Although this is an on-going project, I have been working on it for over a year before publishing anything, so I'll try and capture some of my approach, experiences and lessons learnt to bring this site up to date.

One thing I have been doing over the life of this project is trying to do it on a shoe-string budget. That's just the way I am - a bit of a tight-arse who likes to think he's got a bargain of some sort! However whilst being budget conscious is great in nearly any setting, it has to be with some level of pragmatic trade off and risking the saying of 'buy cheap, buy twice' coming in to effect. So I learned by trial and error the right budget trade off! I don't know how much I have spent on this hobby, but I am sure it has to be approaching £250-300 by now.

In the beginning..

I started off with literally nothing - no components, no tools, no experience. But I did a lot of reading. The internet is fantastic, so many people willing to put their experience and knowledge out there on the web for the rest of us to benefit from. I can't possibly remember or list every resource I have examined, but over the course of this project www.6502.org has been a fantastic well of knowledge.

Initially I needed to get some kit together. I bought a bunch of basic components (resistors, capacitors, transistors, 555 timers, crytals, LEDs etc.) with some wire (solid core, multi-colour). I also decide to use breadboards to try out my experiments (as they were right at the beginning). I love breadboards, but I'll mention some of the problems I have with them later, they're not all good!

Despite working in IT professional services - where I often have to spend weeks or months architecting and designing solutions - I spent virtually no time thinking about the design features of my computer (perhaps it was a back-lash to my day job!). I had some key features in mind, some mandatory otherwise it wouldn't be a functional computer, and some that I wanted to make it useful. What useful meant was the ability to program some 1980s style rudimentary video games, complete with sound effects. So the features:

• It had to have a CPU (!), a 6502 one (see below)
• It had to have memory (!), but how much I hadn't decided
• It had to have some IO (!), but for what purposes I hadn't decided
• It had to have video graphics output. I had homebrews which used small LCD panels or no video at all - this wouldn't for me as I wanted to be able to play video games on it!
• It had to have sound output output. Again, I had seen homebrews which didn't care much about sound, but I wanted this.

CPU

So, I had some basic electronic components, but no complex ICs like a CPU, memory etc. Looking through a number of existing homebrew sites, I could see a variety of CPUs had been used including 6809, 8080, 68000.

However fascinating these other projects were, I had already resolved to use the 6502 at the heart of my project. The reason is straightforward nostalgia. My first computer was a 6502 based micro called the Oric-1, bought for £80 in a knock down sale in 1984. I was 13, and knew nothing about computing, but convinced my hard-up parents that I needed one, and this was in the right price bracket (i.e. really, really cheap). But this computer took me from never having written a BASIC program to being able to code assembly in hex (because I couldn't afford an assembler - eventually wrote one for myself). When I grew out of the Oric-1, I got an Atari 800XL in 1985, which was also 6502 based.

Aside from nostalgia, I could see that with everything else I was going to have to learn to be able to build my own computer, having something I was already familiar with would be a good thing.

The first one I bought was off ebay (where most of my acquisitions of components etc. were made) - a CMOS based Rockwell model. However, during my investigations, I also found out that a company called the Western Design Center was still producing 6502 CPUs which could operate at much higher clock speeds than the original 1-2MHz in the 80s. For this main reason, I ended up using a WDC 6502 piece (there is a story behind why I thought I needed the higher clock speeds - will go in to that in another log).

Memory

Choosing RAM was straightforward enough. On ebay UK, there are a couple of sellers who provide 'hombrew computer kits'. These kits provide a CPU (typically a Z80), some glue chips (typically PLDs) some support chips (video, RS232, IO), some ROM and some RAM.

The supplied RAM is typically SRAM and the ROM is typically EEPROM. Neither SRAM nor EEPROM were very common in 80s microcomputers. However, DRAMs are difficult to both obtain and use - DRAMs need refesh circuitry which would add complexity (6502 has no refresh capability built in, although I understand the Z80 may well have). So I decided that in the interests of pragmatism, SRAM would be fine - I'm not trying to recreate a 100% 80's replica.

The supplied ROM is typically EEPROM. However, before purchasing one of these kits, I had decided that I would buy some EPROMs, a very cheap multi-format PROM programmer and a basic UV eraser. This is because of experience with my first job after graduating, working as an embedded systems programmer. In this job I would burn EPROMs and erase with a UV unit - I figured this would be how I would do things in 2014! Anyway, I will cut a long story short, due to the unreliability of probably 10+ year old EPROMs and the long build/burn/erase cycles, I turned to EEPROM. The so-called multi-programmer I had couldn't burn EEPROM, so I had to invest in some proper German (and so expensive - around £130) kit that I could rely on.

