Vacuum Tube RF Amplifier Anode Choke Inductance

[W2DML]

Hello,
I'm currently  designing a grounded cathode tetrode power amplifier. I was wondering the importance on the inductance of the Anode Choke. Typically with solid state amplifiers I've learned to pick an inductor that has a reactance > 5* the load impedance. When looking at some similar tetrode designs I see a much lower inductance. IE. a 28 MHz amp has a output load of 1k, inductance of 4uH which @ 28 MHz = ~703 ohms. My simulations show power loss with such low inductance. In Eimac's "Care and Feeding of Power Grid Tubes" it states that if you want the RF current to be limited to 1% through the choke, it should be 100 times your tank L, which is ~40uH. Is this really the case or do things like skin effect help improve this value? I believe at 28MHz your real impedance will be higher since the majority of current will flow on the outside of the wire. Additionally to help my AWR Microwave Office simulate properly I am using a RF tank to go from the Load impedance to 50 ohms, in actuality my amplifier will be driving a RF cavity with impedance of ~900 ohms on the amplifier side at the Fc.

I calculated for my design a inductance of 40uH but it is not possible due to multiple constraints:

Voltage Rating needs to be at least 40kV
DC Current needs to be at least 30A
SRF needs to be at least 49 MHz

[Andy Chee]

Is this a multi-band amplifier? If so, your biggest concern would be constructing an anode choke that has sufficient impedance at all bands, whilst simultaneously lacking self-resonance at those bands.

[W2DML]

It is for two narrowbands, 24.6 MHz & 49 MHz. I would assume going for a higher srf can kill two birds with one stone.

[szoftveres]

In case you're asking about the anode VHF choke:
The choke typically comes in between the plate and load, and it is heavily damped (i.e. there's a low value resistor across it).
The purpose of it is to cut the RF path between the plate and the dangling wires (anode cap) at VHF frequencies, and the purpose is to extinguish VHF oscillations.

[W2DML]

In case you're asking about the anode VHF choke:
The choke typically comes in between the plate and load, and it is heavily damped (i.e. there's a low value resistor across it).
The purpose of it is to cut the RF path between the plate and the dangling wires (anode cap) at VHF frequencies, and the purpose is to extinguish VHF oscillations.

I'm talking about the choke that is between the Anode DC Power Supply & the Anode of the Tetrode. Similar to solid state where you have a choke from the Drain supply to the drain of your transistor.

[W2DML]

To help aid discussion I'm talking about the choke labeled RFC in the example schematic I found below. They also have two more stages of chokes and filtering caps.

[T3sl4co1l]

RFC in a narrowband design is kinda neither here nor there. It acts in parallel with the tuning choke, so some combination of the two achieves plate tuning.  How this affects the overall network (impedance, tuning, tunability for a specific target frequency or band switching) is something you'd want to check (simulate or measure).

BTW, the first resonance is parallel resonant, i.e. impedance peak.  The choke doesn't get any "chokier" than this. It's only the second resonance (series) you need to worry about (and subsequent modes -- which, neat fact, are not harmonic, because the helix is a dispersive transmission line structure).

(When you need a broadband choke, higher modes can be damped to some extent with ferrite or powdered iron, at the cost of core loss of course.)

Since this is a cavity structure, you might also just feed it in that way.  For a coaxial resonator with the tube at one end, the bypass caps are how the middle cylinder mounts (and RF short-circuits) to the other end.  This isn't going to be trivial at 30kV, but, I mean, none of this will be.

Unless those frequencies are harmonic in a given length resonator (I'd have to check?), you'll need some manner of telescoping resonator to run both bands.  This may encourage a lumped-equivalent design instead, but, that will be hard to do while respecting voltage and power ratings.

