On Mon, May 18, 2009 at 07:30:06PM -0400, Adam Maas scripsit: > On Mon, May 18, 2009 at 7:12 PM, Graydon <o...@uniserve.com> wrote: [snip] > > It would be surprising if all the chips involved could make it to 10 > > years; 20 would be well past surprising and into "how on earth?" > > territory. Running a microchip will eventually break it, as the > > electrical current causes atoms to move and this eventually makes a > > transistor unable to trans. Relatively low power, and relatively low > > use (no one has their camera on 24/7, unlike a server) camera chips are > > going to last relatively well, but there's a certain amount of passive > > diffusion involved, too. All that cleverness isn't really stable; it's > > just that the house of cards takes some years to fall over. > > Design lifetimes on that sort of failure are in the decades. Server > failures are generally heat or mechanical failures (something > overheated or spindle/fan failures) and often both.
Or current spikes, yes. But chip failures, particularly memory and processor failures, do happen. Managing materials migration is still something of a black art, especially as new process nodes are developed and theory meets practise. > I've got functional systems dating back more than 20 years (Mac II, > circa 1987) and 2 systems currently up 24x7 that have been running > since the late 90's (one PowerMac B&W G3, one PC) There is very likely a certain amount of long-right-tail going on there, and also a certain amount of 'larger process, less problem'; the Mac II probably doesn't have any migration issues to speak of, it's not miniaturized enough. Current ~40nm processes have much more of a problem with this sort of thing, there's just less distance involved. Camera parts won't be on that kind of process, though I'd suspect they're all down around ~100 nm by now; power is going to be a major driver for process selection in anything battery powered. And in a camera, the most vulnerable chip is the sensor. There are 2009 ACM papers presenting improved modelling methods for predicting sensor node (which if I understand the abstract means "pixel") lifetime, so I suspect the 'black art' part is still at least somewhat the case. > 20 years for a camera is easily achievable under light/moderate. I won't go for easy; possible, sure, but this is something that gets carried around and regularly rattled, a wide (or at least potentially wide) thermal environment, and where the most important chip is frequently directly exposed to daylight. > The biggest issues will be batteries on those bodies that take > proprietary batteries and overall shutter life. Chances are that > getting replacements may become an issue and battery lifetimes are > 18-24 months. Of course the Pentax bodies that take AA's lack the > battery replacement issues. No arguments whatsoever on the batteries. -- Graydon -- PDML Pentax-Discuss Mail List PDML@pdml.net http://pdml.net/mailman/listinfo/pdml_pdml.net to UNSUBSCRIBE from the PDML, please visit the link directly above and follow the directions.