Board lossing comm. @ -40F

Oscillator starting is usually the prime suspect in cold temperatures.

Did you measure the oscillator output at -55 C or just at -40 C ?

Exactly where did you measure the clock ? From the crystal terminals or after some buffering stages ?

Hitting one of the crystal terminals with the oscilloscope probe may be enough to generate the initial transient to start the oscillation or the stray capacitance of the probe may be enough to let the oscillator running while the probe is connected :-).

Using systems with components below their rated _storage_ temperature range is one thing (expansion coefficients etc.),

However, the component specifications are guaranteed for a more limited temperature range and outside that range, the behavior is unspecified.

When working in areas with cold temperatures and out door equipment are powered down for the night or for the weekend, powering up the equipment in the morning may take some time, until the heat generated by the electronics itself will warm the equipment to temperatures within specifications.

Since the most power dissipation in modern electronics is due to charging and discharging stray capacitances, it is imperative that the clock starts to work even at extremely low temperatures. Thus, if the built MCU internal oscillator is unreliable, say, below -40 C, one should consider using an external oscillator guarantied to -55 C.

For non-critical applications, it does not make sense to pay aerospace prices for every component or arrange the transportation of fuels (often several thousand of kilometers) to keep the electronics all the time within the limits specified by the manufacturer specifications.

While in such environment, a permanent failure would be a very bad thing (replacement parts could be shipped next summer), transient out of specs operation during the warm up period is not usually a bad idea.

On a sunny day (Tue, 28 Dec 2010 15:22:28 +0200) it happened snipped-for-privacy@downunder.com wrote in :

Did anybody mention the effect of condensation at these low temps? Especially if you cool it is some cooler, and then put it on the bench. Condens water can makes a good cunductor.

Condensation is a bitch to get around. I used some very thin paint on coating to get around the problem. The odd thing was the finished product was sold without a coating and was often used in sub zero celcius environments without issue.

On a sunny day (Tue, 28 Dec 2010 21:52:57 +0800) it happened "Dennis" wrote in :

While the condensation definitively is an issues at ice swimming temperatures (0..-20 C ) but at temperatures below -20 C, there really are not that much humidity to worry about.

I suggest that you never fly in any airplane again, ever. There is not anything flying that is not filled with parts rated to -40 that are expected to work at -55.

If you've decided to use parts below their functional temp range, is insulating the IC and adding a heater to get it to at least -40 quickly an option?

NT

Many parts will work fine to lower than their rated temperatures provided you are not pushing them to the limits.

If your microcontroller is rated to -40F and you expect it to work to

-65F then DON'T RUN IT AT THE MAXIMUM CLOCK SPEED, stay towards the middle of the operating voltage range etc. The manufacturer's don't really test most products at the limits anyway, they infer operation at limits by the margin at test temperature. That said, you're essentially characterizing the units yourself, which is not easy to do well.

There have been some fairly serious temperature-related errata on some micros, such as some instructions not working even within the rated range-- needless to say this is not easy for the user to debug, since the micro "sort-of" works. You should check into published errata.

As others have mentioned, oscillator startup is a commonly reported failure mode, but there are others. The oscillator may keep running, if the temperature is lowered with it running but won't start up when the unit is dead cold.

Do you have any electrolytic or Z5/X5 ceramic capacitors in the unit? They can do kind of a disappearing act at temperature extremes, which could cause side effects (for example, if you used such parts in a charge-pump based transceiver).

Best regards, Spehro Pefhany

I often run onto this kind of issue. It's driven by cheap-ass (disguised as cost-saving), so crystal oscillators have no external parts.

I'd use a separate oscillator built by pro's in the crystal business.

...Jim Thompson

Old trick: Hang a small loop onto a coax and that into the climate chamber. Then see with a spectrum analyzer or receiver whether oscillators quit. That way you can monitor all oscillators in the box. Oscillators not only refuse to start when too cold, they can also quit below a certain temperature.

Trick #2: Let the uC run a fast software loop out of flash, pipe to a pin and solder a piece of wire to that. Monitor this frequency as well as it may tell you whether anything in the flash area hangs it up.

Anson, since you are posting through Google I won't see your responses until someone else responds.

When you're designing at extreme temperatures, using at least some parts outside of their specified temperature range is not uncommon. You need to know what you're doing, though, and it sounds like this guy hasn't jumped through the requisite hoops.

Is there a bandgap reference involved?? Some have problems starting below -40c. regards, al

I have tested 3 boards earlier today after reading about condensation at extreme cold temp. My test group were 1 good board used for reference (works all the way down to -69F) , 1 bad board that outputs random temp. at -40 then dies (loss communication) at -65F, and 2rd bad board that outputs random temp. at -57F. After baking the boards at 240F for over an hour, I put them through the cold again by going down to -40 then -65F. The board that started to go bad at -40 still does the same except it now start to acting up around -42F and still dies at -65F. However, the one that use to give out random output at

-57F actually passed the test at -65F and it was all good until the thermalcouple reading was actually -67F. So I am happy that by baking it at 240F, it seemed it have vaporlized some moisture or whatever it is that's saving at least one board.

Now, how can moisture get onto the board while it doesn't show any immediate affect above -40F? What is a good and safe temperature to bake these boards to without damaging or melting PCB, ICs, SMD, Solder? Also, what kind of humidity should I store these boards at to avoid future failures? I heard conformal coating like HumiSeal is good way to go to protect boards again humidity, is that a good way to go?

You are getting all kinds of condensation problems when you open the cooling room to insert the cards. The air at room temperature contains a lot of humidity, which is condensated at quite mild sub-zero temperatures. At -40 C there is very little moisture left in the air.

The extremely dry air is also a good insulator and there are going to be a lot of static electricity build-up, which can cause malfunctions, if zapped into wrong places.

Since you used tin-lead solder, assuming good soldering techniques, there will be no solder joint problems (and little tin pest) down to at least -70C. Furthermore the proof of that pudding is the fact that the failure / restore is repeatable in the -40C region. Start point: "Regulated" voltages could give problems, as almost all voltage "references" are based on a band gap technology which is sensitive to temperature. Just look at that "nice" upside-down parabola voltage VS temperarature curve IN the max temp spec range. Every manufacturer does their circuit differently, and National Semi seems to have one that starts to die in the 160C region (useless above as the I/V curve looks like a tunnel diode). So, look very carefully at _all_ "reference" voltages; bring them out to the HOT room temperature for monitoring while cycling the beastie.

Yeah thats what I thought, except most of the gear is inside IP68 plastic enclosures, "office air" at say 40% RH is trapped inside and then the box is deployed into a sub zero (C) environment.