[comp.os.cpm] Fast PCPI Applicards

nsayer@uop.EDU (Nick Sayer) (06/27/89)

I recently found this old file and set about speeding my Applicard up
by 40%. I now have a 10 Mhz CP/M card running inside my 7 Mhz Apple IIgs.
It's not too tough, so read all about it!...

---------- cut here ----------

Revised 6/26/89
---------------
 
     When is fast not fast enough?  When it's easy to go faster!
If you ever compared a Softcard to an Applicard you know what a
difference 4 MHz can make. It'd be great if you could make that 
same 200% increase in speed over the Applicard but that's not so 
easy.  What is easy is a 20%-33% percent increase.
     When the Applicard was designed, I doubt the designer ever
considered anything faster than 6MHz.  He did consider sound
design practices that not only make the normal card "bullet-proof"
but makes acceleration possible.  I have modified 4 cards and
have obtained reliable results at 7.160, 7.5 and 8 MHz.
    How is it possible?  Simple, just change the CPU, crystal and
depending on your memory's access speed, the DRAM too!  Here are 
three combinations I tried:
 
- 7.160 MHz:   1- change crystal from 12 to 14.31818 MHz
                  (this is a Apple crystal).
 
               2- change CPU from a Z80B to a Z80H.
 
- 7.500 MHz:   1- as above but use a 15MHz crystal.
              
               2- select 150nS RAMs until you get reliable
                  operation.
 
- 8 MHz:       1- as above but use a 16MHz crystal.
 
               2- install 120nSec RAMs.
 
 
for the brave:     
 
above 8Mhz:    1- change crystals.
 
               2- select RAM's.     
 
               3- change 7404 chip to maintain the phase 
                  relationships of the clock.
 
               4- keep things cool and cross your fingers!
 
 
     What does this effort get you?  Obviously,  if there's a
lot of I/O, the Apple will pace things.  If it's just the Z80
rolling along on its' own.... you'll get the full effect.  For
example, the Sieve runs in 91 seconds and Calculations in 42
seconds. (Standard Byte Benchmarks) at 8Mhz.  These pretty
much are direct 33% increase over 6MHz times.
 
 
Where to get the parts:
 
Z80H: try JDR or BG Micro or if all else fails, let me know.
 
120nS DRAM:  Microprocessors Unlimited
 
Crystals: JDR or BG Micro.
 
 
          A warning about DRAM's for Z80 applications:
 
    These parts must be 7-bit (aka 128 cycle refresh). The common
8-bit (256 cycle) parts used in IBM's WILL NOT work.  This rules
out TI and Matsushita.  If you get your DRAM's from uP unLTD specify
Mitsubishi.
 
The bad news..........
 
- if you have a communications program that must run at 6MHz
  or the timing is wrong, you could have problems.  MEX does
  NOT have this problem (STAT CLOCK 93, check with SLEEP 60).
 
 
Jim Lill                      or via "Sparrow" 
POB 10098                             716-377-1113
Rochester, NY 14610                   ------------
                                      GEnie [JIMLILL]
                                      ------------
                                      Potpourri
                                      ------------
                                      Z-Node Central
                                      ------------
                                      FortFoneFileFolder
 

NEW NOTES, 8/12/86:
 
- I've found that the clock won't always start. Selecting a
  new 74LS04 fixed the problem. Actually, I used a 54LS04
  with full Mil 883 burn-in. The first one worked fine at
  8Mhz.
 
- Bussman makes a very nice socket for their new Radial lead
  PC mounted fuses. Part number PCS. It fits the existing holes
  in the board and reliably takes wire lead crystals.
 
- Fujitsu makes a 100nS 7-bit DRAM. I got some for $2.70 from
  Marshall Electronics, try uP Unlimited.
 
- with the 100nS parts and the sockets I'm gonna go for 10Mhz
  but will probably have to change more "glue" chips.
 
- I've done four 8Mhz boards and four 7.159MHz boards to date
  all still working...

 
NEW NOTES, 8/19/86:
 
- it looks like a better part for the clock is a must. Try a
  54LS04 (same specs except for temp. but seem to be better),
  a 74 or 54ALS, a 74 or 54S or perhaps the best yet a 74 or 54F.
  The latter has the speed of a "S" and the power consumption of 
  a "LS".
 
- using a "F" clock chip and the the rest mentioned in the 8/12 notes
  above, I have things running at 9.4+ MHz. When I get a 20MHz rock
  I'll go for 10MHz.
 
 
NEW NOTES, 10/5/86:
 
- this file has been renamed to be consistent with the trend to name
  Applicard related files staring with PCPI making them easy to find all
  in one spot on a directory.
 
- I've had the 10MHz rock in for better than a month now and all works
  fine. Keeping things COOL is a MUST, even at 8Mhz.
 
- I have confirmation that the Applicard card will work with the //gs
  in the //gs high speed mode.


NEW NOTES, 6/26/89:

- This section written by Nick Sayer. I've pumped mine up to 10 Mhz without
  any trouble.

- Apparently the Z-80H is now called the Z84C0010-PEC. This chip is rated at
  10 Mhz.

- When chip manufacturers came up with 256Kx1 RAM chips, they gave up on the
  64Kx1 configuration. However, the pinout of a 256Kx1 differs from that of
  a 64Kx1 only by 1 extra address pin. The makers of the PCPI board
  conveniently tied this pin high for us, so using 256Kx1 chips instead of
  64Kx1 chips works just fine. You'll be wasting 192K of RAM, but it should
  not be too tough to use this extra space to emulate the "piggyback" RAM
  expansion board and get a 192K RAMdisk in the bargain.

-------------------------------------------------------------------------------

---------- ereh tuc ----------

------------------------------------------------------------------
Nick Sayer  |  nsayer@uop.edu  |  N6QQQ @ WB6V
Disclaimer: The BBC would like to appologise for that announcement