New VCO,linearity!

[René Schmitz]

At 08:10 22.05.00 -0600, Ian Fritz wrote:

>Again -- personal choice. Some people have been happy using a differential
>pair converter without the compensation for the 1/T in the exponent. This is
>3300ppm/K. If you just throw in a tempco resistor this drops to the range of
>about 500-1000ppm/K, since the tempco is not exact compensation without
>trimming and because there are other sources of drift in the circuitry. Many
>people live with this level of drift, although it requires frequent
>retuning. I've been trying to do better mainly because it's an interesting
>technical challenge. It doesn't take too much work to get down to the
>200-400ppm/K level, but it takes some time and careful measurements.

I'd say we need to differentiate between different sources of "drift" here.

The first is the temperature dependant scale factor of the expo circuit.
In its uncompensated version it changes ~3300ppm/K, messing up your V/oct
setting, so any octave would become stretched/squashed by 0.33%/K. This can
be almost perfectly reduced by using a tempco (or some other means to
change the input to the expo). Having 1000ppm of drift in the expo
convertor is unacceptable IMO.
I think one can get this below the 100ppm margin with ease.

The second (and mostly never discussed) source is the drift of the current
controlled oscillator that inevitably follows the expo convertor. Here the
effect is to shift the oscillators frequency, while not affecting the V/oct
setting. So 1V stays an octave. Maybe you had tuned it to 440Hz and it
drifted to 480Hz, but the octave above that is 960Hz. So all you have to do
is to tune it again. Here 500-1000ppm are not that bad, all it requires is
to retune the instrument when it has warmed up, and a roughly constant
ambient temperature.

Bye,
 René




-- 
my keyboard gets   | uzs159 at uni-bonn.de
flakey so here are | http://www.uni-bonn.de/~uzs159
the missing "M"s:  | http://members.xoom.com/Rene_Schmitz
mmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm