Eimac® 3-500Z Tube Linear Amplifier

In 2008-2009, I approached a friend Jim Fish (K7NCG) with a goal to build a linear amplifier.

At the time, I was probably using a SB-200. It was a fine amplifier and helped me get a big signal out.

I acquired some parts from Jim and set out on building the 2 KW Linear based on a pair of 3-500Z tubes.

Most of the parts I acquired would help me complete the project, but I had some work to do for building a chassis and completely re-wiring a new amplifier. Most of the parts needed some care for being refitted and through the whole process I had to ``break down'' the amplifier into the basic systems.

Fast forward 12 years and now I am still using the same 2 KW linear that I made. Along the way the amplifier had been refitted with some new parts and improved circuts for controlling the amplifier.

Old Notes

At the time the amplifier was finished, I wrote up all the notes and put them on these pages.

These old notes are here for reference and mainly document what was the original plan/design of the amplifier.

New Notes

May 2020

I counted the QSO from the log book to see just how many contacts I've made with the amplifer over the years. Turns out, quite a few. Roughly 4,500 contacts or so since running the home-brew linear.

In that time I had been using the same pair of RF Parts 3-500ZG grahite tubes ordered back in 2009.

In May 2020, after being away from the radio world for a while, I began using the amplifier -- mostly making contacts on 20 meters and 40 meters when suddenly a glitch occured and I lost one of the tubes. Short between the filament and ground. The instant it happened, it was preceded by the destruction of the 1 Ohm "protection" resistor in series with the B+ to the plate of that tube.

New tubes were installed in May 2020, and I think the amplifier will be fine for years to come.

Comments about Construction

There were a lot of changes made to the amplifier after it was built. The changes were caused by problems in the design and implementation of the linear.

The baseline to recall is that the amplifier was based entirely on two sources of information.

  1. The ARRL Handbook, specifically older handbooks from the late 60s and early 70s. I visited the Seattle Public Library and scanned through all the older QST articles and Handbooks they had on the stacks for reference. I advise new builders of "old" equipment to take advantage of the treasure of old articles in QST. Visit your library. It will be useful!
  2. The notes provided by Jim (K7NCG).
Most of the "helpful videos" and other articles on the internet (maybe even this one you're reading) are not quite the best source for detailed and accurate technical information. I strongly encourage you to read old Handbooks and scan through the schematics of units that were designed back then to understand what the issues were for design and implementation of a 2 KW Linear.

Main Defects and Solutions

HEAT Number one is heat. The heat has to be removed from the tube component. A very good air cooling system must be the first thing that is planned out as the chassis and layout of the amplifier is designed. Ask yourself -- how will I move as much air as I can past the surface of the tube and out of the chassis? The reason why heat is your enemy with the amplifier design is that it can degrade the capabilities of the amplifier to remain linear during operation. When the amplifier goes non-linear, it causes all sorts of problems not the least of which is a distorted signal and "buck shot" across the band causing QRM.

Key steps to solving the heat problem:

  1. Chimney. By using a glass chimney around the tube, the air flow around the tube is directed to flow along the glass surface and out from the top of the tube. Without them, the motion of the air (the fluid flow of the air) will be distrupted and diffused. The whole tube and especially the areas of the glass and plate will not receive as much air cooling without chimneys around the tubes. Get Chimneys. For the amplifer I made, that is a SK-406, for Eimac 3-500Z. Luckily I had a set of chimneys, so I used them.
  2. Fans. Yes, of course you'll need a fan, but what kind of fan and how many fans? Let's talk this through:

    Most designs I read about involve two basic kinds of air flow management:

  1. Directed air-flow around the tube.
  2. Indirect air-flow around the tube.

By Directed, I mean where the output of the fan is literally driving air past the tube because the fan is just inches away from the base of the tube and the fluid flow of the air is in the direction towards the plate of the tube. Base mounted muffin fans that direct air from under the tube socket up along the side of the tube body and through the chimney is the kind of Directed Air-Flow I am refering to.

