I recently watched the movie Fat Man and Little Boy, a 1989 great movie about the Manhattan project and the development of the atomic bomb. Paul Neman is the lead character as an army general assigned the responsibility of delivering the bomb. John Cusack plays the role of a very bright mathematician and scientist from the University of Chicago. Toward the end of their efforts many of the scientist, especially Cusack's character, begin to think about the moral issues of what they are about to achieve. The movie centers around the collaborative effort to affect the end result with a lot of very independent minds and ideologies.

The film reminded me of how we amateurs work together exchanging ideas and helping each other with all sorts of issues - antennas, radios, projects - etc. So it has been with this LDMOS amplifier project for me. (Silicon-based LDMOS FETs are widely used in RF power amplifiers for base-stations where requirements for high output power with a corresponding drain to source high breakdown voltage ratings are required - typically cabable of very high frequencies.) Exchanging ideas with many of my ham friends for this project has been the key for its success. After two failures with LDMOS chips it was pointed out to me that the source to drain voltage ratings of the devices being tested was not high enough to take the 55VDC analog power supply being used. As a rule of thumb the rating should be divided by a factor of "2", and that would mark the upper limit. The original unit we used had a rating of 100 volts so anything over 50 volts was likely to result in a "failure". The NXP BLF 188XR has a rating of 135 volts...high enough to use a 50VDC source.

NXP BLF 188XR

Hard to imagine how this tiny LDMOS transistor can output 1.2 KW! Viewing the NXP device was scary when thinking about heat dissipation. In dealing with this issue some designers have channeled out the copper plate used to disperse the heat and soldered the units within the indentation. This seemed extreme to me so I just went with the conventional application by bolting it directly to a copper spreader with a good dose of heat transfer silicon. This is working okay as the heat gets moved away from the transistor sufficiently to stay within its design parameters - the junction temperature limit of the NXP transistor is 225C (I would think it would melt at this level!). Some hams have used a clamp on top of the unit but I did not see the advantage and worried about possible damage. LDMOS chips are "Gemini" - literally translated as "multi faced" - and have two transistors incorporated within the same chip. This is the perfect combination for a "push-pull" design and significantly reduces the cost. One can buy the Freescale models or the NXP ones for about $200 in single units!

Having built two MOSFET amplifiers in 2006 (see QST June 2006) I wanted to incorporate all of my knowledge and experience into the new LDMOS project. The NXP engineering group provided their ideas for a HF schematic as shown below along with a list of materials. (The feedback resistor looked a little light to me - FYI they don't show the bias supply and regulations circuit.) Note: Rock Qiu at NXP's research division will probably help with questions and ideas for HF applications.

I could have "rolled" my own board following the schematic provided by NXP. However my objective at the outset was to use offerings from third parties to lessen the project's obstacles. Most of us have busy lives and don't have the time to put it all together - including me. Because of my growing relationship with Baruch Zilbershatz (4Z4RB), and his know-how with LDMOS devices, I decided to purchase one of his hand made boards. Using Skype and email we communicated regularly about the design and application I had in mind. His experience with other hams and overall expertise has been a big plus.

Building a solid- state MOSFET amplifier requires several basic components: (1) a power supply that can provide about 50VDC at 30 plus amps (either a switching or analog design), (2) a circuit board with bias voltage regulation, input and output transformers, (3) a low pass filter assembly to attenuate the harmonics, (4) a copper plate usually " to 3/8" thick to allow the heat from the devices to spread out, (5) a heat sink (usually aluminum) to dissipate the heat from the copper, and: (6) relays and auxiliary equipment to manage the amp. Basically the circuits for solid-state LDMOSFET amplifiers follow the same basic design parameters as those for MOSFET devices. In simplest terms the transistors have very low input and output impedances so the drive has to be lowered and the output increased to feed the 50-Ohm source and load. The input and output transformers are usually engineered with 3:1 (sometimes 4:1) and 1:3 ratios, in RF terms - 9.1 and 1:9, in and out (the square of the turns). The data sheet for the BLF 188XR shows the following impedance at 108 MHz in push pull at full output - 2.94 j9.64 Zi and 2.74 + j0.57 Zl.

