Dear LF Group,
There seems to be some potential for BPSK modulation methods
on LF, but the big difficulty is generating a TX signal with
acceptable sidebands - many people find the high-power amplitude
modulator, or linear transverter/PA required rather daunting. It has
been suggested that similar reduction in sidebands could be
acheived by having "soft" phase transitions with constant signal
amplitude, rather than the abrupt phase keying and envelope
shaping of normal BPSK. This would eliminate the high power
modulator, while allowing use of high efficiency, non-linear
amplifiers, such as class D types.
In order to test this idea, I have breadboarded a variable phase
modulator, together with waveform-shaping circuits. The circuit
implements 2 approaches - ON7YD's linear phase transition idea,
and G4JNT's raised-cosine phase transitions.
The phase modulator is essentially a voltage controlled pulse width
modulator running at 136kHz. The leading edge of the PWM output
pulse is delayed relative to the 136kHz clock signal by an amount
proportional to the input voltage. This leading edge is then used to
trigger a monostable, which generates a constant duty cycle,
approximately square wave, output whose phase lag relative to the
input signal is linearly dependant on the input voltage. The phase
modulation range is up to about 270 degrees; by clamping the
input signal level, a 180 degree shift is obtained.
Rik's linear phase transitions are generated by feeding the
incoming data waveform into a "bounded integrator" circuit; the
integrator charging current and capacitor control the duration of the
phase transition, the bounds on the integrator output voltage
control the phase excursion. The resulting trapezoidal waveform is
fed to the phase modulator. I found experimentally that a fair
aproximation to Andy's raised-cosine waveform could be produced
by feeding the trapezoidal signal through a 3rd order Bessel low
pass filter, and playing around with the time constants.
I looked at the resulting spectra by feeding the low level phase
modulator output to a RX with 1 kHz bandwidth, and examining the
audio output using DL4YHF's Spectrum Lab analyser software. I
tried linear transitions ranging in duration from 5ms, up to 100ms,
ie. the entire bit period. Transitions that were a small fraction of the
bit period did not offer significant improvement over the "raw"
BPSK, with sidebands extending to several hundred Hz. However,
with transitions of 50ms or 100ms, the bandwidth of the signal was
considerably reduced, with no significant sidebands beyond about
+/- 100Hz, which is about the same as a 12wpm CW signal. The
spectra had "spikes" at 10Hz intervals, presumably due to the
abrupt changes of gradient in the phase modulation.
As you might expect, adding the low-pass filter to produce the more
rounded raised-cosine transitions largely got rid of the 10Hz
spikes, and gave very clean spectra. I tried 50ms and 100ms
transitions; bandwidths were about 80Hz and 40Hz respectively.
The 100ms transitions gave a spectrum of very similar bandwidth
to my Decca TX BPSK signal; the main difference was that it was
more "peaked", presumably because there is some carrier present.
I have a .jpg file of the various spectra I obtained, plus 12wpm CW
and "conventional" BPSK for comparison; it is 107kB so a bit big
for the reflector, but available direct on request. The circuit isn't
really much more than a lash-up at the moment, but I will get some
details together if anyone is interested.
So having gradual phase transitions of 0.5 -1 times the bit period
certainly seems to produce acceptable transmit signals - the next
experiment is to see how compatible these "Variable PSK" signals
are with the existing BPSK demodulators that are available, such
as "Wolf" and "Coherent".
Cheers, Jim Moritz
73 de M0BMU