Overview
When I was designing the PC104 Dragonball project, I wanted to design a simple (cheap!) PC104-based module to check I'd got the footprint and mechanical properties correct.
This project is based around the RTL8019AS ISA Ethernet Controller. It's mostly very similar to a very typical RTL8019 reference design, but I was able to add support for 10Base-2 and AUI.
10Base-2 support is implemented via a DP8392C (again, closely following the reference design).
There's also a jumper-selectable option to add a DS2431 1-wire EEPROM to the ISA IOCHK pin (somewhat nonstandard - it's based on an idea I had a while back for adding Plug'n'Play support to boards that didn't support standard ISA PnP).
A few things I noted while designing this board:
- There wasn't enough space on the board to add DIP switches to select the address/IRQ, so Plug'n'Play or EEPROM mode must be used
- During some initial bringup tests, I wasn't able to see an Ethernet link at all, but programming the EEPROM resolved this
- 16 bit PC104 connectors are quite difficult to solder, as the inner row of pins tends to be hidden by the outer row.
- I had tried to allow for push-fit PC104 connectors on this board, but the diameter I used for the connector holes was incorrect
- KiCAD's default PQFP-100 footprint doesn't fit the RTL8019 very well.
- I'm not sure if it's due to the RTL8019 or the KiCAD footprint being incorrect, but the first revision of this board was almost impossible to assemble
- The revision attached to this page is OK.
- The footprint for the BNC connector seems to be slightly incorrect, although it's still just about usable.
- There's a jumper (JP103) which allows a 512kB Flash to be used as either 1x512kB (fitted on pins 2-3), or 2 separate 256kB images (link disconnected to select A18=0, or link fitted to pins 1-2 to select A18=1).
Status
An earlier revision of the board had a few small issues, but was able to successfully send and receive Ethernet frames over the 10Base-T / 10Base-2 / AUI interfaces when tested with an x86 based PC104 processor board running DOS.
Annoyingly, this processor board is currently damaged, so I've not been able to fully test this revision.
A few basic tests with PC104 Dragonball have seemed to be reasonably promising - the board can be detected, the 10Base-T link seems to work and I was able to program/verify the onboard 93C46.
LED polarity for TX and RX might be incorrect.
It's not been tested any more than that so far!