How to make electronic systems which work below −40°C (−40°F)?

Question

I notice on a lot of electronic products that most of them work down to −40°C. Glaring example with Microchip microprocessors page: https://www.microchip.com/ParamChartSearch/chart.aspx?branchID=1012 (check last columns).

What I want

I would like this electronics to work down to −80°C (or ≈ −110°F)

What I cannot do

I cannot "heat" the case of my circuit or even the circuit itself.

Some examples

• weather balloon (−56°C from 10km to 15km excluding extremes)
• electronic thermometer for Pfizer vaccine freezer (−60°C to −80°C)
• Perseverance rover (on average −65°C on Mars excluding extremes and excluding SuperCam)
• rockets
• military air planes
• satellites
• etc.

My case

Operate a fairly basic electronic system in a remote-controlled air balloon at very high altitude thanks to Dihydrogen or Helium. The circuit should work for about 7 hours.

So I have 3 questions:

• Why "generally" −40°C? Why this particular value?

  Why not −39°C, −41°C?

• Can we add additional protection to withstand its low temperatures (until −80°C or ≈ −110°F)?

  I sometimes see on certain circuits a thickness of silicone poured on the circuit. I understood that other alternatives exist such as urethane, varnish and acrylic. Does it help withstand the temperature? Let's assume we want to run a board in a −80°C freezer.

• So how do space electronic systems work?

  For instance, on Perseverance rover, SuperCam need to be heated to maintain a temperature above −40°C. However the remaining parts are not heated. So if it does work, how?

Warning

I think trying to found a specific "patch" to my case is not a good idea. The real answer could be very useful and would allow me to modulate my needs and truly understand how it could work.

The question is about TEMPERATURE. Please AVOID long developing on anything relating to:

• electromagnetic compatibility (CEM)
• pressure
• rapid temperature change
• UV / infrared / light destabilising electronic circuits
• efficiency

We just stay focused on the temperature. Can come into the line of sight:

• humidity or condensation
• ice formation

Answer 1

Most semiconductors will function fairly well at −80°C, maybe reduced gain and perhaps some increased speed that could affect a marginal design. The reduced gain can cause oscillators to not, although if they are already going they may continue as the temperature drops. I've run plenty of electronics immersed in liquid N₂ and not much unexpected changes from 300K to 77K. −80°C is about 200K so not even halfway to 77K. Liquid helium at 4K is another matter. There is some information available in the literature, but getting guarantees from the manufacturers is probably not in the cards due to a lack of market that would interest them. Things that are guaranteed to work at cryogenic and very high temperatures tend to be pricey. Characterizing them yourself is easy for simple parts and very difficult to do with 100% assurance for very complex parts.

Most passives like resistors won't change much either – a bit of shift from the temperature coefficients. Ceramic capacitors may shift a lot more.

Things that don't work well, or at all, as fraxinus says, are batteries and electrolytic capacitors, and they may well be limiting. Other things like DPSS laser tend to have a very narrow range.

As I understand it, the Perseverance rover has a long-life heat source on board (a Pu 238 radioisotope RTG), so getting too cold is not an issue. Including a chunk of weapons-grade plutonium is not an especially practical approach in most applications.

Answer 2

The −40 °C value is more or less a matter of a standard. So is −55 °C for military specs. −40 °C is "good enough" for most civil use and −55 °C is acceptable for general military use.

On the other hand, most electronic components will work more or less acceptably all the way to at least −100 °C.

You just need to find out yourself what works and what doesn't. The manufacturer of the part may or may not help you and you may or may not afford the help from the manufacturer.

On the other hand, last time I checked, a kilogram of dry ice (solid carbon dioxide) is less than 10 euro. Get a good ventilation (or at least know the risks of the evaporating carbon dioxide) and freeze your electronics all the way down to −79 °C. Liquid nitrogen is used in extreme overclocking projects, it is way colder but is more dangerous as well.

Things that are almost guaranteed to fail are the electrolytic capacitors and the batteries.

---

edit: one more thing that is pretty much easy to go wrong: quick cooling and quick warming back. Don't, or else something will crack (even invisibly) and will stop working either right away, or after a while (e.g. when the condensed water enters the crack)

Answer 3

Be careful that you don't misinterpret what those temperature ratings mean. When a part is rated to −40 °C, that doesn't mean that it won't work below that point. It simply means that the manufacturer has tested it down to that temperature and will only certify/warranty the device's operation at or above that point. Manufacturers usually leave a decent safety margin around what they'll warranty, so odds are that it will work at least a modest level below the official specs. This hasn't received official testing, however, so the manufacturer won't provide any guarantees.

