Electronics in rockets, reliability and best practices.

[Emil]

I recently designed a magnetic apogee detector circuit (http://www.zeptobit.com) and this has made me think about best practices for making reliable electronics for rockets.

In particular I'm thinking about model rockets, HPR rockets and amateur/experimental rockets using solids, hybrids or liquid bi-props. Such rockets typically reach altitudes from a few hundred meters to tens of kilometers. Parts of the flight may be at significantly reduced pressure or near vacuum. Acceleration will can be anywhere from a few G to several hundred. Solids and hybrids will usually have unstable combustion resulting in vibration.

Will oscillators (ceramic or crystal) drift or have other problems because of acceleration or vibrations? After watching mikeselectricstuff's video about pressure sensitive PICs, I'm worried about seeing similar effects in flight caused by acceleration or vibration. Are there particular types of oscillators which should be preferred or avoided?

Is the micro phonic effect in ceramic smd capacitors a practical problem? (Can my decoupling cap turn into a noise generator?)

Many SMD components require baking before soldering if not stored in air tight bags. If I understand the correctly this is to avoid outgassing during soldering. Could this also happen during short exposure to vacuum, causing the components to fail?

And now for the part that worries me the most.. wires and connectors.
What types or connectors are likely to work reliably for signals and power without being excessively large or heavy?
Screw terminals seems to be by far the most common option in electronics for HPR. They are certainly convenient, but are they reliable?

What types of cables should be used for signals and power?
Based on this quote from Wikipedia, I think wires with a minimum of 19 strands should be used, but many more may be better. "For application where the wire moves, 19 is the lowest that should be used (7 should only be used in applications where the wire is placed and then does not move), and 49 is much better. For applications with constant repeated movement, such as assembly robots and headphone wires, 70 to 100 is mandatory." Any thoughts on this? Are there any disadvantages to using wires with more strands other than cost and availability in smaller diameters?

Any important considerations I have left out?

[zimzom]

There were some V2 test flights on youtube. Considering that in the 40s they managed to get the first human objects into space. Its a pretty good indication on what could be done.

This was achieved with an analog computer guidance system. You can see from the picture its pretty delicate in places but overall the enclosure is rock solid. But look at the wire size. and the gyroscopes. I can't remember exactly but I think these had a primitive form of radio telemetry system.

[Dago]

Everything that flies usually has redundancy for the electronics. Usually meaning multiple flight computers for airplanes etc. If I would make something for a hobby rocket that should be reliable I'd look in to making two identical units and then somehow combining the signals. Maybe for the apogee detector that both of them could eject the parachute to make sure at least one of them manages to do it. Also try to minimize single points of failures (meaning use multiple wires and connectors for all cabling etc.). All connectors should naturally be locking. Maybe use something like micro-D connectors that have screws or something.

[Emil]

I'd look in to making two identical units and then somehow combining the signals. Maybe for the apogee detector that both of them could eject the parachute to make sure at least one of them manages to do it. Also try to minimize single points of failures

I think the easiest way to combine the output from two units is to connect one e-match/igniter to each and place both e-matches inside the same deployment charge. This will work with identical or different units.

[vvanders]

We rarely ran redundant system in most of the HPR I used to do. Altimeters we're ~$130 a pop which wasn't cheap and was even more expensive if you lost two of them because of a bad ematch.

That said it's a pretty good idea if you can actually swing it.

[Emil]

We rarely ran redundant system in most of the HPR I used to do. Altimeters we're ~$130 a pop which wasn't cheap and was even more expensive if you lost two of them because of a bad ematch.

That said it's a pretty good idea if you can actually swing it.

The backup could be something less expensive than an altimeter, such as a magnetic apogee detector or a timer.

I think there is an advantage to using two different systems in any case, to avoid situations where two identical systems would both fail. This could happen using (barametric) altimeters if the electronics bay is not vented correctly or if using accelerometers which don't filter the signal correctly (early accelerometers with hybrids).

E-matches should definitely be redundant.

[tszaboo]

I guess, that potting the electronics assembly would be advisable (after baking it). I could be really wrong.
Even if potting is wrong conformal coating is probably a must.