Baruch appeared to follow the NXP model using toroids hand wound for the output stage. The board features a high power transmission line transformer wound on stacked ferrite toroids modeled after a design by ON9CVD - resulting in lower IMD and less saturation at higher frequencies. Baruch noted that the length of the cables was kept under 1/8 wavelength of the upper frequency limit. The mix for the toroids is typically 43 material. The front end of the boards should have good voltage regulation. A potentiometer is used to adjust the voltage on the gates and turns the LDMOS "on" - this nominally happens at around 1.9-2 volts DC. A good idle current setting is 200ma. (Note: since the LDMOS device has two transistors built-in they come on concurrently!) Because of the very high gain delivered by the NXP unit (around 30db in the HF range) I decided to add a feedback circuit consisting of two 25-watt Caddock non-inductive resistors (450 ohms) and two blocking capacitors (0.01ufd) - soldered to both sides of the drains to gates.

The circuit board mounted on the copper spreader and aluminum heat sink. Note the thermal disconnect switch at the bottom of the copper spreader.

After getting the board mounted on the copper (with a lot of silicon heat transfer grease) I tested it on the bench. Setting up the bias was routine. True to form the unit started to conduct with about 1.9VDC on the gates. Using a dummy load on the output and relatively low voltage on the drains with a small RF signal there was a marked gain on the output as viewed on my oscilloscope. It appeared to be ready to go.

Having tried an analog power supply that was the cause of the earlier failures I decided to use a switching supply where the voltage could be set to lower levels. Luckily I found a "slider" built by Hewlett Packer (HP 226579 ESP 120) for computer application that outputs nearly 3,000 watts at 50 plus volts. It looks like the lock boxes used in old bank vaults. For $40 plus $15 for shipping it arrived in good shape but without a schematic or any information on how to get it operational. The supply voltage was 240VAC and the output voltage set at 53.5VDC - wow, a beast with an output rating of 50 amps. I wanted to get the voltage down because of the issues with my earlier failed attempts.

The HP Slider

There was very little information on the net as to how it could be reduced. Thanks to Bruce Clark (K0YW) he was able to direct me to an article where the author had lowered the voltage to about 48, and sent me information on the critical pin connections that turns the unit "on". The bottom side of the unit is shown in the picture below...notice the pin marks in red. The three pins need to be shorted together for the supply to activate. I soldered the 240VAC three wire feed to the right side with the common lead to the center and tied them so they would not pull lose. I applied a generous portion of liquid insulation to all three leads. On the left side I put together a cable with different colored wire for the plus and minus DC output using number 10 wire to handle the current. Male and female plugs were added to the cables to disconnect the supply from the amp when needed.

The article identified the location of the surface mounted potentiometer that needed to be replaced with a higher resistance. Finally after hours of trying I was able to get it reduced to 47.5 volts - perfect for the NXP transistor.

The bottom view of the "slider" showing the three pins to be shorted, the 240VAC input connections on the right, and the output contacts on the left.

Picture of the "slider" with cover removed and activated after changing the potentiometer to lower the voltage.

The HP design seems to be excellent with filtering on the input side to stop any RF emissions out the back end - a problem with many switching power supplies. Since I did not want the 240VAC anywhere near the amplifier I wired it to the disconnect source through two "hockey puck" relays that are activated from the front panel of the amplifier. The only voltage entering the amp is the 47.5 DC generated from the switching supply! The hockey pucks are activated with 5VDC provided via a switch on the front of the amp.

I routed the 5VDC leads that activate the relays through a thermal switch mounted on the copper spreader that opens at 70C. If it heats up to this level the relays are deactivated and the amp shuts down. By experience heat is one of the major causes for device failure! (I made a short video for watching the action on YouTube - : https://www.youtube.com/watch?v=EnYWs_wteSk.)

I built three low pass filters (80M, 40M and 20M) for the board. There were spaces for seven filters but I only use an amplifier on the lower bands. The relays for these three are activated from the front panel where LEDs indicate the band in operation.

I hand wound all of the toroids and tested them to make sure they were the correct value. I used 3KV rated capacitors (Mouser) due to the significant currents in the filter circuits for high power amplifiers.

The Ukrainian's board - I only installed three low pass filters for 20/40/80 Meters

For the filters the values on the schematic shown below are on the money. I would suggest testing the wired board when done. I used an MFJ antenna tuner and a scope to measure the input and output - a signal generator and RF probe would work well too. Increasing the frequency to the attenuation point on each of the three filters showed their cutoff point, and measuring the input and output showed the insertion loss. It worked great - the three filters had very little attenuation at their operating frequency and each one cut of completely at about 2 MHz above the upper level of their respective bands.