Previously, for reader reference: https://www.eevblog.com/forum/rf-microwave/sourcingdesigning-high-power-edge-wound-inductors/msg5522953/#msg5522953

Tim

[W2DML]

RFC in a narrowband design is kinda neither here nor there. It acts in parallel with the tuning choke, so some combination of the two achieves plate tuning.  How this affects the overall network (impedance, tuning, tunability for a specific target frequency or band switching) is something you'd want to check (simulate or measure).

If the RFC is high enough wouldn't it be negligible when it comes to output tune? It's why EIMAC suggests RFC to be 100 times the tank L? This seems to fix my simulation when I increased RFC from 4uH to 40uH, the RF power matches my load line computations. Talking to vendors it may be hard to design an inductor with 40uH inductance. When this inductance is essentially an RF load to gnd, it will present itself in parallel with the RF Cavity circuit, taking power from the cavity? At resonance the cavity will present itself has an impedance = (Rshunt/Beta*n) where Beta is the coupling factor and n is the coupler transformer ratio. So it would be that Zcav in parallel with Zrfc.

BTW, the first resonance is parallel resonant, i.e. impedance peak.  The choke doesn't get any "chokier" than this. It's only the second resonance (series) you need to worry about (and subsequent modes -- which, neat fact, are not harmonic, because the helix is a dispersive transmission line structure).

Exactly why I'm concerned with SRF (Series Resonant Frequency)

Since this is a cavity structure, you might also just feed it in that way.  For a coaxial resonator with the tube at one end, the bypass caps are how the middle cylinder mounts (and RF short-circuits) to the other end.  This isn't going to be trivial at 30kV, but, I mean, none of this will be.

Yeah in the past we used a dielectric film between two plates in the cylinder to have the anode and  the DC Blocked rf out side. That rf out side formed an air coax connection directly to the window of the RF cavity.

Unless those frequencies are harmonic in a given length resonator (I'd have to check?), you'll need some manner of telescoping resonator to run both bands.  This may encourage a lumped-equivalent design instead, but, that will be hard to do while respecting voltage and power ratings.

Both frequencies of interest are harmonics of each other but will be separate RF cavity systems. My goal is to have one amplifier design for both systems  that can be tuned or have a few components swapped depending on which system it is driving.

[T3sl4co1l]

Well, if you don't need any tuning through the RFC itself, it can be fixed, and the value isn't very important as long as it can be tuned for the two frequencies of interest, and the attached output network.  Maybe you keep the matching/tuning network specific to the cavity and bolt on the amplifier unit as needed; maybe the amplifier should be wideband so you only need a coax line out of it or whatever, and then maybe the RFC doubles as a transformer.

For RFC as transformer, the output impedance likely won't be 50 + j0 Ω, but the network will present some conjugate impedance to account for the transformer's "bad transforming".  And also TL length, since VSWR will be above 1.

RFC as transformer might not be very workable depending on wavelength and all that.  Likewise, given the dimensional scale, you might not be able to avoid series resonance in large values, forcing a smaller (tuned) value.

As long as the Q factor is high, little power is taken from the cavity.  Mind, out of 500kW and say 2MVA reactive, a Q of 100 is still 2M/100 = 20kW, that'll need a fair bit of forced air or preferably water flowing to keep cool.  Targeting a Q of some hundreds seems wise.

Offhand, 3.5uH or so seems feasible?  Say 150mm coil dia., 6 turns, 150mm coil length, 12.7mm dia. tube, series resonant at 63.2MHz, parallel at 36.3MHz.  So it'll be capacitive in the high band and inductive low.  Q of thousands seems typical, power dissipation shouldn't be a problem.  Best calculator: https://hamwaves.com/inductance/en/index.html#input

Tim

[Andy Chee]

This article might be handy

https://www.w8ji.com/rf_plate_choke.htm

[CaptDon]

I think you are over estimating the current that the choke will see. What plate voltage will you be using? You mentioned a 4CW150,000. You wanted 120KW output power so figure a very poor 50% efficiency and 20KV plate supply, 20 amps would give you 240KW of plate input power. A 30 amp choke would be very conservative. 120KW at a 900 ohm load is indeed a lot of R.F. voltage!! I would imagine you'll be wiping out anything radio controlled at 49MHz for a mile or so in every direction. The tank coil along with the tuning and loading capacitors will be hellish expensive.

