• Please review our updated Terms and Rules here

Dec KA-10

tradde

Veteran Member
Joined
Apr 30, 2003
Messages
2,137
Location
Katy, Tx
Don't ask me why, but I have decided to try to use Eagle to draw the backplane and
components to a KA-10 pdp-10. I used one of the machines at Penn State and still
have fond memories of it. This is a long (very long) term project. I got the idea from
Vince Slyngstad who is redrawing components for the pdp-8 series. I don't know
how far I will get or if it will be and on again off again thing. It's a huge undertaking.
I wish there were some KA-10s in museums, but I know of none.
 
Tim,

Yeah, after much looking around the internet over recent years, it sure looks like there are no surviving KA systems. I think the computer history museum has two consoles, one of which is in the console bay. I never saw any recent photos of any machine having bay 1 or 2. I am even curious just where all the known consoles are that survive.

The KA backplane wiring is probably the biggest of all the challenges you will have in this project. If the already-wired backplanes from bay 1 and 2 were in existance somewhere, I think the rest of a system could be pieced together (but it would likely have to have semiconductor memory.)

Lou
 
There are at least three that I know of. One in Germany, Sweden, and the US

I have heard there might have been one in Sweden and one in Australia. I never heard of any being in the US.
Would love to see some pictures.
 
There are at least three that I know of. One in Germany, Sweden, and the US
I'm guessing the one in Sweden is complete (if the owner is who I think it is). What about the others?

Tim: What is your goal? Do you want to be as close to the original as possible or will you take any modern "shortcuts". I'm guessing the backplane could be substituted with a few big multilayered PCBs
 
I'm guessing the one in Sweden is complete (if the owner is who I think it is). What about the others?

Tim: What is your goal? Do you want to be as close to the original as possible or will you take any modern "shortcuts". I'm guessing the backplane could be substituted with a few big multilayered PCBs

My goal is to keep the backplane and boards as original as possible working from the original drawings. Yes, it would be
a multi-layer board, but huge. Actually several huge boards as it's 2 big backplanes and one smaller. It's a daunting
task, but I always will remember the KA-10 from Penn State with fond memories.
 
First, I must say it saddens me to think that the KA10 from the University of Queensland's Museum of Computer I.T. may no longer be out there. It was operational in the early 2000's, from what I understood, and was sent to an undisclosed fate by the time the Museum's doors were shut.


  • It makes sense that museums in general, would rather have the "ultimate" incarnation of the line, the KL10 - and it is reassuring that several survive.
  • It is not unheard of to find a KS10 in private collections.
  • I think perhaps the first of the line to go MIA was the KI10 - a disappointing family member that never gained popularity, probably due to it's disappointing performance compared to projections.
  • However - the idea that all the KA10s would go extinct, was a surprise to me. If there is one remaining, it seems likely that it's boxed and perhaps forgotten, or unrecognized.

Tim -

I don't want to discourage you, but you do realize what you're undertaking here? If you were able to identify the various parts, it's very likely to be cheaper to buy them individually on the curio market, than to build. Finding an original backplane or frame would indeed be fortunate. I think these would have been 30ga or 28ga Kynar Wire-Wrap [silver] and given the skyrocketing price of metals over the last 10 years, wouldn't expect any to survive unless part of a larger identifiable assembly - or as a relic in a misplaced crate. Maybe one could begin with a few phone calls to Sweden? [I too, heard about that destination]

As for conversation of "multilayer boards" this confuses me. I'm pretty sure the backplanes would be wire-wrapped - not PCBs. Individual boards ~ less certain, but probably double sided G10.

Do we have documentation of actual boards? I haven't looked - maybe later tonight. I thought I had images, but looking yesterday after your post - I was wrong. (?)

So you'd try to make a system faithful enough to accept genuine boards intermixed with those of your own creation? Am I understanding that correctly?

Picking a KA10 would put you in a pretty reliable family member, except for their power supplies - which were notorious. CompuServe re-designed them completely, making them stone-cold reliable and some 60% more efficient. There are many tales of this still on the internet.

