The Commodore 64, or C64 from here on out, was worth every penny. I had enough money left in my budget to purchase one Commodore 1541 floppy disk drive to go along with the C64. It was either that or else use a cassette tape recorder, and I’d come to absolutely loathe using cassettes for external storage due to my exposure to them with the TI 99/4a. If I’d tried to purchase the functional equivalent in Apple hardware it would have set me back between two and three times as much as I spent on Commodore equipment.

I didn’t purchase my C64 for games. I purchased it to act as a development system for other 6502-based DIY computer designs. To support development I wire-wrapped an external expansion board for my C64 that would program EPROMs and implement a real serial RS-232 interface. That board consisted of two MOS 6522 Versatile Interface Adapters (VIA) and an MOS 6551 Asynchronous Communications Interface Adapter (ACIA). I used an ACIA because I was absolutely horrified by the bit-banging the C64 used to do serial communications; the ACIA guaranteed serial communications to at least 19.2 kbits/s. While I started utility programming with the built-in C64 Basic, I quickly transitioned to a 6502 macro assembler for both utility tools native to the C64, and then later to do programming for my other 6502 DIY designs.

I purchased my C64 and 1541 in March of 1982. By the end of that summer I’d finished my C64 expansion board for programming EPROMs and had begun to development on the board you see below.

The design you see now is what’s left of a five year on-and-off evolution, before I stopped to work on other paying designs and began to pull parts off that board to support those paying designs. At the height of its operational glory my 6502 board had two RS-232 serial ports, 48K of static RAM, and space for 11K of EPROM. There’s 1K of the address space set aside for all the peripheral chips on the board. Before I explain how I did that, let’s go through a list of the components on the board.

• W65C802P-4 — A Western Design Center 65C802 at 4 MHz. This chip replaced the original NMOS chip I started with back in 1982. The manufacture date is 8542, so this chip was made in 1985, which fits with the time I purchased and installed it on the board. It was my intention with the original board to push the CPU clock as much as I could. I never went any faster than 4 MHz.
• R65C22P2 — A Rockwell 65C22 VIA chip. I swapped this in for the original MOS device to correct a logic bug with the original MOS 6522.
• WD1770 — A Western Digital single-chip floppy disk controller. The disk interface connector is on the far left edge. I managed to write 6502 assembly routines that could read and write MS-DOS formatted 3.5″ 720K floppy disks. The VIAs above the 1770 provided additional drive selection and control. I pulled the 1770 for another project where I built a stand-alone prototype 3.5″ floppy controller board using a Motorola 68HC11 micro controller for a customer.
• SCC2692 — A Signetics Dual Asynchronous Receiver/Transmitter, or DUART. It replaced an MOS 6551 ACIA. It came with two serial channels with on-chip ring buffers that could go up to 115 kbit/s. There were also some GPIO pins as well on the chip.
• HM6264 — An Hitachi Memory 6264 organized as 8k x 8. I had six to start with that ran contiguously from 0x0000 to 0xBFFF.
• DS1225 — Dallas Semiconductor 1225 non-volatile memory organized as 8k x 8. I used two to replace the Hitachis from 0x8000 to 0xBFFF. The monitor I’d written would take S-record files from the RS-232 port, convert them to binary, and store them at the starting address defined in each S-record. I used those to store my version of Commodore Basic I’d disassembled to source from the C64. Once the source was assembling to matching binary in the C64, I modified the source to call my hardware I/O routines, and was able to program my board in C64 Basic.
• TBP28L22 — Texas Instruments TBP28L22 programmable read-only memory organized as 256 x 8. I programmed a bit pattern that would produce a single high bit output dependent on the address input. It was my I/O device address decoder. The three highest address bits of the 6502 were decoded with a 74F139, giving me 8 x 8K regions. The 28L22 decoded all the I/O on the lowest seven output bits, with the highest bit being the final address decode for an EPROM that started at address 0xD000.
• HN462532 — Hitachi HN462532 EPROM organized as 4K x 8. I originally had four but pulled one for another project. The highest 8K was the full monitor and DOS, while additional applications were in the 7K that started at 0xD000. I’d intended to program my hacked C64 Basic into that 7K area, but never got around to it.

