(Updated in August 2010)
We "hams" seem to have a common ailment of wanting to always change things, even when they work okay. Think of the many antennas we have constructed over the years trying to find the perfect radiator. You might surmise that this is the basis for which new ideas and technologies are born. In the summer of 2004 I conceived the idea to build a solid-state amplifier using power FET enhancement mode transistors. It was a great project (www.k0gkd.com/ssamp.html - QST June 2006) and the amplifier continues to chug along easily putting out 600 watts. There have been so few issues with it that I am beginning to think that it will run along forever.
Shortly after I had it up and running I begin to dream about building a one kilowatt amplifier. Why not? It was like moving up the food chain. Once you have the experience why not climb the ladder? To get there I would need a device that would handle a lot of power. I was intrigued with the MRF-154s manufactured by Tyco Electronics - MA-COM. These were the only enhancement mode FETs that would come close to outputting this kind of wattage without going to combiners. Wow - it is hard to imagine that these babies will dissipate 600 watts each - perfect for my second home brew solid-state amplifier.
MRF 154's mounted on a copper spreader.
After mulling over the many hurldles for building an amplifier of this magnitude I begin to formulate a plan. While the MRF-154 amplifier concept is similar to the MRF-150 (600 watt) project it was apparent that achieving 1,000 watts was going to present a raft of new design problems. For starters the power supply specifications change dramatically. The main issue would be the transformer selection that would provide the voltages and amperages required. The specifications in my plan were for a primary of 240VAC, and a secondary voltage of 39vac (rms) that would deliver 30 amps continuously - 50 amps on SSB voice peaks. Nothing off the shelf was available so I decided to get a quote on a tailor made unit from Toroid Corporation of Maryland (www.toroid.com). In outlining what I needed to their engineering department they came up with an 8 ½ inch toroidal core design that matched my specifications - with some input from their designers. The completed unit weighed 33 lbs. and cost $220 plus freight.
Toroid's engineers suggested that I would need at least 5,000ufd of filter capacitance for every continuous amp. At 30 amps this worked out to 150,000ufd. Mouser offered a Sprague capacitor that fit the bill at 75wvdc and 40,000ufd, and I purchased four. Hooked up in parallel I would have 160,000ufd. At $35 each I was now up to $400 in parts and I was just getting started.
I used the same basic design for the rest of the power supply, including the very necessary surge circuit. The bridge rectifier had to handle the excess amperage and I was able to find one that would handle 50 amps. The basic full wave rectifier is indicated below. R1, and R2 were selected more out of necessity as I had four 22 watt 75 ohm sand power resistors on hand.
Main Power Supply Schematic
T1 - Toroidal Transformer (see text) C1-C4 40,000ufd @75vwdc
Note: The bottom side of the primary of the torodial transformer is hooked up via a 15amp fuse to the other side of the 240vac.
R5=20 ohm 20 watt wire wound resistor; RL1 is a solid state 25 amps w/3.0vdc activation relay
("Hockey Puck" solid-state relay with paired SCR output). T1 and the four filter capacitors as noted in the text.
Surge Delay Circuit
R9=47ohms; C1=2,200ufd @25VDC; D2 = 5VDC to 9.1VDC Zener diode @1/2 watt
The power supply capacitance of 160,000ufd acts like a direct short on startup so I put together a surge-delay circuit using a solid-state relay (Hockey Puck) that is rated at 25 amps, and triggers with a minimum of 3.0VDC. The circuit shown above has about a two second delay before Q2 is turned on. The delay allows R5 (the 20 ohm resistor) that is across the AC leads of the solid state relay ("Hockey Puck" SSR - solid-state relay series - w/paired scr output) to conduct when the power is initially turned on feeding the primary of the transformer, and inhibiting the surging current. The reduced power activates the bias supply which is used to feed D5. Current than starts to flow through R8 and R9, pulled by the Zener diode D2. This allows C1 to charge which keeps the voltage on the base of Q2 below its cutoff. Once C1 becomes fully charged the transistor (Q2) turns on sending 7.5VDC via its emitter to activate the "Hockey Puck AC" relay shorting the AC across R5 and allowing full power to T1. (Note: increasing the values of R8 and C1 will further delay turning on Q2.) It works great and you can see the action visually. On power up the B+ voltage hesitates at about 30VDC for nearly two seconds, and then jumps up to 55VDC - its idle state.
The other necessity was for metering. So I purchased a volt and an amp meter from All Electronics. The latter had the matching shunt included which is necessary due to the high amperage, and is rated at 50 amps. The voltmeter goes to 100VDC. As noted, both sides of the primary windings are protected with 15 amp slow-blow type fuses.
Bias Power Supply Schematic (28VDC)
(See text for detail)
I used two 12VDC fans also purchased from All Electronics CAT# CF-152 at $5.00 each. Both were mounted on the aluminum heat sink.
The design for the unit was outlined by Helge Granberg in an application note AR347 published in 1990 Communication Concepts offers a circuit board and all of the necessary parts for the MRF154s - www.communication-concepts.com. The basic schematic is shown below:
I made a few changes that I will describe later that improved the suppression of unwanted oscillations. Also I did not use the voltage regulator as shown but substituted a much simpler device. The output is routed to pin 8 in place of the voltage regulater on the circuit board with the middle lead to ground.
The 7812T works great and supplies a regulated 12vdc to the bias circuit and is fed off of the bias supply.
The 154's have to be mounted on copper to dissipate the massive amount of heat. I could not find a thick enough piece of copper so I bought two 3/8" thick spreaders and bolted them together along with an aluminum heat sink. I then mounted the transistors and the circuit board as shown below.
