W8SDZ@mit-mc@sri-unix.UUCP (07/17/83)
From: Keith Petersen <W8SDZ@mit-mc>
The following is forwarded from my RCPM system. Replies to the author,
please, not to me.
--Keith
--forwarded file--
AA6PZ
Paul Zander
86 Pine Lane
Los Altos, CA 94022
Fellow Computer Hobbiest:
Several friends and I have been trying to use the popular CCS2422
Floppy Disk Controller board in our S-100 systems. We encountered a few
problems which cost us substantial frustration until the solutions had
been worked out. We had tentatively thought that we had bad EPROM's and
bad disc controller IC's before making the modifications described
below. The IC numbers refer to the rev. B board, the changes apply to
the rev A board as well.
1. The buffer which drives the data-in bus, U40, has to drive 8
lines at the same time. Depending on the configuration of the rest of
the system, it may have a peak current of 1/2 amp or more for a few
nanoseconds. Although this IC is close to the S-100 ground pins(50 and
100), this current has to flow through traces which are common to the
rest of the board. I have actually seen this current surge cause a
voltage spike of 2 volts. Needless to say, this can and does cause
false inputs to many of the other IC's on the board. The solution is to
cut the trace on the top side of the board to U40, pin 10. Next, on the
back side, connect U40, pin 10 by a piece of insulated wire to the
ground trace near pin 100, and continue the wire around the edge of the
board to the top side near pin 50. Number 28 or 30 wire-wrap wire is
suitable. The inductance of the wire is primarily a function of its
length. There is little to be gained by using an 18 guage wire.
2. Although the ground traces appear at first glance to be
connected together, there are several places where a ground trace starts
at one side of the board, goes to several IC's and then stops in the
middle of the board. Your DC ohmmeter will measure continuity, because
the traces are connected by long paths around the edge of the board.
More reliable operation of the 2422 can be achieved if these "dangling
ends" are connected together. The candidates for connection are as
follows:
U7 pin 7 to U8 pin 20
U20 pin 7 to U21 pin 8
U13 pin 10 to U14 pin 7
U35 pin 7 to U44 pin 8
U39 pin 7 to S-100 pin 50
If your system uses the S-100 pins 20 and 70 as grounds, the should be
connected in as well.
3. This one corrects one of the symptoms of the inadequate
grounding mentioned in 1 and 2. Historically, it was found first,
because the board would often hang up in an "Auto-wait" state when using
the monitor EPROM. Short spikes on SOUT or SINP propogate the the logic
to the clock U42A when the CPU is really accessing a memory address of
F733. This change prevents short spikes from clocking the Auto wait
flip-flop but has no effect on desired operation.
Cut the +5v trace going to U42 pin 2. It is on the back of the
board and you do not have to remove the IC or the socket to get to it.
(Hurray!) Now connect pin 2 to U35 pin 11. This is the logical OR of
SOUT and SINP. In the case of a short spike, this point will be back in
the logic low state before the clock signal gets to the flip-flop.
4. Some of our boards did not work with some of the Western Digital
IC's. The problem was most pronounced with 5 1/4 inch drives.
Eventually we found that the problem is that the board design requires
the WD1793 to operate outside of its specified range. In particular,
the data sheet requires that the data inputs be held for 70 ns after the
end of the write pulse. However, the 2422 actually switches off the
data bus buffer before sending the write pulse to the 1793 or the
control latches. It is only the PC board capacitance which causes the
circuit to work! The solution is to keep U39 turned on a little longer.
There are several solutions, but the most reliable is to substitute an
unused section of open-collector NAND gate U5 for the section of U20
which drives U39 pin's 1 and 19. Connect U20 pin's 4 and 5 to U5 pin's
13 and 14 respectively. Cut the trace to U20 pin 6 and connect a wire
form that trace to U5 pin 12. Connect a 1K pullup to U5 pin 12, and
somewhere along that trace connect a 80 to 100 pF capacitor to ground.
This will slow down the rising edge of the signal to U39 pin 1.
After these changes, the CCS2422 works quite well. With some
experimental changes to the wait-state circuitry, one board has
successfully done disk operations with a CPU clock speed in excess of 6
MHz. More work remains to be done before claiming that a complete
speed-up can be reliably achieved.
The problems with inadequately engineered ground connections and
buffers which are turned off too early are not unique to the CCS 2422.
You might keep them in mind when you have problems with other circuits
which give erratic behavior. I hope this letter helps make your
computer experience more enjoyable.
Permission to copy this letter or its contents, in whole or in
part, is granted provided that credit is given to this writer.
Sincerely,
Paul ZanderCC.Otken%utexas-20@sri-unix.UUCP (07/19/83)
From: John Otken <CC.Otken@utexas-20> Paul Zander's fix contained a few typos: 2. S100 pin 50 to U39 pin 10 instead of 7 4. U20 pin 4 to U5 pin 12 instead of 13 U20 pin 5 to U5 pin 13 instead of 14 U20 pin 6 to U5 pin 11 instead of 12 1K pullup to U5 pin 11 instead of 12 Before I installed his modification, my CCS 2422 barely functioned. It seems to work MUCH better now. However, I haven't thoroughly tested it yet. Also, one plated through hole did NOT conduct through the hole. -------