Power dissipation at high ambient temperatures

[Demishrike]

Hello,

This is my first post on this forum and I better warn you that there maybe some poor grammar and spelling mistakes. Given what I had to aggree to in order to register I'd like appologies in advance to all those people who are offended by such things.

I'm looking at a problem with a PSU that may end up operating with a high ambient temperature in the enclosure and I'm looking for some advise on calculating expected junction temperatures for various components.

The high ambient temperature is cause by a high current, (200 A), circuit that shares the enclosure. I've done some testing at 23^{\circ} and get around a 30^{\circ} rise with 100 A in the high current circuit. The unit needs to operate with an ambient temperature of 70^{\circ}, so I believe this gives me an enclosure ambient of 100^{\circ}. I can't supply 200 A so can't know the temperature rise at this level but I assume that the temperature rise will be two to four times greater than that at 100 A. Which means that before I even energise the PSU the junction temperature rating of the components could be exceeded.

I know about themal resistance and calculating junction temperatures as normal; e.g. T_{j}=T_{a}+(P \times R_{ja}).

It was put to me that the junction temperature can be lower than ambient and the heat flow will then be from ambient to the junction. Hence the junction will not rise above ambient.

Does anybody know if this is correct?

This does seem to make some sense in that generally things flow from hi to low. However, my basic understanding is that the energy dissipated by the circuit can be view as an ideal current source, (where Watts is equivelnt to Amps), such that no matter how high the ambient temperature gets the junction temperature will always be higher.

 

[MrSparkle]

With a steady state temperature and the circuit turned off, Tj = Ta. So yeah, Tj will always be higher when the circuit is turned on, excluding some conditions like when you take the circuit from a cold basement outside in the midsummer heat. Tj will be only lower than ambient temporarily.

 

[Demishrike]

Thanks for the reply, I will have to go back and have the discussion again as some people at work don't think this is the case; as they are my superiors I wanted to make sure my own understanding is correct.

 

[jim hardy]

I know about themal resistance and calculating junction temperatures as normal; e.g.
Tj=Ta+(P×Rja)...
... generally things flow from hi to low.

Heat flows from hot to cold.. your intuition is correct.

It was put to me that the junction temperature can be lower than ambient and the heat flow will then be from ambient to the junction. Hence the junction will not rise above ambient.

? How would heat then get out of the junction ? Is it a Peltier cooler ?

Sounds to me like somebody is "messing with the new guy "..

 

[meBigGuy]

Junction temperature is always lower than ambient when the junction is turned off and ambient is rising :)

Also, you can think of it in a twisted way, that when the ambient rises it transfers energy to the active junction causing its temperature to rise. But the reality is that when the ambient rises, the junction has to rise to maintain the same energy transfer into the ambient.

For a short period of time the packaging material between the junction and the ambient might receive some energy from the ambient since it takes time for the system to equalize if the ambient rises suddenly.

 

[Demishrike]

Heat flows from hot to cold.. your intuition is correct.

? How would heat then get out of the junction ? Is it a Peltier cooler ?

Sounds to me like somebody is "messing with the new guy "..

I wish I could say I was the new guy :) I don't think I'm being messed with I trust the people who put this idea to me and they are trusted by the company; they are the head technical people that guide the whole company. Apparently the same argument was used to defend another product that has similar problems. When people at this level put something to me I can't ignore it even if it doesn't quite seem right.

If you tell me 2 + 2 = 5 I'll consider it a possibility I wasn't aware of; (I don't everything about everything). Hence why I'm trying to find the truth before I go back saying you are wrong 2 + 2 = 4.

 

[AlephZero]

Presumably there is some air flow through the enclosure. Even if there is no cooling fan, the temperature difference will create some flow unless the enclosure is sealed.

So I suppose you could design the setup so that the external air flows over or through the PSU first, before it is heated by the other circuits. Maybe that is what is meant.

But you can't mess with the basic laws of thermodynamics: heat flows from hot to cold, not the other way round.

 

[jim hardy]

I wish I could say I was the new guy :) I don't think I'm being messed with I trust the people who put this idea to me and they are trusted by the company; they are the head technical people that guide the whole company. Apparently the same argument was used to defend another product that has similar problems. When people at this level put something to me I can't ignore it even if it doesn't quite seem right.

If you tell me 2 + 2 = 5 I'll consider it a possibility I wasn't aware of; (I don't everything about everything). Hence why I'm trying to find the truth before I go back saying you are wrong 2 + 2 = 4.

That's the spirit ! I used to work for a playful genius who enjoyed challenging me to explore. If that's your situation you are lucky indeed.

Your response prompted a search on peltier pn junction. You're aware of peltier coolers...
There is apparently such a phenomenon for the ordinary pn junction but I've never encountered it in practical circuit work.

From a paper on lasers:
https://quantum.soe.ucsc.edu/research/old/bpe.html

The thermoelectric effects due to both majority and minority carriers in pn-junctions can be optimized so that the junction temperature could be less than the surrounding material. The figure below shows the internal temperature distribution in a HgCdTe p-n junction under different forward bias conditions. By optimizing the device structure (thickness of different layers and dopings), cooling by ~30C can be achieved over submicron distances.

Parsing your sentence above:

It was put to me that the junction temperature can be lower than ambient and the heat flow will then be from ambient to the junction.

Nothing illogical there, just counterintuitive.

Hence the junction will not rise above ambient.

That sounds improbable. But that's just what my old mentor would do - make me investigate and prove it right or wrong.

The proposition is way outside my experience base so i think i'll become an observer on this one.

I've learned something already. A hearty "Thank you" to both you and your mentors.

old jim

 

[Demishrike]

Thanks Jim for pointing me in a useful direction. I've not been able to follow up further than reading the article you quoted. Unfortunately I need to read a lot more to know whether or not this is happening in normal PN junctions rather that the the rather specific application mentioned. I don't think I'd like count on a MOSFET behaving in this way and use something that will over heat; especially without more information from the manufacture.

Anyhow measurements I've made show that the system behaves exactly as you'd expect: heat flows hot to cold and as ambient rises so does everything else.

I'll put this on my ever growing list of things to read up on when I have time.

 

[mfb]

You won't get a cooling effect from your electronics. That is possible in principle, but "by accident" you won't have a design where the cooling effect exceeds the ohmic heating and heat conduction, and even then you always have a hot side you have to cool (how?).

If the high-current system is like a resistor, power will increase by a factor of 4 with the higher current. That will probably increase the temperature rise by a similar factor (~3-4), so you have to change something I guess.
There are MOSFETs rated for up to 200°C, for example.