More Info on Imbedded Chips

greenspun.com : LUSENET : TimeBomb 2000 (Y2000) : One Thread

Found this link the other day-- it's some of the best writing I've seen so far about the problems of imbedded systems. The most interesting part IMO is about why chips may fail even though the application doesn't use the time/date function-- seems like that would greatly increase the percentage of malfunctions over the present estimates of 1 - 5% predicted now.
"Thus, even when only relative time is required by the OEMs, this may often be derived from chips that keep absolute time internally. Those chips that represent absolute time using two digit dates are subject to Year-2000 failures just as with computers and software as has been more widely reported. The logic "It does not NEED to keep dates, therefore it does not keep dates." is not based on what is actually happening within the chip. This has resulted in a number of systems being declared Year-2000-compliant when in fact their chips have not been tested. Examples of systems containing unassessed chips include remote control load management switches installed at consumer sites by electric utilities, automobile power train transmission control modules and major household appliances. Please see reference [4]. In the case where no date is set by an external agent, the chip defaults to its "epoch" date. This could be the design date, the date of manufacture, or some other, arbitrary, date. Non-compliant systems are subject to failure when the internal date reaches 1/1/2000, which in general will not be in step with actual time (since there is no means, or need, to input actual time at the turn-on point). In general, such a chip will reach 1/1/2000 internally AFTER 1/1/2000 actually occurs. This is due to natural delays introduced by the production life cycle, shelf life and possibly the duty cycle (the fraction of the time the chip is "powered up"). For non-Year-2000-compliant architectures, these delays increase the likelihood that most of their failures will occur AFTER 1 January 2000. One manufacturer has released documentation to its customers that some of their systems will not fail until 2006. [5]"

The OTHER Year 2000 Problem

-- Max Dixon (Max.Dixon@gte.net), July 12, 1998

Answers

Printed both articles this morning, and my husband and I read both articles, "Year 2000: who will do what and when will they do it?" and the one you mentioned. It kinda leaves you speechless! Thanks for telling us about it!

-- Barb-Douglas (bardou@yahoo.com), July 12, 1998.

Interesting information. But I still need some convincing about one point. That embedded systems measure relative time by reading two absolute times. So if there are so many systems out there with absolute times... where does the absolute time come from? They must need to have a back-up power source to maintain time - like on a PC motherboard. Otherwise when they are powered up wouldn't they just start from some epoch time? This will just make them like incrementing counters. The value of the absolute date being unimportant - hence no y2k issue. Right? There must be an answer to this. Even embedded systems were source code during development. Is it not just a case of getting hold of some typical source code and seeing which system call is used? ie. "get_absolute_date" or "get_seconds". How are these programs developed? Maybe there are only a few time libraries used thus reducing the task of assessing whether the actual methods used are going to cause problems after y2k or not.

Mark

-- Mark (aag78@dial.pipex.com), July 13, 1998.


Moderation questions? read the FAQ