[Lowfer] Re: 136k amplifiers

[James Moritz]

Dear Bill, Lowfers,

At 04:01 29/08/2002 -0400, you wrote:
>1) When one drives a coax with a 0 ohm source the load at the other end has
>to be really close to the Z of the coax or the waveform becomes very
>distorted. This applies even when the wavelength of the signal is 50 times
>longer than the length of the coax.

Any reactive load (this includes antennas in general) will have an 
impedance which varies widely with frequency - so if you drive it with a 
square wave consisting of a wide range of harmonic components, there are 
bound to be large mismatches at some frequencies, and therefore complex 
waveforms depending on the load impedance and the transmission line 
parameters will be created - in other words, trying to drive a reactive 
load on the end of a transmission line with a square wave is going to cause 
trouble. In theory it is possible to construct a network that will be 
matched over a wide frequency range, but in practice it will be difficult.

Some simplification can be achieved by eliminating the transmission line - 
in this case, you are driving effectively a series resonant circuit 
consisting of the loop inductance and tuning capacitance with a square 
wave. The series resonant circuit has high impedance at all the harmonic 
frequencies, but resistive at the fundamental frequency, so the current 
flowing in the load is more or less sinusoidal , in-phase with the driving 
square wave, and generally well behaved. The trouble is, the loop antenna 
is only approximately modelled by a series resonant circuit over a limited 
frequency range - in the HF range and above, the loop will have multiple 
resonances, so significant harmonic currents can flow and cause radiation 
at high order harmonic frequencies - driving an antenna with a square wave 
is always going to be a lousy idea from a QRM point of view.

These problems are greatly reduced by putting the series resonant circuit 
at the transmitter end - in fact, it should be considered as an essential 
part of the TX PA. Since a series resonant tank has a high impedance at all 
the harmonic frequencies, it effectively isolates the PA output from the 
load impedance at all harmonic frequencies, so you only have to worry about 
matching at the fundamental frequency. It is just possible that, if the 
load somehow has a high capacitive reactance at HF, an unwanted resonance 
could occur at a harmonic frequency - however, connecting a pi section low 
pass filter of the normal type between the tank circuit and the load 
eliminates this possibility. The combination of series tank circuit and 
low-pass ensures that the PA always sees a well-behaved load impedance, and 
is also very effective in filtering out harmonics. It also helps to improve 
the efficiency, because by ensuring a sinusoidal load current, the current 
is always near zero when the output devices switch, reducing the transient 
power dissipation in the transistors.

As an experiment, I would take your existing square-wave PA, and add a 
series resonant circuit in series with the output. A loaded Q around 5 is 
about right; for a 50 ohm load, this means XL = XC = 250ohms, ie about 
215uH and 3.44nF at 185kHz. You should find that this gives "nice" PA 
waveforms provided the load is resistive at 185kHz, however much coax is 
involved.

Cheers, Jim Moritz
73 de M0BMU