(Note -- I am writing the following mostly from memory, with some supporting reference materials, over 35 years after the fact, so hope I got all the details correct.)
Why did I design the ST-2900 Computer System? After a few years of using my SwTPc system, I decided I wanted a more modern, much more compact, and more economical 6809 system for myself, that supported 64KB RAM, with I/O and ROM occupying as little as possible of the address map (especially beneficial for OS-9), and could handle double-sided, double-density 5¼" floppy disk drives. I also thought that such a system would be of interest to other people, and since one like it wasn't available on the market at that time (1982), it could provide a business opportunity for me.
The computer that provided some of the inspiration for the ST-2900 was the low cost Radio Shack Color Computer (CoCo 1) that came out in 1980. Most importantly, it used the same Motorola 6809 microprocessor that I was currently using and really liked. The base system was all on one circuit board, and it used inexpensive DRAM, along with a support chip that made DRAM easy to use.
But there were many things that I didn't like about the CoCo, which is why I never owned a CoCo 1 or 2. It ran its 6809 processor at only 895 kHz -- I wanted a 1.000 MHz system, which would make it easier to calculate execution times of instructions, besides being slightly faster. (This was important for double-density floppy disk I/O.) The CoCo 1 could only display 16 lines of 32 uppercase characters. The RAM used by the video display left less memory for the operating system and user programs. The keyboard left much to be desired. Its only built-in I/O ports were an audio cassette interface, joystick ports, and a serial port that was bit-banged in software, requiring interrupts to be masked while a character was transmitted or received -- not good for multi-tasking under OS-9. Expanding a CoCo system (in 1982) was awkward -- if a person wanted to add "real" serial ports, parallel ports, and a floppy disk controller, they had to buy 3 separate cartridges. But the CoCo only had one expansion slot, so only one of these cartridges could be used at a time -- the Multi-Pak Interface wasn't available for the first few years. Although the new system I was designing would have limited expansion capability (to keep it compact and low cost), it would have all of the most important features built-in.
Other computers on the market at that time that may have also inspired me were several S-100 bus Single Board Computers (SBC) such as the SUPER/NET (1982) from Advanced Micro Digital Corp., the Big Board (1980) that many of my fellow computer club members used, the Chandler 6809 CPU (1982), and maybe some G-64 boards from Gespac. There were ads for a 6502 SBC (p.34) and a 6809 SBC (p.62 (p.22 of catalog)) in the November 1982 issue of the Canadian magazine Electronics Today International (ETI). National Semiconductor produced a system of small (3.3" x 1.6") stackable modules with 52 pins called the MA2000 series (pages 12-3 to 12-24) (1982), which didn't use any backplane. There probably also were others that I can't remember. Boards that came out after I had started designing the ST-2900 are the Ampro Little Board Z80 (1983), and especially the Megatel QUARK/100 (1983?) -- but I'm not sure if I saw ads for them before I had already finalized the design of the ST-2900.
I decided to split my new computer system into two (or more) smaller boards instead of having a single larger board, to allow various configurations. Putting the 6809, EPROM, DRAM, and 2 serial ports on the CPU board would allow it to run all by itself for some simple applications. The CPU board could be paired with a custom I/O board for control systems, or an FDC board could be plugged into the CPU board to add a floppy disk controller and parallel ports to make up a complete general purpose computer system.
A sheet of my notes and drawings (undated, but probably made in late 1982) shows the proposed layout of the CPU board on a 4"x4" (or 10cm x 10cm) board. It suggested five possible expansion boards (of the same size), and called the system the "Model 4x4". It also listed a "family of single board computers attached to a local network". But since Ethernet (or ARCnet or Token Ring) networking was still very expensive and not common at that time, I'm sure I was referring to ULCNET (Ultra-Low-Cost Network) as described in the October 1981 issue of BYTE magazine. ULCNET just used a modified RS-232 port to connect two or more systems in a multi-dropped configuration, albeit at fairly low speeds and short distances compared to modern networking. (However, I never implemented ULCNET on the ST-2900.)
When I subsequently realized it would be difficult to fit the CPU board onto a 4"x4" (or 10cm x 10cm) 2-layer board, I increased the size to 100mm x 160mm (3.9"x6.3"), which is the standard single Eurocard size. This was fortunately also a good match to the new 3½" floppy disk drives that were introduced shortly after I decided on this form factor. Because I didn't want a separate backplane to connect the boards, I omitted the Eurocard's DIN 41612 connector, instead eventually deciding on 60-pin (2x30) male and female board-to-board connectors/headers. Connecting the CPU and FDC boards together back-to-back, sandwich-style, would make a very compact system, yet allow easy access to all components.
