Shelf life of assembled electronics

[pcm81]

Photography is a hobby of mine and often when discussing high end nikon cameras and lenses the comments get made like "it will outlast you". Which is true at-least about lenses, i own a Nikon lens that is 2 months older than me and unlike me is in a perfectly usable condition. But unlike old style film cameras, the large number of very small components (transistors) in the DSLR cameras got me wondering: what is a shelf life of a package of electronics, such as a DSLR? I am not talking about wear and tear from the use, that varies greatly; i am also not talking about a camera being rendered useless due to new technologies being developed. What i am interested in is the actual expected shelf life from aging of electronics. 20 years? 30 years? More? Less?

[james_s]

Electrolytic capacitors tend to have a shelf life, anywhere from several years to several decades depending on the specific type and environmental conditions. Other than that electronic components don't really have a definable shelf life, there is a lot of 40 and 50 year old solid state equipment that still works. Rubber, foam and plastic parts can deteriorate with age.

[unknownparticle]

Very difficult to define. As stated above, caps are the component most likely to fail over just time, even in storage, followed by resistors, their most likely failure mode is going high resistance. Semi conductors can also fail in storage, due to package issues. Surface mount components are less reliable than through hole.  All the forgoing is the reason Mil Spec components are produced for the military.  Regarding camera's, I don't use or even have a digital camera, but have many 35mm types.  The earliest of these which have active electronics are from around 1978, and are still working perfectly. However, I have a Nikon F5, a late 90's model that developed an auto focus fault, as in, it doesn't!!  I've never had it repaired as it would cost serious money, it would be cheaper to buy a good used one. It has a huge amount of electronics as it was one of the most sophisticated 35mm film cameras ever produced, so more likely to fail. It isn't the focus motor, as that can be heard hunting. It is impossible to predict when components will fail, but they are spec'd with a theoretical MTBF life, Mean Time Before Failure. Taking the most unreliable, shortest lived component, the electroleaktic crapacitor, they are lifed in thousands of hours when used at max voltage and max temp, the number of hours depending on the quality of the cap. Even then, they can well exceed that theoretical life, or, fail well before!!!

[pcm81]

Everything stated above makes sense to me, but there are also additional factors like solder, oxidation etc.
Mil-spec items are actually rated for harsh environment MTBF, they are not really rated for shelf life. I work in aerospace.
Hence is this oddball question. So far we lived in a world where new features of new electronics outpace lifespan of older electronics, hence the upgrade cycle. I am actually typing up this post on a computer that is 11 years old and i used it every day.
Naturally an assembly of components, like pcb, cannot have longer lifespan or shelf life than the individual components used on it; however it would seem like the assembly should have shorter lifespan than shortest lifespan of individual components, since assembly is an item of higher complexity and may be more prone to faults or additive behavior of faults.

That's why i was curious if anyone did any research into lifespan of assemblies.

[Kasper]

Flux can cause problems over time if not managed properly during production.  I'm guessing that's not too common in the type of company that makes cameras.  Probably more likely in smaller operations.

I once worked in a messy little company that didn't believe in this or in ESD.  They had problems with both.

[Berni]

Electrolytic caps are the typical component with a shelf life.

However these days firmware is stored in flash memory and it has a rated data retention time. The electrons gradually leak out of the flash cells and flip bits. This is so rampant that pretty much all high capacity NAND flash memory made in the last 15 years has built in ECC to hide how unreliable it is. However if enough bits flip even ECC won't recover it and flash data retention times are getting worse and worse due to the flash cells being made ever smaller to pack more capacity in, at the same time multiple analog levels of the cells are introduced to pack more bits into a cell.

So there is a real possibility that the firmware could become corrupt after 30 to 50 years of sitting around. If you are lucky it might happen in the application part, so you just have to reflash the firmware to get it back to working. If you are unlucky it might break in the bootloder, bricking the whole device.

[TERRA Operative]

Some HP and Tek gear from the '80's has the bit-rot problem already.with sone UV erasable EPROM chips.
I wonder if one-time programmable PROM chips are more reliable long term?

