SM 5 BSZ - Power Supplies for High Power Amplifiers
(April 24 1997)

Variable Class Amplifiers

For weak signal communication, CW is superior to SSB by orders of magnitude (20 dB or so). Therefore it seems natural to me to optimise high power amplifiers for CW, and accept slightly less than maximum possible output in SSB. Class C amplifiers give better efficiency than class B or AB, but a class C amplifier is far from linear and useless for SSB. Even for CW care has to be taken to avoid that the nonlinearity of a class C amplifier causes problems with severe keying clicks.

A good solution is to run the amplifier in variable class by the use of a resistor in series with the control grid. When current starts to flow through the control grid, the DC component on the control grid will automatically go more negative and bring the tube into class C. Just by limiting the drive level for no or very little grid current, the tube is operated linearly and will work fine for SSB. Just watch the grid current meter, and make sure it does not move (much). For this solution to work without creating any problems to your fellow hams, the absolutely necessary precaution is to make sure that the time constant for the DC voltage on the control grid is very much faster than the rise and fall times of the SSB signal. Otherwise an occasional pulse of grid current will bring the tube into class C for a long time with disastrous splatter as a consequence. The time constant on the control grid has to be well below 0.1 ms.

A two tone test of an amplifier in variable class will show bad results as soon as grid current starts to flow, but for real voice signals, the performance is better. The distortion is caused by the amplifier going into saturation, and this happens gradually with increasing grid current if the RF impedance is held low on the control grid. The SSB peaks are amplitude limited in a soft and gentle fashion that will create low order intermodulation products only. The signal bandwidth is only increased marginally and no splatter is generated.

An amplifier that is optimised to be very linear often saturates very abruptly and therefore causes severe wide band splatter at drive levels when the two tone test still is not too bad.

Power Supplies for High Power Tetrode Amplifiers

I have been using tetrode amplifiers operating in variable class since about 1970, and I find it very easy and straight forward. There is no need for any "difficult to find" components if you have access to a reasonably stable source of mains power. When the mains voltage goes up, all voltages increase. The effect of more positive voltage on plate and screen is nearly compensated by more negative voltage on the control grid, leading to an idle current that is more or less unaffected. When the mains voltage goes up by 10 %, the output power goes up by 20%. You just have to apply slightly more drive power before the control grid current starts to flow. The tuning is completely unaffected.

The Heater Voltage

When the cathode is cold, the resistance of the heater wire is in the order of 5 times lower than during normal operation. The switch on current that flows through the heater wire may be very large, if the heater is fed from a low impedance source. In some tubes this makes part of the heater wire very hot for a while, until all of the wire is warm and the current has gone back to normal.

If you switch your rig on several times each day, just to check if there is someone there, it may be a good idea to limit the heater current by use of a resistor in series with the heater transformer. Just allow some voltage drop, 5 or 10 %, in the cables between the transformer and the tube or use a resistor in series on the primary side. (Obviously you need a transformer with a slightly higher nominal voltage)

I have a feeling that current limiting is good for tube life. A possibility to adjust the resistor (or wire length) to bring the heater voltage exactly to the nominal value is also valuable particularly for thoriated tungsten cathodes. Both overheating and underheating may be harmful to the tube.

Negative Voltage for the Control Grid

The design starts by selecting R2. Look in the data sheet for control grid current and voltage in class C operation. These numbers give R2 directly. Typical data is:

     Tube           Ig1 (ma)     Ug1(V)      R2(kohm)
QQE06/40 (5894)      4(2x2)       -80           20
   4CX250B             20        -100            5
  QBL5/3500            80        -400            5
Once R2 is known, the value for R1 follows because R1 and R2 form a voltage divider for the voltage across C1. For a QBL5/500 AC in can be taken directly from the mains. With 230 V AC in, there will be -325 V on C1, so R1 has to be 10.5 kiloohm to give -105 V on G1 when there is no RF in.

The time constant (R1+R2)*C1 has to be very much larger (100 times) than the charging time interval. In a single phase system where only D1 is used the charging frequency is 50 Hz (or 60 in the US), so (R1+R2)*C1 should be about 2. With R1 + R2 = 15 kohms C1 should be 150 microfarads or more. In a three phase system with R1 + R2 = 15 kohms, 50 microfarads is sufficient for C1, but there is no disadvantage in making C1 larger as long as diodes and fuses cope with it. C1 can of course be charged from a transformer with a bridge rectifier or by two diodes from a transformer with a centre tap.

