Organization:
NCRI
To:
iaem-y2k@new-focus.org
References:
1
Robert Spencer - EMERMGMTX wrote:
My question about embedded chips that do not have a specific
date/time call
is...how do they know what time/date it is, even if they have
somehing
permanent burned in if they are not constantly powered? In other
words, if
they have their electricity shut off because they are not run
constantly,
they would not be synchronized with any clock/calendar. A chip
cannot keep
time if it isn't continually hooked up to a power sorce.
Often such products are designed with an integral battery to provide a
tiny amount of electricity for the chips even when the
product is not plugged in. I suppose you could accomplish the same
thing by storing the information in nonvolatile memory, but
I would have to check with one of our electrical engineers to see if
that option is ever used.
Either the
burned-in information means nothing at all, or we are going to have
chips
failing for the next 50 years as they reach some arbitrary end
time/date.
We will continue to have some embedded chip failures for years after
y2k. Fortunately, as I explained in my previous post,
when such failures occur spread out over a long time they are
considerably less likely to lead to disruptions.
How will we even know if it was date related since thousands of
chips fail
every day right now?
Why do we have to know if the failure was date related or not? I
would guess that quite often we would not know the cause of
the failure. A chip will fail, and the personnel at the facility will
have to replace the part. Unless the part is especially expensive
or important, or the failure is unusual or part of a larger pattern, a
thorough analysis of the failure mechanism will probably not be
conducted.
This latest scare doesn't make sense to my simple mind.
The problem is not the drawn out failures of certain types of timer
chips. The problem is that computer chips are rarely custom
designed for each application. Engineers will pick "off the shelf"
chips that may contain date codes even if they aren't needed for
that particular application. This means that there are likely to be
more y2k vulnerable embedded chips than originally believed.
When those chips have a properly set clock they may add to the already
high rate of failures on Jan. 1, 2000 When the chips
do not have a properly set clock they will add slightly to the
existing rate of failures; unless a large number of those failures are
concurrent they will not pose the same type of threat that the Jan. 1,
2000 failures do. Also, the paper in question (Embedded
Systems and the Year 2000 Problem, available on the NIST Web site at
http://www.nist.gov/y2k/embeddedarticle.htm)
explains that proprietary date coding methods may not be detectable
with testing equipment unless the testing equipment and the
part being tested are made by the same company. This will mean that
some y2k susceptible date sensitive embedded chips may
be falsely thought to be y2k compliant.
To simplify, there are probably more embedded computer chips
vulnerable to failure on y2k than originally thought, and the
existing test methods may not be as effective at finding these chips
as we originally thought. There will, therefore, be more
embedded chip failures than originally predicted on Jan. 1, 2000.
This higher rate of failure means that disruptions in service on
or around Jan. 1, 2000 are more likely. This is the point that is of
most interest to the emergency management community.
(Unfortunately, I do not know how significant these factors will be in
increasing the rate of y2k failures; NIST did not give, and
probably does not yet have, specific numbers regarding these
problems.) The failure of some more embedded chips for years
after y2k is certainly bad news to the facilities with these chips,
but it is probably not something that will affect emergency
managers.
Chris Taylor
Y2k Policy Committee Member
National Crisis Response Institute
www.public.usit.net/jupiter
615-386-3626
-- Charli Claypool (claypool@belatlantic.net), December 02, 1999.