Last update October 6, 2008
Items for Discussion
(topics to promote thought and conversation)
  • There is a book that was produced by Eimac back in 1967 called "Care and Feeding of Power Grid Tubes". The complete book is available, in sections, in pdf format from the following address (click on the "Care & Feeding" link on the left):

    This should be required reading for anyone contemplating building an amp. Section 4 is of particular interest and Section 4.1.1, page 19, applies to the triode tube.

  • This is a complete scan of the original Care and Feeding document in PDF. It is 174 pages and over 40 megabytes - CLICK HERE

  • IMD - InterModulation Distortion - What is it and how do I prevent it? IMD is the result of mixing within an amplifier. Some transceivers have poor IMD characteristics and are poor to use when testing your power amp for IMD because they will be feeding IMD from the rig to the amp and you will not know where the IMD is coming from. Read more HERE.

  • How do I tap my coil in the plate tank? What size caps should I use for C1 and C2? This is a question that many builders never understand. One of the biggest mistakes made is thinking that the amp should operate with the plate tuning capacitor, C1, set to mid range on each band. There are several good resources for determining the right place for your coil taps and the value of the plate tune, C1, and load, C2, capacitors. Before you begin, remember that these things are determined through a process that insures that your tank circuit will operate with proper Q. The importance of maintaining the proper Q can't be stressed enough. If your Q is too high, you will have very high circulating currents in the tank. These high currents can heat the components to the point that they melt the solder right out of the connections. If your tank Q is too low, tuning will be extremely broad, efficiency will be low and suppression of harmonics and other spurious emissions will be poor. Design your tank for a target operating Q between 12 and 15 for best operation. The actual resulting Q may fall between 10 and 20.

    Here is a relatively simple method to determine the tap points on the Pi Net coil:
    • Note that it is important to be able to identify the values of your capacitors and inductors. There are some that say they have good results using the MFJ analyzer for this but I prefer the AADE (Almost All Digital Electronics) L/C meter for this purpose. Without such a tool, you will find it very difficult to get good results from the following procedures.
    • It is important to understand that you have to attack this problem backwards.
    • The first step you must do is to know what your plate load impedance is at your expected power output with the known plate voltage and plate current. If you don't know these, you will have to approximate them. The amplifier operating in the transmit mode with drive applied and it out of tune will approximate the maximum plate current you are going to see at the loaded plate voltage.
    • Use your calculator or one of the links below to determine your plate load impedance. If you use my spread sheet, you will also have the values for the next step and that is the actual values of c1, L, and C2 for the particular target frequency.
    • Note that the values indicated INCLUDE stray capacitance and inductance. For retrofits using the GI-7 or GS-35, the internal capacitance of these tubes will contribute a little to the capacitance (be sure to check the data sheet of the tube you are going to use). Tubes like the YC-156 will contribute a LOT more capacitance.
    • Now all the following operations will be done with the amplifier OFF and Disconnected from the power. Be sure that the plate supply has completely discharged before you do anything else.
    • Pick yourself a 1/2 Watt carbon or carbon film resistor (not wire wound) that is the same or as close to the same as your plate load impedance. Connect this resistor between the anode of the tube and chassis ground. Leave the tube in place.
    • The next step is to force the antenna relay to the transmit position. You can do this mechanically with something like a tooth pick or you can electrically close the relay with the proper voltage applied to the relay coil. Caution, do NOT do this with the primary power applied to the amp!
    • Your next step will to be to preset C1 and C2 to give you the proper capacitance indicated by the spread sheet for the particular band of interest and set the band switch to the proper band. Note that at this point, that INCLUDES stray capacitance so the actual values of these capacitors will be slightly less than the values on the sheet. Be careful here to set these accurately. DO NOT set these capacitors to an arbitrary physical position! Some people will tell you to "just set them to a third or half their mesh". That advice is WRONG!
    • With the caps set, use your antenna analyzer like the MFJ-259 or equivalent set to the frequency you want to adjust to. Be careful to pick a point that will put you in the band and on a band like 80 meters, you may want to use two different points to verify that you have an acceptable tuning range. Feed the analyzer into the OUTPUT RF jack (this is backwards!) and check the SWR. You will want to adjust the tap point on the coil to achieve minimum SWR at the frequency you did the calculations. At the higher frequencies you may have to be very careful to get the tap point correctly placed because there isn't much inductance change for big frequency changes there.
    • What we are doing here is matching the RF source at 50 Ohms to the plate load impedance. This is exactly the same thing you're doing when you tune the amp, just in reverse.
    • Repeat the steps for each band and then double check the tap points. Once you've completed this process, your taps will be as close as you can expect them to be. Once done your plate tune and load controls should come very close to the same points when you tune the amp into a dummy load.
    • Remove the resistor from the anode to ground. Then put the antenna relay back into its normal operating mode. Be sure to disconnect your antenna analyzer.
    • You should be ready for a test of the amp. This system works well if you take your time and pay attention to detail.

