Sat Nov 10 15:12:58 CET 2007
i'm not going to waste time on trying out a pure square wave
modulation. let's stick with some simple demodulator, and have a look
at the numbers.
i currently have the debug net running at 8kHz. this should also host
the filter tick, which consists of:
- read adc + update filter state
- once every x samples, wake up the detector tasklet
what if i start out with using a FSK, because it requires no
synchronization, and use a square window where the 2 frequencies are
placed each other's zero.
so.. square window does have perfect rejection for the harmonic
frequencies. it's only the stuff that lies inbetween that is
problematic. ok, this is obvious.
the problem can be entirely moved to synchronization and linear
distortion due to transitions. if the receiver listens during a steady
state part of the signal, perfect rejection is possible. so the main
- how to synchronize?
- how to limit transitions?
which brings me back to PSK.. maybe it is just simpler to use? as long
as the start of a symbol can be detected (threshold) and the phase can
be corrected (preamble) the rest seems not so hard really.
again, from a different angle: demodulating PSK is a synchronous mixer
followed by a low pass filter. i assume that a rectangular window is
going to be good enough as an LPF, which just leaves the problems of
signal detection and synchronization.
if i leave the non-synchronized receiver on constantly, it outputs a
24 bit complex number. during synchronization this needs to move
toward zero. the phasor will rotate once per window length. which
direction? if the direction is known, it's possible to detect a
crossing. the direction is determined by the rotation direction of the
i need to up the frequency: 2 MIPS is not enough. maybe i should do
that first.. then the output stage then the receiver then a decoder.
* have a look at PSK31 demod code
* build the output stage (either PWM or SD)
* build board + move to 40 MHz (monday: can't find xtals, maybe test
on the dude?)