Referring to John KD4IDY's note -

I cannot help but feel a bit uneasy about the use of class D
amplifiers followed by conventional shunt C series L pi-section
filters - although the designs using them do seem to work OK.  The
shunt C is surely going to increase the circulating current at the
harmonic frequencies, stressing the active devices more an lowering
efficiency.  I would have thought that the first component of the
output filter should be an inductor - either giving a Tee section
filter, or Pi section with an even number of elements.  Dave / Dave

can you comment further ?.

This approach is preferred by G8GSQ who has built high power HF Mosfet
PAs and finds the initial L in the output filter improves device
stability considerably.  QEX ran an article on HF PA filters a few
months ago, and these ended up as quite complex diplexing units in
order to present the devices with respectable out of band impedances.
A tank circuit actually sounds quite attractive on first thoughts just
as a filtering element, but unlike a valve, we are dealing with much
lower impedances.  Back in the valve class C PA days, (I'm just old
enough to remember these from several Practical Wireless articles)
seem to recall a loaded Q of 12 being taken as optimum.  To get this
sort of Q at a 50V rail 10A current needs a seriously small value of
inductor in a conventional tank.

A back of envelope calculation :

If  Rload = 50V / 10A = 2 ohms and   Ql = 12  then     Xl = 2 / 12 =
0.17 ohms  = 194nH at 137kHz.  Which is one turn (or less) on a decent
sized iron dust former or a few turns of VERY thick wire in air.
Furthermore, with I = 10A and Q = 12 then circulating currents
approach 120 Amps.  Enough said.    In practice, tapping into the tank
circuit would be used and the easist way to achieve this is to use two
capacitors rather than try to tap into turns with high current wire,
so now a choke is needed to feed in DC.  An alternative to tapping is
to use a transformer, which also has the advantage of isolating the
antenna circuit from the PA at DC.   We now have a transformer PA
followed by a resonant network just providing filtering - Beginning to
look familiar ?

In the early days of 73kHz I tried this approach, but the high Q
output network was too prone to destroying devices when mistuned, and
particularly when the antenna went off resonance.  The ultimate design
was a half bridge circuit running off 340 Volt rectified mains, output
transformer and C/L tank using an air wound inductor rather than
conventional Pi filter.  For a time it delivered 700 Watts into a
dummy load of multiple 150W light bulbs quite happily - until I keyed
the circuit a few times.  Remember what the resistance of light bulbs
are when cold and the current surges when switched on.  There is a lot
of energy in 1000uF of capacitors charged to 340V.  Never did like CW
!   After blowing four IRF450 fets spectacularly plus two bridge
driver chips I decided this was rather dangerous and kept with
conventional class B designs from then on.   But now that really high
power is wanted, it may be time to look at using rectified mains
again.  The output transformer provides one side of a safety barrier
and input isolation can come from a transformer or opto isolators.
An input transformer is preferable as it protects preceeding devices
from large DC voltages/currents generated as a result of FETs
exploding.  Switch mode PSUs have been doing this for a long time with
a good safety record and there is a large base of SMPSU components
both surplus and in the catalogues.  Everyone uses SMPSUs these days.

I don't have a death wish, so will leave such a PA  to others to
develop.

On the last point of John's note...

At the moment I'm developing a Pulse Width Modulator for supplying
Class D PAs.  This will accept a binary word (which can of course be
the output of a A/D converter) and control the supply voltage to the
PA.  The aim is for a supply which will operate with 50V input and
supply 0 - 50V out at currents up to 10A.   Software authors could
hopefully be persuaded to make phase and amplitude available
separately, digitally, or these could be regenerated as per John's
suggestion.  The PWM scheme is open loop so can operate very fast -
I'm using a switching frequency of 50kHz but this is not critical.
Modulation up to a few kHz should be feasible.  Works OK so far at up
to 3.5 Amps output, but I haven't tested at full current yet not
having had any suitable switching devices in the 'junk box', but I
hope to in a few days time, so watch this space.........

I already have a PIC/EEPROM based  PSK31 beacon which supplies
separate amplitude and phase, and have tested this successfully on
1.8MHz at 8 watts, but with a 12V supply rail just being very
inefficiently modulated with an emitter follower - thus negating all
the efficiency advantages of a class C/D PA !  It proved the concept
though.

Andy  G4JNT

From  John KD4IDY


>The output of any dual-ended or bridged switching amplifier does not

require >the flywheel effect of a traditional tank circuit. The output of the

amplifier is simply a square wave, from which everything but the

fundamental can and should be stripped (with impedance transformation, where necessary). Unlike single-ended switches, which require resonant circuits with a certain minimum loaded Q to opreate at all, dual-ended switching amplifiers can be built to work over roughly an octave without retuning.

Side note: If a "linear" amplifier is desired with extremely high

overall efficiency, one solution at medium power levels is to use envelope-elimination-and-recovery. (That was the late Helge Granville's name for it at Motorola, though I suspect there are other terms in use as well.) The signal to be amplified is passed to both an envelope detector and a hard limiter. The hard limiter output becomes the RF drive for the switching-mode final amplifier, be it Cass D or E or something else. The baseband/envelope voltage is amplified linearly, although this can be done at high efficiency too, in an amplifier switched with pulse duration modulation or other techniques. The amplified envelope becomes the source voltage for the final RF amplifier. Assuming there are no significant time errors between the RF and envelope paths, the output can be a highly accurate representation of the input, as the RF path preserves the phase information impressed on the

carrier and the envelope contains the amplitude information.  All

modulation schemes involve varying amplitude and/or phase, so the technique works for CW, SSB, PSK, AM, QAM, etc. There are some practical complexities, but I know amateurs who have used it successfully at levels of several hundred watts.

--
The Information contained in this E-Mail and any subsequent correspondence
is private and is intended solely for the intended recipient(s).

prohibited and may be unlawful.