Variable ramp again

[mbartkow at ET.PUT.Poznan.PL]

Oh, my fault I didn't explain how it works.
It's very simple, though. Consider a nonsymmetrical pulse wave, e.g with PW=10%.
Removing the DC offset means that a constant value is added which causes that
the areas under the negative "half" and the positive "half" are equal. In other
words, the integral of the wave calculated over whole period is zero. In our case,
the negative part should have 1/9 of the amplitude of the positive part. These
different amplitudes cause different slopes of the ramp observed at the integrator
output. Since ideally the pulse wave has zero DC, the integrator should not drift.
As this is not achievable (for example, the comparator output levels may not be
strictly symmetrical) a correcting feedback is needed (thanks JH for the interesting
point here)

MB
(I am in Hannover for two weeks, and replying remotely)