• Please review our updated Terms and Rules here

About re-producing PDP-11/70

theor

New Member
Joined
Aug 4, 2020
Messages
9
Location
Lyon, France
alphaaxp, you should get in touch with Raphael Jaquot; he's recreating the 11/70 boards from the schematics and documentation. I'm not sure if he's making good progress lately, but I've seen at least one board with my own eyes.


The CPU is what you really want. The missing devices, including memory, can either be emulated with a Unibone, or you can find some compatibles boards on eBay (get at least a DL11). Another option for memory would be to re-implement it with modern ICs such as static RAM with a bit of glue logic, provided you get the timing right. You don't need the front panel; a Unibone can emulate an M9301.

As someone mentioned, the backplane will be more annoying. Best is to purchase new connectors from Douglas Electronics, and make a four layers PCB from the backplane schematics. Don't bother wrapping, DEC only did so because of the lack of four layers back in the days. The BA11-K box will also be annoying to re-create but I assume any 19" rack could be adapted.

IMG_3387.jpeg
 
Last edited:

theor

New Member
Joined
Aug 4, 2020
Messages
9
Location
Lyon, France
By the way, I think creating a brand new PDP-11 clone for hobbyists, similar to what the Xi 8088 and its open-hardware boards did for the IBM PC/XT, is a sound idea.
It is almost impossible to get one now and the PiDP-11 just doesn't cut it. There is a real demand for an actual working replica especially from the computer hobbyists born in the 80s-90s who wish they had known this era and dream of owning a mini.

A modern replica using CMOS components and a smaller switching supply would also be much more reliable and power-efficient than an old clunker sold at an obscene price. Cosmetically, the inside of the modern replica could be reworked to fit a modern 19" rack and the blinkenlight panel could also be adapted.

This would probably involve a small 4 figures investment, but it would be cost-effective. All the expenditures would be directed towards acquiring actual modern, working hardware instead of filling the pockets of an unscrupulous scrapper for something that will then also require a lot of investment.

However I think the 11/70 sets the bar too high, and probably is the worst model to pick for a replica because of its complexity. Some more realistic candidates could be:

- The 11/20, as the original PDP-11. Historically significant, probably the most sought after by hobbyists. Straightforward CPU which doesn't use microcode, uses very generic TTLs with little integration, can address 28 kilowords and has roughly the power of a /40, /34 or /60. It's also the easiest to debug, and it can be emulated at cycle accuracy on a Unibone for troubleshooting (https://github.com/aap/pdp11)
- The 11/05 and 11/04 would be fairly realistic targets too. The 04 uses a single board CPU but a lot of combinational logic is integrated with bipolar ROMs for which we have no dumps. Also, those models are less significant than the 20 and 70.

Some PDP-11 clones have already been made by various individuals. Here is a homebuilt 11/05:

IMG_03191-1024x764.jpg



However, this is a substantial undertaking and would be best done as a community project, with contributors each working on a part.

Would such a project interest the community?
 

theor

New Member
Joined
Aug 4, 2020
Messages
9
Location
Lyon, France
This has already been done with the J-11, F-11 and their soviet counterparts.

The way I see it, using a VLSI chip defeats the point of an authentic replica, and won't solve the problem of long term supply for the next generations of hobbyists. We may as well use an FPGA then, or SIMH + front panel à la PiDP-11.

Part of the joy of owning a mini, especially a PDP-11, instead of a micro is having a CPU made of discrete TTL chips plugged into a Unibus backplane. The low scale integration is a definite part of the "old school" experience. This is like owning the replica of an actual classic car with all the quirks and smells, versus having one playable in a video game. Sure, the latter is more convenient, but it definitely isn't the same experience for an enthusiast.
 

cchhrriiss11

Experienced Member
Joined
Oct 19, 2019
Messages
261
Location
Duncan, BC
This has already been done with the J-11, F-11 and their soviet counterparts.

