What is the lifetime of common parts/circuits these days ?

[Phil1977]

Are they doing anything about the ultrasonic sensors, radars, cameras, and so on, which have been driving up insurance repair costs?

I do not know details, I just know from the glasses that they have statistics from the insurance companies and try to keep costs down. Though frankly spoken it´s for sure not a very high priority at many parts.

[tszaboo]

Are they doing anything about the ultrasonic sensors, radars, cameras, and so on, which have been driving up insurance repair costs?

I don't think so. Tesla driving into a concrete post at 3kmh, 5000 EUR repair.
Mercedes matrix headlights, apparently 7000 USD.
https://www.audiworld.com/forums/q5-sq5-mki-8r-discussion-129/2018-matrix-headlights-%247-000-us%24-per-pair-2928438/
BMW tire pressure sensor for one wheel cost more than a set of Continental tires for my car.
I have no idea how anyone can offer insurances for these cars. Oh, wait, I know. Insurance companies just get everyone else to pay, driving up my insurance.

[Phil1977]

Oh, wait, I know. Insurance companies just get everyone else to pay, driving up my insurance.

That´s quite exactly half true. Most new cars and especially leased cars have full insurance and not only liability.

But yes, if I could change the law I´d propose to limit liability to standard rates. But I can't change the law and that´s now absolutely offtopic

[radiolistener]

Look, I'm sorry, but I just think that entire text is completely wrong and nonsense. 

Marking entire answer of your opponent as "completely wrong and nonsense" is a classic manipulation of fraudulent peoples. The purpose of such manipulation is attempt to "cancel" everything that you don't like without giving any reasonable arguments.

It seems like you can't or won't provide any authoritative references to support your claims or you'd have done that by now, so until you can do so it seems pointless to engage with you any further on these particular claims.

You asked how planned electronics lifetime can be done. And I answered you for that question.

If you want to organize lawyer bullying for those who said things that are unpleasant for you and your business, do it with someone else.

And now I understand your attempt to impose an opinion that planned short lifetime of electronics is impossible which is very stupid opinion from my point of view. Now I see why you well understand reliability testing processes, but at the same time "don't understand" pretty simple things. It appears that this is your business to get money on that. So I think this is completely pointless to continue discussion with you, because I'm sure you already well understand what I said and now I'm sure you know very well that things which I said is true.

I just want future readers of this thread to know that your assertions are in dispute by me. If you do provide the appropriate references, and I'm wrong, I will gladly say so.

Don't worry, this is technical forum, so readers here are not stupid as you imagine and have already seen and well understand the real nature of your business.

Those who come to such forums do not need orders on what to do, so just keep your orders to your subordinates.

Your mistake is that you came on a technical forum, maybe on housewives' forum your attempt could be successful.


And by the way attempt to use "authority reference" instead of logic is another manipulation which is often used by fraudulent peoples. That's pretty obvious when someone use logic explanation why 2*2=4, so it's very strange to ask for "authoritative references to support claims". It just show that you're not interested in logic explanation, but instead of that you're looking for a way to cancel claim that 2*2=4 despite the fact that you very well understand it's logically true claim.

Those who really looking for answer will not say "all what you wrote is completely wrong and nonsense" in response to answer. They can ask for authority reference if they don't have enough knowledge in the area and want to use "faith" instead of knowledge to understand the logic explanation. But this is obvious not about you, because you already shown that you know reliability testing process very well.

It just shows that internally you well understand and agree with all what I said, but this thing is not good for your business, so you decided to start use low manipulation techniques to "cancel" and hide it... 


I think Giordano Bruno can tell you many interesting things about "authoritative references" as a "science method" to approve something...

[SteveThackery]

My reply will be brief. You have not come even close to explaining exactly how you can design an electronic device to fail after one year, to an accuracy of weeks or even days (your claim!).  You've just made some vague assertions about "missing out certain manufacturing stages".

If you want to be taken seriously you need to come up with a far more detailed description of how you would go about delivering on such an extraordinary claim.  Don't just say it can be done - show us the evidence.

Until you do, I will continue to say that your claims are unfounded. The basis of my saying that is that I cannot find any evidence anywhere to support your claim. Of course you can shorten the life of any product just by under-specifying the components, as I've already said more than once. But to ensure it fails after exactly a year (within days of that) - well, the entire internet cannot offer any suggestions on how that might be achieved.

