Published on

Building my own 8 bit computer

Authors

This post contains some of the work in progress during my build of Ben Eater's inspired 6502 8-bit computer similar to typical machines of the early 1980s. If you haven't seen Ben's videos, I would strongly suggest you start there: Ben Eater's 6502 project

The full source code for this project is available on GitHub

KW-6502

Table of Contents

Philosophy

After watching Ben Eater's videos, I was intrigued by the idea of building a 8-bit computer. I have always liked building hardware projects, and this would be my biggest one yet.

But most of all, I was super excited to start building software for this thing. I always wanted to learn assembly and this was the perfect way to do that.

What is different

After building Ben Eater's 6502 project, I was very excited to start changing and optimizing it. I changed some things to the address decoding, added serial communication and changed a few other things.

Address decoding

As Ben explained in his video, there are many ways to handle address decoding logic and he opted for model with 16K RAM, 8K VIA shadow and 32K ROM. He does note that it's a bit of a waste, but given the simplicity of the project it should not be a problem - and he is absolutely right.

I wanted more with my build. I wanted to test my understanding of how address decoding works. My build provides 24K RAM, 4K VIA shadow, 4K ACIA shadow and 32K ROM.

Address Decoding Comparison
Segment BE6502 KW6502
RAM 0x0000-0x3fff 0x0000-0x5fff
ACIA 0x6000
VIA 0x6000 0x7000
ROM 0x8000-0xffff 0x8000-0xffff

As you can see, less address space is wasted which means more space is available to RAM. I have never managed to fill up the RAM or ROM, but it's a nice change anyways. To achieve this I added another Quad NAND gate (74HC00).

ACIA chip for serial communication

Thanks to the ACIA chip all the input/output can be handled by the serial port. This expansion allows you to output infinite amounts of data to your serial monitor, so you're no longer limited by the tiny LCD. You can input all sorts of data, like keyboard input or even files using the XMODEM protocol.

Installing the ACIA chip was a bit tricky, there's a lot of wires to connect. But once you have it, it's pretty easy to use.

Other changes

I chose to migrate from VASM to CC65. As much as I liked VASM for small, simple projects, it became quite cumbersome to use when I wanted to introduce different addressing models. CC65 is basically much better compiler for anything 6502. It also enabled me to use C code in my project.

A small change I did was swapping the 1 MHz clock to a 1.8432 MHz clock. I did this because the ACIA chip needs this specific clock speed anyways, so why not use it for the whole system. At first I was worried the system would become unstable, but I was wrong.

Software

Most software for the KW6502 is written in assembly. I got C code working, but I just haven't used it that much yet. The software should work on any reasonable clock speed and can be easily configured for any similar system. Most of the software makes use of the serial expansion, so a lot of software can't be run on the default Ben Eater 6502 project.

I wrote examples for every piece of hardware I could interact with. The lcd, serial port and buttons all have an example program that shows them working. I split the code for interacting with the lcd and serial port into separate files, which can be used in other programs.

The source code for the software is available on GitHub

MicroChess

I got the MicroChess game working. It's a simple chess game, but it's a good example of how powerfull the serial port is. Microchess for the Kim-1 was the first game program sold for home computers. After six months of development, the first copy was shipped on December 18, 1976.

MicroChess

Getting C code working

It was very difficult to get C code working. I struggled for a long time because debugging the 6502 at runtime is basically imposible.

I think it's pretty cool that it's possible to write both C and assembly code in the same project. It enabled me to use my assembly lcd routines in my C code. Only thing I had to do was export them, specify them in a header file and use them in my C code.

C code is a lot easier to write, read and understand than assembly, because seemingly simple things like creating conditional loops are actually pretty difficult in assembly.

The only downside is speed and space. Assembly code is much more space efficient and faster to run. Small price to pay for sanity, I guess.

BASIC integration hell

I was very excited to start writing in BASIC. I knew it would be quite difficult to get it working, but getting BASIC to work on the 6502 was one of my end goals, so I persisted.

There are a lot of BASIC variations, and you can even create your own. I found a modified OSI 6502 BASIC program that included some ACIA routines to make it work with a serial monitor. It just wouldn't work, I had loads of trouble with it. I changed the addressing and it still didn't work. After a lot of changes to the ACIA routines and countless hours of debugging, I finally got it working.

Future plans

I'm planning to add a few more features to the KW6502.

It would be fun to add the TMS9918A video chip to the KW6502, I heard it's not that difficult, so this may be my next expansion.

Some other ideas:

  • Adding an SD card reader
  • Audio beeper with multiple tones
  • Potentiometer to digital input
  • Writing a shell
  • Writing a simple math interpreter
Discuss on Twitter • View on GitHub