Many modern Amiga cards use Xilinx chips for everything from a basic FPGA for auto configuration to full blown CPU replacements. My Terrible Fire 534 accelerator has two of these chips and they are field programmable using JTAG. It turns out a Raspberry Pi can be turned into a programmer for these, so this is how to do it.
First of all you need a Raspberry Pi, preferably not version 4. There are ways to make this work with v4 but I haven’t tried them, I’m told they are slower too. I happened to have a spare Pi Zero W which I bought for another project. It needed the GPIO pins soldered on and I have some spare right-angle ones from a previous Amiga repair, so I soldered these on.
The SD card was flashed with Raspbian and setup with SSH access so I could remote into it from my laptop to manage things.
We then need to build the programming software, this is a fork of xc3sprog that supports Raspberry Pi:
sudo apt update sudo apt install build-essential libusb-dev libftdi-dev libgpiod-dev wiringpi git cmake git clone https://github.com/matrix-io/xc3sprog mkdir xc3sprog/build cd xc3sprog/build cmake .. make sudo make install
Finally we have to wire everything up. Use this table for the wiring map, I also used some very large wire tag labels to label each end.
|1 (or pin 17)||3V3|
|6 (or any Ground pin)||GND|
When wired up to the TF534 it looks a little like this:
Once we have both ends wired up, it is a case of running it. The following command should list all the chips in the JTAG chain, the TF534 has two of them. The “matrix_creator” uses WiringPi to talk to the JTAG:
xc3sprog -c matrix_creator -j
If you wish to download a Xilinx binary to your hard drive you can run the following (assuming you want to talk to item 1 in the chain). The “:r” part at the end means read the chip to file instead of write from file:
xc3sprog -c matrix_creator -v -p 1 myxilinx.jed:r
Finally to write a new image to the Xilinx (again to item 1 in the chain):
xc3sprog -c matrix_creator -v -p 1 newxilinx.jed
You should see something similar to this (screenshot from programming the TF534):
It all worked out really well and I now have a Xilinx programmer I can use for an upcoming project!