Looking around the shack for a proper cabinet I came across a Heathkit SB104 I had purchased for parts. The case was perfect. To start with I needed to include a low voltage power supply for the relays, bias and other ancillary parts. I used an old filament transformer coupled to a full wave bridge rectifier to supply the main voltage. Its output was high so I installed three regulator ICs that yield 5VDC, 12VDC, and 14VDC - the latter for the bias. In order to monitor the copper spreader temperature I bought a digital probe with a led readout from China. I glued the probe into a small hole drilled into the copper spreader to make sure it reads accurately. I also found a dual readout meter for the output and reflected power readings, and meters for the amperage and voltage.

The inside view - top left the T/R relays and low voltage transformer, the low pass filter board in front left, and the current meter shunt to the right.

I bought a 12 VDC six pole high current relay which switches the antenna, the drive to the gates, and the bias voltage. When activated by the driving transceiver - my K3 - the antenna is switched to the amplifier output, the transceiver drive to the gates, and the bias to the circuit board. Two "L" brackets found at Home Depot were perfect for mounting the board assembly to the chassis independent of the main amp body. I had a 120VAC fan in my junk box and mounted it on the backside of the aluminum heat sink.

The fan, aluminum heat sink, copper spreader and circuit board as assembled.

With everything wired up and tested to the extent possible it was time to put it to the test. I hooked the antenna connection to a pair of 20M phased verticals that have a very low SWR (1.1:1), connected the "slider" cable that feeds the 47.5VDC to the amp, and hooked up the transceiver (a K3) to its connections. Before turning the amp on I tested the output of the K3 and SWR. It performed normally. I lowered the output to 5 watts and then took a deep breath and turned on the amp and the hockey puck relays to activate the "slider". The voltage came up and the amp was ready to go. I cleared the frequency and called CQ. I got an immediate response from a ham in Georgia with 5/9+ 10db. I watched the amperage and output meter while talking - the current was peaking at 20 amps and the output meter was bouncing up to 500-600 watts. The data sheet indicates an efficiency of about 60+% at less than full output. From the graph below the output would likely be peaking at about 570 watts at 20amps (47.5x20*x0.6). During the QSO I kept my eyes on the copper spreader temperature. It was climbing up to 40C during my longer transmissions and then dropping back slowly. Toward the end of the QSO I turned the amp off and cranked the K3 up to 100 watts and asked my contact if he noticed a difference. He said that the signal dropped 5-6db. I recently added a muffin fan on the inside, which blows directly on the circuit board and LDMOSFET.

The plot above shows gain and efficiency as a function of output power. The signal was a CW signal at 30MHz, 27.12MHz, 13.56MHz, 4MHz, 2MHz.

From the above graph it is clear that the NXP device will output over 1KW...a point on the chart where the gain and efficiency meet. To run it up to a kilowatt however the additional heat would have to be removed. I have thought about soldering copper tubing around the skirts of the copper spreader and hooking it up to a liquid cooling system. The "gamers" have their CPUs cooled with small radiators, cooling fans, and circulate the fluids through their CPUs. This might be the key to driving it to its maximum output. Meanwhile I will be content to loaf along at 600 watts on voice peaks - hopefully 24/7. I will probably raise the voltage to 50VDC, which would still be on the safe side of the drain/source rating.

Working with many ham friends to make this happen has been a real joy. In reflection on this everything I have learned over the years I have been a "ham" (as a non-engineer) has come from others - reading, talking on the air, emails, Skype et. Al. What a great hobby. I want to give special thanks to my friend, now a SK, Art McBride (KC6UQH). His vast knowledge of electronics was a huge benefit in helping me think through the various issues for this project.

Little Boy with a new paint job...ready to go! (See it in action on YouTube - : https://www.youtube.com/watch?v=Zp2PTM_gXgM

PS I use a check off card every time I start up my solid-state amplifiers. MOSFETs are very unforgiving if you put too much drive on them, they get too hot, or feed them into an open load. My list:

a.)Check the driving radio to make sure the output is low - often used "barefoot" at 100 watts it would take the gates out in a New York minute if left at this level.

b.)Make sure the low pass filter selection is for the band you plan to use.

c.)Check the SWR with the driving radio before turning the amp on...anything below 1:7/1 is probably okay.

d.)As you operate monitor the temperature of the copper spreader.

Following these simple rules will keep the amp running for a lifetime.

My hope is that many of you who want to "build your own" will collaborate with ham friends. There are infinite options "to do your own thing". Accept the fact that most projects don't work on "day one". My friend Art used to say, "if it works right away it is a bad deal because you have not learned very much!" Cognitively working through the issues will allow you to make the necessary changes to get it "right". As time goes on it becomes a "moving target for improvements" and it evolves to a better end.

(1)CD-101-13 BLF188XR 2-30 MHz 1000 W power amplifier Rev. 1.0----March 10 2013 Application lab report