Yes, you want to *start with* the triangle wave. You can always make a square later, via nonlinear process (comparator). That is, you'd never make a square by differentiating the triangle, that'd be silly. But you have a whole lot harder time making a triangle from a square, without using some bizarre compensated integrator.
If you absolutely must (e.g., to sync with an external clock), you might be better off with:
a. Triangle oscillator, PLL locked to source
b. Integrator with peak amplitude detection, OTA controlled gain (amplitude locked loop, I guess you could say)
They're more-or-less equivalent because of the direct ratio (frequency and voltage being proportional), which should have interesting consequences actually, as a PLL is generally considered more complex than a mere integrator and gain stage. Or conversely... the amplitude loop should be as complex as the PLL, since after all, it's trying to regulate on a widely variable signal, so the compensation will never be perfectly well defined.
Tim
Home
Help
Search
Login
Register