In November 1982, when I was working on the detailed design of the ST-2900, 32Kx8 static RAM chips were not yet available. Even 8Kx8 SRAM chips were either unavailable or very expensive, and in any case, 8 of them would have occupied too much board area because they were initially only supplied in .600" 28-pin packages. The alternative was to use 8 of the newly available 64Kx1 DRAM chips, which came in compact .300" 16-pin packages, and were more reasonably priced. But these required address multiplexing, periodic refresh, and tricky RAS and CAS control signals.
Fortunately Motorola's Synchronous Address Multiplexer (SAM) chip (MC6883/SN74LS783 and later SN74LS785) was already available and used by the Radio Shack Color Computer. It provides DRAM address multiplexing, automatic refresh, and generates the RAS and CAS control signals. It also generates the E and Q clocks for the 6809E microprocessor, implements address decoding, and allows switching between ROM/RAM and all-RAM memory maps, all of which would otherwise take many chips to implement. Together with one other chip, it provides three chip selects of 32 bytes each, perfect for the three I/O chips I was planning to use. Most importantly, the I/O is located at $FF00 and above, leaving 63.75KB of contiguous RAM, which was much more than most other 6809 systems of the day.
I chose to not add the MC6847 VDG chip to generate video, instead planning to connect a separate CRT terminal via a serial port. This would provide a higher quality 24x80 (text-only) uppercase and lowercase display instead of 16x32 uppercase only, and would eliminate using precious RAM space for a video buffer.
I selected the new Signetics SCN2681 DUART chip for the serial ports, even though it wasn't software compatible with the Motorola 6850 ACIA typically used by 6809 systems of the day. I chose it because it provided two serial ports in one chip, included a 16-bit counter/timer and several spare input and output lines, and had 4 built-in baud rate generators, all of which helped reduce the chip count of the system. Other benefits were that both receivers were quadruply buffered, and could use automatic RTS/CTS hardware flow control to prevent receiver overruns at higher baud rates. It even allowed configuring a different baud rate for the transmitter vs. the receiver, and supported fractional stop bits. (A few years later RFM Microplex sold an SS-30C card with the 2681, the M108 Dual Asynch Serial Card.)
But when I tried to obtain some 2681 parts for my prototype I discovered that the chip was so new it was only being sampled to selected customers! (The downside of choosing parts announced in an advertisement!) Fortunately a fellow computer club member worked for a local company that had already obtained samples. When they received new engineering samples, he gave the older samples to me. They proved very difficult to get working -- I don't know whether some of it was due to hardware bugs in these early samples, or incomplete or incorrect information in the data sheets? Eventually, with lots of trial and error using different programming sequences, and getting newer samples to try, the 2681 started working.
I selected the MOS Technology 6522 VIA chip instead of the Motorola 6820/6821 PIA chip because it provided two 16-bit counter/timers and an 8-bit shift register (in or out) in addition to the two 8-bit parallel ports (with handshaking lines). Almost 40 years later this chip made it possible for me to add the new NVRAM/RTC board to the ST-2900.
I chose the Standard Microsystems Corp. (SMsC) FDC9229T/BT (p. 497) to partner with the Western Digital 1793 floppy disk controller chip to provide adjustment-free digital data separation and write precompensation for ease of use and reliability for both single- and double-density operation. I designed it to allow four 5¼" and/or 3½" drives to be attached.
All this design work was being done in my spare time while employed full time as a senior computer programmer. The company I worked for was struggling financially, and in February 1983 I was laid off. (They went bankrupt a few years later.) I was off work for about 6 weeks before being hired by another company, and used that time to accelerate my progress on the project. By March or April I had finished a preliminary design, built a wire-wrapped prototype, and got much of it working. (Unfortunately I can't find any pictures of it.) However, comments in some source code files I found indicate that the initial prototype only had 16KB DRAM and one 6850 ACIA instead of the 2681 DUART, and only ran test and diagnostic software, not yet the FLEX operating system.
