mosfet operating temperature

Old datasheet but figure 11 might give some insight:

Personally I'd feel uncomfortable if it went over 90C for any extended time period. One must also consider usage. A hot summer day somewhere in Arizona and either a thermal shutdown engages if provided (nuisance) or ... phut ... *POP* (major nuisance).

Add a little dust, and a high ambient, and maybe a corner case on power output/line voltage/whatever changes things, and it could get really hot.

What is the gadget?

On the other hand, power mosfets are pretty rugged. We tested some TO247's: they quit working at 300C but recovered; they died about 330.

John

Yeah, but by 200C they're depletion mode again. ;-)

Tim

They were working in that direction. At 300C, they were hard tunned-on at zero gate bias.

John

I think it's a matter of poor design. The mosfet is operating well within it's rated power limit, but the switching signal is not at all clean - hence heat generated whlst the device is partially on. Lack of circuit diagram and lots of surface mount components packed into a small area makes me disinclined to investigate further. All of the units run at well over 100 deg, so I think it's a case of labelling them as unacceptable and contacting the manufacturer.

Steve

the scope shows an untidy switching signal. Otherwise the mosfet should be operating well within spec.

switch mode power stage of a control unit, low power - under 1 amp. I think I can label these units as unfit for purpose as 130 deg does seem to be too near to the limit.

Steve

First, if you want low leakage at temperatures up to 185C the voltage spec should be no more than 600V; 400v or less ideal. STM FETs are superior in that respect; "typical" measured: SCT6NF30V: 23uA @ 10V, 185C. STD3NM60T4: 23uA @ 200V, 185C. STD4NK50ZT4: data not kept; maybe

Prolly the die solder melted..

Pretty unlikely as a rule but it depends !

You need to learn about thermal interfaces and use use the figures on the data sheet. There is no simple figure and it's also pulse width sensitive too.

Motorola AN1040 ( Now On-Semi ) provides classic info even though very dated by now.

It's not complicated. Trouble is, this stuff isn't taught any more. Come to think of it, I wasn't taught it either, I learnt it for myself.

Graham

-- due to the hugely increased level of spam please make the obvious adjustment to my email address

As a rough rule of thumb I use 90C max Tc for plastic power devices in linear applications and up to 120C for metal can TO-3.

Max junction temps are respectively 150C and 200C.

Your point about Tamb is also VERY valid. For that kind of use heatsink temp MUST be monitored.

Graham

-- due to the hugely increased level of spam please make the obvious adjustment to my email address

It doesn't sound to me as if you understand the data sheet or thermal design in that case unless your initial post was misleading.

Do you understand what theta(j-c) means for instance ? And Tj(max)

Graham

-- due to the hugely increased level of spam please make the obvious adjustment to my email address

Graham

-- due to the hugely increased level of spam please make the obvious adjustment to my email address

That's thermal runaway for you in spades !

Graham

-- due to the hugely increased level of spam please make the obvious adjustment to my email address

Your problem may not be the device per se but how you are cooling it.

Graham

-- due to the hugely increased level of spam please make the obvious adjustment to my email address

It isn't true, either, not when you get down to very low thermal resistances. (It's okay as long as thermal diffusion inside homogeneous materials is the limiting factor.)

Heat is conducted by two mechanisms: electron scattering and phonon (very high frequency sound) scattering. Both of these have issues with mismatch at material interfaces--just as in transmission lines or optical refraction. This leads to an effective thermal resistance contribution from each interface between materials, which can easily dominate when you get below 1 cm**2*K/W.

The theory of this is not well developed even today--most of the theoretical predictions have the effect going away at low temperatures, whereas in real life it's closer to a constant.

Some friends of mine developed the liquid metal thermal interface system that's used in high-end Apple machines (and even a couple of IBM ones, I think), and they ran into this problem.

Cheers

Phil Hobbs

by

Hmmm... would here be any equivalent to anti-reflection coatings for the phonon component? What's the average phonon wavelength anyhow?

John

by

The electron reflections are worst at metal/insulator boundaries, naturally. Phonon wavelengths are very short--in thermal equilibrium the phonon spectrum has a Planck spectrum just like blackbody radiation and for the same reasons. However, a phonon of energy h*nu has a wavelength of v/nu instead of c/nu, i.e. something like 10**5 times less than the equivalent photon.

The room temperature Planck peak is around 20-30 THz iirc, corresponding to an optical wavelength of 10-15 um (the spectrum is two octaves wide anyway, and the exact peak position depends on whether you're talking per-wavelength or per-frequency units). That means that the phonon peak has a wavelength of around 1 angstrom in rough numbers--i.e. atomic dimensions. (Of course aliasing sets in when the wavelength gets that short--elastic waves in crystalline solids are the original sampled-data waveforms.)

Cheers

Phil Hobbs

dated by

So antireflection coatings are out. A materials taper might help, but that occurs naturally if metal is soldered to metal.

Metal-to-silicone-grease sounds like a bad interface all-around.

This is discouraging me from specifying solid diamond heat sinks.

John