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

Forums

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

« Previous topic | Next topic »

Monitoring Heatsink Temperature...

previous 1 2 3 4 next last

Move Thread

LAN_403

2Spoons

Fri Aug 24 2012, 09:13PM

Registered Member #2939 Joined: Fri Jun 25 2010, 04:25AM
Location:
Posts: 615

I like the AD590, which is a current output of 1uA/K. Only two wires, which could be any length as its a current output - wire and connector resistance have no effect. In your case you would either rescale your meters to read ~400uA FS or add a 273uA offset (AD590 outputs absolute temperatue)

Ash Small

Tue Aug 28 2012, 09:49PM

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

I've been away camping and watching power boat racing over the bank holiday weekend so I've not got around to replying until now.

Thanks for all the replies. I'll consider each in detail before making a final decision, however, in the meantime I've come up with the following idea to (possibly) overcome the issue mentioned by Klugesmith (and others), and would appreciate any comments:

klugesmith wrote ...

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.

If you build the following circuit, with both transistors biased to be switched on halfway:

Then connect a thermocouple accross the bases, one will switch more current than the other:

If you then connect a moving coil meter accross the two collectors, the meter will deflect. The Vbe variations will cancel out.

Comments please (I'm sure the solution can't be this simple).

2Spoons

Tue Aug 28 2012, 10:33PM

Registered Member #2939 Joined: Fri Jun 25 2010, 04:25AM
Location:
Posts: 615

For that to have even a slight chance of working the two transistors must be extremely well matched, in the same physical package, preferably on the same die. Such things do exist but are not that cheap. e.g MAT12 made by analog devices, is over $20 at Digikey.
The base bias resistors are going to have to be high values (they load the TC) and very accurate.
Also there is no cold junction compensation.

What you have drawn is very similar to the differential front end of a bipolar op-amp (minus one or two important features). So why not just use an op-amp?
Or better still use the purpose designed chips that include cold junction compensation (available at ~$2 up). There's plenty to choose from.

Dr. Slack

Wed Aug 29 2012, 07:15AM

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

Let's spell it out. The TC is putting out uV, microvolts. Each transistor has a VBE tempco in the order of mV, millivolts, 1000x bigger. As the TC makes the transistors take different currents, the transistors will dissipate different powers, and so end up at different temperatures. How different will depend on the thermal detail of how the circuit is built, drafts, the rail voltage, the meter load. If you build one circuit, and mount it in a thermally stablised container with an accurately stablised power supply, you could perhaps calibrate it. If you build another one, it's likely to be different. If you use it with a different meter, it will be different yet again.

If you buy a TC conditioning amplifier, then it doesn't need calibrating, will work with any meter, at any voltage, and if you build another it will be the same.

Ash Small

Wed Aug 29 2012, 09:58AM

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

I found a paper here:

that I'm currently reading.

The non-linearity of type 'T' thermocouples is a bonus for this application.

The only thing I need to accurately calibrate for this application is the approximate temperature that the meter goes into the red, so I dont really need to compensate for reference junction variaton either.

Steve Conner

Wed Aug 29 2012, 12:51PM

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

Congrats on your almost-reinvention of the long-tailed pair, but it won't perform well enough as Dr. Slack explained. I have some experience designing input stages for audio power amps, and it is tough to get the offset voltage below some tens of millivolts, let alone microvolts.

Ash Small

Wed Aug 29 2012, 04:31PM

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

Steve Conner wrote ...

Congrats on your almost-reinvention of the long-tailed pair,

Thanks Steve

.

I've been reading up on the AD8497.

Although it is designed for type 'K' tc's, it should be ok for my application as I don't really want linear amplification. It looks fairly simple to use.

Ideally I want this output:

so that at lower temperatures within the operating range the meter doesn't move much, and moves more as the temperature approaches ~100C.

From what I can gather from the output, 5mV/degree C through a 100k Ohm resistor, I may need to either further amplify the signal or use two or three thermocouples in series, unless I can use a lower value output resistor, as my meters are 200uA FSD.

They are cheap and available, RS have them for Â£3.59 each or Â£27 for ten.

Steve Conner

Wed Aug 29 2012, 04:37PM

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

Yes, I'm pretty sure you can use a lower value output resistor. 100k was chosen to get a low cutoff frequency for the filter without an excessively big capacitor. A moving coil meter doesn't need the filter, it reads the average value so it is filtering the signal already.

If this were my project, I'd take a step back and question the choice of a type T thermocouple in the first place. The solid-state temperature sensing ICs (LM35DZ, AD590 etc) are cheaper than the thermocouple amplifier chips.

Ash Small

Wed Aug 29 2012, 05:43PM

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

Steve Conner wrote ...

The solid-state temperature sensing ICs (LM35DZ, AD590 etc) are cheaper than the thermocouple amplifier chips.

The AD590 range are all a lot more expensive from RS.

The LM35DZ is half the price but is limited to 100 C.

The diodes I have are rated for 300 A at 125 C, but will run at 180 C.

I want the meters to enter the red @ ~100-120 C.

Also, I'd need to drill larger holes in the heatsink to accomodate the transistor type packages.

It looks to me like the AD8497 is a better solution for this application.

Avalanche

Wed Aug 29 2012, 07:13PM

Registered Member #103 Joined: Thu Feb 09 2006, 08:16PM
Location: Derby, UK
Posts: 845

I built a TC amplifier a few months ago around an LTC1050 chop-amp

which was a bit OTT in terms of performance and price, but it was a one-off project and I needed it to work!

If you're interested in why you can't just use a single thermocouple to measure temperature, I'd recommend reading the physics of the 'seebeck effect' - here's a link

It's actually impossible to only have one junction anyway, a 'thermocouple' will only give a voltage proportional to the difference between 2 junctions. The TC amp I built at work around the LTC1050 had an arrangement of thermocouple wire as shown in the diagram on that wiki page, the hot junction was attached to an IGBT baseplate, and the cold junction was on the melted surface of a block of ice. With software compensation for non-linearity, we got a ridiculous accuracy! With one junction embedded into the IGBT and the other in the heatsink, the TC amp gave a voltage proportional to heat/power flow through the baseplate

previous 1 2 3 4 next last

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