15: Notes on RF Linear Amplifier Operation
Many radio amateurs purchase a commercial linear amplifier to increase the
effectiveness of their station. Whereas this is often a good investment, it brings with it the necessity to understand the complexity and the required manner in which the amplifier needs to be operated.
For some amateurs who have had no previous experience with valve equipment it can be a recipe for disaster! Many amateurs have only experienced the solid-state world and to them valves are a foreign subject. I have been involved in many post mortems on valve amplifiers and some of the catastrophic failures I have observed are purely due to the operator not using the equipment correctly.
I am sometimes approached for advice when an amateur is considering purchasing an amplifier, either new or used. The question most often asked is “Which make and model would you recommend?” I hate these sort of queries, as firstly I haven’t had experience of every amplifier on the market, and some of those which I do have knowledge of sometimes leave me doubting the sanity of some amplifier designers! It seems some never even bothered to study the valve manufacturer’s data sheet or read the engineering bulletins put out on correct operating conditions.
Over the years there have been many different manufacturers who have entered the amateur valve linear amplifier market. Some produced reasonably priced products and look attractive, from the outside at least. But when you dig down under the covers and look at what is offered for the price it can sometimes be a poor investment, unless you are skilled in using these beasts. A linear amplifier is not a cheap object. Very often it is one of the major costs in a station. Some cost more than others, often much more.
If I were really pressed for an answer, there is really only one manufacturer I would consider laying out a huge sum of money on. That would be Harris Electronics Inc in the USA. But these types of amplifiers are not amateur grade or priced. To purchase a new 1.5kW Harris amplifier would need a second mortgage. The reason is that everything about them is right. They are widely used in the military field to provide rock-solid reliability, 24 hours a day, at full rating. In the South African navy installation in Durban there are at least six Harris HF amplifiers I am personally aware of. These stand in 19in racks and are taller than I am. They run 1kW PEP for lengthy periods. I was acquainted with the service and technical people who maintained these amplifiers for many years and in all that time not one required a valve replacement.
The amateur grade amplifier in military service would not last very long before they gave trouble. There are noticeable exceptions, for example Collins Radio, whose amplifiers are really slightly downgraded military spec equipment and also not cheap. A good second hand Collins 30S-1 (not the 30L-1, which uses the 811 valves) would provide ample output and run forever, if the operator knew how to tune it and interpret the meter readings correctly. The cost of a replacement 4CX1000, however, should you ever need to replace it, is a major factor.
The basic problem with amateur grade linear amplifiers is always the cost factor. If the manufacturer puts all the right bits into it, and used suitable valves operating within the valve manufacturer’s recommended ratings, there would be far less ‘splatter boxes’ on the bands. Unfortunately, this would put them out of reach of all but the well-heeled amateurs. Instead we have a mixture of so-so amplifiers, none of which are truly linear at the output power claimed by the makers. The only amplifier I have personally tested that is almost up to the claims is the Kenwood TL-922, which uses two Eimac 3-500Z triodes.
However, there will always be the ‘Walter’s’ of the world who, no matter how much care has gone into the design and construction, would manage to mess things up. These are the amateurs who never bother to read the instruction manual, until something blows up. Consequently, most of the things they believe to be true – because another Walter at the radio club said so – are in fact just plain wrong!
High Voltage Transformer Tapping Problem
One of the worst mistakes these amateurs make is to mess with the high voltage tapping on the mains transformer. Many commercial amplifiers are made for different mains voltages to suit different countries, typically 115V or 230V are common. However, some manufacturers also have the high voltage secondary winding provided with different taps as well. There are several popular makes of HF linear amplifiers which use the 811, 572B and 3-500Z grounded grid Class B triode which have this stupid feature. The amateur thinks that by dropping the anode voltage from the maximum to something much lower they are being kind to the valves and they will last longer. Nothing could be further from the truth!