[ahbushnell]

Hello,
I'm currently  designing a grounded cathode tetrode power amplifier. I was wondering the importance on the inductance of the Anode Choke. Typically with solid state amplifiers I've learned to pick an inductor that has a reactance > 5* the load impedance. When looking at some similar tetrode designs I see a much lower inductance. IE. a 28 MHz amp has a output load of 1k, inductance of 4uH which @ 28 MHz = ~703 ohms. My simulations show power loss with such low inductance. In Eimac's "Care and Feeding of Power Grid Tubes" it states that if you want the RF current to be limited to 1% through the choke, it should be 100 times your tank L, which is ~40uH. Is this really the case or do things like skin effect help improve this value? I believe at 28MHz your real impedance will be higher since the majority of current will flow on the outside of the wire. Additionally to help my AWR Microwave Office simulate properly I am using a RF tank to go from the Load impedance to 50 ohms, in actuality my amplifier will be driving a RF cavity with impedance of ~900 ohms on the amplifier side at the Fc.

I calculated for my design a inductance of 40uH but it is not possible due to multiple constraints:

Voltage Rating needs to be at least 40kV
DC Current needs to be at least 30A
SRF needs to be at least 49 MHz

Does the inductor use Litz wire? 

[CaptDon]

Your schematic example is Grounded Grid. Driving the cathode requires a lot of R.F. drive although it is claimed most of the drive power is passed through so not a complete loss. Also, are you building grounded grid according to your schematic example? If so where on earth will you find a filament choke capable of handling the massive filament current of a 4CX150,000? What are you using in the driver stage? 4CX10,000 maybe 4CX15,000? Those are usually driven by the very finicky 5CX1500B. The early 5CX1500 and 5CX1500A were very troublesome. You are looking at 3 very expensive tubes, sockets, blowers and loss of airflow protection circuits. Actually, I thought most of the bigger 'external anode' tetrodes were really not recommended for grounded grid? Usually I only see triodes in grounded grid.

[W2DML]

Well, if you don't need any tuning through the RFC itself, it can be fixed, and the value isn't very important as long as it can be tuned for the two frequencies of interest, and the attached output network.  Maybe you keep the matching/tuning network specific to the cavity and bolt on the amplifier unit as needed; maybe the amplifier should be wideband so you only need a coax line out of it or whatever, and then maybe the RFC doubles as a transformer.

For RFC as transformer, the output impedance likely won't be 50 + j0 Ω, but the network will present some conjugate impedance to account for the transformer's "bad transforming".  And also TL length, since VSWR will be above 1.

RFC as transformer might not be very workable depending on wavelength and all that.  Likewise, given the dimensional scale, you might not be able to avoid series resonance in large values, forcing a smaller (tuned) value.

As long as the Q factor is high, little power is taken from the cavity.  Mind, out of 500kW and say 2MVA reactive, a Q of 100 is still 2M/100 = 20kW, that'll need a fair bit of forced air or preferably water flowing to keep cool.  Targeting a Q of some hundreds seems wise.

Offhand, 3.5uH or so seems feasible?  Say 150mm coil dia., 6 turns, 150mm coil length, 12.7mm dia. tube, series resonant at 63.2MHz, parallel at 36.3MHz.  So it'll be capacitive in the high band and inductive low.  Q of thousands seems typical, power dissipation shouldn't be a problem.  Best calculator: https://hamwaves.com/inductance/en/index.html#input

Tim

Tim,
3.5 uH would have a reactance of 86 ohms at 24.6 MHz, wouldn't this cause RF current to flow towards the capacitor to ground on the other side of the choke?

