Electronics are stupid reliable

[Fried Chicken]

And I say this having just thrown myself into electronics repair to fix stuff that's broken, built stuff, or improve stuff.

I'm having a hard time processing just how unbelievably reliable the miniaturized ICs actually are.  What's expected of them, and how hard their reliability can be pushed.
It's actually wild.

I just put together an SMD solder practice board this one.

I put it together and it worked... I mean great, but there were still some flaws in there.  Then I think about the motherboard in my computer: hundreds if not thousands of components in a massively complex thing, produced by the hundreds of thousands, failures fleetingly rare, with decades of expected hard use in the worst possible environments.

How is this actually possible?  Most other industries, automotive for example, will happily accept failure rates orders of magnitude higher.

I can't put it into words.  It literally blows the mind: it's beyond comprehension.

[electrodacus]

How is this actually possible?  Most other industries, automotive for example, will happily accept failure rates orders of magnitude higher.

I can't put it into words.  It literally blows the mind: it's beyond comprehension.

A modern automobile has way more electronics than your computer and it works in much tougher environment for decades.

[tautech]

40 yrs ago electronic ignition/fuel injection systems were notoriously unreliable.....just saying.
In that time we have come a long way .....when/if end products are properly designed.

[Haenk]

How is this actually possible?  Most other industries, automotive for example, will happily accept failure rates orders of magnitude higher.

Mechanical stress and wear is simply the worst thing that happens, but on mechanical parts this is expected and considered during construction (corrosion is even way evil). That's why you need regular machine maintenance - just to check for growing defects or by replacing them before they fail.
Electronic parts mostly come with a very low failure rate, by multiple orders of magnitude compared to mechanical parts. And then it's simple math.

[SteveThackery]

We should be aware of the difference between reliability and durability. Reliability revolves around random failure rates, durability is about aging mechanisms. In most cases, both are very low for electronic systems.

In fact only a few electronic components have known aging mechanisms: electrolytic capacitors, LEDs, FLASH memory, switches, connectors, and so on. Likewise, failure rates for almost all components is astonishingly low.

The big exception is electrical and/or thermal stress. Operating components close to, or beyond, their safe operating area increases the random failure rate and any aging mechanisms. I should also mention thermal cycling, even when it is within the specified temperature range for the component. It can loosen lead-offs from integrated circuits, or cause encapsulation failures.  Finally, mechanical vibration can cause fatigue failures in component wires.

So the secret to a long and reliable life is to ensure that all components are operated well within their parameters, don't subject them to unnecessary thermal cycling (so manage heat extraction carefully and consider leaving it switched on), and isolate the device from mechanical vibration (also consider fixing heavy components with silicone adhesive).

[ejeffrey]

It's possible to run mechanical devices in  a regime where there is negligible wear, nothing flexes outside it's true elastic limit, and composed entirely of materials that are not subject to oxidation or other environmental degredation.  It's just not very practical for most applications.  And the mechanical side of a device is most exposed to the outside world: a car suspension can't really control how fast you drive over speed bumps.

 Electronics can vulnerable to outside environment.  This can be ESD, heat, chemical attacks, dist and grime, and so on. but by the time the uncontrolled environment makes it to the electronics the mechanical part has already failed.

[IanB]

Just to provide a contrast, I have had a laptop that spontaneously failed (twice) in the same way while still under warranty. The first repair was a board replacement, and the second repair was to replace (again) the same board.

I have had two TVs that spontaneously failed.

I have had a dishwasher control board that failed, an oven control board that failed, and a furnace control board that failed.

I have had a car console display board that failed.

Many of these failures were bad solder joints.

So reliability of electronics may be a matter of how lucky you are.

[Miti]

Just to provide a contrast, I have had a laptop that spontaneously failed (twice) in the same way while still under warranty. The first repair was a board replacement, and the second repair was to replace (again) the same board.

I have had two TVs that spontaneously failed.

I have had a dishwasher control board that failed, an oven control board that failed, and a furnace control board that failed.

I have had a car console display board that failed.

Many of these failures were bad solder joints.

So reliability of electronics may be a matter of how lucky you are.

Or unlucky, luck goes with plus and minus sign.
I have all those appliances and only one washer had a board with a broken solder on a power part.

