Comment? email to:
Many of us have been waiting for a practical 70cm solid-state amplifier design. Freescale Semiconductor's recent offerings of 1.25kw and 600w transistors have been getting a lot of attention, and this resulted in the development of kw amplifiers for 6m, 2m (F1JRD) and 222MHz (W6PQL).
In the 1/2012 issue of Dubus magazine, F5FLN showcased his KW amplifier for the 70cm band. This amplifier combines a pair of 500w RF decks, using input and output couplers; the device used is the 600w Freescale MRFE6VP5600H. RFHAM is currently offering board kits for these 500w amplifier units, which contain the more difficult-to-get parts unique to the design.
I bought one of these kits, and after making and testing an RF deck using these parts, there were a few changes I had to make to get the amplifier to play as advertised; still, the original kit is worth buying, and the changes I made to it were not complicated.
The RF deck shown on the right is my own version of the original kit, with a few enhancements (more optimum matching, temperature-compensated and balanced bias circuit, simplified layout and corrected output balun connection). Blank boards, basic kits and assembled/tested RF decks for this version are available on the parts page.
If you have one of the RFHAM kits, you may want to make the same changes I did...here's what I did and why:
The cut-out for the transistor is a bit cumbersome to work with, so I trimmed the board into two halves.
4 additional mounting holes were drilled near the transistor
body; you can see how I did this by comparing the two photos above. The
two original mounting holes at bottom center, and the one at top center serve no
useful purpose if the 4 new ones are used, so they were not used. The 4 new ones
are used to hold the board to the copper spreader where it matters (close to the
transistor body); two larger holes, suitable for #8 screws, are drilled through the spreader in the
transistor slot. These provide the main torque for binding the spreader to a
heat sink (not shown); the four screws at the corners are only
tightened enough to hold board and
spreader to the heat sink.
After quite a bit of fussing with multiple capacitor positions and values, it turned out that none of the original matching capacitors (C18, 23, 24) produced the best results. The value of C18, it's type, and it's position as the lone discrete matching component did the trick. As can be seen in this enlarged inset, a 20pf type MIN02 metal mica, placed across the end of the bridged gaps in the output traces, did the trick. The giga-trim variable capacitor, the 18pf and 10pf matching capacitors were not used. Output was now 550w.
This may not work for every builder, and it may be possible
the changes I made to the board were the reason I had to change the output
Here's another change I made, which brought the output up to 580w. The schematic in the Dubus article showed the bias feed connected in such a way that only half the transistor (it's a push-pull unit) was biased on; the opposite side was almost on, but was essentially still class C, not drawing any idling current. This unbalances the circuit, and may even account for some of the lower IMD numbers in the KW version of the tests (23db) as described in the magazine article; this is only speculation on my part, I did not make IMD measurements myself.
The fix was an easy one; just jumper the two gate decoupling pads
together as shown here (orange wire from one pad to the other). This connects
C1, 12,13 and 14 together, and biases the two sides equally. IDQ was readjusted
back to 1 amp, and this bias equalization provided a better balance between the
two halves of the device, resulting in higher output power.
What we are doing here is connecting the grounded shield of the rg402 output balun to the top transistor output. On the input side, the grounded shield should be connected to the bottom transistor (it is connected correctly as shown above).
The Dubus article had this output connection reversed, and I missed it completely; however, Fred (WA7TZY), who worked on this type of amplifier in the commercial world, caught it right away as he was assembling his kit.
Initially, I had some odd stability issues with the amp; on some antennas, it was fine, but connected to others it would show some instability; I just attributed it to the very high low frequency gain of the device. To solve this, I added a small inductor across the gates of the device to kill the low frequency gain, and this is still a good idea. However, the root cause of this "sometimes unstable" problem was the output balun connection error.
If you already assembled your kit, it may be easier to reverse
the connections on the input balun, it's more flexible. It doesn't matter if you
make the change to the input or to the output balun, so long as one of them is
changed so the grounded shields of the baluns connect to different transistor
halves from the input side to the output side.
Here are the test results; measurements are + or - a bit, as the wattmeter was not being read in the most accurate part of it's scale.
If you are building this amplifier from a kit I supplied, here is the link to the assembly instructions.