About re-producing PDP-11/70

[alphaaxp]

I have always been very interested in the PDP-11 series, and recently I really want a PDP-11/70. I am wondering whether it is possible to re-produce a new machine, although it may not be possible to make all components exactly the same.
First, use verilog describe all the SSI components used in the machine
Then integrate each CPU/FPU/MEM board and the entire system according to the circuit in the engineering drawing, and run the simulation
Draw the PCB of the CPU/FPU/MEM board/backplane according to the original size and pin assignment, manufacture and solder each 74 series components, guarantee that these are exactly the same as the original design
Use modern power supplies, fans, and SSDs to emulate tape/disk and other external storage components
Custom Chassis/Front Panels
Integrate all the above components to complete the whole machine
Is such a re-produced machine valuable?

 

[tradde]

You do realize the 11/70 CPU was 8 full sized card (hex?) in that backplane? I can't imagine how much this would cost. Cool, Yes.

 

[Scrub]

You could also get yourself a PiDP-11 kit (uses a raspberry pi), or if you dont care about a front panel an 11/73 can be built relatively cheap using ebay parts.

 

[NeXT]

You do realize the 11/70 CPU was 8 full sized card (hex?) in that backplane? I can't imagine how much this would cost. Cool, Yes.

We had someone here many years ago who was dead-set on making a repro straight-8.
In comparison, fabbing 8 hex-width boards seems easy.

 

[AK6DN]

I have always been very interested in the PDP-11 series, and recently I really want a PDP-11/70. I am wondering whether it is possible to re-produce a new machine, although it may not be possible to make all components exactly the same.
First, use verilog describe all the SSI components used in the machine
Then integrate each CPU/FPU/MEM board and the entire system according to the circuit in the engineering drawing, and run the simulation
Draw the PCB of the CPU/FPU/MEM board/backplane according to the original size and pin assignment, manufacture and solder each 74 series components, guarantee that these are exactly the same as the original design
Use modern power supplies, fans, and SSDs to emulate tape/disk and other external storage components
Custom Chassis/Front Panels
Integrate all the above components to complete the whole machine
Is such a re-produced machine valuable?

Possible to do? Certainly. DEC did it 45 years ago.
Costly to do? Most certainly. In terms of manpower and cost of parts and assemblies.
Cost effective? Not likely. Doubtful you could ever sell enough systems to recoup the parts cost, much less pay the development cost.
The dozen or so CPU boards are the easy part. The backplane is a killer. It was a huge wire-wrapped assembly.
Much of the reason why DEC discontinued the 11/45, 11/70 (and follow-on 11/74) was due to the expense of producing the huge backplanes.

You are better off starting here: https://github.com/wfjm/w11 if you want to build a PDP-11/70 class CPU from logic (VHDL or Verilog) today.

Or if you are a real aficionado of the 11/70 take the printset and recode each board as a verilog module using 74xx building blocks and wiring.
Then build a verilog netlist of the backplane, combine all the modules together, and poof, you have a simulate-able synthesize-able PDP-11/70 CPU.

Then all you need to do is implement all the RH disk and tape controllers, and of course various useful UNIBUS boards like communications etc.

Others in the past (QED and MENTEC IIRC) both did part of the problem, producing equivalent boards using FPGA technology.
But they kept the same 11/70 backplane and peripherals.

Your original idea could keep a whole engineering team of 8 to 10 busy for a few years, full time.
Of course if you are just Elon posting under cover to remain anonymous and money is no object then go for it.

 

[vrs42]

Or if you are a real aficionado of the 11/70 take the printset and recode each board as a verilog module using 74xx building blocks and wiring.
Then build a verilog netlist of the backplane, combine all the modules together, and poof, you have a simulate-able synthesize-able PDP-11/70 CPU.

Kyle and I did this for the 8/I a couple years ago. We took the 8/I Eagle CAD drawings which I had done, and exported the netlist as a set of module interconnections, then Kyle modeled the modules, and presto, an 8/I Verilog model. He went on to debug it in simulationa and in an FPGA development board. He also took my TC08 DECtape controller drawings, wrote some Verilog to simulate the additional modules used there and the TU56 drives, and had that all integrated and working. I fixed the problems he found, which generally traced to a few transcription errors when I did the Eagle drawings. (So now my 8/I and TC08 drawings are more accuracte.)

He's been after me to redraw the PDP-12 drawings, but that hasn't reached the top of the stack.

A similar approach should be doable for a PDP-11, though if there are were a lot of programmable parts used (as, for instance in some of the boards in an 8/A), it can be a pain to reconstruct the logic that was hidden in there.

These days, the expensive bits are the interconnect (backplane) and the front panel. Logic and PCBs are cheap.

