is actually to
use a full Gilbert cell so that one input is the normal triangle sign swapping
and the other is the direction (pos/neg frequency) sign swapping.

>    > The VCO works, but the exponential generator suffers from
>    > linearity problems. 
>    > http://home.attbi.com/~sbernardi/elec/og2/tempcomp_vco_theory.htm
> 
> (Wow, that's a complicated exp amp!)

It has raised the complexity level yes, but the prize also has an interesting
gain attached to it. If you start to compare with all the waveshaper stuff that
comes along with it, I don't think it raised the complexity THAT much.

> I would think that the linearity of the OTA would be a major issue.
> OTA's aren't especially linear... in the sense that they make great
> triangle to sine converters.

This is why Scott and Jim used two different forms of linarizations. Scott used
the builtin linearization diodes where as Jim used the other half of a LM13600
to do the linearization. The tanh non-linearity is due to the input diffpair
exponential properties, properly setup linearazing diodes will effectively do
an arctanh distorsion such that they cancel into a semi-linear expression of
Iout = Iabc * Id / Ic where Id is the diffrential current on the inputs and
Ic is the common mode diffrential current on the inputs. For the compensation
scheme we are talking about Ic should be a sufficiently large DC current and
Iabc and Id should be the kT/q and input signals respectively.

There is a scale issue in the quality of the OTA current mirrors. The LM13600
has pretty good current mirrors. However, if your looking at 1.7 ppm/C there
are many other places to look at.

Cheers,
Magnus