fast switches for a boxcar integrator

That number is impressive, amazing actually, but is measured with Rgen=0 and Vin=0 and a 100 pF output load. I'd love to see some more practical data. These logic-type switches are usually terrible for charge injection.

John

The Rgen=0, and Cload (typ 1 to 10n) condition is the usual one for switches and most switches are given for pC to 10s of pC under such conditions. But even with tuned to 0pC injection, most switches have weird behavior on high load impedance and I got caught once on this (1G circuit impedance on one side and C load on the other). Charge injection isn't necessary what we'd expect at first cut: some (most) switches have worse behavior than simple gate channel capacitive coupling since they actually *draw* current when switched on/off on a RC load, that is:

_/ .------o/ o-----+------. | | | --- --- .-. --- --- | | | | | | | | '-' | | | === === === GND GND GND

not only give you an AC component but also a substantial DC one, depending on the switching frequency. It appears that when OFF, some internal part on the switch path is tied to some supply/ground pin, or maybe some internal servoed voltage, which AD didn't deny when I asked. (It was for an ADG444 IIRC). I finally found one (the ADG1211 family IIRC) that was free from this leakage. Some trick allowed a net charge (after the transient glitch has settled) about the order of a fC. (not pC, but yes, fC)