The way I see it, using a VLSI chip defeats the point of an authentic replica, and won't solve the problem of long term supply for the next generations of hobbyists. We may as well use an FPGA then, or SIMH + front panel à la PiDP-11.

Part of the joy of owning a mini, especially a PDP-11, instead of a micro is having a CPU made of discrete TTL chips plugged into a Unibus backplane. The low scale integration is a definite part of the "old school" experience. This is like owning the replica of an actual classic car with all the quirks and smells, versus having one playable in a video game. Sure, the latter is more convenient, but it definitely isn't the same experience for an enthusiast.
I couldn't agree more!

-Chris
 

alphaaxp

Member
Joined
Jul 14, 2023
Messages
11
By the way, I think creating a brand new PDP-11 clone for hobbyists, similar to what the Xi 8088 and its open-hardware boards did for the IBM PC/XT, is a sound idea.
It is almost impossible to get one now and the PiDP-11 just doesn't cut it. There is a real demand for an actual working replica especially from the computer hobbyists born in the 80s-90s who wish they had known this era and dream of owning a mini.

A modern replica using CMOS components and a smaller switching supply would also be much more reliable and power-efficient than an old clunker sold at an obscene price. Cosmetically, the inside of the modern replica could be reworked to fit a modern 19" rack and the blinkenlight panel could also be adapted.

This would probably involve a small 4 figures investment, but it would be cost-effective. All the expenditures would be directed towards acquiring actual modern, working hardware instead of filling the pockets of an unscrupulous scrapper for something that will then also require a lot of investment.

However I think the 11/70 sets the bar too high, and probably is the worst model to pick for a replica because of its complexity. Some more realistic candidates could be:

- The 11/20, as the original PDP-11. Historically significant, probably the most sought after by hobbyists. Straightforward CPU which doesn't use microcode, uses very generic TTLs with little integration, can address 28 kilowords and has roughly the power of a /40, /34 or /60. It's also the easiest to debug, and it can be emulated at cycle accuracy on a Unibone for troubleshooting (https://github.com/aap/pdp11)
- The 11/05 and 11/04 would be fairly realistic targets too. The 04 uses a single board CPU but a lot of combinational logic is integrated with bipolar ROMs for which we have no dumps. Also, those models are less significant than the 20 and 70.

Some PDP-11 clones have already been made by various individuals. Here is a homebuilt 11/05:

IMG_03191-1024x764.jpg



However, this is a substantial undertaking and would be best done as a community project, with contributors each working on a part.

Would such a project interest the community?
I found him, maybe he hasn't been online recently

KB11-C, FP11-C, MK11, even RH70 and SPC are all what I want. I feel that there is no way to replic the tape/disk/etc outside, can only simulate it with a modern sd/cf card.

And I think the blinkenlight front panel and the use of discrete TTL are also necessary. Using VLSI always feels a bit lacking in taste.

Investigated the local PCB manufacturer on the weekend, the backplane of PDP-11/70, 5 pieces of 58x42 PCB with 4 layers cost $600, 16 layers cost $1200, so it seems that the price is quite good.

The cmos version of the 74 series components basically needs to be SMD, which is more energy-efficient. I also wonder if it is better to use the S/AS/F series of DIP, which can solder the base to the PCB, and hobbyists can even insert components into the board and build the entire mini from discrete kits.

PDP-11/70 is a personal ideal. I also believe that people always tend to fill up the configuration. Can everyone really agree on another model?
 

alphaaxp

Member
Joined
Jul 14, 2023
Messages
11
J-11、F-11 及其苏联同类机型已经做到了这一点。

在我看来,使用 VLSI 芯片违背了正品复制品的意义,并且无法解决下一代爱好者的长期供应问题。那么我们不妨使用 FPGA,或者 SIMH + 前面板 à PiDP-11。

拥有迷你电脑(尤其是 PDP-11)而不是微型电脑的部分乐趣在于将由分立 TTL 芯片制成的 CPU 插入 Unibus 背板。低规模整合是“老派”经验的一个明确组成部分。这就像拥有一辆真正的经典汽车的复制品,具有所有的怪癖和气味,而不是在电子游戏中玩。当然,后者更方便,但对于爱好者来说绝对不是同样的体验。

From the perspective of replicas of actual work and long-term availability of enthusiasts, another feasible method is FPGA.

