Newsgroups: comp.sys.transputer
From: paul@walker.demon.co.uk (Paul Walker)
Subject: Re: Strobe skews
Organization: Paul Walker Consultancy
Date: Thu, 2 Jun 1994 14:29:22 +0000
Message-ID: <770567362snz@walker.demon.co.uk>

In article <CqptKw.7HJ@tbit.dk> mmj@tbit.dk "Mads Meisner-Jensen" writes:

>Hi
>
>I'm working with hardware design, using transputers, but as the
>transputer is rather new to me, I have a few questions that I could
>need some help on. Here goes:
>
>1. Why are all strobes notMemRd, notMemS1-S4 referenced to notMemS0?

Because if you are connecting to a memory device, you want all the 
memory signals referenced to the memory Chip Enable signal, which
is most likely the notMemS0 signal.

> 
>2.  There is a skew between ProcClockOut and notMemS0, such that
>notMemS0 changes state within a specified window around a clock edge
>(see [1] below).  Furthermore, all other strobes possess a skew
>relative to notMemS0. Now, my question is about this skew; Is this
>skew meant to be added to both sides of the notMemS0 window (see [3])
>or is the strobe transition window *within* the notMemS0 window.
>
>
>                   ________          ________          ________
>ProcClockOut    __/        \________/        \________/        \__
>                _______________
>notMemS0                 \_____\__________________________________ [1]
>                _______________
>notMemS1-S4              \_____\__________________________________ [2]
>                ___________________
>notMemS1-S4         \______________\______________________________ [3]
>

Strictly yes as defined in the T805 data in the 3rd Edition databook ('92),
(page 94). 

In fact, as the S0 skew is characterised rather than tested, it is
probably valid to use the same figures as S0 for the other strobes,
and if you add 25% or so to the figures you should be pretty safe.

Do you need to know the relationship for a wait-state generator? or
for something else? And do you really need to know the relationship?

Paul Walker




