[sdiy] Temperature Compensated VCO attempt - help?

[Magnus Danielson]

From: Scott Bernardi <sbernardi at attbi.com>
Subject: Re: [sdiy] Temperature Compensated VCO attempt - help?
Date: Sat, 01 Feb 2003 16:26:22 -0800

> Ah, I must correct myself. It is the CA3280 which has the Id current
> doubled. However, Serge also went into analyzing the LM13600, which he said
> showed more distortion as you approached the maximum input signal levels
> than the CA3280 counterpart, especially for the single ended input as shown
> in the National datasheet (and which my circuit used).
> Magnus, this also falls into your discussion about common mode errors with
> the National scheme. Jim Patchell's configuration with the two LM13600's
> doesn't have this problem.

Well, looking at Jim's schematics I realize that I need to exercise my paper
and pen some more. However, my conclusions where that with a differential feed
into a linearized OTA (assuming a perfect current mirror, which doesn't change
with temperature) I had a perfect cancelation on paper. The differance and
sum of input currents annoyed me when I hit the lamp goodnight but the morning
shower cleared the toughts on that. The OTA I used for my figuring is the
minimal 4-transistor thinge with two additional transistors for input
linearization. The output feed to a virtual ground IV converter (resistor over
an opamp) and then into the traditional diffpair. That would mean that
8 transistors and 4 op-amps would build you one of these thingies. In the ideal
case would the linearizing transistors, the OTA diffpair and the expo diffpair
all be on the same piece of silicon. If the current mirror is crammed in there
that would be a big plus, but it would not be as sever if they where not
thermically tight to the other pairs, but they must be tight to each other.

Hmm... how about an analog full-custom chip anyone? ;O)
I know where one can get them made... ;O)
But I think we have a few guys from that industry online here... ;O)

<SNAP!>

Oh, sorry... day.. eh... night... eh... late evening dreaming... so sorry!

Should be possible to achieve. A pair of MAT-04s glued together should be
pretty good. The MAT-02 is better for the diff pairs, but a PNP pair is needed
for the current mirror anyway. For good compensation lack of thermic connection
between the parts can be devistating or at least what sets the thing back.

Things doesn't have to be heated, but tightly connect and away from heat
sources and flows which causes big thermic flow vectors trough the transistors
such that they are at different temperatures.

I've got myself a pair of ovenized chrystal oscillators (oh, three when I come
to think of it, but currently only two of them heated) so I think I've learned
some of the story by now... ;O)

Cheers,
Magnus - has my temperature stabilized 6,834682608 GHz reference heated