Also, there are two temperature limits here: a "storage" temperature and an "operation" temperature. For some components, these can be quite different. I've seen components that can be stored at −40 °C but require +5 °C while in operation. For some devices, these limits are the same. Make sure you're reading the datasheets carefully.

Manufacturers don't test products to find out where their specific thermal limits are. These tests are expensive and time consuming. Instead, they test at a small number of standard test points. This tends to be −40 °C for standard components and −55 °C for "military grade" or "extreme temperature" versions. There's not a lot of demand for parts that go lower than that, so running tests at lower temperatures is rare.

Why these numbers? MIL-STD-810 covers testing standards for military equipment. According to Method 502.7 – Low Temperature, "most applications with normal development cost considerations" are tested to a 20% "frequency of occurrence", meaning that 20% of the hours in the most extreme month are below that temperature in the cold part of the world. The 20% frequency of occurrence is listed as −51 °C, and the 10% frequency of occurrence is −54 °C. Those values are close enough that instead of stopping at the 20% threshold, you might as well go down to −55 °C and cover practically every use case for standard military equipment. Similar specs exist for non-military equipment as well. For example, IEEE 1156.2 "Standard for Environmental Specifications for Computer Systems" lists −40 °C as the low temperature test point. See also IEC 60068-2-1.

You can certainly test components to whatever levels you want. The manufacturer probably won't do that for you (at least not for free), but there are testing companies that will test components in whatever sort of environment you want. You can run or commission your own thermal tests covering whatever range of conditions that you expect to see. Talking to the component manufacturer can also be helpful. They may have parts available that have already been tested to meet your needs but due to limited demand are only available via special-order. They may also be able to indicate which parts other customers have successfully used in extreme temperature environments.

Answer 4

First, consider what will happen if you exceed the temperature range of a part. Silicon, copper, etc, are tough, they aren't going to be damaged directly.

1. Parameters will shift so much that the circuit won't function anymore.

2. Parameters will shift, but the circuit still works.

3. Unequal thermal expansion/contraction will physically crack the parts.

And, probably other bad things will happen, but lets concentrate on these.

Components are generally manufactured in three temperature ranges, Commercial (0 – 70 °C), Industrial (−40 – 85 °C), and Military (−55 – 125 °C). These ranges cover the majority of applications.

The silicon die is often the same for all three ranges. The difference is the packaging and testing. Parts won't magically fail if you exceed their temperature range, I have run commercial parts to 100 °C without issues.

So, to exceed the military temperature range, pick the best package to minimize thermal issues, and have a robust design that can withstand parameter shifts. Characterize the parts and develop custom specifications for the new temperature ranges.

Temperature cycling is usually worse than just going to an extreme just once. Reliability will be reduced, but may be acceptable for a 7 hour balloon flight. Determining the reliability will be difficult, there are experts that can help determine what will happen.

If the application is important enough, parts can be created with higher guaranteed temperature ranges. They will be extremely expensive. Special parts have been designed to operate at the bottom of oil wells where the temperature can be 200 °C.

Answer 5

• Can we add additional protection to withstand its low temperatures (until -80°C or ≈ -110°F)? I sometimes see on certain circuits a thickness of silicone poured on the circuit. I understood that other alternatives exist such as urethane, varnish and acrylic. Does it help withstand the temperature? Let's assume we want to run a board in a -80°C freezer.

I don't think the conformal coatings can protect against the cold (I mean, the temperature is going to make it's way through after all, even a super thick thermos has it's limits). What they might protect against is condensation. Silicone conformal coating is only good until -40C while Paraxlyene which needs to be deposited via vacuum chamber is good down to -200C. I am also unsure if silicone is able to handle the low pressures you might encounter at high altitude, but Paraxlyene definitely can. Paraxylene is satellites and space probes use.

Polyurethane and acrylic can handle -65C. Again, do not know about low pressure.

Answer 6

Many circuit parameters are dependent on temperature. Resistance, resonance frequency, threshold voltages for FETs, etc.

If your question is "Why does stuff only work down to xx temperature", it's because the market only requires it to work down that low. IC designers, in particular, work very hard to make designs that cope with parameter changes, both on-die and in the supporting systems. Making things robust through wider temperature ranges adds significant effort to device design and characterization.

