measureing temperature of power resistors using current.

[Tonkabot]

I have a TO220 power resistor used to heat a gas manifold.    The resistor is 50 ohms, we have a fixed 24v supply.
I can turn on and off the power to the resistor, and I can measure the current through the resistor.   I can also calculate the number of Ohms the resistor currently has.

I want to use this current feedback to get an idea of how hot my gas manifold is, because of course we know the resistor value changes with temperature.

How accurate temperature measuring can expect to get out of this system?   16 bit ADC on the current.

Thanks!

[Benta]

because of course we know the resistor value changes with temperature.

How do you know this? Normally, power resistors are optimized to keep a constant resistance over temperature. Is this a special type?

[Tonkabot]

Well, I don't know it.   I assumed it...
But looking at the power resistors specs, your right, they design for a low temperature coefficient.

I would need resistors that are not optimized for a low Tc.   Or better yet,  something like a TO220 PTC or NTC  power thermistor.

[TimFox]

It would be far easier to use a calibrated thermistor probe, or a thermocouple or RTD sensor, or a semiconductor device (for lower temperatures).
See  https://www.omega.com/en-us/resources/integrated-circuit-sensors  for a range of solutions.

[Benta]

I agree with TimFox. Keeping heating and sensing separate is a much better and safer solution.
BTW, which kind of "gas manifold"? That could mean anything.

[wraper]

The issue is not just a low Tc but how resistive element is formulated for that. Depending on proportion tolerances of components, resistors of the same model may have negative or positive Tc which is not stable across temperature range and may even reverse from positive to negative at some temperature.

[Tonkabot]

We don't want to change the circuit board, there are hundreds in the field already.  So we only have the two wires to the heating elements. 
It occurs to me that I could make a copper spiral on a PCB that is the perfect size to heat the manifold, and make it the right length and width to give me a useable resistance (I need probably 15 ohms or so).  plus it will have the Tc of pure copper, so I can make a crude temperature element by reading the current at my known voltage.   There is a current sense amp on the circuit that gives me a good resolution at 16 bits.

The gas manifold is actually a CO2 regulator.  We want 35 PSI, but when supplied with a 700 PSI bottle the regulator does its job and also sucks heat out of the regulator.   I have around 50 watts of power I can use as heat, to keep things from freezing up.   The only temp sensors are on the main board, and there are 3, so if things get too warm we can have an emergency cut off.

[Kleinstein]

How good the temperature measurement from the resistor change works depends on the TC of the resistor and also the TC of the resistor used to measure the current. Adding the circuit to measure the current (and compensate for voltage changes) would already need a circuit revision. The circuit of choice would be a kind of asymmetric bridge and difference amplifier. The ADC resolution would likely not be the limiting factor, more the resistors.

A problem may be the the TC of the resistors is not all the same. Worst case some resistors can have a more parabolic curve instead of a more linear that is suitable for a temperature measurement. For power resistors the TC is usually not well defined and can scatter quite a bit, though they may still be all on one side. It may still be good enough for a crude regualtion (e.g. to some +-10 K) to avoid freezing / overheating.
Another problem is possible long term drift of the resistor, especially if the resistor at times also runs hot. Ideally one may have to do a check periodically at room temperature and no heat / gas flow.

Things would be much better with a defined TC and in this case high temperature resolution would be possible.

[Berni]

The temp co. of power resistors is not consistent from batch to batch and can be significantly affected by the wires it is hooked up with (copper has a pretty positive temp co.)

This can be done with some heating elements that are specifically designed for a very high positive temp co. . That way the heating element can crudely regulate itself, since as it gets hotter the resistance climbs and so it draws less current and heats itself less.

Still this is the easiest to do by just sticking a NTC temperature sensor on it. Those are designed to have a huge and consistent temp co. making it easy to read the temperature using just simple a voltage divider circuit.

[Marco]

We don't want to change the circuit board, there are hundreds in the field already.  So we only have the two wires to the heating elements. 
It occurs to me that I could make a copper spiral on a PCB that is the perfect size to heat the manifold, and make it the right length and width to give me a useable resistance (I need probably 15 ohms or so).  plus it will have the Tc of pure copper, so I can make a crude temperature element by reading the current at my known voltage.   There is a current sense amp on the circuit that gives me a good resolution at 16 bits.

It's possible, but will require some calibration. Traces aren't precision instruments.

