thomas@ssc-vax.UUCP (Thomas Uczekaj) (07/28/87)
Would anyone be interested in discussing the effects of radiation (choose your environment: Nuclear Weapon, Space etc....) on semiconductors and radiation hardened techniques for designers (engineers using software, such as VLSI design)? For example would you have any information concerning the idea of Linear Energy Transfer (MeV/gm-cm) to characterize energy loss by cosmic rays? How is this concept used to give valuable information concerning the degradation of the semiconductor.
karn@faline.UUCP (07/31/87)
Yes, I would find that an interesting discussion, for a very practical reason. AMSAT, the volunteer group that builds satellites for amateur radio communication, is facing this problem. We built and launched a satellite (Oscar-10) in 1983 aboard Ariane into an elliptical orbit (inclination 26 degrees, perigee 4000 km, apogee 35000 km). Oscar-10 carries an onboard computer and quite a bit of random CMOS logic. SSI logic is no problem; most RCA CD4000B series ICs are hard to a megarad. The microprocessor was the rad-hard Sandia version of the 1802. However, the Achilles heel was the 16K bytes of Mostek 4116 dynamic RAM (I know, I know, but it wasn't *my* idea...) At first, the memory performed fine; the (8,12) Hamming code error correcting hardware had no problem correcting the one or two soft errors that occurred on each orbit during passage of the inner Van Allen belt. However, less than 3 years later the memory hard-failed due to accumulated radiation damage. The next spacecraft in the series has been redesigned to use Harris HS-6564RH static CMOS memories which are claimed to function after 100 kilorad exposures. It will be aboard Ariane V21 (the first Ariane 4 launch) late this year or early 1988. While we hope we have solved our radiation susceptibility problems for the present spacecraft, it is pretty obvious to everybody that the 1802 is an outdated microprocessor. The 3-axis stabilized spacecraft now being designed for geostationary service will have much greater onboard computing requirements, so we are considering the Harris 80C86 microprocessor. (No, we do *not* plan to run MS-DOS on it!) Has this chip proven itself in radiation environments yet? How about larger, cheaper and/or faster rad-hard memory chips? Phil Karn, KA9Q Asst VP, AMSAT
kopaz@rdlvax.UUCP (John Anthony 'Echo' Kopaz) (08/01/87)
Posting-Front-End: GNU Emacs 18.47.1 of Thu Jul 9 1987 on rdlvax (berkeley-unix)
if you want to discuss some aspects of emp or hem i would be interested.
but these are emr as opposed to nuclear decay.
echo.
--
john a. kopaz [aka echo.]
associate research scientist / software engineer / test specimen
voice: 213-410-1244 -- fax: 213-216-5940 -- corporeal: rdl
arpa : kopaz@rdlvax.rdl.com 5721 w. slauson ave.
uucp : ...!{psivax,csun,sdcrdcf,ttidca}!rdlvax!kopaz culver city, ca 90320peter@ethz.UUCP (Peter Beadle) (08/04/87)
In article <107@rdlvax.UUCP> kopaz@rdlvax.UUCP (John Anthony 'Echo' Kopaz) writes: > >if you want to discuss some aspects of emp or hem i would be interested. >but these are emr as opposed to nuclear decay. "do you want to talk about X" messages seems to comprise the entire contents of comp.lsi. In an effort to get some sort of discussion going and to learn something I pose the following questions: 1. What are the primary causes of radiation (in the largest sense of the word) induced circuit failure. 2. What proportion of the effects are permenant and what proportion are transitory. 3. What makes a radiation hard circuit "hard". We have already heard about some quantitive measures of "hard", what I would like to know is what I have to do to design a "hard" circuit. 4. What effects do process parameters have on radiation sensitivity. 5. I hear a lot about alpha particle hits on memories. What does an alpha particle do as it passes through a 1Mbit D-ram.
buyno@voder.UUCP (Matthew Buynoski) (08/07/87)
In article <169@bernina.UUCP>, peter@ethz.UUCP (Peter Beadle) writes: > In article <107@rdlvax.UUCP> kopaz@rdlvax.UUCP (John Anthony 'Echo' Kopaz) > > 1. What are the primary causes of radiation (in the largest sense of > the word) induced circuit failure. > > 2. What proportion of the effects are permenant and what proportion are > transitory. > > 3. What makes a radiation hard circuit "hard". We have already heard > about some quantitive measures of "hard", what I would like to know is > what I have to do to design a "hard" circuit. > > 4. What effects do process parameters have on radiation sensitivity. > > 5. I hear a lot about alpha particle hits on memories. What does an > alpha particle do as it passes through a 1Mbit D-ram. Answer to 1 (as I know it). The chief causes are: hole-electron pair generation in the oxide, and outright smushing of the semiconductor lattice. To the former: energetic particles in the oxide layers over the semiconductor create hole-electron pairs in the oxide. The electrons so generated are mobile enough to enter the semiconductor, but the holes are not. This leads to a build-up of positive holes along the Si-SiO2 interface. That in turn affects the electrical properties of the semiconductor near-surface region. Typical effects are the inversion of p-regions (bipolar bases, e.g.), shifts of threshold voltage in MOSFETs, lowered breakdown voltages due to field-induced junctions. The second effect is the outright damage to the lattice by the energeticparticles. These events tear up everything...all the diodes start to leak, oxides break down, etc. Usually the first mechanism occurs first. Answer to 2. They are mostly all permanent. Alas. The alpha-particle problem in memories is an example of transitory radiation "damage". In this case, many more hole-electron pairs are created in the holding capacitor of the DRAM cell, which effectively messes up the stored signal. A single alpha can create about 10e6 such pairs. The same alpha may well have also created a few hole-electron pairs in the oxide above the cell, doing permanent-type damage but at a slower rate (the oxide is thinner, and it is more difficult to create pairs in it than silicon, since the bandgap is considerably larger in SiO2). Actually, some of the effects are reversible, but with annealing at, say, 250 to300 degrees C. Not very practical in a working circuit. Answer to 3. Most of the things done to "rad hard" parts are done to reduce the surface state and fixed charge densities. Surface states are very nice homes for the radiation-generated holes and tend to bind them right on the interface, which is of course the absolute worst place for them. IF the holes don't have as many places to "settle in", it is possible that some of them may wander around and eventually find an electron to combine with..poof goes a bit of your trouble. Reducing fixed surface charge helps in that it is positive and if it is less, then it takes more holes to get you up to the point where the total positive charge is too much. Or, it pushes back the day of reckoning for a while. I think the answers to 4 and 5 are somewhere in the bulk of the answers to 1,2, and 3, so I won't repeat them here. Hope this helps.