[755-1] Setup and hold time measurements

[Konstantinos Dovrolis <dovrolis@cs.wisc.edu>]

A couple of people asked about the setup and hold
time measurements. 

First, remember that the setup time for a flip-flop
is the minimum time *before* the latching clock edge
in which the data should be stable in order to be
latched correctly (and without the risk of metastability,
but forget this for now).

The hold time is the same thing, but it refers to the
minimum time *after* the clock edge in which the
data should be stable.

The question is how to measure these two parameters.
The standard approach is to do "crash tests", i.e.,
simulate waveforms in which the data change value
before or after the clock edge, and then see if the
right value is latched. If yes, decrease the spacing
between the clock and data transitions; if no, increase
it. Doing so in a binary search fashion should give
you an answer after a few iterations with a precision
of say 100ps. 

And to make the measurements more consistent, please use 
the next values:

* Your clock rise time should be, say, 100ps (the fall
time does not reall matter).

* Your data rise and fall times should be say also 100ps.
Note that in general, the answers that you will get can be
different in the low-to-high and high-to-low transition
of the data signal.

C.