Nov 25, 2022
The scheme of using a pair of fast-response thermistors to measure CT^2 works well. The hardware implementation needs an iteration.
Issues include:
1. thermistors are exposed and subject to damage. Needs a low-obscuration protective screen.
2. circuit implementation is unacceptable. Plugboards and wiring and data acq unit are rattling around, unmounted, in box. OK for test setup but not as a final version.
3. There are tens of degrees of vertical temperature gradient in the dome. Need a local determination of temperature to get from voltage differences to temperature differences.
Good end-cap screen:
Ordered 12 from Amazon
Ordered more fast-response thermistors, 150K room temp, Digikey 235-1625-ND
Use Sparkfun-ordered TMP36 analog temperature sensors. They need 2.7 to 5.5V. Provides 750 mV at 25C with dV/dT=10 mV/C.
We want no more than 1V across two-thermistor voltage divider. Mouser sells good part from Analog Devices: 
Instrumentation amplifier, AMP-04: Run at +-12V to get good SNR at other end. 
For thermistors that are 150K at room temp, expect about 10 mV of delta-V for 1 degree temp change. So we could run at a gain of around 1000, get 1V of signal per deg C and dynamic range of 10 C/m gradient.