TADD-2 Frequency Divider Tempeature Stability

These are some preliminary results of the temperature versus phase response of the prototype TADD-2 frequency divider (a PIC running code written by Tom Van Baak and modified by Richard McCorkle, with some glue logic attached).

The short answer: I measured a phase change of about 3.3 nanoseconds over a temperature range of 53 degrees Centigrade, or 62 picoseconds/degree.

I don't have a proper environmental chamber, so I had to improvise. I do have a quartz infrared pre-heater that's designed for soldering. It has an adjustable set temperature with a thermostat and feedback loop that monitors the actual temperature of the working area. It doesn't regulate too well, but it provides even heat across the board and seemed usable for this task. Here's a picture of the TADD-2 mounted on the heater:

I used an HP 5370B time interval counter (resolution 20 picoseconds; accuracy <100 picoseconds) to record the time difference between PPS from a Z3801A GPS discplined oscillator, and PPS from the TADD-2 when driven by the 10 MHz output from the same GPSDO. Since the TADD-2's PPS is derived from the same source, there should be no frequency offset between the two pulses, and under constant conditions the results should show only noise, as seen in this plot:

After logging the results for an hour or so to get a baseline, I turned on the heater (set for 125 degrees F) and continued to record the phase, while taking the board temperature once per minute using a hand-held infrared thermometer (I aimed the spot from the thermometer at an unused area near the center of the board). Here is what the phase record shows over a several hour period with the heater experiment starting at about 80 minutes into the run:

You can see the phase shift as the heater overshoots the setpoint, shuts off while the board cools, and then starts cycling every 12 minutes or so. Here's a zoomed-in plot with the temperature superimposed over the phase data during the initial heating period:

This nicely shows the impact as the heater comes to temperature, and attempts to maintain 125 degrees F (37.2 degrees C) as well as the fact that it significantly overshoots and maintains a pretty crude idea of the set temperature (it overshot by about 50 degrees F on the initial cycle, and 25 degrees in steady-state).

The starting temperature was about 21 degrees Centigrade, and the peak was about 74 degrees, for a 53 degree range. The ambient temperature in the lab remained fairly constant during the experiment.