As I wrote earlier, all of this was started back in 1982. According to what notes I still have, I stopped completely around early 1988. My C64 along with my hand-built EPROM programmer was stolen Christmas 1984 while my wife and I were living in a rental in downtown Orlando. I’d gone out of town on business and my wife drove up to see her mother in Ocala. I eventually rebuilt my 6502 programming environment around a DEC Rainbow under MS-DOS. I purchased a commercial EPROM programmer which I used for nearly a decade and now sits on a closet shelf (how I wound up with a DEC Rainbow is a story for another time).

At this point in my life I’m still building embedded systems, but instead of a 4 MHz 8-bit microprocessor I’m working with ARM, Tensilica, and RISC-V 32- and 64-bit embedded MCUs running at hundreds of MHz; instead of EPROMS all of my development boards come with megabytes of reprogrammable FLASH and megabytes of external SRAM. And instead of assembly, I’m programming them in Micro/Circuit Python, C/C++, and Rust. Each MCU comes equipped with I/O, timers, ADCs and DACs, and so much more. I’m glad for my 6502 experiences from the 1980s, but I have absolutely no desire to try and bring any of that back up again. In 1982 I’d have given my right arm to have 1/10 the capabilities I currently have with today’s low-cost boards, which run from $4 for the M0/RP2040-based boards to $10 for RISC-V-based boards, to $20 for the dual core LX7-based boards and M7-based boards.

There is no going back for me, technically nor romantically.

I have carried this with me since I created it, starting in 1980, through my dating with my future wife, our marriage, and on down to my current home in Orlando. It’s been sitting in its case in the garage until I went out today, pulled it out, and took these photos of it. It is a single board computer (SBC) designed and built around a 6502 processor, the same processor that wound up in the Apple ][ and the Commodore computers of the era (PET, VIC-20, and C-64).

And believe it or not, I can still power it up and it still works. More or less…

This top-level view gives a better idea of the components and the density of the components. Again, keep in mind that this was built on a proto-board from the company I worked for at the time, Digital Communications Associates (DCA) of Atlanta. I was an engineer working for them (first a customer engineer, then a software engineer, then a field engineer; that last position is how I got to Orlando). It was interesting building this board because DCA used the Z-80 as the core of their SBC, then later, the Motorola 68000. I wanted to recreate, in essence, a cross between the Apple ][ and the C-64.

One reason I hang onto this board is the one lone chip from Rockwell International (you can see the logo on the lower 6522 chip), the conglomerate that made the Space Shuttle. Rockwell also had a chip division, and I picked up a kit of their industrial grade semiconductors back in the day. This board was initially meant for harsh industrial environments, not shirt-sleeve office areas.

Here are the specifications:

• 4MHz 65C02 (before I put the board away in the garage, I upgraded the processor to a 65C802 and began to dabble with the extended 16-bit instructions).
• Full 64K decoded address broken up between 48K SRAM, 16K EPROM, with 1K I/O. The SRAM addressing lost 1K to the I/O block.
• Zero wait state/non refresh memory allowing for full 4MHz performance.
• Western Digital WD 1770 IBM format compatible single-chip floppy disk controller (long since removed for another project). The Rockwell 6522 provided additional I/O controls on the floppy connector (see far left edge).
• Phillips Electronics SCC 2692 Dual Universal Asynchronous Receiver/Transmitter (DUART). These chips came with separate input and output ports as well as full blown UART control pins, baud rates up to 38.4K, quad buffer registers, and timers. I had two on the board at one time, one was taken off for another project.

When it was fully populated with all its parts and running at full bore it was capable of handling up to four 3.5″ floppy disks, four separate serial channels, and capable of running with either the kernel and monitor I wrote, or a hacked version of Commodore Basic (yes, I said Commodore Basic) from the C-64. I’d disassembled C-64 Basic with an application called Sourcerer, modified the I/O routines to use my serial routines in my kernel, then loaded it into the SRAM from the floppy drives.