[james_s]

I've seen bit rot in EPROMs and I've had mask ROMs fail in old arcade games. SRAM seems to be the least reliable part in those, followed by ROM, followed by jellybean logic.

[Berni]

Yeah i did also had a ROM chip go bad in an old HP semiconductor analyzer from the 80s. Luckily a kind user on this forum gave me the binary image of its original contents. Then i found a guy that specializes in fixing old arcade machines who has the equipment to burn me a new copy of that obscure ROM chip and mail it to me. Quite the adventure that one.

Oh and this brings us to the other ticking timebombs in older equipment. Battery backed SRAM memory. Be it those Dallas SRAM chips with a built in battery that runs out in a few decades, or even worse external memory backup batteries that leak corrosive crap over the PCB and eat away the traces.

[hans]

Indeed, FLASH memory can be relatively unreliable. Cheap USB drives and SD cards have over-provisioned FLASH storage, so e.g. a 120GB drive may have 128GB of FLASH. Some sectors can be marked as bad so the drive remains at capacity. The controller must do ECC on the fly or verify data when it's written. That's why some of these cheap storage devices can have a horrendous write speed that fluctuates a lot as well, especially when a device is long unformatted or not properly TRIM'ed (if that's supported). Sometimes if you do a full format the device "fixes" itself. Very similar thing happens of SMR hard drives (or even bitrot in CMR drives), but because of different reasons.

Regarding components: I would say anything mechanically related is prone to fail. Switches, knobs, dials, connectors, etc. can get bad contacts. Motors, relays, or even wiring etc. can suffer from moisture/water issues and get stuck or have bad connections. Some assemblies may subject PCBs to some constant mechanical load which can fail over time. Heat cycles can crack connections, especially some lead-free solders that are brittle. Liquid caps (like aluminium non-polymer) can dry out and have worse specs. Small-node semiconductors running extremely hot for long durations can suffer from electromigration, degrading their performance and eventually fail otherwise 'stable' specs. Rechargeable batteries degrade over time and use, especially when poorly managed (e.g. the mfgr used up the whole lower and upper end of the battery cell range to get a certain battery life)

Now I do think that 'quality' can help for a lot of these aspects. Self-cleaning contacts can help. Components subject to moisture or even water may need coating or sealing to prevent oxidization. A device should be constructed mechanically sound. Heat stress should be minimal or with the least amount of 'thermal shock' as possible. Capacitors can be overspec'ed to account for degradation. Batteries should be well managed instead of abused heavily.

But of this list, I think only the first 2 I can say are convincingly issues related to shelf life.

[Siwastaja]

I call bullshit about Berni's comments about the flash.

If the camera designers are not total morons (which they could be, though, in which case I will stand corrected), for the firmware, they use the kind of flash which
1) is NOT unreliable, and does NOT hide the unreliability by ECC. Might not have ECC at all.
2) has data retention of at least 50 years or so guaranteed by manufacturer, and this is at elevated temperature.

This is usually internal to the microcontroller(s), sometimes external, with size in hundreds of kilobytes to maybe a few megabytes.

This is totally different to USB flash drives or SD cards which use flash technology with far higher storage density, the focus being least $ per GB, with (sometimes) questionable retention time, unless you are careful what you buy and who you believe.

But, by default, I would give the microprocessor manufacturers benefit of doubt and assume their guarantees hold. What else can you do, after all. Same for other parts with shelf-lifes.

[AndyC_772]

Photography is a hobby of mine and often when discussing high end nikon cameras and lenses the comments get made like "it will outlast you". Which is true at-least about lenses,

I've just been through, let's call it, a relevant experience. I decided to get back into photography after a few years' hiatus.

I was planning a holiday which would involve wildlife photography, so I figured it would be a good idea to test my equipment and re-familiarise myself with how it all worked.

