The attached circuit is adapted from an article originally published in EDN magazine. In the article titled "Test your analog-design IQ" by Jim Williams, this was question #14 in conjunction with Figure 9. The challenge was to find the temperature sensor in the circuit. The instrumentation amplifier in the figure had no components with values so I just made one up. I've never had to use or design one before so I was a bit unsure of the scale. I got a reasonable gain of ≈ 126 on the second try. This was also the first time I've done a temperature sweep in LTspice. I thought some other simulation aficionados might enjoy it as well.

 

I had a play. U3 is one candidate.

 

The diode.

 

The instrumentation amplifier is looking at a difference, What is different about the two things it is looking at and what is the same?

 

This is a bit of a puzzle.

From the 1:2 ratio of R4:R5, it's evident that the Q1-Q3 string are operating at roughly twice the emitter current as Q2-Q4. As a result, the difference in Vbe between Q1 and Q2 (that is, V1 - V2) will be proportional to absolute temperature, provided BOTH Q1 and Q2 are subjected to the same temperature. But that's two components sensing temperature, not just one.

On the other hand, if only one of them, Q1 or Q2, were sensing temperature while the other was held at constant temperature (V1 - V2) would be changing at the rate of approximately 2.2 mV/°C. And that rate would vary somewhat from unit to unit.

My take: the Q1-Q2 pair are the temperature sensor. That's the best I can do right now (maybe need more coffee).

 

This is a bit of a puzzle.

From the 1:2 ratio of R4:R5, it's evident that the Q1-Q3 string are operating at roughly twice the emitter current as Q2-Q4. As a result, the difference in Vbe between Q1 and Q2 (that is, V1 - V2) will be proportional to absolute temperature, provided BOTH Q1 and Q2 are subjected to the same temperature. But that's two components sensing temperature, not just one.

On the other hand, if only one of them, Q1 or Q2, were sensing temperature while the other was held at constant temperature (V1 - V2) would be changing at the rate of approximately 2.2 mV/°C. And that rate would vary somewhat from unit to unit.

My take: the Q1-Q2 pair are the temperature sensor. That's the best I can do right now (maybe need more coffee).

Bing..Bing..Bing! Give that man a cigar. They are operating at different currents and the gap between them increases. The change is about 5 mV over 0°-85°

 

Bing..Bing..Bing! Give that man a cigar.

Can I have a Hershey bar instead? Obie don't smoke...

 

Can I have a Hershey bar instead? Obie don't smoke...

Carnival Barkers are one-trick ponies. My supply of things with sugar is well...limited. I don't smoke either BTW. Will an "attaboy" do the job?

 

Carnival Barkers are one-trick ponies. My supply of things with sugar is well...limited. I don't smoke either BTW. Will an "attaboy" do the job?

Works for me. Thanks.

In one of his Linear Technology "circuit collection" appnotes, Jim Williams showed a design for a ΔVbe thermometer that worked on this same PTAT principle; it used one section of an LTC1043 analog switch to alternate the emitter current of the sense transistor (which was diode-connected, with base and collector shorted) between two fixed values and another two sections to demodulate the changing Vbe into a DC voltage proportional to temperature.

The key feature of his design was that it didn't matter what transistor you used in the circuit-- accuracy was almost completel independent of transistor characteristics.

The guy did lots of good work.

 

He certainly did. I wish I had had an opportunity to meet Mr. Williams and some of his contemporaries. There were other bright lights I missed along the way.