SM 5 BSZ - Building High Power Amplifiers
(Sept 15 1997)

Introduction

Contrary to semiconductors, vacuum tubes do not change their interelectrode capacitances when voltages are applied. Therefore, a high power amplifier can be tested and optimised for tuning and loading without applying the VERY DANGEROUS voltages needed in full operation.

Here is my attempt to share my experience with vacuum tube power amplifiers. Although my practical experience is limited to tetrode amplifiers on 144MHz (and a few on 432MHz) I think some of the points below are quite general, and maybe of some interest also in other contexts.

This page deals with the output side of the amplifier. For the grid side, look here: Tuning the grid circuit.

Efficiency

It is a good idea to avoid losses in the output circuit when you design a high power amplifier. The precious RF energy should not be wasted as heat.

Keeping efficiency high is a good idea for several reasons.

You need less DC current, so the difference between loaded and unloaded supply voltage becomes smaller. If the amplifier has a total efficiency of 70%, which is possible, instead of 40% which is common, the DC power needed is only 56%.

This may be used for a higher plate voltage at key down, without exceeding limits with the key up. Higher voltages give higher electric fields inside the tube making the electrons move faster. This will improve tube efficiency on 144MHz if your tube is one of those allowing full data up to 110MHz only, provided that your tank circuit still has low losses at the higher impedance levels.

With more plate voltage, the screen grid voltage can be decreased, allowing slightly larger AC voltage on the plate without excessive screen grid current. This again leads to improved tube efficiency (at a higher load impedance).

If the tank circuit losses are large, which I think is quite common in amateur high power amplifiers, the tube should be operated with as much current as possible at a relatively low plate voltage. By operating the tank circuit at half the RF voltage, the losses in it will be reduced by a factor of four. Loosing a little more in the tube may then be less important.

If you can produce the RF power at half the load on the mains, your family and neighbours will be (much) less irritated by the lights going up and down in synchronism with your keying!!

The Tank Circuit

The tank circuit is at VHF frequencies best seen as an inductor that has to be added in parallel with the plate to ground capacitance to cancel the large capacitive current flowing from plate to ground and thereby creating a resonance circuit. The resonance circuit thus placed between plate and ground allows very large AC voltages to build up between plate and ground at the desired frequency.

Using air and vacuum as the dielectric in the plate to ground capacitance, the capacitance causes virtually no losses so all the losses of the tank circuit originate in heating of the conductor constituting the inductor.

Of course it is also possible to create losses in the by pass capacitor that has to be inserted somewhere in the inductor between the plate and the RF grounded grid of the tube.


Figure 1. Equivalent circuit diagram of the tank circuit of a power amplifier


If the tube is a QBL5/3500 the capacitance from the anode to the screen grid is 8.4 pF. With an AC voltage of 3.5kV (RMS) on the plate, the current flowing through this capacitance is about 25 A. These 25 amperes have to flow through the coil from a to b, then through the decoupling capacitor from b to ground at c.

The resonant current also has to flow from ground at d through the screen grid decoupling capacitor into the screen grid.

Everywhere around this loop the ohmic losses have to be small. The current flows in a thin layer on the surface of the metal structures used to form the equivalent circuit above. The conductivity is important, but it is also important to avoid steel anywhere where the magnetic field due to the current is large. Steel, besides having a high resistivity, is ferromagnetic and converts alternating magnetic fields efficiently to heat!! BADLY PLACED IRON MAY LOWER THE Q BY A FACTOR OF 10!!!

It is a very good idea to cover any steel near the tube by copper or low resistivity aluminium (some hard aluminium alloys have three times higher resistivity than pure aluminium). Regard the copper or aluminium as a screen which is inserted to prevent the RF fields/currents from reaching any iron. Of course it is a good idea to build the amplifier entirely from aluminium and copper, maybe with a few more complicated things made from brass.

The large current flowing out from the screen grid needs a good contact. Use many contact points all around the tube.

The screen grid decoupling capacitor is conveniently made from some copper sheet on which the screen grid connector are soldered in the form of finger stock around a hole. On both sides place some good and thin dielectric and then a grounded aluminium sheet on each side. I am using 1 mm polypropylene, but the area needed to get enough capacitance is a bit large. 0.25 mm teflon would be better.

