home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
CP/M
/
CPM_CDROM.iso
/
cpm
/
z280
/
new280.txt
< prev
next >
Wrap
Text File
|
1994-07-13
|
8KB
|
125 lines
Some Ideas for a new Z280 product
---------------------------------
Soon after the Z280 became available I acquired two samples and began
experimenting. At this moment I have an S100 board working of sorts and am
beginning to give serious consideration to bringing out a product based on the
280. The idea behind this document is to get ideas from prospective users on
what form it should take and what features should be included to be of most
use.
Currently I am using an ICD XLM180, a 64180 based S100 system with 2
floppies, hard disk, tape backup and 512k memory, to which I have added an
Illuminated Technologies 1024x1024 color graphics card. Any new system should
have at least the capabilities of this, and preferably more.
Bus Structure
-------------
The present system uses IEEE S-100 bus which I find very satisfactory.
However I would not use it for the 280 system. Why? Because first, the 280
cannot generate strict IEEE spec signals, specifically the sM1 signal. Though
it is possible by means of external logic to distinguish fetches of operands
and opcodes and thus generate a sort of sM1 signal, it would be useless as
because of the cacheing and pipelining in the 280, fetching and execution are
wholly asynchronous events, thus none of the usual uses of this signal would
work (hardware debuggers and the like). Second there is the lack of choice of
really up-to-date peripheral cards for the bus. After lookink at several
alternatives, I have chosen the IBM AT bus.
The AT bus is lousy. It is sadly deficient in a number of areas, but most of
these do not matter to a single user system. In its favour there is an enormous
wealth of cheap and technically advanced peripheral cards available, and the
mechanical hardware needed for building an AT-like system is cheap and readily
available. The 280 can easily be made to deliver 286-type signals to give
wholly compatible bus timings. As far as I can see the only peripherals that
would not work without modification would be hard-cards and the like which
contain ROMS full of 8086 code, which would have to be changed. But as the
system would come with a hard disk as standard I see no need for hard-cards.
Disks
-----
I propose two 3.5 inch floppies (800k each) and a 40MB hard disk. Though 3.5
inch disks are not widely used outside 68000 systems, they have surely proved
themselves and are a big improvement over 5.25 and (heaven forbid!) 8 inch
floppies. Also they will help to keep the size down. The AT, and my XLM180, are
huge beasts, and I don't want my new machine to be any bigger than an XT. But
the disk controller used will support 4 floppies so external 5.25inch drives
can always be used to read older disks. The hard disk controller would be SCSI
and easily replaceable for future upgrades.
I/O
---
Two serial ports, Centronics parallel, and SCSI ports would be included on
the mother board, as well as a 2400/1200/300 baud modem. Additional I/O can of
course be added on plug in boards in the six expansion slots provided.
Keyboard and display
--------------------
A normal AT-type plug-in keyboard would be used. The display would be wholly
unlike anything on an AT, being a bit-mapped 1024x768 color screen of 15 inch
diagonal. The dispay would use its own processor, a TMS34010, and its own 768KB
of memory which would be independent of and not encroach the main address
space. Graphics cpu program memory would, however be accessible to the Z280 so
new graphics primitives could be developed and loaded. This memory, accessible
to both processors would be accessed on an alternate-phase system permitting
both cpu's to access at full speed without using DMA. Graphics commands would
be passed through a small bloc of this memory. The bit-mapped display would
enable anything that could be done on a Mac to be done on this machine, and in
full color. And the resolution is sufficient for decent CAD applications.
Memory
------
There would be room on the motherboard for 32 memory chips, giving 1MB using
256k chips or 4Mb using 1Mb chips. This could be expanded on external memory
boards to 16Mb, less that used for ROM and the 128k or so of graphics
primitives.
Power Supply
------------
A 200 watt supply using a special circuit of my own which gives very high
efficiency and large power/volume ratio. This will help keep the size of the
unit down.
Operating System
----------------
I'm currently thinking of some form of ZRDOS, either a modification by me,
Echelon, or a third party, or a wholly new clone of same. To be useful in such
a system, ZRDOS needs adding to it (a) large memory model support (b) graphics
support, and (c) multitasking would be very useful. Large memory models need
careful thought; I am already using large (by this I mean >64k) programs on my
XLM180, but to do the addressing I have to manipulate the MMU registers from
within the applications program, a highly unportable approach, and one
impossible on the 280 as changing the MMU registers is a privileged instruction
and can only be done in system mode. So we need a set of standardised OS calls
for this purpose, as well as a means of loading a large program (cf. EXE files
in MeSsyDOS). As far as the graphics are concerned, why not take a look at the
way the Mac does it. On the IBMs, it's a hopeless mess and I don't see much
point in emulating that. Also we need an operating system call to
enable/disable the cache. Again it's a privileged instruction but occasionally
a user's program may want to disable the cache if it's executing self-modifying
code in tight loops, though my own experiments suggest that the pipeline is
probable more to blame than the cache. In any case, despite rumors to the
contrary, 95% of self modifying code WILL run correctly on the 280. One very
special case where the cache must be off is where code is loaded via DMA and
then immediately executed - the cache will still contain the old code. But no
USER program is likely to do that sort of stuff - much more likely in a BIOS or
IOP - which would be in system mode anyway.
So these are the ideas I currently have for the 280 system. Please let me
know what you think - any suggestions for improvements - etc. Please keep it
reasonable - I know you'd all like 100MHz systems with 16Mb memory, 1GB optical
disks and 4096x3072 pixel realtime color graphics, but it would be so
preposterously expensive that the customer base would be zilch. This system has
to be cost effective compared to an AT - thats the reason for using cheap AT
clone parts in it to start with. And don't destroy its compatibility - I see no
reason we can't have a system that concedes nothing to IBM, Apple, Atari and
Commodore in up-to-the-minute features yet will still run all our old favorite
Z80 programs - even faster than before.
I can be regularly contacted on Canada Remote Systems BBS 416-232-0442. Less
regularly on Znode Central, ICBBS Ottawa or Holly Park RCPM. Or you can
phone me at home (416-299-0502) after 6pm EST or at work (416-439-4273) 9am-5pm
EST Mon-Fri on voice. Or even FAX (416-439-3875). Or write to: 1131 Sandhurst
Circle #111, Scarborough, Ontario M1V1V5, Canada.
Greg Trice, Dec 1987.