The burden of proof lies with the person making the claim.  Over to you....

[radiolistener]

My reply will be brief. You have not come even close to explaining exactly how you can design an electronic device to fail after one year, to an accuracy of weeks or even days (your claim!).  You've just made some vague assertions about "missing out certain manufacturing stages".

If my answer was not clear. I can repeat.
In order to design electronic device which will fail after one year, you need to use components which will fail after 1 year.
Is it clear enough? 

If you're interesting on how its possible to predict failure accuracy of weeks or even days, that's is easy. The components are manufactured in a large quantities and all series has very close quality and manufactured from the same materials and using the same equipment and the same process. Usually it was manufactured by robots with minimal human intervention. So their lifetime and probability to failure is within pretty small confidence interval. And if you know which component will fail after 1 year in average, you can use it to design device with predictable lifetime...

Of course some device may work longer and some may fail earlier, that is allowed variation. As you understand, in total their average lifetime will be 1 year.

If your plans for some device lifetime appears incorrect, you can fix that by replacing reliable components with less reliable or not reliable components with more reliable components. In that way you can extend or shorten electronic device lifetime as you wish. And use that case to investigate root of cause of a wrong estimation and implement measures to avoid it future. That's how quality control management works.

The same if your product has too large lifetime variation you can investigate root of cause and apply process fix in order to get more predictable value.

If you want someone share such plans for public, I'm afraid it is almost impossible. Because such information is usually confidential and covered with NDA, so it can't appears in public documentation.

[SteveThackery]

And is it still your claim that "most" electronics are engineered in this way?  To fail after a year or less? That it is routine practice for the designers to select components with a one year lifespan, test them to confirm that, substitute others if they don't fail quickly enough, and then push the result out to manufacturing?

By the way, I know of no components with a lifespan of 1 year, unless you include old-fashioned things like filament bulbs.  What type of components do you have in mind?

[radiolistener]

And is it still your claim that "most" electronics are engineered in this way?

Yes, I see that the most of modern electronics has very short lifetime. That's true.
So, it seems that shorten electronic product lifetime is becoming a common practice.

I didn't said that all electronic manufacturers using it. But most of them.

[CatalinaWOW]

I will jump in the middle of this argument.

It is trivially easy to design something to fail after a fixed period of time.  In today's world of ubiquitous microcontrollers it is merely adding a counter.  But older technologies like the plating based hour meters would work.  But anything so engineered is incredibly dangerous to a business.  Just look at what is happening to Boeing where a hint of poor quality and reliability may kill the company. 

Designing one or more components to fail in a year or to, but work through manufacture and warranty, and also to fail without an obvious preplanned failure mechanism is hard.  I won't go so far as to say it is impossible, but I can't imagine how to do it, especially if cost is an object.  Making something last anywhere from x to infinity requires far less process control than making it last x plus or minus delta.

[radiolistener]

By the way, I know of no components with a lifespan of 1 year, unless you include old-fashioned things like filament bulbs.  What type of components do you have in mind?

The estimated product lifespan is not equal to the lifespan of a single component. It will be shorter because the product consists of many components, and when they work together, the probability of failure increases. So if you're planning a 1-year warranty period, you need to use components with a longer lifespan. How much longer depends on product complexity, component count, environmental conditions, etc.

At my times, such estimation was a typical task for a student of digital circuit design course. Aren't such things studied today?

[HarryDoPECC]

Late to the party, and maybe bringing it back nearer where it started.

I'm interested in electronic calculators between 1960 and 1975 and have restored many to working condition. Transistors, early generation ICs, linear power supplies.  See dopecc.net if you are interested.

These machines are generally over fifty years old, were expensive in their day and could use best quality parts, but some of these were of course bleeding edge tech. To the original question - I am generally impressed at the longevity of these machines over a 50 year period.