[Niklas]

Crystal
The crystal itself vibrates mechanically from electrical excitation, but with a much higher frequency. I am not sure, but mechanical vibrations in the MHz range seems unlikely to be a big problem. From the pictures at your website I can see no crystal on the PCB, are you using the internal RC oscillator instead? Is there anything that is critical with respect to timing that could not handle a 5% drift?

Isolation distance vs air pressure
At altitudes above 2000 meters over sea level, the isolation distances are usually increased to compensate for the less dense air. Perhaps not applicable for the 3.3 V supply but maybe for a higher battery voltage for the ignition?

Thermal effects
When travelling with a commercial jet at 12 km altitude, the outside air temperature is usually quite cold and could pose a problem. There is a teardown video of a weather balloon transmitter at mikeselectricstuff's Youtube channel. Styrofoam was used as a thermal insulator.
Have you checked the lower operating temperature limit of the components? The electrolytic capacitor for instance?

Connectors
Screw terminals sounds like a very bad idea in combination with vibrations. Spring loaded terminals might work as they can adapt to rearrangements of the copper strands and keep the pressure.
A better solution would be connectors with crimped terminals that has multiple gold plated contact fingers. Don't forget that both the PCB and the wire connectors should have the same plating material.

Wiring
Try to avoid excess wiring that can lead to sharp bends and loops. Squeezing in the cables and the shut the lid is not a good option.

Potting or conformal coating
Seems like a good idea to at least give the PCB a thin protective coating. Be aware that potting underneath components is not advisable as the potting compound usually has a different thermal expansion coefficient than the PCB laminate. The result can be broken soldering joints, not during the first thermo cycle but in the long run.
That electrolytic and perhaps also the connectors could require something that dampen vibrations.

[SeanB]

With crystals check the manufacturers datasheets for max allowed acceleration. The vibration can add a dither to the oscillator, but too many G can shear the crystal from the inner mountings. Typically use as small a crystal as possible, as it will be more rugged. Ceramic resonators can be used but need a conformal coating or a small potting to keep them from tearing off the board, as the legs typically are weak. They are also less precise, fine as a clock but not as a transmitter reference. You probably will have to at a minimum conformal coat the board both sides, bot to hold the components down but also to reduce water condensation from causing issues. Connectors have to be either latching types, or you will have to have a hole either side to tie them to the board. Use crimp contacts as a soldered one will fatigue from vibration on a stranded cable. Lace cable looms so they are both secure from vibration and so they will not put force on connectors with acceleration. Use a cable support for long cables, nothing complex just a rod to tie the cable to and mounted firmly at top and bottom.

[dannyf]

I'm worried about seeing similar effects in flight

Don't over-think. Just do it.

If there is a problem, then deal with it.

[SeanB]

With your design the only flaw I can see is the big capacitor. That is easy to solve using just a blob of electronics grade silicone ( non acetoxy kind that will not corrode copper) or flexible 2 part epoxy to hold it placed around it after assembly and test. Then it will not pull off the board with repeated use. Either that or use a heatshrink sleeve over that part of the board to hold it down. Will double as a mounting method as well, and should not interfere with the programming pads.

[Nerull]

Crystal
The crystal itself vibrates mechanically from electrical excitation, but with a much higher frequency. I am not sure, but mechanical vibrations in the MHz range seems unlikely to be a big problem. From the pictures at your website I can see no crystal on the PCB, are you using the internal RC oscillator instead? Is there anything that is critical with respect to timing that could not handle a 5% drift?

Isolation distance vs air pressure
At altitudes above 2000 meters over sea level, the isolation distances are usually increased to compensate for the less dense air. Perhaps not applicable for the 3.3 V supply but maybe for a higher battery voltage for the ignition?

Thermal effects
When travelling with a commercial jet at 12 km altitude, the outside air temperature is usually quite cold and could pose a problem. There is a teardown video of a weather balloon transmitter at mikeselectricstuff's Youtube channel. Styrofoam was used as a thermal insulator.
Have you checked the lower operating temperature limit of the components? The electrolytic capacitor for instance?

