metal vs plastic transistors

[jneutron]

Someone needs to do some test on nominally identical devices in different packages with various mounting methods (direct, thin insulator, thick insulator, screws, clamping bar, clip). All that needs to be measured is the heatsink temperature, which will allow relative thermal conductivity die-to-heatsink to be calculated. No need for absolute measurements. I might do it if someone can suggest some suitable devices that are cheap.

Interestingly, Exicon do their MOSFETs in TO3, TO247 and TO264 cases. TO3 devices go up to 250W. Equivalent TO247 up to a mere 125W. TO264 up to 250W (same as TO3). Their datasheets say 1.0C/W juntion-to-case thermal resistance for all three packages, but I don't think I trust their numbers.

Your method doesn't measure the die temp, just the reaction of the heatsink to the power influx..

What you have to do is measure the die temperature while the device is operating..

I did this a while back...well, ok...15 years ago...what you do is run the device in active mode, and every second or so, shut the power off for 100 uSec or so, keeping 10 milliamps base to emitter forward direction. This junction will have a temperature dependence, and since this junction is directly above the collector base junction, will provide a very good indication of the active junction during dissipation. You have to use 100 uSec because any longer, and the thermal constant of the silicon to spreader or base will cool the junction down.

Then, you get smart, and use that measured temperature to control the power dissipation. That way, you can force the die to a specific temperature, and measure the power required to do so. The neat part here, is if you are pushing the device to 150 C, and then you disconnect the heatsink, the circuit will know it by the die temp, and the power will drop instantly, keeping the die at setpoint temp..

To cal this setup, just heat the device to the desired temperature, like 150 C....turn on the 10 milliamp measuring current....measure E-B voltage, then have the circuit try to maintain that E-B voltage as the device is coming down in temp..as the sink cools, the power will rise.this will give you an exact derating curve for the heatsink attach used.

If anyone wishes to build such a contraption, I can give more details..

Oh, btw...it'll work for any active device..all you have to do is measure a temp dependent variable, like the drain-source body diode, collector-base. I'd stay away from gain as a measured variable, as that isn't a very good measure of temp for different operating currents.



Cheers, John

 

[Mr Evil]

Your method doesn't measure the die temp, just the reaction of the heatsink to the power influx..

What you have to do is measure the die temperature while the device is operating...

You're right. I was thinking of something else.

 

[Nelson Pass]

Keep in mind that fatigue is a major cause of long term
failure, and as you cool the case of the transistor, creating
a greater internal temperature differential, the mechanical
stress increases.

So maybe -65 C. for the case and 150 C. for the die is not
such a good idea after all.

 

[jneutron]

Keep in mind that fatigue is a major cause of long term
failure, and as you cool the case of the transistor, creating
a greater internal temperature differential, the mechanical
stress increases.

So maybe -65 C. for the case and 150 C. for the die is not
such a good idea after all.

Ummm, it was 200 C junction , and 379 watts...what's wrong wit dat??? Dry ice neva hoit nobodies...

Wimp..

Hi Nelson...

Yah, cold is the worst for the die attach...and as the die heats, the differential shear forces will increase if you really force the case to remain cold.

At least, the relative TCE difference between the copper and silicon are in the right direction..the silicon gets hotter than the copper, but it expands less per degree C, 3.3ppm vs 16 ppm.

I don't know at what power density the silicon will cleave from the vertical thermal gradient, either. But I would suspect you'd have to go up into the kilowatt regime to find out.

Cheers, John

 

[jacco vermeulen]

Keep in mind that fatigue is a major cause of long term
failure, and as you cool the case of the transistor, creating
a greater internal temperature differential, the mechanical
stress increases.

That would suggest:

* an amplifier output stage is better off with many devices and a high heatsink temperature.
* a class A output stage is longer lasting because of constant
temperature, or is temperature level still more dominant ?

 

[jneutron]

That would suggest:

* an amplifier output stage is better off with many devices and a high heatsink temperature.
* a class A output stage is longer lasting because of constant
temperature, or is temperature level still more dominant ?

Semiconductor die attach has advanced sufficiently that the stresses that are there at room temperature are not significant.

Just tossing more devices at the problem and running them hot will create other reliability problems..

Keep em as cool as reasonable, and just make sure the SOA isn't approached under any conditions.

Cheers, John