The following is an edited version of a post I recently made to alt.sys.pdp11 and vmsnet.pdp-11:
We’re trying to set up an 11/24 for dumping tapes and disk packs. The plan is to run ULTRIX-11 V3.1 from a ZuluSCSI emulating a (huge) RA81 on an Emulex UC18 MSCP SCSI controller, and then dump the media to that.
At the moment, there is also a TS11 controller in the machine, and we want ideally to have RK05 and RP06 controllers too.
Unfortunately, we haven’t been having much success so far. The CPU won’t come out of halt properly, and we appear to lose bus continuity.
Either something is broken, or there is a configuration issue. Several of the cards have to be installed installed in a slot with the NPG jumper removed. The machine has “DMA slots”, which I take (though I may very likely be wrong) to mean slots where this has been done.
Before we started, the machine was booting RT-11 from an RL02 pack just fine. After failing to make the new configuration work, we attempted to return it to its original configuration, and it would no longer boot RT-11 from the same RL02 pack. We thought that the pack may have become damaged. When we tried to boot it, the “PROC” light flashes slowly. After replacing the RL02’s ribbon cable, we are able to boot RT-11 from an RL02 pack on the machine in its original configuration. So the 11/24 itself still works.
There is not much clearance between the cards, which have bits of cardboard document folders between them for insulation. We did let out some magic smoke when trying to make the new configuration work, but I believe we have found the culprit.
I am posting here just in case there is something obvious preventing it from working. I’m the software guy on this project, and I don’t know much about PDP-11 hardware, so my description of the problem is based on what I have seen and heard.
It seems to me that we have either bad hardware, bad configuration, or (more than likely) a combination of the two.
Thanks in advance for any advice you may be able to provide.
To elaborate from the above- I am working on this project as well.
We ended up finding the RL02 ribbon cable failed during our reseating of cards and the unit is back to running RT11 from an RL02.
The issue is now the Emulex UC18 SCSI card and settings and compatibility.
When it’s installed in a slot (7) the unit either doesn’t run the processor or is stuck in halt. (@ command or nothing at all)
Trying to understand if it will work straight in a PDP-11/24 with a stock M9312 boot card or if we have to do something to the unit to allow it to work.
Jumpers are all default and the Slot its in was previously a DMA (M7819 card) which apparently will have the NPG set correctly.
There’s the other thought - having plugged these into a QBUS machine in the past- can they have been damaged? We didn’t realise they were Unibus initially due to the UC-18 and UC-08 being so much the same
Are they just bad cards? One did lose a bus transceiver on plugging it in for testing.
Update: after posting on Facebook, we have discovered we were plugging the UC18s into the wrong half of the slot. At least one of them has a fried bus transceiver, which should be replaceable. Hopefully (fingers crossed) we should be able to make one work soon now.
1 Like
This is correct.
Strangely- the users manual doesn’t explicitly state this well- the installation portion of the physical card starts on page 45 and apart from CDEF being printed microscopicly on the card, I don’t feel as bad as I would have finding out from another Facebook poster that they delivered 20 of the cards for 4 of them to be blown up in the same way.
The online copy of manual is in very poor form. I will scan our original, add an addendum to prevent future issues by others.
There is a lot more PDP-11 technical activity on DEC | Vintage Computer Federation Forums and on Discord
There are several different types of slots in a Unibus backplane. In a Unibus expansion backplane, like a DD11-DK, the AB part of the first and last slots is Unibus. These two connectors are used to interconnect processor backplanes to expansion backplanes and interconnect expansion backplanes. The M9312 Bootstrap/Terminator board goes at the end of the Unibus in the AB connectors. The CDEF part of the first and last slots of the backplane will accept quad-wide SPC I/O boards like serial ports, clocks, RX01/02 controller, etc. The remaining hex-wide slots are usually Modified Unibus Device slots. You can put hex-wide SPC controllers in the ABCDEF connectors like the RL02, DZ11, UDA-50, etc. You can also put quad-wide boards the CDEF connectors of a MUD slot.
If a controller like the RL11 and UC18 needs to DMA data to and from a peripheral and memory it needs to kick the processor off the Unibus. It sends a signal to the processor on the NPR bus (pin FJ1 on the backplane), and when the processor can give up the Unibus it sends a signal back on the NPG bus (pin CA1). Now comes the fun part.
