How would we design electronics to last 100 years?

[Electronpusher0]

A post by Radio Wrangler recently reminded of something I have been meaning to ask for a long time.
"The electrolytic capacitors have finite lives as they lose water content from their electrolytes. So a fresh set will re-start the ageing clock"

Any piece of vintage electronics we aquire (assuming it has not already been restored) will require some work before we dare apply power.
Electrolytics will have dried up, resistors changed value, "that capacitor" and others like it will be changed on sight, not to mention transformers that may be breaking down, if it's a transistor set there is always the tin whiskers to consider, the list goes on.

The question is, knowing what we know now, how would we design a piece of electronics to last, say 100 years or why not 500?

Mods, if this is in the wrong section please move.

Peter

[stevehertz]

Greater integration keeps out the effects of the elements as does encapsulation. Does not make for cost effective repairs though - or does it?! In general, if longevity and reliability are the goals, look at what the military and space agencies are doing.

[Radio Wrangler]

For another pointer: Undersea phone cable repeaters have been around a long time and have been designed for longer lifespans than space hardware. Also date back to the valve era.

David

[Heatercathodeshort]

Most electronic equipment is scrapped due to becoming obsolete rather than breaking down. A check of the flat panel TVs at the dump shows that more than 50% work perfectly.

Design life used to be 10 years, not bad for the cost of a replacement being about £70 for a mono set and £300 for colour [1980s]

Cost is of course a leading factor. That is why military and aircraft construction costs millions. You could probably design for 50 years but who wants to watch a 2022 TV receiver in 2032 let alone 2072? John.

[kellys_eye]

Over complication doesn't help. The KISS principle must surely be kept if longevity is a requirement? I'm sure it could be possible to engineer out any requirement for electrolytics in a circuit - they manage in integrated circuits - and the non-electrolytic versions seem to survive time.

Innovation does of course result in obsolescence but how much of that innovation is a result of 'manufactured' demands on the existing principles? i.e. VHF radio is still working and very popular despite the efforts to change to DAB (I mis-spelled that as BAD initially) which seems more of a legislated demand for change than anything else - bandwidth demands etc.

[Hybrid tellies]

In theory you could keep a vintage analogue radio and older CRT televisions going for decades, as we already do, providing you can source spare parts and are able to feed an analogue Radio or TV signal to it from whatever digital platform is around in 100 years time.

[jjl]

Polymer electrolytic and multi layer ceramic (MLCC) capacitors are now in wide use in modern electronic items where traditional electrolytic caps may have previously been used. These avoid the "drying out" mechansim of traditional electrolytics but have their own failure mechanisms and foibles.
I doubt that they could be expected to last 100 years.

John

[Welsh Anorak]

Re post #4 - not necessarily obsolete - I pickedup (with permission) a 42" Samsung LED TV, 4k, full HD and smart in excellent condition from the recycling centre. Worked perfectly, and probably only replaced as they wanted a bigger TV. Many people think their TV or whatever is going to be repaired and given to someone less fortunate. Others don't think at all...

So there's no point in designing consumer goods to last - why would you? Looking at premium makes of white goods that cost several times the own-brand equivalent, if they go wrong the customer is understandably furious and takes to every available platform to tell the world of his misfortune. If you've bought el-cheapo you shrug and buy another.

Seeing a forum member making much improved copies of TV circuit boards got me wondering if his boards were fitted with top-quality components and made into a replica TV how much longer it would last than the original.

At Uni, we had a module telling us about MBTF (mean time between failure) of components. Like hard drives, it's not if, but when they fail. After the lectures, few of us ever wanted to get into a plane again!

[G6Tanuki]

While things *could* be made to last 100 years I am quite happy for them not to.

100 years ago a radio would likely have been a crystal set only able to give poor quality reception on headphones... Record players were wind-up things and definitely not hi-fi. There was no TV or Internet, and the telephone was still very much a rich person's thing.
As to things like surgery or dentistry, they were both almost barbaric by today's standard.
Even if you had the desire to produce something that would last 100 years there are still unforseen pitfalls, tin whisker problem, brittle brass syndrome, expensive mil-spec Tantalum capacitor failure, and Lockfit transistor issues to name but a few.

Give me cheap and cheerful, design life of 10 years to be replaced by something faster and more feature rich. OK it's nice to play around with vintage electronics as a hobby, but for everyday use, when you don't want to spend time fixing a leak on an old washing machine, you just want things to work... Give me the latest technology!

[kalee20]

Packaging definitely makes a big difference. Some of the glass-encapsulated, vacuum-packed items of hardware are approaching 100 years old now, and they're still working to spec.

It of course highlights 'to last 100 years' rather than 'to work for 100 years,' the latter is a big ask!