Vince

 

[Al Kossow]

Talk to Dale Luck (he's on here)
He has 11/70 cpus without boards because they had been retrofitted with replacements
At one point I was going to build one up, and bought some board sets.
Who knows, maybe he'd sell you something in the $5000 range, which would still be
a hell of a lot cheaper than attempting to build one from scratch.
I suspect, though, that you don't even have the money to do that.
The other insane thing would be period-correct memory.
Boxes and boxes of 10 1/2" chassis to get all the memory
you could address.

There was a reason companies existed to retrofit all that power-hungry STLL and 4K DRAM memory.

 

[cchhrriiss11]

I have always been very interested in the PDP-11 series, and recently I really want a PDP-11/70. I am wondering whether it is possible to re-produce a new machine, although it may not be possible to make all components exactly the same.
First, use verilog describe all the SSI components used in the machine
Then integrate each CPU/FPU/MEM board and the entire system according to the circuit in the engineering drawing, and run the simulation
Draw the PCB of the CPU/FPU/MEM board/backplane according to the original size and pin assignment, manufacture and solder each 74 series components, guarantee that these are exactly the same as the original design
Use modern power supplies, fans, and SSDs to emulate tape/disk and other external storage components
Custom Chassis/Front Panels
Integrate all the above components to complete the whole machine
Is such a re-produced machine valuable?

Well, it's obvious that you really like the PDP-11/70... honestly, who doesn't? There's a bunch of us here that have been looking for a real 11/70, or perhaps already have one (or two) of their own! What you're proposing will cost well into the six-figure range, if not more. You'll have much better luck finding a turn-key PDP-11/70 for a fraction of that cost, in-fact, there are people selling them out there (for very reasonable prices) - if you know where to look.

As mentioned in an earlier comment, other people/companies have done just this -- Quickware Engineering is one of them, they did combo CPU/Memory replacements for the 11/70, the 11/44, and other machines. The 11/44 replacement QED-95 for example is only 3 boards, plus the console interface - much less than a "real" 11/44. I'd imagine the 11/70 version is similar. I just sold a set to another member here... I just need to get my butt in gear and ship it!.

If you're that ambitious, I'd suggest finding a "rough" 11/70 (or 11/60 for something really interesting) and restore it back to like-new condition. It can be done (and -has- been done) -- I've got a few PDP-11's now that I'm working on getting to "like new" condition and my 11/44 is close, the DECLAB11/03, 11/04 and 11/34A isn't far away. On the other hand, I've got an 11/60 in very rough condition that sat out in the elements for quite some time. I'm dealing with a corroded chassis, rusty boards, and all sorts of issues - but given how rare the 11/60 is, it's worth some effort to restore. This is the type of project that will (most likely) take many years of effort to get the machine not only running again, but looking in presentable condition.

-Chris

 

[pbirkel@gmail.com]

You do realize the 11/70 CPU was 8 full sized card (hex?) in that backplane? I can't imagine how much this would cost. Cool, Yes.

16 Modules (one quad; remainder hex) for the KB11-B/C CPU and then 4 modules (hex) for the FP11-B/C FPP. Add more modules (7) per MASSBUS controller ...

 

[jplr]

An interesting idea!

Maybe it could be a single board computer based on a variant of the M7264 LSI CPU board with beautiful PiDP-11 keyboard and LEDs.

It could combine a simulated CPU (it could be done with cheap controllers), memory (some had 4KW, just extend it to 4M) , just add a serial line interface, a PIA for interfacing with the board from the PiDP-11, and indeed a QBUS interface. Let people handwrap the QBUS if they want (I did it, it's not a so hard task, certainly requiring much less dedication than most projects done by forum's members).

The LSI-11 CPU was made with ICs from MCP-1600 family which ran at 3MHz.

MCP-1600 - Wikipedia

en.wikipedia.org

A modest MSP430 has a similar programming model and runs at 8MHz and for some models at 25MHz.

TI MSP430 - Wikipedia

en.wikipedia.org

or it could use Russian CPUs :

1801 series CPU - Wikipedia

en.wikipedia.org

A pseudo "Jaws" with MSP430 maybe would look like this ;-)

 

[pbirkel@gmail.com]

Just note the XOR between a DEC (or most any other) LSI processor and a bllnkenlights front panel. Unless you're willing to write special driver software and modify the OS/Applications to accommodate it. The only exception that I know of for DEC is the SpareTimeGizmos SBC6120, and it accomplishes that feat courtesy of a design feature of the HD6120 that provides a transparent 32Kw of memory to use for the necessary peripheral emulations (including the front panel). It's a really nice approach/solution!

 

[jplr]

> Unless you're willing to write special driver software
Yes you are right as I wrote about a PIA.

