Heatsink temp approximations ?!

[Al (astral highway)]

I’ve just ordered an infra-red thermometer after too much sweating over heatsink temps.

But does anyone have any ready estimates? I know my domestic radiators are 65C. But what on earth does 125C feel like ? I know it’s a seething that will still coo food in an oven!

I wonder if anyone has measured the temp of a high power amplifier’s output valve envelope/ line output valve or glass-envoleoped transmitting valve for comparison ?

[trh01uk]

125 degC is going to b****y painful! Well it will be if you can keep your fingers there for more than a few hundred millisecs!

Ive used an infra-red thermometer many times to measure hotspots in electronic equipment. It works a treat. You can see whether your power transistors are properly losing heat to the heatsink for one thing - because if the transistor case is at 120 degC and the heatsink is only at 100 or less, then you know the thermal transfer between the two is looking a bit dodgy.

Richard

My "good enough for engineering" method is to spit on the heatsink (either directly or via a wet finger) a fair sizzle is about 110C, much hotter than that it is too hot and it pops rather than sizzles. Lower temperatures lead to a swift evaporation rather than sizzling.
In the end if it only sizzles or less silicon (silicon carbide transistors work at red heat) semi conductors will be fine, valves on the other hand, if it isn't glowing (the anode) much, fine too.

Infra red thermometers assume the surface is a black body radiator and do give errors on other surfaces. One of the best I have found is masking tape, OK it looks off white but is blackish at IR, stick a bit on for measurement. At the very least measurements will be consistent.

[Electronpusher0]

I always used to reckon that 50 deg was the temperature that you could just keep a finger on. Anything over 100 deg would boil water so a drop of water should sizzle and steam.
Don't forget that 125 deg is the max junction temperature of most silicon devices, you have to factor in the thermal resistance junction to case, case to heatsink and heatsink to air. Without wishing to teach my grandmother to suck eggs these are best thought of like series resistors. The thermal resistances are in degC/Watt so given the power dissipation in the device multiply this by each of the thermal resistances and add them to get the total temperature drop junction to air.
By the way it is a fallacy that there is a temperature gradient across a heatsink. While there will be a very small gradient for all intents and purposes the whole heatsink will be at the same temperature (for any commercial heatsink that is, a flat sheet of thin ally will have a gradient.
Peter

[G6Tanuki]

Rather than instantaneous temperature, what really matters for both valve and semiconductir stuff is the peak-temperature-reached.

When I was building power-amps professionally, they would always go out with non-reversible temperature-indicating labels stuck in strategic places - these sort of things: https://www.omega.co.uk/prodinfo/tem...re-labels.html

which meant that if any of them failed and I could tell they'd been 'cooked' then in the absence of an obvious fault it became a chargeable repair rather than a warranty-one.

I've used similar labels in all sorts of places, like measuring peak oil/coolant/brake-disc temperatures on race-cars and detecting intermittent overheating of central-heating systems.

[Al (astral highway)]

Originally Posted by merlinmaxwell

My "good enough for engineering" method is to spit on the heatsink (either directly or via a wet finger) a fair sizzle is about 110C, much hotter than that it is too hot and it pops rather than sizzles. Lower temperatures lead to a swift evaporation rather than sizzling.

That's really useful! I was feeling anxious that I was on the edge of ok and tried the sizzle test. The drop evaporated quickly. If you hadn't posted this, I would have guessed that evaporation happened at higher than sizzle temp, so this is great to know.

I've been cautiously keeping the heatsink cool after each test with a freezer block (one of those blue things).

Infra red thermometers assume the surface is a black body radiator and do give errors on other surfaces. One of the best I have found is masking tape, OK it looks off white but is blackish at IR, stick a bit on for measurement. At the very least measurements will be consistent.

This is also really helpful, thank you!

[Al (astral highway)]

Originally Posted by Electronpusher0

Don't forget that 125 deg is the max junction temperature of most silicon devices, you have to factor in the thermal resistance junction to case, case to heatsink and heatsink to air. W

Sure, good points though.

Junct temp of some is 150 degrees, astonishingly.

[Al (astral highway)]

Originally Posted by G6Tanuki

Rather than instantaneous temperature, what really matters for both valve and semiconductor stuff is the peak-temperature-reached.

When I was building power-amps professionally, they would always go out with non-reversible temperature-indicating labels stuck in strategic places

Now these are good to know about, too! Cheers!

[Herald1360]

Originally Posted by astral highway

Sure, good points though.

Junct temp of some is 150 degrees, astonishingly.

Metal can stuff like TO3 can go to 200C Tj!