Understanding degredation and wear-out mechanisms is key. Electrolytic capacitors have been mentioned - drying-out can be fixed by hermetically sealing them and having a generous dose of electrolyte. But there are also corrosion reactions happening, so the aluminium would need to be ultra-pure. Alternatively, solid tantalum capacitors could be used (expensive, but possible).

Other capacitors - I take comfort from the environmentalists who claim that discarded crisp packets thrown around the countryside could take 50,000 years to degrade. I think, Great! That's the same polyester that's used in the capacitors I buy! So as long as the aluminium metallising is protected against moisture ingress, I'm OK!

Semiconductors... Hot-running power devices sometimes fail due to die-bonding fatigue, caused not by hours at high temperature, but by the number of thermal cycles. So paradoxically, a TO3-encapsulated transistor may well be OK if operated continuously for that time, but not if operated only an hour a day.

Resistors... Carbon composition types are rubbish. But carbon-film types, and metal-film, both on glass or ceramic substrates, are likely to last a very long time (my money would be on the carbon, as even if the coating fails, it just doesn't oxidise unless it gets super-hot).

[Paul_RK]

Originally Posted by Electronpusher0

Any piece of vintage electronics we aquire (assuming it has not already been restored) will require some work before we dare apply power...

Ideally, and I'm as ready as anyone to counsel new members that way to avoid their getting a nasty surprise: but some of us are lazy. I've a 1935 Bush SAC25, for instance, which cost me two shillings off the back of a van in 1969. Took it home, plugged it in - anything else wouldn't have occurred to me as a ten year old with no one to proffer contrary advice - it worked with a pleasant mellow tone from the triode-driven 10" Rola speaker, and received, I think, quite a lot of use over the next few years. After that it could be at rest for one, two, five years, then I'd plug it in and use it again. On one occasion the wavechange switch was a tad intermittent, so I took some emery paper to its contacts (not the usual sort of wavechange switch at all, the contacts make or break according to the flats on the control spindle). It's had several years' rest now, so I've promised it I'll do a couple of tests before next using it...

Originally Posted by kellys_eye

Over complication doesn't help. The KISS principle must surely be kept if longevity is a requirement?...

You'd think so, but the most elaborate radio I know of - the Panasonic RF-9000, with 41 ICs, 174 transistors plus 21 FETs, and capacitor and resistor counts that must each be nudging if not actually into four figures - is about forty years old and my example, like most of the few others I've heard of, works delightfully so far. I hope it continues to do so, because the service manual scares me.

There's no way of knowing that a vast array of components will carry on working indefinitely, so, yes, simplicity has to help. I suspect that much of the time designers are aware of, or at least have opinions about, how well specific components from specific manufacturers can be expected to last: and of course designers can, and often don't, design such that all components run comfortably within their specified voltage, current or temperature ratings. The RF-9000 must have received generous attention in that regard - if the design lifespan had been seven or ten years, if just the electrolytics had been of the sort Sony used in various of their world-band radios, the sets would practically all have failed long before now.

Know the components you're using, select with care and use them in a minimally stressful way, and longevity should be about the best that can be achieved unless and until better parts become available.

Paul

[Craig Sawyers]

When the idea for the code breaking computer Colossus (functionally rebuild at TNMOC) was first put forward by Tommy Flowers, there were many gainsayers who pointed out that radios regularly needed valves replaced. They said that there were thousands in Colossus and it would need valves to be replaced every minute.

Nonsense, said Flowers - it is all to do with turning them on and off frequently, and it is thermal shock that kills valves. Never turn the machine off, and it will be just fine.

That turned out to be the case.

Of course the rebuilt machine uses a dwindling supply of the correct WW2 valves. So it uses the biggest variac I've clapped eyes on, to slowly bring up the heaters, after which the HT is applied.

Interestingly, the Tektronix valved oscilloscopes use a similar scheme. They actually turn on the heaters all at the same time. But there is a thermal delay switch (it just looks like a valve) that waits 45 seconds or so for the heaters to stabilize before it turns on the HT. Now they have been in use, now by entusiasts, for over 60 years. And still perform like they did when new.

Interesting Tek story. With the valved scopes, they were concerned how robust the front panel lettering would be. So they made front panels using different methods. Engrave and fill, screen print etc. And then screwed them into the factory entrance, and required everyone to walk over them every day. Engrave and fill was the choice (although later machines were silk screened)

And one heard about when visiting Atlas Copco in Sweden way back in the late 80's. They made production line air tools, and the traditional tool had an aluminium housing. Copco wanted to move to a plastic housing - and marketing were horrified - the chunky aluminium housing they said was the recognized house standard.