Connectors
Screw terminals sounds like a very bad idea in combination with vibrations. Spring loaded terminals might work as they can adapt to rearrangements of the copper strands and keep the pressure.
A better solution would be connectors with crimped terminals that has multiple gold plated contact fingers. Don't forget that both the PCB and the wire connectors should have the same plating material.

Wiring
Try to avoid excess wiring that can lead to sharp bends and loops. Squeezing in the cables and the shut the lid is not a good option.

Potting or conformal coating
Seems like a good idea to at least give the PCB a thin protective coating. Be aware that potting underneath components is not advisable as the potting compound usually has a different thermal expansion coefficient than the PCB laminate. The result can be broken soldering joints, not during the first thermo cycle but in the long run.
That electrolytic and perhaps also the connectors could require something that dampen vibrations.

Spring loaded contacts may be prone to opening under high G forces/extreme vibration, potentially causing a flight computer to reset and not deploy anything - the worst result possible for a high power rocket.

Screw terminals have been used in thousands of flights for decades and are pretty reliable, don't take up critical space, and are easy to use in the field. Any connector is going to need to meet those specs to be adopted.

[Niklas]

Spring loaded contacts may be prone to opening under high G forces/extreme vibration, potentially causing a flight computer to reset and not deploy anything - the worst result possible for a high power rocket.

Screw terminals have been used in thousands of flights for decades and are pretty reliable, don't take up critical space, and are easy to use in the field. Any connector is going to need to meet those specs to be adopted.

True, good point that I missed.

In trains, with a lot of vibrations, screw terminals can cause a lot of problems. Either the screw can unscrew itself or the copper strands can rearrange their positions, both with less contact force as a result.

[IanB]

On external surfaces heat can be a problem (sensors, cameras, etc.). Check out this video at about 5m30s:

http://youtu.be/rvDqoxMUroA

The external camera housing melted!

[Emil]

@Niklas

You are correct about the crystal, I use the internal RC oscillator. There is an I2C bus, I don't think 5% drift will be a problem for I2C(?). There is also a timer feature to fire the second pyro channel, but I don't think 5% drift during the burn would be a practical problem. However 5% could be a problem in many cases, such as when reading data from a GPS module over a serial bus.

The electrolytic capacitor is rated for -55C to +105C. The other components are rated at -40 or -55 to at least +85C. I don't think extreme temperatures are a problem, except for sensors which need to be exposed to the ambient air. In part because the time at extreme temperatures will be short (unlike a balloon) so only thin insulation would be required and in part because heating from atmospheric drag and the low ambient temperature would partially offset each other.

How do you prevent the potting compound from getting under the components? Just use a high viscosity compound?

[vvanders]

We rarely ran redundant system in most of the HPR I used to do. Altimeters we're ~$130 a pop which wasn't cheap and was even more expensive if you lost two of them because of a bad ematch.

That said it's a pretty good idea if you can actually swing it.

The backup could be something less expensive than an altimeter, such as a magnetic apogee detector or a timer.

I think there is an advantage to using two different systems in any case, to avoid situations where two identical systems would both fail. This could happen using (barametric) altimeters if the electronics bay is not vented correctly or if using accelerometers which don't filter the signal correctly (early accelerometers with hybrids).

E-matches should definitely be redundant.

Redundant ematches require more current draw though so there's upsides and downsides.

Most of the stuff that used electronics was a dual stage deployment. Drouge parachute/streamer at apogee and large parachute at ~800FT AGL. Otherwise you could be driving many miles to pick up your rocket.  Cases like that make it a bit hard to use different redundant system.

I'd actually love to get back into the sport again but we've just got too many damn trees here in Washington, unless you want to go over the mountains to eastern WA.

[Niklas]

You are correct about the crystal, I use the internal RC oscillator. There is an I2C bus, I don't think 5% drift will be a problem for I2C(?). There is also a timer feature to fire the second pyro channel, but I don't think 5% drift during the burn would be a practical problem. However 5% could be a problem in many cases, such as when reading data from a GPS module over a serial bus.

The I2C bus will work fine as it is synchronous (clock and data) and if you have taken +5% into account for the clock rate. For shorter time periods the drift will not cause problems. Check if you have something in the code that is more critical, like timeouts, watchdogs etc.

