paul@dual.UUCP (Baker) (10/09/85)
> > All Sun VMEbus cards have good arbiters, with metastability calculated > to be a problem once in a zillion years or so (based on the mfr's specs > on the flipflops that capture and synchronize the signals coming in from > the bus). I would be very interested in seeing some manufacturer's specs on metastability for, say, a 74S74. There do seem to be rules of thumb about such parts, but it is only in the last couple of years that I have seen a manufacturer even admit that a problem might exist. Fairchild are the only company that has put out anything of any substance on the matter, and only about their 74F series. The academics, as usual, are silent about the problem as it is just something messy to do with producing working equipment, and therefore totally uninteresting. Also, how many is a zillion? I would suspect in this case about five to ten. Paul Wilcox-Baker.
agn@k.cs.cmu.edu.ARPA (Andreas Nowatzyk) (10/11/85)
Unlike Paul Wilcox-Baker believes, the academics were far from silent
about this topic. Thomas J. Chaney (among others) fought many battles
to allert places like DEC and TI to the very existence of this problem.
The article "Measured Flip-Flop Responses to Marginal Triggering",
IEEE transactions on Computers, Vol. C-32 No. 12, Dec 1983
has a table of the data for a collection of common Flip-flops.
Ouoting said table:
MTBSU(20ns) MTBSU(40ns)
74S74 10 s 7x10**9 s (worst case, 74-75 dated
devices)
7474 -- 10 s
74LS74 -- 0.4 s
74F374 2x10**13 2x10**35
MTBSU is the Mean Time Between the Synchronizer events where the FF is
still Unresolved for a given synchronization time. Please look up the
article on the precise set-up conditions and how to scale the results
for a particular circuit. Talking about "Zillions" is not unreasonable
given that these figures span 35 (!) orders of magnitude.
--
-- Andreas Arpa-net: agn@cmu-cs-k.ARPA
uucp-net: ...!seismo!cmu-cs-k!agnjohnson@uiucdcsp.CS.UIUC.EDU (10/11/85)
/* Written 12:47 pm Oct 9, 1985 by paul@dual.UUCP in net.arch */ The academics, as usual, are silent about the problem as it is just something messy to do with producing working equipment, and therefore totally uninteresting. /* End of text from net.arch */ "Introduction to VLSI Systems" by Mead and Conway, a very famous and well-read book among academics, discusses metastability problems, and makes references to a report by TJ Chaney and CE Molnar (of Washington U) in the April 1973 issue of IEEE Transactions on Computers that discusses these issues. In other words, these problems have been known by academics for some time. There is actually a lot of theory that can be applied to these problems.