If you want to understand what sort of changes occur in electronics with temperature, look at the temperature coefficients of various passive and discrete semiconductor devices, and understand that many of the same fundamentals apply (more or less) to ICs.

Tips on designing systems to work at low temperatures:

1. Avoid cold starts. If you can start your system up at 'normal' temperatures, and then launch it, it's more likely to work. (We had all kinds of issues cold starting systems below -40 for arctic boreholes. Not sure what it was exactly.)

2. Calibrate oscillators (and other things) throughout your whole temperature range. Make sure that you take advantage of whatever means you have to tune things, and develop lookup tables based on temperature for your whole operating range. This is especially important if you have radios.

3. Test, test, and test again. Dunk things in LN2, Dry ice slurries, freezers. Make sure you're testing EVERYTHING, not just the PCB. Do it exactly as the system will be deployed, especially things with clocks (so...everything. Radios, power supplies, microcontrollers...) and batteries!

4. Use proven designs for your particular application. Don't re-invent the wheel if you can avoid it.

Answer 7

There are already plenty of good answers, but I have not seen any which includes some experience from the domains which deal the most with your issue so here are my two cents.

I am a satellite mechanism designer, local specialist for electronics within the team. I have been involved in a well-known mission relying on a Martian rover, and we had to tackle these issues with extreme temperatures.

I cannot talk for those who designed Supercam, but on Mars just as in orbit, "heating up" is not some kind of hacked-up patch in case the environment is too cold: this is the baseline. Only the components which absolutely cannot be heated up to temperatures within their operational temperature range are insulated from those which are, and properly passively (preferentially) or actively cooled. The reason is that cooling is much more complicated than heating (which only requires heater control loops), especially for active cooling (which basically require high-tech fridges with special heat pipes + fluid). So, generally speaking, the temperature of all electronics is directly or indirectly heated up, unless specifically required not to – for infrared imagery frontends for example.

Therefore, for the majority of electronics, the issue rather lies in the non-operational temperature. We had to switch from an electronics-based integrated position sensor to an electromechanical position sensor acquired by an electronics acquisition system located further inside the rover because of this: the component we wanted to use only went down to −55 °C and the manufacturer did not recommend initiating a qualification campaign to test its survival to −130 °C (which, I think I remember, was the coldest temperature measured on Mars plus −15°C of qualification margin, see figure below for margins scheme as per ECSS standards [more below on that]). Generally this issue is difficult to tackle because all platform and payload equipment are switched off before commissionning in operation (except sporadic health checks) to save power while power generation is not yet on (solar panels deployed and tracking, radioisotope generator switched on, etc.).

In your case, I do not think you have these restrictions – so your temperature control system may be on from the beginning of the flight. You "just" have to make a proper thermal analysis (as accurate as you can with what you have, of course) in order to estimate the power needed to warm up your components to bearable temperatures, considering all radiative (from Earth's albedo, the deep space black body temperature, and temperature gradients on the balloon itself), convective (depending on what you mean by very high altitude), and conductive (on your balloon) paths. The more you will use highly reflective of emissive paint or surfaces, the more you will need heating – but the less hot you will go during hot cases. For unpainted mechanisms components in geostationary orbit, 5–10 W average per critical zone is generally enough, for example. The range of the temperature in the satellite itself is actually much narrower that we would think, thanks to conductive interfaces averaging the heat flux from all faces – so that is where most of the electronics is located. Here is an example of finite element analysis modelling on a random satellite I found on the web, but you can reach some pretty accurate results using thermal network modelling as equivalent electrical models (using thermal resistances, specific heat capacitors, voltage sources for temperature boundaries and current sources for power injection).

In a nutshell:

• Assume a component cannot be used at a temperature outside of its non-operational and operational ranges, unless you are ready to test it thoroughly (as part as some "qualification" campaign) at the required temperature extremes, with some margins. For components which are not dead-simple, it will be necessary to define in collaboration with the manufacturer the required tests, sequences and conditions of tests, to ensure the qualification campaign actually proves compliance of the component to the desired temperatures (some issues may not be understood by the user, or some others may arise randomly).
• Do not neglect thermal analyses, however crude they are. Place your components in function of the results and iterate, so that you achieve compliance with all operational and non-operational temperature ranges passively as much as possible (using coatings and paints if necessary). Favour being too cold over too hot – place heaters whenever necessary and wherever the most efficient, and design the temperature control system according to the thermal analysis.

