Over the last weekend VCF held a small workshop for people to bring and work on some vintage computers. So I figured it would be as good a time as any to finally sit down with the SGI 3130 and really debug the issues I was having. This is the second time working on it, the first was out of the blue after realizing the supply I had for it was working. Previously there was one rail that refused to come back up but after throwing the supply back together to clean up the mess I tested it again and it seemed to be working again. Left it on some tail light bulbs to ensure it was stable and threw it into the machine and it began to come up. I got stuck with random bus vector errors.

So I brought the machine to the workshop along with my HP 16700 and Rigol.

setup

Coming to the workshop I thought that maybe a bus driver was going bad on the main processor board from symptoms. After getting the logic analyzer probed up to a multibus proto board to break out the entire bus I was able to find the vector reset problem came down to the bus attempting one access then throwing the problem. It looked like it was reading something. I knew it couldn’t have been the ram as having one or the other in the system caused the same symtoms. So I tried removing both the disk controller and network controller from the backplane and the issue suddenly went away. Turns out the Excelan NX200 had some kind of problem.

excelan_nx200

Thankfully I know someone with another one of these lan boards with the manual so we can try that out and possibly the docs will help in debugging the issue with the card I have. Next I was on to attempt to adapt a later Personal Iris keyboard as many in the SGI community insisted that it used the same protocol and interface as the older 68K IRIS boxes.

After tracing back the connections to the top DB25 and some guesswork I am pretty sure the RX/TX for the keyboard is in the same spot as any RS-232 port would be. Though they are using a slightly different standard RS-423 which is supposedly backwards compatible with RS-232 and was an enhanced version of RS-232. But it clearly didn’t catch on. I tried the default RX/TX pairs and no cigar, keyboard would not initialize and the machine did not seem to recognize it. I tried another seemingly RX/TX pair on the DB25 as-well but it gave the same result. Moving on I was hoping to just see if we could get the machine booted and poke around on the drive.

framebuffer

I had forgotten that the graphics subsystem would give me an initialization error on boot. A stray interrupt 43 or 44. 43 occurs when you don’t have the ribbon which interconnects the processor and one of the graphics boards. 44 if everything is in place as it should be… I found I had jumpers different from a photo of another board on the internet on one of the graphics boards. I tried that config taking photos before hand and no dice.

My current suspects are either: A) Bad graphics subsystem, needs debug and will probably need a working set of boards for debug purposes to at-least narrow the issue down to one of the 3 major boards and or bitplanes. B) A keyboard is absolutely needed to boot this system with graphics installed. With a keyboard the system should initialize the video and draw an on-screen terminal for booting. Then the OS should take over. Its clear that the OS isn’t seeing something it likes with the graphics and is getting a stray interrupt. If the PROM doesn’t init the graphics without a keyboard this could be leaving it in a random state causing the OS to try and talk to the graphics and end up spewing what I see above sometimes.

I can not test B until I have a keyboard. I know someone with one but they have not been able to get to it and get it to me to loan. This will be continued when I can get either access to a keyboard, another system, or both…

-Connor Krukosky