Some general observations:
 - transistors age well, failures seem random and uncommon, I have not found a plague of tin whiskers (yet)
 - early DTL and TTL ICs have aged well. I typically find plastic packages and these are surely not hermetic.  eg plastic Fairchild 99xx DTL chips from 1960s aged badly in a machine stored many years in a barn, but interestingly the plastic TI parts in same machine did well.
 - old electrolytics are most often OK, but I recognise some types of Sprague from 1960s that are often dried out. But my experience is that wholesale recapping is not indicated
 - likewise old tantalums, mostly OK but a small proportion do go short. I fix if/as they fail and leave the rest
 - mechanical connections are one of the largest failure groups, mostly oxidized/corroded connectors. Wire wrap seems to live up to its reputation for longevity (if done to standard in first place)
 - and remember that these are mostly digital circuits, so tend to be tolerant of maybe up to 30% drift in value of composition resistors and ceramic caps. If I were playing with vintage radio, these effects may be more significant

So my experience is that over a 50-year time frame, if in reasonable conditions of temp and humidity, the semiconductors and components generally seem to age very well.

Also maybe of interest are the early computers I have.  These may have UVEPROMS and my experience in machines from the 80s is that the data survives after 40 years, far longer than the original specifications suggested.

Small and specific samples I know, but maybe interesting observations from actual vintage machines.

[SteveThackery]

It is trivially easy to design something to fail after a fixed period of time.  In today's world of ubiquitous microcontrollers it is merely adding a counter. 

I've already addressed that further up the thread. Of course it is trivial to build a software bomb into anything with a microcontroller. But this way outside the scope of the OP's question, which was about how long electronic components and systems can be expected to last these days.  All of my posts have been in that context. 

Designing one or more components to fail in a year or to, but work through manufacture and warranty, and also to fail without an obvious preplanned failure mechanism is hard.  I won't go so far as to say it is impossible, but I can't imagine how to do it, especially if cost is an object.  Making something last anywhere from x to infinity requires far less process control than making it last x plus or minus delta.

Indeed, we are as one on that.  I believe radiolistener is entirely mistaken in his assertions.  In fact, I'm as certain of it as I can be.  I've tried hard to get a full explanation of how it would work, but without much luck.  His most recent post artfully avoids the question by saying "You make a system last one year by using components that last one year."  Well, duh, you don't say.  So how do you make a resistor or a microchip fail after one year?  And where on earth can you buy a component like that? Where can I buy a capacitor with a one year lifespan? 

Can you imagine how utterly crazy Microchip or Intel would be if they had a part in their catalogue that said "Here is a microcontroller that will fail after a year"?  And as you say, radiolistener still hasn't offered any credible description of such a reliable, and rapid, ageing mechanism. 

[radiolistener]

Designing one or more components to fail in a year or to, but work through manufacture and warranty, and also to fail without an obvious preplanned failure mechanism is hard.  I won't go so far as to say it is impossible, but I can't imagine how to do it, especially if cost is an object.  Making something last anywhere from x to infinity requires far less process control than making it last x plus or minus delta.

As I said before, its not so hard. For example, just manufacture a product batch and measure its lifespan. If it's appears longest that you plan, then investigate the root of cause. Replace very reliable components with more cheap and less reliable, replace materials to less reliable. Manufacture a new product batch and measure again. If it's lifespan become too short, then find the root of cause and replace not enough reliable components with a new more expensive and more reliable. Manufacture a new product batch and measure again. etc.

It become hard if you don't have stable component quality. But that is also fixable. Just investigate why your component lifespan confidence interval is too wide and fix it. Manufacture it again and measure results. Investigate root of cause. Fix. Manufacture again. Investigate. Fix. Manufacture again..... until you get desired result 

[SteveThackery]

The estimated product lifespan is not equal to the lifespan of a single component. It will be shorter because the product consists of many components, and when they work together, the probability of failure increases.

Yeah, obviously.  Tell us something we don't know.

So if you're planning a 1-year warranty period, you need to use components with a longer lifespan. How much longer depends on product complexity, component count, environmental conditions, etc.

At my times, such estimation was a typical task for a student of digital circuit design course. Aren't such things studied today?

I see what's going on.  I think you are confusing lifespan and MTBF.  Manufacturers of components will, often reluctantly, quote an MTBF for their components because Reliability Engineers keep asking for them.  Said engineers then convert MTBFs into failure rates, add them up, and convert the system failure rate into a system MTBF.

You need a seriously enormous number of components to bring a system MTBF anywhere close to one year, though.

And none of this has anything to do with ageing mechanisms and life expectancy of components. In fact, I don't know of any modern components that specify a service lifetime apart from LEDs, and that's because they have a well-understood ageing mechanism.  I expect there will be others, but I can't think of them.

