How does solid state hardware just die?

[elfs]

I got home this evening to have Omaha tell me that we had no wireless. The two wired machines still had Internet access, so the router was fine. The old bridge, a Linksys WAP11 v1.1 showed a solid light on above its ethernet port, and no light at all on the wireless port. Every attempt to communicate with it, either via ethernet or the USB connector, was rebuffed. The thing was dead.

After dinner I drove out to Frys and in an hour had a new 802.11g Linksys WAP54G up and running. I even reprogrammed it to use DHCPC so I won't have to scan the entire network looking for the damn thing next time, and told my DHCPC server to preserve its MAC-to-IP address across reboots. It has better security now. Yamaraashi-chan was having a nastier case of Internet withdrawl than either Omaha or I.

But how does this happen? Linksys WAPs are solid state. It's sat in our pantry for the past eight years (the last time I downloaded drivers for it, the most common version of Windows was 98!), quietly ticking away, doing its job. It has no fan, no hard drive, no moving parts at all. I've only updated the flash four or five times in its life. Why would it suddenly just keel over and die in the middle of a transaction like that?

Comments

[acelightning.livejournal.com]

[(Anonymous)]

It's actually pretty likely that instead of the solid-state components failing, you had a capacitor flow out, or some other part of the power system fail.

[acelightning.livejournal.com]

[mouser.livejournal.com]

These and other reasons. It's an electrical product generating heat and having parts smaller then a human hair and relying on household current. Eventually it WILL die. They all do.


Like all good software geeks, just tell yourself "It's a hardware problem."

[elfs.livejournal.com]

Yeah, I know. It's just annoying. Anyway, we're upgraded to 802.11g now, which is nice to have. Omaha and I can now blow each other up from anywhere in the house!

[phred1973.livejournal.com]

An excellent ability to have ;)

Don't forget the interconnects!

[lemur123.livejournal.com]

There's also the solder joint side of the equation. If the unit was built recently, I'd suspect tin whisker growth (it looks really cool!). An eight-year-old unit would be pre-RoHS, and would have enough Pb present to passivate the Sn. In that case, since it's an IPC Class 1 electronic device, I'd suspect ionic contamination from the initial manufacturing process (late 1990s was when manufacturers were first switching to no-clean fluxes) leading to dendritic growth. In either case, it's nothing you did - molecules actually moved around and migrated in different directions, driven by the electropotential differences between charged circuits ("electrochemical migration").

http://nepp.nasa.gov/WHISKER/

http://www.electronicsproductionworld.com/articleView~idArticle~71680_173413417372008.html

[acelightning.livejournal.com]

Fortunately you've only been breathing flux fumes

[ideaphile.livejournal.com]

If you'd been vaporizing the lead (1749 C) or tin (2602 C) in the solder itself, you wouldn't be here, and neither would those PC boards. :-)

. png

[acelightning.livejournal.com]

[lemur123.livejournal.com]

the fluxes and binders that you heated up when soldering are pretty nasty, too - breathing in charred rosin and scorched organic acids is worse than smoking cigarettes, and the isopropanol vapours attack mucous membranes. our ever-paranoid site Health & Safety team made me get a $2,000 fume extractor and air purifier for each of our soldering stations. i think it's overkill, but i agree that the fumes can be pretty nasty.

[acelightning.livejournal.com]

[lemur123.livejournal.com]

*blushes* same here - don't tell my eh&s coordinator!

oh - use aluminum foil on top of the newspaper, or else squirt a bit of water on it. that will help keep down the static so you don't fry the electronics.

[acelightning.livejournal.com]

Re: Fortunately you've only been breathing flux fumes

[lemur123.livejournal.com]

Tin-lead solder is usually at the eutectic point (melts at about 186C), and the nastier bioavailable oxides start forming at around 900 F, so our site Health & Safety team makes us limit our solder pots to 500 F and our solder irons to 700 F. 900 F is still too low for toxic vapours, of course, but the residues get everywhere, inevitably get ingested, and bioaccumulate.

Small form factors suck!

[lemur123.livejournal.com]

ooh yes, it's a pain dealing with part with hundreds of leads with a spacing of less than 0.016" centre-to-centre. We've had to start using binocular microscopes just so our operators can see the workspace, and few people have hands steady enough to hold the iron still. Our solder iron tips are so fine that they have to be replaced each shift - that's $12 per workstation per day!

As if that wasn't bad enough, the designers just specified some surface-mount capacitors that are only 0.002" x 0.001"! I can't even SEE those without magnification!

Re: Small form factors suck!

[acelightning.livejournal.com]

Re: Small form factors suck!

[lemur123.livejournal.com]

Tip 1: make sure you aren't trying to solder through the solder mask. The copper traces are usually covered with a thin acrylic coating that solder won't stick to. If you try soldering without removing the coating first, it will SEEM as if it's taking a while for the trace to heat up and the solder to flow - really, it's already pretty hot, and it's the solder mask that is keeping the solder from flowing. By the time you've burned through the acrylic, you've also burned the epoxy holding the copper trace to the PWB. Careful use of a sharp razor blade or Dremel tool will remove the mask, hopefully without damaging the trace.

Tip 2: get a soldering iron with temperature controls, and check the tip temperature before use with a calibrated three-wire thermometer. You can also get a Metcal station - they self-calibrate from line voltage. Never let it get above 700 F, even for huge components. A good soldering iron is expensive, but will pay for itself in reduced scrap. (note - I'm not following my own advice; my home soldering iron is a Radio Shack cheapie)

Re: Small form factors suck!

[acelightning.livejournal.com]

DIE COSMIC RAYS! I HATES YOU!

[lemur123.livejournal.com]

or even a stray cosmic ray.

oooh! i HATE debugging those! especially when it isn't strong enough to actually cause permanent damage, and is just enough to cause a bit to change state.

Re: DIE COSMIC RAYS! I HATES YOU!

[acelightning.livejournal.com]

Re: DIE COSMIC RAYS! I HATES YOU!

[lemur123.livejournal.com]

and the smaller and smaller dimensions on the silicon aren't helping! with an old 286 processor, several cosmic rays need to hit the same transistor to add enough electrons for it to change state and cause a single-bit error; on a modern processor, the charge from a single cosmic ray is enough.

Re: DIE COSMIC RAYS! I HATES YOU!

[acelightning.livejournal.com]

[qtplatypus.livejournal.com]

Flash memory has a limited lifetime perhaps that ran out. Caps are the other option as Anon said, also a high voltage spike can kill solid state stuff.

[featheredfrog.livejournal.com]

Given the above comments (and I agree with 'em), it would be interesting to open it up and see if there are any interesting marks, or odd charring* on the mobo.

--
* Or if the government censorbug finally had enough wit'cha.

[shunra.livejournal.com]

Don't you wonder just what data was going through it when it crashed? Anthropomorphize it for speculative amusement.

[wookie-cd.livejournal.com]

Well, those comments were fascinating to read - I'll add my two cents.

Whenever I look at gear that "just died", the first thing I always look for is a bulging electrolytic capacitor. Those things just go pop. I even recall buying a motherboard where one of the marketing points on the box was that it used all-solid capacitors instead of electrolytic types.

In a a way it's all speculation - certainly the other commenters' posts are equally valid.

If you take the black box that is a wireless router and break it down into the major functional groups, you usually have a 5-port switch, with one port internally wired to the actual "router" portion, which in turn has another port to the wireless hardware. The software running on the router performs bridging between the wireless and wired LANS, provides services like DHCP and DNS relaying to the LAN side, and performs NAT between the router's LAN and WAN ports.

The WAP54G is similar in that it has a LAN port and a wireless port, and a cpu bridging the two. It seems pretty clear that something is preventing the CPU from finishing its initialisation of the lan port, or from doing anything to the WAN port. There's a strong chance that the LAN light is the hardware's power-on default, and the thing is completely brain-dead.

[gromm.livejournal.com]

Our tech support desk (of which I used to belong) gets that very same question all the time.

Except you *know* that their eyes will glaze over the moment you give them any kind of technical explanation like those listed here.

Here's a better explanation: because your bridge cost $79.95, not $495. It has cheap components in it that will likely fail sooner than later. Your bridge lasted 8 years, and that's far beyond its useful lifetime. As you've noted, it's long past obsolete.

More than likely, one of those cheap components was a capacitor, as previously noted. It's also worth noting that to find and replace that component would cost more in labour than it would to replace the whole thing. Electronics are like that.

[ewhac.livejournal.com]

Don't feel too bad. I've gone through two inexpensive Linksys WAPs -- an 11 and a 54. The 11 packed up without reason; I assumed it had taken a power spike. The 54 just slowly started dropping out all the time.

I'm currently running a medium-priced NetGear WG302v2 WAP. So far it's been okay, although trinsf insists it still drops connections to her laptop. I've never managed to work out how to diagnose the local WiFi neighborhood.

Parenthetically, I'm finding it harder and harder to find a WAP qua WAP. They all seem to be WAP plus router/firewall/DHCP server/cable modem/print server/espresso machine.