In the late 80s and early 90s, a group of IBM-PC clone
manufacturers, including Compaq and Hewlett-Packard, among others,
defined an extended ISA bus that included a 32-bit data bus and very
fast DMA capabilities. This bus was named EISA (for what else,
Extended ISA). This bus was built into machines for a few years, but
never became a major player. By the mid-90s, PCI had driven EISA
pretty well into oblivion.
I had designed an
EISA-VME Bus Interface for a client, who specialized in
Bus-to-Bus-Bridges and Bus Extenders. Shortly after my EISA board
development project, my client was asked to develop an EISA
Extender. Bus Extenders are used to provide more slots for
peripheral cards than are available in a normal backplane. The size
of a backplane and the number of cards that can be plugged into it
are limited by electrical issues. A bus extender consists of two
cards, one plugged into the main backplane and the other plugged
into a physically separate backplane, although the secondary
backplane is usually located close to the primary. A cable connects
the two Bus Extender cards. If the bus extender is truly
transparent, then every card in the secondary backplane may be
accessed just as if they were plugged into the primary backplane.
Furthermore, the timing of the bus cycles to access the secondary
backplane do not violate the specifications for the bus, although
the secondary cycles need not be as fast.
Since I was up to speed on EISA bus design, my client asked me to
analyze the design of an EISA-EISA Bus Extender to see if it was
possible. Essentially, I performed a preliminary functional design
and then analyzed the timing. The key to an extender is the
round-trip delay in the connecting cable and interface
drivers/receivers, and I assumed a realistic time as the basis for
my analysis. I made progress through the first set of programmed I/O
bus cycles, and I figured out a way to handle interrupts. But when I
got into some of the DMA and bus master protocols, I found that it
was impossible to meet all of the EISA specifications even with some
creative design concepts. I wrote my findings as I went along, and
when my client saw my analysis of the bus master and DMA timing,
they agreed that a fully functional EISA Extender was not feasible.
It is interesting to note that at that time, there was a company
advertising an EISA-Extender, but based on my analysis we decided it
could not meet the full EISA specification. This turned out to be
the case.
My client abandoned the idea of developing an EISA Extender based
on this preliminary study.
