Diode connected transistor as temp sensor

Nice. Could be useful as a reference for some of us. What environment did you use for the lower temperatures - liquid nitrogen?

Too bad you can't go down to 2K maybe. Below 20K, you get "carrier freeze-out" and the voltage drop goes way, way up, and part-to-part variations get bad. Good cryo diodes can cost hundreds of dollars. It would be interesting if some cheap transistors could do as well.

John

Yeah liquid nitrogen.

George H.

................

I don't know either, but I'm guessing it's a nice round figure at a level that's low enough to avoid significant self-heating and high enough to swamp out common noise sources.

Self-heating? Somewhat, but mostly low enough current that bulk re effects don't cause errors. ...Jim Thompson

I don't see any reason that a cheap trasistor/diode couldn't work down to 2K. (I've not get any liquid helium here.) With cyro-diodes I think mostly you are paying for the nice package that they put it in. But yeah $200 for a diode from Lakeshore cryo and another $200 to calibrate it.

George H.

It might also have to do with interference effects. I run these as transdiodes so I can also do order of magnituce current switching and use the Ebers-Moll equation as a secondary temperature measurement. When I do this at lower currents 10-100 nA I find you get errors if the diodes aren't electrostatically shielded.

George H.

=A0 =A0 ...Jim Thompson

=A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 | =A0 =A0 et =A0 =A0 =A0|

=A0|

=A0 =A0 =A0 |

quoted text -

Low temperature types will worry about micro-watts of power. The re effects, I think, will be taken care of by the calibration.

George H.

text -

The exotic ones take measurements at three currents, gets rid of IS and bulk-re effects. ...Jim Thompson

text -

Yeah, most thermal conductivities, and specific heats, drop like a rock below roughly 40K. We usually run cryo diodes pulsed, at low duty cycles, to minimize self-heating.

John

A little hum or RF pickup is bad news for these things.

John

I think the real problem is the wild variation in voltage drop from part to part, below 20K. One of my customers went out for bids and wound up with SI instead of Lakeshore. Every batch had a different cal curve; they didn't change the part number or identify the batches, they just changed the data tables!

The Lakeshores are all pretty much the same. I don't know how they do that... maybe they just throw a lot of diodes away.

SI did us the honor of cloning our CAMAC cryo-diode scanner module. I don't think they sold many.

John

xt -

=A0 =A0 ...Jim Thompson

=A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 | =A0 =A0 et =A0 =A0 =A0|

=A0|

=A0 =A0 =A0 |

quoted text -

OK you are "Mr. Vbe" after all. I've seen the re effects but with currents up near 1mA.

George H.

I've seen the diode curves at low temperature, but I've never tried commercial diodes/ transistors down there. Perhaps Lakeshore picks their diodes for the low temperature performance.

I got our calibrated diode from Lakeshore. I 'goofed' a bit and only got it calibrated from 77 to 325 K. (it was like an extra $100 or something to get it caled up to 400K.) I figured that above 325 the accuracy would just be worse, but still OK. What I found was a 'step' change of a several millivolts between 325 and 330. I had to 'fix' the calibration myself. They also do this multiparameter Chebychev polynomial fit to the data. I would have thought it better to try and fit the data with some 'real' physical model.

Who is SI?

George H.

text -

Don't miss the implications of Jim's comment about Is.

Jon

text -

xt -

Hi Jon, I think I got it. I can look at the difference in Vbe at two different currents and pull out the temperature.. or if I know the temperature I think I can then get Is. With three measurements, I can also determine re. (Or is there something more subtle that I'm missing?)

But mainly I'll be using the diodes at constant current and reading the temperature off a calibration curve.

George H.

text -

I often need to measure chip temperatures, so the devices are small and have large bulk-re. ...Jim Thompson

I'm sure I'm being stupid here, but what's wrong with using integrated circuits that have been made for the job, like the Analog Device AD590 or the National Semiconductors LM35?

By the time you've spent time - and thus money - on calibating your parts the price difference between the transistors and the special purpose chips isn't going to be all that significant.

Robert Widlar's application note for National Semiconductor makes interesting - if perhaps irrelevant - reading

Of course none of the parts is specified for operation down to 77K.

-- Bill Sloman, Nijmegen

"Additional information will be available in the near future."

I interfaced to these roughly 12 years ago.

John