It didn't start well. My Canon 100-400 L IS threw intermittent error codes whenever I turned IS on. The repair centre diagnosed a faulty IS unit, but said the lens was no longer supported by Canon. No spare parts available. I sold it as-is and bought the mk II version to replace it, which turned out to be a much better lens anyway, but it was an expense I could have done without.

A few weeks later I went out with my 24-70 f/2.8 L, and came back to find some shots significantly mis-focused at f/2.8. Turns out this is quite common and is due to mechanical wear causing slop in the mechanism, and should be easily sorted with a routine service. But you've guessed it, no longer supported and no parts available. Once again, I sold it as-is and replaced it with the mk II, which is slightly smaller and lighter but otherwise barely any different as far as I can tell.

Feeling neglected, my 50mm f/1.4 prime decided to take a flying leap out of the cupboard where I store my kit. It has a focus mechanism that's oddly fragile, but fortunately spares are readily available and inexpensive, and there's a really good video on YouTube showing how to fix the exact focus issue that it developed as a result of its drop to the floor.

In the meantime I sent in my 5D mk III for service, whilst service is at least still available for that model. I still remember buying it new, and there's barely a mark on it... can't quite believe how old it actually is already. A dodgy rotary encoder needed replacing, which was surprisingly inexpensive, but it was still over £200 all-in to get it shipped, cleaned and calibrated. Still, I have a camera that should last well now, which is just as well as it can't be too long before spares for that model also dry up.

I also cleaned the focus screen of my 1DX mk II, and thought I'd scratched it in the process. (Note: this is *very* easily done - though the 'scratch' in this case actually turned out to be a fibre from a cotton bud, which was a relief). Nevertheless I was dismayed to find that the standard focus screen for this camera is - you've guessed - already discontinued, and the only place I could find a replacement was an Ebay seller in China. IMHO it beggars belief that such an easily damaged, critical part of a 1D-series camera should already be unavailable as a spare.

Finally, at least for now, another unfortunate accident befell my 70-200 f/4L. This was by some margin the oldest of all my lenses, and surprisingly, it's still under service support. Except in this case, that's irrelevant, as the cost to repair the apparently minor damage to the focus mechanism exactly equals the cost to buy a good used replacement. Odd, that.

My advice? Don't expect kit to last forever, because even if you don't use it much, even a minor problem could easily write it off if spares are no longer available. I've spent a small fortune over the last few months, on kit which should have been easily fixed, but couldn't be due to the lack of spare parts.

Instead, go out there, use your camera, and get the best value out of it that you possibly can.

[Berni]

I call bullshit about Berni's comments about the flash.

If the camera designers are not total morons (which they could be, though, in which case I will stand corrected), for the firmware, they use the kind of flash which
1) is NOT unreliable, and does NOT hide the unreliability by ECC. Might not have ECC at all.
2) has data retention of at least 50 years or so guaranteed by manufacturer, and this is at elevated temperature.

This is usually internal to the microcontroller(s), sometimes external, with size in hundreds of kilobytes to maybe a few megabytes.

Often bigger SOC chips are being used and those boot from external flash that often ends up being some form of NAND flash.

For example Keysight had issues with there most popular X3000 series scopes where early stage boot code would bitrot away rendering the scope completely bricked (official fix is sending it back to be serviced for free). So if it can happen to one of the worlds largest test equipment manufacturers, then id argue it could happen to one of the many digital camera manufacturers too.

[tom66]

It very much depends on whether the firmware engineers have done their homework and know what they're doing.  And the hardware engineers must also appreciate the limitations of the hardware itself.

I was involved in an engineering project for a set top box that trick-played to the internal eMMC.  For those unaware, trick play is the function of being able to play and pause live TV, skip ads and so on.  In order to support that, you need to record the active stream of video to a memory device.  This is usually the demultiplexed compressed stream.