[Siwastaja]

It's the magic of "solid state".

Semiconductors specifically reached a point where they only die out of abuse, not age, within a few decades of their invention. After that point was reached, any new idea producing unreliable semiconductors would simply not gain traction.

It's just that the "wear region", the design margin between semiconductor aging more rapidly than normal, and dying instantly, is narrow. It is nearly impossible to design a thing where semiconductors die after 5 years of use.

[Fried Chicken]

Just to provide a contrast, I have had a laptop that spontaneously failed (twice) in the same way while still under warranty. The first repair was a board replacement, and the second repair was to replace (again) the same board.

I have had two TVs that spontaneously failed.

I have had a dishwasher control board that failed, an oven control board that failed, and a furnace control board that failed.

I have had a car console display board that failed.

Many of these failures were bad solder joints.

So reliability of electronics may be a matter of how lucky you are.

Think of how many transistors there are in your CPU, think about how many times they've been actuated, and then think about how few failures there have ever been... now multiply this by the other CPUs in the world.

It's so next levle, and I wish other industries would take note.

[tom66]

A transistor doesn't actuate, there are no moving parts.  It is just electrons modifying the movement of other electrons.  Very cool technology but nothing to fundamentally wear, aside from general issues like electromigration.

[Siwastaja]

Similar thought: think how many electrons go back and forth in a copper conductor of a power cable. It's gazillion trillion mega multi billions. And think about how many copper atoms there are in series! Yet, if the copper is of good quality and not exposed to environmental acids or humidity, it lasts hundreds of years moving these electrons.

Transistor is not that different, conceptually. It has a wear mechanism of electromigration, but that is easy enough to avoid/mitigate.

And further: think about how electrons move here and there from just temperature (of more than zero kelvin) alone, in any atom. Like an atom in a rock. Yet the rock will be there in 1000 years. Transistors are not much more than just clever use of the sub-atom scale movement we have all around in our universe all the time.

[schmitt trigger]

Whenever I come across a home with a beautiful garden, I know this didn’t happen by luck. It means that the owner has put some serious time, money and effort into it.
Electronics reliability is similar. If one of the project goals is reliability, it has to be designed and produced accordingly.

Therefore;
-It must be conservatively designed. Meaning not operating the devices near its maximum limits, at least not continuously. Derating is your friend. Read and comprehend the datasheet.
-Understandi the environment where it will be used, protecting it against humidity and chemicals, and having sufficient thermal dissipation for operation at the extreme temperature ranges. Passive cooling is best, as fans themselves are a failure point.
-Related to the environment, will it be subjected to vibration/shock? What can be done for protection? Are the heavy components properly secured, as to not rely on solder joints?
-What sort of electrical stresses will it be subjected to, and what protection devices should be included?
-Test the hell out of first production samples, don’t dismiss any failures as random, unless you fully understand the root cause. Be ready to implement the required corrective actions.
-Last but not least, the components, the PWBA have to be sourced from reputable suppliers. Perform due diligence

This is only from the hardware side. Software is another can of worms, which I will allow someone more knowledgeable to dissect.

[coppice]

We have got to the point where electronics properly implemented is amazingly reliable in the short term. However, we are still really really bad at being able to put things together that will last a long time. When you look at things like travelling to the stars, we have no clue how we might be able to build things that will operate with minimal support for a very long time.

[stretchyman]

No moving parts only electrons whizzing around, hence why its so reliable!

[coppice]

No moving parts only electrons whizzing around, hence why its so reliable!

Rust never sleeps

[coppercone2]

But is it? If you are really gonna build some good stuff like active controlled structures, ultra complex robotics, space craft.. then its not reliable enough.


Look at the amount of redundancy you need for a traffic light to work, and their not even that good.

Lets say a hypersonic jet that will blow up and crash if a ADC makes a mistake. Or super conductor systems (like LHC) that cost 10 years of progress and billions of dollars from a single electrical parts failure.

The things we have are still really simple and not very useful compared to what they could be.

[SiliconWizard]

I think the things we could have should be unburdened by what has been.

[hans]

Lets say a hypersonic jet that will blow up and crash if a ADC makes a mistake. Or super conductor systems (like LHC) that cost 10 years of progress and billions of dollars from a single electrical parts failure.