Many valve manufacturers have covered this topic in their technical literature. The problem is that Walter doesn’t read technical stuff, so he is totally oblivious to the facts. The other Walter at the radio club told him it was the best thing to do, so he went ahead and lowered the anode voltage to the lowest tap setting. Now instead of having the recommended 2.4kV on the anodes it only has 1.2kV.
The manufacturer provides this facility when the necessity to operate at a far lower output power is required. Walter doesn’t know this and still thrashes his amplifier to get his full-entitled power output. He did, however, notice that the drive needs to be much higher and the anode current meter almost hits the end stop. He never looked at the grid current meter, because he doesn’t understand what it is telling him anyway!
Valve manufacturers are universal in the opinion that a transmitting valve is most efficient at the maximum rated anode voltage. Many have issued engineering bulletins, but Walter doesn’t read this sort of stuff. A valve is a low current – high voltage animal. To attain the same input power when the anode voltage is only half the recommended value needs twice the anode current.
Excessive cathode current can be very damaging to some valves, particularly the 811 and 572B. It causes scouring or erosion of the cathode oxide coating, which leads to premature failure. It also causes severe overheating in the anode structure. Examination of a severely over stressed anode, due to abnormally high current, shows that it looks as if it has been shot blasted, this is due to the electrons slamming into it at a velocity of nearly the speed of light. The high temperature causes metal migration and abrading of the coating on the nickel base material. Not only that, to induce the anode current to get to where it needs to be for the required input power, needs much more drive power. Modern amateur transceivers normally have an excess of drive available so that isn’t an issue. What is of more concern is that when a grounded grid triode is operated on an anode voltage well below its rated voltage, the grid current becomes excessive. This causes accelerated degradation of the grid structure due to high temperatures.
The RSGB Radio Commu-nication Handbook of early days covered this factor and produced a nice graph showing efficiency versus anode voltage, which sadly no longer appears in later handbooks. It is reproduced here as Fig 15.1, and every linear amplifier user should have a large copy stuck on the shack wall to remind them of this vital fact!
Fig 15.1: Anode efficiency versus anode voltage graph for Class B.
However, what wasn’t adequately covered in the early Radio Communication Handbooks is why this is so important. I cover this topic in detail in RF Design Basics [1], in the chapter on the pi tank network. To explain the basic facts, it is all to do with the pi network loaded Q figure.
Loaded Q and Why it is Important
In an amplifier pi network the component values required are heavily dependent on the selected loaded Q. From this starting point the value of the anode tuning capacitor required for the anode load resistance to achieve the required loaded Q is first ascertained using the standard formula. The next component to be defined is the loading capacitor required to transform down to the load impedance, and finally the series inductance required. All of these calculations use proven formulas. The valve has a particular required anode load resistance it needs to be matched to, the conjugate match criteria. At this value it will provide maximum power transfer to the following transformation network, and work most efficiently. If the anode voltage and current are changed, the anode load resistance also changes.
When the anode load resistance is radically different the component values required to ensure the correct matching are also quite different. Often the component that needs the greatest change is the series inductor. The anode tuning and loading capacitors, because the operator can adjust them, often have sufficient range to find a tuning point that allows some maximum power to be produced. But this is not the most efficient operating point or providing the maximum possible power output. Consequently, the valve will experience higher dissipation than if it were correctly matched. The tuning is also a bit ‘soggy’ and imprecise, although a power output peak can often be established.
It just so happens that when the anode voltage is lowered to half the original value, the component values required are exactly the same as the higher voltage. At this setting the valve will be matched to deliver exactly 25% of the power with these component values as when it is operated at the higher voltage, with the same low anode current. But that is not the case when Walter taps down the high voltage and then drives the anode current to twice the normal value. Power is the result of V2 / R, so if V is reduced to 50% the power is now 25%.
Lowering the anode voltage without re-optimising the pi network values, especially the series inductor value, changes the anode load resistance and hence the loaded Q value and the overall efficiency. The valve is in fact far greater stressed under this condition than when it can loaf along on the higher anode voltage it was designed for. It draws less anode and grid current and will last many times longer under the high anode voltage condition.
