2 x 3-500Z  /  80-10 meters

( NOTE: Please do not ask for schematics or a how-to for this amp. This article is meant to give you ideas for your own custom design. Every bit of documentation in existence is presented in this article. I build them as I design them and keep no notes!)

This was another "junque box" project and had several objectives, some rather unique. I started thinking about this design when it became obvious that I needed to re-build my old trusty 2 x 4-400A amp. That amp deck had never been upgraded to cover the WARC bands and was a drive hog. With a 100 watt exciter, it wasn't possible to achieve a full 1500 watts output (at the available anode voltage).

So the first objective was to create a design that would produce a minimum of 1500 watts output on all bands 80-10 with less than 100 watts of drive power (and use the existing power supply).

The second objective was to wind up with an amp deck that integrated into my existing facilities with a minimum of effort and to be easy to use in my HF operating environment. This meant that the tubes selected had to be "instant on" and precluded the use of the Russian triodes or 8877, etc.

Third was a self-imposed mandate to NOT buy anything to build this amp. I have been gathering and saving components for 50 years, time to use them up! If some of the parts in the photos look old and crusty, it’s because they are! Included in the final design are parts from TV sets, salvaged military and commercial radios, computers, and various other electronic devices. Some date back at least 50 years!

Fourth was to make the thing "look better" than most of my projects. I always start with good intentions but after all the "mods" and "improvements", my projects usually look like the junk they were made from.

The fifth (but not last) objective was to try out some circuit "tricks" and "do-dads" that have been on my list of improvements needed on some of my other amps.

The first step in any amp project is the selection of the major components. Using the above objectives as a guide, I rummaged around in the old "junque box" looking for the right parts. I had a NIB pair of Amperex 3-500Z's so that part was easy. I also had a pair of Eimac air system sockets that I had picked up over thirty years ago so that was a match. Digging deeper produced a chassis, some coils and caps, a few switches and a bunch of smaller parts. A long time was spent placing the possible parts on the chassis and moving them around for the best compromise layout. Several times it was necessary to toss a component back in the "junque box" and look for a better fitting part.


Here’s a photo of the top of the partially assembled chassis.

In the above photo, you can see the tank components, tube sockets, filament transformer, and fans. The black object just in front of the filament transformer is an air duct which directs the air flow from one of the two fans down into the chassis bottom. I was worried about cooling the tubes during RTTY operation and this was the solution I came up with. It provides ample air flow while meeting all the necessary mechanical requirements.


In this next photo, we are looking at the back of the partially assembled chassis:

In the rack where this amp will reside, there is no clearance on the top of the cabinet. Air must be taken in from one side (or the back) and exhausted out the other side. The chassis base I had “in stock” was only 3 inches high so that prevented a fan from being mounted below. The duct was needed to allow the fan which cools the tube pins to be placed above the chassis where there was room to mount it.

On the right you can see the screened exhaust port. In operation the ducted fan pressurizes the bottom and forces air past the tube pins. The other fan moves a lot of air across the tube anode seals. Both air streams exhaust out the right side. The fans are speed controlled, more about that later.

The sides, top and bottom of the cabinet are made from salvaged aluminum sheet cut to size. The chassis base is a commercial Bud product that I had “in stock” as well as the 19 inch rack panel. The pieces are screwed to ½ inch aluminum angle stock.


Here’s a shot from the side:

In this shot, you can see the feed-through caps which the cathode network relay control lines pass through. The small coax is RG-142/U and connects the PI network to the output connector located below the chassis. Both this coax and the HV lead (seen between the tube sockets) caused me some grief. I wanted to drop them straight down but there was no easy way to do that due to components below the chassis. They had to be routed across the top to a place where I could pass through without mechanical interference.


This view shows the top of the chassis with the tubes installed. Except for a few minor components, it is complete:

The silver plated inductor on left is the 10 meter tank coil. It actually has one turn too many for optimum "Q" but since it worked out OK, I just left it as was. The edge wound inductor used for the remaining bands actually has one or two turns too few for optimum 80 meter tank "Q", but it too worked out OK so I left it alone as well. The edge wound inductor actually has a lower measured Q than a same size same inductance coil wound from heavy wire or tube. It was used it because it was on hand and it was very easy to install the band taps with the matching clips.


Here's a close up of the tubes. The parasitic suppressors are copied from the Eimac 3-500Z data sheet and had to be replaced after the first time 10 meter operation was attempted. The inductance was way too high. I wound new inductors with approximately half the inductance and replaced the metal film resistors with Ohmite OY's. The small inductor you see just above the left hand tube is used for the "L" network for 10 meters. This compensates for the high output capacitance of the tubes and layout. Without the "L" network compensation, the tank "Q" on 10 meters is too high.


Moving on to the bottom of the chassis:

 This photo was taken before the cathode matching network board was installed. Top left is the adjustable power resistor used for the filament inrush current limiting. Below that is a 6.3 volt filament transformer connected as a line boost. The filament voltage at the 3-500Z pins was right at the low end of the spec, so I put in the boost transformer to raise it to the middle of the spec, allowing for minor live voltage variations. Below the boost transformer, you can see the hole and air duct for the top mounted fan.