I think most people on this forum can see the issue already.  If you assume an MPEG-4 bitstream of about 10Mbit/s (1080i30 Freeview DVB-T2), 16GB memory device, after 1 year of usage 8hrs/day you are at 13TB written.  That is already 800 cycles across the NAND.  The Toshiba eMMC devices we were using were based on QLC NAND, a new technology at the time, and had a life time of about 1500 cycles.  However that lifetime was given at Tj=25C.  At Tj=50C, the lifetime was only 750 cycles.  The NAND got quite hot under continuous write and read operations, so 50C was a lot closer to reality.

To make matters worse the memory device was located about 5cm away from the main SoC, and when it died, it would bring the whole system down as the kernel debugger was left switched on printing kernel debug messages on an inaccessible UART about how the eMMC was not able to read or write blocks because of a device error.  The result is the box would take 10 mins to turn on and change channel 30 seconds after the remote control button was pressed.

And this could happen 18 months after you bought the device with relatively average usage.  I believe in the end a firmware update was shipped that turned off the debugger and we sent free USB HDD's out to customers who asked for them because trick play was now dead on their less than 2 year old box.

For what it's worth as an intern I tried to highlight this as a design flaw, but I was ignored.  The company pulled out of the UK market a few years later and focused on LatAm and so on.

[PlainName]

LCDs can go off in various ways. Some stuff I am looking at currently has been in storage for between one and two decades. The displays suffer from blooming, missing columns, permanently dark splodges and being completely unresponsive (I guess that could be all columns missing).

[lasmux]

Most devices out there, including flash storage etc aren't really tested for 'shelf life' when the device isn't powered on, as it's normally so long to make it an irrelevant problem, so it's a difficult question. Anything with a chemical process will have a more defined shelf life, such as already mentioned electrolytics, but also things like gas sensors etc. Another factor is dendrite growth due to ROSH compatible solders. Since everyone started using SAC305 for everything, we may in future years start seeing a lot of devices starting to fail in unexpected ways. That said the device has to be powered on for this to happen, so it's not fully applicable to your question.

[Kasper]

I call bullshit about Berni's comments about the flash.

If the camera designers are not total morons (which they could be, though, in which case I will stand corrected), for the firmware, they use the kind of flash which
1) is NOT unreliable, and does NOT hide the unreliability by ECC. Might not have ECC at all.
2) has data retention of at least 50 years or so guaranteed by manufacturer, and this is at elevated temperature.

Are you assuming designers make 50 year lifespan at elevated temperature their top priority?  How many consumers do you think care about 50 year reliability?  The number of people who purchase phones without replaceable batteries is evidence that many consumers do not care for even a 10 year lifespan.

[iJoseph2]

This is an interesting thread... I really hope the flash in older cars lasts over 50 years.
The engine ECU is obviously important, but the ABS units have the ability to actuate the brakes ... I can imagine a failure mode could even actuate the brake on just one wheel and be very dangerous.

[tom66]

The good news is that I can't think of an ECU that would not checksum its main flash memory and bail out if the checksum fails to match.  In the ABS case, that would light the fault lamp and land you a repair bill for a new ABS computer, but probably not end up with you in a ditch.

Also the flash in these systems tends to be of a higher grade and much lower density, the program for an ABS computer or ECU is probably under 1MB and would be stored in NOR flash rather than the higher density NAND.  NOR flash has much better lifespan.  If it is NAND, it will be infrequently written and will be of a higher standard, SLC or pseudo-SLC or similar.

Automotive systems have much higher standards generally but then you have manufacturers like Tesla using eMMC in important infotainment systems (big touchscreen that controls the speedo and defogger) which can still go bad.  I am most worried about those because they are probably quite expensive to repair and nowadays they control a lot about a car, I know for instance on an EV you wouldn't be able to change charging settings if the head unit went bad.

[Siwastaja]

Are you assuming designers make 50 year lifespan at elevated temperature their top priority?  How many consumers do you think care about 50 year reliability? 

It's not about 50 years, it's about the fact that if designers fuck this up, it happens like tom66 described - snarky, overconfident designers go on dismissing the problem interns bring up by going "oh well, 50 years at elevated temperature is not our top priority" strawman, while failing to actually define what their lifetime target is or provide any calculation of it whatsoever, and as a result, it lasts only a year, and company loses a lot of money fixing the problem, maybe resulting in failure of the company.

