diode as a temperature sensor for liquid nitrogen

[Chris Smith]

If the voltage increases, it means the resistance is dropping. Voltage is a follower, not a leader.

[rshayes]

The characteristic that is used for temperature measurement is the forward voltage of the silicon diode, not its resistance. The forward voltage can be viewed as being a constant plus a term proportional to the natural log of the current density minus a term linearly proportional the temperature. If the current is fixed by the external circuit, then the voltage decreases as the temperature increases.<p>As the temperature is lowered, the forward voltage increases about 2.2 millivolts for each degree kelvin drop in temperature. As you can see from the graph that jwax referred to, this is true until the temperature drops below the temperature of boiling liquid nitrogen. Below this point, the temperature coefficient changes rather drastically.<p>Semiconductors are not superconductors. As the temperature of a semiconductor is lowered, the resistance increases, and can become quite high at cryogenic temperatures. This can interfere with the measurement of the diode voltage, since the voltage drop through the bulk semiconductor can become significamt.<p>Metals show the opposite behavior. As the temperature of a metal increases, the resistance usually also increases. This is often used to measure the temperature rise in transformers.<p>The current at which the diode is operated should probably be in the 10 microamp to 1 milliamp range. At currents much above 1 milliamp, the bulk resistance of small signal diode tends to become significant. Currents much below 10 microamps will require a hgih impedance measuring device, and leakage currrents may become significant.<p>You will have to calibrate at least one point. The boiling point of liquid nitrogen is as good a point as any. Use fine wire to connect to the diode to avoid conducting heat down the leads.<p>If the case of the diode doesn't crack when it is immersed in liquid nitrogen, you are probably OK. If the thermal mismatches are large enough to cause trouble, the damage will probably occur in a few seconds. Metal can transistors may be more rugged in this respect, but a good glass case diode may be good enough.

[labview1958]

I just dip the silicon diode in liquid nitrogen. The DMM read 930mV. At room temperature it was 480mV. An increase of 450 mV. The change in temperature was 223 K. Thus that gives a change of 2.02 mV/K. Is that correct. From the website it should have been 2.7 mV/K. Any comments?

[L. Daniel Rosa]

Keeping the constant surrent stable won't be a problem as you won't be dipping that into the nitrogen- it'll be kept at room temperature. On that note, the LM134 (available as 234 or 334) is a temperature sensitive current regulator that be stabilized for use with the diode.<p>OTOH, the '134 comes in a metal can. While it's operating conditions are -40 to 150C, you might play with the idea of using it in the soup- perhaps as 1uA/degreeK.

[rshayes]

Textbooks give a nominal temperature coefficient of -2.2 millivolt/degree kelvin at room temperature. For some reason, the measured value in the liquid nitrogen range seemed to be closer to -2.0 millivolt/degree kelvin. Your results seem to be pretty close.<p>The diode may not have actually reached 77 degrees kelvin. Heat conduction down the leads may have kept the diode a little warmer than the liquid nitrogen. If you have some wire in the #34 to #36 range, you might try using that to connnect to the diode and see if the results change.<p>Presumably the diode didn't crack, so it appears the the diode packaging is adequate for this purpose.

[Dean Huster]

"If the case of the diode doesn't crack when it is immersed in liquid nitrogen, you are probably OK. If the thermal mismatches are large enough to cause trouble, the damage will probably occur in a few seconds. Metal can transistors may be more rugged in this respect, but a good glass case diode may be good enough."<p>That was one of my concerns. After all, most components are rated within what humans can stand with the proper clothing, on the low end anyway. I was wondering if a glass diode would crack. A plastic/epoxy diode might have enough of a shift in expansion/contraction to goof up the junction. Metal case transistors more rugged? That would give you one more metal-to-glass seal to compromise, 4 total rather than just two!<p>I think that's why thermocouples are the preferred choice. They're about the most rugged of all sensors.<p>Dean

[philrosenberg]

I've just read an article about just this topic printed in a uk science journal. The article is available online at http://ej.iop.org/links/q80/+9ZpDA6X06K ... i2p104.pdf but you may need a subscription to view it. It is called "Low temperature thermometry using inexpensive silicon diodes" and details callibration of a IN4148 diode from 60-300 Kelvin. They seem very happy to use this sensor repeatedly in a University student lab and comment on the need for durability. They used a constant current source but commented that changes in room temperature affected their source and hence temperature measurements. You only wanted 5K accuracy though so the circuit they use should be fine. If you can't access the article online then email me at [email protected] and I'll send you a copy.