[sdiy] Finding the correct IC...

[Neil Johnson]

> Where does that say "D" stands for "data"??  Besides, "data" isn't much
> of a distinction among flip-flops.  It's all data if it stands for
> something :)

True!

> "Another flip-flop available in an integrated circuit form is the clocked D
> (delay) flip-flop..."
> _Fundamentals of Logic Design_, Charles H. Roth, Jr (1975, 1992).

Me thinks this was around the time that these terms were being defined,
hence the two variations.  Now it has settled on "_D_ata" because of the
single-bit data memory function that it implements.

Suffice it to say, we all know a D-type!!

> >Again, from Lenk, if both inputs are set high, both outputs go low.
>
> Then Q and Q' would be the same.  Since they are supposed to be
> compliments, R = S = 1 is "not allowed".

Actually, the two outputs are normally the complement of each other, but
that's only a naming convention.  It might be better to think of them as
"P" and "Q" that just happen to be the complements of each other under
certain conditions, and both go low when both inputs are high.

> Right, if both inputs are changed to 0 at the same time it might
> oscillate, or have other unpredictable results.

Actually, in this instance its pretty clear what happens, as already
mentioned.  What is _not_ so predictable is what happens when both inputs
go from 1 to 0 as we won't be able to predict the outcome.  It will depend
on so many factors that it would be difficult to predict.

> Compared to modern digital the gate delays of these older logic families
> are quite large, so "simultaneously" seems possible depending on what is
> driving it.

Actually, the faster things get the harder they are to control and use --
slow logic is/was quite immune to short spikes on the inputs or supply
lines as they were just too darned slow to notice.  Now this fast stuff
would quite happily notice the noise and process it like any other signal.
As speeds increase we start to see physical parameters affect the logic
circuits, like transmission lines on PCBs, overshoot, termination pads,
etc.  Sometimes speed is a curse :(

Remember that the speed of light is finite, the point being that signals
travelling along a PCB trace will take a finite time to propagate along
the track.  For example in copper the speed of light is approx 0.67c, so a
signal will take 1ns to propagate 20cm.  If you're running at 1GHz (1ns
cycle time) you'll want a rise/fall time much less than 1ns -- lets say
0.1ns.  So you can't have a trace any longer than 2cm.  And it must be
properly terminated otherwise you'll get ringing (reflections) which will
(a) over/undershoot the inputs causing damage, and (b) take much longer to
settle, so screwing your 0.1ns transition time.

Who said RF was a dead subject?  Its only been waiting its turn, watching
the approaching CPU clock speeds and smiling in a "my-time-will-come"
grin.  Respect is due to the engineers at Cray -- they really knew their
stuff!

Neil

--
Neil Johnson :: Computer Laboratory :: University of Cambridge ::
http://www.njohnson.co.uk          http://www.cl.cam.ac.uk/~nej22
----  IEE Cambridge Branch: http://www.iee-cambridge.org.uk  ----