Note that if the components suppliers are following the same standards (one would have to check the associated MIL-STD to be sure), if the rated temperature is −55 °C op, it means at least one unit made it down to −70 °C, which for electronics is not really far away from −80 °C (for mechanisms, thermal expansion could eat up a gap and jam well within ±10 °C). Pay attention though that a rule of thumb is electronics life/reliability figure is halved every 10 °C in hot cases –so I would expect it to be similar when cold as well–, and that all kinds of parameters that you checked before selecting the component will still need to be re-assessed taking into account any sensitivity to temperature. All of this is generally not worth the trouble compared to heating.

For more information on guidelines specific to the space domain, look up the ECSS standards – these are the standards we need to comply to when working for the European Space Agency, and generally for any european customer. ECSS-E-HB-31-01, ECSS-E-HB-31-03, and ECSS-E-ST-31C come to mind.

Answer 8

Extreme temperatures in electronics are a matter of parameter margins and testing more than anything else. Some parts such as elcos with liquid electrolyte and various sensors may not like deep freezing, but most parts will keep working at low temperatures, perhaps with different parameters.

Essentially, there are three approaches to make sure your design works:

• Testing that your actual device indeed works at the desired temperature
• Making safety margins for parts parameters large enough to cover the desired temperature range
• Use redundancy for critical components which cannot be made reliable enough

There's nothing impossible w.r.t extreme temperatures if you use the right approach. Even consumer-only electronics such as Raspberry Pi keep working at pretty extreme temperatures.

Answer 9

One thing I haven't seen mentioned is solder. Pure metallic tin is unstable under 13.2 C, preferring to convert to non-metallic grey tin, though the tin pest is slow to start. Below -30 C it can turn into an actual issue.

You may not want to spring for indium or silver solders or the like, but at least make sure you're not using lead-free solder.

NASA has done some research related to this because the outer planets and moons get pretty cold. The main findings are about ductile-to-brittle transitions.

This paper tested ingots of various alloys for over a year at 255K (which is just -18 C) and found Sn-0.5Cu to be seriously affected, while no change was observed on Sn-37Pb.

Answer 10

There's another couple of interesting things about temperature requirements that I don't think has been mentioned, particularly as it relates to MIL or NASA missions.

An example here helps. Let's use the -40C low temperature point that has been batted about. I will assume that this is the lowest temperature at which a part/assembly/system has to operate and meet all of it's performance requirements, when deployed (on Mars, in orbit, etc). This usually means that the assembly has to be tested at a lower temperature, -45C (acceptance test temperature), in order to provide some margin against the -40C requirement.

Then, in some circles there's this thing called "thermal uncertainty", which is basically a catch-all bucket for all the unknowns in your thermal model. A common value for thermal uncertainty is 11 deg C. At the low temperature end, this 11C is subtracted from your acceptance temperature (-45C in our example) to give us -56C. When selecting parts for this type of application, this is the low temperature point at which the part has to meet all of it's performance numbers, and so parts have to be selected guaranteed to operate at -56C, and not the -40C

Answer 11

• Why "generally" −40°C? Why this particular value?

General standards probably. So they tested it till -40 C, so it must work till that temperature. But it won't break if you reach -41 C. You can probably reach quite below that, just test your systems for they may act different(different gain or what not).

• Can we add additional protection to withstand its low temperatures (until −80°C or ≈ −110°F)?

Condensation(the water) is probably your main problem even with the -40 C.

• So how do space electronic systems work?

Systems in space are more rigidly built because take off is a massive hit. Vibrations, soundpressure, acceleration in all axis, heat(take off), radiation and vacuum. So just wrap those things up real nice and cozy and you got working electronics in space. Also when in space you have to figure out how a long term vacuum affects your hull, but that is an entire different story.

Long story short: You can most likely use the -40 C electronics in your freezer.

Answer 12

Silicon nanocoatings are each one temperature specific. Specific to the temperature from which their crystal is stable in structure and conductivity. thus to handle larger temperature ranges you would need enhanced coatings. the lesser temperature the more compacted the coatings get and -40C(-73C) is generally the temperature where materials have a deformation such as if you couldn't manipulate them anymore from the schematics and need a whole different silicon, solderings, pcb materials and passives;