10 Apr
2002
10 Apr
'02
10 a.m.
On Wed, 10 Apr 2002, Tony Duell wrote:
Actually It's not that bad, recompiling my Sweet16 16 bit CPU takes
only about 5 minutes using Xilinx tools (on a fast machine though) The tools
are available free (webpack). The tools do require Windows, but some people
have reported success using them under WINE on Linux.
The main
reason I have never played with them is I could never find a
Fuse Prom Burner schematic that looked reasonable. I still would like to
That's probably because fusible PROMs are notoriously difficult to 'blow'
correctly. They're also getting hard to find now, and as they're strictly
one-time-programmble, I really wouldn't want to use them in a new design.
Some of the modern electricallly eraseable technologies are getting
pretty fast -- certainly fast enough for experimental processors. And
they're a lot easier to program. OK, the devices are massive by
comparison with the old bipolar PROMs. but they're also not expensive (at
least not compared to 2 or 3 of the old bipolar fusible-link PROMs, and
you need to reckon on buying at least 2 for every one you will use as you
will make programming mistakes), so it's worth 'wasting' most of the
capacity of the chip.
do a TTL computer with fused based proms ( or
EEPROM's as modern
substitute )for control logic. I am just finishing up a nice FPGA design
but thinking this is going to be a pain to get a serial prom and have it
burned too. 50,000 sure ! quantity 1, HA-HA you must be kidding.
(EPC1441LC20 altera 440,800 bits -- any place in Canada). While TTL is
low density you don't need to pay $$$ for a license for modern
programiable logic, have the software needed 5+ years down the road! (
That is assuming TTL is still around 10 years from now )
For experimental and educational projects I much prefer TTL (including,
of course the CMOS versions of the TTL chips, like 74HCxxx parts). It's
easier to prototype with, easier to test (you can clip the 'scope or
logic analyser wherever you like), and easier to see what's really going
on. It's quicker to make small changes to the circuit as well (on an FPGA
design I did about 5 years ago, a full compile of the main chip took
overnigh (OK, PCs have got faster since then, but FPGAs have also got
larger!)). That meant every small change took a day to test. A soldering
iron and/or wire-wrap tool is a lot faster for changing a few connections
:-) You also aren't tied to a proprietry program running on some
computer/OS that I don't have...
Please don't attempt to convince me that FPGAs make more sense for
production. I don't need convincing of that...
I wouldn't call those SSI parts.
> 74HCXX chips as 74LSXX is harder to find and lots more power. How ever a
> 74LS382 style alu would be used rather t
han a 74LS181 if LS parts were
> The PDP11/05 was 2 full hex-height boards
just for the CPU, so around 200
> chips. It used mostly TTL, but also some PROMs containing the microcode.
> And the TTL included chips rather more complicated than just gates --
> things like 16*4 RAMs, 4 bit latches, multiplexers, etc.
This was ball park figures. A 4 bit shift register ,4 bit up counter
,dual 4-1 muliplexer would be the typical ALU parts. I was looking at
Why? The '181 has many more operations, some of them useful....
Nah, there 100's of them in a $19.00 FPGA...
used. A 8 bit x 1 register file is a 8 bit
addressable latch and a 8/1
multiplexer. The alu design is for a undefined 24 bit processor with a
Sure... I guess 74x170s are really hard to find now :-(.. All these
wonderful chips I grew up using are discontinued :-(
-tony
Peter Wallace