[Info-vax] OT: Steve Wozniak

[David J Dachtera]

"Richard B. Gilbert" wrote:
> 
> David J Dachtera wrote:
> > [snip]
> > Transitions in the "strobe" signal should take care of that.
> >
> That's just the point!  When do you assert the "strobe" signal?
> The only valid time is after each bit in the parallel port has settled
> to an unambiguous 0 or 1.  On a typical parallel interface, there can be
> a fairly long delay for all the bits to settle and the delay can be a
> little different for each bit.  The delay is not necessarily symmetrical
> either; it may take longer to set a one than a zero or vice versa!

I guess that's what I was referring to when I mentioned being "stuck" in
the classic Centronics-parallel paradigm.

Consider a CPU executing a "move from memory to register" instruction:
how does the CPU know when all of the bits at the source address have
"settled" to their final values?

There should be no more "settling" then there is in the process of
assembling a serial bitstream. All you're doing is breaking the bits out
into separate channels. This is done at the same frequency as the strobe
signal. Think of it this way: eight parallel channels of serial data,
all running at the same bit rate ("baud rate", if you prefer the older
terminology, though the application is somewhat inaccurate). "Strobe"
provides the clock.

Remember when serial interfaces used a separate clock signals for
transmit and receive? (Think synchronous vs. asynchronous.)

Same deal. It could work.

D.J.D.