[time-nuts] Designing and building an OCXO and GPSDO
holrum at hotmail.com
Mon Aug 11 20:40:13 EDT 2008
What you can do is calibrate the AD537/sensor(if used)/freq counter as a unit. Things like temperature offsets, chip dissipations, oscillator shifts, component tempcos, widget wiggles, etc. get compensated for. Ideally in an ocxo you would want the sensor and freq counter in the oven (or at least the freq counter reference oscillator). Hmmm, could get wickedly interesting if you clocked the freq counter with the OCXO output you were controlling.
In one system we used a 20th order polynomial to do the conversion from reading to temperature. After a couple of years in the field (actually under the field) they retrieved, the data was dumped, and a new calibrations were done. The calibrations had typically shifted less than 100 microdegrees... and that was with a crappy oscillator in the freq counter.
As far as sensitivity... a lower res unit (around 0.01 degree resolution) could easily detect the body heat from people in the lab. From temperature gradients you could reliably count people entering and leaving (a couple of particularly ample people counted as two bodies).
The AD537 has too much drift and too little sensitivity to reliably
maintain the oven temperature (at the sensor) to within a few millidegrees.
The internal dissipation of other circuitry within the chip will
increase the sensor temperature offset from the temperature being measured.
A thermistor bridge is a better idea as it can have high stability and
sensitivity (about 10x that of an RTD).
Get more from your digital life. Find out how.
More information about the time-nuts