Recently I made a PDP-11 instruction list. Finally, I want to complain that CIS has really become popular ( must)? It is basically not suitable for modern implementations of splitting instructions into uOPs
 

Attachments

  • PDP11.zip
    15.2 KB · Views: 5
Last edited:

AK6DN

Veteran Member
Joined
Aug 23, 2010
Messages
1,279
Location
Silicon Valley USA
From the perspective of replicas of actual work and long-term availability of enthusiasts, another feasible method is FPGA.

Recently I made a PDP-11 instruction list. Finally, I want to complain that CIS has really become popular ( must)? It is basically not suitable for modern implementations of splitting instructions into uOPs

You mean an FPGA PDP-11 implementation like this ... https://github.com/wfjm/w11 ? Or this ... https://pdp2011.sytse.net/wordpress/ ?

CIS for the PDP-11 has become popular? Really? What software are you using? AFAIK COBOL-11 was the only language system to ever use it.
ISTR that some versions of RSTS would use the block MOVC instructions, if available, to move memory around, but that was about it.
I have a CIS accelerator in my 11/44 and would love to find some real software to actually use it. So far all I have is the XXDP diagnostics LOL.
 
Last edited:

alphaaxp

Member
Joined
Jul 14, 2023
Messages
11
Sorry, I may not have made it clear. My other idea is to make a PDP-11 that can execute out of order, which is a bit similar to P6 for x86.
For me, this design may be a little less difficult than replica PDP-11/70. Writing RTL is the same as my job. The basic instructions and FP11 are relatively conventional, but CIS is really very different.
I haven't experienced the era of PDP-11, so what's the problem with PDP-11 without CIS extension, in various aspects
 

DDS

Veteran Member
Joined
Sep 5, 2013
Messages
644

"About re-producing PDP-11/70"​


Get two PDP-11/70s.

Put them close together.

Put on some mood music & dim the lights.

============

Years ago "Byte" did one of their first issues focused on AI. A few issues later some of the readers' comments showed up in the letters column. One that tickled me a bit was a guy who wrote that he and his wife had found a fun way to create Artificial Intelligence. But just about the time they thought they had debugged the programming the AI would move away from home.
 

AK6DN

Veteran Member
Joined
Aug 23, 2010
Messages
1,279
Location
Silicon Valley USA
As someone mentioned, the backplane will be more annoying. Best is to purchase new connectors from Douglas Electronics, and make a four layers PCB from the backplane schematics. Don't bother wrapping, DEC only did so because of the lack of four layers back in the days. The BA11-K box will also be annoying to re-create but I assume any 19" rack could be adapted.

Do a full 11/70 backplane in only four layers and no wire wrap? I don't think so. Have you ever seen a real 11/70 backplane and looked at the wiring pattern?
It is not bussed like a QBUS backplane. It is random point to point connectivity. Easy backpanel connection on each of the boards meant the backplane had to undo the heavy lifting.
Look at the picture in post # 18.
 

AK6DN

Veteran Member
Joined
Aug 23, 2010
Messages
1,279
Location
Silicon Valley USA
Sorry, I may not have made it clear. My other idea is to make a PDP-11 that can execute out of order, which is a bit similar to P6 for x86.
For me, this design may be a little less difficult than replica PDP-11/70. Writing RTL is the same as my job. The basic instructions and FP11 are relatively conventional, but CIS is really very different.
I haven't experienced the era of PDP-11, so what's the problem with PDP-11 without CIS extension, in various aspects

