Reverse engineered PCB of the never-released 1984 Dragon Beta machine from Dragon Data
The Dragon Beta was a machine in development at the time of Dragon Data's demise in 1984. Only three boards are known to have existed - and only one can be located now. The one board which remains is in the care of the Dragon Data archive at http://www.dragondata.co.uk. It is this latter board which has formed the basis for this current reverse engineering project. Richard, from the Dragon Data archive had made some very clear photographs of his Dragon Beta board before, during and after it's restoration. That board was returned to working order, but understandably Richard was not keen to continually use such a rare piece of computing history. The board was apparently rescued by the Dragon Data M.D. as he left the building for the final time. The Dragon Beta disks were still held by Duncan Smeed of 'Inside the Dragon' fame, and he supplied those to Richard for imaging.
Initially, and somewhat secretly, I (John Whitworth) began to trace the circuitry using Paint Shop Pro and KiCad. That is, I'd have the high-quality images of the desoldered PCB on the screen, and would slowly build the circuit in KiCad, whilst drawing coloured bold lines over the board traces that I'd already dealt with on the schematic in KiCad. The process took, in all, about 6 months.
During that time, as it started to feel less like an impossible mission, I revealed to a gradually larger and larger group exactly what I was doing.
It's ended up being a very closely knit Facebook Messenger team who have come together to get this board working - mainly Phill Harvey-Smith, Richard Harding, Mike Miller and myself. The original boards I ordered had a few mistakes on them. Some Vcc and GND pins not connected on ICs, as well as one of the select pins on the 74LS138, which is used to drive many devices on the board.
I managed to get my board built quickly, and it came up with the OS9 is loading screen - but nothing else happened. At this point, Phill also built his board, and we were to start diagnosing the issues together. There was very little actually wrong with the board as it was made. Just the power connections, the LS138 configuration, and the fact that I'd missed a little Dragon Data bodge on the back, which was a 4.7K pull-up resistor on one of the expansion port control lines. This was effectively throwing an interrupt request spuriously and stopping the machine from running. Mike has been a huge support to me throughout all of this, being my sounding board for many issues. I lost count of the amount of logic analyser images and videos I sent him!
Many of the components used on the Dragon Beta are obsolete - to the point of being almost unobtainable. Phill has gone to great lengths to create some good add-on boards which can be used to replace the archaic SRAM chips that handle paging, as well as replacing the 32 DRAM chips which provided the original 256K. These are replaced by a small daughterboard. He has also created a 512KB expansion RAM which plugs into the expansion port and effectively takes the machine to 768K of RAM.
Another archaic part of the Beta was its use of the original Sony 26-pin floppy drive connector. This is different to the Amstrad 26-pin connector. Phill has put together a 26-pin to 34-pin adapter, which allows the connection of normal floppy disk drives.
The last area for now is improvements to the keyboard interface. The original Dragon Beta used a matrix scanned/bit serialised keyboard. As with many other old machines, Phill put together a PS/2 keyboard adapter, and in addition now has a USB keyboard and mouse adapter (yes, the Dragon Beta had a mouse - it also had a light pen - but we will likely leave that back in 1984!)
The boards which are available at this repository are the slightly updated ones, which contain some ease-of-use modifications. Some of those are to allow Phill's add-on components to be plugged in more easily (e.g., header pins for control signals etc.), whilst others are about completing the expansion port wiring up and adding a few useful control signals to that as well.
Please note that this final board revision has not been produced or verified physically, but all electrical rules have been validated in KiCad, and although the changes have involved some detailed re-routing, electrically the changes are minor. Thus, it is believed that these boards will work just fine if produced and built - but this is without any guarantee or warranty from any of the development team.