Update (14 Feb 2018); assembled/tested RF decks are now available for shipment. The latest technical update is shown at the end of this page. If you purchased an assembled/tested RF deck, here are some observations and the instructions for making the connections.
I've been waiting for an LDMOS device to become available which might be easy to design for, could be driven with a 10w radio, and would deliver as much power as my LDF4 coax cable could handle. That seemed like quite an unrealistic dream, but in August my favorite supplier asked if I'd be interested in a new device from NXP, the PRF13750H, capable of more than 600w output at 1300Mhz. Naturally, I was delighted to hear the news, and once a sample device became available I began to experiment...and here are the results from the final prototype I built around this device:
The NXP data sheet lists it this way, and after building 20 RF decks to date, I'm finding less than 0.3 db variance; 70% of them will produce up to 700w, the other 30% a bit less, but none less than 645w.
On the ones I'm producing, I'm setting IDQ at 1.7 amps for linear operation; I'm finding the IDQ setting affects input match, but 1.6 to 1.8 amps seems to be about right.
I was told the PRF13750 is the "pre-production" device issued by NXP to it's distributors for initial sampling and sales; there were a limited amount of them made available, with full-scale production of the MRF13750 (note the prefix change) scheduled for late January or early February, though sometimes these dates tend to slip a bit. Pricing on the device is expected to be in the $330 to $360 range from most of the major distributors.
Here are two of the other particulars I measured on the prototype; this first one is gain, almost 20db:
This last one is input return loss, translating to roughly 1.25 to 1 SWR at center frequency:
Both gain and input return loss can be moved down in frequency by use of small pads located at the input and output of the PCB, which can be connected in or not, depending on which parts of the band you wish to optimize. For example, if your main interest is ATV, you may want your optimization in the lower part of the band instead of the upper as I have it here.
Here's a photo of the prototype; it's a bit on the rough-looking side because I made it here with humble tooling, as I do with all prototypes before investing in having a commercial board house make them in quantity. The board with the device installed will easily fit on a standard 3 x 5 copper spreader...the spreader in the photo is oversized, used only for experimenting. The amp itself needs to dissipate no more than the VHF KW amplifiers (400-500w), and the standard 3 x 5 spreader appears to be just right for the job.
Oh yes, this part can also produce 750w at 900MHz. Different PCB, etc. still
to be prototyped when time permits.
So the project begins...I have on order an initial 50 sets of the PC boards shown on the left (not to scale), and plan to have fully assembled and tested RF decks available using the PRF13750 in late January (I purchased a modest quantity of the devices when they were still available). These will be mounted on a 3 x 5 copper spreader, ready to bolt down to your heat sink. Expected pricing for the RF decks is $850
Shortly thereafter in April-May, complete custom turn-key amplifiers (w/o power supply), the same physical size as the VHF KW amplifiers, will be available. These will have the usual safety features (SWR and overtemp protection); expected pricing is in the $3100-$3300 range (depending on options ordered).
By the beginning of February the MRF13750 should be in full production, and
if there is sufficient interest I can have an additional run of boards made
ahead of time, included in kits available for those wanting to build their own.
The amplifier does use some rather expensive ATC capacitors, so the expected
pricing for the kit without LDMOS or copper spreader is $150. Fully machined
(for this rf deck) 3 x 5 copper spreaders will be available at the normal price
($100). Those wishing to machine their own can do so
using this drawing; not on the
drawing, but important, you will need to fly-cut the back surface of the copper
spreader for flatness so it mates well with the heat sink you will be using (the
surface of which should also be fly-cut for flatness).
A heat sink drilling template is
That's all I can think of for now; if you have interest in either the RF deck, the kit, or the custom turn-key amplifier, please drop me an email, as there will be a limited number of these available until I'm able to determine whether or not to continue investing in the project.
best regards, Jim
Technical update (14 February 2018):
The commercial PC boards arrived, and completed RF decks are now available on the parts page.
Board kits with all necessary components (except for LDMOS) will be available in about 8 weeks, as I must re-order boards without solder mask specifically for the kits. The reason for this is because the board maker applied solder mask where it wasn't supposed to go on the present batch, and this caused arcing at high output unless the offending area was relieved of solder mask around the output connection.
This is something I do here as I construct the RF decks, but I'm reluctant to put these boards in kits because of the difficulty in making the needed change; kits will need to wait until the re-order arrives in late April (it takes about 8 weeks). The MRF13750 is just beginning to show up at distributors, so timing will be about right for those waiting for kits.