[Boulevardier]

Rod Elliot has one of the most comprehensive discussions of heatsinks on his site - http://sound.whsites.net/heatsinks.htm

I did quite a bit of experimenting - tried using a thermocouple probe thermometer, and the problem with these is that they are really intended for immersion in a liquid. Nevertheless, there does seem to be a "sweet spot" near the probe's tip which will give a reliable reading if you can accurately locate it. Trouble with IR thermometers is that they depend on the emissivity of the surface, which varies with texture and colour. I seem to recall that the posher models even give correction factors for different types of surface. As others have said, a piece of masking tape stuck to surface gives reasonably accurate readings, but what you measure will also vary significantly according to which part of the heat sink you measure it on.

At the end of the day, I think it's a very imprecise business, with big margins of error often built in to the calculations. Suck it and see seems to be the best advice (though don't take that too literally!)!

[Refugee]

I have seen a heat sink with a little hole drilled into it with a blob of heat sink compound in it for placement of a thermocouple.

[Diabolical Artificer]

I've seen some good reports of the Flir temp device, the one that clips into a smartphone and uses an app. Not cheap though.

Andy.

[ionburn]

I use a contact thermometer for getting an idea, but generally try to over-engineer to avoid the problem when possible. I find rate of change of temperature useful, but the total heat path and bottlenecks are important to realise. Heat generally means reduced reliability.

[mole42uk]

Originally Posted by Diabolical Artificer

I've seen some good reports of the Flir temp device, the one that clips into a smartphone and uses an app. Not cheap though.

It is an IR camera. I use one at work to find hot spots on vintage synthesisers. It’s quite a handy tool and the phone app also indicates the temperature at the focus of the screen display.

I frightened pals in the Triumph Stag owners group by measuring the temperatures in the engine bay of my car, I was asked to desist!

[MrBungle]

Spit on the finger works. If it sizzles it’s above 100 degrees. If you burn your finger it’s way above 100 degrees. If it burns and hurts ten times more than the hot ones it has RF on it

I have a decent DMM with thermocouple as well. Just blob some Vaseline on the end of it and touch it on the heatsink for an accurate measure.

I’d live an flir camera but too expensive. Will wait until the Chinese have worked out how to knock one out for a pony.

[G0HZU_JMR]

You could also make a cheap and basic temperature sensor from a BJT B-E junction or a diode PN junction and permanently attach it to the heatsink. It would need to be crudely calibrated but it could be used with a comparator (or two) to set green/yellow/red LED status for the heatsink. Or you could read the PN sensor(s) with a basic Arduino PCB and a £3 LCD display.

At work we tried several IR 'spot' temperature meters for general electronics work but ultimately threw them all away because they weren't reliable. Some have large spot areas and some only work well at a set distance and it was never clear if the surface being tested was giving a reliable result. However, these meters will probably be OK for a large heatsink as long as you stick some tape on it if the heatsink has a bare metal (reflective) finish.

We upgraded to a thermal camera at work and this was much better as a small pad could be stuck on any surface and this could be used with a precise marker on the camera display to measure the temperature. But for something like a heatsink the PN junction sensor or the 'temperature strip' indicators already mentioned can be very useful and they are cheap and easy to use.

[Refugee]

Small audio amplifiers often used a PN junction as a bias stabiliser.
Obviously the transistors had to be the same material as the bias diode.
It was very common with germanium transistor amps in things like record players.

[Al (astral highway)]

Jeremy , thanks , that’s an ingenious idea , I’m going to try that out.

Everyone: thank you, too.

[G0HZU_JMR]

It's a well known method of making a cheap sensor and I think it's worth experimenting with the B-E junction of a transistor rather than a regular diode like the 1N4148. I think you could start by biasing the diode junction from 5V via a large resistor to set it at a light (but essentially constant) current and then just read Vbe. I'm going to guess you end up with a y=mx+c calibration line where the gradient will be something like 2mV/degC. So you need to be able to measure down to a few mV and an 10bit AVR ADC referenced to 2V should be able to do this. It will go a bit finer with decimation. However, it's best to read up on a few app notes where people have done this for real.

At work we often use the LM50B or LM50C temperature sensor and these are cheap and are pre calibrated up to +125degC. It will also give a higher gradient/slope so it will be easier to use with a basic comparator and a couple of LEDs. Not as much fun as trying to calibrate a cheapo homebrew PN junction sensor but they do work well.

[trsomian]

There are also TO92 "one wire" temperature sensors from Dallas that have 1/2 degree resolution. Obviously a readout unit is required, but designs for those and the code is available freely on the internet