You'll want to consider memory, peripherals and storage before you begin. As Lou points out - original Core memory is not easy or cheap to re-create. Disk drives, MASSBUS and UNIBUS peripherals will also need to be thought out. Some will probably still be obtainable so you might want to begin here and do so before they aren't.

My personal experience at such "collections" is that once you get started, there is a phenomenon of "Critical Mass" which when achieved, causes the effort to avalanche and get easier thereafter. The important thing is to get going, think ahead, and do a little publicizing along the way. You never know what forces will come to your aid. Even so, I hope you have significant resources to call on for this endeavour, you'll need them.

My advice: Begin with air conditioned real-estate. About half an acre under roof would do nicely. Be sure to include 200 amp 3-phase power. Have a source for a large truck ready at hand - if you run into a system, you'll want to be able to grab it immediately. Keep a passport current.
 
Last edited:
My advice: Begin with air conditioned real-estate. About half an acre under roof would do nicely. Be sure to include 200 amp 3-phase power. Have a source for a large truck ready at hand - if you run into a system, you'll want to be able to grab it immediately. Keep a passport current.

And be prepared to pay more than Paul Allen will.
 
I don't want to discourage you, but you do realize what you're undertaking here? If you were able to identify the various parts, it's very likely to be cheaper to buy them individually on the curio market, than to build. Finding an original backplane or frame would indeed be fortunate. I think these would have been 30ga or 28ga Kynar Wire-Wrap [silver] and given the skyrocketing price of metals over the last 10 years, wouldn't expect any to survive unless part of a larger identifiable assembly - or as a relic in a misplaced crate. Maybe one could begin with a few phone calls to Sweden? [I too, heard about that destination]

As for conversation of "multilayer boards" this confuses me. I'm pretty sure the backplanes would be wire-wrapped - not PCBs. Individual boards ~ less certain, but probably double sided G10.

Do we have documentation of actual boards? I haven't looked - maybe later tonight. I thought I had images, but looking yesterday after your post - I was wrong. (?)

So you'd try to make a system faithful enough to accept genuine boards intermixed with those of your own creation? Am I understanding that correctly?

Picking a KA10 would put you in a pretty reliable family member, except for their power supplies - which were notorious. CompuServe re-designed them completely, making them stone-cold reliable and some 60% more efficient. There are many tales of this still on the internet.

You'll want to consider memory, peripherals and storage before you begin. As Lou points out - original Core memory is not easy or cheap to re-create. Disk drives, MASSBUS and UNIBUS peripherals will also need to be thought out. Some will probably still be obtainable so you might want to begin here and do so before they aren't.

My personal experience at such "collections" is that once you get started, there is a phenomenon of "Critical Mass" which when achieved, causes the effort to avalanche and get easier thereafter. The important thing is to get going, think ahead, and do a little publicizing along the way. You never know what forces will come to your aid. Even so, I hope you have significant resources to call on for this endeavour, you'll need them.

This is a modern replacement for the original KA-10 backplane which was wire wrapped. This would be a multi-layer board to accept the
DEC backplane modules, but would not be wirewrapped but a modern PC board. This is similar to what Vince is doing for the pdp-8
line. All I am planning to do is the drawings for the backplane and boards as a schematic. Nothing further as it would cost
a small fortune to build it. A board that big might not even be producable (at least without some errors for being so large). I plan
to use the existing schematics in the back of the maintenance manuals to redraw those as close to how they are now. No, there
won't be any updates to the original as I do not know of any source for patches/fixes for a KA-10 backplane. Testing it would be
a nightmare I think.

I am not worried about power supplies, memory, storage. Just the CPU. If the CPU existed I think the rest
could be cobbled together using exisiting stuff.

I know this is a huge project. It will be an on again off again thing. It may never get done, but what
I end up with will be saved somewhere that others may decide to continue it. If nothing else it will
help to document the KA-10 CPU.

The backplane is huge, as it has space for about 1600+ flip chips.
 
Well, that's what I'm talking about...