In a different proposition:
Wouldn't it be possible to interface the PiDP-11 panel to the system console (thanks to some controller) and translate each manual stroke in ODT commands and light up the LEDs accordingly?

 

[pbirkel@gmail.com]

Well, sure, for some commands. But once you hit RUN your panel will be divorced from knowledge of, e.g., register contents. Only bus signals will be accessible.

 

[alphaaxp]

Not long ago, a friend told me that he saw 2 PDP-11/70s from an electronic recycler, but they had been dismantled, in terrible condition, and the asking price was $40000. I hesitated for a day, and when I asked again, there was none, this is the closest time to this machine so far.
I don't want to wait for a machine to show up in front of me and sell it at a reasonable price. I hope that these machines, built with STTL logic, can be bought whenever he wants.
Theoretically speaking, PCB or 74 series chips are not a problem, but non-standard components are difficult to reproduce as they are, such as the front panel. After all, it is too expensive to re-mold. I don’t think much about it. Is there any other way, such as CNC? It can be produced in small quantities at a slightly cheaper price, and this part will be resolved later.
I also thought about the time, whether the cost is worth it, I don't know, people are always tossing something, so be it
Talk is cheap. At present, I have counted the chips used in each board of pdp-11/70. All SSI components have been described in verilog. And the first board M8130 is being integrated. I hope that the description of all boards can be completed by the beginning of next year, I will build a github link later.
I have some ideas, I don't know if it's right, I also want to ask
1. I understand that the original backplane is the edge connector soldered on the PCB. The PCB itself has no winding, and the winding is done manually with wires later. If this is the case, when design re-produced backplane, the winding is done on the PCB, which can avoid manual winding and instability. Can the interconnection lines on the backplane also use modern shelf products?
If the idea is feasible, then the re-product board and backplane are compatible with the original, if I'm not mistaken.
2. PDP-11/74 is the most famous never. It is basically based on the modification of KB11-C. Whether it is possible to obtain KB11-Cm/E from the verilog model output above through the relevant documents on bitsaver, I see the comments It is said that the reason for the cancellation of 74 is that the interconnection of CPU/MEM is not stable enough, so can we solve this problem by redoing the backplane and using modern PCB wiring? According to the number of 70 boards, CPU+FPU=16+4, The 44-slots backplane has the opportunity to put two KB-11s, as a double-layer plane, and memory and cpu/fpu can be placed on the same backplane. Two such double-layer backplanes can achieve 74 4 cores back to back.

 

[pbirkel@gmail.com]

I suspect that you'll need a 4-layer PCB for your backplane; many of the wire-wrap connections are twisted pairs, not simple point-to-point wires as seen in most other DEC backplanes ("system units"). Routing may be an issue as well.

 

[Radix]

On a similar but different idea.. I have to say that I like running the PDP11 software and O/S on "real" hardware - rather than SIMH or other emulators, but of course, real hardware is getting rare - especially backplanes - and is usually quite bulky by modern standards...

What I think would be an interesting group project would be to "reimagine" Q-Bus, for example - keeping the original timing and signal specs but on a smaller modern and obtainable connector - say fitting a 3U rack chassis, so it could be as small or large as required...

Then, have probably MCU based replacement boards for CPU/MEM (most likely on the same board) DL/DZ/DH lines - but keeping the same BERG connectors and wiring as DEC used, and various disk emulations, and other IO, so rather than the Q-Bone all in one board, we have a modular and modern DEC compatible hardware platform that is extensible like the real systems but practical to build from scratch.

Critically - keeping the DEC bus signals and protocol give us a working and known standard from the start that we know works, so we have a baseline to compare the new re-designed boards to.

Also - a hybrid backplane would be possible to use both real and new boards - most likely using a Q-Bus PCB to join to the new backplane and bus grant cards as required.

So - Mini-PDP11 anyone?

Robin

 

[AK6DN]

I suspect that you'll need a 4-layer PCB for your backplane; many of the wire-wrap connections are twisted pairs, not simple point-to-point wires as seen in most other DEC backplanes ("system units"). Routing may be an issue as well.

Just 4 layers? Highly doubtful. I suspect more in the 16 to 24 layer range given the complexity of the interconnect.
The wiring pattern is not regular like a QBUS or UNIBUS backplane, but mostly random point to point diagonal wiring.
Mostly signal layers, but there will be ground plane layers required to help mitigate crosstalk (which wirewrap handles remarkably well) between parallel traces.
I would assume you would want to maintain the exact same pin/slot connectivity as the current backplane.

And you would probably want to find a pressfit style connector rather than having to hand (or wave solder yikes) each connector pin.

In any event, not going to be a simple engineering task nor simple and inexpensive to produce.

