DIY Q22 Backplane

[Steve Toner]

PDP-11 Qbus cards are plentiful and affordable, but you still need a backplane, power supply (that supplies a Line Time Clock signal) and enclosure which are somewhat harder to come by. I picked up a chunk of an old DEC backplane from the free table at VCF West. Don't know what it originally came from, but it had wirewrap pins and a PC board interconnecting them in a non-Qbus fashion. By cutting a few traces and wrapping some new wires, I was able to turn it into a 4-slot Q22 backplane:



Note that +5V (red wire) and GND (black wires) buses were already provided. The yellow wire is +12V, and the little busboard is for front panel connection. The pin marked with a white heat shrink is for BPOK H, which DEC did not put on the ribbon cable.


So now I've put together a minimal 11/23 system with these boards:

(L-R: M8186 with KEF11-AA floating point chip, M8059 RAM, RLV12 Emulator designed by Peter Schranz, M8043 serial card). The RLV12 emulator provides a virtual boot ROM. Otherwise, a crippled (RAM disabled) M8047 might have replaced the M8043...

I still need to put it in a case with a cooling fan, but for now I'm just testing it all splayed out on the workbench:

The blue board with the buttons and LEDs is the front panel - same as I used in my 11/53 build. It uses a PIC16F1615 to generate the power sequencing signals. LTC is generated with an optoisolator and can be seen enshrouded in heat shrink tubing to the upper right of the front panel board.

Running RT-11 (XM monitor):

 

[hunterjwizzard]

That's pretty cool!

 

[Radix]

Oh Nice work! - I'm surprised the FPF11 doesn't show as "Floating Point Microcode" rather than FP11 though - Out of interest, where did you source the RXV12 emulator? - I have seen someone doing the RX02 end of the cable but interested in getting a set together - options are always good! - I take it the 40 way interface could connect to a real RX02 drive - or vice versa?

Most of the SCSI controllers also have a minimal (self booting) bootstrap, so you could get bigger disks online with a ZuluSCSI, which works well

Robin

Dump of resorc.sav blocks 6/7

 

[AndrewZ]

This is brilliant! I have most of the parts to do the same thing, although I still need a PSU and ROM card. I'll just mention that the PDP-11/23 is mostly software compatible with the DEC PRO 350 series.

 

[Steve Toner]

Out of interest, where did you source the RXV12 emulator?

Most of the SCSI controllers also have a minimal (self booting) bootstrap, so you could get bigger disks online with a ZuluSCSI, which works well

It's an RLV12 emulator, not RXV12. Emulates RL02 drives. Info is here. I downloaded the gerber files from the download page and sent them to JLCPCB to get the boards.

And yes, I could do SCSI but I don't have a spare SCSI controller and they are expensive. Using boards I have on hand at for now. I do have a couple of RQDX3 controllers I could use as well, but at this time no drive to attach... I also have a spare M8192 if I want to upgrade the CPU at some point

 

[czunit]

This is brilliant! I have most of the parts to do the same thing, although I still need a PSU and ROM card. I'll just mention that the PDP-11/23 is mostly software compatible with the DEC PRO 350 series.

Um................. Not as such. While the Pro/350 does have a 11/23 type CPU (and yes you can put a FPF11 on it if you're nuts) the rest of the system is not really very Q-Bus-sy. CTI bus handles DMA very differently, the cards autoconfigure, etc. So P/OS is not going to run on your Q-Bus 11/23.

I think of an 11/23 as a good solid pdp11 CPU. Similar to the 11/34, slower than an 11/60, bit better than the 11/35, missing the split I/D of the 11/44 and the dual register sets of the 11/70.

Oddly enough I think you can switch into Supervisor mode, just no split I/D.

 

[AndrewZ]

Um................. Not as such. While the Pro/350 does have a 11/23 type CPU (and yes you can put a FPF11 on it if you're nuts) the rest of the system is not really very Q-Bus-sy. CTI bus handles DMA very differently, the cards autoconfigure, etc. So P/OS is not going to run on your Q-Bus 11/23.

I think of an 11/23 as a good solid pdp11 CPU. Similar to the 11/34, slower than an 11/60, bit better than the 11/35, missing the split I/D of the 11/44 and the dual register sets of the 11/70.

Oddly enough I think you can switch into Supervisor mode, just no split I/D.

I talked to a guy who ported BSD 2.9 from a 11/23 to the Pro. I think BSD is set up to auto configure at boot. But yes, Q-bus and CTI are different by design.

 

[Steve Toner]

Mounted it in a Hammond 1453C case. It's a little tight but everything fits. I'm not sure about the fan - there's no direct airflow across the top card (CPU). I'm hoping the airflow is chaotic enough in there that it'll be OK, but I'll have to monitor it and may need to install a baffle of some sort...

The case comes apart in two halves, and all connections between the two halves are pluggable.

Bottom half of Hammond case contains power supply, front panel & RLV12 breakout board, power inlet and (2) RS-232 ports:



Top half of Hammond case holds backplane and cooling fan:


The backplane is attached to the case with screws. I drilled & tapped some holes in that steel piece that runs along the top & bottom of the backplane. It's only about .100" thick, so I used metric screws with a .5mm thread pitch to give enough threads that I felt comfortable with it...

Front panel:



Rear panel:



I had to do all the drilling and cutouts by hand, as the U-shape of the Hammond case halves prevented me from using my (small) drill press and mini mill.
I considered painting the case a DECish blue, but I left it the factory color...

