HF PA-to-antenna matching. Different thinking!

[G6Tanuki]

The 'traditional' way of matching a HF PA valve to its antenna has, for over half a century, been something like the 3-element Pi-tank circuit: two variable capacitors and an inductor.

I've never liked the combination of impedance-matching and filtering by the same components that this approach implies: the L:C ratios must inevitably be a compromise, and it always seems to me that there are multiple combinations of L and the two "C"s that give a dip of some form in the anode current but no indication which is the optimum one for either impedance-matching or harmonic-minimisation.

Transistor PA stages generally don't have user-variable tuning in their equivalent circuit (which often consists of a matching transformer to scale-up the very low collector-impedance to 50Ohms, followed by a 5- or 7-element Pi LPF with fixed component values).

My 'Different thinking' is - why not do something similar with a valve PA?

A multifilar transformer (toroid) with something like a 6:1 or 7:1 turns ratio to present the required 2K to 5K impedance (for typical PA valves in the up-to-100-Watt class) to the valve but step-down to 50Ohms, then a fixed-values Pi LPF.

You get the dual benefits of no need for manual tuning, and the easy ability to switch in band-specific LPFs whose impedance is the same at both ends so they can be built more-predictably.

Am I on to something or am I deluding myself?

[Radio Wrangler]

At those frequencies and impedances, the stray capacitances in the transformer become a problem. It can be done, but its difficult to get enough bandwidth for the HF bands.

Have a look at distributed power amps. An interesting way to get plenty of power and bandwidth out of a fair number of smaller valves. Racal did some.

David

[Jon_G4MDC]

Funny I was thinking on this the other day too.
My version was for something like a QQVO6-40A for a single VHF band (4m).
It ended in a 4:1 coax balun to 50 Ohms with PI or L sections between the input end of that and the anodes.
Would it have any advantages over the coil and butterfly capacitor with swinging link winding? Probably not but it would be different.

[G6Tanuki]

The issue of transformer leakage is understandable - hence my idea to use a toroid and multifilar windings so I could get the necessary transformer-effect coupling while at the same time using the fewest-possible-turns (so the lowest unwanted capacitance between the windings).

I've always found it odd that the typical HF transmitter has used a Pi-tank two-capacitors-one-inductor matching network followed by another 5- or 7-element Pi low-pass-filter [usually with a cutoff at about 35MHz] to constrain the TVI/FM-BCI-generating harmonics. A transformer-to-50Ohms at the source-end followed by band-specific Pi LPFs is surely simpler and more-predictable to design. The band-specific LPFs can provide a lot more attenuation of higher-order harmonics than a 3-element-Pi followed by a 'generic' nothing-beyond-30MHz LPF?

And it's a lot easier to design good LPFs when you've transformed down to 50 Ohms for both the source and sink impedances.

Another thought: make the RF 'power-amp' a cathode-follower. Its output-impedance is then already an order-of-magnitude or two closer to the antenna-impedance. In times-past I built a bunch of line-driver amplifiers intended to shove ten Watts or so into a 75-Ohm coax over the 35-70MHz range. It was a 'distributed' amplifier with multiple triode-connected EL86 valves interconnected along a L-C network in their cathodes, with a similar L-C network in their grids. It worked rather well, which I consider a credit to my several days spent sweating over Smith-charts aided by my then 'leading-edge' HP33C calculator!]

[Terry_VK5TM]

Back when valve transmitters were king, the filtering aspect was a side effect of the output matching, not a design requirement.

And yes, there are several combinations of L & C that would give an Anode dip, but experienced operators knew which one was the right one.

Also, the "50 ohm" antenna did not come into pervasive usage/terminology until solid state gear came on the scene (yes, I know it was around in WW2, but it took a while to filter into everyday life, coax was hard to get and expensive back then).

[Terry_VK5TM]

By pure coincidence, there is a Bernard's book, #61 Amateur Transmitter Construction from 1947, that has just appeared on the American Radio History site.

Interesting to look at the arrangement of valve PA stages and aerial coupling at that time (no Pi coupling arrangements there).

No mention of output filters at all either, whether the designs presented would comply with todays requirements of signal purity would be interesting to investigate.

[G8HQP Dave]

Originally Posted by G6Tanuki

Another thought: make the RF 'power-amp' a cathode-follower. Its output-impedance is then already an order-of-magnitude or two closer to the antenna-impedance.

This would not help. The optimimum load impedance - which is what you are matching the antenna to - for a cathode swinging 500V at 100mA is exactly the same as an anode swinging 500V at 100mA.

What might help is reducing impedance by paralleling lots of valves, and then using a little series L to compensate for shunt C by acting as a short lumped transmission line (although this can limit the max frequency). Also use a lower HT voltage than is common for PAs, and suitable valves such as TV line output which can pull quite high currents.

[G0HZU_JMR]

I've always found it odd that the typical HF transmitter has used a Pi-tank two-capacitors-one-inductor matching network followed by another 5- or 7-element Pi low-pass-filter [usually with a cutoff at about 35MHz] to constrain the TVI/FM-BCI-generating harmonics. A transformer-to-50Ohms at the source-end followed by band-specific Pi LPFs is surely simpler and more-predictable to design. The band-specific LPFs can provide a lot more attenuation of higher-order harmonics than a 3-element-Pi followed by a 'generic' nothing-beyond-30MHz LPF?

I'm afraid I know next to nothing about valve/tube PA design but I'd be very scared if someone asked me to try and design the high Z to 50R transformer you are describing. By contrast, I think I'd be reasonably comfy having an initial stab at throwing some numbers at a conventional Pi tank design even though I have no practical experience of actually doing this.

If 2000R was the desired anode loading then a basic L match to 50R would have a loaded Q of less than 4. A typical Pi tank might manage an unloaded Q of 6 or 7 but that is just a guess. But if the main inductor in the Pi tank had an unloaded Q of 350 the insertion loss with a loaded Q of 7 would be less than 0.2dB. Note that there are several interpretations of what loaded Q means and mine is the classic version used in modern RF design/simulation. So this assumes a 2000R source driving the network and this might not be the right thing for a high power valve design if it can behave more like a current source. So the loaded Q might actually be higher than my numbers above. If it was 12 then the loss might be just under 0.35dB.

For a typical design at 7MHz I'd expect to see maybe 30dB rejection of the second harmonic in the Pi tank network if the loaded Q was about 7. It might be closer to 35dB with a loaded Q of 12. So if the valve could manage 15dB rejection on its own then another 30dB or so would mean a t least a -45dBc 2nd harmonic might be a typical result. Maybe someone with some practical experience could massage these numbers to be more realistic?

To compete with this you would have to be able to design a 2000/50R transformer and a series of switched LPFs and try for an overall insertion loss of maybe 0.3dB. I can't immediately see how you could design a low loss, wideband, high power, Hi Z transformer like this but maybe someone else has been here before and can advise?

[G8HQP Dave]

It often turns out that the way that something was traditionally done is actually the best way to do it, in the absence of some new game-changing technology. Ferrites have of course improved since the 1950s, but the laws of electromagnetism are still the same.