Note that the centre part of a screen grid capacitor made like this is a resonator in itself. If the diameter is large enough (in my case with polypropylene about 60 cm) the resonance frequency is 144 MHz, and then there is a parallel resonance between screen grid and ground making it impossible to use the tube as an amplifier. The diameter of the screen grid decoupler has to be made much smaller (or larger).

Current flow in the tank circuit


Figure 2. Idealised tank circuit with screen grid decoupler


The (resonant) current flowing out from the screen grid will flow radially out on the copper plate onto which the fingerstock constituting the screen grid connectors is soldered. The current will flow through the dielectric and come out on top of the grounded (aluminium) plate which is on top of the screen grid capacitor assembly. No current will flow out on the lower grounded surface because it has nowhere to flow. A few good connections between the upper and lower grounded plates close to the tube is good for mechanical stability and may lower the inductance of the capacitor slightly.

The current is following the upper surface of the chassis and then it goes vertically up to, and through the plate decoupling capacitor. The tank circuit is finally closed by the current returning into the tube through the anode.

The current is flowing in a thin layer on the metal surface. The more surface the current can flow on, the lesser current density will be produced. If the conductor area is doubled, the current density will be halved and the losses per unit of surface reduced by a factor of four. (loss is current squared) With twice as much surface, still the losses are reduced by a factor of two.

The lowest possible losses are obtained if the current is allowed to flow radially in all directions from the tube. A cavity, which would then be the tank circuit will have a diameter of about 1.2 metres and a unloaded Q of about 3000 with the tube in place (aluminium at 144MHz with a QBL5/3500). This is actually the design I am currently using, (although I might have gone for something slightly smaller if I had known in advance what Q to expect.)

If the current from the plate is allowed to flow in only one or two directions, the current loop has to be closed much nearer the tube because the higher current density corresponds to higher inductance per length unit.

If the plate line is made from sheet metal, say 1 decimetre wide, the length will be in the order of 2 decimetres. Then the current will flow more or less equally on the upper and lower surfaces on the plate line. Such a tank circuit has to be screened in order to avoid losses of power by radiation. (And for safety reasons too!!!) If the compartment around the plate line is close to it, and made from steel, severe losses will be introduced.

If the screen is placed close to the plate line, the inductance is lowered, but if the screen is a good conductor, the Q is not reduced dramatically, but still it is a good idea to place the screen at a good distance. Increasing the distance between the plate line and the chassis will allow a shorter conductor for the same inductance with correspondingly lower losses.

The optimum cavity will carry the current radially out from the anode to grid capacitor (G1 / G2 for triode / tetrode) for a short distance, then the distance between "floor" and "roof" should gradually increase to some maximum value, and then follow a more or less circle like curve until they meet. Maybe such things would be of some use at 1296 MHz. At 144 MHz it is certainly not worth the trouble.

Theoretical aspects of tuning and loading of the tank circuit

For tuning, an extra capacitor is added in parallel with the anode to ground capacitor. The tuning capacitor should be small, an amateur VHF power amplifier is designed for a fixed frequency, and tuning is only to compensate for very small variations that may occur when a tube is replaced. The tuning capacitor causes an extra current in the tank coil, and an extra x% current a to b in fig 1 above will increase the losses in the coil by 2x%. It is a good idea to keep x small. A small tuning capacitor will also give a larger bandwidth with better thermal stability as a result.

Conventionally the output is coupled to the plate by an adjustable capacitance. At 50 ohms, this will add 5 more (essentially capacitive) amperes (at 1250W) to the plate current with a corresponding increase of the current through the inductor. With a capacitive coupling at the highest impedance point, the loading adjustment will have a strong interaction with the tuning adjustment, both are essentially capacitive impedances from plate to ground.

Much less interaction (or nearly none) can be obtained by magnetic coupling through a link or by capacitive coupling to a tap on the plate inductor where the impedance is not so high, perhaps 100 ohms or so. Inductive coupling in combination with a small tuning capacitor may easily double the circuit efficiency for the tank circuit.

