The 10 or 20 year electronics project

[Griffin]

Hi every1

Just an idea I want to put out there if anyone is interested. After listening to The Amp Hour podcast #132 Dave and Chris talked about the idea of electronics being inherently more reliable because they where allot less complicated in the olden days and it made me think:

What if we took some of the leading hobby platforms like Arduino, Raspberry Pi … and designed a project for them aimed at working for the next 10 to 20 year uninterrupted. Sort of a statement for future generations.

What do you think? Any ideas, goals or just plain wacky things we could do

[Ed.Kloonk]

20 Years? Nice idea if you could make the arduino run for just 20 hours without a reboot.

 

[JoeyP]

That sounds like a good idea. Maybe something as simple as a perpetual clock powered by scavenged RF power - though AM radio stations will probably be entirely replaced by the internet within 20 years.

I bought a digital clock for $0.50 at a garage sale more than 30 years ago, which has been running continuously since that time (9V battery back-up during power outages - but no display on backup).

[Griffin]

Figured it would need to be powered by something else than divine power but that’s where the trick comes in. Make it something powered by Solar, Heat, Interchangeable batteries something like that.

[SLJ]

I've got digital meters older than that that still work untouched and tube radios from the 1920s.  20 years is nothing.

[EEVblog]

The basic project would be a clock of course. Just tick over the seconds until last reboot.
The hard part is ensuring the battery source lasts, and you have a fail proof change-over mechanism.
And then if it does reboot, what does that actually tell you? Has it actually failed in some way, or is that just some outside transient event that the Arduino/Pi/Whatever was never designed against? (e.g. cosmic rays)

Dave.

[Griffin]

20 years for old school tech yes..... but what about an embedded system from the current year. How reliable with they be in 20 years time.

[Griffin]

I apologize for double posting this at another form but wanted to see the responses. JohanEkdahl over at AVRFreaks.net suggested a time capsule idea. Something with an embedded message that gets lost if the device looses power or something interrupts the main loop.

[G7PSK]

Run for twenty years continuously without external power supply and battery changes, first find some plutonium, the rest should be fairly easy just a 555 flashing an led would do.https://www.eevblog.com/forum/Smileys/default/smiley_laughing.gif

[AlfBaz]

It would be interesting to see if the flash in these embeded things would last the claimed time if they were continuously on

[Griffin]

@G7PSK i think your missing the point here. The objective is to design something on a hobby level using a hobby based embedded system add battery backup, mains/solar power or all of them and design it to work for 10 - 20 years. Think of the considerations that would need to be made. Part lifetimes, battery choice, power decay, backup backup power, standby time. Everything a million dollars would get you at NASA but built by hobbyists.

[westfw]

the idea of electronics being inherently more reliable because they where allot less complicated in the olden days

Isn't that a lot different than "run on a 20y power source"?  I mean, power sources are still about as reliable as they ever were (assuming, I guess, a design other than those counterfeit switchers), and I think both batteries and grid power are better than they used to be.

The main reason that electronics doesn't last is that we know it's going to be "obsolete" in 5 years anyway, and we treat it (and build it) accordingly.

[free_electron]

20 years ? That's not a measure of quality. There's HP 200 oscillators out there that were built in 195x that still work today... They survive power surges , dropping them, shorting the output , overloading the output. Whatever. You're gonna need some serious protection on that hadwino. Start by finding a rad-hard atmel cpu...

Take a look at the following project : the clock of the long now. It ticks once a year.. Designed to run for 10000 years. Designed by Danny Hillis ( yes, THE Danny Hillis that invented the connection machine. The worlds craziest supercomputer )

Longnow.org

[Griffin]

The idea is not to design a system that will run of a single charge for 20 years or to measure quality but to build a device from ordinary off the shelf parts that will trigger something stored and kept active in memory at the end of that time period. This would then force a design that can keep that device up and running for 20 years just to complete that 1 task at the end of the time period.

With a deliberate failsafe build in that if power is interrupted what was in memory is unrecoverable or change the outcome.

[JoeyP]

20 years ? That's not a measure of quality. There's HP 200 oscillators out there that were built in 195x that still work today...

I don't think he's talking about something that could be used occasionally, and still be able to power up after 20 years. I think he's talking about continuous uninterrupted operation for 20 years. Those old HP 200 oscillators may still be functional, but they  have not run 24 hours a day with zero maintenance since 195x.

Just don't build it lead free

[Alana]

Are we talking about continuous running for more than 10 years or are we talking about device that will not brake if actively used for those 10-20 years?
I'd concentrate on the latter - to make something that would be useful after those years. For example - i have a digital multimeter - clone of  M890G that i got more than 10 years ago and its still working. And i wonder if my unitek that i got last year will work after same period of time.

[TerraHertz]

Why think so small? Go for 10,000 years. Like these people:

  http://longnow.org/clock/
  The 10,000 Year Clock


Personally I'd like to make something solar powered that would last 100 years.
Last time I tried something like this, I made the mistake of including batteries, that got charged by the solar cells. BIG mistake.