I/O

Without I/O there is no point to a computer of course. So I had decided originally that a 6522 (ebay sourced) would be fine with an RS-232 interface for serial comms. What I didn't do was to realise that RS-232 was a pain due to the voltage differences and the fact that my modern computer doesn't even have RS-232! I was also unaware that the 6522 serial capabilities are both flawed and limited, so that I may have to 'bit bang' the I/O without some kind of serial interface IC like a 6551. In the end, I ditched the 6522 in favour of two 6526 (close relation to the 6522) out of a non-working C64. At the time I didn't have access to a 6551 so had to do a bit of funky wiring to get that working. Although I will add a 6551 at some point, it is working well enough for now so it will be a lower priority than things like sound.

However, the serial data capabilities on my development PC (modern laptop running Win7) only suppports high speed USB. I had to buy a USB serial adaptor cable - which are commonly used for Arduino development. Not expensive - less than £3.

The 6526s together currently provide the following functions:

• TImer A in both 6526s is used for the serial input output timing. When I get a working 6551 in, this timer will become free
• In 6526-1, all 8 bits of Port A are used to enable the keyboard row and column addressing
• In 6526-1, 3 bits of Port B are used to control the leds on the BBC keyboard
• In 6526-1, 1 bit of Port B is used to sense the key pressed for a row and column
• In 6526-1, /FLAG is used to sense when serial input is available. Will become when I used a 6551
• In 6526-2, all 8 bits of Port A are needed to provide a bi-directional databus to communicate with the sound chip
• In 6526-2, 2 bits of Port B are needed to control the select signals to the sound chip
• In 6526-2, /FLAG used to sense when the BBC keyboard hardware strobe senses some key has been pressed.

When I use a 6551, then I will move the keyboard signal from 6526-2 to 6526-1. What way all keyboard access is through a single CIA (somehow feels a bit neater / elegant, although has no material effect on the code structure!).

I have now replaced the 6526 based I/O with an ACIA. And also added a keyboard from an old BBC Micro. The BBC Micro keyboard is actually very cool - it has some basic circuitry which means it scans for a key press in hardware, generating a flag which can be then used to cause an interrupt. The interrupt service routine then needs to check each row and column - but this is really fast and does not need to be done very often (relative to a 2.7MHz 6502!), so a whole key scan is not done any more often than 0.05 of a second (3 vertical blank intervals).

Further update - as noted elsewhere, I found that he 6526s are not reliable at 2.68Mhz - in nearly everything they appeared to operate as expected. Only when I was trying to read the joystick inputs from the AY-3-8910 was it clear something was wrong. After some confusing with chip select and Phi2 timing, I have managed to replace the 6526s with 65c22s - so now everything works fine at 2.68Mhz.

One thing I have observed with many homebrews is the lack of video output - they get by with serial I/O only or interfacing to hex LED or dot matrix LCD panels. This would not be enough for my needs, as I wanted something which would be able to drive a TV, ideally with graphics capabilities to render simple games.

A reason for lack of video on many homebrews is that old graphic chips are hard to design for (at least for someone of my low level of expertise) and hard to come by. Modern graphics chips are way too complex to be able to interface to my homebrew - and totally would dilute the 80s feel that I was after.

I looked at a number of options. One of the early options was to 'bit bang' video! Yes, I calculated that if I drove the 6502 at 8MHz, I could probably drive a display with about 176 pixels horizontal resolution. Anyway, I didn't spend too much time on it - the cycle counting was critical and even if the display did work, there was absolutely no room for future expansion (e.g. colour, sprites, resolution). I was partly inspired by Quinn Dunki (see http://quinndunki.com/blondihacks/?p=955) - although she was using an ATmega and outputting VGA, so it was a much more sophisticated bit bang solution than mine. So, due to effort vs limited reward, I ditched that approach [Edit - there are cleverer ways to use the 6502 address lines to generate video signals, but would have required additional logic which I never considered and got too complicated to generate colour etc.]

Next, I looked at 6845 chips. These were used in a few popular computers in the 80s including the BBC Micro and Amstrad CPC. A highly configurable chip, but it still needed a bunch of supporting electronics (character ROMs and additional custom logic for colour) - it just looked too complicated for someone of my limited experience.

So finally, I settled on the Texas Instruments 99xx VDP, as used in the MSX computers of the early to mid 80s. I acquired both the TMS9918 and TMS9928a - almost pin compatible but the '28 produced component video whereas the '18 provides composite. These are great graphics chips for their era (late 70s) - however the integrated nature comes with a big con. They are design with the assumption that they connect to DRAM. As I mentioned before, DRAM is a pain and hard to get hold of - and it was one of the reasons why I was originally put off by using the TMS99xx. But there are so many clever people willing to share their knowledge, and I found a resource on the web to adapt SRAM for use with TMS99xx VDPs - a fantastic write up by Tom LeMense entitled "SRAM Replacement for TMS99x8 VRAM".

Sound

I have added an AY-3-8910 made by General Instruments for sound. This is another nostalgic chip for me as it made an appearance in my first home computer, the Oric-1. I've described more about it in the sound blog. I will use the 10 input/output lines to add 2 joysticks (each joystick needs 4 direction signals plus the fire button).