• Please review our updated Terms and Rules here
  • Exhibitor application for VCF West 2022 is now open! If you are interested in exhibiting, please fill out the form here.

Change CPU in a PDP-8A from KK8A -> KK8E

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
I wander if a PDP-8E would run with a non working bus-load board? Given a small back plane with few other boards.
 

bobaboba

Experienced Member
Joined
Sep 7, 2013
Messages
389
Location
Scotland
Depends quite how non-working I suppose. Many (most?) of the omnibus drivers are open-collector so it might not take too much leakage here and there before the 1-level wasn't strongly enough asserted. TTL does usually work without input pull-ups but it isn't very immune to noise.
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
I guess that what I really ment was: Can a it run with the puls shaping circuitry broken?
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
I have hooked up my logic analyzer to the omnibus now. I have concluded:

Write into memory works. Memory timing looks alright. After deposit correct data appears on the MD-bus.

First I load address 5555
Deposit-this 5656
Load address 5556
Deposit-this 6565

If I then do
Load address 5555
Examin-this

I can see the correct data coming out on the MD-bus from the memory. But when the CPU shall drive the MD bus so the same data could be written back the lines are all high.

If I examine the same address twice the data is gone, of course.

If I then examine the other address i can see the correct pattern once.

On more interesting thing, if I do:
Load address 5555
and then press examine-next the address gets 7777 instead of 5556 (same goes for deposit-next).

Looks like CPU-problems to me. I think its a problem how the registers are loaded from the register bus. I also noticed that MD6 is stuck to "1" (held low on the bus). Could be a missing pull up, bus driver or the MB-flipflop.

8a-debugging.jpg

I've made some handy cards for measuring in the back plane. But I only have three double extension boards, this makes it impossible for me to put the CPU-major register card in an extension.
 
Last edited:

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
I figured out what the problems was.

1.) MD6 stuck to "1" was a shorted diode on the bus load board. Changed to a "D664" to a 1n4148 (the best I got home). 1N3606 is listed as replacement, what other diodes can be used? 1N4148 seams to work in some way.

2.) The other problem was that one of the control signals that decides what should go on the register bus was stuck to 1,2V (signal name EN1 on pin EA2). It turned out to be bad contact in the connector block between the M8300 & M8310 cards.

There is a problem to fit this connector block since it's impossible to push it all the way in. The sliders that holds the card in the cardcage collides with it. I have a cracked one that I cut a bit off on so it was possible to push it all the way in.

I have run two small toggle-in programs that works. It doesn't boot OS/8 from the RL01 disk though. Is there versions of OS/8 that requires KK8A?

I'll run a couple of more small programs, then it's time for diagnostics!
 

vrs42

Veteran Member
Joined
Nov 24, 2009
Messages
628
Location
Beaverton, Oregon
1.) MD6 stuck to "1" was a shorted diode on the bus load board. Changed to a "D664" to a 1n4148 (the best I got home). 1N3606 is listed as replacement, what other diodes can be used? 1N4148 seams to work in some way.

In general, I believe D664 are used where a low forward voltage drop is required. D662, on the other
hand are use where a relatively high forward voltage drop is required. (Some of the circuits that use
D662 require the forward voltage drop to be quite near 0.75V.)

I think the D664 is likely being used as a clamp here, so lower drop would mean less ringing.
In general, I suspect D664 circuits are less picky than D662 circuits, though, so you are probably OK.

The 1N4148 samples that I've tested actually turn out to be similar to the D662 (about 0.72V). They
vary by manufacturer, though, and most suppliers don't seem too picky about forward drop, so
you'd probably have to measure from the same lot as the ones you're using.

Vince
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
I actually measured it before I solder it onto the board. I'm not 100% sure that is was a 1N4148. I measured the forward drop to 0,60V, the ones that's mounted on the board measured to 0,58V.
 
Last edited:

vrs42

Veteran Member
Joined
Nov 24, 2009
Messages
628
Location
Beaverton, Oregon
I actually measured it before I solder it onto the board. I'm not 100% sure that is was a 1N4148. I measured the forward drop to 0,60V, the ones that's mounted on the board measured to 0,58V.

That should do nicely to replace D664, then!

I've seen a lot of switching diodes sold as 1N4148 with forward voltages all over the range
from 0.6V to .72V. It seems you have to sort them yourself. Worse, the 1N914 and such
also vary, so there doesn't seem to be a part number that you can reliably order. (Actually,
most are toward the lower voltage, so it's the D662 substitution that is most difficult.)

I've also wondered if Schottky diodes would be appropriate for D664, as might be used for
clamping diodes in a modern design. That would be less "period authentic", though.

Vince
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
Did some more testing. Many things works now, but group 6 instructions (IO) doesn't work. Can't load data from DATA-bus -> AC.

However,it works occasionally. I run a console local echo program as test. May it be the bus load board again? I'll hook up the logic analyzer (and oscilloscope) again to do some measurements.
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
I had some more time to spend on fixing the CPU. I found out that the signal that puts AC on the data bus was always asserted. I have isolated the problem to a failing SN7412. Unfortunately I didn't have any of those home. I ordered a couple from a local distributor and they will hopefully show up before the weekend.

