Monitoring Heatsink Temperature... / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Monitoring Heatsink Temperature...

1 2 3 4 next last

Move Thread

LAN_403

Ash Small

Thu Aug 23 2012, 04:42PM

Registered Member #3414 Joined: Sun Nov 14 2010, 05:05PM
Location: UK
Posts: 4245

I picked up some Constantan wire this morning for the type 'T' thermocouples required for this 'Projects' thread:

All was going fairly well until I realised that the moving coil meters I'm planning to use as temperature guages have an internal resistance of ~750 Ohms.

As the thermocouples only produce ~5milli Volts @100 degrees C and I require @100micro Amps this (I think) means I need to amplify the thermocouple signal using transistors. (I was hoping I could connect thermocouples in series (thermopiles) and get a sufficiently strong signal that way)

I've only been able to measure micro Amps generated by the thermocouples when heating with a lighter (significantly higher temperature than 100C) but I assume this is due to the internal resistance of my DMM, and not a reliable indication.

Can anyone suggest suitable transistors to try for signal amplification?

(BC109, maybe?)
]bc107_108_109_4.pdf[/file]

klugesmith

Thu Aug 23 2012, 06:36PM

Registered Member #2099 Joined: Wed Apr 29 2009, 12:22AM
Location: Los Altos, California
Posts: 1716

Any reason you can't use different meters, with coil resistances in the 10 to 100 ohm range?

A thermocouple's voltage output is very small, but it can generate current limited only by the resistance of the circuit.
Consider a gas pilot light safety valve, with a single copper-constantan junction pair. When the hot junction is kept at a few hundred degrees C, the thermocouple directly powers a solenoid that holds a gas valve open. A solenoid wound for low voltage, i.e. with relatively few turns of thick wire.

I made a handheld pyrometer using type K metals in the form of solid wires, about AWG10 (coat hanger wire size).
Chose an appropriate meter from a bin at the surplus store. Had gone there armed with a battery and voltage divider, designed to put out about 10 millivolts.
The pyrometer's full scale temperature was set to about 1200 degrees C by choosing a series resistor, IIRC in the middle tens of ohms, to bring the total resistance to a calculated target value.

Good luck.

[edit] Just spent some online search time looking for data about meter movements. Harder than I expected. Here's one line of analog panel meters, DC, available down to 15 mV full scale.

Steve Conner

Thu Aug 23 2012, 07:34PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

If you can't find a suitable meter, the best bet is a precision op-amp with low offset voltage, maybe the OP07. Believe me, you don't want to try making a discrete transistor amplifier. If you're curious, one of Bob Pease's books has a good circuit. Or maybe it's in the LM394 datasheet, I forgot.

I believe the old-school way was to convert the thermocouple signal to AC with a mechanical "chopper" and then step it up with a transformer.

Sulaiman

Thu Aug 23 2012, 07:40PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

For a thermocouple to read temperature accurately
a high impedance very low offset voltage VOLTMETER
and a reference 'cold junction' are required
together with some (non-linear)correction factors.

For heatsink temperature monitoring I would use an NTC thermistor
or my old favourite wan an LM35 - or a new version.

Dr. Slack

Thu Aug 23 2012, 08:08PM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

DIY metering of thermocouples is hard work. The two best options are a) buy a dedicated thermocouple pre-amp, search RS for "thermocouple amplifier" and you'll find several Analog Devices parts with varying accuracy for around Â£5 +/- . They do cold junction compensation and give an output with defined volts/degree scaling, just add a multi-meter or b) hit fleeBay for a Chinese knock-off thermocouple meter. I've just bought one for Â£15, it delivered better than +/- 1C out of the box, and can be calibrated with ice and steam to the probes. Don't buy a multimeter with thermocouple input - horrible accuracy.

NTC thermistors are good, as their output is huge. There is a cunning trick to get a substantially linear voltage output from them over a 20C range, and perfectly usable outside that. Choose a temperature range, for heatsinks you might pick 60C to 80C. Measure the resistance of the NTC at the mid-scale temperature, 70C in this case. Put a fixed resistor of that value in series with the NTC, and feed them from a constant voltage. Measure the voltage across the resistor. The voltage will be offset + scale*temperature. Or measure the voltage from the wiper of a pot across the voltage source to the R/NTC junction, which will allow offset to be adjusted to zero. Or you can measure the voltage from mid-voltage-source to the R/NTC junction. The voltage will be +/- scale*(temperature-mid_temperature).

Nice to see the heatsinks in use so soon

Sulaiman

Thu Aug 23 2012, 08:36PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

Most industrial temperature monitoring systems use type 'J' or 'K' thermocouples,
or a PT100 resistor.

