Have you designed early life failure into a product?

[SirNick]

RIM is a special case, I think.

If I had to guess, I would pin at least part of the blame on the Blackberry Enterprise Server licensing model.  When you have the option of paying $$$ per seat for mobile email access, or connecting an iPhone / Android to your Exchange server for free, well.... *shrug*

I'm sure the long-term usability of the phone does prolong the life for the casual user (BB fans tend to be a loyal bunch), but if you've been the one responsible for provisioning new phones in a corporate environment, you'll see that CxOs are the consumer-iest of the consumers when it comes to being the first kid to have a new toy.    Theirs always seemed to start "acting up" a lot when a new model got released.

[miguelvp]

I've always been able to connect my phone to my many e-mails including work's Exchange without the need for the Enterprise Server. I can accept meetings and it's all in my calendar. Never saw what value the Enterprise Server brought, maybe it allows you to do more things, but you never had to pay just to connect it to an e-mail address.

I have 9 e-mail addresses and text messages going to the same inbox. Of course I can view them separately

[ejeffrey]

I have wondered, because I don't know what they could possibly stand to gain.  (Drivers are a free download, there's no subscription model or anything.)  But, what reason does a kernel driver have to know the date and time?  Puzzling.

There is practically no failure pattern so bizarre that it can't be the product of an unintentional bug.  The standard example is the 500 mile email.  There are plenty of reasons why a driver for a multi-channel audio interface might do something with date/time.  There are smart things, like keeping track of time stamps for audio streams, or dumb things like logging how old the firmware build is when the driver is loaded (oops, the firmware version is 1000 days old, but we only have a 3 digit field!).

[T3sl4co1l]

Time is a weird thing... the Patriot missile system had a peculiar bug where its clock would drift out of sync and lose tracking by the same distance.  Now, why they'd ever need to use disparate clocks in the damned thing to begin with, I have no idea... design by committee?  More lessons to avoid.

Tim

[bwat]

Time is a weird thing... the Patriot missile system had a peculiar bug where its clock would drift out of sync and lose tracking by the same distance.  Now, why they'd ever need to use disparate clocks in the damned thing to begin with, I have no idea... design by committee?  More lessons to avoid.

Tim

Looks like it was a problem with different representations of time in the software and they mixed them when subtracting: the subtrahend in one representation, the minuend in the other. I got this from: http://www.ual.es/~plopez/docencia/itis/patriot.htm. I have the "clock drift of 57 us/minute" reason cited in a book [But], but I think the linked paper has a more plausible reason.

[But] Giorgio C. Buttazzo. Hard Real-time Computing Systems. Kluwer, 1997

[SeanB]

Many years ago when cellular phones were both new and expensive we gave our bike messenger a Panasonic phone. Due to the nature of SA driving the bike regularly was involved in accidents, and one day of course we got the call that he had come off it and hit something. The phone was the point of contact, with the only protection being his suit pants pocket between it and the road as he slid down it to a stop. We sent it into Panasonic, and it was repaired. The only parts that they kept of the old phone were 8 small screws, the entire phone was replaced with new parts aside from these screws. IIRC this was about $20 cheaper than a new phone at the time.

[SirNick]

There are plenty of reasons why a driver for a multi-channel audio interface might do something with date/time.  There are smart things, like keeping track of time stamps for audio streams, or dumb things like logging how old the firmware build is when the driver is loaded (oops, the firmware version is 1000 days old, but we only have a 3 digit field!).

Duh, of course... SMPTE time codes.  I'm sure those make it to the kernel driver level on that kind of interface.  It had a lot of intelligence in the PCI card to control multiple modular interfaces and such.  Time sync would be essential between interfaces.  I clearly didn't think that statement through before posting.

I actually really like that screen, and although it isn't much of a step up from full HD it is noticeable.  Forget all the nonsense Apple put out about "retina" displays and the limits of human eyes. Having a super high resolution display means you can scan web pages when zoomed out, rather than zooming in and scrolling.  Your brain only needs word shapes to get a vague sense of the content. It's handy for PDFs too.

OK, that's fair.  I guess I'll withhold judgement until I try it.  Apple marketing notwithstanding, I find my phone's screen sufficiently detailed to cause eye-strain before running into issues with inadequate resolution.  Now the 5.5" of screen real-estate may be genuinely useful.  (Although I'm more interested in something small than something with an expansive screen.  For that, I have a tablet or computer.  To each his own.)

