2N2907 PNP transistor as temperature sensor

[alex.martinez]

Howdy!,

I have been recently into temperature sensing applications and I'm testing different methods to read this magnitude. Thermocouples and dedicated IC's are cool, but using a diode-connected BJT is even cooler (no pun intended). From what I have read in Analog Devices application notes, a current has to be injected into the diode-connected transistors. I achieve this by using a sketchy current source consisting of another 2N2907 and an opamp (I'm using an LM324).

In order to read the temperature I probe the Vbe through a voltage buffer (to isolate it) and then I try to subtract a DC offset of 500 mV for latter amplification. Point is, that I cannot get to amplify it correctly.

The current source is set to output 250 uA (probed, and it does maintain such level) and the Vbe stays at around 580 mV. Whenever I heat it there is a variation in temperature (drop in voltage due to a negative temperature coefficient). However, when it comes to probing and amplifying it, I'm getting some unknown issue to me. That is why I wanted to ask in the forum if there is any wrong assumption in the schematic, attached below.



Thanks in advance,
Alex.

EDIT1: Fixed opamps on schematic.

[StillTrying]

The center op amp is connected in a nonsense way, input should be to its non inverting input.

[alex.martinez]

The center op amp is connected in a nonsense way, input should be to its non inverting input.

Thank you, I assembled the schematic quickly on LTSpice, the opamp is wired correctly on the breadboard (I'm using an LM324N with 4 opamps).

[StillTrying]

The feedback on the VOUT op amp is wrong! After that work out what the inputs of 580mV and 450mV should do.

[alex.martinez]

The feedback on the VOUT op amp is wrong! After that work out what the inputs of 580mV and 450mV should do.

Yup...I messed up when orienting the opamps in the draft... Im an idiot. Just fixed it. with the resistor names in the re-uploaded schematics, VOUT = -Vbe*(R6/R5) + Vcc(Rb/(Ra+Rb))(1+R6/R5) = -4.54Vbe + 2.6

This amounts to -0.06 V at the output and I guess it's outside the rails of the opamp...Thanks.

[StillTrying]

Assuming inputs of 454mV and 580mV I make the output -118mV, but I don't use the formulas.

[David Hess]

Check out figure 11 of Linear Technology application note 45 shown below for how delta-Vbe temperature measurement works without calibration.  The Tektronix DM501, DM502, and 7D13 all used this technique.

Triple Vbe measurement is also possible to remove the effects of series resistance.

[alex.martinez]

Quote from: David Hess on Today at 10:46:35 am

Check out figure 11 of

Linear Technology application note 45 shown below for how delta-Vbe temperature measurement works without calibration.  The Tektronix DM501, DM502, and 7D13 all used this technique.

Triple Vbe measurement is also possible to remove the effects of series resistance.


Thank you! I had already read this application note, together with AN137. I am intending to perform a deltaVbe measurement with the schematic I attached on the original post, simple by chsnging the voltage reference on the opamp that is part of the current source to inject another known value, then mwasure both Vbe and obtain the change in potential.


However, first I wanted to optimize the full scale voltage for the ADC that I am using.

[Kleinstein]

250 µA is quite a lot of current and will heat up the sensor quite a bit.  An advantage of the diode as a temperature sensor is that it can work with a low current and still have a relatively low impedance. So something like 1-10 µA should be sufficient.

The amplifier can be non inverting  -  so no need for the extra buffer.

Using digital decoding for the difference is a good idea, though it might need a little more resolution for the ADC: the difference is proportional to the absolute temperature. So a 8 Bit ADC would only get steps of a little over 1 K if the gain and offset are well chose.
Just direct voltage reading could give higher resolution as higher gain could be used.

[dzseki]

Professional industrial diode temperature sensors exclusively run at 10uA bias current and as it seems they prefer the CB junction instead for whatever reason, these are meant for cryogenic use manily down to only a few Kelvins.

[David Hess]

Thank you! I had already read this application note, together with AN137.

Tektronix includes a detailed circuit description in their service manuals for the DM501, DM502, and 7D13 which is also worth reading.

[jackthomson41]

You are using 2N2907 as a temperature sensor, that's kind of a new thing for me and I still can't understand this how you are gonna implement it?

Btw when I need temperature sensor I use DS18B20. I think you are trying to design a new temperature sensor may be. You can also check this DS18B20 Simulation in Proteus.

[alex.martinez]

250 µA is quite a lot of current and will heat up the sensor quite a bit.  An advantage of the diode as a temperature sensor is that it can work with a low current and still have a relatively low impedance. So something like 1-10 µA should be sufficient.

The amplifier can be non inverting  -  so no need for the extra buffer.

Using digital decoding for the difference is a good idea, though it might need a little more resolution for the ADC: the difference is proportional to the absolute temperature. So a 8 Bit ADC would only get steps of a little over 1 K if the gain and offset are well chose.
Just direct voltage reading could give higher resolution as higher gain could be used.

True, but according to the datasheet if the transistor is at 250 uA the temperature coefficient is the most linear. I'm still tinkering with it, so I will try 10 uA as well.
What I want to do in terms of ADC is to obtain the 100°C ∆Vbe and he 0°C ∆Vbe. Then subtract that offset and amplify to the full scale of the ADC (Arduino's 10 bit).

Thank you! I had already read this application note, together with AN137.

Tektronix includes a detailed circuit description in their service manuals for the DM501, DM502, and 7D13 which is also worth reading.

Gonna check it this afternoon.

You are using 2N2907 as a temperature sensor, that's kind of a new thing for me and I still can't understand this how you are gonna implement it?

Btw when I need temperature sensor I use DS18B20. I think you are trying to design a new temperature sensor may be.

It's just a bit of an experiment. Thilese parts are absurdly cheap when compariled to a dedicated temperature chip. Plus, a the end of the day, a diode connected BJT is what most silicon based chips use, but with different kinds of control systems.

[StuartA]

Re the use of transistors as temperature sensors, am I right that they offer a more-linear response than a simple diode?

I built a thermometer using a simple diode (silicon signal) many years ago, and that has linearity adequate for my purposes over 0-100'C when calibrated. The advantage it may offer over a transistor is that you can configure it so that one lead projects forward and is then folded back again, thus forming a very compact 'sensing pad'.

I used it recently on a flanged resistor, together with K-type thermocouple sensor right along side. During fast heating, the diode was really fast to respond, always showing an appreciably higher reading than the thermocouple. During slow cooling, the two probes gave the same reading.

S

[Kleinstein]

Transistors are sometimes a little closer to an ideal diode than most real diodes, as the material tends to be a little cleaner. However there are plenty of diodes to choose from, so some can also be good enough.

In some cases the transistor case may be more convenient (e.g. TO126).