I’d gone to the trouble to write 6502 assembly routines to read, write, and manage the floppies using the MSDOS file system. That meant I could use an IBM PC with 3.5″ floppies installed (720K only) as an intermediary to set up the floppies, writing files I could read via my SBC. I hated how Woz had created the floppy controller and totally incompatible floppy format for the Apple ][, and I was no fan of the Commodore 1541 floppy drive either. When it came time to add disks to my SBC, it was IBM format compatibility with MSDOS support all the way.

In this age of multi-gigabyte, multi-gigahertz cell phones, something like this limited and slow look ludicrous. But back in 1983 when it was finished it was a decent tool for driving other computers for testing. And when I finally added my modified C-64 Basic to the toolkit, it because even more flexible. It was the computer I wanted, something that was powerful and flexible and anywhere from four to eight times faster than the C-64 and the Apple ][ at that time. But I didn’t push to productize it because I already saw the writing on the wall with the 16- and 32-bit processors that were hitting the market. For example, in 1985 I picked up a Compaq Computer 80386 computer as a loaner for evaluation. Just two years after I finished my SBC, and one year after adding a hacked Basic to my SBC. And keep in mind that the IBM PC was introduced back in 1981, and it was based on a 4.7MHz 8088, a 16-bit internal processor with an 8-bit buss.

Just to show you how much a prototype it really was, here’s the backside, all wire-wrapped. Yes, I wrapped every wire on every post using a commercial wire-wrap gun.

It’s fun to pull out, and I could probably re-create this board with up-to-date parts and probably my old firmware source code reassembled for the new hardware. But that would detract my attention away from the 6502’s spiritual descendant, the ARM processor and all the products built with it, like the majority of today’s smartphones.

And I’ll be writing about that in fairly short order.

If you’re wondering where this came from, it’s a copy from my old blog on Blogger. The post was originally published 6 October 2013. I’m pulling this post over because it’s hardware related around the 6502, a chip I was heavily into during the late 1970s and early 1980s. And it helps to explain why I still do what I do today. I’ve been a hardware and software developer for many decades, going back to the late 1970s.

You’re looking at something I’d completely forgotten about, something I built around 1980. It’s a custom perf board I built on a slab of aircraft-grade aluminum, on which I wired a complete 6502-based embedded computer. It has 4,096 bytes of static RAM, 16K of EPROM (2716), a peripheral chip that included two serial ports and two eight-bit-wide parallel I/O ports, a fully decoded keypad and a six digit display broken into four digits for an address and two for data, all in hex. I was proud of those displays. They were special HP multi-segmented alphanumeric displays that read ASCII bytes. I had them display hexadecimal and special words, and then got them to scroll text. Although the keypad looks like it will only handle hex digit input, I had added software so that if you held the key down it would give you a display of the function and an alt function. Primary and alt key functionality was toggled by the tiny switch on the top left.

The calculator sitting to the right is a TI Programmer (hexadecimal calculator) for scale.

I used it for 6502 software testing as well as driving external devices. Software development was on an Apple II using a source assembler. The resultant binary file was written to the 2716 with an EPROM programmer. I was proud of the fact that the 6502 was clocked at 2MHz, twice as fast as the Apple II’s 1MHz clock speed. That and the fact its memory was fully static meant the system ran with no wait states. I used it for driving stepper motors at one point, and then built a light tracker and then combined all that into a simple wheeled line tracker. I had big dreams back then. I was going to build Asimov-like robots. I wonder what happened to all those dreams…

The image at the top a black and white Polaroid I took of the project all and sent off to someone I haven’t heard from in 32 years. He scanned that Polaroid and sent it back to me this evening.

If you’re wondering where this came from, it’s a copy from my old blog on Blogger. The post was originally published 1 November 2012. I’m pulling this post over because it’s hardware related around the 6502, a chip I was heavily into during the late 1970s and early 1980s. And it helps to explain why I still do what I do today. I’ve been a hardware and software developer for many decades, going back to the late 1970s.