Somewhat OT: Long term design

[Sylvia Else]

A recent episode of Stargate Atlantis prompted me to think about how
would could design equipment that's intended to function far into the
future. The episode required stuff to function 48,000 years after
construction, but perhaps we could be less optimistic.

Say 1000 years.

Note, the requirement is not that the equipment function *for* 1000
years, but that when it is turned on, 1000 years from now, that it will
work.

It seems to me that semiconductors are out due to effects of difusion
and radiation.

But how about thermionic valves? They're not very reliable, but do they
age when not in use? Would they hold a vacuum over that time?

Obviously electrolytic capacitors are a no-no, but can resistors and
capacitors be made stable enough that they'd work?

Would it help to enclose the entire circuit in a vacuum tube? Again,
could the tube sustain the vacuum over such a period?

An energy source is a problem. Perhaps a cell where acid is added (how?)
at the appropriate time?

Sylvia.

 

[Joerg]

I'd expect that most semiconductors and passives would last 1000
years, given a conservative design. There's not much radiation around
at sea level. The gadget could be stored in vacuum or dry nitrogen to
prevent corrosion and wiskers and such.

It shouldn't be hard to keep a vacuum tight for 1000 years. A decent
flange-sealed vacuum vessel hardly leaks at all. If it can do 1e-12
torr for a minute, it leaks to atmosphere in (linear extrapolation) 2
billion years.

I think solar cells would stand up well. I bet that a Casio solar
calculator will work 1000 years from now if properly stored. The
biggest hazard would probably be polymerization of the plastics in the
keypad, or maybe leakage from a poorly sealed LCD.

I still use my original HP35 calculator, purchased in 1972.

HP11C over here. I still use my grandpa's drill from around the 1920's.
Works fine. You just have to keep the grease reservoirs packed by
tightening the caps once in a while and refilling when at the peg.

Oh, and the church we were married at goes back about 1200 years, the
organ in there is probably well past 500 years. I guess a pipe organ
fulfills the definition of "equipment". It can be done.

 

[Sylvia Else]

HP11C over here. I still use my grandpa's drill from around the 1920's.
Works fine. You just have to keep the grease reservoirs packed by
tightening the caps once in a while and refilling when at the peg.

Oh, and the church we were married at goes back about 1200 years, the
organ in there is probably well past 500 years. I guess a pipe organ
fulfills the definition of "equipment". It can be done.

Well, I'm not sure the organ qualifies, even if it reaches 1000 years.
Has it never been repaired?

The requirement is that the equipment be built, be left untouched for
1000 years, and then work.

Sylvia.

 

[Joerg]

Well, I'm not sure the organ qualifies, even if it reaches 1000 years.
Has it never been repaired?

The requirement is that the equipment be built, be left untouched for
1000 years, and then work.

It could do that if not played, except for the bellows which are
leather. Of course, nowadays you could make those from some newfangled
material. Also, this organ only needed the bellows when there was no
wind so technically it would still qualify as operational with cracked
leather. Unless you say the rules don't allow wind

 

[Tim Williams]

Aluminums fail by drying out, through water vapor leakage through the
rubber seals. That's a wearout mechanism.

Al Po's?

Generally considered as good as tantalum and fairly indestructible, aren't
they? Kind of new to use for millenium hardware though.

Tim

 

[Joerg]

Reliability folks generally assign component failure rates in FITs,
namely one failure per billion hours. Most passives have numbers near
1 FIT, and lots of semiconductors are in the single digits.

So a 100-part gadget that uses average 1 FIT parts will have an MTBF
of 1e7 hours, a bit over 1000 years. That assumes the parts have no
wearout mechanism. In practise, field failure rates can be quite a bit
better than specs like MIL-HBK-217 or Bellcore predict, ie better than
1 FIT average per part.

In Europe the problem with much of the Roman stuff was that there were
usually 5-10 fierce wars over a period of 1000 years. That kind of puts
a crimp into the MTBF.

 

[Spehro Pefhany]

HP11C over here. I still use my grandpa's drill from around the 1920's.
Works fine. You just have to keep the grease reservoirs packed by
tightening the caps once in a while and refilling when at the peg.

Oh, and the church we were married at goes back about 1200 years, the
organ in there is probably well past 500 years. I guess a pipe organ
fulfills the definition of "equipment". It can be done.

I have a hammer that must be 400 years old. The handle has been
replaced a dozen times, and the head maybe 6 times. ;-)


Best regards,
Spehro Pefhany

 

[Tim Williams]

They are kind of new, so it's hard to tell. Apparently one failure
mechanism is water vapor leaking *in* and wrecking the polymer.

If it ain't one thing, it's something else.

I've been using polymers lately and they seem fine. ESR is low,
leakage is low. They seem to fail suddenly at ballpark 2x rated
voltage or so with no "warning" leakage or anything like you'd see in
a tantalum or regular 'lytic. They don't detonate like tantalums.

