-55C operation of power conversion modules.

I've dissected some cheapie unregulated sip converters. Many seemed to be a classic bipolar inverter

The base bias resistor seems to be picked to allow some (short) duration of load short before the transistors blow up. But cold, there may not be enough gain to run. The cap near the bases might be a resistor in some designs, which would be nasty for cold start. Or something.

More expensive regulated converters probably use ICs and maybe mosfets.

-55C is a lot of stress for all sorts of circuits.

y prevent the module from starting at -55C. Vicor, SynQor, Gaia, Astrodyne. ....

oubt I would find any electrolytic caps. I'm not really sure what could ca use problems. Has anyone done any work in this area?

Yes, I am familiar with those free running power supplies. They usually bl ow up when they turn on. If you can get them running they usually stay run ning. Sometimes they blow up when you turn them off. I once got the boss to buy me a nice LeCroy when his self-oscillating supplies started blowing up. I told him if he bought me a LeCroy I would fix his little prob :)

Back to my question - I don't really see what would keep a supply from tur ning on at -55c if it would turn on at -40c. You say that -55c is a lot of stress. Can you give me a couple of examples?

Thanks John.

We temperature test the first article of all of our products, and we seem to have the biggest problems cold; we usually go to -40C to check margins, even if the box is spec'd for 0 or -20.

One recent problem was some SPI-interfaced delta-sigma ADCs that made data errors when they got cold, -25 or some such.

Why not get some dc/dc converters and test them?

We have a 7 GHz 4-channel LeCroy scope that cost about $50K. It's basically a Windows PC with some ADCs attached. Only one guy here knows how to operate it.

It has channel-channel RMS jitter below 1 picosecond, which is impressive. Trigger-to-channel jitter is worse, which is common in digital scopes.

Beta tanks at low temperatures, electrolytic cap ESR goes through the roof, differential thermal expansion stresses the solder joints, which become brittle at low temperature, ....

For a good read on what low temperatures do to materials, there's the Encyclopedia Arctica,

Cheers

Phil Hobbs

Things go wonky. CMOS delays go down, but not uniformly. Bipolars get slower. Parameters slide all over for analog circuits. I've only ever designed for -40C, and I had help.

Basically, you can't trust chips to work within specs below the promised temperature (depending on the manufacturer, sometimes they won't work _at_ the promised temperature), and you can only trust discrete circuits to work if they've been designed for that environment (you can build chips to work all the way down to 77K, and probably beyond -- but they tend to blow up if you power them up at room temperature).

I would expect that if I were handed a bunch of supplies that were designed to work at -40C, some would work down to -55C and some wouldn't. I know from experience that when a team builds circuits that are specified to work at -40C, some project managers will sign off on things if you turn them on at 0C and bring the temperature down, while others will want to see things working when you turn them on at -50C after a good long cold soak.

I do know that when you get below about -20C, with each five or ten degree drop in temperature it gets harder and harder to keep things going; so I'm not terribly surprised at stuff that works at -40 but not

-55.

(And electro-mechanical is worst -- EVERYTHING wants to seize up as the temperature goes down).

Those transformer-coupled bipolar-device inverters are crap, at any temperature. ...Jim Thompson

At really low temperatures the art of discrete design comes back into play. There are transistors which are rated down to -65C and they aren't even expensive. This one is under $0.10:

Low power ones can be had under $0.02:

Even some versions of the trusty old 2N7002 can be operated at -65C. What one really has to watch out for is temperature gradients and fast temperature changes. That can crack stuff such as MLCC.

Or that lone bottle of beer someone plopped outside for a quick chill and forgot ... POOF :-)

I don't think we ever exploded a beer bottle in the environmental test chamber. That sounds like production-stupid or service-stupid. We were engineering-stupid. We (well, a very few of us, but the mud sticks to all) did things like leak testing systems and then leaving them half full of water, to be discovered months later by the engineering tech, or sending them off to be vibe tested at higher-than-normal levels with all of the expensive lenses just tacked in with glue dots instead of fully glued.

I recently finished a new design, including 2 microcontroller, smps etc

Lower operating temperature was -70 and upper 165 degrees. If you design for a robust system at minus 40, then most likely it will work at lower extremes

Cheers

Klaus

We've used thousands of the CUI and Murata parts, and they work great. They do usually blow up if you short their outputs, so we add a current limiter of some sort when that might matter.

This series is cool:

I often use them even when I don't need the isolation.

electrolytic capacitor?

I'm seeing polymer electrolyte (aka solid) aluminium capacitors in my converters here

sample data sheet:

look at that! -55C, right on the money.

If you specify -55 to +125 degree operation in a datasheet, for some products this puts it into the category of military or dual-use goods for export control purposes, which can require you to complete much more paperwork to sell it. I'm not saying that this is the case here, but sometimes it is easier to sell the product if you make the temperature range worse.

Yes - ITAR.

Thanks Tim.

Do they use spiral coreless xformers or planar?

Cheers

Klaus

The ones that I've dissected use tiny torroidal transformers.