There are still useful features being added to phones. NFC is getting more and more useful, and always-available voice control handy. Indoor navigation (not using GPS) has also improved a lot in the last few years.

No argument there -- we're not quite done yet.  However, I do think we're a while away from revolutionary changes the likes of the BlackBerry, iPhone 1, etc.  Yet, people still expect something new that will blow them away every release cycle, with the inevitable disappointment of "merely" incrementally improved hardware.

[Richard Crowley]

You COULD make a notebook computer that would last 20 years. Buy why WOULD you do that?
The market for consumer computers with a 20-year expected life cycle is essentially zero.
It is quite likely that the "notebook" form-factor won't even exist (or be remembered) in 20 (or even 10) years.

[SirNick]

Sure it will.  My oldest brother had an 8088 laptop in college.  That was something like 25 years ago, and they're still prevalent today.  Until technology somehow makes keyboard input a thing of the past, it's a form-factor that can't be beat for portable computers.

[David Hess]

You COULD make a notebook computer that would last 20 years. Buy why WOULD you do that?
The market for consumer computers with a 20-year expected life cycle is essentially zero.
It is quite likely that the "notebook" form-factor won't even exist (or be remembered) in 20 (or even 10) years.

I doubt the notebook/laptop form factor will be going anywhere because it serves as a portable replacement for the desktop with a keyboard and monitor.  The market will be smaller because people who only consume data can get by with tablets but it will still be large enough to support itself.

As far as laptops, what I want is something that is ruggedized, takes standard batteries, has a matte screen, and no chiclet keyboard even if it has to be thicker and heavier.

[luky315]

I once had to use 85°C Electrolytic capacitors and not 105°C parts because "management" thought they would last longer then the warranty period and that is long enough.

[wagon]

I once had to use 85°C Electrolytic capacitors and not 105°C parts because "management" thought they would last longer then the warranty period and that is long enough.

The best way to deal with this is to trap management into demanding something too cheap and creating a lot of in-warranty failures. Once they have been burned and maybe one of them made into a scapegoat they will listen to you next time.

You and I both know that's not how it works in the real world.  The turds who demand the cheaper parts will destroy you to save their own skins..... better to argue the point and win before the product goes to market.

[David Hess]

I once had to use 85°C Electrolytic capacitors and not 105°C parts because "management" thought they would last longer then the warranty period and that is long enough.

The best way to deal with this is to trap management into demanding something too cheap and creating a lot of in-warranty failures. Once they have been burned and maybe one of them made into a scapegoat they will listen to you next time.

They will just argue that it is your fault for not being persuasive enough.

[wagon]

I once had to use 85°C Electrolytic capacitors and not 105°C parts because "management" thought they would last longer then the warranty period and that is long enough.

The best way to deal with this is to trap management into demanding something too cheap and creating a lot of in-warranty failures. Once they have been burned and maybe one of them made into a scapegoat they will listen to you next time.

They will just argue that it is your fault for not being persuasive enough.

You can't win against management ar$eholes, salesmen & accountants.

[Refrigerator]

Well, as a hobbyist who gets 90% of his electronics parts while dumpster diving i can't complain, the sooner the failure, the newer the parts i get.  After that i can can use those parts to repair my broken electronics.

[Phaedrus]

My company wants to sell a non-UL certified (or certifiable!) power supply that is so poorly designed it presents immediate danger to life and limb. My report and analysis basically amount to, "This product cannot legally be sold in the US, and even if it could, we shouldn't sell it, and if we do it anyway we will regret it when we suffer a million dollar lawsuit." I can't even use my lab right now because one of the samples burned, and I don't want to inhale cadmium fumes (it's non-RoHS as well).

Let's see if common sense prevails. My resume is being sent out this weekend regardless. If anyone's got an electronics technician job in SoCal available let me know.

[SirNick]

Curious... Was that the product of cheapskating, or merely a poor design?

[Phaedrus]

Curious... Was that the product of cheapskating, or merely a poor design?

Someone reverse engineered an ATX 1.3 PSU, probably an FSP or Antec, from the late 90s. Then they started removing parts experimentally until every part not required to turn on was removed. Then they put it in the cheapest possible housing on the cheapest PCB with the cheapest parts possible.

BOM cost is probably under $7/unit. I'd be impressed if it wasn't a lethal hazard to the operator. Seriously--there are holes on the underside of the unit near the primary section that are large enough for a screwdriver or a careless finger to fit through. And that's just the beginning.

