OwenDuffy.net 

PA0V described a 300W linear amplifier module for 144MHz using the Philips BLF278 dual FET.
If we assume that the peak RF voltage swing at one drain is 44V, the peak RF voltage draindrain is 88V. Allowing for a loss of 10% in the output circuit, the FETs need to develop 330W with 88V pk, so Rl=88^2/2/330=11.7Ω draindrain. A 1:4 balun is a candidate for the output circuit.
Above is a plot of the recommended series load impedance from the datasheet from 75 to 150MHz (centre=5+j0). It can se seen that the points fall approximately on a constant G(=0.12S) circle. The points suggest that the drain has a nearly constant shunt capacitance of about 235pF. So, additionally to providing the load resistance calculated above, the output circuit needs to ‘tune out’ the 235pF or so of shunt capacitance at each drain. (The datasheet gives a recommended load for 300W out with 50Vdd of G=0.12 per section or Rp=16.6Ω draindrain. The 11.7Ω calculated above is for 330W with 48Vdd.)
PA0V has used a nominal 1:4 transmission line transformer using 50Ω coax for the transformer. In that case, the coax sections will have standing waves which will alter the transformation ratio somewhat and compromise symmetry.
Another approach is to design for symmetry by striving to achieve uniform current in the inner conductor, therefore I=Vl/50A where Vl is the load voltage. For uniform current, Zo of each coax section must equal V/I for that section.
Considering then an ideal 1:4 transformer, the draindrain voltage is half the load voltage, and the drain voltage is a quarter of the load voltage. It can be seen that the differential mode voltage of the:
The Zo required for uniform current is V/I, which for each section respectively is:
The 12.5Ω section could be fabricated by parallel 25Ω coax, and the upper section could use 35Ω coax.
Having dealt with the design of the inside of the transmission line sections, let us look at the outside, the common mode current. The outer surfaces of the lines form a s/c transmission line stub that shunts the drain with an inductance that can be chosen to offset the drain’s capacitance. An inductive reactance of about 4.7Ω is needed for each drain. The electrical length of the stub is given by le=atan(X/Zo) where Zo is the characteristic impedance of the line formed by the coax outers and tracks soldered together, and the PCB ground planes.
Modelling a trefoil of UT141 coaxes on a 10mm wide track on double sided 1.6mm FR4 board with atlc, indicates Zo of 21Ω and εr=2.7. It is not clear whether atlc properly accounts for proximity effect, so the value for Zo might be low. Assuming it is correct, for a reactance of 4.7Ω, le=atan(4.7/21)=12.6°. Length=le/360*λ/εr^0.5=12.6/360*2083/2.7^0.5=44mm. Choice of a slightly shorter stub would allow a trimmer capacitor in series with the 50Ω output to tune the output stage for maximum power output at the operating frequency, but note that the trimmer capacitor current at 300W is 6A.
This is a ball park design, the estimate of effective shunt capacitance of the FETs is not of high accuracy. Nevertheless, it explains the operation of the output circuit.
A design along these lines operates the coax sections with VSWR (internally)=1, giving uniform current independent of frequency and a matching scheme that is broadband. Obviously, the method used to tune out the FET output capacitance is frequency dependent.
Note that PA0V’s design appears to have used 50Ω coax for both sections which must result in significant standing waves on the sections and they will work a little differently. The effect of the standing waves will be to transform the external load to a more inductive load at the drains, and the length of the stubs and series tuning capacitor is part of that design.
© Copyright: Owen Duffy 1995, 2017. All rights reserved. Disclaimer.