The electrolytic capacitor is rated for -55C to +105C. The other components are rated at -40 or -55 to at least +85C. I don't think extreme temperatures are a problem, except for sensors which need to be exposed to the ambient air. In part because the time at extreme temperatures will be short (unlike a balloon) so only thin insulation would be required and in part because heating from atmospheric drag and the low ambient temperature would partially offset each other.

For shorter times and with some heat dissipation on the PCB, that could be ok even at higher temperatures. LCDs sometimes have power resistors as heating elements to prevent lag. I have seen -56'C outside air temp display on the infotainment system on commercial flights. How about the battery?

How do you prevent the potting compound from getting under the components? Just use a high viscosity compound?

From the pictures of the PCB I would say that none of the components are in danger. Perhaps the electrolytic capacitor, if you have too much solder paste, could float up from the PCB surface. However, connectors are more sensitive. For example, SMD pin headers with 2.54 mm pitch have approx 1.5 mm between the PCB surface and the plastic former.

[Emil]

Redundant ematches require more current draw though so there's upsides and downsides.

Most of the stuff that used electronics was a dual stage deployment. Drouge parachute/streamer at apogee and large parachute at ~800FT AGL. Otherwise you could be driving many miles to pick up your rocket.  Cases like that make it a bit hard to use different redundant system.

I'd actually love to get back into the sport again but we've just got too many damn trees here in Washington, unless you want to go over the mountains to eastern WA.

More current draw was the reason I programmed my device to never fire both channels at the same time. One channel will always be delayed until the other channel has finished. I think two batteries would be a good idea if using two recovery devices, to avoid any single point of failure.

I didn't think about dual recovery.. A timer might work, but it would be hard to beat an altimeter for that..

[acbern]

I used to design spacecraft electronics and the question raised initially is a pretty board one. so it takes books to anser them. a few key guidelines nd simple, and not costly rules however can be observed to achieve good reliability:
-use conservative design rules (do worst case analysis, avoid failure propagation, e.g. one board failing and outputing ax voltage shall not damage subsequent board...)
-use proper derating. there are military standards out there that give guidance. this applies to parameters such as voltages, power, thermal...
-keep electronics cool. every 10-12K temp. increase reduces lifetime by about factor of 2.
-some part technologies are known for bad reliability and should be avoided: electrolytics, mechanical parts such as relays+switches, batteries. now sometimes this is not possible. but space qualified batteries are costly... crystals are not critical.
-try to avoid that the PCB is in the resonance frequency of anything around (motor...), bod heavy parts to boards, but without stressing leads
-avoid single point of failures (e.g., if you have a diode protecting a batteriy to discharge into some part of a circuitry, use two diodes in series). redundancy in general is regularily used in spacecraft.
-when doing the layout , and when soldering, avoid stress to components (bend components far from body...).
-in vacuum, rule of thumb, any component cunsuming more than 100mW should have conduction cooling and be cooled through the system, other parts are usually kept cool eoung through their leads. inner copper layers help distribute heat.
-moisure helps part aging, avoid it, conformally coat the board (but di not etrap agressive fluxes, these must be avoided anyways)

this is by far not exhaustive, obviously, and there is a reason why spacecrafts cost what they cost, but for non-industrial applications, a lot can already be done by proper design and observing simple rules as above.

[LukeW]

As with everything in engineering, it's always a tradeoff - how "good" do you want to engineer it, and how much do you want to pay?
Do you want to use military-grade aerospace radiation hardened components for everything, and have it cost $100,000 per board, for a hobby rocket project? The answer is probably no.

On the ZeptoMag board seen on your site, there are a couple of issues that immediately jump out at me:

1) What is the purpose of the pushbutton switch? Is it possibly susceptible to being accidentally actuated due to strong acceleration or vibration?

Obviously a pushbutton switch is not needed during actual flight although it may be needed during user configuration or software development.
Consider mounting it off the board on wires, or using a header or pads to solder wires to etc. Or just don't solder the button onto the boards for production flight use, or desolder them before flight.

There was a famous issue on Apollo 14 where a ball of solder came loose inside the abort switch, causing an unwanted abort signal to be sent to the guidance computer... a real headache at the time.