At full RF drive (CW) G1 is at -400 V in the QBL5/3500 case. The purpose of D2 is to make sure that C1 is not charged through R1. Particularly at lower AC voltages this is very important because otherwise the tube looses its idle current between dots and dashes with faster rise and fall times on the CW parts as a consequence. This means a considerable broadening of the signal spectrum (=keying clicks) and D2 is a very easy way of avoiding that.

C2+C3 represent the sum of all coupling and decoupling capacitors around the G1 drive circuit. C2 includes any decoupling capacitor in the power supply in case R1 and/or R2 are placed there. The time constant (C2+C3)*R2 has to be well below 0.1 millisecond. With R2 = 5 kohms C2+C3 should be below 20 nF.

L1 represents the resonant circuit on the grid side. Do not add large tuning capacitors on the grid. The capacitance of the tube itself should be most of the capacitance of the resonance circuit. Additional capacitance has to be balanced by lowered inductance which makes the bandwidth smaller. A large bandwidth is good for thermal stability and uncritical operation in general.

C3 represents the coupling of RF to the grid circuit. The coupling should be tuned for power match at maximum drive power when full current flows through the control grid. For details tuning the grid circuit

R3 is a resistor that prevents the Q of the grid resonant circuit from going very high when the circuit is not loaded by current in the grid. This resistor helps to make the amplifier saturate softly, which is essential in avoiding keying clicks and splatter. For details tuning the grid circuit

S1 is a switch that can be opened to bring the tube into stand by mode.

The meter is essential. You have to watch it when running SSB. You will soon find out how much current, if any, you can allow in SSB. The meter is also good to have to make sure not to destroy the tube by overdriving G1 when running CW.

The Screen Grid Supply

The screen grid of a tetrode can be used as a probe for the AC voltage on the plate. If the instantaneous plate voltage goes below the screen grid voltage, the screen grid becomes the most positive electrode within the tube. The electrons will then be collected by the screen grid to a large extent, giving rise to a large screen grid current. It is a very good idea to protect against screen grid overcurrent. Excessive AC voltage on the plate may cause a lot of problems, and the protection is not really to protect the screen grid itself from overheating although that also comes along as a bonus.

If the load disappears, for example due to poor contact somewhere along the path to the antenna, very high voltages may build up in the tank circuit at the same time as the screen grid current becomes very high. If the tube is switched off rapidly, the damage will be limited and some cleaning and polishing of the failing contact will be all you have to do to restore proper operation.

If the tube is allowed to run in this fail condition, you can easily have a puncture of the glass due to dielectric heating if your tube is a QBL5/3500 or a 4X250B. You may also have to replace a lot of other things. The kW you put into the tube will go somewhere - and if it can not get out through the antenna......

K1 is a relay that will close as soon as the screen grid current is above normal. I am using a reed relay with a home made coil in order to have a very good isolation in case the tube fails and delivers plate voltage out on the screen grid. It is a good idea to design as much as possible of the screen grid circuitry so it will survive if you get a short between plate and screen grid. (Many kilovolts on the "DC for relays" line would certainly cause very much damage.) The tubes we buy cheaply at ham fests were once replaced (maybe) and the reason may have been problems with arcing.

A1 is an air gap, formed by a piece of uninsulated wire bent near to the chassis. Hopefully it will limit the voltage on the screen grid system in case there is arcing from the plate.

R2 is a resistor selected for K1 to close at the desired screen grid current.

R1 is a bleeder. It is very important because during some operating conditions tetrodes may draw current "the wrong way". R1 should be designed to always carry as much current as the maximum expected "negative" screen grid current of the tube. 25 mA through R1 is reasonable for both 4CX250B and QBL5/3500. If R1 is too large, the screen grid voltage will start to rise due to the "negative" current. This increases the negative current and the anode current and the tube "rushes". When the tube "rushes" nearly all power will heat the anode. I have seen this happen in many amplifiers where the screen grid bleeder was omitted, but I am not aware of this problem causing any serious damage. It usually happens during amplifier tuning, and several seconds of 5 times the rated plate dissipation will not destroy a tube.

