Conference rusure::math

1116.0. "Metastability and Chaos, Setup Time in Logic" by ANT::JANZEN (cf. ANT::CIRCUITS,ANT::UWAVES) Fri Aug 25 1989 14:31

    Is there a chaos conference?  i thought there was.
    
    In the specification of an MSI logic integrated circuit, 
    if it has a clock and flip-flop or registers, a parameter called
    Setup Time is given.  Setup Time in the data book is the time
    a logic circuit must wait after setting up data for the chip
    before sending the clock.  If the clock comes too soon after the
    data is set up, the data outputs will be wrong because wrong data
    was clocked in.  Because of extra delay on the clock in the chip,
    setup time is almost always negative at the pins on ECL.
    
                       ____________________
    clock^____________/                    \_________________
                        ___________________________
    data in____________/                           \_________
                          ___________________________________
    data out_____________/?//////////////////////////////////
    
    Setup time is measured by sliding the data change edge into the
    clock edge until the data go bad.  If you keep moving the data,
    you go into the metastabiliy region, a a region of, that's right,
    Non-Linear behaviour.  It is treated statistically.  I think
    that setup time could be treated statistically, also, because the
    definition isn't very precise.  It could be, the measured setup
    time is the time at which the probability is, say, 95%, or as close
    to 95% or higher as you can get that the data is good.
    Something like that.  Maybe 99%.
    
    The metastable region is currently characterized by measuring the
    probability at a given clock/data intpu time that the output is
    good.  I will be measuring that on a chip this afternoon.
    
    On the other hand, if this is a non-linear unpredictable region,
    maybe it is also self-similar, fractal, and chaotic.
   
    What treatment from chaos would be profitable?  The information
    theory approach?
    
    Sorry to bring up computer science so much in here ;-)
    Tom