SUMMARY:
Mildly Bad News: We are pushing our function generator (pulser source for timing calibration) to its limits at the current settings and are consequently getting sub-par performance.
Definitely Good News: Our measurements from mTC indicating that the calibration pulses were consistently closer to ~8.2ns rather than 8.0ns have probably been right all along!

* * * *

We had a fairly long phone conversation with a Tektronix engineer (Steve) today. He was quite helpful. Below is a summary of each of the main topics we discussed with him.

- Definitions of "pulse width" and "rise/fall time":
We confirmed that we are all using the same definitions & conventions for these terms. Rise & fall times are determined by the 10% & 90% crossings*, and pulse width is the time difference between the 50% crossings, as illustrated by the diagram 'mTC_Calibration_Pulse.png' attached here.

- Interface for our scope:
We were hoping that a module could be swapped out in our scope to provide a more up-to-date interface than the current floppy-disk setup (the function generator, for example, has both USB & ethernet interfaces). Unfortunately, no such upgrade for our model is available.

- Custom / arbitrary waveforms:
Our generator has some capability to output custom-designed waveforms (uniform sample rate only). There is a free (unfortunately Windows-based) program for creating these called "ARB Express." There also appears to be a menu for doing so on the generator itself, but the interface is clunky & unclear. Uses of this could include, for example, 1) simulating some mTC output from the generator so it could be plugged into a scope, since we can't get a scope probe inside our actual electronics; or 2) using a custom waveform to squeeze the output waveform a little closer to our desired parameters (see  below).

- THE BIG ONE -- PULSE WIDTHS:
Especially after Slava noticed that we've been getting pulse widths of ostensibly ~8.3ns, we decided to plug the function generator directly into the scope to see exactly what we've been feeding the mTC. (For today's tele-meeting with Tektronix, we actually used the borrowed 40-GS/s Mega-scope). We saw a good frequency match, but the pulse widths  were consistently wide by about .1--.3ns, and the rise/fall times were consistently narrow by about roughly .4ns. Furthermore, our supposed 1.8Vpp amplitude was coming in a few 10's of mV too small. Steve pointed out that we'd cranked not just one but all of the time settings all the way down to their minimum values and were therefore running the generator at the very edge of its operational range, most likely causing the problems we were seeing. Sure enough, after we relaxed (only) the pulse width by a few ns, the pulses became both more stable and more accurate*. Steve's recommendation was to leave the rise/fall times where they are but back off on the pulse width until we get the performance we want. (He also suggested some details for custom waveform generation; I'll write these up if we go that route).
UPSHOT: We can, of course, calibrate to 8.2ns just as well as to 8.0ns provided that we know the actual value, but it may be worth using a couple-ns-wider pulses in order to get 1) a more uniform calibration signal and 2) the desired 1.8Vpp.


So all in all, a very informative meeting; and those of us worried that mTC was stretching out 8.0ns into 8.2ns can rest a little easier tonight.

* This is an adjustable setting on our scope ("reference levels"), but I checked and its settings were correct.
** See attached screenshots for details on all these measurements. "DPO" is the borrowed ultrafast scope; I've also included screenshots of some measurements taken by our scope, the "TDS." [At present I do not have an explanation for why our scope agrees with DPO about widths but not about rise/fall times, other than that these photos were taken on different days and perhaps the straining generator was behaving a little differently].