Cold adjustment of tuning and loading of the tank circuit

Assume you have a new/rebuilt power amplifier for 144MHz, and that you do not know anything about how tuning and loading works. Most amateurs would apply voltages, and a little drive power, for the tube to operate at a modest power level, and tune/load for maximum output. The problem with this approach is that the tuning/loading controls usually will come into an end position (Murphy's law), or that the air gap becomes too small with arcing as a consequence. Of course the procedure works, but convergency is slow because one has to be very careful in making sure all dangerous voltages are gone and dismantle all screening to be able to make necessary adjustments.

There is a much easier way, fast and safe, and at the same time you will also get information on the tank circuit efficiency. If efficiency is not good, it is not difficult to locate the sources of the problems and find a cure.

First decide what voltage and current you want to run the tube at. Then calculate the corresponding load impedance for the tube. For a 4CX250B at 2500V and 250mA in class C, the load impedance is about 5 kiloohms.

When you know the load resistance, connect a resistor with this impedance at the design frequency between the plate and the screen grid. The tube has to be in place, but no voltages should be applied to it. For a grounded grid triode, the resistor should be connected between plate and control grid. The idea is of course that the resistor is connected from plate to ground from a RF point of view.

Connect a 50 ohm signal source at the power output connector. In this way the amplifier is run backwards, and when the tuning and loading is correct, the signal source will see SWR=1 because then the amplifier is tuned to transform from 50 ohms to the load resistor between plate and screen grid. At the same time the RF voltage on the plate is at a maximum.

It is easiest to use a wide band signal source, a normal signal generator that will give maybe a -20dBm signal. Start by finding out at what frequency the signal on the plate has a maximum. Adjust the loading control while checking that the signal generator still is at the frequency of maximum.

To measure the plate RF voltage you need a wide band level meter. At 144 MHz the easiest may be an oscilloscope, but you may also use a wide band amplifier followed by a diode. Of course a spectrum analyser with tracking generator or a network analyser makes everything very simple, but such equipment is not available to most of us.

Measure the plate RF voltage by probing the electric field near the tube with the wide band level meter as indicated in figure 3. Start with the probe tip a few millimetres away from the tube and when you have found a signal, and tuned it to a reasonably large level, reduce the influence of the probe by shortening the length of the probe wire. If you feed -20dBm into the power amplifier, you want nearly all the power to heat the load resistor so do not try to extract more than -40dBm from the probe. Otherwise the result maybe misleading.


Figure 3. Probe the RF field by capacitive coupling near the plate. Get the ground point for the cable to the RF meter on the outside of the amplifier, and use a hole not very far from the anode. Make sure that the distance from the probe to the output connector and loading arrangement is much larger than the distance to the tube (use opposite side)


Since the whole procedure is without voltages and blowers it is possible to try various modifications on tuning and loading by use of very thin sheet metal. The plate and screen grid decoupling capacitors may be shorted or simply added at a later stage (but they may alter tuning slightly).

When the design/modification is finished, you have a nice maximum with both loading and tuning controls in sensible positions that will be safe also when high voltages are applied. It may be wise to check that you can tune also for half and double load resistors without any problems.

Remove the resistor and measure the 3dB points when changing the frequency of the signal generator. The tank circuit is now loaded by the signal generator exactly as it will be loaded by the antenna. The loaded Q of the tank is the operating frequency divided by the 3dB bandwidth.

Finally couple the signal generator very loosely to the output connector. Bend a piece of metal to hold the cable from the signal source with the screen in good contact with the screen of the power amp output connector, but with a few millimetres distance between the inner conductors. The metal piece should be bent to make the inner conductors reasonably well screened. In this way, the signal from the signal generator is coupled very loosely to the tank circuit so the 50 ohm output impedance will not load the resonance. The coupling to the level meter may have to be reduced also. The idea is that the tank circuit should be loaded by its own losses only.

The resonance should now be much narrower, and the unloaded Q of the tank can be determined as the operating frequency divided by the 3dB bandwidth.

To verify the correctness of this measurement, repeat it with a different coupling (10 dB weaker or stronger signal at the peak) and make sure the bandwidth is unchanged. Do this at both sides to make sure that neither the signal generator nor the level meter is loading the tank circuit.

You can now calculate the ratio of loaded Q to unloaded Q. This ratio gives directly the proportion of power you are loosing in the tank circuit. An aluminium cavity on 144MHz easily gives this ratio larger than 100 corresponding to less than 1% losses. I have seen 2 meter amplifiers with this ratio as low as 2, where half the RF energy was used to heat the tank circuit.