That's interesting. Do they just 'tick' like films then? Not self-healing
yet I suppose. Mmm, that would be interesting, self-healing electrolytics.

I'd like to try a few, but when I see the price, I just think... meh, I'll
beef up my cans with some ceramics and call it a day. And e.g. Mouser's
available product line isn't nearly as diverse yet either.

Gee, this sounds like something Joerg would say. I'm betting he's never
even touched one. ;-)

Tim

 

[Sylvia Else]

Reliability folks generally assign component failure rates in FITs,
namely one failure per billion hours. Most passives have numbers near
1 FIT, and lots of semiconductors are in the single digits.

So a 100-part gadget that uses average 1 FIT parts will have an MTBF
of 1e7 hours, a bit over 1000 years. That assumes the parts have no
wearout mechanism. In practise, field failure rates can be quite a bit
better than specs like MIL-HBK-217 or Bellcore predict, ie better than
1 FIT average per part.

John

However, there is a caveat that these rates apply during the design
life. That is, a part with a failure rate of one per billion hours
cannot be expected to function, on average, for one billion hours before
it fails, unless the design life is itself a billion hours. Which it
probably isn't.

In the context of this thread, the issue is further complicated by the
question of whether a component is using up its design life while doing
nothing. It may, or may not, depending.

Sylvia.

 

[Sylvia Else]

OR....

go all mechanical

http://www.longnow.org/

(How long does a weight suspended in the air keeps its potential
energy? Makes a good battery, no?)

Yes....

But now you have to design a mechanism to extract the energy that will
work after 1000 years.

Sylvia.

 

[Raveninghorde]

I'd expect that most semiconductors and passives would last 1000
years, given a conservative design. There's not much radiation around
at sea level. The gadget could be stored in vacuum or dry nitrogen to
prevent corrosion and wiskers and such.

It shouldn't be hard to keep a vacuum tight for 1000 years. A decent
flange-sealed vacuum vessel hardly leaks at all. If it can do 1e-12
torr for a minute, it leaks to atmosphere in (linear extrapolation) 2
billion years.

It's not that easy. My only real job was for STC, then part of ITT on
undersea telephone systems. The amps had to sit at the bottom of the
sea for 25 years without failure.

One of the big concerns was outgassing of corrosive gases from
components. All resistors were plain, no coating. By 1977 they were
testing PCBs for long term reliablity but stuff was still built on
tags rivited to perspex. They were also working on fibre optics for
new projects.

 

[Martin Brown]

A billion hours is a long time, 114K years, but 1000 years is a mere 9
million hours. Unless there's a long-term wearout mechanism
(diffusion, corrosion, radiation damage) I'd guess that most parts are
still in the flat part of their bathtub curve at 1000 years. If one
were designing a 1000 year product, you'd certainly want to look for
potential wearouts.

You also have to be fairly cunning to weed out infant mortality in any
components that might be vulnerable early on. It would be really
annoying to wait 1000 years to find that bad handling had wrecked or
weakened a component that failed at switch on.

I am fairly sceptical of anything with wet chemistry inside surviving on
these timescales. Electrolytic capacitors have a bad habit of drying out
or otherwise developing defective internal behaviour when left alone for
a long time. Or for that matter anything using tin, zinc, cadmium or
other metals inclined to form whiskers.

http://www.aciusa.org/leadfree/leadfree_verdi-11-5-04.htm

Aluminium chassis and heatsinks might not be all that good longer term
either. The metal is just too reactive for its own good even if the
oxide coat is inert. Copper coated with gold ought to be OK though.

That's an issue in calculating equipment MTBF. The general rule is
that if you don't hit your reliability target doing straightforward
calculations, then you toss in a use factor.

There are old clocks and watches and scientific instruments around
that work after hundreds of years, without benefit of exotic storage.
I'd guess that WWII-vintage electronics, stored in military wax-sealed
cardboard boxes, usually still works.

Depends how many bites the rodents have taken out of it. Inductors and
galvanometers survive remarkably well - the mirrors may tarnish though
and that is over mere couple of hundred years. Clocks that are not well
looked after seem to have a half life of about fifty years.

I reckon Zambonni piles might be OK as a maintainence free power source
(for tiny currents). The oldest one at the Clarendon lab is still going
strong after 170 years running the Oxford Electric bell.

http://en.wikipedia.org/wiki/Oxford_Electric_Bell

I have one somewhere ex WWII image intensifier.

And I am pretty sure there is (or was) something similar at Leiden with
a ball of sulphur that is gradually wearing away as it rings. But a
quick search failed to find it.

What would fail in a conservatively-designed electronic gadget after
1000 years? Barring corrosion, I can't see a wearout or diffusion
mechanism for thickfilm resistors or ceramic caps. Given the
observable stability of bipolar transistors and ICs, there doesn't
seem to be much carrier diffusion or radiation damage going on at room
temperature. I'd avoid CMOS type parts where a little charge
redistribution could cause problems.