So they tied an aluminium tool and a plastic tool to the back of a car, and drove it around the car park for an hour. Then asked marketing to use the resulting tools. The aluminium one was covered in sharp edges and was unusable. The plastic one was the worse for wear, but had no sharp bits and was easy to use. So plastic won the day.

Craig

[emeritus]

Military-grade components used to be burnt in as part of the testing regime, which took time and money. Subsequently it was found that thermal cycling, unpowered, was just as effective in weeding out defective items, but this still takes time and resources to do effectively, so eoild not be fone on consumer products.

In the early 1970's I was designing custom ICs for a military project. Circa 1974 the costs were approximately as follows:
Set of 5 masks for IC fabrication:-£500
Cost per chip from 3" wafer: 50p
14 pin military gold-ceramic flat pack for chip:- £5
Cost of electrical and environmental testing to mil. std.883 :- £50

[Maarten]

Originally Posted by kalee20

Understanding degredation and wear-out mechanisms is key. Electrolytic capacitors have been mentioned - drying-out can be fixed by hermetically sealing them and having a generous dose of electrolyte. But there are also corrosion reactions happening, so the aluminium would need to be ultra-pure. Alternatively, solid tantalum capacitors could be used (expensive, but possible).

There's also a difference between theory and practice. It's easier to make a cheap conventional electrolytic last 50 years than a cheap solid tantalum. While some say tantalums will last forever when sufficiently overrated, I'm not convinced at all. Normal electrolytics have already proven they can last forever when optimally rated.

I think the best chance to have a solid state electrolytic last 50 years or more, is to use a solid aluminium (Philips patent, sadly not manufactured anymore) or a polymer type.

[Maarten]

Originally Posted by Craig Sawyers

Nonsense, said Flowers - it is all to do with turning them on and off frequently, and it is thermal shock that kills valves. Never turn the machine off, and it will be just fine.

That turned out to be the case.

I expect that's how myths (that continue to exist to this day in the computer world) come into existence. I don't know about the numbers and the research this particular case, but in general, measures like that are more about being able to predict the wear than to lower the wear per se.

Of course a valve will wear out when it's operated continuously, but it's easy to predict that after for example 50.000 hours, 80% of the valves will still work. Derate that accordingly and find a preventative maintenance interval.

On the other hand when operating valves intermittently, combined wear and failure due to inrush current will probably not be higher on average, but the spread will be larger. Also, valves might work loose from their sockets more easily - failure without wear.

[The Philpott]

Software programmers have to take their share of the blame for scrappage of assemblies- think, as an individual, how many times in your life you have thought 'Why did they make it do that?

If the software and firmware are well thought out then the next future-proofing aspect that comes to mind for the hardware is not over-miniaturising. My physics teacher expounded on the potential problems when we had already got down to 'tiny'...that was 33 years ago and things have got tinier. He simply would not have believed the way things have progressed since then.

Dave

[Craig Sawyers]

Originally Posted by Maarten

I expect that's how myths (that continue to exist to this day in the computer world) come into existence

It is not a myth. I don't trade in myths. For the record Colossus was designed to break the code from the German SZ42, an on-line cypher machine that transmitted at 80 baud.

I restored one of the only three known remaining SZ42 machines for Bletchley into full operating capability for the first time since it was captured in 1945.

And that is true and not also a myth.

Craig

[G6Tanuki]

One could argue that in the best of all possible worlds all the parts in an assembly or device would be lifed such that they all failed at the same time.

So when the thing was cast into landfill there were essentially no usable parts left in it.

Alas, I am not Dr Pangloss and we do not live in the best of all possible worlds, so we live with external influences and perturbations to mess up our bathtub curves.


Mil-spec parts aren't always all they are cracked up to be, in times past I supported a series of encrypting Modems for the financial institutions, these used dedicated crypto chips made by a major US defense contractor initially for military purposes. Ceramic cases, Gold plated leadouts, some serious pre delivery testing... But they failed after a couple of years, both in banks and aircraft!!

[Craig Sawyers]

You of course need to de-gold the leads. The gold is only applied as a corrosion inhibitor during storage. If the gold is not removed before soldering into the circuit board it forms a brittle intermetallic with tin, compromising the joint over the short to medium term.

This is also done with space grade parts. There is usually a de-golding solder bath, with an automatic dipping mechanism. There is then a tinning bath to finalize the lead preparation. This process is even done with surface mount parts.

The composition of the degolding bath is usually monitored for gold content and discarded when it above a critical percentage (I cannot recall off-hand what that is)

After degolding and tinning, parts are stored in desiccators until used. Standard practice for mil spec and space grade parts.

Craig

[Craig Sawyers]

This is the ESA document with a section on de-golding and pre-tinning http://esmat.esa.int/ecss-q-st-70-08c.pdf , page 42

And the answer to the maximum gold content of the de-golding bath - <1%

Craig