>I have been
considering the key click problem when using BPSK with switching
transmitters. Traditionally, amplitude shaping is >used to limit the
transmitted bandwidth of PSK signals, but this is not a valid solution for class
E Txs . Dave's idea of going part >way there by ramping the PSU to the driver
chips to generate Pulse Width Mod being only a temporary solution.
PWM of the >drive waveform does not linearly control the amplitude of the RF
component which varies with pulse width in a SIN(X) / X
manner.
>Has anyone any
experience of systems that ramp the phase slowly in lieu of the amplitude
? My feeling is that this will give a >similar result to
amplitude shaping, but the text books don't cover PSK bandwidth control
using a constant amplitude waveform.
Class C, D or E, or
any other nonlinear power amplifier, is fundamentally a wrong choice for handing
a PSK signal. There will inevitably be a "jump" in waveform at conduction
threshold and cuttoff. As Class E is specifically designed to spend least
time switching between "full on" and "cut off" then clicks (transient sidebands)
then clicks occur at all phase changes. Even if all driver stages are nice
and linear, the PA will "digitise" it to on and off.
For amateur
operation, one way could be that key clicks are acceptable, for high
efficiency power amplifiers, and weave this in to a band plan, having
a segment for "dirty transmitters". Those who generate key clicks should
be happy to exist alongside other clickers.
Another approach
could be to develop a more sophisticated power amplifier, with a Class AB1
section to handle all small signals, and Class E for higher level parts of the
emission. In New Zealand we have enough bandwidth to run SSB on LF, so
anyone building a big amplifier would look at including linear mode of
operation. I am contemplating building an amplifier with n+n pushpull
transistors, with at least one pair working in Class AB1 and others in unbiased
AB2 mode. The value of n is say 10 or 20 (no mucking around, at least 1 kW
is a target output). The idea is that an AB1 stage is active at all times
and easily handles low level, but when a bigger level arrives it is sufficient
to automatically commutate the AB2 devices. I have already run my existing
bipolar amplifier with no bias and it still gets readability 5 reports on SSB,
even though a local can notice a moderate drop in quality and dynamic
range. When hard driven in CW mode, the efficiency is very high (much
better than theoretical Class C). I have a good stock of bipolar switching
transistors, so I have a preference for using them rather than power MOSFETs,
but the suggestion of having mixed classes of amplification in a single
final unit should be able to be contemplated. There could even be a
mixture of bipolars and MOSFETs, or other devices, sharing the same power
supply, so long as an AB1 stage can operate in linear fashion. There could
even be different backoff in bias for clusters of devices, to get piecewise
cut-in and cut-out over the drive waveform. Some questions
are:
- how many of the
devices need to be biased to give sufficiently linear operation to reduce PSK
key clicks to a satisfactory level?
- is the phasing of
Class AB1 and Class E output signals such that the AB1 devices do not get
interfered with, or fried?
- does any reader
have large signal modelling software to assess the viability of the
suggestion?
73, Bob ZL2CA