If the slot does not have a DMA capable board installed pins CA1 and CB1 on the backplane are interconnected with a wire-wrap jumper. If the slot does have a DMA capable board installed, like the RL11 or UC18, the wire-wrap jumper on pins CA1 and CB1 must be removed.
If you have an empty slot with the CA1 and CB1 jumper installed you need to install a little G727 jumper board in connector D. Be careful which way it is installed because you can put the jumper board in backwards. If the CA1 and CB1 jumper has been removed you can put a dual-wide G7273 board in the empty slot. If you don’t have a G7273 you will need to replace the wire-wrap jumper on pins CA1 and CB1 and install a G727.
DMA capable controllers like the RL11 and UC18 connect the NPG-in signal to the NPG-out signal when they are not using DMA, This lets a controller further away from the processor do DMA. If the chips on the controller that interconnect the NPG-in and NPG-out signals then no controller further away from the processor will be able to do DMA. You can measure the voltage on the NPG, NPG-in, and NPG-out signals to see if the are in the correct state when the processor is not running. The NPR signal should be high and the NPG-in and NPG-out signals should be low.
hi @m_thompson see attached.
We did learn that two of the three uC18s we hold are bad - horrid to say at least 1 had a bus transceiver smoke up because of my incorrect understanding of the SPC slots being CDEF and NOT ABCD - which would have been a smart inclusion in the UC18 manual as a sidenote!
Now we have a good UC18 installed in slot 9 in place of the G7273 shown and the 7772150 vector is now shown in RT11 as INSTALLED.
We are however unable to boot from the card either due to the incorrect ROM pack on the M9312 card or some other issue.
Showing the original card layout.
Showing the UC18 visible within RT11.
Good progress. Have you run F.R.D to format and verify the SCSI disk and configure the controller?
We have not been able to enter FRD.
We have tried to boot using the ODT commands from the manual, and we have managed to make the ZuluSCSI activity light light up, but we haven’t been able to boot.
I may be completely wrong, but since that boot procedure is trying to read the first block of the disk into memory, presumably using DMA, perhaps its failure is indicative of bus grant problems?
The MSCP emulating controllers make the SCSI drive look like one or more DEC drives. If you had a 100 MB SCSI disk the controller could be configured so it offered 10x RL02 drives or 4x RK06 drives, or some other combination of DEC drives that would total to 100 MB. The F.R.D. tool can scan the SCSI bus to see what drives are there, and let you configure what drive emulation you want.
I imagine that it writes a partition table on the drive that matches the drive emulation, and also saves the configuration settings in NVRAM. If you have not run F.R.D the controller is probably looking for the configuration information written to the drive and not finding what it expects.
What we have plugged into the UC18 is a ZuluSCSI with an RA81-sized image on the SD card.
The image has ULTRIX-11 installed on it, configured for an 11/24 with an RA81 on a UDA50 controller.
I don’t think that it is that simple with an MSCP controller that is emulating DEC drives. The manual here: https://bitsavers.org/pdf/emulex/UC1751001-C_UC17_Dec90.pdf says that you need to format the drive from the UC18. Section 4.5.9 tells you how to start F.R.D. on an 11/24.
The MSCP emulating controllers make the SCSI drive look like one or more DEC drives.
I can find no reference to this mode of operation in the UC18 manual.
The manual here: https://bitsavers.org/pdf/emulex/UC1751001-C_UC17_Dec90.pdf says that you need to format the drive from the UC18. Section 4.5.9 tells you how to start F.R.D. on an 11/24.
We have been following those instructions. We know that ^J sends an LF. We haven’t been able to start FRD yet.
Did you look in memory at 200 to see if F.R.D. got loaded into memory?
When you run the commands the 200 or 200g can not be inputted as the ODT has effectively locked up to a @ symbol. This does not seem to be a matter of the drive connected as I would expect it to work when no drive connected.
Can you examine and deposit memory when the system is first powered on?
I imagine that the UC18 is using DMA to put F.R.D. in memory @ 200. You could connect an oscilloscope to the NPR, NPG, and MSYN signals to see if the UC18 is actually putting F.R.D. in memory. If not, you may have an issue with the UC18 not getting the NPG signal so it can’t do DMA.
Not sure if this means anything but if I run 200G i get back 000202.
Im sure there is an expectation of me to load the FRD into the memory, I just simply wondered if it auto dumps it in there.
I will have @rchipp look into the NPG lines, however considering it had to have a M9132 card in it to terminate the NPG surely it is there unless being blocked by the RL02 controller.