The other thing that changes is the saturation voltage. Any valve requires a certain minimum anode-cathode voltage to work correctly. At each anode current setting it exhibits a certain anode load resistance. As the anode voltage swings down lower towards the cathode voltage the resistance changes. It also becomes lower and it is this that determines the anode current amplitude. When the anode voltage approaches the cathode voltage, the resistance needs to be driven much lower to support the current flow demanded. When the anode voltage becomes very low the attraction of the electrons falls away and they are not as easily attracted in sufficient quantity to support the anode current flow. The valve anode current saturates at this low anode voltage level and the signal becomes grossly distorted, which generates a horrendous amount of splatter. This shows as severe flat-topping of the anode voltage waveform.
The other factor is that the original RSGB chart shows the theoretical efficiency value for a true Class B amplifier when operated under ideal conditions. Eimac also supplies the same data and it can be proven that in an ideal Class B amplifier the maximum possible anode efficiency is only ~78%, for the ideal valve. This is the theoretical calculated efficiency figure. It does not consider possible losses in the transformation network to get from the high anode resistance to the impedance of the antenna. In the real world we can deduct about 10 to 15% from this utopian situation for the perfect valve, and when also including the potential power loss in the tank network, in practice 65% is about the best we will achieve with a practical HF amplifier when running in true Class B. But a Class B amplifier at full output in a single ended stage (not push-pull) will have a distortion figure of between 5 and 10%. As one valve manufacturer stated in an engineering bulletin, “Nobody today actually uses Class B in RF amplifier service, because there are much better types, such as Class AB1 or AB2”.
This is why Class B is best used for audio push-pull amplifiers, as the distortion produced in one half is cancelled by an equal and opposite distortion product in the other half – at least in theory. Again, it is assumed that each valve in the push-pull amplifier has identical characteristics and each delivers exactly 50% of the total power and uses the full 180º input voltage swing. In practice this never happens in the real world.
Incorrect Loading Technique
Another major error is not loading the linear amplifier heavily enough. In days gone by when Class C amplifiers were in vogue for AM and CW, the method to tune up was firstly to fully mesh the loading capacitor and then apply enough drive to get about 25% of the required final grid current. The anode current would show about 25% of the full value when the drive was correctly set. Next the anode tuning capacitor was swung until the biggest dip in anode current was found, which indicated resonance of the tank circuit. The loading capacitor was then unmeshed about half way and the anode current would rise and also power would be indicated on the output power meter or antenna ammeter. The drive was further increased until the required DC input power was achieved and the anode tuning capacitor re-dipped to keep the network on resonance. The loading capacitor and the drive level were adjusted to set the required anode current coincident with the RF power output peak.
Linear amplifiers do not work in the same way. The amplifier should always be adjusted for maximum RF output power and not by looking for a dip in anode current. If the output is too high the drive can be reduced to the required level, whilst peaking the output power for this setting. In this condition the valve is loaded as heavily as possible and this gives the best efficiency and linearity. Eimac covers this in their engineering bulletins and they are well worth reading. In an old ARRL book on single sideband they have a cartoon picture that is reproduced as Fig 15.2. (There is also an excellent chapter by Warren B Bruene, W0TTK, who designed the Collins 30S-1 amplifier and invented the modern SWR meter that is so widely used today.)
Fig 15.2: Cartoon from the ARRL book ‘Single Side Band for the Radio Amateur’.
In the grounded grid amplifier the most sensitive indication of correct loading is the grid current. When optimally adjusted the grid current should be a minimum. Typically, the grid current meter should be not more than ~30% of the anode current. If it is much higher then it indicates that the amplifier is either not resonated correctly or not loaded heavily enough.
It is a sad fact that some of these problems can be laid squarely at the door of the manufacturer for not having sufficiently good handbooks for the equipment. Some of these manufacturers are Japanese and although the original Japanese version of the operating manual may have been correct, they lose a lot in translation to English. I have come across some classic mess-ups in the translation and when I do I always send a correction, for which they normally thank me profusely and change the documentation to correct the errors. It seems proof reading is fairly low on the priority list of some manufacturers. Unfortunately, it seems that all the ones you find on the various sites on the Internet for free downloading are the earlier uncorrected versions.
