OK, very interesting. Now is it possible to measure/verify this? I think that using any test equipment, the comparator-style approach is unavoidable: the trigger of the scope or the counter cannot be an amplifier/limiter. How to tell what is up to my design under test and what is the trigger contribution? Maybe only by comparison: test design A then design B and see which is better...
On Fri, Jul 20, 2012 at 2:31 PM, Magnus Danielson < [email protected]> wrote: > On 07/20/2012 07:42 AM, Chris Albertson wrote: > >> On Thu, Jul 19, 2012 at 5:47 PM, Rick Karlquist<[email protected]> >> wrote: >> >> >>> Hysteresis does nothing to eliminate jitter or temperature >>> >> >> Maybe, but it is absolutely needed if there is any noise on the >> signal. A perfect comparator with zero hysteresis would dither on >> every zero crossing. >> > > Yes, and this dither is due to the additive noise on the signal. The > slew-rate at and about the trigger point will determine how much of that > additive noise is converted into time-noise. The schmitt trigger is there > to make sure that you surpress the dither around each transition, but it > will not help you to remove the time polution, as the first time the dither > occurs, is bound to be early and bound to be controlled by the noise. > Those, the noise will shift the trigger point. > > You can view the schmitt trigger detector as having a state, and when in > proximity of the trigger point, you let the noise control when the trigger > point occurs. > > If you noise is pure gaussian noise, this is not so bad, since the trigger > point will be shifted by the noise RMS, but it will be noisy. > If you have say a sine signal, then the non-linearity of the trigger point > will act like a mixer and it will cause the time jitter to be spread out, > and the peak-to-peak amplitude of the signal will when divided by the > slew-rate of the trigger point convert to the peak-to-peak time modulation > at that frequency. The distribution has a very steep bath-tub look, since > the sine spend most of it times at its extremes (where it's slew-rates are > low) but very little time in the middle (where it's slew-rate are high). > The sine signal would modulate the trigger point up and down on the slope > it's at. The schmitt trigger action doesn't help to protect this behaviour. > > Schmitt trigger is a nice tool, but it can do you great harm if you do not > understand what it does help you with and what it doesn't help you with. > > You need to gain yourself to slew-rates where a schmitt trigger would do > no harm, and when you are there it will do essentially no good either, as > you are looking at a high slew-rate square signal. > > So, you *can* do better than a Schmitt trigger. A schmitt trigger can be > sufficiently good. A schmitt trigger can work well if you have filtering in > front of it to significantly reduce unwanted systematic noise. > > Cheers, > Magnus > > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. > _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
