What’s going to keep a clock running for a century, unattended? Well, whatever’s running it will have to sip power, and it’s going to need a power source that will last a long time. [Jan Waclawek] is looking into solar power for daytime, and capacitors for nighttime, to keep his clock running for a hundred years.
This project carries on from [Jan]’s previous project which looked at what kind of power source could power the gadgets around his house for a century without needing intervention – ie., no batteries to replace, no winding etc. [Jan] whittled his choices down to a combination of solar power and polypropylene film capacitors. Once the power had been sorted, a clock was chosen in order to test the power supply. The power consumption for a clock will be low during the night – it would only need a RTC circuit keeping track of the time – so a few low-leakage capacitors can be used. When daylight returns or a light is switched on, the solar circuit would power the clock’s display.
At the moment, [Jan] has a proof of concept circuit working, using the ultra-low-power microcontroller on a STM32L476 DISCOVERY board and a few 10 μF 0805 size capacitors, when fully charged by the solar panel, the clock’s display lasts for about two minutes.
Take a look at [Jan]’s project for more details, and check out his previous project where he narrowed down the components for a hundred-year power supply. [Jan]’s prototype can be seen in action after the break. Also take a look at this master clock that signals slave clocks and runs for a year on a single AA battery.
53 thoughts on “How To Run A Clock For A Century”
Seems like a more tested approaches are: https://en.wikipedia.org/wiki/Oxford_Electric_Bell or https://en.wikipedia.org/wiki/Beverly_Clock
I agree, the Beverly Clock approach is the most sensible, and nothing made from silicon semiconductors is going to last anyway due to dopant migration etc.
You could always just build a system with discrete components… running at low frequencies and voltage it would probably operate for a century many electronics from the 70’s still operate today just fine. One contributing factor to older devices operating longer is no use of flash memory in the 70’s… calculators from then up into the 80’s used masked roms. Eproms are speced up to 40-100 years also these days, so you could probably get away with a redundant pair and a low frequency CPU. I’ve actually had a few older flash drives die just lying in a drawer didn’t quick working it just bit rotted. FRAM might be a good fit for such a device since it has longevity and can act as ram… and might outlast a regular SRAM while at the same time being nonvolatile. Another idea might be to use a CPU with a static core and async sram that can run down to DC like the W65C802 and use a 555 to generate a clock signal to it every time a second passes using a mems oscillator to drive the inputs to 4 4bit binary up/down counters 4 4 bit comparators and a 7421 quad input AND to drive the 555 as well as clear the counter. Probably some holes in that idea though… but it would have the most fragile part the CPU running relatively infrequently to maximize it’s life span as well as making the CPU frequency independent of the clock.
Yes very careful selection of parts would be required, like for a deep space probe.
Dopant diffusion is incredibly slow at room temperatures, electro migration would be a bigger concern though such a circuit will not be driving a lot of current which minimizes this effect. These ICs should be able to stand the test of time.
Meh, it is the principle, that level of tech is not made to last so one sort of component or the other is bound to fail on you. What is the mean shelflife of modern electronics, 20 years at best? If we stopped making computer gear how long before we were in trouble and started running out of computers, not long at all I suspect.
Ain’t nobody gots time fo dat..
Some sort of RTG would be perfect :-)
How about a solar cell and a very accurate gnomon…then take away the solar cell. Not much good at night nor on stormy days, but it’ll run for centuries.
I’ve got a Casio Wave Ceptor solar. Had it for at least 10 years, never changed the “battery”, and never had to adjust the time on it. Best watch I’ve ever had, beats my £300 Seiko Kinetic hands down, which I ended up giving to my in law while I still use the Casio. I even had to change the cell (cap? battery? can’t remember) on the Seiko.
I presume that the thing that may prevent my Casio or this clock from running for a century would be the solar panel efficiency going down to the point when it’ll stop generating enough electricity?
Me too. Best watch ever.
Leclanche cell. And multiple power sources OR’d with diodes. Solar – meh…..
Won’t dust accumulation end up preventing it from charging ? I don’t see anything short of a radioisotope producing decent power for a hundred years unattended.
A microchip seems a poor choice for a century clock. I’m of the understanding that sublimation of the relatively reactive elements in an IC occurs over time. That in essence the chip will chemically decompose over several decades, particularly with the fine resolution of modern chips, and that the shelf-life of an IC is supposed to be about 15-20 years. Optimistically it might make it to 50yrs.
Two orders of magnitude more here: http://longnow.org/clock/
My dive watch is “Solar” powered, never needs a battery, keeps time via the naval weapons lab signal, water proof to 30 meters, altimeter, Navigation compass for above and under water, displays tide information, (For surfing) time zones around the world. Moon phases, no telling what else. Many timers, lap, elapsed, and interval, $130 USD Amazon
We’re all happy for ya, but that won’t last 100 years. Semiconductors need to be avoided in this application. Solar cells won’t produce useful energy after 100 years, so alternate sources need to be used or added later.
Everybody is thinking solar but I was given to understand that solar cells only have a lifetime of a few decades at best
But perhaps that is for X percent of charge, maybe if you over-provision a few times it could keep going.
They degrade over time much like any other silicon device… so at 20 years they are guaranteed to operate at 80% most often. But would probably exceed that a bit, not sure they would make it to 100 years though. Any silicon components in such a system you would want to run at as low a voltage and frequency as possible.