There's a good reason for this: resistors, capacitors, inductors, semiconductors.... the ageing mechanisms have almost been eliminated and what's left are extremely slow.  I don't believe any manufacturer specifies a lifespan for such components, because it is so difficult to identify and quantify them.

radiolistener's argument rests entirely on the availability of these very short-lived components.  I have yet to see any evidence that they exist.  Unless he's getting mixed up with MTBFs, which now seems probable.  Even then, he would need components with an inconceivably short MTBF for his scheme to work.

[radiolistener]

Can you imagine how utterly crazy Microchip or Intel would be if they had a part in their catalogue that said "Here is a microcontroller that will fail after a year"?  And as you say, radiolistener still hasn't offered any credible description of such a reliable, and rapid, ageing mechanism.

Again, if you use MCU which can work for example 5 years. This doesn't means that if you use it in your product, then your product will work for at least 5 years. No, your product lifespan will be less, because it also uses other components. And if your mechanical button fails after 1 month, then your 5 years warranty MCU cannot help to extend your product lifespan to 5 years as your MCU... 

And even if your product components all have more than 5 years lifespan it doesn't means that your product will not die within 1 month due to exceed flash memory resource or due to work at overheating conditions. Even if your MCU lifespan is 10 years.

[radiolistener]

You need a seriously enormous number of components to bring a system MTBF anywhere close to one year, though.

Here is example. 5 month ago I bought electric stove. It has warranty period 12 month. After 4 month its power regulator suddenly broken, and stove don't works anymore. The heater still works, the mains socket works, but the stove lifespan is expired and it cannot be used anymore... Just 5 month lifespan. For a dumb product with no complex electronics. It is even hard to expect what may happens within 1 year warranty period...

[CatalinaWOW]

Designing one or more components to fail in a year or to, but work through manufacture and warranty, and also to fail without an obvious preplanned failure mechanism is hard.  I won't go so far as to say it is impossible, but I can't imagine how to do it, especially if cost is an object.  Making something last anywhere from x to infinity requires far less process control than making it last x plus or minus delta.

As I said before, its not so hard. For example, just manufacture a product batch and measure its lifespan. If it's appears longest that you plan, then investigate the root of cause. Replace very reliable components with more cheap and less reliable, replace materials to less reliable. Manufacture a new product batch and measure again. If it's lifespan become too short, then find the root of cause and replace not enough reliable components with a new more expensive and more reliable. Manufacture a new product batch and measure again. etc.

It become hard if you don't have stable component quality. But that is also fixable. Just investigate why your component lifespan confidence interval is too wide and fix it. Manufacture it again and measure results. Investigate root of cause. Fix. Manufacture again. Investigate. Fix. Manufacture again..... until you get desired result 

In principal that process would work.  But it requires time and money.  And narrowing the confidence interval on failure is the same story.  Time and money. 

The other problem with the approach is that it will take a few loops through the cycle to get the desired result, hence a few years.  Which is long with respect to consumer electronics product life cycle.

A far more productive way for manufacturers to get product turnover is feature change.  In TVs this means things like 3D, hi res, 4K, 8K

[radiolistener]

I have yet to see any evidence that they exist.

Did you hear about the case when Apple deliberately slowed down their old smartphones?

If you want to see any evidence, then you just need to open your eyes and not ignore them... 

[radiolistener]

In principal that process would work.  But it requires time and money.  And narrowing the confidence interval on failure is the same story.  Time and money. 

Yes, if you starting from zero and don't use other approach for optimization it may take a lot of money and time. But if you already in this business and already has all required statistics and tested different way to control the process, the resource which you will spend on this will be almost zero in comparison with profit which you can earn from that. 

[SteveThackery]

As I said before, its not so hard. For example, just manufacture a product batch and measure its lifespan. If it's appears longest that you plan, then investigate the root of cause. Replace very reliable components with more cheap and less reliable, replace materials to less reliable. Manufacture a new product batch and measure again. If it's lifespan become too short, then find the root of cause and replace not enough reliable components with a new more expensive and more reliable. Manufacture a new product batch and measure again. etc.

It become hard if you don't have stable component quality. But that is also fixable. Just investigate why your component lifespan confidence interval is too wide and fix it. Manufacture it again and measure results. Investigate root of cause. Fix. Manufacture again. Investigate. Fix. Manufacture again..... until you get desired result 

Exactly as I thought: radiolistener is conflating reliability and lifespan, almost in every sentence.  That explains his misconception. 