If you look closely at the tube sockets, you can see that the grid pins do not run directly to the chassis. They are bypassed to RF ground with three different sized capacitors on each pin.  There is a 250 pF Unelco mica UHF capacitor connected between the pin and a copper strap bolted to the chassis. These are the same capacitors that are used for solid state UHF amplifiers and have extremely low stray inductance and a series resonance in the microwave region. In parallel with each of the Unelco caps is a 1500 pF disc ceramic cap and a 0.01 uF disc ceramic cap. This configuration eliminates all the usual problems associated with bypassed grids and is RF-wise equivalent to a direct ground connection. Having the grids DC isolated from ground makes bias and metering much easier.

At the lower edge of the image, there is a 28 volt transformer which provides power for the fans, the control board, and the cathode network relays. This transformer has AC power applied at the same time as the filament transformer.

Unlike all the rest of my amp projects, this control board has multiple functions. Usually I make separate modules for the filament inrush limiter, bias, etc. This time I thought I had my act together enough to put them all on one board. This proved to be a mistake and in the future I will put each function in it's own module.

Filament inrush current limiting is accomplished with a power resistor in series with the line side of the filament transformer. A set of relay contacts shorts out the resistor after a preset time delay. The value of the resistor was adjusted to have the initial inrush current be the same as the secondary inrush current. Next time I will use two stages to get better control of the inrush current.

The relay coil is supplied from the 28 volt supply through a series resistor and a parallel capacitor. The relay is a "TV power" type and has a 9 VDC coil. The capacitor value was selected to allow enough time delay for the filaments to reach equilibrium before the realy energizes shorting out the current limiting resistor.

The 28 VDC supply is a positive ground supply. I did that to make the grid bias circuit easier. With the grids above DC ground it made sense to have a grid bias system instead of the usual cathode bias arrangement. The bias regulator consists of a common adjustable three terminal regulator and a darlington pass transistor configured as a shunt regulator. It has sufficient voltage adjustment to allow the Eimac or the Amperex tubes to be used. It will handle several amps of grid current.  The grids are routed to the bias supply through a 1.0 ohm resistor. Leads from that resistor are brought out to jacks on the rear panel allowing grid voltage and current to be measured. I do not
normally monitor grid current during operation.

The two fans are 28 VDC muffin fans. Each one will provide sufficient air through the two air system sockets to meet Eimac's cooling requirement for 500 watts dissapation for each tube. With two of them running the noise level is pretty high. It turns out that with only the filaments running on the tubes a lot less air is required for cooling so I decide to slow the fans down when max air flow was not required. The DC type fans allow the speed to be easily reduced by simply reducing the applied voltage. A series resistor is placed in the 28 VDC line and switched out with an opto-coupled relay. The tube cathodes are returned to ground through three 3 amp diodes. Not only do these povide some measure of safety bias in the event of a bias supply catastrophy, the small voltage across them triggers the opto-coupled relay. The fans go to full speed anytime current flows through the tubes and drops back to slow speed when the tubes are not conducting.


The next shot shows the cathode matching network board in place:

A printed circuit board was contemplated for this module and rejected due to the fact that it would be a one of a kind board. I could not justify the time and expense required for that. A piece of double clad copper board was cut to size and the relays glued to it dead bug style. Not only is dead bug cheaper, it allows shorter leads between the coils and caps and has lots of ground plane to connect to. The relays are SPST 9 volt coil jobs that I had procured some time ago for another project and didn't use. Each pair of band related relay coils are connected in series and routed to the proper band switch position. The -28 VDC source supplies power through an appropriate dropping resistor for 18 volts across the series coils.

The PI matching networks are designed for a Q of 2 and are constructed with small torroid cores and small silver mica capacitors. Several small value caps are put in parallel to give the correct capacitance. Load resistors were temporarily connected from the tube cathodes to ground and the individual matching networks are adjusted for best return loss (SWR) by squeezing/spreading the turns on the torroids and/or altering the value of the capacitor on the tube side ot the network. Because the correct torroid material and silver mica caps are used, the network efficiency is very high. If a builder tries to shortcut by using slug tuned coils and/or ceramic caps, the input efficiency will suffer and more drive will be required. An input SWR of 1.1:1 or less is obtained across all 8 bands except for the top end of 80/75. That band is just too wide!


The next photo shows the completed amp deck installed in the rack:


 And the full rack with some of my other HF amp decks. Top one is the 3x4-400A deck for 40 meters. Next is the 4x813 deck used on 160. Below the meter panel is the 2 x 3-500Z deck and bottom is the YC-156 deck. At the very bottom is a HV supply (for the 6 meter GS-35B amp not shown here).


And here is the 2 x 3-500Z deck in the rack from behind:

At the time of this writing, this amp has been in operation for several months. It produces at least 1500 watts output, key down CW, for 65 watts drive power from 8-10 meters. It is stable, easy to tune, and quiet during RX periods. It fits my requirements quite well and I consider it to be a success. Although I did a few unconventional things in the design, the results have proven the usefullness of thinking "outside the box" when designing a custom amp.

Addendum: Since the amp requires only 65-70 watts drive for optimum performance, and since I can never remember to turn down the power on the exciter, I made up a 1.5 dB attenuator and placed on the input.

With the attenuator in place, I can leave the exciter run at full 100 watts output and not worry about overdrive. The resistors are 3 watt metal film and the assembly can run 100 watts input power key down. The return loss from 1.8 through 55 MHz is more than 30 dB.