[time-nuts] Precision temperature measurement - Time-nuts style

Mark Sims holrum at hotmail.com
Fri Jul 11 19:20:53 EDT 2008

In the beforetimes,  I built several precision temperature systems using the Analog Devices AD537 voltage-to-frequency converter.  This chip has an on board temperature sensor and makes a very nice precision thermometer.  Due to its size and thermal mass it is good for fairly slow response curves.  For faster response you can use a small thermistor as the sensor.  You can get microdegree resolution in fairly simple systems.

One system that I built was a long term data logger that measured temperatures in mines, caves, and lakes.  They were built in iron plumbing pipe sealed with teflon tape and pipe ends.  The original unit, built by a friend, used discrete logic to wake up, do a sample, save it to eeprom, and go back to sleep.  The frequency measurement was a synchronized gate counting scheme.  It has a couple dozen chips and would record for a couple of years.  It's replacement used two PIC chips.  One was the wake-up timer on a 32KHz crystal,  the other was clocked with a 16MHz xtal and did the frequency measurement and logging to a serial EEPROM.  It gave millidegree resolution.

Another system was a fast response system with a tiny (almost microscopic) thermistor.  It would log temperature changes as it was lowered into wells and boreholes.  From the temperature changes one could deduce the rock strata, etc of the hole.  The counting scheme was basically recording time tags from a high frequency clock/counter.  It had microdegree resolution.  The thermistor was mounted in a piece of stainless hypodermic needle tube protected by a slotted steel shroud.  You could only log a hole once in a long time because just lowering the probe through the hole disturbed the temperature profile.

Both units were calibrated to absolute temperatures as a complete system.  One group thought that that they had a problem with a unit.  They mounted it in a (supposedly) well nsulated box and were seeing unexplained temperature gradients.  Turned out to be body heat from people entering and leaving the lab.  You could easily count bodies in the data.
Need to know now? Get instant answers with Windows Live Messenger.

More information about the time-nuts mailing list