Dr Jefyll wrote:
Another factor is thermal cycling. A chip that, in the normal course of business, gets hot and then cools down again is at risk due to the resultant expansion and contraction. My understanding is that after thousands of cycles this can lead to mechanical failures such as separation of bond wires from the die.
Someone asked Bob Pease about that, referring to the different coefficients of thermal expansion. He's the famous industry guru and analog chip designer who worked for National Semi and who was killed in a car accident a couple of years ago. His response was that the temperature cycling on the bond wires where they contact the die wasn't a problem. The temperature most ICs' dice run at is not very high though. When I was working in applications engineering at the VHF/UHF power-transistor manufacturer in the mid-1980's, I sometimes did infrared scanning of uncovered transistors while they were running under abusive loads. The maximum I ever saw actually
running was over 350°C. At that temperature the die might not have lasted even an hour, because of migration of the metalization and other factors; but it does put a new perspective on temperatures.
Quote:
But of course
we all use cool-running CMOS chips, right?
Hmmm. I once had a VIA that got so hot it burned my finger! -- aka my "digital" wattmeter.
That was a CMOS chip, but I'd made a bus contention boo-boo and was drastically abusing the part (which, incidentally, survived
).
I've had a CMOS Rockwell VIA get super hot because of latch-up. Fortunately, it didn't hurt it. It was feeding a printer across the room that was powered by a different circuit, and I guess some kind of spike produced the latch-up.
Quote:
BigDumbDinosaur wrote:
I would avoid parts that can't be positively identified as new. You don't know what pulls were subjected to before you get them.
Fair enough. Like you, BDD, I'd rather have a new chip than a pull. But preferences may yield if price or availability is a problem. Myself, I'm still keeping a bunch of chips that are essentially pulls from my own previous projects. Frankly, many of them have been handled & stored somewhat carelessly. But it doesn't bother me to reuse them -- I just make a mental note in case there's trouble. But I don't recall any.
Of the thousands of unused ICs in my inventory, a few hundred are from a source whose handling I'm not as confident in, so they're kept separate.
Quote:
BigDumbDinosaur wrote:
[...] PLA/PALs and GALs from way back. Most of those seem to lose data retention after about 20 years or so.
Good point about data retention. What I'd worry about most are EPROMs and similar chips which rely on a minuscule charge on a floating gate (although chips that use fuse technology aren't perfect either). Of course if the part we're buying is blank (unprogrammed) there's no data to worry about retaining.
I've refreshed the data in EPROMs of commercial products I have from the 1980's to prevent that problem. I also kept a copy of them on disc.
Quote:
Going slightly OT, I suspect data retention may be a far more serious problem in modern equipment -- particularly Flash memories that use analog voltages to store more than one bit in each individual floating gate. A possible instance of this involves a digital camera I purchased last year. One day it inexplicably lost the ability to turn the reset of file numbering on & off -- the UI no longer lets me determine that particular option. I suppose maybe I could download the firmware and do a refresh. But meanwhile I view the new, ultra-high capacity thumb drives with some distrust...
I don't like to be the guinea pig on anything, including new high-density memories. Our son was one of the first to get a 4GB SD card for his camera, and it went bad right away. I guess now they can do 4GB without having to store multiple bits in one cell. Single-bit-per-cell flash however usually has data-retention specifications of anywhere from 40 to 200 years, which is much longer than EPROM. EPROM was usually guaranteed for ten years and usually went at least 20.
Quote:
ps- Garth, were those tantalum caps you were referring to?
No, just the inexpensive aluminum electrolytics, not even high-temperature. We've never used tantalums in the products. When I worked at TEAC in the early 1980's, the only electrolytics that went out were power-supply capacitors where they used too low of a WVDC to try to save money and space.