    The links below will give you tools to calculate values for your tank circuit:

  • How do I test my power transformer? Most folks can't test much beyond output volts without some kind of test equipment. I have built a High Voltage Dummy Load which comes in very handy. Here are some pictures of a test performed on a hamfest transformer. This transformer was tested to 1.5 Amps output and load tested for ten minutes at 1 Amp output. CLICK HERE
    CLICK HERE for information on the dummy load.

  • How much capacitance do I need on the HV supply? This is a common question and not many folks have a good understanding of it. First you need enough capacitance to eliminate any appreciable hum modulation of the output signal due to ripple on the HV. Second, although you may have enough capacitance to prevent noticeable hum from the power supply, you may still wish to add capacitance to improve the dynamic regulation of the supply. Improving the dynamic regulation will improve the linearity of the amplifier and can yield higher peak power output because the voltage doesn't drop as much on voice peaks. Some people will argue both sides of this discussion. There are those that say "you don't need more capacitance" and others that claim they can "hear" the difference. Just be sure to use enough to prevent undesirable hum and add more if you want to. Be careful, larger capacitors store more energy and stay charged for long periods of time. They can be charged and dangerous for days after removing the power from the mains.
    • Here is a discussion of filter capacitance by Tom, W8JI, from the Amps Reflector <>:

      Actually the ESR of the AC source, the load resistance, and the rectifier type determine the ripple and dynamic regulation.

      In short form: The higher the transformer ESR (equivalent secondary resistance), the less important capacitor size is for anything. This is because the supply starts to act like a R/C filter with the R distributed throughout the AC system feeding the rectifier.

      There will be a point where going beyond a certain capacitance does nothing at all. That point is very low with a high resistance transformer.

      As the transformer and power mains get better and better, an improvement can be had using more and more capacitance.

      At 2000V and 500 mA with a transformer ESR of 200 ohms 25uF is acceptable

      At the same voltage and current with a transformer ESR of 20 ohms values up to 250uF would continue to reduce ripple.

      73 Tom"

  • What happens if B+ shorts to ground? Where does that place B - (which isn't directly grounded)? This will cause B- to rise above ground to the full High Voltage power supply potential! CAUTION! This will blow up meters and metering circuits and bias supplies; possibly cause cathode to grid (ground) arcs, destroying tubes; plus, it can kill you! Remember, the B- is NOT normally at ground potential but is below ground potential by some amount determined by your bias circuit. There is a good bit written about providing protections for this event. Most of them detail using a diode with a large surge current capability to ground from the B- rail. This may actually be a string of diodes but the most important thing is that it carries a large surge current, shunting the rising voltage to ground and blowing a fuse. There is a high likelihood that the diode will be shorted as it sacrifices itself to save you and the rest of your amp. Expect to have to replace the diode, your glitch resistor if it is undersized and fuses in the HV primary and / or secondary.

  • Are we going to use a HV interlock? Where? Is it going to short B+ or just turn it off? I think that an interlock that does any more than turn it off is a waste and dangerous. Shorting the HV with any kind of interlock is asking for damage to the supply. This doesn't mean that an interlock shouldn't be used. On the contrary, interlocks can help save your life but you can't depend on the interlock to replace your common sense. Most importantly, keep your hands out of the hot box!

  • Are you going to put a meter on the power supply indicating B+? I am... and a bright RED LED too.

