Zener accuracy over temperature range

[Dave Turner]

I'd suggest a peltier but they seem to be very poor with shock resistance. I sit to be corrected though.

The issue stands - if you can't obtain economic parts for the working environment then adjust the micro-environment to to available parts.

[Simon]

No no crystals, they won't take the shocks. I don't know what the internal clock does at -46 but at +50 it speeds up by about 2% which is not too much bother for this application

[qno]

Do you know what the problem is with low temperature devices?

Usually it is crimping and expansion of the chip enclosure that put stresses on the silicone and bond wire.
This can make them break or crack.
Military or automotive devices have the widest temperature range.
Military devices have been cycled at least once over the specified temperature range.
This makes them so expensive.


A semiconductor manufacturer will be able to help you.
Check out the general specifications of the IC's.
It is usually not in the data sheet.

You can also go to a test lab. I have done Highly Accelerated Life Test on avionics
that cycle temperature from -100 to +200 °C in Mannheim Germany.
 Normal components did survive but it is not cheap.

 

[branadic]

Do you know what the problem is with low temperature devices?

Under a certain temperature the semiconductor becomes an insulator and the device skips its function.

[Simon]

Interesting and one of the reasons I'm reluctant to just say "I'll be ok"

[T3sl4co1l]

But that varies.  Majority carrier devices (FETs, schottky diodes) work at most any temperature.  FETs (GaAs or GaN I guess?) are routinely used in LHe for physics front ends.  BJTs and junction diodes "freeze out" because they depend on thermally generated minority carriers.  And for that matter, I'm not sure of schottky diodes; they are majority carrier devices, but obey the same rules as regular diodes.

Zener or avalanche should work down at absolute zero, give or take what voltage it ends up going at (due to tempco).  These are due to breakdown effects which should depend on the material being *made of* electrons, not the relative availability thereof.

But by all means, don't take my word for it.  I don't know this stuff.  Find the literature and go from there.  If nothing else, phone a few manufacturer reps...

Tim

[SeanB]

Just remember as well your solder might not perform at low temperature if it is lead free.....

http://en.wikipedia.org/wiki/Tin_pest

[Simon]

I'm allowed to use leaded solder

[lewis]

I already suggested heating the whole enclosure but my boss was not happy with the idea.

He might not be happy with the idea, but it's really tough tit. Unless you want to use some really exotic and expensive components.

The thing with heating is that I don't want to start the MCU's at -46 as then they may behave strangely and I still don't get heated, so I'd use a thermistor and comparator to kick in various resistors around the board.

Components rated for -40 operation often have a storage temperature rating down to -55. As you say, if the device will be unpowered while at -46, the heater control might not start up. So use some sort of passive heater control to get the enclosure up to -30 or something before operation begins. MOSFET + Thermistor + heater resistors, or even put a load of series connected PTCs directly across the power input (use them as heaters) for some rough temperature self-regulation. Make it as simple as possible.

Does potting compound transfer heat that well ? I'm using scotchcast N8.

Very well - orders of magnitude better than air. Not as good as copper or ally, but good enough to ensure your board will be relatively isothermal, and to use as heatsinking for the voltage regulator.

linear voltage regs are a pain, you will actually struggle to find any that do more than 0.5W even in bigger packages, they seem to end up needing heatsinks, so I'll either use multiple regulators as I can divide the circuit up or a switchmode even if i make it myself

What are the input/output requirements of the reg? D2PAK on a 1 square inch pad with thermal vias through to another 1 square inch pad underneath, and potted, will easily sink 0.5W You'd get away with 2-3W

[David Hess]

But that varies.  Majority carrier devices (FETs, schottky diodes) work at most any temperature.  FETs (GaAs or GaN I guess?) are routinely used in LHe for physics front ends.  BJTs and junction diodes "freeze out" because they depend on thermally generated minority carriers.  And for that matter, I'm not sure of schottky diodes; they are majority carrier devices, but obey the same rules as regular diodes.

I have used bipolar devices including operational amplifiers and bandgap references down to LN2 temperatures without problems but had to qualify the parts myself.

[Edwin G. Pettis]

Motorola published the Zener Diode Handbook, 1st edition in 1967, it had a lot of information on zeners and using them.  Some years later, Motorola republished the handbook but I do not consider it to be as good as the original (I have both of them).  These books have become difficult to find and while they are probably available in PDF form somewhere (ON Semi), they were not meant to be precision voltage references with the exception of certain types:

The old 1N827/A voltage reference zener which dates back to the 1950s from Motorola, this special zener could exhibit a TCV of 5 PPM/°C but this wasn't the only Motorola zener that could do that; the 1N4569/A, 1N4574/A, 1N4579/A, 1N4584/A, 1N4779/A, 1N4784/A, 1N939/A/B, 1N4769/A, 1N4774/A, 1N2624/A/B, 1N2169/A, 1N2170/A, 1N2171A and the 1N945/A/B.  These zeners all had 5 PPM/°C TCV and voltages between 6.2V and 11.7 V nominally.  The zener currents varied between 0.5mA and 10mA for best performance and the temperature extremes could be as wide as -55°C and +150°C, pretty impressive I'd say.

Some of these diodes may still be available.

[Simon]

Getting a comparator for -55 is fairly easy I can use it to drive a mosfet and drive some resistors to heat the board, We don't usually expect -46 but being safety critical as it is supposed to be speced for -46 that is what I'll do. I need to drop 30 down to 5V with about 50mA of current so 1.25W SOIC8 regulators only do 0.5W and you can supposedly get more with a 2oz copper ground plain but it needs to be significant and I won't have the room for that, what I can do is use a basic SMPS made with discrete parts to pre-regulate to say 7-8V so tempco on a zener won't be a problem and then the linear regs, that way I don't have problems with trying to adapt the circuit to work at 20-30V as what i normally do is use a series resistor on the regulator to help take some of the load but with 20-30V input that still makes life hard.

[lewis]

LT3470 is an excellent switched-mode regulator I've used many times, and you can buy them from Farnell. Works up to 40V too, so your high input voltage requirement is taken care of. The internal reference is very stable, and it's effing tiny.

http://cds.linear.com/docs/en/datasheet/3470fd.pdf