See pics. It's a kind of sound-light sculpture. Solar cells charge battery, and also the light level is sampled regularly and the sample makes the latest 'note' in a ring buffer that plays a tune while the thing is illuminated. Well, more of a droning-chirping, but anyway... It takes a few days to overwrite the ring buffer, so the tune playing at any time is a kind of reflection of the recent sunlight profile. The more sunlight, the more cheerful it sounds. The 'speaker' is a piezo disk glued to an inside face of the sealed glass cube.
Also, there's a flash tube that fires at random times, about once a day. Including in the dark. Quite amusing when it goes off nearby in the dark unexpectedly.

Too bad the whole idea was sabotaged by shitty NiCad battery technology, and the thing is sealed so the batteries can't be replaced without cutting it open. And since I know of no battery technology that would last a respectable number of years, there's no point. Idea = fail.

[TerraHertz]

JohanEkdahl over at AVRFreaks.net suggested a time capsule idea. Something with an embedded message that gets lost if the device looses power or something interrupts the main loop.

Why would you want to make it lose the data if power was lost?

Also, it should be something that is continuously interesting, for as long as it lasts. Not just a one-pop wonder.

How about this:
Solar powered weatherproof object. Power is stored in a supercap so it can operate through brief intervals of darkness. It has a microphone, and a speaker (both impervious to weather), a VOX circuit with software analysis to detect human voices. Over the course of a day it records in CMOS ram just the first 30 seconds of any voice sequences that contain at least 30 seconds of continuous speech. At the end of each day it compresses and transfers ONE of those sequences (randomly chosen) to a large flash memory. Typical flash memories should be able to hold centuries of such once-a-day fragments. Organized as a ring buffer, the data doesn't start to get overwritten until full - in hundreds of years? Also the system should once a year or so, do a 'rewrite' of the entire memory to refresh the cell charges. Minimize the wear on the flash cells, but don't let the data 'evaporate' over time.

During daylight hours, every day, it will also now and then randomly select fragments from its sounds archive, and play them back. Maybe two or three times a day. Or maybe it would do it as payment for someone talking to it.

I'd imagine this as some kind of very solid stainless steel pillar, set into bedrock in scenic, not-so easy to reach places, that not so many people would visit. The idea would be that people could speak to it, knowing their voices *might* be preserved and sometimes played back, for centuries. If only one person visits in a single day, their words are guaranteed to be preserved.

[G7PSK]

@G7PSK i think your missing the point here. The objective is to design something on a hobby level using a hobby based embedded system add battery backup, mains/solar power or all of them and design it to work for 10 - 20 years. Think of the considerations that would need to be made. Part lifetimes, battery choice, power decay, backup backup power, standby time. Everything a million dollars would get you at NASA but built by hobbyists.

I was being facetious I will admit. The problem with building any electronic item that will survive ten or twenty years is not only the battery but also such things as capacitors, will will often go toes up without use, the fact that some items of electronics are running after many years is more to chance than design in most cases. It takes the type of component that NASA uses in its deep space probes to be sure of a long life.

[ignator]

Small nanometer silicon geometries, and very small and close spacing of metalization layers, are already leading to failure of consumer electronics.  It does not take high current densities (this accelerates the process), especially if constant power is applied for 20 years.
This is a problem in avionics today (my domain, this specific area is not), and it's an issue especially when aircraft life can exceed 50 years.  Avionics upgrades are done in a 20-25 year period, but only now is this industry dealing with nanometer geometry.
Now mix in the small geometries that have large cross sections, with single neutron upset events, and you will have a state machine reset, even without component failure.
If you want to do this, go back to the earliest generation of electronics that are low power enough for 20 year operation.

[JackOfVA]

One of the more challenging "design for long term reliability" projects was the Bell System's first US-Europe voice cable system (TAT-1) --all vacuum tubes, of course. And, it's not a trivial task to replace a failed vacuum tube when it's several thousand feet deep in sea water in the middle of the Atlantic.

The January 1957 Bell System Technical Journal is devoted to TAT-1 http://www3.alcatel-lucent.com/bstj/vol36-1957/bstj-vol36-issue01.html -- makes for interesting reading.

[lapm]

I would say this idea of 20 years project starts to go dangerously close to something nasa engineers does. Systems with multiply failsafes, etc... In fact thats probably how it should be treated..  2 decades starts to be long enough time frame that ageing effects of components and materials starts to play big role on desing. Not to mention few other things i cant event dream about...

But nasa has one trick that might help to get this started, use the oldest tech you can find. Chips that are no so highly integrated, etc... Those now retired spaceshuttles had really really old tech in flight computers.

I cant even imagine how current plastic capsulated few nanometer line width microcontrollers would even work for 20 years from now. Thermal cycles, oxygen creeps in, etc... Would be interesting to find out. Anyone got time machine? 

[GK]

My K-mart cheapo "Audiosound" digital alarm clock has been going since about 1994.  Must have been designed by NASA

[G7PSK]

The only sure way of having something working in twenty years time is to build several million of them, then the law of averages might work for you.

[george graves]

20 Years? Nice idea if you could make the arduino run for just 20 hours without a reboot.

 

What makes you say that?