Postato su sci.electronics.design, provo anche qui perche' non si sa mai. Scuasate per l'inglese ma ho poco tempo e faccio giusto un copy-paste:
Any idea for a voltage controlled current source (linear of course) to be implemented in a CMOS process? The difficult part is to get rid of process variations (the most dispersed component seems to be the poly resistor)
Variants from standard textbook bias sources (Vittoz, PTAT, Wilson...) seem not to work properly; I get changes of more than +/- 40% from centered value. Temperature compensation is not important, nor PSRR.
I'm looking for solutions which compensate resistors variations, for example making use of a differential pair. I could make use of a bandgap, if necessary, though I prefer to avoid it. Do you know any solution?