I have 74LS12 at home but those can't sink enough current.
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
Changing the LS12 solved the problem with the AC always driving the databus. OS/8 doen't boot yet but I got CHECKMO-II to run.
 

pontus

Veteran Member
Joined
Aug 13, 2008
Messages
698
Location
Uppsala, Sweden
Interesting read, good work as always Anders. I might try this myself sometime in the future, so it's a good reference.
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
Interesting read, good work as always Anders. I might try this myself sometime in the future, so it's a good reference.

Thanks.

Finally I got some time again to work on this again. I have now ran all CPU tests maindec-8e-d0a to maindec-8e-d0j without any problems at all. So now the CPU seams ok.

When I try to boot OS/8 the RL01 drive tries to read on the disk (the white ligth goes out for a short while), then the error light (the red one) flashes and nothing more happens.

I wander what this can be. Remember that the computer boots with the KK8A cpu...
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
Ok, the KK8E works as good as the KK8A. Ran the memory test just to be sure and those works fine. But the computer having issues with the RL8A disk controller.

I tried the AJRLADO diskless controller test with the KK8E. It didn't even start. So I tried to swap to the KK8A cpu and got the exact same behavior. So maybe I did something wrong when I started the test? So I tried to boot OS/8 with the KK8A which used to work but now it stops in the same way as with the KK8E...

The test stopped @ address 5714, MD buff = 7402

Does some one have the source code to this test?
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
Os/8 build

Os/8 build

I ran the diskless controller test again, with the drives powered on and unloaded. It works! I also ran all other drive tests and pack verify test and all passed with both processors.

But I can't boot OS/8 so I suspect that the only bootable disk pack I have is broken. I never got to make a copy of it. It could be the boot-prom as well, I should probably test to manually enter the bootloader into memory just to be sure.

I need to build a new disk pack with OS/8 and I read about it in the OS/8 Handbook (1974 edition, which is to old to cover RL disks). I guess I can build it from paper tapes, I found those on bitsavers: http://bitsavers.trailing-edge.com/bits/DEC/pdp8/papertapeImages/setA_20040525/tray5/

I think they are to old and I need a version that can handle "BUILD" to a RL-diskdrive.

I guess that I also could boot from serial server with a emulated RK05 disk. I guess this is the easiest way to transfer files from my PC to the PDP and something I want to setup anyway. There is also an option to write thing onto floppies, but I don't have hardware to write 8" RX01 floppies from my PC (yet).

Which is the easiest way to go?
 

jackrubin

Veteran Member
Joined
Mar 12, 2004
Messages
832
Location
Chicago, IL and Buchanan, MI
SerialDisk is definitely worth setting up for lots of reasons (especially being able to load and run diagnostics easily). Dave G's 'diagpack2.rk05' image (http://www.pdp8.net/pdp8cgi/os8_html?act=dir;fn=images/os8/diagpack2.rk05;sort=name) has a version of BUILD that includes the RL handler and also has the handler on the disk, along with RL8A diagnostics. I think it has everything you need to build an RL system disk.

Jack
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
I have silently followed the SIMH/os8diskserver/PDP-8 thread and i must admit that I'm more worried about learning OS/8 than setting up SerialDisk.

I have found the same rk05 image, all the RL01 test I have actually used is from that.

I have also another problem, my power distribution board is broken. It's not possible to turn the PDP-8A off with the switch on the frontpanel. It's been broken for a couple of weeks now and I feel that I need to fix it first. The problem is that the computer is built around the BA8-C...
 

anders_bzn

Experienced Member
Joined
Jan 27, 2016
Messages
232
Location
Sweden
I started with the broken backplane. I have isolated the problem to one or maybe two components.
ba8c-schematics.png


I suspect that the optocoupler is broken since current passes between emitter and collector in the photo transistor even if no current passes trough the LED (the interlock is broken up). But it can be the transistor as well. When I connect the interlock I measure 5V on TP with the power switch in off state. When I turn it on I get 10V. But the 5V is more than enough to ignite the triac.

The transistor i easy to find, lots of MPS6531 on Ebay. But the optocoupler is harder. I can't even find a trace of a data scheet on it so I can substitute it with something else.

I took a photo with my mobile phone (I couldn't get my DSLR in there) of it, what could it be?
BA8C.jpg


Markings "Q2132" or maybe with a leading "O" och "0" doesn't give anything. "7629" is the date code.

One more interesting thing is that there are one resistor and two capacitors patched onto the board that is missing on the schematics.
 

vrs42

Veteran Member
Joined
Nov 24, 2009
Messages
628
Location
Beaverton, Oregon
Markings "Q2132" or maybe with a leading "O" och "0" doesn't give anything. "7629" is the date code.

According to this document:
http://www.pdp8online.com/pdp8cgi/query_docs/view.pl?id=246
the DEC part number was 19-14194, "high ct" (high current?), if that's of any help. There are a pile of opto-isolators made in DIP-8 packages, and I don't know what recommends one over another.

According to the BA11-A_RevE engineering drawings on bitsavers, there was also another opto-isolator, a 19-23364 "hi isolation 8kv" in that device.

Vince
 

Chuck(G)

25k Member
Joined
Jan 11, 2007
Messages
39,136
Location
Pacific Northwest, USA
Early optos often brought out the base of the phototransistor so that the sensitivity could be adjusted. 4-pin isos came out later. I suspect a plain-Jane 4N28 will probably work fine--it brings out the base of the transistor. Datasheet.
 
Top