The probe is model DTU6024C show below. The probe combines a precision, calibrated thermistor with a 14 bit analog to digital converter and a simple USB communication interface. The probe is encapsulated in a waterproof thermoplastic elastomer (TPE) jacket. Specifications are accuracy +/- 0.1°C from 0°C to 100°C, resolution +/- 0.01°C, operating temperature range -20°C to +110°C, current draw less than 100 mA. The probe measures 0.8" in length, 0.2" in diameter and is connected to a 48" length USB cable. The probe's current list price is $300 from Redfish Sensors and its calibration expires April 21, 2014.
I zip-tied the probe to the bottom of the OTA between the mounting rings. I then added an insulating block of 2" foam covering the probe, squeezed and held in place by the OTA above and dovetail saddle plate below. My intension with the foam is to insulate the probe from nearby ambient air and allow it to track OTA tube temperature more accurately. The probe is connected to the USB hub mounted on the top of the OTA. The zip ties and foam are visible in the photo below just to the left of the rightmost ring. The probe itself is not visible.
I updated my focus drift compensation software to read the probe using its simple serial protocol.
Focus position and temperature measurement data and linear models for 22 nights at 4 sites are shown below. The dotted lines above and below the fits indicate the +/- 10 micron critical focus zone. Red lines correspond to nights with the OTA dew heater turned on, blue lines correspond to nights with the heater turned off. Linear model residuals are smaller each night and linear model parameters are more consistent across nights than previous results using the FLI Atlas' internal temperature probe. The median slope of the data is 44 +/- 5 microns/°C. I will update this chart from time to time with additional data.
The following chart shows a linear model fit to the entire data set using weighted least squares. Light grey lines join measurements on a single night. The weighted least squares model slope equals 45 microns/°C.
Regarding the focuser image shift problem, another workaround was suggested by Fred Miles. Immediately prior to refocusing, close the camera shutter without terminating the exposure, and once the image shift is corrected by the autoguider, open the shutter. I am currently investigating a software implementation of this scheme.
The QSI camera driver does support closing and opening the shutter during an exposure, but does not support multiple clients. Also, the driver is implemented as a plug-in to MaxIm, so adding shutter control does not appear straightforward. I am putting this idea on hold for the time being.
I employed temperature compensation using the Qti probe on my July and August 2013 photograph. None of the 52 subexposures were rejected due to focus drift. The image shift problem does not seem to compromise star shape. I think the reason for this is that the duration of the uncorrected image shifts is small compared to the 40 minute subexposure duration.