Metering
Some commercial linear amplifiers have poor metering features. In a grounded grid amplifier, it is important to be able to measure the grid current and the anode current simultaneously. Unfortunately, in an effort to cut costs, many commercial amplifiers have only a single meter that is switched between different functions by a front panel switch. Some amplifiers do not have a grid current position at all. This is certain to cause problems. One commercial amplifier I examined has two meters on the front panel, one is dedicated to the anode voltage and the other can be switched to cater for RF output, anode current and grid current. This is wrong; the first meter should read anode current and not voltage. Many have just RF output, anode current and anode voltage, selectable using a single meter, and that is all. In days gone by, when amateurs constructed amplifiers and meters were freely available on the surplus market, often as many as four meters featured on the 19in rack front panel.
The other factor which Walter and his clan seem to not understand is that a moving coil meter is heavily damped and will not follow a rapidly changing signal accurately. When using SSB or CW signals the meter lags a fair bit behind what the actual current is doing. The input signal is varying at a low audio rate, especially for SSB operation: the human voice is an average of about 30% of the actual peak value. Although the anode current meter reads correctly with the amplifier keyed up in the idle mode, where just the valve quiescent current is flowing, or when the key is down with a carrier, and it really is a true DC condition, on a speech-input signal it reads incorrectly. The average of a true sinusoidal waveform is 0.637, so if the speech signal were considered the meter only reads about 60% of the true current. Because the meter is heavily damped, the real average is more like 33% and that is what it displays when driven by a speech waveform. Although the meter tells us it is peaking up to 300mA, it is actually much higher. The peak anode current under this condition is about 1A. This may well be too much for the valves.
The same argument applies to RF output power meters. Not only is the moving coil meter heavily damped but the RF detector is reading the average RF power and not the true peak power. Some simple and inexpensive modifications can be made to convert the average reading power meter to a true peak reading wattmeter. The way that some amateur equipment manufacturers attempt this is to fit a large electrolytic capacitor to try to hold up the DC-detected signal. However, this also does not work correctly. By the time the capacitor has begun to charge up, the signal has dropped again, so it never really approaches the true peak value. All they achieve is to hold the detected DC voltage a little higher, perhaps as high as the RMS value, but certainly nowhere near the true peak power reading.
To catch the waveform at the peak needs a very short time constant detector and a low value peak holding capacitor. An excellent peak hold detector can be made using a dual op-amp and a few components. Although it is simple, it gives a reading that is ~99% of the true peak value, which is more than good enough for most purposes. The basic circuit is shown in Fig 15.3. This simple add-on circuit provides a true peak reading meter and maintains the same sensitivity. Therefore it can be inserted between the detector diodes and the meter without altering the calibration. D1, D2 are 1N4148 / 1N914 diodes. C1 should be a polyester low leakage capacitor. Resistors are 1/4W. To increase the hang time make R3 larger, not C1. IC1 can be any dual op-amp or even two 741 types will also work fine. Current drain is a few mA.
Fig 15.3: Peak reading adapter for wattmeter.
Bird Electronics make a version of the Type 43 wattmeter which is a true peak reading wattmeter, but they are expensive. They will, however, work down to as little as 1µs pulse length and give a true peak reading. I used one when I was developing high power pulse transmitters for military jamming systems. The model number is the Bird 4391A and it takes two of the type 43 slugs to measure forward and reflected power.
