Hello Chris and group,
I failed to store all the postings of this thread - so sorry if some
aspects of the following comments should have been already discussed or
could be boaring to all of you succesful users of semiconductor-minded
D- and E-designs.
Chris, your problem seems to be typical when trying to use real TX tubes
on LF/MF: the Vp/Ip ratio is rather high resulting in high impedances
and bulky air cored coils for the tuned circuits. If you want to
overcome this by using ferrite cores you run into new trouble: the high
voltages and currents in resonance mode due to the necessary Q require
large core cross-sections and can cause the danger of insulation
breakdown between the windings or between windings and core.
Single-ended circuits will cause detrimental DC premagnetization. The
fact that the relative permeability of ferrite depends strongly on its
temperature will force you into retuning the selective circuit again and
again as it gets warm by the dissipation.
About three years ago after some bad experiences with MOSFET-PAs I was
sick and tired of those nervous semiconductor parts and started to
construct a reliable and good-natured tube PA from scratch. Wanting a
linear PA and keeping in mind the above-mentioned stumbling blocks I
chose an alternative approach: broadband push pull class B without grid
current and with transformer coupling like an audio amplifier using 4
paralled (8 in total) TV line tubes (PL 519) in each branch. So I got a
low plate-to-plate resistance (about 1 kOhm) at a plate voltage of 600 V
DC and RF voltages about 1 kV between the two groups of tubes. Using a
stack of 4 recycled colour TV line output cores with a total
cross-section area of 610 mm² I got by on 2 times 28 turns primary
between the plates and a secondary winding of 14 turns with switchable
taps and contra windings for adjusting the load. To improve coupling and
reduce stray inductivity the windings were twisted and interleaved
similar to good quality audio transformers. For the windings I used RF
Litzwire TEX-ELZ 7x 0,3 mm from Furukawa with a dielectric breakthrough
voltage of 10 kV. At the end of the day I came out with a linear PA
giving a continous output of 800 W RF maximum on LF and MF at 50 Ohms
load with a plate efficiency of 73% on LF and 69% on MF. Without any
further filtering the antenna (and dummy load) current is purely sinus
shaped. A little disadvantage of the concept is the need for individual
bias to be set for each tube. Via the voltage drop at an 1-Ohm-shunt in
every cathode line I adjusted the individul cathode current to 250 mA DC
at full load - the setting kept stable since the beginning. Because the
ferrite core is used as transformer only with ample allowance there is
no problem due to variations of permeability when the temperature rises:
the transformation ratio keeps constant. Class B allows to form rise and
fall of the keyed CW signal according to its speed to avoid key clicks
in adjacent channels and can be used for future linear modes. Pictures
can be seen on my QRZ site www.qrz.com/db/dk1is.
If there are data for audio or RF push pull class B for your tubes or
you can deduce them from the caracteristic curves you will certainly
find higher plate-to-plate resistances as with my TV line tubes. But it
should be possible to make a suitable ferrite output transformer anyway
as long as you don´t use it as resonant tank circuit. Take care not to
run into self resonance due to the higher primary inductivity. Perhaps
you have to omit twisting and interleaving of the windings to keep the
self capacity low. At higher impedances the stray inductivity wouldn´t
be so important. The concept works with push pull only, single ended
circuits will cause problems due to the DC component magnetizing the
ferrite core.
73 es gl,
Tom, DK1IS
|