1) In scrutinizing those photos, one can clearly see major components of KL10s [that "s" denotes plural - not a model variant]

2) Although it's a fairly oblique angle, I'm certain this is a console bay from a KI10

dscn2941.jpg

3) However, I see only one cabinet in that collection that appears to be KA10 vintage. It can be seen in the upper right corner of this image, facing away from us offering us a view of it's "right" interior and profile. [note the annunciator panel [white] is all the way at the top of the cabinet, unlike succeeding generations]

dscn2938.jpg

Reference this image from the UQ Museum of IT collection:

23 - Sharon, Colin & Laurie monitoring a recoding (1024).jpg



Thanks for posting that link. That is a sizable collection, and although it looks lonely [and crowded] down there, they seem to be appreciated. Maybe now you see what I meant by needing "half an acre under roof"?
 
Last edited:
What, wait! What have I missed!

I was about to suggest that you start with a small machine, like the 8/s, to get the hang of it.

Vince is drawing the 8s currently, and has plans for the straight 8 later on. He already has most
of the M series modules drawn, along with many others. I decided to dive right in and do the
hardest. Vince will do what he can to help me get up to speed with Eagle (the CAD system I
will be using).
 
Drawing PDP-8 stuff

Drawing PDP-8 stuff

Vince is drawing the 8s currently, and has plans for the straight 8 later on. He already has most
of the M series modules drawn, along with many others. I decided to dive right in and do the
hardest. Vince will do what he can to help me get up to speed with Eagle (the CAD system I
will be using).

The 8/S is basically done. There are two major pieces to that. The backplane drawings (8/S and PT08 ), and the flip chips. I have finished drawing an 8/S and PT08 backplanes (used a substitute for the core plane, which I also drew) some time ago, and did the last of the flip-chips needed by the 8/S fairly recently. The PT08 uses some old school RTL logic for the TTY shift registers, though (chips -- I could hardly believe it!), so you'd need to locate those on eBay or wherever. Alternatively, you could design some modern equivalents with HC logic (which can handle the funky supply voltage) for the W706 and W707.

Currently I am slogging on the drawings for all the straight-8 flip chips. According to my Perl script, I have 14 more module types to draw to finish the flip chips for that. I figure to finish that early next month. (I also have the TC01 and TU55 flip chips done, but I haven't drawn backplanes for those yet.)

After I finish these flip chips, I had planned to revisit the Mxxx series and make sure I had done all the flip chips used in for the 8/L and 8/S, and probably the TC08 and TU56. I haven't counted how many of those are missing, yet. Later in the year I hope to back-fill with the backplane for the straight-8.

Tim's work has made me aware that the parts library for drawing backplanes is is behind the times. Currently the only modules that are there are the ones I've used in a backplane somewhere. There are a *lot* of flip chip types that now have schematics for the module itself, but no part in the library to add to a backplane. There's probably a week or so's work to draw symbols and connect pins to pads for all those modules.

Vince
 
_very_ cool. at first I thought it was Vince Briel that was making some sort of kit that would totally ruin me :)

Will you consider sharing your efforts. At some point I would like to build a PDP-8 (haven't decided which model) with modern TTL SMT components. It would help to have a mapped out system.

What documentation are you using as a base? I haven't found a wiring diagram for the 8/s backplane. Admittedly I haven't looked very hard.
 
_very_ cool. at first I thought it was Vince Briel that was making some sort of kit that would totally ruin me :)

Will you consider sharing your efforts. At some point I would like to build a PDP-8 (haven't decided which model) with modern TTL SMT components. It would help to have a mapped out system.

What documentation are you using as a base? I haven't found a wiring diagram for the 8/s backplane. Admittedly I haven't looked very hard.

My work on individual flip chips can be found at
http://www.so-much-stuff.com/pdp8/flipchip/flipchip.php
and the backplane stuff can be found at
http://www.so-much-stuff.com/pdp8/cad/cad.php

For the 8/S backplane, click on "The boards". Look for "./DEC/PDP8I", and the 8L and 8S are right after that, a little over halfway down. (It's a long list.) Those links, in turn point into the SVN repository.