You could cheat and store the gear in Antartica. Most degradation
mechanisms follow the Arrhenius relationship.

A salt mine is a pretty safe location provided you hermetically seal the
kit in dry nitrogen before it is taken down. Steady temperatures also
help longevity just as thermal shock accelerate decrepitude. One of the
advances neutrino detector experiments is down our local potash mine.

Regards,
Martin Brown

 

[baron]

George Herold Inscribed thus:

What does the machine have to do? Mechanical stuff (gears, cams,
punch cards) lasts a long time. It could be powered by gravity.

George H.

The Ancients created mechanical traps in the pyramids. AFAIA they still
worked after how many years 4500 or so.

 

[Sylvia Else]

Centuries-old weight-powered clocks still work. Surely we can do
better with modern materials.

I don't think 1000 years is a long time for good materials.

OK, you're just convincing me I should have said 10,000 years

Or indeed, the 48,000 years in the TV program.

Though in Stargate Atlantis, they do have the advantage of using
"naquita" (sp?) for their power source, which seems to be an element
oddly overlooked in the periodic table.

Sylvia.

 

[Rich Webb]

A recent episode of Stargate Atlantis prompted me to think about how
would could design equipment that's intended to function far into the
future. The episode required stuff to function 48,000 years after
construction, but perhaps we could be less optimistic.

Say 1000 years.

Say 10,000 years? There's a group working on such a project now.
http://www.longnow.org/clock/

 

[Sylvia Else]

George Herold Inscribed thus:
Well, if way take the TV program as an indication, and including my own
interpretation of what happened (which wasn't that clear).

It appears we need a radio receiver on some frequency that will be
turned on after some number of thousands of years, and then run for
perhaps one hundred years. If it detects a radio signal, it has to start
up a computer, which is connected to a transmitter on that same
frequency. The computer doesn't have to run continuously, but has to be
able to run intermittently for a further thousand years. The transmitter
is not required to be able to run for more than a few hours once turned on.

There was also a signficant power supply that could be turned on after
48,000 years, and then actually run for a further 1000. I had thought
this was a big ask, but perhaps not. U235 has a half life of 700 million
years, but is clearly usable to produce power in nuclear reactors.

I suspect that building a computer that would still be functional after
even the first 1000 years would be a challenge, even if it were not
running during that time.

Sylvia.

 

[Sylvia Else]

Say 10,000 years? There's a group working on such a project now.
http://www.longnow.org/clock/

Not quite the same problem. They're looking at something that will run
for 10,000 years, but with some maintenance, and a constant supply of
power (from humans).

I'm thinking of something that could be built now, and be secreted away
from human interference, only to perform my biding a thousand years, or
perhaps 10 thousand years hence.

Sylvia.

 

[Tim Williams]

You just put them on the outside of the Space Ship. That way you would
never lose the vacuum.

Not quite -- interplanetary space is around 10^-6 torr IIRC, which is good
enough for crappy triodes, but you won't get a 6L6 working quite well enough
out there. There's also a mixture of gas and particles which would be
better removed with a getter than the electrodes. A solar powered diffusion
pump (wait, do diffusion pumps work without gravity?) would be a low
maintenance solution.

Tim

 

[Adrian Jansen]

A recent episode of Stargate Atlantis prompted me to think about how
would could design equipment that's intended to function far into the
future. The episode required stuff to function 48,000 years after
construction, but perhaps we could be less optimistic.

Say 1000 years.

Note, the requirement is not that the equipment function *for* 1000
years, but that when it is turned on, 1000 years from now, that it will
work.

It seems to me that semiconductors are out due to effects of difusion
and radiation.

But how about thermionic valves? They're not very reliable, but do they
age when not in use? Would they hold a vacuum over that time?

Obviously electrolytic capacitors are a no-no, but can resistors and
capacitors be made stable enough that they'd work?

Would it help to enclose the entire circuit in a vacuum tube? Again,
could the tube sustain the vacuum over such a period?

An energy source is a problem. Perhaps a cell where acid is added (how?)
at the appropriate time?

Sylvia.

Maybe you could ask how far in the future you have to go before your
'device' becomes unrecognisable ?
If the original function is still needed, just publishing the ideas and
design is probably the best way of preserving it. After that time,
either it is still needed, and could be built, or its totally
irrelevant, and not needed.

 

[Sylvia Else]

Solar.

The sun is unliklely to disappear over these sorts of timescales.

But will the solar cell even still be exposed to it? We usually have to
dig down into the ground to find prehistoric sites, even from a few
thousand years ago.

And if the cell manages to remain exposed - how many severe hailstorms
do you get an a period of several thousand years? Even if you choose a
place that's currently unlikely to get hailstorms, will that remain true
for the next thousands of years?

Sylvia.