Variable ramp again

[jhaible]

Haven't thought about your proposal in depth yet (but I'm looking forward
to see the schemos !).
Nevertheless, if your remaining problem is just that of a tiny offset
voltage
that will build up in the integrator over time, we can probably asume that
this building up happens quite slowly, compared to the actual waveform
genereation. So I'd say what you need is a *slow* control loop that will
correct for the offset at the integrator input. In other words, removing the
(or part of the) DC component at the integrator output.
How can this be done ? Removing DC means AC coupling in the signal path
(which is a no go here, of course), or an integrating feedback negative
loop.
Can't tell you much about stability of such a circuit without seeing the
implementation. But even if there would be stability problems (the loop
comprises two integrators, after all !), one might go for a certain
"controlled instability". In German, that would be a "Zweipunkt-Integrator"
("two point integrator" - but I doubt this is the right technical term in
English).
The idea is to monitor the DC offset of your output signal (RC lowpass with
long time constant), and determine when its lower then a first threshold
(say,
-10mV) or higher than a second threshold (say, +10mV). As long as the
RC-filtered integrator output is between -10mV and +10mV, do nothing.
If it goes below -10mV, send a very small offset current into the integrator
that will slightly push it into the positive direction. Must be so small
that
the result is a very slow integrator drift - just a little faster than then
your
offset error would cause an integrator error in the opposite direction.
Same thing if your RC-filtered output goes above +10mV - just send a current
of the opposite direction into the integrator.
This method will result in a controlled instability", i.e. the whole system
does oscillate, but the amplitude of this self oscillation signal (that's
added
to your desired signal) is low (you choose it with the thresholds you
choose,
+/-10mV in the example), and it is way below audio (you choose it with
your time constants).
Probably you can find a real stable solution as well. Just in case you
don't,
there is always this "two point" method. You might even use a sample & hold
stage to sync the correction currents to your VCO frequency (this would
result
in a slightly imperfect waveform, but no independent oscillation).

Hope this is of some use,

JH.



----- Original Message -----
From: <mbartkow at ET.PUT.Poznan.PL>
To: <synth-diy at mailhost.bpa.nl>
Sent: Sunday, August 01, 1999 3:29 PM
Subject: Variable ramp again


> Dear list,
>
> Few weeks ago there was a discussion on variable ramp generator I had
> needed for my own project. In fact, I needed even more than adjustable
> triangle/saw wave - I wanted to have it voltage controlled.
> The conclusion was that it's better to built a kind of wave shaper than
> to affect the charge/discharge time ratio within a VCO.
>
> Many nice ideas came, I especially appreciated the Don Tillmann/Martin
Czech
> approach to appropriatelly scale two ramps going in oposite directions and
> take the min of the two as the output. The problem was with the scalling,
> as it required some nonlinear relation between two scaling factors. I have
> built such circuit which almost works, but it is very complicated and I
thought
> about an alternative approach and at last I got a simple idea. As many
ideas
> this one would never come out without the stimulating lecture of this very
> list, and for this particular one I owe thanks to Juergen who mentioned
> integrating the pulse output of a regular VCO.
>
> The idea follows:
>
> Take the saw wave from the VCO and compare with an adjustable threshold
(or
> a voltage input for VC ramp). At the output of the comparator one has a
variable
> width pulse train. Now, remove the DC offset (which is equal to the
threshold
> voltage BTW) and input it to an integrator. A variable ramp signal is
observed at
> the output of the integrator with the symmetry (ramp ratio) dependent on
the
> voltage threshold. In order to prevent the amplitude of the signal from
changing
> with frequency, the pulse train may be appropriately scaled which is very
easy-
> just run it through an OTA controlled by the same pitch controlling
current the
> core VCO is.
>
> I have built this very simple circuit (note: no nonlinearities or complex
depen-
> dencies!) and it works great with the one obvious problem of integrator
drift
> resulting from some residual DC and its offset voltage. The question is
now,
> how to keep its output swinging around 0V ?
>
> I have an idea again, but this does not work perfectly: take two peak
detectors,
> each consisting of a diode and a capacitor and measure the highest and the
lowest
> voltage at the inegrator output. Now, offset the integrator input by the
difference
> between these two.
>
> Needless to say I would love to hear if you have some better ideas
> or even some general comments/criticism. When finished, I am going
> to post a complete schematics here.
>
> yours,
>
> mb
>
> --
>
> Maciej Bartkowiak
> ========================================================================
> Institute of Electronics and Telecommunication     fax: (+48 61) 8782572
> Poznan University of Technology          phone: (+48 61) 8791016 int.171
> Piotrowo 3A                             email: mbartkow at et.put.poznan.pl
> 60-965 Poznan POLAND               http://www.et.put.poznan.pl/~mbartkow
> ========================================================================
>
>
>
>
>