By Indirect, I mean causing a fluid flow of air to pass by the tube, even through the chimney, but done so by pressure of the volume of space between the tube base and the top of the chimney. For example, running a "squirrel cage" fan to blow air INTO the chassis box but leaving only one path of egress of the air -- the tube socket base is indirect air-flow. The air will find it's way past the tube socket base and eventually past the tube and out the chimney, but it's a long path and there's a lot of resistance (obstacles) within your chassis to contend with.

The air is dispersed, diffused, and the fluid flow of the air does not easily reach the tube. In addition, it's an issue albeit small one -- leakage of air pressure of the chassis. Every hole, crack, seam that isn't sealed up will also help defeat the passive Indirect air flow.

One can argue convincingly that with enough C.F.M. output either method (direct or indirect) will suffice, but I tried that. Indirect cannot beat direct air flow. To overcome the limitations of indirect air flow you need to push more air which means more C.F.M., which means a louder fan operation. It makes noise. More noise.

There are commercial vendors for linear amplifiers that get the best of both scenarios -- they pressurize a very small cavity, so they minimize the losses of the air-flow resistance in the chassis, but they gain the conveniece of plasing the fan system in an ideal location. They also employ very good chimney solutions so that even with the indirect air-flow from pressurizing the volume below the tube, the efficency of their design is good enough to warrant indirect flow.

That's another thing about the trade off between direct and indirect - Where does the fan go? For direct air flow solution, the fan must be right below the tube socket and for some designs that may not be easy to do, or possible with the way the rest of the amp is setup. This is why designing the fan/air-flow issue before anything else is so critical. It's much harder to go back and retro-fit a fan/air system AFTER all the other parts of the amplifier are setup and placed in the chassis.

Figure out the air-flow problem first

For all the reasons stated, figuring out how you will cool the tubes is job number 1. After that, the jobs stack up. Don't worry, there's a lot more to do, but the cooling system affects so much of the layout of the amplifier, so best to get that out of the way and designed well.

Metering Circuits

When the amp is in use, the monitoring of the plate current and grid current are essential. There are ranges of plate current and grid current that are expected and they tell you information about the operation of the amplifier. The other measurement usually is the Watt meter of either the tuner or other Watt meter attached to the output. But, that's up to you.

Measuring plate and grid current isn't that difficult, but you'll need a few parts to make it easier. The two measurements are based on the voltages on the B- signal. Your linear amplifier has two HV outputs of interest. One is B+. That goes to the plate. That's the "plate voltage". It's the voltage that is going to pull those electrons from the cathode to the plate controlled by the voltage put across the grid. The names of these voltage sources harkens back to when the primary source was via batteries. A, B and C batteries were the names of these sources.

The name "B" for that voltage goes back to the historical design of triode amplifier design schematics. The B voltage is the voltage that governs the Plate voltage. The "A" voltage is what is usually referred to when the voltage across the cathode (filament) is described. And, the "C" voltage (C for control?) is what is used to describe the voltage across the grid for controlling the current between the cathode and the plate. More on grid later.

But the voltage line B- (the negative side of the B voltage source), is what is used to measure Plate Current and Grid Current. B- is not driving the plate. B+ drives the plate. But the B- voltage is what is used to bias the filament.

How to make a good metering circuit: Refer to the schematic below

Notes:

Program the Meter Circuit

By "program" I mean just select R1/R2 ratio that gives the full-scale reading on the 1 mA device for the expected full amount of current presented to the meter.

How to program R1 and R2: Suppose te meter is a 1 mA meter. It means that the meter reads full-scale at 1 mA. So, when 1 mA of current flows through the meter, the meter reads full scale. How do we interpret a plate current then? We typically see plate current between 0 and 600mA perhaps. Ideally, per the data sheet of the 3-500Z, the maximum rating is 400 mA of plate current. So, how do we arrange R1 and R2 values such that full scale of 1 A (1000 mA) occurs on the 1 mA meter? We see that the face of the meter has a right-most marking of 1 A and the scale is divided into units so that the user can read off a value between 0 and 1 A (1000 mA).

We divide the current. Notice that R1 and R2 are current dividers. Suppose the MAXIMUM current is presented to the node at R2 and R1. We want to reflect that maximum value as full-scale reading on the meter. What is full scale of the meter? It's 1 mA. So we want 1 mA of current to go through the meter. The rest of the current? We want to put that current through R1. So if the actual range of current to measure is 0 to 1 A (0 to 1000 mA), then 1 mA goes through the meter and 999 mA goes through the R1 when the maximum current is presented.

