Some links:

* capturing radio signals
  http://www.stevenhale.co.uk/main/2013/08/home-automation-reverse-engineering-a-worcester-bosch-dt10rf-wireless-thermostat/
  https://damn.technology/controlling-british-gas-wr1-receiver-arduino
  http://rossharper.net/2015/11/decoding-a-siemens-rcr10433-thermostat-signal-to-control-a-boiler-from-a-raspberry-pi/

IMPORTANT: the transmitter prefers 3.3V, it stops working after a
while on 5V.

Reveng
======

Sampled at 192k samples/second, found that the transmissions had 3
types of (presumably square) pulses:

======== ======= ========
name     samples duration
======== ======= ========
narrow       200  ~ 1ms
middle       300  ~ 1.5ms
wide         400  ~ 2ms
======== ======= ========

The two sequences are:

* ``nnnnnnWnWnnMnnMnnnnnWnnnnMnMWn``
* ``nnnnnnWnWnnMnnMnnnnnnnnnnWMMWn``

Actually no, the pulses are not ½ high and ½ low, they're shaped
weird. The source code has the correct widths.

Without Arduino
===============

The GPIO / XIO pins on NextThing's CHIP, even when driven via the `slow
sysfs interface`_, seem to be fast enough for our purposes.

See the ``sender-chip.pl`` test program.

If we keep the ``*/value`` file open, we get good enough timing. The
``sender-chip.c`` program works.