I think you are over estimating the current that the choke will see. What plate voltage will you be using? You mentioned a 4CW150,000. You wanted 120KW output power so figure a very poor 50% efficiency and 20KV plate supply, 20 amps would give you 240KW of plate input power. A 30 amp choke would be very conservative. 120KW at a 900 ohm load is indeed a lot of R.F. voltage!! I would imagine you'll be wiping out anything radio controlled at 49MHz for a mile or so in every direction. The tank coil along with the tuning and loading capacitors will be hellish expensive.

I will be running class AB1, my current calculations & simulations estimated an efficiency of ~70% based on the load line, the DC current should not exceed 15A. I try to follow a 1.5-2* Max operating Value for my ratings which is why I selected 30A. The plate voltage will be between 14-20kV, most likely 16kV.  This amplifier will be in an accelerator tunnel directly driving a cavity so it will not interfere with anything. The risk is the gamma radiation from the RF cavity.

Your schematic example is Grounded Grid. Driving the cathode requires a lot of R.F. drive although it is claimed most of the drive power is passed through so not a complete loss. Also, are you building grounded grid according to your schematic example? If so where on earth will you find a filament choke capable of handling the massive filament current of a 4CX150,000? What are you using in the driver stage? 4CX10,000 maybe 4CX15,000? Those are usually driven by the very finicky 5CX1500B. The early 5CX1500 and 5CX1500A were very troublesome. You are looking at 3 very expensive tubes, sockets, blowers and loss of airflow protection circuits. Actually, I thought most of the bigger 'external anode' tetrodes were really not recommended for grounded grid? Usually I only see triodes in grounded grid.

The example was grounded grid but I will be running grounded cathode. The purpose of the example was to show where RFC is.

[Andy Chee]

The risk is the gamma radiation from the RF cavity.

You mean X-ray radiation, right?

[T3sl4co1l]

3.5 uH would have a reactance of 86 ohms at 24.6 MHz, wouldn't this cause RF current to flow towards the capacitor to ground on the other side of the choke?

Yes, that's how a decoupling network works.  Check the link and calculations: it's actually j677Ω at that frequency (and -j308Ω at 49MHz). You need a filter/coupling network to complement that, it will be a resonant network and I gather bandwidth is irrelevant so you just need to provide impedance to suit.  Supply bypass needs to be large enough that ripple is low, and then V*I = S is small in the capacitor (doesn't need to be any special materials like C0G), just rated for the current.

If say 10kV RMS is on 677Ω that's 14.8A, maybe a lot for a decoupling network but comparable to load current / impedance so you should still expect pretty wide bandwidth (Q ~ 1), give or take the rest of the system.

With a Q of 5000 and on the order of 100kVAR, it should only dissipate 20W, no big deal.

Tim

[W2DML]

The risk is the gamma radiation from the RF cavity.

You mean X-ray radiation, right?

Yes, mental fart there.

[W2DML]

3.5 uH would have a reactance of 86 ohms at 24.6 MHz, wouldn't this cause RF current to flow towards the capacitor to ground on the other side of the choke?

Yes, that's how a decoupling network works.  Check the link and calculations: it's actually j677Ω at that frequency (and -j308Ω at 49MHz). You need a filter/coupling network to complement that, it will be a resonant network and I gather bandwidth is irrelevant so you just need to provide impedance to suit.  Supply bypass needs to be large enough that ripple is low, and then V*I = S is small in the capacitor (doesn't need to be any special materials like C0G), just rated for the current.

If say 10kV RMS is on 677Ω that's 14.8A, maybe a lot for a decoupling network but comparable to load current / impedance so you should still expect pretty wide bandwidth (Q ~ 1), give or take the rest of the system.

With a Q of 5000 and on the order of 100kVAR, it should only dissipate 20W, no big deal.

Tim

Tim,
I think I'm following what you're saying. We need to account for this choke in our tank calculation. Doing so made the lower value choke work in my simulation.


Also yes, its j677, I missed the 2pi term when calculating.