more VCO cores

[Martin Czech]

:::This  all sounds plausible.  I'm not sure how you could check which of these
:::are occurring and to what extent without having probes within the IC.  And
:::I'm not sure I see how the mistracking would change sign as a function of
:::drive level.

Ohh, voltage probes can give you a hint, current measurements are not
really possible...  On the IC level we're really blind, I mean we can
make a hypothesis do some experiments, try an improvement in the next
tape, and so on, it is more Sherlock Holmes then God. In discrete land
you can get hold on any (external) node voltage and current, you know
everything, the designer is omniscient (God).

But that's part of the fun ;->
I can't wait till we get down to 0.25u technology with analog circuits...
5 metall layers above... debugging will be much fun...


No, really, one could try to measure the error at low and high frequencies,
using a much larger integration cap, so only leakage and mirror errors
will have an effect. Effects 1 and 2 steal current, make the circuit slower.
Effect 3 will influence symmetry (slower?).

So the test would be:

Measure frequency for every octave (wide range, 10 or so).
Change ota overdrive, measure again.

We assume that a certain overdrive level is standard, and normalize
to the respective curve.

Display the deviation to the standard case in a table,
rows are overdrive level, columns are frequency.
Or do a graph.

Measure the triangle symmetry as well (i.e. square output
low time, high time). Most frequency counters can evaluate
time intervals, too.


Of course, heating, opening the window etc. will give thermal
effects that may make the whole measurement a nonsense.

The Iabc should come out of a precision current source,
a resistor, a MOSFET and opamp (low Vos-drift).
Now you may know why I tried to get 1 1000 Megohm resistor? :->>




:::Or maybe because of the exponential response the high side of the excursion
:::has a larger effect than the low?

Yes, that's what I mean with rectifying, a dc offset will remain.
As long as the junctions don't cross the conduction/leakage
border this is the only rectifying way (it's a small signal).

:::
:::> I always though that oscillation is critical with the feedback
:::transistor,
:::> so it is compensated with a cap across emitter/collector. This
:::> should then happen when the transimpedance of the tranny is high.
:::>
:::> Now it seems to me that oscillation comes when the transimpedance is low,
:::> it is the opamp output stage that is loaded...
:::>
:::
:::Let's see ... you need more compensation when the gain is low, right?

As I said, this depends on the effect, if it is sluggish feedback,
or loading the op amp....

Any experts?


m.c.