XRF-286 Amplifiers for 23cm

The XRF-286 is a wonderful part for use on 1296, and one of the few choices available for developing significant solid-state power on that band. It is an unmatched 26v LDMOS transistor that was once made by Motorola, and can still be found in surplus PCS amplifiers made by Spectrian. A copy of the data sheet can be found on the internet in several places.

The Spectrian amplifier output boards contain three of these devices, one 286 driving the other two at 2.4 GHz. There are three of these output boards in those amplifiers, all combined for a total of 40w out. I suppose you could say they were barely used, considering a single 286 is rated at 60w out at 2 GHz.

Unmatched parts like these are the most useful, mainly because the matched ones are set up for different frequency ranges outside of the amateur bands. You can’t really get inside the device package to change the matching network, and trying to re-compensate externally usually results in high losses and inefficiency if it works at all.

On 23cm, an amplifier using just one ‘286’ can be driven into saturation with as little as 3 or 4 watts, and a pair with less than 10, making that last configuration a close match for many of the 10w radios in use today. With 8 to 10 watts drive, a pair of them will easily deliver 150 watts.

On the web, one can also find amplifier designs that appear to have been originally developed by F1ANH, F6DRO and others, implemented on FR4. The basic design is very good, though I did make some changes to the bias circuit and the input and output matching transformers. Our French brothers deserve a lot of credit, though, for pioneering the original design.

Although FR4 is inexpensive and easy to work with, it can be problematic; losses create a practical limit to its power-handling capability on 23cm, and it tends to be variable in specs from one lot to the next; thus the decision to port it over to a more reliable material. I tried two low-loss substrates: Rogers 3006 for the prototypes, and Rogers 4003c for the later versions. Both of those materials produced consistent results.

When prototyping this design, I discovered that the input matching line was a bit short for 23cm, probably due to the variability of the FR4 used in the original design. The line needs to be a bit longer, but I left it alone in favor of the small high-Q ceramic trimmer capacitor you can see shunting the input stub. For the microwave purist, I realize this is heresy; however, I like to be pragmatic, and left it that way, appreciating the ability to optimize input match without having to strap in snowflake tuning stubs to do the same. This input stub is 4.32 ohms, and 52 degrees long as shown.
I handled the output stub differently, however, due to the higher RF currents there. It also needed an adjustment to work properly. The first section is 4.07 ohms and 58 degrees long; the second section was the one requiring the change; the original design had this at 22.7 ohms and 52.4 degrees long, but I found that it needed to be lower and longer. 14.2 ohms and 58.4 degrees produced the best results.

This is one of the 2-device amps, also on 4003. It's basically two of the single-device designs, combined using a pair of branch-line hybrid couplers, all etched onto the same substrate.

The hybrid couplers offer isolation and protection for the individual amplifier units; if one should fail, the other is protected by the 150w termination at the dump port of the output coupler.

If you look at the input matching stub on each device, you'll see those rectangular "snowflake" trimmer stubs, there to fine-tune the input match. Again, I chose instead to use a small high-Q trimmer capacitor instead of the stubs. It's up to the individual builder, but I find it easier to adjust the trimmer than to fuss with the stubs; one can do it either way.

The board is mounted onto the heat spreader shelves using the 4-40 screws shown, and spaced above the heat sink using 1/4" aluminum bar stock spacers. The 100w termination also requires a 3/16" spacer for proper positioning.

Here is the typical performance of this amplifier.
Now let’s look at how to mount these devices. If you looked at the data sheet, you probably noticed that the ‘286’ comes in two form factors; one is the type with a mechanical mounting tab, and the other is a direct-solder type designed to be soldered directly to a heat spreader. I haven’t seen the tab-mount type, just the solder-in variety. If you are intending to harvest these from Spectrian amplifier boards, you will be getting the solder-in package.

Removing them from the Spectrian board requires the use of a hot plate and hot air gun; the board is heated from the bottom; then the air gun is used to re-flow the solder around the part. At that point, the part can be picked up with tweezers and deposited elsewhere to cool gradually.

The process of soldering the device to the new spreader requires even more courage, and is best accomplished in an oxygen-shielded environment with special equipment, though I hear from some of the boys in the EU that they have been successful without that precaution. A friend of mine in the semiconductor business handled that latter part for me; after that was done, I was able to treat them like a more conventional transistor, and just bolt them down to the heat sink.

The photo at the right shows one of these parts in the form I like to use them; secured to a thick copper heat spreader. The spreader is 1.75” x 1.75”, and ¼” thick. There is a slot milled into the spreader to space the part at the correct height to accept the PC board, which slides under the lead (s) and onto the shelf of the spreader as shown in the previous photos.

These thick spreaders do a great job of drawing heat away from the devices and passing it into the heat sink below. Even at full sustained output, I can detect no drop in power level, which is a common problem when the heat spreaders are made from thinner material.

Of course, amplifier boards can be combined to make some EME-capable amplifiers. I built two of them so far (600w out); the first requires about 40w for full output. The second has a driver stage and needs only 2w drive. Pictures of other amplifiers built with the XRF kits can be seen courtesy of KD5FZX, VE4MA, and RA3AUB.
For those of you capable of making your own boards, here is the latest artwork for the 75w board (positive) (negative), and and for the 150w board (positive) (negative). The larger holes in the boards are .125", and the smaller ones are intended for tin rivets, the hobbyist equivalent for vias. Board dimensions are given on the positive artwork copies. The boards must be cut to remove the center piece separating the gate and drain halves. A paper cutter works best, but a steady hand with large scissors will also do the job.

Corrective adjustments will be necessary if poor construction techniques are used. Leaving too much room between the board and the transistor body (it has to be right up against the ceramic), substituting components (I've seen plastic capacitors used for the input trimmers, for example), leaving wires from coax jumpers too long, and not flow-soldering the devices to the spreaders will always introduce variables. Some of these variables can be overcome with tuning tabs at various spots on the RF traces, but all of them are just fixes for construction errors, so follow the recipe as closely as possible and you'll have no troubles. I built several hundred of these, and found them to perform very consistently; a total of two of them wouldn't make power, and in both cases it was sloppy construction on my part causing the problem (gaps between the boards and transistor bodies).