I was inspired by Beth’s avrfid.S project to try to build a replacement for the multiple HID Prox cards that I carry for work. Her design is simultaneously a technical tour-de-force and an example of how badly we can abuse the Atmel chips. Here is the entire schematic:
There is no connection to power and ground: the chip is powered through leakage current from the input pins. The AC waveform is fed directly into the pins: the internal protection diodes rectify it. During negative parts of the wave the silicon die’s inherent capacitance maintains state. The CPU clock is driven by the AC as well and depends on the ability of the coil to drive more current than the chip when DDRB is configured to pull the pins to the same potential. It’s truly amazing that this works at all.
The firmware she wrote in macro assembler is easy to understand and straightfoward, but filled the entire 8 KB flash on the ATTiny85 when compiled for HID Prox cards. Unlike the CW modulated EM41xx cards that just load the coil for thirty RF cycles to send a baseband one and don’t load the coil to send a baseband zero, the HID cards use Frequency Shift Keying (FSK) modulation. In FSK a baseband zero is sent by cycling the load on the coil for 50 cycles at a frequency of 4 RF cycles, and a baseband one is sent by cycling the load every 5 RF cycles. Beth’s code loads the coil by setting the two bits in DDRB to 1 while holding PORTB at 0, which places a short across the coil by putting both ends at the same potential.
While it turns out that my dream of automatically selecting the right RFID card doesn’t work, read on for details of how to build your own HID compatible RFID devices and some overview of the hand-tuned assembly necessary to fit the RFID timing.