Hacker News Reader

> I know an engineer who spent two years of his life making sure the new water pump designs would fail at warranty + 50 percent, but only in an annoying, non catastrophic way.

Wow what an abhorrent waste of talent and life.

This can be a pretty blurry line.

If the engineer took some existing product and did nothing for 2 years other than design in additional failures that show up after the warranty has expired, I agree that's awful. However, I'm very skeptical of that. I've worked in some manufacturing orgs who try to pinch every possible penny they can, and I've never seen anything resembling that level of mustache-twirling villainy in engineering. Also, 2 years seems like an incredibly long time for someone to spend doing this and nothing else.

It's much more likely that the company asked the engineer to reduce the cost of the pump while still hitting the product's requirements (e.g. no failures during warranty period plus some margin), and while doing that, the engineer found the design was overspeced for those requirements and made some changes which reduced the pump's expected lifespan while still meeting that requirement.

Unlike what the OP is suggesting, this happens constantly...it's what engineering is.

Obviously I'm not OP and I'm reading between the lines here, but it's very easy to imagine someone hearing this story second-hand and completely misrepresenting it.

I think it’s something in between. The new pump actually cost more to make than the old one it replaced. The old pump would 90 percent make it through warranty, but sometimes failed early.

If the old pump failed, it often did so in a way that caused additional warranty expenses and reduced reliability perception. The problem was that a seal failure could cause catastrophic bearing failure.

So, the new pump was designed so that a seal failure would not cause bearing damage. Good sealed Bearings properly specced are very reliable these days, and seals on a true shaft wear very little. The problem was that the new pump would invariably make it to EOL.

So after spending a few weeks designing a no-fail pump, they had him redesign it so there was criticality in one of the bearings, so that one was effectively overloaded and would develop enough play to cause seal wear, (and dripping) while the other bearings would prevent the shaft from actually failing. The thing is, it’s hard to spec a bearing for “will wear a little out of tolerance but won’t fail” so it took two years and a lot of testing to get it right.

In all, there was a considerable amount of moustache twirling going on. But not as much as in those damnable disintegrating plastics or the unneeded battery in the thermostat.

In mechanical design, you typically target an order of magnitude of stress cycling for product design. In a pump, everything stress cycles, so it all needs to be simulated. If this was a pump you were going to make ten million of, then suddenly every fractional 10th of a cent you save becomes the difference between, e.g., having another coworker or not. So to balance that with offering a reasonable warranty such that there's actually demand for this product makes sense.

Before anyone brings it up: sure, a lot of companies just pocket this savings or whatever, but a lot don't; you just don't hear about them because they're typically midsize companies that nobody complains about.

I'm sorry, but I feel like you're reading my message extremely uncharitably. I'm obviously onboard with the idea that nothing lasts forever, and time spent figuring out how to increase something's lifespan to "at least X years" is obviously valuable, as is time spent figuring out how to cut costs without making a product break "before X years". But none of that is the same as spending time to figure out how to ensure that something breaks at the right time.

It matters whether the focus is on reducing costs while keeping track of the impact on the product's lifetime, or the focus is on reducing the product's lifetime. What I'm calling a waste of talent and human life is spending time on the latter, and that's what it sounds like my parent comment describes their engineer friend doing.

> none of that is the same as spending time to figure out how to ensure that something breaks at the right time

Yes it is. At the absolute limit, balancing cost and reliability is this exact optimization exercise; we are talking about the exact same thing taken to different extremes. The difference between your ideal scenario and the practice outlined in the OP is the difference between 97% BOM efficiency and 99.99% BOM efficiency. At extreme scales, that gap can mean the difference between a company laying people off vs. surviving.

Side note - this same discipline is what ensures your widgets fail in a safe, controlled manner and not in a catastrophic way.

> I feel like you're reading my message extremely uncharitably

I saw an naive take on how mechanical design for failure works in practice and wanted to educate. I mean no disrespect, it's just seemed like it might be an unknown unknown for you. Your comment had at least a few upvotes, so there are others out there who may also lack the same information.

To your charitability point, though, my first take on the OP above is probably the most charitable take. The least charitable scenario for the OP is that the company wants to make sure they have higher annually recurring revenue from replacing their widgets breaking down but need to _barely_ beat or match the other guys in a warranty. Reality is probably somewhere in between.

Amen, intent matters. And it sounds like in both of the cases the OP cited, the intent was to force otherwise-unnecessary replacement, which is bad.

Oh i think I saw an AvE video where they explained a cheap part being omitted resulted in this exact sort of problem.

While I don’t doubt the behavior, I have a hard time believing that a product manager would let an expensive part like a battery get added to the BOM if it’s only purpose was to fail after ten years. Which suggests it performs some sort of function, which means this could be a bug rather than a malicious act. We will likely never know but I’m always willing to assign stupidity rather than malice

> While I don’t doubt the behavior, I have a hard time believing that a product manager would let an expensive part like a battery get added to the BOM if it’s only purpose was to fail after ten years.

It depends how much of the BOM price was the battery.

When the battery is removed and the pin pulled up to the 3v bus, the thermostat works flawlessly.

