hams on classiccmp

This machine was one of a pair bought from a collector who bought from a collector who bought from...? It was interesting noting the differences between the two: for instance, on one the black foam was in fairly good shape (replaced it anyway), but on the other it was falling off in chunks. Both seemed in good mechanical condition. The next step was to ensure electrical integrity. So I replaced every electrolytic capacitor of consequence in the switching power supplies - five in one machine, six in the other - plus the LSI-11 boot machine and RX01 boot floppy. This was expensive! We used Mouser as our source because their stock and prices were the best we found among the established vendors. Even so, there were a couple of instances where it just wasn't possible to buy a large electrolytic with the exact same form factor as the original - the newer ones were smaller! Curse progress.... In one instance (the main filters in the switchers), it was necessary to ever-so-slightly reshape the holes on the circuit board so the new capacitor's posts would line up. Of course, I visually inspected each and every board in both machines to look for heat marks, obviously damaged bypass caps and the like. There were several instances of tin-plated IC sockets, so I did the rock-and-reseat maneuver just as a precaution. After all of this, we carefully brought up the machine. We had a challenge because 120V three phase doesn't seem to be usual practice in US wiring - we had 240V three phase, but that obviously wasn't going to do us any good! Carefully looking through the power distribution unit's engineering drawings, it became clear that DEC used three-phase simply to balance the current load among the legs - in fact, everything runs on 120V. So we used equal care in reviewing the wiring of the warehouse where we keep these machines and found three outlets that were (a) on the same side of the 240V mains and (b) not sharing a breaker and circuit. Those were connected to a three-phase outlet, the VAXen were plugged in and voila! NOTE: we have a team member experienced with commercial power circuits. Don't try this at home - or if you do, be very very careful and be certain that, from any of the three live blades to another, you don't have more than 120V. Now that the machines would power up, I scoped all the power supply voltages to ensure they were really DC, i.e. that I hadn't missed an important filter cap anywhere. All good, so I tried booting from the floppies we got with the machines. I was able to get the basic console to boot - hooray! - but was unable to get the microdiagnostics to run. We had agreed that successful execution of the low-level diagnostics was a precursor to any attempt to install the OS, so this was a roadblock. I learned a lot about the interaction between the LSI-11 boot device and the VAX-11 CPU. The Console Interface Board actually 'lives' in the LSI-11's memory space - it holds the boot ROMS, for instance. So if you remove the LSI-11 to test it, you can get an ODT prompt but that's it. There is a channel between LSI-11 I/O space and the VAX-11, too, and the two machines communicate during normal operation, which is how the LSI-11 serves as the console device. One thing that seemed odd was that when I tried to execute the microdiagnostics, the LSI-11 would halt at 070000. I finally found a listing for the microdiagnostic that showed a macro statement basing the program at that address - but after poking around a bit, I found out it was loading at 071000. I tried starting from there - no go. What I finally figured out was truly bizarre. THERE IS MORE THAN ONE VERSION OF THE MICRODIAGNOSTICS FOR A /785, and the floppy and program names are the same. The only thing that distinguishes them is the 'stock number', i.e. AS-A123 blahblah, and I never found a shred of documentation that associated one or another with a given machine or serial number series. I infer that DEC assumed you'd get your set of floppies when you bought your machine, duplicate them as needed for security, and that was that. They never anticipated people like us, doing 'archaeology' on obsolete machines.... After some investigation, I discovered a key characteristic: the version that didn't work with my machine had a zero-filled first block in the program files, and the ones that did work had some code in that block, beginning with a JMP instruction. We discovered the repository maintained by trailing-edge.com and perused those, together with the EVNDX file from bitsavers.org, and eventually found a full set of diagnostic disk images. We transferred those onto physical floppies using a PDP-11/03 and the TU-58 emulator program I found through Will Kranz's website. Now we were running microdiagnostics - but failing some. I started focusing on one of the two machines because it seemed in better health, but the memory test chains were failing. Of course, they identified themselves on the console as testing MS780A through -D, and we had a -H in each of our machines. It seemed quite possible, from the errors I was seeing, that it was that simple - but I still wanted a good memory diagnostic run. Searching some more, I found that there was a floppy