No need to be "top priority". Normal design practices are enough - use internal MCU flash for firmware, check the ratings of any external flash -, and flash life is not a problem. And if you fail to do that, who knows how badly you failed it?

50 years was just an example of being completely normal rating by nearly all MCU manufacturers, nothing expensive or special about that, contrary to what you seem to think.

[Siwastaja]

For example, a cheap commodity microcontroller from ST, is guaranteed:
Data retention: 1 kcycle (2) at TA = 85 °C: 30 years
(https://www.st.com/resource/en/datasheet/stm32f401re.pdf page 87)

And if they guarantee the parts for 30 years at +85, it is obvious to the designer a shelf-life will never be a problem, actual problems start appearing long after 50 years, unless ST produces crap.

You get similar guarantees for all flash based microcontrollers. No one would buy a microcontroller if the flash was rated for 15 years at 25 degC.

[harerod]

...
I've just been through, let's call it, a relevant experience. I decided to get back into photography after a few years' hiatus.
...

What a sobering report. Thanks for the heads-up. I wouldn't have expect this from Canon.
While everybody with enough dough goes mirrorless, I expected to use my DSLR gear for the next decade or so. 5DmkIV, 6DmkI, 700D, 600D, same 100-400mm as you, gave away an inexplicably broken 70-200mm F/4 (aha!). My better lenses are Tamron these days (among others, a 150-600mm G2, for the birds).
Have you checked the situation with independent repair shops? Can they get hold of, for lack of a better term, Chinese replacement parts?
Canon maintenance has always been expensive and Tamron Germany won't even touch my lenses, because I had the audacity to buy them (new) in Japan.

At any rate, your report should prove that investing in a mint condition camera equipment collection may have its hazards.

[Kasper]

Siwastaja, you called bs on a comment about FW retention of 30-50 years.

Then you linked an MCU that is rated for 30 years.  Does that mean Berni's comment was not bs?  Or are you saying anyone that uses that is a "total moron"?

You originally said at least 50 years at elevated temp or you're a total moron.  I guess what you meant was "about" 50 years? 

You've used a comment about cycle life related to streaming video to support your argument.  Does FW typically get rewritten hundreds of times? 

[Siwastaja]

Siwastaja, you called bs on a comment about FW retention of 30-50 years.

Sorry for failing to quote the exact part - the claim that flash is becoming more and more unreliable and embedded devices "mask" this unreliability by ECC is BS - unless the designer puts the firmware on some random cheap SD card or similar unsuitable flash technology, which is, I think, moronic and a mistake.

The 30-50 year part of the comment is not BS IMHO.

I guess what you meant was "about" 50 years?

Yes, thank you.

You've used a comment about cycle life related to streaming video to support your argument.  Does FW typically get rewritten hundreds of times?

I commented about the attitude of ignoring such obvious issues, even when pointed out to the designer. It doesn't matter if the exact mechanisms is wear out by too many erase-write cycles (the video time shift example), or due to aging by choosing completely inappropriate flash technology when the right technology is right in the front of their eyes - in either case, designer failed to do their job. And for the firmware part, it's not rocket science, choosing a normal microcontroller gets you into the guaranteed 30-50 years retention. And now this is relevant to the OP - if the flash is guaranteed to 30 years at +85 degC and being written 1K times, actual retention at +25degC and 1 write (at production) is going to be significantly more, possibly 100 years.

Sorry for lowish-quality post which was argumentative for no reason. I really just wanted to comment about the ECC part since the higher-end MCUs which have the ECC at all, this ECC can be configured to generate a failure interrupt, where the failure can be logged and handled - quite the opposite of masking the "increasing unreliability" of flash.

[Kasper]

Siwastaja, thank you for clarifying.  When you try to discredit people's posts, it's good to include a clear explanation.