Computer History - Mini-Computers

Rick York

The mention of computer history and the book, "Soul of a New Machine" made me remember the old days and mini-computers. They were called "mini" because they weren't like the old room-filling mainframes. Most of these were about the size of standard refrigerator. That's the era in computing that I am most fascinated with and one that the details of are mostly unknown to most people today. I especially like how the machines were built. They used what is known as the "bit-slice" design. The CPU usually took an entire two-foot square circuit board and it was built with circuits that were initially just one bit wide and they essentially stacked them together to achieve the word width they wanted. Many of which were 32-bits and, as I recall, DG's was 36. AMD was probably the biggest company making the chips at the time and eventually they came up with 4-bit wide chips. All of this was before microprocessors became useful enough to compete with the bit-slice designs. An amusing story from the tail end of this era : I worked at a company that had a room full of mini-computers including several VAXes. We were doing a project that controlled the North Shore Pipeline and talked to several RTUs using the Modbus binary protocol. It uses CRC-32 CRC-16 to compute a checksum for every packet and that would bring the VAX 750s to its knees. The CPU usage would spike every time because this was before the table-driven algorithm had been publicized. DEC's answer was to implement the CRC-32 CRC algorithm in firmware. To implement this a new set of microcode on EPROM was installed and the CPU board was re-wired. The local service guy, Ed, came and had to change the wiring on the CPU board which was ALL wire-wrapped. I couldn't believe it. It was NOT a printed circuit board - it was all wire-wrapped. Those modifications made a huge difference and CPU usage was normal afterward. This was in around 1985 or so and it still amazes me.

"They have a consciousness, they have a life, they have a soul! Damn you! Let the rabbits wear glasses! Save our brothers! Can I get an amen?"

Gary Wheeler

Rick York wrote:

several VAXes

Before Ravi falls over in a dead faint, I'll let you in on a small correction. The plural form of "VAX" is "VAXen" :-D . I worked on several VAX-11/780's back during the 1980's. The 780 would let you load microcode at startup time via an 8" floppy drive attached to the PDP-11 they used as a console/boot device.

Rick York wrote:

the wiring on the CPU board which was ALL wire-wrapped

Reminds me of a project I watched in the same facility during that era. It was a graphics engine, 1024x1024 resolution, with 32 bit planes. Four cabinets of 8 bit planes each, 4 boards per bit plane. Each board was densely populated and looked to be about 24" square. All. Wire. Wrap. I really felt sorry for the poor schmucks installing the nightmare. After months of work they finally got one cabinet to power up successfully and run for a demo. I don't think the Air Force ever accepted the system, however.

Software Zen: delete this;

PIEBALDconsult

I still insist that the plural is "VACes" -- "VAXen" is too Germanic in my opinion.

raddevus

Gary Wheeler wrote:

Each board was densely populated and looked to be about 24" square. All. Wire. Wrap. I really felt sorry for the poor schmucks installing the nightmare.

I have a small project I created that uses a small tft screen and there is one data line for each channel (8 total) plus about 8 or so other lines I have to wire to the arduino and just dealing with those few and running into issues where something isn't making a great connection can drive you insane, so I can only imagine. Best Debugging Tool Ever: Continuity test setting on multimeter. :rolleyes: Great stories, thanks for sharing

Gary Wheeler

You're welcome.

Software Zen: delete this;

CPallini

Why CRC32?

"In testa che avete, Signor di Ceprano?" -- Rigoletto

Rick York

That is part of the Modbus RTU standard. The Modbus ASCII protocol does not require a CRC but apparently they added it for the binary RTU version. Back then, communication was mostly 9600bps and 19600 was considered fast. RTUs were usually separated at far distances so repeaters and modems were frequently used. In the case of the pipeline, the RTUs were separated by about a quarter mile between them so some were miles away. Today, the vast majority of communication using Modbus is over TCP/IP and no checksums are used - TCP/IP's error handling is relied on.

"They have a consciousness, they have a life, they have a soul! Damn you! Let the rabbits wear glasses! Save our brothers! Can I get an amen?"

CPallini

RTU MODBUS standard uses CRC16.

"In testa che avete, Signor di Ceprano?" -- Rigoletto

Rick York

OK.

"They have a consciousness, they have a life, they have a soul! Damn you! Let the rabbits wear glasses! Save our brothers! Can I get an amen?"

rjmoses

I supervised the CAM group in the early 70's at GTE where we built PABXes. Our group took circuit data from the CAD group and produced the wire wrap data for Gardner-Denver wire wrapping machines. The data was created and stored on a pair of IBM 370 models 158 & 168. It was accessed from three IBM 1800's via 9600 BPS bi-sync links and stored locally on 2311 disk drives using a MRU algorithm. We drove 30 such wiring machines and 10 testing machines from two of these computers with the third as backup and running low priority applications. We got more productive work done out of 32K mini-computers than most people get out of 32GB machines nowadays. OF course, we didn't have to worry about fonts, font sizes, colors, windows, icons and all the other glitz.