That aside, he describes a theoretical process that requires an enormous amount of time and effort, involving several iterations of the design and manufacturing process, simply to make sure something breaks after one year. It is too theoretical - I don't believe it is done, or would even work.  Does he not realise that each iteration would have to last a year? That's how you measure something's lifespan - you manufacture hundreds or thousands of the component, put them under test, identify their mean failure rate, and then keep the test running until the failure rate starts to rise - that's how you identify the service lifespan of a component.  Oh, I know, he'll now start talking about accelerated ageing tests (something he hadn't thought to mention until now).

Even if we forget about ageing altogether and just talk about intrinsic failure rates (so reliability, not longevity), they are so low that thousands of components would need to be tested in order to find ones with a high enough failure rate to make our widget fail reliably after a year.  You need a LOT of components to bring a system MTBF low enough to make it fail after a year.  The motherboard in my PC - a highly complex device - is working just fine after 12 years. In fact I've never had one fail.  In fact I can't remember the last time I experienced the failure of anything electronic (putting aside batteries).

I think he is talking from a purely theoretical standpoint with no experience of this in the real world with real components. He also seems unclear about the relationship between reliability, longevity and ageing mechanisms.  Finally, he actually asserted earlier that "most" electronics are designed to fail after a year, a claim that anyone with dozens of electronic devices in their home will know to be profoundly wrong.

[SteveThackery]

I have yet to see any evidence that they exist.

Did you hear about the case when Apple deliberately slowed down their old smartphones?

What on earth has that got to do with it?  They were trying to mitigate the effects of failing batteries, a component we have already discussed.

[SteveThackery]

Again, if you use MCU which can work for example 5 years. This doesn't means that if you use it in your product, then your product will work for at least 5 years. No, your product lifespan will be less, because it also uses other components. And if your mechanical button fails after 1 month, then your 5 years warranty MCU cannot help to extend your product lifespan to 5 years as your MCU... 

And even if your product components all have more than 5 years lifespan it doesn't means that your product will not die within 1 month due to exceed flash memory resource or due to work at overheating conditions. Even if your MCU lifespan is 10 years.

You keep repeating what we already know about combining failure rates to achieve an overall failure rate.  It doesn't need endless repeats.

[radiolistener]

They were trying to mitigate the effects of failing batteries, a component we have already discussed.

I knew it, you have a reason for everything... 

Of course, you can come up with some reason and just close your eyes on obvious reason. But it just shows that you don't want to know the root of cause why it happens. Or you know it well and trying to hide for some reason.

[Bud]

As I said before, its not so hard. For example, just manufacture a product batch and measure its lifespan. If it's appears longest that you plan, then investigate the root of cause. Replace very reliable components with more cheap and less reliable, replace materials to less reliable. Manufacture a new product batch and measure again. If it's lifespan become too short, then find the root of cause and replace not enough reliable components with a new more expensive and more reliable. Manufacture a new product batch and measure again. etc.

It become hard if you don't have stable component quality. But that is also fixable. Just investigate why your component lifespan confidence interval is too wide and fix it. Manufacture it again and measure results. Investigate root of cause. Fix. Manufacture again. Investigate. Fix. Manufacture again..... until you get desired result 

Why would anyone go thru that many troubles instead of just using the cheapest components possible. This is the same ridiculous claim as that Rigol or Siglent intentionally make their scopes hackable. I said to those idiots and I am going to repeat this here: doing such thing will make it a Feature of your product, it will make it an engineering Requirement, just like other Requirements, meaning you have to assign people and funding to develop it, test it, document it, integrate it with other features, if anything changes in the product you will have to perform regression testing to make sure that feature still performs (or rather Not performs in this case), and support it. Hope you see how ridiculous these type of clames are.

[SteveThackery]

Yes, if you starting from zero and don't use other approach for optimization it may take a lot of money and time. But if you already in this business and already has all required statistics and tested different way to control the process, the resource which you will spend on this will be almost zero in comparison with profit which you can earn from that. 

No, because you won't find components with sufficiently low MTBFs.  Of course you can increase the failure rates by deliberately overloading the components - is that what you are claiming?

Point us to some electronic components with an MTBF of 5 years or less.  Most manufacturers will publish MTBF data.  Remember, we are talking about MTBFs, not life expectancies of consumables such as batteries.