Saw to triangle problem - NOT ANYMORE :)

[Martin Czech]

> After two nights of bugging with this rounded top of the triangle waveform I solved the problem! Yes way!
> Finally diodes got the boot, because of their knee. Away with them. New design uses two opamps (they are cheap anyway) and a transistor (ordinary NPN) and a bunch of resistors. The schematic is available at:
> http://imola.uni-mb.si/~bozicko/saw2tri_v2.gif
> I am very proud on this converter since both peaks of triangle are razor sharp (as it should be in theory). The only thing to watch out is, that you use high slew rate/bandwidth opamps (TL082 roughly does it), otherwise there could be some spikes, when sawtooth changes from higher to lower peak. It think Burr-Brown's OPA604 would be ideal, since its is low cost and has 25V/us slew rate and 20 MHz of bandwidth. You can remove spikes, if you use low amplitude sawtooth (lower than the slew rate). The triangle remains perfect... I am impressed!
> 
Maybe you can even save the comparator, if your vco pulse output has the 
right shape, 50% duty cycle etc.

m.c.