# thermistor compensation

Discussion in 'Electronic Design' started by daveatfernie, Apr 26, 2006.

1. ### daveatfernieGuest

Hi everyone,
I have some 0402 surface mounted thermistors which are causing me some
bother. I am measuring an object in contact with the thermistor but am
obviously measuring the temperature inbetween that of the PCB and the
object. I would like to use another thermistor placed in the circuit
(that will measure the temperature of the PCB) to allow me to offset
this middle temperature to compensate. Can anyone help with a circuit
example or ideas?

David

2. ### Phil HobbsGuest

Given the very small thermal mass of an 0402, and the huge thermal
signal error is going to be dominated by the quality of the thermal
contact between the 0402 and the object being sensed. Given that,
you're going to have real problems getting a good calibration from a
second thermistor.

Think about it as a voltage divider problem: analogically speaking,
your sensor has a 1-ohm resistor go ground across its input, and you're
connecting it to an unknown voltage (the object temperature) across a
resistor of poorly controlled resistance that is certainly a good deal
larger than 1 ohm (the thermal contact).

This is not a recipe for a good measurement, regardless of how closely
matched the two thermistor curves may be. You ought to figure out how
to insulate the thermistor from the board.

Cheers,

Phil Hobbs

3. ### daveatfernieGuest

At the moment my jig is conducting the heat to the sensor well and
repetitively. (I have a look up table to convert the resiatance into
degres C and I'm getting a value approaching 0.1 degree C Accuraccy)
I can read the temperature accurately with the sensor but it does vary
with ambient.

I have tiny PCB tracks (minimum my manufacturer will let me have 0.127)
to try to minimise heatsinking.
Unfortunately insulating an 0402 from the PCB is nigh on impossible.
However I can improve the thermal conductivity of the ambient
compensation sensor to the PCB (to accurately measure the temperature
of the PCB).

The offset of ambient is the final problem really. My thermistors have
less than 1% variation.

The layout is somthing like below:

Object
Thermal paste
[ Real sensor ]
================================PCB
[ambient sensor]

Do you think I'm fighting a loosing battle?

4. ### John PopelishGuest

Do you have room to route a slot on each side of the sensing
thermistor? I think if you could do that, and add a patch of thermal
insulation (foam tape) to the back side of the board (and move the
ambient sensor outside of that thermal insulation) you could improve
your accuracy by an order of magnitude or more.

5. ### John LarkinGuest

If you add the second thermistor and calibrate the fudgefactor
somehow, you should be able to improve accuracy by, I'm guessing, 5:1
or so. You'll have to do the thermistor curve thing to each thermistor
independently, giving two temperatures, and then apply the fodge.

I've seen similar stuff done in poorly-designed ovenized crystal
oscillators, like from S**.

John

6. ### daveatfernieGuest

Many thanks to both Johns,
I'll look in to routing some of the PCB away. It is the fudge
suggested by John Larkin I would prefer to add but electrically using
perhaps a potential divider circuit but am having trouble getting my
head around what I'm after. As the ambient increases the target sensor
reads closer to the correct temperature so i need to further decrease
the resistance of the target sensor using the ambient sensor.
Many thanks again

David