  • Metering - What are my choices?
    • Plate Current - In the old days we used to measure the plate current, Ip, with a meter connected directly in series with the high voltage. These meters worked very well. They didn't require any calibration and you always knew what was going on. The downside to this kind of metering is that the entire meter is at high voltage potential. In broadcast transmitters it was common practice to mount the plate current meter behind an insulating window in the meter bridge. There have been guys killed by tapping on the front of the plate current meter to see if it was stuck. The high voltage arced right through to their fingers killing instantly. Another problem was a meter that was directly in series with the high voltage would vaporize if there was an arc inside the tube.
    • Grid Current - In days of old and in some amplifiers still in service you find that the grid, instead of being directly grounded to the chassis, is floated off of ground for DC, RF bypassed, and the grid current/voltage is metered through leads to a meter that eventually makes itself back to ground. Once again, we have problems. First is the fact that amplifiers with floating grids are more prone to oscillate than those with grids that are firmly bonded to ground. Second is the same problem as we had with the plate current meter during a big glitch. The grid meter can smoke.
    • Plate Voltage - Measuring plate voltage with a DC voltmeter should be easy, right? Well it is. There are just a couple of problems that need addressing. First is the fact that most high voltage meters have a string of high voltage resistors in series with a meter movement tied to ground making a good voltmeter. The problem with this is that if the ground side of the meter lead is opened for any reason, the entire meter multiplier string AND the meter all soar to the full plate potential. This can be VERY dangerous. If the meter movement itself opens, you have the same problem, high voltage in the meter and a danger to everyone.
    • Proper Metering Today - Take a look at THIS SCHEMATIC.
      In this example you will see the more recent method for metering a cathode driven (grounded grid) amplifier. The key to this entire method of metering is the fact that the negative side of the high voltage is not connected directly to ground. Instead, the B- is carried through a low resistance resistor, R1, which will become the plate current meter shunt, through the keying contacts, the bias circuit, the filament choke and to the cathode of the tube. There is another resistor, R2, which is also very low in resistance, that is connected from the tube side of R1 to ground. R2 is the shunt for measuring grid current. There are diodes placed in parallel with these resistors to limit the amount of voltage that can appear across them which helps to protect our metering. We calculate the value of these resistors to give us a voltage to measure at the maximum expected current. I like to use 1 Volt as the target voltage. Each meter is set up as a voltmeter to measure the voltage across a corresponding shunt. In this case I chose to use 1mA meters. A 1mA meter is a 1000 Ohm per Volt meter. Therefore by placing 330 Ohms in series with each meter (to this point making it a .33 Volt meter), and then adding a 1K Ohm pot in series, we can adjust the meter calibration to read voltages between .33 Volts and 1.33 Volts (approximately). Protection diodes are placed between the junction of the 330 Ohm and the 1K Ohm resistors to the other side of the meter. This limits the meter overload to 100% and avoids smoking the meter movement.
      • Plate Current - In this case, the desired plate current meter is 1 Ampere full scale. If we compute the value of R1 to give us 1 Volt across it when 1 Ampere passes through it, we get 1 Ohm of resistance. At the maximum current of 1 Ampere through 1 Ohm the maximum power dissipated is 1 Watt. Given that we want to design for maximum reliability, I would choose a 10 Watt resistor. This resistor doesn't have to be a precision resistor as our voltmeter is set up so we can calibrate it.
      • Grid Current - Here we have R2 to ground. I have chosen to make this grid meter a 0.5 Ampere meter. Computing the value of R2 to give us 1 Volt with 0.5 Ampere flowing through it we get 2 Ohms. The total power dissipated in this resistor at maximum current is 0.5 Watts. In this case I would choose a 3 or a 5 Watt resistor. Once again, this resistor doesn't have to be a precision resistor since our voltmeter can be calibrated.
      • Plate Voltage - Metering the plate voltage is similar to the old fashioned method with two notable differences. First, we will measure directly to B- instead of to ground. Second, we will meter across a resistor placed in series with the meter multiplier string tied to the B-. Here I chose to make this plate voltage meter measure 5 KV full scale. I chose to use 20 each 51K Ohm, 3 Watt, MOF resistors in series with a 180 Ohm, 3 Watt, MOF resistor. At 5KV, the total current through this multiplier string will be 4.9mA. The voltage across the 180 Ohm resistor with 4.9mA flowing through it is 0.882 Volts. This is well within the adjustment range of our meter.
    • As you can see from this example, you can safely and conveniently meter your amplifier without risk to either you or the meters. You can change component values to suit your particular meter movements. Example would be 50uA meters instead of 1mA meters. A 50uA meter is a 20,000 Ohms per volt meter so you can scale accordingly

  • BANG! What's the significance of these pictures? These pictures show a tube that was damaged by an internal arc. This arc could be from not conditioning the tube properly or from another situation like parasitic oscillation. In either case, the tube is fried. A glitch resistor may help prevent this damage but it will not prevent the cause.

  • What is skin effect? The vast majority of the RF current flows on the surface of the conductor. That is why a large conductor is needed in the plate tank to prevent heating losses. A #12 wire may handle 20 amps at 60HZ but not nearly that much at 14 MHz. In high power RF environments, bigger is better. Silver plating helps lower the IR problem at VHF and UHF but has little effect at HF other than to make the tank circuit look nicer BUT it is a good surface protector. It is better to use flat strap wound thin edge to thin edge in a coil to increase the current handling capability without increasing the capacitance between turns. The down side of this in winding a coil with large inductance is that it takes a lot of length.