designated RX3 that claimed to be a microdiagnostic diskette - the documentation for the microdiagnostics (from bitsavers.org) pointed you at RX1 and RX2. My study of the microdiagnostic system told me that it was smart enough to allow for extension, because it would look for files with names of a given structure when you told it to proceed. So I dropped in RX3, gave the DIAGNOSE instruction, and wonder of wonders: a memory diagnostic for MS780E-H began running! We passed that one just fine, and began working toward getting the OS installed. We had CDROM media for OpenVMS 6.2, the latest version certified for the VAX-11/785 - but of course, we didn't have any sort of adapter to talk to a CDROM drive. But we'd earlier made a judgement call to use the CI interface for the system device, and had ordered an HSJ50 setup from PSDS in Woodinville, Washington - really cool guys, by the way, I'd recommend them anytime. I anticipated we'd hook a CDROM into one of the SCSI ports, a shelf of RZ drives into another and we'd be off. However, I did not count on the embarrassment of riches we received from PSDS. They fully populated the HSJ with six shelves and configured RAID arrays up and down! It was great, but it didn't leave anywhere to plug in my CDROM. I played with it for a while, loathe to disassemble it, and finally developed a solution. One of the RAID arrays was 'short', with four drives instead of six; the other two were hot spares. So I pulled one of the spares, plugged it into a StorageWorks shelf on one of the MicroVAX 3100s in my personal collection, and copied the installation media onto it with BACKUP ddcu:/IMAGE/NOALIAS target:. I plugged that back into the HSJ, set it up as a single disk unit, and it booted into the install! I found that I actually needed TWO of these: one was the boot device AND target, the other served as the source media under the install program. When I finally installed the OS onto one of the RAID arrays, I used the same BACKUP command to install an image of the media, booted it, and used the single-disk copy as the source media. One diversion that is worth noting: I tried using the floppy based standalone backup floppies we're gotten with the machines. While I could go through the series, I could not get them to recognize the CI-based drives. Given all my other challenges with versions, I suspected this was just another such issue. I finally decided to simply use the CIBOO.CMD file on the boot floppy. However, I needed the 'station number' and 'device number' for the boot device. Station number was easy: our HSJ dual-redundant controller was set up as IDs 0 and 1, so I used the value 0x100 in R2. But device number had me stumped for a bit. I finally figured out the key: the unit numbers defined in the HSJ's configuration correspond to the device numbers! So 'D6' in the HSJ menu - what I'd called the single-disk unit with the install media copy - translated to 0x6 in R3, and things booted fine. Now we were cooking with gas. I had a valid VMS license but not a UCX license. We also had MultiNet media, but licensing looked like it was going to be a challenge. So I tried CMUIP. It installed with no problems, but wasn't finding an Ethernet adapter. I did some troubleshooting on this: the adapter was, of course, housed in a Unibus expansion cabinet, a DELUA. We had a spare and I tried that - no joy. The LEDs on the DELUA did not change with boot, but all stayed on. I scratched my head... ...and noticed that we had two Unibus cabinets, since I wasn't paying any attention to the other machine. I swapped them around - non-trivial, but not too difficult - and the Ethernet card fired up just the way Ken intended. Now CMUIP found the card and everyone was happy! I suspect the problem with the other cabinet is cabling - you know how ribbon cables can be. After a bit of scrabbling around in the wire closet, we had live Ethernet, plugged in a static IP address, and left the room - to go to another room and telnet into the VAX-11. Happy Dance! Overall, I'd have to say that the biggest challenge was overcoming the shortcomings in extant documentation. As I said, I don't think DEC ever anticipated someone like me would be trying to piece things together years after the machine was EOL. When diagnostics fail, the course of action is either 'replace the device' or 'call field service' - perfectly reasonable for the era, but not always a solution today. I also found that some key diagnostics were not in evidence - or rather, appropriate versions of them. For instance, I found a copy of the Ethernet diagnostics, but they would not run under my version of the Diagnostic Supervisor. It's quite possible there's another version of the DS that would run them, but I don't know if it will run on my hardware - perhaps one of these days when I'm really bored I'll find out, but for now I'm just happy to have a living VAX-11. -- isk ________________________________________ From: brad at heeltoe.com [brad at heeltoe.com] Sent: Saturday, January 17, 2009 4:11 AM To: General Discussion: On-Topic and Off-Topic Posts; Ian King Subject: Re: hams on classiccmp Ian King wrote: very nice. Can you describe how it went? Was the 780 previously in moth balls? -brad