Because this was my first microcomputer design, I had several experienced electronics engineer friends look at the schematics -- they didn't find anything significantly wrong with the design! It helped that I had spent lots of time carefully studying all the relevant data sheets and application notes from the chip manufacturers, analyzed the timing of the signals, and tried to design with safety margins in the timings. Work continued on the monitor firmware and a port of the TSC 6809 FLEX operating system.
In September 1983 the design was mature enough that I hired Jon Rescorla of Quantum Designs to produce the PCB artwork for the CPU and FDC boards, since I had no experience (or patience for) creating hand-taped artwork for PCB layouts. (PCB CAD was very expensive in those days, so was used mostly for very complicated boards.) I took a one month unpaid leave of absence from my new job to assist with this task. Jon had just started the layout when I received the '68' Micro Journal magazine issue with Peripheral Technology's ad for their new PT-69 single-board computer. I felt quite discouraged, knowing that their board would be stiff competition for mine, but decided to keep going, as mine was different enough that some people would prefer it.
After receiving a few prototype PCBs and assembling and testing them, I found that only a few minor changes were required to get them ready for quantity production. I over-optimistically started advertising in the Canadian magazine Electronics Today International (ETI) -- see page 61 (p.77 in .pdf) in the February 1984 issue -- and in the July 1984 issue of '68' Micro Journal. But there was still lots of work to do to finish writing and testing the ST-MON monitor code, the FLEX Conversion Package code, and writing four user manuals, and all this took much longer than I expected.
In February 1984 I quit my job to become self-employed on a full-time basis, which I did for the next 5½ years.
I received the first small batch of production CPU and FDC bare boards on 1984-May-25. In June I finally shipped the first bare board sets to a few customers, but with minimal documentation. The manuals (CPU, FDC, ST-MON 1.01, FLEX Conversion) weren't ready until the beginning of September.
Microware's OS-9/6809 Level I operating system became available for the CoCo near the end of 1983 for only U$69.95 -- and that price even included an editor, assembler, and debugger! This was a real bargain compared to what OS-9 cost for other systems. I bought a copy and hand disassembled various modules such as Boot, Init, SysGo, Clock, and much of the kernel, to see how they worked, and what modifications I would need to make to them to work on the ST-2900.
I decided to license SDisk from D.P. Johnson for the floppy disk drivers and format program. This reduced the time required to write the drivers, provided support for several OS-9 disk formats, and provided features to support other software that could read and write other floppy disk formats, such as MS-DOS. I also hoped the SDisk drivers would be a good selling feature.
Like I had already done for FLEX, I decided to create a conversion program to make the process of adapting the CoCo version of OS-9 to the ST-2900 easy for the user. Since ST-MON 1.01 could only boot from FLEX format disks, the first version of the OS-9 Conversion Package (December 1984) was actually a bootable FLEX program.
An improved version of ST-MON (2.04) was created that could directly boot OS-9. This allowed me to create a second version of the OS-9 Conversion Package (August 1985) that was on a bootable OS-9 disk, and could create bootable OS-9 system disks configured for the ST-2900, which was much more convenient.
Along the way I received lots of encouragement from fellow members of the WCCS computer club that I belonged to, as well as from members of the Motorola User Group (MUG) in the Seattle WA area. They also provided technical help, corrections and suggestions for improvements to the software and user manuals, and purchased some board sets from me. (If anyone has any information on Will Wilgus, who lived in Seattle in the 1980's, and was a great help to me for both the ST-2900 and DMC projects, please let me know, as I have lost contact with him. I believe his full name is Lee Maurice Wilgus, born December 30, 1949 to Maurice K. and Jane (Cobb) Wilgus.)
An ST-2900 system became my main development system, replacing the SwTPc system I had been using for the previous 6 or 7 years. Once I became familiar with OS-9, I started using it more than FLEX, as OS-9 was a far superior operating system.
Using a system with only floppy disk drives was becoming a pain, so I designed and built a prototype of a 512KB RAM-Disk board (but I can't find any pictures of it), and wrote FLEX and OS-9 drivers for it. (Fortunately I had included the EXXX signal in the ST-2900 CPU board design to accomodate it.) Jeff Peters did an excellent job of creating the hand-taped artwork for the PCB layout. Adding a production RAM-512 board to my main ST-2900 development system made a huge improvement to my productivity and enjoyment. However, I sold very few RAM-512 boards, so it was a financial flop.