I've tested hundreds of models of ATX PSU, including some really nasty ones I found in night markets in Taipei, and this unit is probably the worst I've ever tested. And they want to sell it here.

[wraper]

You and I both know that's not how it works in the real world.  The turds who demand the cheaper parts will destroy you to save their own skins..... better to argue the point and win before the product goes to market.

That's actually the best possible outcome. You will of course have saved every email and document you wrote pointing out how badly things were about to go wrong. You can then take them to an employment tribunal, win and get a free all expenses paid year off work and a glowing recommendation for your next position.

In fantasies... Only rare one will want to hire someone who goes against the system and collects evidence instead of doing a fucking job.

[Sigmoid]

"If that transistor fails for any reason in the next 15 years (life of system), an onsite service visit will be needed. This may involve me buying a plane ticket for $2000 to fly across country on a day's notice, grab a $200 rental car, $300 hotel bill, at least $100 in meals and incidentals, and about a $300 dollar two way round trip for FedEx charges to ship the 50 lb part out overnight freight.  So i say spend the extra $4 to save $2500 cost to the company in service costs."

R&D response was "you are not looking at the bigger picture. We have a price point to bring this system to market and we must hit that mark".

I mumbled something about who is not seeing the "bigger picture" and told the interviewer I could never work for him under those constraints - not with the "bigger picture" I have come to see from working in the field.

Later that year I was paged out to service one of those systems. A "fix it before the sun comes up or get it out of here" scenarios. I met my further wife to be that night. She was a technician at the hospital and saw me working on the system at 2 in the morning and let out a string of curses I can't repeat here.

We have been married going on 35 years now. That's my story and I'm sticking to it.

So you were introduced to the love of your life by a cheap component? It seems they ARE good for something after all.

[wagon]

"If that transistor fails for any reason in the next 15 years (life of system), an onsite service visit will be needed. This may involve me buying a plane ticket for $2000 to fly across country on a day's notice, grab a $200 rental car, $300 hotel bill, at least $100 in meals and incidentals, and about a $300 dollar two way round trip for FedEx charges to ship the 50 lb part out overnight freight.  So i say spend the extra $4 to save $2500 cost to the company in service costs."

R&D response was "you are not looking at the bigger picture. We have a price point to bring this system to market and we must hit that mark".

I mumbled something about who is not seeing the "bigger picture" and told the interviewer I could never work for him under those constraints - not with the "bigger picture" I have come to see from working in the field.

Later that year I was paged out to service one of those systems. A "fix it before the sun comes up or get it out of here" scenarios. I met my further wife to be that night. She was a technician at the hospital and saw me working on the system at 2 in the morning and let out a string of curses I can't repeat here.

We have been married going on 35 years now. That's my story and I'm sticking to it.

So you were introduced to the love of your life by a cheap component? It seems they ARE good for something after all.

I met my future wife after  discussion about why it cost $100 to replace a 40cent part..... in my case, an electrolytic. It happens!

[XFDDesign]

In fantasies... Only rare one will want to hire someone who goes against the system and collects evidence instead of doing a fucking job.

Which is why I'm pretty sure I'll be getting the Axe come the December Layoffs.

My day job has me working on a part for the medical space.
This part, has some serious problems.
Design is saying it's not a big deal (I'm an apps engineer), and it'll only affect a "small" number of customers who will then simply not use our part.
Marketing is simply holding up the rug while the designers sweep the problems under it.

Lucky me, my role has at least the authority to stop release until I'm happy (or fired and replaced with someone else).

[wagon]

I was thinking about this some more today.  A bad habit of some is putting batteries like nicad or NiMH on a PCB. The battery leaks at about the ten year mark, about a year after all parts are obsolete.

[David Hess]

Batteries are always a problem; either they have to be soldered in or use a relatively unreliable mechanical connection.  Parts with integrated batteries like non-volatile memories are even worse.

It gets even worse when the battery is used to backup critical information like firmware stored in SRAM, calibration constants, or purchased options.

[wagon]

Batteries are always a problem; either they have to be soldered in or use a relatively unreliable mechanical connection.  Parts with integrated batteries like non-volatile memories are even worse.

It gets even worse when the battery is used to backup critical information like firmware stored in SRAM, calibration constants, or purchased options.

Mounted off-board is a better way of doing it, if you must have a backup battery, so if it leaks the damage is largely contained. (although, I have seen corrosion 'wick' its way along wires then rotting the pcb.)