If the Q is too low, the reason may be heating of a conductor or some dielectric. Energy may also be lost by radiation or conduction from the tank circuit. In any "normal" design losses by radiation should be excluded a priori by the tank circuit enclosure. You may use aluminium foil to improve screening and remeasure the unloaded Q to verify that radiation losses is not the problem. RF energy may also be lost from the tank circuit by improper decoupling of the DC voltage for the plate. Just temporarily remove the DC wires, RF chokes and any other wires that may pick up RF from the tank circuit and measure the unloaded Q again.

If there is steel somewhere, just cover it with aluminium foil to find out if you have to replace it or not. The aluminium may have to cover several centimetres of surrounding aluminium/copper surfaces, very tightly to get capacitive coupling, or it has to be in good contact with the surroundings. If the Q improves a replacement steel to aluminium is a good idea. Or permanent screening with thin aluminium or copper.

The losses may be in the anode decoupling capacitor (maybe improper plastic was used for dielectric). That is easily seen as a dramatic improvement of the Q when the dielectric is removed and the tank is reassembled without it, or with it replaced my a piece of metal. You may also wrap the dielectric or the entire capacitor in aluminium foil. The idea is to replace the decoupling capacitor by a low inductance short circuit.

If the ratio of unloaded Q to loaded Q is still too small, an increased LC ratio will improve as explained above. Going from capacitive to inductive coupling may be a good idea.

The efficiency of a tank circuit is a function of the geometry and the resistivity of the metals used when all other possible sources of losses are eliminated as described above. To get better Q, more area for the resonant current to flow on will reduce losses, but the inductor length has to be increased to compensate for the reduced inductance. Avoid current concentrations, make sure the current has a large area to flow on when it goes from the plate line into the tube or to ground, and remember that the current also flows from ground into the RF grounded grid.

Screens too close to the inductor will reduce the inductance per length unit, which has to be compensated by a longer conductor, in turn adding more losses. The current on the screen also adds losses in itself (particularly if it is a nearby steel wall).

Finally, choice of metals may be very important, particularly if you want a good amplifier in a minimum volume. If the losses are X% when everything is made from copper, an identical amplifier made from some other metals will have losses as indicated in this table:

   Material             Relative losses
Silver plating              0.96*X
Copper                      1.00*X
Aluminium (pure)            1.25*X
Brass                       1.87*X
Aluminium (worst case)      2.17*X
Steel                       25*X or more

Steel with a yellowish surface treatment, common in electronics, has losses similar to brass. I do not know what this surface is, but it is easy to verify the dramatic reduction of losses that it gives.

Verification of efficiency

When the amplifier has been in use, and is properly tuned for maximum output in a 50 ohm load, you can measure the loaded Q of the tank circuit just by replacing the antenna cable with a cable from a signal generator with the same impedance. The procedure is as described above, and of course without any voltages applied to the tube.

You get the loaded Q, QL when the tank circuit is loaded exactly the way you were loading it when really operating the amplifier at full power. The theoretical impedance comes from a "rule of thumb" and has an unknown (at least to me) accuracy.

The unloaded Q, QU is then determined and gives tank circuit efficiency factor:

Tank=( 1 - QL/QU )
multiply by 100 to get it in %

The total efficiency is simply Power out divided by power in, but by convention power in is defined as the anode to cathode DC voltage multiplied by the anode current. Additional power is used to heat the cathode, some RF power is used to drive the tube, and for a tetrode, some power is used to heat the screen grid. These losses are usually not considered at all, or they are treated separately.

The total efficiency is the product of the tube efficiency and the tank circuit efficiency, so the tube efficiency factor is:

Tube = Total / Tank

A 2 meter class C amplifier with 4CX250B has a tube efficiency well above 80% while a QBL5/3500 will give between 60% and 75% also in class C (higher plate voltage gives better efficiency). In class B or AB the tube efficiency is considerably lower, but it is easy to run a tube in variable class, although then the SSB output power has to be limited to a slightly lower value than what is possible from an amplifier designed for pure class B.

To SM 5 BSZ Main Page