Doing a fully pipelined out-of-order issue PDP-11 design in verilog suitable for an FPGA implementation would be an interesting project.
Forget CIS-11 (and maybe even FP-11) to start. And forget some of the more esoteric base instructions (like MARK LOL).
Until the 11/44 (and the never released 11/74) CIS options appeared CIS did not exist in any form. It was an optional add on to the later 11/23 chip series.
But CIS was too much too late to make a difference to the PDP-11.
 

cchhrriiss11

Experienced Member
Joined
Oct 19, 2019
Messages
261
Location
Duncan, BC
Doing a fully pipelined out-of-order issue PDP-11 design in verilog suitable for an FPGA implementation would be an interesting project.
Forget CIS-11 (and maybe even FP-11) to start. And forget some of the more esoteric base instructions (like MARK LOL).
Until the 11/44 (and the never released 11/74) CIS options appeared CIS did not exist in any form. It was an optional add on to the later 11/23 chip series.
But CIS was too much too late to make a difference to the PDP-11.
Don't forget about the KD11-F (LSI-11) had a CIS option as well...

-Chris
 

alphaaxp

Member
Joined
Jul 14, 2023
Messages
11
blackplane_slot.png
There are a few questions
1. What are the FPT maint and CP maint functions of slot 1, and what signals do we need to leave on the backplane?
2. I see that DL11-W has the functions of DL11-A and KW11-L. Replacing DL11-A with DL11-W, slot0 can be used without inserting KW11-L?
3. Is the A/B of the SPC slot functional? Is it only SPC or Modified UNIBUS Device on the backplane
 

alphaaxp

Member
Joined
Jul 14, 2023
Messages
11
There are also some compatibility issues.
The 5411086 module of the power supply generates LTC-L, which results in a line interrupt compared to KW11-L.
Modern power supplies do not exhibit such behavior at all, and it feels like they need to be hacked
And are+15v/-15v really necessary?
 

pbirkel@gmail.com

Veteran Member
Joined
Apr 7, 2013
Messages
1,081
Location
Silver Spring, MD, USA
1. What are the FPT maint and CP maint functions of slot 1, and what signals do we need to leave on the backplane?
See documentation for the KM11 Maintenance modules. Schematics in KM11_Maintenance_Panel_May70.pdf PDP-11/70 documentation excerpt for their use attached.
3. Is the A/B of the SPC slot functional? Is it only SPC or Modified UNIBUS Device on the backplane
Yes, although it's typically (probably only) third-party vendors that will offer modules that use it exclusively. Suggest studying PDP11_BusHandbook1979.pdf.
 

Attachments

  • PDP-11_70 How to Use KM11 Maintenance Cards W131+W133.pdf
    1 MB · Views: 5

pbirkel@gmail.com

Veteran Member
Joined
Apr 7, 2013
Messages
1,081
Location
Silver Spring, MD, USA
There are also some compatibility issues.
The 5411086 module of the power supply generates LTC-L, which results in a line interrupt compared to KW11-L.
Modern power supplies do not exhibit such behavior at all, and it feels like they need to be hacked
Yes. This problem has been solved in various ways; this days I think folks just use a small MCU to generate signals. Here's a Qbus example using an Arduino: https://avitech.com.au/?page_id=1657
And are+15v/-15v really necessary?
Typically used by communication modules with RS-232 interfaces and analog modules with ADC/DAC circuitry and/or op amps. Sometimes CPU timing generators have dependencies. So whether you can ignore those rails depends on what peripherals you intend to support (or your CPU requires).
 

jonk

New Member
Joined
Nov 21, 2012
Messages
3
Location
Oregon
Sorry, I may not have made it clear. My other idea is to make a PDP-11 that can execute out of order, which is a bit similar to P6 for x86.

I worked at Intel on the P II and the BX chipset. Are you thinking of creating something similar to the instruction decoder that translated CISC into RISC (ROB aka 're-order buffer') with a "retire unit" used to ensure in-order behavior at the end?
 
Top