Wing it with a small MOSFET, as Mr. Larkin suggests? Maybe backed up with some measurements?
My bet is that once the thing exceeds the datasheet specifications, the actual behavior is all over the map, not just from manufacturer to manufacturer, but from date code to date code.
I think if I were using such a critter and repeatability was important, there'd be a known resistance in between the FET and the cap. Come to think of it, I'd probably have it there anyway, for chip-preservation.
One issue is discharging the cap to millivolts in nanoseconds, or something. Another is whether the nonlinear capacitance of the switch distorts the charging curve. It's not easy to make a fast ramp that's linear to 10 or 14 bits.
John Larkin Highland Technology, Inc
picosecond timing precision measurement
"Jim Thompson" wrote in message news: email@example.com...
There's a then-Philips library out there, implemented in discrete. But it's crap: no capacitance so it's infinitely fast, output transistor dimensions are wrong, no input protection diodes. Not to mention only the most basic gates are implemented; MSI chips (like '74, if that even counts as MSI yet!) are blank, or a dummy inverter, or something.
Anyway, I took it upon myself to measure a couple real parts and fit the transistors and everything properly. Note that the differences between real parts are pretty crude, so the model isn't very precise, nor should it be.
I modeled a 74HC7014, but it's easily chopped into an '05.
Yabbut -- if I were that concerned, I'd use some buffer driving the gate of a Really Well Modeled FET, or maybe just an analog switch. Or I'd be sure to demand a payoff from whatever engineer was going to be called on twice a year to tweak the design every time that the fab changed its process.
While I like using logic chips for off-label solutions like this, if the performance is really critical I start thinking of using something a bit more predictable.
I suppose if it's for something that's going to be made 100,000 at a pop the story might be different -- then the engineering time spent to make it work with a schlock part might be worth the BOM cost saved.