  • What is a "glitch resistor"? It may be the one thing that can save your tube(s) when (not if) there is a flashover inside. The glitch resistor limits the amount of energy available at the instant of the flashover and should limit the current to less than 200 Amperes for a tube like the GS-35B and lower for tubes like the GI-7B. The glitch resistor should never fail during a glitch. Placing an appropriate high voltage fuse in series with the glitch resistor can halt the current flow while the glitch resistor limits the current flow. Low voltage fuses should never be used in series with the high voltage.

  • How much iron is enough? If you have enough iron, you can get enough power. The bottom line is that if there isn't enough iron in the core, you can't transfer enough energy to the secondary of the transformer no matter how large your wires are.

  • What if you want to use one power supply for 2 (or more) RF decks? Then you have to consider some things that will help protect you, the amps and facilitate powering both of them. A few of these things might be:
    • Making sure that the blower for the proper amp is running (you do plan separate blowers don't you?)
    • Use a B2B vacuum relay to switch the high voltage to the proper amp; One for each and mounted in the power supply.
    • Lockout in the power supply which senses which amp is "on" so the other one can't be powered up (at least for B +) unless your supply is large enough to handle both at the same time.
    • Will you need a triode control board in each RF deck? You will have to think this through giving consideration to BIAS, METERING, LED STATUS INDICATORS, and control.
    • W4ZT - Some say share the board, some say don't. My gut feel is dedicate a board to each RF deck because of the differing bias requirements of each tube.

  • Physical layout of the RF deck will be limited by the components you have available. Once you have a layout that will work then the front panel layout has to be considered, including meters. If you look at all the pictures, you will see layouts in many different styles like meters across the top, meters stacked on one side, meters side by side in the middle. Try to visualize the placement of the two caps and the coil behind the panel and how they will limit where the meters can be placed. The front panel will also have switches and LEDs. The front panel layout is what you will see and live with every day so it deserves a lot of time making it what you want.

  • How do you test the High Voltage power supply under load without risking damage to your tube(s)?
    Can this be done? Can it be done safely? For how long? Can it be done without spending a fortune? Does it need to be done at all? Why?
    W4ZT - I think it needs to be done to insure that your power supply can and will provide the HV that you expect without a sudden breakdown with your tubes in circuit. It can be done safely and inexpensively with a little planning and work. Can you say high voltage dummy load? Click here for pictures.

  • Do we need to use a tuned input? Some people say that the GS-35B (or take your pick of tubes or combination of tubes) will present your exciter with an acceptable load. But is that all there is to it?

    NO... there's more.
    Eimac's Care and Feeding section 6.1 says: "In a grounded-grid circuit the cathode, or input circuit, is in series with the plate circuit. Because of this, any change made in the plate circuit will have an effect on the input circuit. Therefore, the driver amplifier does not see its designed load until the driven stage is up to full plate current."

    There are TWO important things that come from that:
    • First, the RF current path between cathode and ground back to the anode will be through the tuned input. That means that if you depend on the output tuned circuit of your transceiver, the amplifier RF current must flow through the input coupling capacitor and then the coax to the transceiver. This is the primary reason for cable length problems between transceivers and amps.
    • Second, the cathode input impedance is very dynamic and the tuned input with its flywheel effect helps smooth that out for the driving transmitter while keeping the RF current path short and contained within the amplifier. Keeping the RF current path short and contained within the amplifier reduces Inter Modulation Distortion (IMD) products. a very significant amount. It also IMPROVES EFFICIENCY. This improvement can be quite a few percent!

    • There is one option that a lot of folks don't want to talk about here. That is one of using a swamping resistor to ground from the input of the amplifier. That resistor will provide an RF current path to the cathode without going to the driving transceiver! It will require more drive to the amp and it will not increase the efficiency but it WILL work.

    • In his series of articles "Amplifiers", Part 1 in the section titled Grounded-Grid Versus Grid-Driven, Rich Measures, AG6K, writes the following:

      " What goes on inside a grounded-grid amplifier is not as simple as it looks. The AC component of the anode-current and the grid-current, i.e., the RF cathode current, passes entirely through the cathode coupling capacitor and the tuned-input circuit--so the input circuit is in series with (and out of phase with) the output circuit. The components in the tuned circuit must be able to handle a substantial amount of RF current. Manufacturers of tubes that are designed for grounded-grid operation typically recommend using a tuned input pi-network with a Q of 2 to 5."
    • One of the best discussions of the need for a good tuned input is by Rich Measures, AG6K, in his article about the Heathkit SB-220. You can read the entire document at and scroll down to the "Improving Input SWR" topic about 2/3 down in the page. In part Rich says the following:

      "The job of the tuned-input circuit is more complicated than just matching 50-ohms to the input-resistance of the amplifier-tubes.

