semicon temp specs?

I'm going to be doing a VME card that has to work at -40. We considered adding one pcb layer that was just a serpentine trace, driven by a switcher, LM45 for feedback, to warm the whole board.

John

Diodes and bipolar transistors cave in at around 20K; they call it "carrier freeze-out" or something. The regular 2.5 mV/K diode slope has a sharp knee at around 20K. Diodes are used as cryo temperature sensors down below 2K. As I recall (from some work with Jlab) at 2.5K a small diode has a few volts drop at 10 uA and looks pretty much resistive.

Lakeshore makes good cryo diode temp sensors.

John

This chip itself will generally work below -55°C.

The limitation is packaging. Plastic packages will often crack adjacent to the lead frame, and are also prone to pulling off wirebonds since there is only a silastic coating between them and the plastic.

...Jim Thompson

been looking at a rf widgit

But it is only spec'd to -20C. What are the chances that it will work normally down to -40C ie in Canada and other chilly places.

and what happens to an ic working at low temps?

martin

the ic is so small, I'd put a little heater on it to keep its hands warm, so it can still push all them digits

martin

CCDs develop very bad streaks along the transfer direction at low temperatures (some as high as -50 C, others as low as -100 C), due to traps becoming unpopulated. So different devices show different sensitivities to freeze-out. People needing to work at nitrogen temperature usually stick to majority-carrier devices such as MOSFETs.

Cheers,

Phil Hobbs

Anybody got advice/warnings on using, say, a 900 pin BGA in the -40 to

I think.

John

Phil Allison wrote in sci.electronics.design:

Sure, but if the device has to run at high temperatures too it may not be quite so simple. Also, low ambient temperature may mean outdoor use, which may mean battery power, which may mean that using power for heating is out.

Anno

Phil Allison wrote in sci.electronics.design:

No, just that the OP hasn't given enough details for an unambiguous recommendation.

Anno

I'd probably use a thermistor + transistor or the LTP in a CA3046 thing as a temp comparator to turn the heat on and off.This would save power in hot places I've never had to do a low or wide temp design, and this was just comment someone pointed out in another NG discussion

It just interested me

martin

Bipolar stuff has a fermi surface in the impulse space of the semiconductor lattice. This fermi surface vanishes toward lower temperatures. When this fermi surface has vanished to significant degrees I don't know, guess far lower,

-100DegC perhaps. FETs do work downto basically zero. The next hurdle is the bonding. The bonding wire, usually gold on silicon doesn't necessarily stick to zero. The different thermal expansion coefficients of silicon and gold let it disconnect. Where this happens I don't know either. Then there may be different thermal expansion coefficient between carrier and substrate, making the substrate come loose.

As the others said, give it a try.

Rene

snip

Thanks Jim

martin

"martin griffith"

.......... Phil

And never store it below -20°C and never switch it off at such temperatures...

-- Andy

"martin griffith" "Phil Allison"

Bound to be cheaper than forking out for "mil spec" devices.

........... Phil

"Anno Siegel"

.......... Phil

"martin griffith" Phil Allison

Philips make them as 8mm disks.

One in an old cat ( 2322 661 91005 ) goes from about 25ohms below 0C to

Supply it with about 4 volts rms and voila !

......... Phil

I tested a bunch of TO247 power fets hot. At around 300 C, they spontaneously turned on and quit acting like fets. They were OK when they cooled off. It took 350 or so to permamently whack them.

John

Plastic packages at high temperatures gradually degrade the silicon the plastic is in contact with.

...Jim Thompson