2N3906 transistor as a remote temperature sensor

Thread Starter

Parasentience

Joined Oct 1, 2008
5
Good morning,

I stumbled upon this site a few days ago while I was looking for information on the 2n3906 transistor, and transistors in general. That said, I'd like to pose a question, in hopes that there's someone on these boards that can help me out.

Right now, I'm working on a project to use a 2N3906 transistor as a remote temperature sensor. I understand that this can be done by applying two consecutive currents with a known ratio. (I'm using a ratio of 10, for simplicity.)

There's a non-ideality factor that comes into play as well though. Everything I've read has implied that it is a constant, and that some manufactuerers of the 2N3906 chip even pre-trim the chip to a specific non-ideality factor. While I haven't found any manufacterers that actually do this, I did try calibrating these chips at three different temperatures to see if the non-ideality factor was consistent. (If it did, I could just solve for it.) Well, it wasn't. I got a different value for the non-ideality factor at every temperature.

I was wondering if anyone here has any experiance with this. I'm wondering if I misinterpreted the documentation I've read on this sort of thing. Perhaps the non-ideality factor isn't a constant at all. Is this the case?

There is a bit of noise in the line, but its in the area of 0.000005 volts. Not sure what the frequency is, my oscilloscope is a little old and I can't measure anything that small with any sort of accuracy. Shouldn't matter though with a magnitude that small, am I right?
 

scubasteve_911

Joined Dec 27, 2007
1,203
I am not sure if you read the following article, which pretty much gives everything to you on a silver platter.

http://www.edn.com/index.asp?layout=article&articleid=CA307863

The impact of noise is a function of your acceptable signal-to-noise ratio. So, if you are working with 0-10V signals and you have 5uV of noise, it is insignificant. If you are working with 0-10uV and your noise is 5uV, then, clearly, you have a problem. I don't know what ranges you have to work with and the desired resolution.

Steve
 

beenthere

Joined Apr 20, 2004
15,819
Interesting experiment. Most of us weenie out and just use an LM34 or -5, or even an uncompensated thermistor.

That 5 microvolts could come from any number of sources. What is its frequency, or can you make that out? Does it increase with the lines open and go away with the lines shorted? Got a transmitter nearby? Or fluorescents overhead?

For that matter, what equipment are you using to measure the 5 uV? What spec's does that meter have, and how long out of cal is it?
 

Thread Starter

Parasentience

Joined Oct 1, 2008
5
Fluorescent lights overhead, yes. Do you think its picking up noise from those? I suppose it could be possible.

As for the meter, its a fluke multimeter, though the model number escapes me offhand. I'll check the manual to see what the accuracy of the meter is at the voltages I'm reading. The meter is up to date on its calibration. Tyical voltage measurements are between .2 and .7 volts, depending on current supplied and temperature of course. However, a change of 0.0001 volts causes a change of 1 degree F in my output, so it is rather sensitive. (Edit: For reference, my supplied currents are 20 and 200 microamps.)

As for the frequency, I can't make it out. I managed to get my hands on an oscilloscope, but it couldn't read in the microvolts range. It looks like it shows some noise, but I can't get the resolution to make out what the frequency might be.

I'll check to see how the noise on the meters is effected by an open or shorted circuit and I'll post what I find. Do you have a suspicion along these lines, if it turns out it behaves as you described? (Noise goes away when circuit is shorted)
 
Last edited:

beenthere

Joined Apr 20, 2004
15,819
It is quite possible you are reading an induced voltage. You could try shielded wire to the 2906.

It's also possible you are seeing variation due to stray air currents affecting the transistor, which could be self-heating. Drop it in a beaker of water and see if it settled down. That will increase thermal mass considerably, and should stabilize any drift.
 

Thread Starter

Parasentience

Joined Oct 1, 2008
5
Thanks Beenthere. I'll try using the shielded wire, or perhaps just turning off the lights if that's the problem.

Maybe stray air currents would effect the room temperature readings. Not so sure about the other two. The transistor is placed in a plastic bag and inserted into a calibration bath for those tests. I may try different current combinations to see if I get a more stable reading.
 

SgtWookie

Joined Jul 17, 2007
22,230
I'd put a small cap on it to remove the noise.

I really don't like noise ;)

Try somewhere in the vicinity of a 47pF cap to ground. That should quiet things down a bit. Since you're over time, the small bit of capacitance shouldn't matter much.
 

n9352527

Joined Oct 14, 2005
1,198
How much current are you working with? I am not familiar with the two currents ratio method, and would really like to know further. Do you have a link to relevant articles?

I made a measurement jig for power dissipation characterisation years ago, the temperature was detected through the voltage drop across the junction (BJT or diode). The voltage change against temperature, for a given current, was constant for each process (thus BJT or diode type), around 2mV/degC IIRC, and I had to measure the temperature coefficient for that particular type first.

Wouldn't applying two different currents resulted in two different temperature coefficients?
 

Thread Starter

Parasentience

Joined Oct 1, 2008
5
Currently working with 20uA and 200uA for a current ratio of 10:1. These calculated by measuring the voltage across a 10K resistor. Actual resistance of the resistor doesn't matter, since its the current ratio used to calculate temperature, not the current itself. Therefore, resistance drops out of the equation. Power dissipation is pretty small in the resistor, so I'm assuming its not self heating. If it was, I suppose that might cause the resistance to rise more when the 200uA is applied vs. the 20uA, giving me a superficially high ΔVBE. Only problem is, I'd rather not try to solve for the resistance of a resistor in a closed circuit if I don't have to.

While applying two different currents will indeed give me two different voltages across VBE at any given temperature, its actually the change in VBE from these two currents that I meaure to get my temperature. You can read more about it in the article scubasteve posted the link to. (An article I had missed, but which may or may not help me troubleshoot my accuracy in temperature readings using this sensor.)

As for the noise, I have a suspicion that its not part of my temperature accuracy, but I am interested in cleaning it up as much as possible. Perhaps I'm mistaken though. Once I solve this problem, then I figure the noise will start to come into play as far as precision of my readings goes. Sgt. Wookie, if you don't mind me asking, why 47 pF? Wouldn't a larger cap be better since I'm only trying to get the DC component into the transistor? I'm more of a novice in electronics and circuit design and I suspect that there's quite a bit I don't have a clue about.

Now, another thing I've been wondering about is the temperature characteristics of carbon resistors. I'd always assumed they had fairly linear resistance with respect to temperature. Is that not at all the case?
 
hello there Parasentience,
im trying to use 2n3906 trans for thermal sensing also..
do you have any circuit diagram for the sensor?
your response would be very much appreciated..

regards.
 
hello there Parasentience,
im trying to use 2n3906 trans for thermal sensing also..
do you have any circuit diagram for the sensor?
your response would be very much appreciated..

regards.
 
hello there Parasentience,
im trying to use 2n3906 trans for thermal sensing also..
do you have any circuit diagram for the sensor?
your response would be very much appreciated..

regards.
 
Top