Effect of radiation on computers

[John C White <ohjcw@utxdp.dp.utexas.edu>]

Prior to my current incarnation in the field of radiation protection, I was a
senior Nuclear Effects Engineer in the defense industry, with a supplier of
communication systems.  The phenomenon of radiation damage in semiconductor
devices is well researched.  The response of Silicon and other materials to
ionizing radiation is different than human tissue.  Damage begins to appear in
a device exposed to a radiation field at about 500 Rad(Si), and increases on a
fairly linear basis.  As feature size (density of transistors per cm2 or cm3)
increases, susceptibility to damage or upset increases. Modern error correction
techniques can compensate for most damage or loss of data from upset.  The
susceptibility varies by manufacturer, software type, and systems integration.
The bottom line:  If you place a commercial PC in an environment where it will
be irradiated on a more-or-less constant basis, you need to consider moving it,
replacing it with a military radiation hardened version (Various mfrs.), or
downloading all data to a remote location on a regular and frequent basis.  The
best source for published information on this topic is the annual publication
of the IEEE Nuclear and Space Radiation Effects (NSRE) conference, which
is published each year.  There are a number of con$ultant$ in the field who
will be glad to help you with your problem.
John C. White RSO
The University of Texas at Austin OEHS
304 E. 24th St.,  SER 202
Austin, TX 78712-1024
(512) 471-2042
(512) 471-6918 FAX
OHJCW@UTXDP.DP.UTEXAS.EDU
-------------------------------------------------------------------------------
Here is a question which I noticed on a usenet newsgroup and which the
poster  has asked me to send to radsafe:

From: Jim Ronback (jronback@direct.ca)

What is the effect of background ionizing radiation on commercial computers?

How often will high energy particles from solar
flares, cosmic rays, residual radioactive decay in solder (lead used be
uranium millions of years ago) and in encapsulating materials cause
anomalous behaviour in a PC / workstation / net server? Is there any data
available which would allow me to estimate the liklihood of having one or
more bits flipped in memory or processor registers in
commercial computers which do not have radiation hardened devices?
If the likelihood is high enough then perhaps one should not use single
bit flags in safety critical software.

Any advice or pointers would be appreciated.

Jim Ronback, P.Eng.
Senior Safety Engineer
Hughes Aircraft of Canada Ltd.,
13951 Bridgeport Road
Richmond, B.C
Canada  V6V 1J6

Tel: 604 279 5789



Michael Williamson
                                        ph +61 7 3365 4504
Radiation Protection Adviser            fax +61 7 3365 1577
Occupational Health & Safety Unit       m.williamson@mailbox.uq.edu.au
University of Queensland
Brisbane 4072
Australia