I found an old chassis that had been some kind of IBM device and got everything to fit inside. For the antenna change over I used Russian military 12vdc coaxial relays (note: now I know why the Russians lost the cold war - the relays lasted about three months and have been replaced with more expensive Japanese versions).
On the input for the excitation drive I added three two-watt 120-ohm carbon resistors in parallel to ground, and two-two watt 100ohm resistors (in parallel) feeding T1 in series. These were chosen based on my available inventory. This is the same circuit I used in the 600-watt unit and does a good job of protecting the gates of the MRF154's from excessive RF drive. It takes about 30-35 watts to drive the amplifier to 1,000 watts. There is enough attenuation that if you inadvertently leave the excitation at a typical 100 watts you will have some protection for the MR154's. I have actually done my mistake without harm to the devices.
To input on circuit the circuit board at junction between the series resistor and the grounding resistor.
Granberg's analysis of the harmonic amplitude indicated -30 to -40 dB for the 2nd harmonic, and the highest 3rd amplitude at only -12dB at 6.0 to 8.0 MHz carrier frequencies. This obviously is a problem. The FCC mandates that harmonic output needs to be at least -40dB! Since I planned on using the unit on 80, 40, 20, 17, 15, 12 and 10, I would need separate filters for each band.
Circuit board, copper spreader, heat sink, and muffin fan mounted on the side of the chassis (talk about "jammed in").
Originally I put the filters in their own cabinet and mounted it on top of the amplifier. It looked so bad I took it apart and moved the filter board to the inside of the amplifier. I am in the process of installing small relays to I can switch the bands from the front panel. Meanwhile I have the 20M filter soldered in place only.
I Used the following schematic and table to calculate the components for the Chebyshev filters. Be sure to measure the high voltage capacitors as they tend to have very wide variances in their values. They almost always test lower in capacitance relative to their stated values - sometimes as much as 25%.
Inductive Filter Layout
Using the following table you can calculate the components:
For a 40 meter filter I chose a cutoff frequency of 9.01mHz - No. 2. This will provide plenty of attenuation for both the even and odd harmonics. All you need to do is multiply the Fcc by 10 and divide the component values by 10. Therefore you get 9.01mHz, L1/L5=.56uH, C2/C4=430pF, and L3=1.27. I used Amidon toroids for the inductance; T-130-6 for L1 and L5, and T-130-2 for L3. Their formula for calculating the number of turns follows:
AL for the T-130-2 is 110, and for the T-130-6 is 96. Doing the math results in 7.63 (8) turns for L1 and L5, and 10.7 (11) turns for L3. In practice this worked out right on the money. Of course, different toroids have different formulas so you will just have to research the factors in order to achieve the correct inductance. Capacitors should be have at least a 5kv rating (disc ceramic) to handle the considerable currents. You should check them for their actual capacitance. I have found that the higher rated capacitors are almost always low by as much as 25% in capacitance relative to their stated values!
When I finally got everything together, and having tested as many things as possible as I went along, it was time for the "moment". Unfortunately I had placed the circuit board in such a manner that I could not get to the potentiometers to set the bias. I had to unbolt the board with the copper spreader and aluminum heat sink attached so I could make the adjustments. I was very careful to ground everything with multiple connections. Making sure that the bias pots were set to ground I placed the 53vdc on the drains of the FET's, and fed the bias voltage to the circuit board through the regulator. Working with the first FET I slowly turned the bias adjustment until I got about 3/4 of an amp of idle current. I then went to the other FET and started to adjust its bias. At this point the current jumped up to 10 amps....something was wrong. I quickly shut down and started to think about what was happening. I tried a second time but started with the second FET - same problem in reverse order! I disconnected everything and shut down.
The next morning after sleeping on the problem it struck me that something was setting the FET's into oscillation. The 154's needed some kind of a parasitic or gate suppression circuit. I wound two small coils of magnet wire on a pencil with four turns, and placed these in series with R14 and R15 to the drains of the FET's. I then replaced L1 and L2 with two small ferrite beads mounted on axial leads to add a little extra inductance on the gates. These would be better placed in series with C7 and C8! After I got everything back together I started the bias setup again. This time the second FET came up just like the first one so that I had about 1.5 amps total without any evidence of oscillation. After placing everything back together I fired up the Kahuna. Watching the output meter swing up to 1.2kw on 20 meters was a huge thrill! I now typically drive the amp to 800 watts on voice peaks. Probably speeding up the fans would allow me to run at 1 KW but at the reduced level everything seems cool and the signal reports are no different.
Sometime after the amp was up and running I started thinking about circuit additions that would protect the devices. For starters I found a thermal switch in my junk box that I mounted on the copper spreader. At 80C it opens so I routed the source voltage for the bias and T/R relays through the device. It works great as on one occasion recently I was talking away and all of a sudden the amp just quit. On examination the copper spreader was very hot to the touch and the thermal switch had opened. A few minutes later when things cooled down the amp came right back up.
The next time I take the amp down I am going to add a high SWR protection detector. I have tested the circuit below on the bench and it looks like it will do the trick. I am thinking about installing it inside my SWR external meter and route the T/R relay source voltage through the SWR relay on the board (Note: the relay is in the closed position until SCR1 fires which will open the circuit and the source voltage to the T/R relays). R3 is adjusted for activation at the SWR limit desired.
The amplifier has worked well to date, now for over four years. It is a fun amp to operate - very stable and consistent with excellent power output. I run it mostly at 600 to 700 watts output so it stays very cool. That seems to be more then enough power to work the world, and it looks like it may outlast me! You can email me with any questions at email@example.com.