For now, the RF decks are playing consistently and well, many with saturated output around 700w, and P1db at 600w. Accurate measurements of efficiency at saturation indicate 58%, a bit better than the data sheet suggests.
After constructing more than 20 RF decks so far, these are some of the observations I made:
All of the amplifier decks were able to
produce at least 645 watts (saturated) with about 70% of them able to
produce 700w. That's a variation of 0.3 to 0.4 db, and is normal considering
manufacturing differences from one device to another, even within the same
lot. All of the RF decks were able to achieve 600w with no more than 10w
drive. From the NXP data sheet, this is what you can expect from the best of
It should be noted this data was taken with IDQ set to 150ma, which is OK for CW, but too low for linear operation. With that in mind, I set IDQ to 1.5 to 1.8 amps, which produces good linearity.
The IDQ setting affected the input match; the higher the IDQ, the lower in frequency the input match moved. VDD had negligible effect on input match (from 35 to 51v).
Total VDD current drawn at full output varied, with the average around 25 amps (50v VDD). Most drew a bit less than that.
When mounting the deck, use all 8 of the supplied 4-40 machine screws and flat washers, or your metric equivalent (the washers will help distribute the mounting forces). Get them firm, but not so tight you damage the board material, which is soft and thin. Use heat sink compound between the copper spreader and the heat sink...not too thick, just a thin layer will do.
All assembled/tested RF decks have already been set up in the manner described below; no additional adjustments are necessary for operation at 1296 MHz.
The power supply used to set up the amplifier was current limited to 3 amps, and that's important when setting IDQ. If you are doing this yourself with any LDMOS device on any band, and it gets away from you, that current limiting will save the device. Fuses are not fast enough. The termination on the output is a 30w cellular type (no drive power for this test, of course).
While making the IDQ setting at 50V VDD, the input match is observed with a scalar analyzer and optimized at 1296 MHz (where most of us will be using this). The two parameters affecting the input match were IDQ and trimming pads. Usually, only one or two of the input trimming pads were needed with IDQ set between 1.6 and 1.8 amps. The inset below shows the measured optimization.
On the output side, most of the decks required
just one or two of the output trimming pads to be connected.
Both VDD traces must be connected together when routed to your 50v rail. Wire size to each VDD trace should be at least #16 AWG passed twice through a ferrite (Laird 28b0562-100 is recommended).
bias is configured for 12v (30ma current is needed for the bias regulator on board). The resistor you see from the VDD trace to the bias pad drops 50v down to 12v for testing. Normally, the 12v for bias would come from a control board or other source, but can be derived in this way from the 50v rail as long as the 50v is switched off when not transmitting.
The output coax should be able to handle the power; SM250-50 is recommended...this is the conformable 0.25 diameter cable you see in the photo, and is available from www.communication-concepts.com .
The input coax can be most anything with good return loss at 1296; rg316 is not very good for this, so the recommendation is RG402 (conformable), shown in the photo above. This is available on EBay and other places at reasonable cost. .086 or smaller conformable coax is also OK, but usually costs more than RG402.
are a couple of ways to connect to the input and output pads. The first is
by direct solder of the shield to the ground foil, which is difficult to do
because of the high thermal conductivity of the PC board material...the heat
is drawn away very quickly into the copper spreader, making soldering
difficult if not impossible. If you choose to do it this way, the method
shown in this photo works well. Prepare the coax so the center conductor
bends down, but does not quite touch the board when the coax outer is laid
onto the ground foil.
Temporarily remove the 4 screws at the end of the board, but leave the two next to the transistor in place. Slip one or two sheets of paper between the board and the spreader, and solder the shield to the foil, but do not solder down the center conductor at this time. That paper will help keep the heat from being drawn away too quickly into the copper. Remove the paper and replace the mounting screws. Now solder the center conductor in place. This same method can be used for the input coax connection.
This alternate method shows the coax pre-mounted to .020 tin mounting strips, and because I must install coax when I set up the decks, and then remove it, it is the method I use most of the time.
Cut your tin strips to .375 by 1.375 (3/8 by 1 3/8). Mark and drill the hole locations for pass-through (for the screws) by comparing the strip to the board hole pattern.
Prepare your coax and solder the shield to the tin strip, aligned as shown in the photo. Now you install your coax using the two mounting screws at either the input or output, whichever you're working on. Once the assembly is fastened in place, you can solder down the center conductor.