Checking Your Wattmeter
To check your RF average reading power meter is simple. Tune up using a carrier to some known power level, say 100W if your power meter has this as one of the ranges. Then switch to CW and set the electronic keyer to maximum speed, and send a string of dots. The meter should now read exactly 50% of the carrier condition. The reason is that the dot-space ratio of a correctly adjusted keyer is 1:1. Hence, it gives 50% on and 50% off. Your reading may be different and there are two possible reasons for this. Firstly, the keying shaping circuitry in the transmitter slows down the rise and fall time of the keyed carrier making the total on time a little shorter than a true 50% duty cycle. The other possibility is that the keyer is not a perfect 50% ratio. However, this is easy to check with an oscilloscope. You can also send a long string of high speed dashes. In this condition the power meter average power should be 75%, because the dash to space ratio is 3:1 for a correctly adjusted keyer. If your power meter reads radically differently it needs sorting out.
To tune up a linear without stressing it, use the fast dots as it will only be running 50% average power but the peak will be 100%. This is the way that I tune up. I have a true peak reading wattmeter, with a fast sample and peak hold circuit, just one dot sends it full scale where it hangs for about 1s before dropping back to zero.
I included details of a deluxe tuning aid in Circuit Overload [2] which has a selectable duty cycle pulse generator for keying a transmitter. This allows any duty cycle from 10% to 100% to be selected in steps of 10%.
What’s this ALC connector for?
This is one of my major topics. I have already mentioned it, but you may have missed it. I offer no apology for repeating this vital information. If you think I am obsessed with this topic, you would be correct. But there is a valid reason why I get hacked off about this.
When I am trying to have a meaningful contact on the band with my friends and Walter parks himself 10kHz off our frequency and then proceeds to splatter over the majority of the band I get really angry. When you move up to see who is making this terrible racket and call them, politely pointing out they are spreading badly, they take umbrage.
Whenever I visit an amateur station that I know has an amplifier I always have a peek around the back of the amplifier to see if they have the ALC hooked up to the transceiver. In more than half the cases they don’t. When I point out that they should really make the effort to fit the correct ALC jumper cable, they look at me with that look!
The purpose of the ALC cable is to prevent the transmitter from over-driving the amplifier. If the amplifier is over-driven it will splatter, no matter how well they think they have adjusted the drive level. The reason is the meter lag story. So if you have an amplifier and do not have the ALC hooked up correctly, I hope you are going to do something about it – pronto.
I have also come across some older commercial amplifiers which do not have any ALC facility provided. I shake my head in disbelief that anyone can be so ignorant to the benefit of having ALC in a controlled loop connected between the amplifier and the transmitter. Fortunately, in the modern day most amplifiers have this feature fitted, but whether the user appreciates the need and how to connect and set up the system is another story.
Valve Failure
If you follow all the correct operating recommendations the likelihood of experiencing a valve failure should be a remote possibility. However, nothing is certain in life and you may be unlucky and have a valve go west. When this happens it can be quite a spectacular event! Loud bangs and blinding flashes, accompanied by lots of smoke and sometimes flames are tales we hear. There is always the strong possibility that a certain amount of collateral damage will also be associated with a spectacular valve failure. This may damage the high voltage power supply or cause irreparable damage to tank circuit components, such as band switches and tuning or loading capacitors. A really big valve failure has the capability of wiping out the entire power supply if you are really unlucky. Provided the correct fuses are fitted this is not likely to trouble you as they should blow if the big event happens.
Some of the things that can induce an apparently healthy valve to decide to commit suicide are under the operator’s direct control. One amplifier I examined, which had a serious meltdown, destroyed both valves and also did a significant amount of damage to the power supply and other parts, but it started off with a stupid mistake.
How to Kill an Amplifier
The operator made a silly mistake when changing bands. The station has a
3-element Yagi for 10, 15 and 20m and also a 40 and 80m trap dipole. In the rush to change bands the operator forgot to change the antenna switch for the correct band. Consequently, when they tried to tune up the amplifier it didn’t like the SWR the wrong antenna presented to it and it arced over the anode tuning capacitor. This caused the band switch contacts to melt and the valves went bang very soon afterwards, taking out a lot of the power supply in the process. Fortunately, the slow-blow fuse in the mains input eventually blew and saved the mains transformer.