Cheap dmm with thermocouple input are not absolutely accurate but usable.
Set the dmm to thermocouple input, short the terminals and read ambient temperature.
If a few meters give the same reading +/- 1C I'd call that good enough.

your type T will have about 45 uV/C sensitivity
a single silicon diode or base-emitter junction voltage varies by about -2 mV/C ... 44x more!

Ash Small

Fri Aug 24 2012, 06:50AM

Registered Member #3414 Joined: Sun Nov 14 2010, 05:05PM
Location: UK
Posts: 4245

Thanks for all the replies.

klugesmith wrote ...

Any reason you can't use different meters, with coil resistances in the 10 to 100 ohm range?

I need four meters, one for each diode in the rectifier. I have these ones, however I may have to consider alternatives if I can't get these to work satisfactorily.

Sulaiman wrote ...

For heatsink temperature monitoring I would use an NTC thermistor
or my old favourite wan an LM35 - or a new version.

Sulaiman wrote ...

you should get a (cheap) multimeter that can use a thermocouple,
make or buy a few thermocouples
(you can use themocouple extension wire to make thermocouples)
and measure the temperatures of
- ambient air
- transistor case (the top part will do for static measurements)
- the heatsink as close to the transistor/heatsink interface as possible
- the temperature of the heatsink far from the transistor
(the last two will require drilling into the heatsink)

It was this post that got me thinking about TCs in the first place.

I don't need it to be linear, or particularly accurate. I just need ~100uA @ 100C (or 80C above ambient), so that the meters go into the red at around 100-120 C.

Using moving coil meters allows me to easily see if anything is amiss.

Sulaiman wrote ...

your type T will have about 45 uV/C sensitivity
a single silicon diode or base-emitter junction voltage varies by about -2 mV/C ... 44x more!

Thanks for pointing that out. It sounds like using silicon to amplify the signal is out of the question then.

Steve Conner wrote ...

If you can't find a suitable meter, the best bet is a precision op-amp with low offset voltage, maybe the OP07.

Will this have the same limitations as silicon, ie -2mV/C ?

I was originally thinking of putting the meter (in series with a resistor) in parallel with a resistor that is in series with the collector/emitter, then putting a third resistor in series with the TC, and connecting to the base/emitter.

If I choose the values of the resistors carefully, and the transistor has sufficiently high gain, small variations in the silicon resistance shouldn't be significant (I think!)

(I'll draw a schematic later, if the above description sounds confusing)

Shrad

Fri Aug 24 2012, 07:45AM

Registered Member #3215 Joined: Sun Sept 19 2010, 08:42PM
Location:
Posts: 780

you should read this appnote from linear

it is really well written, and contains decent examples with complete schematics

klugesmith

Fri Aug 24 2012, 02:44PM

Registered Member #2099 Joined: Wed Apr 29 2009, 12:22AM
Location: Los Altos, California
Posts: 1716

Shrad wrote ...
you should read this appnote from linear

it is really well written, and contains decent examples with complete schematic

I agree, that's great. Don't miss figure 2, which compares TC's with other kinds of electronic temperature sensors. RIP, Jim Williams.

With respect, your single transistor (or 2 discrete transistor) approach with a TC and analog meter would be a waste of time. You have V_BE variations in series with TC voltages.
Re. simple voltage amplifiers using a low-offset-voltage quad op amp: will your TC cold junctions be at a constant temperature, within your desired instrument accuracy?

Since you have 4 meters already on hand, I recommend Dr. Slack's solution using NTC thermistors and a handful of R's. You can use a variable R to calibrate each meter at one temperature point, such as 100 C.

In defense of my three-component (including the TC) pyrometer solution: It served well for many years in my backyard foundry. (Until I broke the hot junction. That had become stuck in Hawaii, when lava at 1800 degrees F crusted over.) Anyway, the type K nonlinearity between 60 and 2400 F can be ignored, or taken care of with nonlinear scale on the meter card. For cold junction compensation, I used the meter movement's offset screw to make pointer indicate the known ambient temperature.

radiotech

Fri Aug 24 2012, 06:25PM

Registered Member #2463 Joined: Wed Nov 11 2009, 03:49AM
Location:
Posts: 1546

The classical method to read thermocouples with traceable calibration was to use a
potentiometer, with a standard cell (Weston) as reference. If you are interested
look up Leeds and Northrup history.

However today with Arduino-like processors, you could set up a four
input data logger for a few dollars. An added bonus, it also could log other
parameters of the rectifier system for correlation.

All you need is a usable signal, i.e. in the millivolt range.

Large processes monitored hundreds of temperatures with miles of
bi metalic signal cables strung everywhere. Like the shield probes on the
Columbia, some were more important than others.

1 2 3 4 next last

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob