Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!henry From: henry@utzoo.UUCP (Henry Spencer) Newsgroups: net.analog,net.audio,net.ham-radio Subject: Re: Electronics wearing out Message-ID: <5689@utzoo.UUCP> Date: Mon, 17-Jun-85 14:52:02 EDT Article-I.D.: utzoo.5689 Posted: Mon Jun 17 14:52:02 1985 Date-Received: Mon, 17-Jun-85 14:52:02 EDT References: <11251@brl-tgr.ARPA> Organization: U of Toronto Zoology Lines: 93 Unfortunately, there really are wear-out mechanisms even in semiconductors. They are less obvious than the ones in tubes, but no less real. Plastic IC packages (used for most ICs in commercial equipment) are not hermetically sealed, and may eventually develop leaks that let air and moisture at the chip; the resulting corrosion kills it. Why do you think the military uses ceramic (truly hermetic) packages almost exclusively? Diffusion does occur even in solids. Various forms of material migration can occur in the presence of electric currents. These things are significant concerns to semiconductor manufacturers. Remember that commercial ICs are optimized for low cost rather than ultimate maximum lifetime. Power spikes, static, and similar forms of overstressing may possibly account for a significant fraction of semiconductor failures in the field. It is very difficult to shield completely against such things, although "99%" protection is much easier and increasingly common. Note that power behavior during powerup and powerdown is rather ill-specified and hard to control completely. EPROMs and EEPROMs inherently have limited lifetimes if they are being reprogrammed regularly, because the erase/program techniques involve pushing electrons through materials that are normally insulators. The technique used to do this is basically brute force. This unavoidably involves some degradation of the insulator. Do it too many times, and it won't insulate any more. EPROMs and EEPROMs will *eventually* lose their contents anyway, because they rely on holding charge on very-well-insulated electrodes. Although the insulation around those electrodes is of truly superb quality, it's not perfect, and the charge will leak off eventually. In the absence of unusual stress, this takes a very long time. Nobody is quite sure how long, because we haven't had the devices that long! Extrapolations from stress tests may not be 100% valid, although they do give hints. Circuit designs which use batteries for volatile memories simply need their batteries replaced occasionally. They don't "become useless" any more than a flashlight becomes useless, although in some cases the battery replacement may be a lot of hassle. (When expected battery lifetime exceeds expected equipment lifetime, it's tempting to just solder the battery in.) Things like semiconductor lasers in CD players are being operated "hard and hot"; this aggravates failure mechanisms like material migration and solid diffusion. Often there is a deliberate tradeoff between power output and lifetime, as with incandescent bulbs. For example, stadium lighting is often run at higher-than-normal voltages; the bulbs burn out quickly but yield a lot more light meanwhile, and this is a reasonable tradeoff in that application. Similarly, commercial semiconductor lasers have very limited power outputs because of difficult technical problems. Getting the necessary power with available components may *require* some deliberate loss of lifetime. Even when higher-power components really are available, it may be a good deal cheaper to opt for shorter life. (Just how much extra are most consumers willing to pay for long-life components? Regular incandescent bulbs heavily outsell long-life bulbs, remember.) > This troubles me. I had thought that we had a reasonable progression of > development of electronics during these days of solid-state gear > replacing tube equipment that meant longer life for any device. ... Longer, yes. But not infinite. Not even the phone company's highly conservative, heavily over-engineered equipment is expected to last more than 40-50 years (and this is *not* a zero-maintenance lifetime, either). > But now I see that designers either deliberately build in inevitable > failures (the use of volatile ROMs and batteries instead of PROMs or > other permanent devices) or the solid-state stuff I was led to believe > had indefinitely-long lives really is as short-lived as vacuum tubes > (the laser diodes mentioned above). You pays your money and you takes your choice. If you are willing to pay double or triple current prices, you could probably get some very long-lived equipment. (The high cost of military electronics is not *all* corruption and overcharging.) (95%, maybe.) > I may be insane, but I really expect electronic equipment to operate > forever. I don't want to buy any that has built-in and inevitable > failure. ... > I realize that consumer products have built-in planned obsolescence. > I'm not talking about small appliances, which I expect to die and be > replaced. I'm talking about multi-hundred-dollar equipment. Am I > totally unjustified in expecting such things to be designed to > continue to work indefinitely without maintenance, except for > mechanical stuff like motors, belts, pots, etc? Basically, yes. There ain't no way to make it work indefinitely. And making it work "a long time" is quite expensive. Semiconductors have made it relatively easy to build gear with far longer lifetimes than tube equipment, but this is a (large) quantitative improvement rather than a fundamental repeal of the Second Law of Thermodynamics. -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry