[Date Prev][Date Next][Thread Prev][Thread Next] - [Date Index][Thread Index][Author Index]

Re: Component selection

We are using commercial memory. It is the memory on the PIC.
If a latchup occurs, the current jumps. We have circuitry to detect the latchup
which will cycle power to the controller board. A cheap trick is to put a 2 to
3 ohm resistor in series between power and the VCC pin of the chip. SEUs which
can be bit flips are harder to detect. Sometimes they don't upset the system.
We plan to generate a checkerboard pattern in unused memory and monitor it for
flipped bits. Also any variables or control registers will be periodically
checked and refreshed. If you have a device with configuration registers, one
way to make sure they are always set right is to periodically rewrite the
configuration values in those registers. 

We also found that we can flip bits in the EEPROM as it is read. The memory
does not get changed. Just the output buffer of the EEPROM. This means a high
energy particle can change the bit pattern for a CPU instruction being read.

We have three levels of latchup detection. One is a current switch which will
cycle power when the current increases dramatically. Another is an external
watchdog timer which when timesout, will also cause power to be cycled. The
third is a current sensor that is read by the processor to detect smaller
current increases that the current switch won't trigger on.

Remember as the electronics are continually exposed to radiation, the current
consumption will continually increase. When we tested out board, it consumed
60ma at the start of the test. At the end, the current just from the processor
being exposed was about 138ma at the time of failure. You need to leave a good
size cushion for the current switch so you don't have it cut out prematurally.


> Thanx for your response to my query.   You mention
> a dose before the EPROM is *all* set to zeros. 
> But a single corrupted bit could cause a crash, right
> ?
> Info I've gotten from elsewhere suggests a probability
> of corruption of 1 bit per 10K of memory every hour
> for
> non-RAD hard static CMOS memory.  Are you doing
> anything to correct for random bit errors,  or
> are you using Rad hard memory ?
> Thanx again for the info,
> Ken  AC1H   
> (BTW-I'm not Karl, I'm just using his email account)
> __________________________________________________
> Do You Yahoo!?
> Yahoo! Mail - Free email you can access from anywhere!
> http://mail.yahoo.com/
Via the amsat-bb mailing list at AMSAT.ORG courtesy of AMSAT-NA.
To unsubscribe, send "unsubscribe amsat-bb" to Majordomo@amsat.org