R3 is a second bleeder. It is there to discharge C1 within a reasonable time to avoid unpleasant surprises when working with the power supply. If you leave K2 closed (see below) when switching off AC in, C1 would stay charged "for ever" if R3 was left out. Design R3 to dissipate 0.5 or 1W at full screen grid voltage.

K2 may be several relays in parallel, or relays controlled by other relays. In my system three relays are used. Two surplus relays for high voltages with 10 mm contact distance switch the screen grid voltage, and also the plate voltage. These two relays are operated from K2 to close when K2 is open. During warm up and in stand by mode these two relays are open. With such a three relay arrangement for K2 it is obvious why R3 is useful.

C1 has to be big enough to produce a screen grid voltage with low ripple. The diodes symbolise any rectifier and/or transformer arrangement that will produce the desired screen grid voltage. (see below)

The plate voltage supply

The high tension needed for the anode should be generated by a transformer followed a rectifying bridge. Full wave rectification is a minimum requirement, but a three phase six pulse bridge is much better since it will give less ripple with a smaller filter capacitor. Keeping the filter capacitor small is a good thing because if you have a discharge, there will be less stored energy that may destroy your tube. With only a few microfarads and a six or twelve pulse rectifier you can omit the 50 ohm current limiting resistor between the power supply and the tube (or you can omit it anyway, just hoping the tube is good and will not arc). The current limiting resistor for a QBL5/3500 is not just a small thing, it has to stand maybe 100 kilowatt for a short time without loosing its resistance in a big arc.

The twelve pulse three phase power supply

This circuit is simply two paralleled full wave three phase bridges. One gives maximum voltage when the voltage between one phase and zero is maximum, the other when the voltage between two phases is maximum.

The capacitor at the output will be charged by twelve evenly distributed pulses during one cycle of the mains frequency.

In the diagram above, six transformers are used, three of each kind. The transformers may be arranged in several different ways and of course a lot of weight can be saved by use of three phase transformers.

There are several advantages of charging the capacitor more often. The capacitor can be made smaller for the same level of ripple. A smaller capacitor makes arcing in the tube less harmful to the tube.

Charging the capacitor more often means that the charging pulses will have a smaller peak value, which means that the ohmic losses will be reduced for a smaller difference between loaded and unloaded voltage output.

For a given power output from your transmitter, the unloaded voltage on the plate is substantially smaller, reducing the risk for arcing.

With a more sine wave like current on the mains, the lights in your and your neighbours house will not darken so much when you push the key - this is part of the reduced ohmic losses. This may make your neighbours less hostile to your hobby if you also handle TVI properly....

I guess most amateurs consider it a bit extreme to use a 12 pulse DC supply for the power amplifier. I was lucky to find the transformers required, so I use now a 12 pulse supply with two three phase transformers, and I am pleased with the low voltage drop at key down.

Variable mains transformers

Variable mains transformers are very useful. If you can find one, use it to control anode and screen grid simultaneously. If you run your amplifier in variable class as described above, keeping a constant ratio between anode and screen grid corresponds to class C anode and screen grid modulation when the key is down. This means that the tuning is unchanged when the output power is changed.

With a variable transformer controlling the voltage simultaneously for both anode and screen grid you can adjust the power by 20 dB or more. To maintain a reasonable idle current, of course the control grid negative voltage has to be reduced also. (You may like to run the control grid supply also from the variable transformer)

The power reduction without detuning or loading change is not really for reducing power output in normal operation. In normal operation, power level is simply controlled by the drive level - the amplifier is linear, except near full power.

The variable transformer is particularly useful when you fire up a new power amplifier, before you know how loading and tuning has to be set (and/or modified). At 25% for the mains voltage, you get 6% of the normal power. You may run the amplifier continuously for ever into a modest dummy load without any heat problems even if tuning and/or loading is completely wrong.

If your tube is not operated near the maximum frequency, like a 4CX250B on 144MHz, the output is very closely proportional to the transformer setting squared. If your tube operates near its maximum frequency, like a QBL5/3500 on 144 MHz, then efficiency is slightly better at higher voltages, and a slight readjustment of the loading may be necessary if the voltage change is large.

To SM 5 BSZ Main Page