straka@cbnewsc.att.com (richard.j.straka) (05/24/90)
In article <15572@vlsisj.VLSI.COM> davidc@vlsisj.UUCP (David Chapman) writes: >In article <322@garth.UUCP> obnoid@garth.UUCP (Michael Kirschner) writes: >>In article <21000076@m.cs.uiuc.edu> kenny@m.cs.uiuc.edu writes: >>> >>>You *can* exceed the `absolute maximum ratings' of CMOS devices if you >>>take care to limit current; the protection diodes will generally handle >>>about 10 mA without any problems. > >Metal migration is bad enough, but the problem is that it's driven by >current density. So if you get necking at some point, the current > >An amusing story (I won't name names because it will embarrass a rather >well-known company): on our old HMOS process, we spec'd a current density Another amusing story. I saw it. With my OWN eyes. It is TRUE. NO joke. Back in the old days of EPROMS (4Kbit P-channel), the product engineer of this animal (company shall remain unnamed, although it shouldn't be very hard to pinpoint with the above information) showed me this cute phenomenon: At wafer probe, the 50V or so of programming voltage was too much for the chip to handle, so they were "pushed-out" with pulses of 20V, 25V, ... for a while, causing a LOT of localized heating, thus actually changing the profile of the metallurgical junction, enabling it to handle the increased voltage without breaking down. Yes, this is bizarre stuff, but read on; it gets better: He said: "Hey, take a look at this!" We went over to a wafer probe station, and during the first probing of the raw wafer, the above mentioned "push-out" got a metal line so hot that when viewing it under the probe microscope (with a probe cloth to block out light), you could actually see the metal line faintly, but distinctly INCANDESCE! Can you say flakey process? Yes, real production-line stuff. Thank God for N-channel EPROMs which are a LOT easier to program (physics helps out a LOT). -- Richard Straka AT&T Bell Laboratories, IH-6K311 ------------------------------------------------------------------------ UUCP: att!ihlpf!straka MSDOS: All the wonderfully arcane ARPA: straka@ihlpf.att.com syntax of UNIX(R), but without the power.