VAX 11/750 Config and Diagnostics, anyone?

On Thu, Nov 15, 2007 at 03:11:36AM -0500, Patrick Finnegan wrote: Pat, you beat me to it, but let me add that in some cases, where you have MUD (Modified Unibus Device) slots, it can be *very bad* to stick a terminator in a peripheral slot. I'm pulling out of the back of my memory that it's really, really bad to do that to an 11/725 or 11/730 - you can blow the power supply. There were, back in the day, configuration guides for various system setups. Unfortuately, all Unibuses are not created equal, especially those in CPU cabinets. There is a great deal of overlap in that a TU80 controller or a DZ-11 or a UDA50 will work in pretty much every instance I can think of, but exceptions do lurk around the edges. One egregious example I can think of is the COMBOARD-I. It works in a DD11-DK (like the Unibus backplane in your 11/750), but *not* in the Unibus slots of the CPU backplane of an 11/730. The fault is with the original layout of the COMBOARD-I, not the 11/730, but it was caused by an incomplete understanding of where the Unibus signals were supposed to live in an SPC (Small Peripheral Controller) slot. I have a "modified COMBOARD-I" with lots of blue wires relocating the signals to the proper edge fingers. Fortunately, the lesson was learned before the COMBOARD-II, which works in all types of Unibus slots. The important thing to remember is what Pat already said - the in/out slots are different from the ones in the middle. Unlike OMNIBUS or VME or ISA busses, the Unibus is not a simple pin-1-to-pin-1, pin-2-to-pin-2 kind of pattern. That should work quite satisfactorily until you have a Unibus controller you'd like to boot from. G727s were very common in the day. What was less common were the dual- height GC7273 (or Software Results' GC747) grant cards that also jumpered the NPR wire. If you are down to the booting stage and you get strange behavior, be sure to check your NPR wires. Disk and tape controllers tend to need the NPR wire off the backplane to be able to make the "Non Processor Request" (to initiate what is called in other architectures a DMA cycle), some comms controllers like the DMF32 also need the NPR wire off, while other comms controllers like the DZ-11 don't need it, so it should be jumpered on the backplane, if the card doesn't happen to jumper it at the card edge (newer peripherals might, older peripherals won't). The "easy" way is to remove all the NPR wires from all nine slots in a DD11-DK, then use dual-height grant cards. The harder, but less expensive way is, to check the needs of your cards and adjust the backplane accordingly. If you don't make frequent changes, it's not as bad as it sounds. If you are changing cards every month (as we did, for development, etc), nine dual-height grant cards is cheaper than an hour or two of head-scratching. We had an 11/750 that couldn't talk to its tape drive out of the box. Turns out we had forgotten to remove the NPR wire for that slot. I think it was 4 hours before we remembered to check (since all of our other machines had *no* NPR wires, we got out of the habit of looking). -ethan -- Ethan Dicks, A-333-S Current South Pole Weather at 15-Nov-2007 at 08:30 Z South Pole Station PSC 468 Box 400 Temp -41.1 F (-40.6 C) Windchill -68.1 F (-55.6 C) APO AP 96598 Wind 9.6 kts Grid 124 Barometer 681.5 mb (10572 ft) Ethan.Dicks at usap.gov http://penguincentral.com/penguincentral.html