 

[pbirkel@gmail.com]

Yes, not much regularity in the FFP/CPU slots to the left. Even the MASSBUS controllers slots would be a bit challenging.

 

[pbirkel@gmail.com]

On a similar but different idea.. I have to say that I like running the PDP11 software and O/S on "real" hardware - rather than SIMH or other emulators, but of course, real hardware is getting rare - especially backplanes - and is usually quite bulky by modern standards...

...

So - Mini-PDP11 anyone?

Robin

Interesting concept, although IMO Qbus backplanes aren't all that R@RE on eBay. Occasionally overpriced, yes. Only modestly bulky "by modern standards".

With your proposed MCU-based replacement boards you're still emulating using SIMH (or equivalent) on "real" hardware. Closer to "real" would be the various VHDL-based projects; for example: https://pdp2011.sytse.net/wordpress/pdp-11/ Substitute those for your MCUs and you'd be running on "real" hardware, albeit using a later-generation implementation approach :->.

That said, I'd invert the design approach by first defining a Qbus extension using a convenient connector, attach that to an existing Qbus system, and then prototype modules on that extension that can then be tested against an operational Qbus environment. Still, all-said-and-done the Qbone (http://retrocmp.com/projects/qbone) has pretty much everything going for it except your desire to emulate HW using HW-specific I/O cabling.

IMO the only Qbus modules that don't show up on eBay "reasonably" frequently at "reasonable" prices are SCSI controllers. Which I/O module functions are you looking for where DEC-specific I/O connectors are important? In general the problem is that the controllers survive (pulled by scrappers), not the peripheral equipment ... so IMO emulating peripheral equipment is the best bang-for-buck at the present time.

Personally, I _really_ like the http://retrocmp.com/projects/lsibox for shear beauty and grace :->. It gets the trade-offs between legacy HW and I/O emulation just right.

 

[Radix]

Interesting concept, although IMO Qbus backplanes aren't all that R@RE on eBay. Occasionally overpriced, yes. Only modestly bulky "by modern standards".

With your proposed MCU-based replacement boards you're still emulating using SIMH (or equivalent) on "real" hardware. Closer to "real" would be the various VHDL-based projects; for example: https://pdp2011.sytse.net/wordpress/pdp-11/ Substitute those for your MCUs and you'd be running on "real" hardware, albeit using a later-generation implementation approach :->.

That said, I'd invert the design approach by first defining a Qbus extension using a convenient connector, attach that to an existing Qbus system, and then prototype modules on that extension that can then be tested against an operational Qbus environment. Still, all-said-and-done the Qbone (http://retrocmp.com/projects/qbone) has pretty much everything going for it except your desire to emulate HW using HW-specific I/O cabling.

IMO the only Qbus modules that don't show up on eBay "reasonably" frequently at "reasonable" prices are SCSI controllers. Which I/O module functions are you looking for where DEC-specific I/O connectors are important? In general the problem is that the controllers survive (pulled by scrappers), not the peripheral equipment ... so IMO emulating peripheral equipment is the best bang-for-buck at the present time.

Personally, I _really_ like the http://retrocmp.com/projects/lsibox for shear beauty and grace :->. It gets the trade-offs between legacy HW and I/O emulation just right.

Hi, I get your point - but I have found backplanes in particular to be very rare and expensive - and a lot of people don't understand the different versions, let alone the need for proper cooling...

A modern and smaller backplane arrangement with functional boards that emulate real DEC boards - like a DLV11-J or DHV-11 would allow people to have the spec they want - more serial ports or different I/O options to suit different O/S while each is relatively cheap and modular.

Keeping the same board connectors for serial cards, for example saves yet another connector standard and would work with the real (if rare) connectors - but alternate chassis kits can be made and they would suit a real PDP - or this idea, so we get more use out of replacement parts.

It would not be using simh as such - but a standalone emulation of a CPU - maybe a low end 11/03 emulation version and a 11/73 version

Apart from combining CPU, memory and maybe bootstrap - I would avoid multifunction boards, making the boards simpler and the system more modular - like the real PDP11 systems with dual width CPU and separate SLU

It lends itself to people contributing different types of boards to the design - and by retaining Q-Bus spec and standards, original software remains compatible as well.

The deliberate LACK of massive integration into a single MCU and modular design also makes it an excellent example of computer design and architecture, as the PDP11 itself was and continues to be.

Simple board configuration options with switches would make it suitably old-school and easy to configure without an app or similar.

The only change to the Bus I would suggest is raising the termination impedance - we don't need a 16' bus length! - then the likes of an ESP32 can directly drive the bus signals in open-drain output/input mode - they are 5V tolerant - greatly simplifying the bus interface...