Now we simply use Ohms law. If the voltage was 3000 volts, then to get R1 to take on 999 mA means:
3000 V = I * R1, R1 = 3000 V / 0.999 A = 3.003 K Ohms.

But we're going to scale this differently. It's really the ratio of R2/R1 that matters, not necessarily the values per se. We're going to re-scale the values so use this formula: R1 = 3000 V / 999 mA and then get 3.003, so, use 3 Ohms.

If the "full amount of plate current" is present, R1 = 3.0 Ohms will mean that .999 A (999 mA) will just bypass the meter entirely. That's OK.

Now for the meter itself. We want that "1 mA" full scale movement when 1000 mA of plate current is presented. We'll not actually ever want to see a full 1 A of plate current, but the markings on the meter will make it easier to read off the actual plate current if we treat the right most numerical value on the meter face as "1000 mA", hence 1 A. Anyway, how to get the full deflection of the 1 mA meter when 1 A? We already took 0.999 A away when we divided the current. What's left is the 1 mA to flow through the meter if R2 is programmed correctly. Back to Ohms law:

(Re-scaled since we care about the ratio of R2/R1):

3000 V = I * R2, R2 = 3000 V / 1mA. R2 = 3 K Ohms.

But, wait. We forgot another part of R2. R2 isn't the only resistance that the metered current will flow. The meter itself has an internal resistance. That could be impactful! So we need to Ohm out the meter and find what the internal resistance is. (Some older panel meters actually print this on the face in small print next to where the needle deflects. Look closely, or just Ohm-out the meter). The equation then becomes:

R2 + R_meter = 3000V / 1 mA = 3 K Ohms - R_meter. If the meter had an internal resistance of 100 Ohms, then R2 = 3 K Ohms - 100 Ohms, or 2.9 K Ohms.

So, now we have programmed the meter circuit. If ever current was presented that would cause the full deflection of the meter (eg., 1 A) then with R1 = 3 Ohms, and R2 = 2.9 K Ohms, then that meter needle will deflect fully. If 500 mA were presented, the meter will deflect to the center (half way), and so on.

The (-) terminal of the meter is the B- signal that carries on into the rest of the circuit (to the Bias Zener actually).. But the point is, you can measure the Plate Current.

To review:

  1. Find your B- value. Your HV power supply is the source of the B- voltage.
  2. Find a "1 mA" meter. Suppose it measured n mA, not just 1? See below:
  3. Determine how much current youw want to represent "full scale". Choose that current value I (above we chose 1 A, or 1000 mA)
  4. Do the math: R1 = V / ( I(mA) / n). I in mA, n is an integer, V in volts. R1 in Ohms.
  5. Do the math: R2 = V / ( 1 - (I(mA)/n) ). We want the complement. 1 - whatever the current is that is going through R1. Subtract the resistance of the meter itself.
Last -- the power dissipation requirement of those Resistors is important. The small value resistor -- the one that is taking the bulk of the current needs to be a fairly heavy duty resistor. 50W would be safe. There are several packages for 3 Ohm, 50W resistor that would work.

The other resistor, the one that is in series with the meter, R2, that can be a 1 W resistor. It's going to read full scale at 1000 mA presented, so a 1 W resistor would be a safe choice.

The grid current meter works exactly the same way. Grid current is measured between the B- potential and ground. The (-) terminal of the PLATE current meter is connected to the grid current input meter schematic.

We expect grid current to be very low, so we can rescale the R1/R2 programming so that full-scale is 500mA of grid current instead of 1000 mA. The user just needs to be sure the interpretation of the meter is based on how the meter is programmed. Use labels or instructive information on your front panel to make that clear.

For grid current, the (-) terminal goes to the input of the grid current circuit (where R2 and R1 form a node). But ground (chassis ground) is connected to the (+) terminal of the grid current meter, not (-) terminal.

