[sdiy] 2164 Expo VCO tracking

[cheater cheater]

Hi David,
for the results to be representative, I suggest using a tuned voltage
source, and doing the test procedure as follows:

-4, -3, -2, -1, 0, 1, 2, 3, 4, 5V

repeat:


-4, -3, -2, -1, 0, 1, 2, 3, 4, 5V

and again repeat


-4, -3, -2, -1, 0, 1, 2, 3, 4, 5V

As per Paul's suggestion, the idea is that to be significant the
measurements need to happen multiple times - if they are spread out
over time they will be less correlated (correlation happens because of
thermal inertia), and the procedure is easy enough to follow through
without thinking.

Doing this series thrice should give you enough information. You could
average them out, but that wouldn't be representative - I think the
average would still contain relative drift error between the items -
really doing interpolation is the best way in my opinion to get
scientific-level results.

Especially if someone starts testing those for production, it should
be easy to make a process which takes such three measurements while
recording the exact time, and then interpolates them with an
exponential curve fit. Then based on this the performance of the VCO
across the board *at the exact time instant* can be reasoned out.

(I am not sure if the exponential curve will be the best? Maybe a
quadratic curve? That's a question for people good with electronics,
so.. not me!)

If you want to make it easy on yourself and just average stuff out to
get some sort of result, maybe try doing this:


-4, -3, -2, -1, 0, 1, 2, 3, 4, 5V


5, 4, 3, 2, 1, 0, -1, -2, -3, -4V

testing the series in exact time intervals. It is only important that
you test the first series in a specific time interval (e.g. one
measurement every 3 seconds with a stopwatch), and then the second
series with that same time interval. If you put a small break between
those the two series the technique will still work.

This way the averages of measurements for each voltage would give you
the temperatures at a single time point, up to first order of
approximation. This means we are assuming that (temp(t_1) +
temp(t_2))/2 = temp((t_1 + t_2)/2) which is of course wrong but the
smaller the drift between the measurements the 'less wrong' this is.
So by looking at the difference between temp(t_1) and temp(t_2) you
can tell how exact the 'central temperature' prediction will be. This
is why a smaller break between the two series can be better.

HTH
D.

On Mon, Apr 12, 2010 at 08:56, David G. Dixon <dixon at interchange.ubc.ca> wrote:
> I spent much of today (when not watching the final round of the Masters)
> slaving over the latest incarnation of the 2164 Expo VCO.  I'm still not
> entirely happy with the tracking, but I would like something else to compare
> it to.  I can't seem to find any independent VCO tracking data on the web,
> except for the writeup of the MOTM 300 on the Synthtech website which
> alludes to 0.05% tracking over a very wide range.
>
> Here is the best tracking I've been able to achieve with my latest PCB, by
> messing around with a couple of resistor values (target values and relative
> errors in parentheses):
>
> -4V:    27.48 (   27.50  -0.07%)
> -3V:    55.00 (   55.00   0.00%)
> -2V:   110.0  (  110.0    0.00%)
> -1V:   220.0  (  220.0    0.00%)
>  0V:   440.0  (  440.0    0.00%)
>  1V:   880.8  (  880.0   +0.09%)
>  2V:  1760    ( 1760      0.00%)
>  3V:  3520    ( 3520      0.00%)
>  4V:  7025    ( 7040     -0.21%)
>  5V: 13755    (14080     -2.31%)
>
> Note that the VCO was set to oscillate at 440.0Hz at a CV of 0V by adjusting
> the panel pots, the CV summing amp feedback trimmer was adjusted to optimize
> LF tracking, and then the HF trimpot was adjusted to eliminate error at
> 3520Hz.  The octave voltages were generated by the six-octave selector
> switch on my custom scanning keyboard, which selects from a ladder of
> hand-matched 100k 1% resistors.  I went from the lowest A to the highest A
> on my keyboard above 880Hz and readjusted the 440.0Hz point to read the
> values above 440Hz.
>
> Obviously, the tracking falls apart much above 7kHz, but you can't really
> discern pitches above that.  The tracking is more or less perfect within the
> musically useful range, except for the octave around 1V, which is always a
> bit sharp (I don't know why).
>
> Tracking was near-perfect on the solderless breadboard up to 16kHz and
> beyond, but is less so on PCB.  I'm still trying to figure out why.  I think
> it must have been a fortuitous combination of stray capacitances, etc.
>
> So, would this be considered "good" tracking?  I presume it doesn't compare
> favourably with the MOTM 300, but I'm curious how it stacks up to other
> VCOs.  If anyone here has any tracking data of their own to share, I'd
> really love to see it.
>
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at dropmix.xs4all.nl
> http://dropmix.xs4all.nl/mailman/listinfo/synth-diy
>