Yes, the spec for release time on the reed relay is 1/5 ms. I'm surprised, of course, that they've increased the performance an order of magnitude since the last time I investigated reed relays. The other side of the coin is the "operate" time which is about a half millisecond. I'd probably press the mfg for more specificity about what these mean.

That is what Suresh intends to determine. The parameters have to be well defined, however, before they can be evaluated. I don't know what they mean by release time, and I don't know what they mean by operate time, but I do note that they've included bounce time.

The reed relay is a spring and a coil, right? When the coil inactive, the spring is relaxed and static. When the coil is active, it takes a short while for a field to build up and the spring to start moving. The force the field applies to the spring is proprotional to the current, is it not? Now, so far, I think the assumption has been that the "operate" current and the "hold" current may, in fact, be different, though that's not how they're specified in the summary to which that link points, but, assuming, for now, that it is, if the integrator capacitor smooths the waveform that's stepped by virtue of the bounce, all it will take is two successive samples that reflect a delta equal to the fastest charge rate of which the integrator is capable. All that has to happen, is that the integrator ramps up/down slowly enough when the relay is NOT connected 100% of the time than when it is, and that the difference in slope is large enough that a converter can reliably detect it with a couple of samples. Since the sample rate is known, the final sample times can be deducted from the duration of the measurement.

Yes, that's more sophisiticated a an approach than some have recommended, but it is probably easy enough to implement, given that Suresh knows a great deal about the devices under test.

There are several ways in which this can be done, but if he has a fairly fast converter, that should be adequate to determine when the delta has reached its maximum. What he essentially has to do is simply to initiate a long sequence of conversions, and look for a constant delta at or very near the rate he interprets as the maximum that his integrator allows. The bounce time, then, is the interval between the first detected delta, and the one at which he's detected the maximal rate.

Now, I suppose he could build a precise differentiator with which to quantify the ramp rate, but if he has a converter fast enough to distinguish between rates, its easier to do that in software, don't you agree?

RE