The common mode propagation delay dispersion is also likely to be significant unless one uses an SiGe ECL/CML comparator.
Calibrating this or actually the differential dispersion between channels is an interesting but not insoluble issue. Bruce > > On 01 April 2017 at 18:49 Scott Stobbe <scott.j.sto...@gmail.com> wrote: > > Also interesting the LTC6752 is rail-rail input. Any rail-rail input opamp > I've used ends up with an ugly bump in input offset voltage transitioning > from the nmos or npn diff pair to the pmos or nmos. I'm not sure how good > or bad a rail-rail comparator may behave when common-mode biased in that > region. > > On Fri, Mar 31, 2017 at 11:22 PM Bruce Griffiths > <bruce.griffi...@xtra.co.nz> > wrote: > > > > > > Attempting sub nanosecond timing with an actual 1Mohm source is an > > exercise in futility. There are very few cases where one would want > > to > > attempt precision timing measurements with such a high impedance > > source. > > The 1M pulldown on the TICC input is merely intended to maintain a > > valid > > logic input should the user leave that input disconnected. In > > actual use > > with PPS signals the source impedance is in most cases a few tens > > of ohms. > > If one wishes to have a 1Mohm input impedance for use with AC > > coupled > > signals then a low noise FET input buffer preceding the comparator > > is > > required. > > > > Protection diodes in this application not only need to have low > > leakage, > > they also need to turn on and off fast enough to be useful. > > > > The propagation delay dispersion (both vs common mode and vs > > overdrive) > > also need to be considered along with the comparator jitter. > > > > Bruce > > > > and overdrive (both vs overdrive and vs input common modeOn 01 > > April 2017 > > at 15:34 Scott Stobbe <scott.j.sto...@gmail.com> wrote: > > > > Fwiw, for a precision comparator you'll probably want a bipolar > > front end > > for a lower flicker corner and better offset stability over cmos. > > For > > high-speeds the diffpair is going to be biased fairly rich for > > bandwidth. > > So you will more than likey have input bias currents of 100's of nA > > to uA > > on your comparator. Which is not great with a 1 megohm source. > > > > On Fri, Mar 31, 2017 at 9:08 PM Charles Steinmetz > > <csteinm...@yandex.com> > > wrote: > > > > Mark wrote: > > > > I thought about using the clamp diodes as protection but was a bit > > worried about power supply noise leaking through the diodes and > > adding some > > jitter to the input signals... > > > > It is a definite worry even with a low-noise, 50 ohm input, and a > > potential disaster with a 1Mohm input. Common signal diodes (1N4148, > > 1N914, 1N916, 1N4448, etc.) are specified for 5-10nA of reverse > > current. > > Even a low-leakage signal diode (e.g., 1N3595) typically has several > > hundred pA of leakage. Note that the concern isn't just power supply > > noise -- the leakage current itself is quite noisy. > > > > For low-picoamp diodes at a decent price, I use either (1) the B-C > > diode > > of a small-signal BJT, or (2) the gate diode of a small-geometry > > JFET. > > A 2N5550 makes a good high-voltage, low-leakage diode with leakage > > current of ~30pA. Small signal HF transistors like the MPSH10 and > > 2N5179 (and their SMD and PN variants) are good for ~5pA, while the > > gate > > diode of a PN4417A JFET (or SMD variant) has reverse leakage > > current of > > ~1pA (achieving this in practice requires a very clean board and > > good > > layout). > > > > I posted some actual leakage test results to Didier's site, which > > can be > > downloaded at > > < > > > > > > http://www.ko4bb.com/getsimple/index.php?id=download&file=03_App_Notes_-_Proceedings/Reverse_leakage_of_diode-connected_BJTs_and_FETs_measurement_results.pdf > > > > http://www.ko4bb.com/getsimple/index.php?id=download&file=03_App_Notes_-_Proceedings/Reverse_leakage_of_diode-connected_BJTs_and_FETs_measurement_results.pdf > > > > . > > This document shows the connections I used to obtain the data. > > > > The TICC doesn't have the resolution for it to matter or justify a > > HP5370 or better quality front end. I'll probably go with a fast > > comparator to implement the variable threshold input. > > > > Properly applied, a fast comparator will have lower jitter than the > > rest > > of the errors, and is an excellent choice. Bruce suggested the > > LTC6752, > > which is a great part if you need high toggle speeds (100s of MHz) > > or > > ultra-fast edges. But you don't need high toggle rates and may not > > need > > ultra-fast edges. Repeatability and stability are more important > > than > > raw speed in this application. The LT1719, LT1720, or TLV3501 may > > work > > just as well for your purpose, and they are significantly less > > fussy to > > apply. > > > > Note that the LTC6752 series is an improved replacement for the > > ADCMP60x > > series, which itself is an improved replacement for the MAX999. Of > > these three, the LTC6752 is the clear winner in my tests. If you do > > choose it (or similar), make sure you look at the transitions with > > something that will honestly show you any chatter at frequencies up > > to > > at least several GHz. It only takes a little transition chatter to > > knock the potential timing resolution of the ultra-fast comparator > > way > > down. Do make sure to test it with the slowest input edges you need > > it > > to handle. > > > > Best regards, > > > > Charles > > > > _______________________________________________ > > time-nuts mailing list -- time-nuts@febo.com > > To unsubscribe, go to > > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > > and follow the instructions there. > > > > _______________________________________________ > > time-nuts mailing list -- time-nuts@febo.com > > To unsubscribe, go to > > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > > and follow the instructions there. > > > > _______________________________________________ > > time-nuts mailing list -- time-nuts@febo.com > > To unsubscribe, go to > > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > > and follow the instructions there. > > > > > _______________________________________________ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.