If I remember correctly, the original source materials for this work were the maintenance manuals for the 8S on bitsavers. Basically what one does, is to re-draw the 8/S schematics, using the Eagle parts library "dec-r", in which I have created flip chips with footprints that look like a backplane slot. (There are also some guide parts, which allow you arrange your modules into connector blocks with appropriate mounting holes, etc.).

I actually think that one of the great things about doing this is "design capture". Once you've got the partslist and netlist, there is all kinds of stuff you can do, and often semi-automatically. Ideas I've explored include mapping to TTL and mapping to FPGA. Things I've thought about, but not tried, would be to fetch the flip chip schematics and blow up the backplane into a monster design using transistors and whatnot directly.

More reasonable would be to create flip chips about half the size of the originals, using SMT and header pins instead of funky edge connectors. Then a fairly trivial transform of the backplane (define a new package that is a header socket; use it instead of the H807 connector) and you're off. (Note that you haven't reduced power consumption, but you have reduced volume, so you may have to think a little about cooling.)

Vince
 
Wow, that is quite an effort you have put into this. I wonder how much you could shrink the flip chip form factor. Make it double sided and use the smaller SMT scales I bet you could get down to a fourth of the original size.

Either way, speaking of Peter Löthberg. A friend of mine got some pictures from Löthberg's Colossal Cave Computer Center and the KIKI computer, a dual processor KI10:

KIKI1.jpg

KIKI2.jpg

KIKI4.jpg

KIKI3.jpg


That is Peter on the last picture. And whether he has a KA10 or not, I'm told he has "one of each" :)
 
Wow, that is quite an effort you have put into this. I wonder how much you could shrink the flip chip form factor. Make it double sided and use the smaller SMT scales I bet you could get down to a fourth of the original size.

Thanks. Some folks do jigsaw puzzles...but I've been doing this :).

I'm not sure how small the SMT versions can(should) be made. You still need the traces, you still have to handle the current and dissipate the power. Another issue is the repairability and skills required to assemble the things. If you use the parts with very fine pitch, you'll need more tooling and/or more skill. I find I can work with the 0.05" stuff easily, but even the SMT resistors are pretty fiddly, and there are sure a boatload of resistors and diodes in the some of those old modules. (I suppose with stencils and an oven it wouldn't be so tedious as the loupe and the rework station.)

Half size should be quite doable. One third might be plausible, but I think 1/4 would be pushing it a little. I'm assuming the same level of integration -- one diode corresponds to one diode, etc. (If you just plop a CPLD down on every module, you'd get much smaller, but then why not just cram the whole thing into an FPGA.)

Also, using inexpensive 2x18 headers instead of gold fingers to keep the cost down wants to force one of the dimensions to approximately 2 inches.

I suppose something like the R210 would be an interesting test case. If that fits, probably the others will fit, too.

Vince
 
Wow, that is quite an effort you have put into this.
Agreed - absolutely Herculean!

That is Peter on the last picture. And whether he has a KA10 or not, I'm told he has "one of each" :smile:

Thanks for the additional photos. Any idea when these would have been taken?

I'm not sure how small the SMT versions can(should) be made.

I feel your pain in this kind of issue.

Miniaturizing the form factor has significant advantages... potential speed increases, cost reductions, reduced electrical noise, not to mention a more practical system footprint. The problem of higher thermal density for the overall system could become an issue, but I feel that's manageable.

However, all that comes at the cost of compatibility with authentic systems. Your work, while applicable to a point - could not serve as a source of replacements for genuine hardware. Nor could a system built to use such modules, benefit from surviving originals.

Like you said - "but then why not just cram the whole thing into an FPGA" - which has been done for some systems.

Then there's the "middle Road" - Putting a CPLD on an original form factor Flip-Chip has merits. It would be possible to make a "Universal" replacement for a range of modules, and simply vary the programming of the CPLD and parts installed to attain the desired function. This would reduce costs [mass production] potentially retaining functional compatibility with originals and be applicable to reduced form factors later.

For me personally, the idea of doing it so that the work can benefit original surviving systems would be the winning point. But then - I keep thinking primarily PDP-10 here.

Anyway - Keep up the great work.

P.S. - I know it's come up before here on the forums, but have we ever heard of a PDP-6 survivor?
 
Back
Top