I think I'm going to call it the "Lunchbox 11"

 

[Radix]

Regarding the cooling - it really should have good forced cooling if you're precious about your boards... The rule of thumb is for every 10C warmer a device runs, its failure rate doubles - or life expectancy halves... If you want this to last longer, you know what to do!

Robin

 

[czunit]

That is really nice, and oddly enough I have not seen much heat related failures on either my pdp8/L or pdp11's. The boards never run very warm, has anyone seen boards to the contrary?

 

[Bitly]

It appears you have plenty of space in the box, why not lower the backplane a couple inches?

CW

 

[Steve Toner]

It appears you have plenty of space in the box, why not lower the backplane a couple inches?

CW

It could come down maybe another 20mm (it's already spaced down about 20mm) if necessary. The upper RS-232 connector sets the limit.

 

[Steve Toner]

That is really nice, and oddly enough I have not seen much heat related failures on either my pdp8/L or pdp11's. The boards never run very warm, has anyone seen boards to the contrary?

Data point:

It's been running for several hours, just idling in RT-11.
Ambient temperature ~23°C, air temp @ top of case ~27°C, CPU chip ~31°C

As of right now, it doesn't look like there is a problem.

 

[Radix]

It's a general rule for everything electronic, and a 11/23 CPU probably doesn't dissipate too much heat - DEC stated that forced air cooling is required at all times for the backplanes - but obviously some boards use a lot more power than others. the KDJ11-B cards and in particular the FPU use a lot of power and get hot rather quickly - as does the RLV12 controller but they are quad boards, so not going to be an issue for you - but a densely packed (eg 2MB memory dual width board) or newer SCSI controller could run a lot warmer and easily go unnoticed. DEC uprated the fans when they went to the 11/23+ chassis with the higher rated H7861 PSU and the BA23 fans are something like 165cfm - but at the other end of the scale, the BA11-VA tiny 4 slot chassis akin to your system still had an 80mm fan directed straight at the backplane and lots of slots in the case...

Enjoy it and have fun!

Robin

 

[AndrewZ]

It's been running for several hours, just idling in RT-11.
Ambient temperature ~23°C, air temp @ top of case ~27°C, CPU chip ~31°C

Just for grins give it a compute intensive task and see if that changes the temp. 10 GOTO 10, or calculate prime numbers or something.

 

[leegleason]

Regarding the cooling - it really should have good forced cooling if you're precious about your boards... The rule of thumb is for every 10C warmer a device runs, its failure rate doubles - or life expectancy halves... If you want this to last longer, you know what to do!

Robin

I think that rule of thumb applies to lots of things, due to the Arrhenius activation energy equation, which says that a 10 degree (Celsius) change in temperature causes a doubling of the reaction rate.

 

[Steve Toner]

Just for grins give it a compute intensive task and see if that changes the temp. 10 GOTO 10, or calculate prime numbers or something.

I don't know why you think it would. This isn't a microcontroller with low power sleep mode. Even idling in RT-11, it's still running code - looking for a task to run, handling clock interrupts, accessing Qbus memory, ...
Unless some chip gets enabled that wasn't before or more transistors switch in a given time period, I wouldn't expect to see any change.

But just for grins I plugged it into a Bluetti AC50B power station. This has a wattmeter that displays the load. And I ran a floating point number crunching program (computing PI) that takes 8 1/2 minutes to run on this machine. Here are the results:

RT-11 idle: 81W
Number crunching: 80W

I don't consider the reported 1W difference significant.

Program:

 

[AndrewZ]

I don't know why you think it would. This isn't a microcontroller with low power sleep mode. Even idling in RT-11, it's still running code - looking for a task to run, handling clock interrupts, accessing Qbus memory, ...
Unless some chip gets enabled that wasn't before or more transistors switch in a given time period, I wouldn't expect to see any change.

But just for grins I plugged it into a Bluetti AC50B power station. This has a wattmeter that displays the load. And I ran a floating point number crunching program (computing PI) that takes 8 1/2 minutes to run on this machine. Here are the results:

RT-11 idle: 81W
Number crunching: 80W

I don't consider the reported 1W difference significant.

Program:

Thanks, I'm always curious about this kind of stuff. And I agree there was no low power sleep mode

 

[Radix]

I would guess at a memory diagnostic like VMSA, constantly driving the bus and selecting RAM chips would be slightly higher consumption, but not much...

 

[Steve Toner]

OK, I need to take a break from this. I was having fun, happily running the system, editing/compiling/running FORTRAN code and my DK: device (DL1 on the RLV12 emulator SD card) got corrupted. I'm seeing multiple instances of ".FOR" extensions changed to ".FOP", ".OBJ" to ".OBH", and "SAV" to ".SAT" -- those are all the result of a single bit getting cleared (e.g., RADIX50 value of J is octal 12, H is octal 10). But there are other corruptions as well that aren't so simple (like a directory entry for a .COM file pointing to a FORTRAN file instead, though I suppose that could still be caused by a single bit error - I don't have any info on RT-11 disk format).

I have no idea where the problem is - my backplane, the RLV12 emulator, the SD card, the memory, the processor, the operating system.
I'm strongly doubt it's any of the latter 3 - the M8059 supports parity, so should report an error if it's dropping a bit, the processor still runs the operating system just fine, and RT-11 is very stable software.

I'll have to think about how to proceed...