I doubt a thermo electric device would last that long either as the expansion contraction from/day night might wear them out. Some sort of solid state peizo/tribo electric device might be interesting.
Only 100 years? Lacks ambition… :-)
Clock of the Long Now
(though I haven’t seen many updates recently)
Might be a good idea to say that it is a 10,000 Year Clock. And I think that it was mentioned here before: https://hackaday.com/tag/10000-year-clock/
[DkE]: they only post updates every 100 years.
If it is for a century, my first thoughts for an energy source would be temperature difference between day and night so very long bar(s) of material shrinking and expanding. I’d probably choose glass for this because it would not oxidise like most metals.
That one has stopped occasionally though, sometimes due to lack of energy.
Maybe geothermal might be better, albeit a bit cumbersome and costly to start.
I was thinking an anniversary clock, but I guess they have to be wound every 400 days. (i.e. you wind it on your wedding anniversary)
My brother received one of those for his wedding. It never worked. Fortunately the marriage is going on 45 years….
Water wheel as a source of power. Or rain collected in reservoir and used as needed to keep the clock moving. Wind can also be used to keep a spring wound as a power source.
There’s lots of ways to do it.
But when you ask the question “how to run a clock for a century”, electronics is the absolute last idea I would attempt. Build a mechanical device, rugged enough to survive the century, and powered by nautre.
I was thinking a weight suspended in an oil well 4 Km deep…
Running a clock for longer than most will be alive to watch it.
Maybe he’ll take a trip in a near-light-speed tesla for a spin around the local cluster at relativistic speeds.
Consumer grade components are engineered to last 10000 hrs, and are affordable. The components you’re spec’ing will cost 50 times that, go for it !
I wonder if you could make a device with excessive redundancy on every single part and automatic switchover. It would be extremely complex obviously, but would it be possible? The problem is that the control mechanism checking for failure would also need redundancy and to self-check, so you’d have to make that as simple as possible I think to not get drowned in an infinite regression.
I remember hearing of a computer that runs 2 cpu’s that constantly compare results…
This project is more of a component engineering issue. Yes there are 100 yr old devices that still operate, with maintence
Maybe you’re thinking of the Space Shuttle’s computers. You can’t use just 2, though, since when they disagree how do you know which one’s wrong?
Good point, that question is above my paygrade, at T-.0001 sec, they’d probably scrub liftoff
I would recommend that we put the unit back in operation, and let it fail … we can certainly afford to be out of communication for the short time it will take to replace it.
The space shuttle used 5 computers…
I would love to see how they can design a system to compare the output of multiple computers, it would seem to be a single-point-of-failure in an already complex system.
I also love how SpaceX do it with their avionics. Standard off-the-shelf Intel multi-core processors with each core running the exact same code on each in parallel with a second physical CPU also doing the same thing, with a separate entire computer assembly also running in parallel. An awful lot of redundancy that is cheaper, more powerful, and easier to develop and test than a lot of Aerospace grade components.
That’s why the Falcon Heavy / Roadster flight that lingered in the Van Allen belts for so long before boosting to the asteroid belt, in order to prove to the USAF/NASA that the consumer grade components could actually survive the environment.
Atmos mechanical clock can run perpetually, at least in theory…
Any device relying on sun wouldn’t be working in case of a major volcanic eruption or global nuclear war.
I think that due to the uncertainty principle we won’t be able to tell if the clock is working or not during the nuclear conflagration. Also we would need to build a sentient surviving robot to even try. I wonder if that was budgeted in, but if not we’ll sneak it in the next defense budget.
Clock based off rate of decay of different isotopes.
STM32L476 is specified to retain the flash memory for 30 years. I guess it is possible that it will last for a 100 years, but some kind of periodic re-flashing algorithm might be useful.
Don’t see how it will run for hundreds of years? The parts only have a life for so many years. Dont a solar only have a life of 25 years ? And the capacitors? How long will they last?. I be shocked if it would last 50 years without problems.
I wonder how long some electronics will actually last.. I’ve got stuff I made in the 70s using things like 555s and 741s that are still working fine today, and my ti59 still works…So does the mosfet amp I built in 79… And my yamaha CD1 still plays Cds.. So that’s 40 or so years so far….
Use tunnel junction based oscillator with GaAs semiconductors: should work fine.
TD’s are known to last 50+ years in storage and with enough buffering should survive a century easily.
Also the idea of using eproms is good, if you used something like a TD based resonant oscillator with thermal feedback via metallic thermistors, comparing two different phased oscillators and cutting out one of them if it fails then comparing readings to get the “best of two” quorum sensing via a summing junction.
Make sure every component has at least two redundant pathways with a means to “cut out” the bad one if it breaks completely and reinforce the good codes into it so bad bits can be rewritten: they should be essentially random so this is also doable.
Also relevant: LEDs though containing semiconductors can be similarly overprovisioned by putting them along a resistance wire. If a few fail the thing should still light up but dimly.
Problem will be the life of solar cells and capacitors. Not gonna go 100 years I think.
Water clock with a ceramic drip-nozzle, fed by a lead-lined supertanker full of distilled water.
I think it would clog up with biofilm.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)