Other accessories were also designed -- serial port expander, parallel printer driver, SASI adapter, SCSI adapter, and one or two RTC (real-time clock) add-ons -- but none of them were put into production. In hindsight I wish I had put more effort into fitting an RTC chip (with battery backup) into the ST-2900 design. (37 years later I designed two new RTC add-ons for the ST-2900, one using the Maxim/Dallas DS2417 chip, and the other using the Maxim/Dallas DS3234, although the latter was eventually replaced by the Micro Crystal RV-3149-C3 on the new NVRAM/RTC board.)
Sales of the ST-2900 boards were slow, but came from all over the world -- I eventually had customers in at least 14 countries on all continents (except Antarctica). Although most were sold as bare boards (plus ST-MON in EPROM, and the FLEX and/or OS-9 Conversion Package on disk), I did eventually also sell several partially assembled as well as fully assembled and tested board sets. One repeat customer bought assembled and tested board sets that omitted the floppy disk controller, and had other modifications, that he used in some kind of telephone system. He typically bought 5 or 10 sets at a time, and the income from this helped keep me in business.
I did preliminary work on follow-on products, such as a 3 MHz HD63C09 system with MMU/DAT (refer to the explanation in the SwTPc MP-09 assembly instruction manual) capable of running OS-9 Level II, that I tried to fit onto 3 or more tiny (business card size) boards! But none of these designs were ever completed or put into production.
In August 1986, as ST-2900 sales were only trickling in, I started developing a non-related product, the DMC "No Halt" Dual Mode Floppy Disk Controller for the Radio Shack Color Computer.
How many ST-2900 boards were sold? Somewhat over 200 of each of the CPU and FDC boards, and maybe a dozen RAM-512 boards, between 1984 and 1992. Not enough to make a proper living, even though I worked at it full-time for a few years, with only a little bit of other income on the side. I was surprised to sell a few more bare board sets years later in 2007, then in 2017 sold all my remaining inventory of ST-2900 boards and components to a customer in the UK, except for three assembled board sets.
Looking back, I wish I would have developed a second-generation ST-2900, with a 2 MHz 6809, 64KB of static instead of dynamic RAM, 4 serial ports, a buffered parallel printer port, a SASI or SCSI hard disk drive interface, and a battery-backed RTC. This would have better competed with Peripheral Technologies' PT69-5. Or maybe I should have tried to squeeze an MMU/DAT onto the RAM-512 board to allow the ST-2900 to support OS-9 Level II? But the affordable (U$79.95) CoCo 3 version of OS-9 Level II wasn't available until late 1986, and it would have been more difficult to port to my system than Level I was.
But wait -- the last chapter of the ST-2900 story hasn't been written yet!! Over the past several years I have been working hard to breathe new life into it. On the hardware side, with a slight upgrade, its serial ports can now run at 115200 bps. The new USB-XR-RS232 adapter lets it easily and reliably connect to computers that have only USB ports. The new NVRAM/RTC board provides non-volatile RAM-Disk and EEPROM-Disk partitions that can be booted from, enabling compact systems with no attached floppy disk drives. It also provides an optional battery-backed real-time clock (RTC) to keep time. On the software side, a new version of ST-MON will allow it to boot from various sources, including from remote disk images hosted on a PC, using my new DSKdrv/DSKserv software, as well as with DriveWire and FLEX-OS9-Net. The old FLEX Conversion Package will be replaced by the new 6809 FLEX Plus 10TM package with a completely new set of drivers and utilities, and modifications to the operating system that support 10 logical drives and file timestamps. The old OS-9 Conversion Package will be replaced by a completely new set of drivers and utilities, and even a modified OS-9 kernel. (Note -- all this software is still under development; I can't/won't promise any completion date, but significant portions of it are already written and working.) I also hope to eventually give the ST-2900 a new baby brother, the 6809-Tiny6, that will share much of this new software.
In case you're wondering, the "ST-2900" name is composed of "ST" for Sardis Technologies, "2" for a 2-board system, "9" for 6809, and "00" to allow 01..99 for names of related products.
If you read this entire story, thank you for your interest!
"FLEX" was a trademark of Technical System Consultants (TSC).
"6809 FLEX Plus 10" is a trademark of David C. Wiens, dba Sardis Technologies.
"OS-9" is a registered trademark of Microware LP.
"Radio Shack" and "Color Computer" are trademarks of Tandy Corp.
"MS-DOS" is a trademark of Microsoft Corp.
Last revised 2024-Jun-16 13:55 PDT.
Copyright 2019- David C. Wiens.