2) There are possible issues with the large electrolytic cap being torn loose from the board at high accelerations, since large SMD electrolytics typically have very small solder area on their legs relative to the size of the can. There are also possible concerns with the behaviour of an aluminium electrolytic at very low or very high temperatures or pressures.

Consider changing the cap to SMD tantalum or SMD ceramic. Or consider adding glue or silicone to provide rigid mechanical support for the capacitor can.

[Emil]

As with everything in engineering, it's always a tradeoff - how "good" do you want to engineer it, and how much do you want to pay?
Do you want to use military-grade aerospace radiation hardened components for everything, and have it cost $100,000 per board, for a hobby rocket project? The answer is probably no.

On the ZeptoMag board seen on your site, there are a couple of issues that immediately jump out at me:

1) What is the purpose of the pushbutton switch? Is it possibly susceptible to being accidentally actuated due to strong acceleration or vibration?

Obviously a pushbutton switch is not needed during actual flight although it may be needed during user configuration or software development.
Consider mounting it off the board on wires, or using a header or pads to solder wires to etc. Or just don't solder the button onto the boards for production flight use, or desolder them before flight.

There was a famous issue on Apollo 14 where a ball of solder came loose inside the abort switch, causing an unwanted abort signal to be sent to the guidance computer... a real headache at the time.

2) There are possible issues with the large electrolytic cap being torn loose from the board at high accelerations, since large SMD electrolytics typically have very small solder area on their legs relative to the size of the can. There are also possible concerns with the behaviour of an aluminium electrolytic at very low or very high temperatures or pressures.

Consider changing the cap to SMD tantalum or SMD ceramic. Or consider adding glue or silicone to provide rigid mechanical support for the capacitor can.

The goal (for this project and a similar one I'm planning) is not to have the highest possible reliability at any price, but rather to have the best reliability which is achievable at a price in the same range as other electronics for hobby rockets with similar features. No $100k boards Mass, size and convenience of use is also relevant.

However my question was deliberately broad, to encourage interesting discussion on the topic.

1) The purpose of the push button switch is to let the user change some settings. This might be done while prepping the rocket on the pad or in advance. I don't think vibration will be a problem as the software will not do anything if the button is pressed for less than 500ms. This time could be increased. I don't think any realistic acceleration could turn on the switch, however that datasheet for the KMR221G LFS switch does not specify maximum acceleration or vibration.

2) The cap is only 6.3mm wide and 5.7mm tall so I didn't think it would be a problem, but I will take you advice and add some epoxy for now and consider replacing it with ceramic caps. If everything else works, the cap is not needed but it powers the board if the battery is temporarily disconnected (maybe from vibrations during flight) or if the voltage of the battery drops too much when fireing the pyro channels.

[Alphatronique]

HI

ceramic cap was good but my break so better to put many small in parallel that a single bigger one
if one break you lost capacity but not all ,and smaller one have less mass

and may good idea to use thinker PCB like 0.93mil instead of 0.63mil ,less flex = less opportunity to break cap

conformal coating right after board was help a 80 deg for few hour was good idea for keep it degassed

as connector military avionic have some really nice circular connector that come whit separate pin
(you order pin separately then push it on connector shell ) that pin alone was not expensive and make good contact
 

[Emil]

ceramic cap was good but my break so better to put many small in parallel that a single bigger one
if one break you lost capacity but not all ,and smaller one have less mass

If I understand you correctly, ceramic caps can actually break if there is some flex in the board?

[Niklas]

...ceramic caps can actually break if there is some flex in the board?

Yes, they can. Both mechanical stress and thermal stress can cause cracks. For some applications, hand soldering using a soldering iron, is strictly forbidden due to the non-uniform heating of the capacitor package. The risk for thermally related cracks increases with package size. There are ceramic capacitors with soft terminations (Flexi-Term etc) that can handle minor bending forces.

There are two fault modes for cracked capacitors: open circuit and short circuit... In some automotive applications you are not allowed to use a single capacitor between supply and GND. Either a resistor or another capacitor must be put in series to minimize the impact from a single fault.
 
http://www.digikey.com/Web%20Export/Supplier%20Content/TDK_445/PDF/TDK_cracking.pdf?redirected=1