Parasitic Chokes

If we go by the vintage ARRL schematics, the parasitic chokes are three 100 Ohm resistors in parallel with a few turns of a copper strip for a coil. 5W resistors would be safe and smart. That's all the schematic says. We can do better.

First, you just need to cut the number down to 2 resistors per tube. Two 100 Ohm 5W in parallel would be OK. The key difference with the amplifier as described in the schematic and the one I built is that prior to the parasitic choke is a pair of 1 Ohm 3W resistors in parallel. These are playing the part of a fuse. If a glitch occurs and a rise in current begins, the 1 Ohm resistors will be the early part to "blow".

That's just what happened to me --- I had a glitch with a tube (a short to the filament) and the tube drew a huge spike of current and that in turn blew the safety resistor. The parasitic chokes are meant to stifle the parasitic oscilliations that may occur in the event of problems and glitches.

Refer to the schematic below:

The coil L1-L4 are 20 AWG nichrome wire. Each coil is 6 turns around a 1/4" form. The gap between L1 and L2 and the gap between L3 and L4 is no more than 1/4". L1 and L2 are in series. L3 and L4 are in series. In parallel with L1, L2, L3 and L4 is R1, R2, R5 and R6 respectively.

R1, R2, R5, R6 are each 100 Ohm 5W Metal Film.
R3, R4, R7, R8 are each 1 Ohm 3W Metal Oxide.

Connect each "string" of the choke (left string for left tube V1, right string for right tube V2) such that the connections are as short as possible. You might find it preferable to solder the parasitic choke with silver solder. If you do so remember:

  1. Use flux meant for silver solder. It is a hazard (the gasses emitted when the flux boils off is not something you want to breath).
  2. Use heat carefuly. Silver solder has a higher melting point than typical rosin core solder and so you'll need to have clean and effective solder tip that can put the heat directly and quickly. I use "freeze-it" canned cool air to rapidly chill the bodies of the resistors after soldering.
Carefully clean off any flux or moisture on the whole setup.

Refer to the picture below for a sample implementation: It's useful to give a bit of slack on the ends (connection to plate and connection to Plate Choke and Tune Capacitor) so that you can move them out of the way if you need to service the tubes.

It's useful to solder to the ends of each string a eyelet that can fit with a #6-32 screw - common for the cathode heatsink and tune-capacitors hardware. But your setup may differ.

Solder carefully to use just enough as to make a good connection. Best solder skills apply: physical and electrical connections, no excess solder, no unnessaryly long leads, turns, loops, etc.

And mount everything clear by at least 1" from ground (chassis walls), the plate choke, or tune/load capacitors.

Other useful notes: Keep the bodies of the resistors out of the path of the venting air from the tube chimneys. Heat kills parts, over time.

Other Improvements

The fans make noise. The sacrifice of more fan noise is that your tubes are going to remain cool. Cool is good.

But, the noise can be a problem. Even with positioning the amplifier in such a way that the noise is aimed a better direction, or lining the shack walls with acoustic foam, there's still that fan noise.

To solve that I designed a fan control system. I programmed an Arduino board to sense the voltage across a POT. That analog voltage is fed to the ADC of the Arduino. The code converts the digital interpretation to a value that sets a PWM on two GPIO's.

The GPIO's (one each per fan) are identical in value at all times. The GPIO signal drives the base of a power NPN transistor which switches on/off the voltage to each fan. All the usual current limiting resistors are employed so that the 12 VDC source can only deliver the rated current to each fan on the pulse. The POT can be adjusted from the front panel of the amplifier.

In idle listening modes -- I can turn down the PWM of the fan to a low duty cycle. (Waiting for a call in a net, or trying to listen for DX, etc..)

In times when I need strong flow of air, I can adjust the POT higher for greater duty cycle. It's a useful add-on to the amplifier to make it easier to hear low signal stations.

And, since the amplifier I built uses two muffin fans (one per tube) driven by the same 12VDC supply that controls the Tx/Rx relay the voltage source is already there. The only tricky thing to remember is to choose resistors to keep the current within limits of the fan and the NPN transistors you've selected.

I don't have a schematic written out for the fan-control yet, but I will try to do that sometime in 2020. I'll also include the very (very) basic Arduino code involved. It's not that complicated actually.