To some extent, modern day solid-state HF transceivers have lulled us into a state of over confidence. If the operator had not been using the amplifier, the very high SWR would have caused the reverse power detector in the transceiver to fold back the drive, so protecting the output transistors. External high power linear amplifiers using valves of the amateur grade never have this luxury. Consequently, if you try this trick it is almost guaranteed to cause some damage.
To destroy an amplifier is dead easy. Simply tune up to full power into a dummy load. Un-key the transmitter, remove the dummy load and key the transmitter again. This is how we used to test military grade amplifiers, but they had lots of protection circuitry including high SWR protection. If the antenna SWR was such as to produce a high reflected power, the amplifier simply tripped out. However, some modern amateur grade linear amplifiers do also have this level of protection – and it is well worth having because, no matter how careful an operator you think you are, sooner or later you will transmit into the wrong antenna or, worse still, without any antenna at all connected!
Tuning Capacitor problems
If the tuning or loading capacitor arc it causes pitting and sharp edges. Good capacitors have all the edges of the plates carefully rounded to deter corona discharge effects. A sharp edge is a stress riser that will encourage tracking at high voltage. So when the capacitor arcs the first time you can pretty well guarantee it will do it again, but at a lower power. The only permanent solution is either to dismantle the capacitor and carefully smooth off the ragged edges and provide enough radius, or to replace it. The quality of the anode tuning capacitors I have seen in some amateur amplifiers leave a lot to be desired. Again, it is a cost driven thing.
To withstand the calculated peak voltage into a poor SWR antenna, the gap between the plates needs to be much larger. You can estimate that at full output power the peak-to-peak RF voltage across the anode tuning capacitor will be almost twice the anode supply DC voltage when supplying a perfectly matched load, that is an SWR of 1:1. Into a 3:1 SWR you need to almost double that figure to have a sufficient safe margin. The problem is that doubling the gap reduces the capacitance to half: to compensate it means the capacitor has to be physically larger and then it won’t fit in the available volume.
Many years ago I was involved with one of the major capacitor manufacturers (Wingrove & Rogers, Liverpool, trade name of Polar) and I learnt quite a bit about tuning capacitors. W&R made some monster capacitors for companies such as Marconi, which were used in high power naval transmitters. The proof testing of these is a specialised job. For normal low power capacitors it is OK to use a 50Hz or 60Hz voltage source. But for very high voltage types RF is better. The reason is that at radio frequencies the proofing test is more accurate. The RF voltage produces a break-over condition at a lower voltage than 50Hz for the same air gap. So although it would pass the 6kV or 25kV AC test using 50Hz it would begin to arc at a lower RF voltage. I never did find out why, but I remembered the factor.
Other Tuning Capacitor Problems
The other disturbing fact about some high power anode tuning capacitors is the quality of the rotor grounding mechanism. This is the rotating joint between the rotor shaft and the grounding terminal on the end plates. At high power the RF current flowing in this grounding strap or bush can be very high. If you consider the reactance of the capacitor and the anode RF voltage it explains the problem.
Assume the anode DC voltage is 3kV and the RF voltage at the anode swings up to almost twice this value at full power. Suppose it gets to 5kV peak-peak at full output. If the capacitor, when resonated, has a value of 100pF and the operating frequency is 30MHz we can calculate the RF current flowing to ground. At 30MHz a 100pF capacitor has a reactance of 53Ω. The 5kV peak-peak voltage is equal to an RMS voltage of 1768V. An RMS voltage of 1768V with a reactance of 53Ω will generate a current of 33.4A. This current is flowing through the grounding mechanism and if the contacts are a bit erratic, because of poor design or lack of maintenance, the possibility then arises that it may suddenly make a bad contact when we have full power generated. This will throw the whole tank circuit out of kilter and is almost guaranteed to cause a flashover in the network. Oxidation of the contacts is a common problem and this can ruin your day!
REFERENCEs
[1] RF Design Basics, by John Fielding, ZS5JF, published by RSGB (2007).
[2] Circuit Overload (the bumper book of circuits for the radio amateur), by John Fielding, ZS5JF, published by RSGB (2006).
