Scor­pio News

  

July–September 1987 – Volume 1. Issue 3.

Page 42 of 67

Upgrading a GM813 to 256K

by W.H. Turner

Introduction

This article describes how to add more memory to a Gemini GM813 CPU board and what to do with it.

There was not a single good reason for wanting to put more memory on my Gemini Galaxy 2 computer! Instead there were several vague possibilities – the possibility of using it for a RAM disk, the possibility of using it for a disk cache (as in CP/M Plus), and the possibility of adding a GM888 board (8088 co-processor), that made me do it.

Whatever the reason, the additional memory boards looked very expensive and used valuable bus slots which made the whole exercise a lot less attractive. My solution was to replace the existing 64k bit RAMs with the pin compatible 256k bit RAMs which quadrupled the memory for an outlay of only £55.

The GM813 board, at the heart of the Gemini Galaxy series of computers. is able to address up to 512k bytes of memory via 19 address lines but only uses 16 to address the 64k bytes of onboard memory. Access to the additional address lines is controlled by two 74LS189 TTL RAMs used as sixteen 8 bit mapping registers between the top four Z80 address lines and the top seven bus address lines. Each mapping register controls a 4k byte block of the 64k byte Z80 address space. Each address generated by the Z80 selects a 4k byte block of memory via one of these registers and the 8-bit value written into the mapping register determines which 4k block out of the 512k bytes is actually accessed. The result is that any 4k byte block of memory can be mapped at any 4k byte boundary in the 64k byte address space of the Z80. In passing I should mention that the bus signal RSFU (reserved for future use) can be redefined and connected as A20 giving a colossal megabyte of address space should anyone ever find a use for it.

I first became interested in extending my memory when everyone else was enjoying RAM disks except me. But the cost of buying an additional memory board end alternatively, the complexity of producing my own board, meant that things didn’t get very far. It was only after an engineering friend suggested replacing the existing 64k bit RAMe on the processor board with 256k bit RAMs that I began to look at the problem more seriously.

For a long time I procrastinated, believing that if I attempted even minor surgery on the processor board I could end up with a very dead computer. Eventually I took a firm grip on myself and removed the GM813 board from the computer to look at the possibilities, besides at that point I had already bought the 256k bit chips! Looking at the board and the circuit diagrams, so thoughtfully supplied by Gemini I worked out a plan of action.

Not being very imaginative when it comes to hardware I felt that I should, as far as possible, duplicate the circuits already in use on the processor board to address the additional memory. Further, by mounting the additional chips directly above the existing chips, the new ones could be easily connected to the power supply and the necessary signals.

The modifications were to be done in a series of short steps after each of which the board would be slid beck into its slot in the card cage, the cables reconnected and the system booted back up to check that it still worked. If it suddenly stopped working then I would only have to check the changes performed in the last stage, looking for pins not inserted in sockets, dry joints, solder bridges etc. Further I would not cut any tracks so everything could hopefully be returned to its original state if necessary.

Page 42 of 67