PS. If the trace is long you might want to put into a non inductive shape (ie. one turn clockwise, one turn anti-clockwise etc).

[Siwastaja]

Copper heating element is usable because copper has large temperature coefficient. You could easily get some +/-5degC reading with 4-wire (Kelvin sensing) and some calibration.

But power resistors do not usually use copper elements, one of the reasons is the high temperature coeff.

[Doctorandus_P]

So what is your range of acceptable stuff? TO220? wires? Anything that has just two connections?

Soldering irons sometimes use resistance wire for the heater in series with a thermocouple.

There are also of the shelf wires for heating. For example carbon fiber with silicone insulation is common for electrical floor heating. I have not checked what the temperature coefficient of that (and similar) wiring is.

[Tonkabot]

Copper heating element is usable because copper has large temperature coefficient. You could easily get some +/-5degC reading with 4-wire (Kelvin sensing) and some calibration.

But power resistors do not usually use copper elements, one of the reasons is the high temperature coeff.

+-5C would  be plenty good enough for our purposes.   although I don't have a kelvin connection, just the two wires.   

A circuit board with a heater trace (in a non-inductive pattern) would be probably as cheap as a power resistor.

I also promise to push harder for putting extra IO on the next board.   I was talked out of it by the customer (who quite possibly is not always right).

[oz2cpu]

with the setup you got, so many factors play in.
I would just take a few normal use caseses and measure temperature, and heater on/off times,
and then perform a "normal" corelation between pwm on/off value and the normally resulting temperature,
and then use that in your product, since supply voltage is constant and known, and resistance is know and not changed a lot,
all you need is the on/off time..
if there is another sensor that could at least tell the ambient temperature to the software,
you can use that value to adjust a bit with.

[TheMG]

Another point to consider is the thermal resistance between the resistive element and the manifold being heated. While the resistor is heating, it will always be some temperature hotter than the manifold itself, and this temperature difference will be greater the more power is being dissipated in the resistor.

This makes the whole thing wildly inaccurate. You would have to periodically turn off the current to the resistor for a little while to let the temperature stabilize before then measuring the resistance using a small test current. Even then you still have to account for temperature coefficient, but at least that way don't have to compensate for thermal resistance, which is going to vary from unit to unit as it is affected by mounting pressure, thermal compound application, etc.

[Tonkabot]

Another point to consider is the thermal resistance between the resistive element and the manifold being heated. While the resistor is heating, it will always be some temperature hotter than the manifold itself, and this temperature difference will be greater the more power is being dissipated in the resistor.

This makes the whole thing wildly inaccurate. You would have to periodically turn off the current to the resistor for a little while to let the temperature stabilize before then measuring the resistance using a small test current. Even then you still have to account for temperature coefficient, but at least that way don't have to compensate for thermal resistance, which is going to vary from unit to unit as it is affected by mounting pressure, thermal compound application, etc.

I don't need accuracy, I just want a rough reading of the temp.   I also have the ability to do exactly what you said, turn off the heater to let it's temp equalize with the manifold, then read the current and calculate the heater resistance before it heats up again - good plan to make it more accurate.

The whole goal here is to just make sure we don't get frost forming on the gas lines, and not have the CO2 so cold that the regulator and solenoid valves still work correctly.  There is a another heater at the end of the gas path that does have accurate temperature reading which is under much closer control .

[Conrad Hoffman]

People measure the internal temperature of transformers all the time based on copper resistance. The same method should work well for you. You could also use iron or aluminum wire for a bit higher sensitivity.

[TimFox]

If a thermostat control (heat on below setpoint, hysteresis to higher temperature for heat off), a simple two-wire fix is a Klixon^TM switch in series with a suitable heating element.
https://www.sensata.com/products/temperature/7bt2
Buried somewhere in their literature is a wide range of standard temperature ratings.

[inse]

My question would be whether you want to measure the temperature of the resistor or the temperature of the gas?
How fast thermal response do you need?

[Terry Bites]

Ha Ha, that takes me back.
See page 48 of the attached [ Attachment Invalid Or Does Not Exist ]
Bad idea then, bad idea now.
I also found in ths pdf a photo of something I designed. But I'm not telling you what page it's on.
Lithium battery tech, wow!

Back to reality. If you want to go this way then there very high accuracy current sense amps like the LTC6102.  A MAX9612 has a digital ouput and a spare opamp to monitor the voltage on your resistor. Its about $4.