Yep.. or any kind of soft sauce on top. Because any electronics these days will have firmware on it to make it do anything at all, either running on a CPU or in some hardware written by code (FPGAs)

Look at software bugs in modern jets. The Boeing 787 had a bug in its electrical power system that would crash after 248 of uptime. Large jets are flown around the clock.. they land, get cleaned for 2-3hrs, and take off again. Some systems never shut down.

Electronics can be highly reliable, but also highly complex to the point these structural oversights can become quite common. Simple projects are still easy to get right, but problems appear when people have to cooperate and communicate to make things work together. In theory every system stays highly predictable if you know all the variables, but thats rarely possible anymore. Unfortunately, there are tons of appliances out there that stay up with periodic resets (even if its something under the hood), because at some point thats economically more sane instead of chasing down red herrings for months trying to fix a "may crash once in a week bug".

And with those red herrings, anything can look intermittent and 'replace everything as precaution' is probably the simplest fix. A lot of car mechanics complain these days they have to be some kind of IT nerd to service modern cars, program options and reset/trigger fault warnings to see if their work is done. So maybe its not the electronics thats unreliable (aside from glitching).. its just too complex to service them. You often hear car tech guys loving 90s cars, because they are modern enough for good fuel economy and creature comforts, but simple enough to be fixed with just mechanical tools.

[EEVblog]

We should be aware of the difference between reliability and durability. Reliability revolves around random failure rates, durability is about aging mechanisms. In most cases, both are very low for electronic systems.

In fact only a few electronic components have known aging mechanisms: electrolytic capacitors, LEDs, FLASH memory, switches, connectors, and so on. Likewise, failure rates for almost all components is astonishingly low.

The big exception is electrical and/or thermal stress. Operating components close to, or beyond, their safe operating area increases the random failure rate and any aging mechanisms. I should also mention thermal cycling, even when it is within the specified temperature range for the component. It can loosen lead-offs from integrated circuits, or cause encapsulation failures.  Finally, mechanical vibration can cause fatigue failures in component wires.

IC's can have it too with electromigration, dielectric breakdown, oxide issues, and bit rot in memory etc.

[BradC]

40 yrs ago electronic ignition/fuel injection systems were notoriously unreliable.

In parts. These days it's the cables and connectors rather than the electronics.

[Fried Chicken]

We should be aware of the difference between reliability and durability. Reliability revolves around random failure rates, durability is about aging mechanisms. In most cases, both are very low for electronic systems.

In fact only a few electronic components have known aging mechanisms: electrolytic capacitors, LEDs, FLASH memory, switches, connectors, and so on. Likewise, failure rates for almost all components is astonishingly low.

The big exception is electrical and/or thermal stress. Operating components close to, or beyond, their safe operating area increases the random failure rate and any aging mechanisms. I should also mention thermal cycling, even when it is within the specified temperature range for the component. It can loosen lead-offs from integrated circuits, or cause encapsulation failures.  Finally, mechanical vibration can cause fatigue failures in component wires.

IC's can have it too with electromigration, dielectric breakdown, oxide issues, and bit rot in memory etc.

Also lead free solder forming whiskers causing shorts

https://nepp.nasa.gov/whisker/

[coppercone2]

Imagine the kind of reliability you would need for a massive space habitat or something though. It just looks good in the simple systems we have.

That's why I always thought Alien looked real. Eventually its going to be monorail man selling you a touch screen interface air-lock. It's just not there. People would be just like WTF why is that not mechanical? And not on a hair trigger either.

[Poroit]

Ants and other critters love electronics!

[CatalinaWOW]

One reason it is so reliable is that electronics are made on a scale that few other things approach.  When you build the same thing millions of times (I believe in some cases it might be billions) you have the chance and motivation to find and fix the things that go wrong.  This applies to the components that go into our boards.  The assemblies also benefit from that kind of scale.  JLCPCB needs to drive the errors out so they can sell at the prices they do.  And have the volume to afford driving the errors out.

This process goes on in many ways.  Switches and pots are notorious for failing in older electronics.  But they were a necessary evil at the time.  Now traditional wipers on a resistive track are all but gone, and panels of switches have been multiplexed into a much smaller keyboard with menus.  So the electronics of today have far fewer of the components that caused much of the trouble.