I traced the circuit, it goes between ground and the uC pin, through a 1 Mohm resistor, no other connection. It’s about right for killing a 300mah battery over 10 years. The pin is held low when the unit is energized by the thermostat signal lines.

> The battery powers nothing, but the microcontroller senses it’s voltage and when it is too low, it changes the behaviour of the thermostat to randomise the temperature cut in/out points by about 10 degrees Fahrenheit , making the thermostat annoyingly unpredictable in a way that is very similar to the typical failure mode of the old thermostat it replaced.

This should be criminally investigated and the person who ordered it be put in prison for at least a decade.

Why not at least 9? Or 5?

I think it would be best to focus on the deterrent effect for the future: we need a law that makes this business strategy not viable. Not on punishing bad behavior that already happened. Maybe such law already exists, but we need more enforcement. Or a better thought out law.

I don't think it's important if _that_ person gets jail time. I would not particularly rejoice at the news. But if somehow this practice was made impossible or impractical, I'd

One of the reasons for punishment is deterrence. It it becomes clear people consistently go to prison for doing something like this that will reduce the likelihood of people doing this in the future

Or at least make it mandatory to disclose such behaviours before purchase. Failure to disclose should result in the vendor and the manufacturer becoming liable for the repair/replacement costs (with the vendor similarly able to push the costs to the manufacturer if it was not disclosed to them either), as well as any actual damages resulting from the failure of the product.

Need the jail time and a safe harbor if the behavior is fully disclosed in advance with all advertising (a simple phrase will do such as: "Useful life limited to ~10 years, details at xyz")

Some people might be fine with a product with a known lifespan, or want to pay more for the unlimited life version

The penalty should be more like corporate death than individual prison, as that often gets fobbed off on some scapegoat rather than on the actual manager responsible

As much as I agree with the sentiment and desired outcome (better/longer lasting products), detecting and enforcing that seems horrendously difficult. A great example is the VW (and others) emissions scandal. They evaded detection for years despite bringing the product for inspection. In the case of this thermostat, you would have to prove it wasn’t a bug and instead malicious intent to send someone to jail. You’d need records of who said it needs to be this way.

We can’t send arbitrary people to jail for bad designs. I don’t think many people would be an engineer if you knew there was possible jail time if you shipped something with a bug.

You can fix some of this by having competent and independent inspection, which this seems to be. The rest - perhaps you can't litigate, but you can publicise, with details, and perhaps something a consumer rights watchdog or public body would pick up.

You can already go to jail for engineering something with a bug, if it has bad enough effects and the prosecution can prove that a reasonable engineer should have fixed the bug or not written it.

This is an incredibly useful investigation! Did you publish it somewhere?

+1 on this. I don't think I've read a story about how planned obsolescence is achieved in such a novel and non-plausibly deniable way. By that, I mean they haven't just cheaped out on materials, they've spent extra to make it fail predictably.

I should have written it up. I could still, but I’d need to order one of those thermostats and do the teardown again.

At the time I was just focused on why my buddy was having to replace 200 thermostats in ten years. It turned out you could just cut out the battery with a pair of dikes and jumper the leftover positive post to 3v to get basically unlimited lifespan, so he did that instead of replacing the thermostats with new ones. Afaik most are still working fine.

Helping him out is also how I figured out the defrost timer disintegrating gears thing. When you do things at scale, things that seem random in onesies practically scream at you. It made me appreciate the value gathering data…you can find the patterns at lower scales.

The car air vent thing I figured out on my 94 Toyota 4runner. The air vent crossbars that keep the vanes in alignment with each other all failed over a two year period. A minor annoyance, but it makes you feel like you need a new car lol. I popped them out and found that of all of the plastic parts, only the crossbars were exceptionally brittle. Suspiciously, the bars also had what looked like date codes molded into them. None of the other pieces had numbers at all. I just printed new crossbars.

Oh well. These comments alone are at least a microblog entry. But if you cross something like that again, and have the time and the inclination, I think a HN submission would be super appreciated.

I noticed the promise of LED bulbs lasting forever were engineered to fail. I replaced all the bulbs in my house with and about 1-2 years in, many have failed.

Though interestingly LEDs in various components or even LED strip lighting seem to last for many many years.

Le sigh.

> The battery powers nothing, but the microcontroller senses it’s voltage and when it is too low, it changes the behaviour of the thermostat to randomise the temperature cut in/out points by about 10 degrees Fahrenheit

This is really neat, if true, but I have to say I'm skeptical. This is just too good a story!

Pick one up and see for yourself. I think they still make them. It’s the basic round dial type thermostat from Honeywell. It uses an at328p MCU and it even has a programming header. The debug was open on the ones I played with, so I downloaded the code, but I never got around to (or had the tools to) disassemble it. I was going to just write an arduino sketch to replace the factory code, but we figured out the “cut out the battery and jumper to 3v “ solution so I never did. Saved about $4000 in thermostats replacements though.

So you know of two cases of companies putting in work to defraud people by selling products that aren’t fit for purpose outside their warranty window. Name & shame.

> It has a battery soldered on the circuit board. The battery slowly discharges while the unit is on. In about 10 years of operation, the battery voltage drops below 1v or so. The battery powers nothing

I am not an electrical engineer. Could the microcontroller use the battery as some kind of calibration? Or could it have another function?