      Here's why: The instantaneous input-resistance of a grounded-grid amplifier fluctuates wildly during the positive and negative voltage swings of the sinewave input signal.

      When the input cathode-voltage swings positive, the grounded-grid looks negative with respect to the cathode, and the current is completely cut-off, making the input-resistance nearly infinite.

      During the negative swing in input voltage, the grid looks more positive, and a large current flows in the tube, making the input-resistance very low.

      For example: a pair of 3-500Zs. When the driving voltage is peaking at negative 117v, the anode-current is at its maximum peak, and the instantaneous anode-voltage is swinging to its lowest point of c.+250v, the total, peak cathode-current is 3.4a.16 Thus, the driving resistance at this point, Rin c. 117v/3.4a c. 34.5-ohms, and, incredibly, Ppeak c. 117v x 3.4a c. 397w.

      Thus, the resistance swing is from near-infinite with positive driving voltage, all the way down to 34.5-ohms. [17] The instantaneous drive power requirement varies from 0w to 397w at the positive and negative peaks of the sinewave input voltage. This is not the type of load that makes for contented transistor-output transceivers.

      During the positive swing in input voltage, there is virtually no load on the driver, so the tuned-input circuit must store the energy until it is needed the most, during the negative crest in the input voltage.

      Thus, the tuned-input circuit's job is to act as a flywheel/energy storage system, and a matching transformer.

      Q is like the mass of a flywheel. More Q makes for a better flywheel, which does a better job of averaging the wild swings in input-resistance, giving a lower input-SWR. The tradeoff is that more Q means less bandwidth. This means that, with a high Q, the input SWR may be near-perfect at the center of the band, but too high at the band edges. Thus, a compromise must be made..." (AG6K)

      This is great reading and everyone should take the time to read the text to get an understanding of the need for a tuned input circuit.

  • What is the input impedance of the GS-35B cathode in Grounded Grid? This is a great question! There is no published data to tell us what it is. There are lots of opinions and you will have to consider each on their own merits. Seems that most folks think the actual effective input impedance is somewhere between 25 and 90 ohms. You should remember one important point. When a tube manufacturer specifies an input impedance for a particular tube that is the effective impedance that you should use to DESIGN a network to interface to. It is NOT to be considered RESISTANCE to drive directly with your exciter. The network that should be designed to match that impedance should be a pi-network. The pi-network provides some energy storage (the flywheel effect) which smooths the impedance match reflected back to your exciter. Your amp should never be run without a tuned input for this reason. See the tuned input values above. See the following thread for more:

  • Should the chimney be removable from the anode side of the chassis? This could aid in tube removal and installation but wont really be an issue with a short chimney. You will have to remove the anode cooler to deal with the grid ring clamps in most mount designs. The exception is my socket/mounting fixture design here

  • Filament / Cathode choke - rod or toroid? This discussion has taken many of us 'round and 'round. bottom line is that you need a core, whether rod or toroid, which will provide the inductance while not saturating when subjected to all the currents through the windings. The common practice of using a rod is certainly the most popular method of making the filament choke. A toroid with a slit cut in it will do exactly the same thing only wrapped into a smaller length. For the 3 Amps of filament current required by the GS-35B, a bifilar choke made with #18 enameled wire on a 3-1/2" long 3/8" diameter ferrite rod works fine.

  • What kind of Blower should I use? Cooling the GS-35B is very important. The minimum air flow is 90 cfm. The data sheets on most blowers will show the zero back pressure air flow and then a chart with air flow at different amounts of back pressure. A blower that will do 100 cfm at zero bp wont be enough. It will take one that will do about 140 cfm at zero bp. A blower of that size will have an air outlet that is near that of a 2" diameter round hole. There are several different types of flanged fittings that can be used for a hose if you mount it separately. The only down side to mounting the blower on the RF deck is noise. Of course that is how a large number are done and it will depend on the blower RPM too. I like a blower that turns around 1500 to 1700 RPM as opposed to a 3000 RPM blower. Grainger lists a large number of Dayton blowers and there are others to choose from as well. Take a look here: Take an example like item number 2C647, click on that number to see the data sheet. For that blower at 1500 RPM you will see that the air output is 134 CFM at zero bp but still 96 CFM at 0.4 inches of water. This would be a good candidate for your blower.