HomeHigh Level DesignControlRF AmplificationCoolingOutput FiltersI/O BoardMechanical DesignAntenna TunerPhoto AlbumContact The Team


1500 watts output
Class AB
Low distortion and high linearity

Design Considerations:

VCC requirement and implications on power supply design / selection
Cooling - ability to spread heat quickly and over a large area
Proven design with low risk
Proven ability to survive poor conditions (high SWR, e.g.)

When we set out to design this amplifier, we assumed (naively, as it turned out), that the hardest part would be the design of the RF amplification system.  In the end, we used an off-the-shelf design.

When considering the selection of the transistor to use and the board design to use it in, we have several important considerations:

First, cooling.  The available transistors range in power output from perhaps 50 watts at the low end to 1500 watts (The ARF-1500 e.g.) at the high end.  A pair of 1500 watt parts seems like an obvious initial choice, until you try to design a cooling system that will be robust and deal with full output power at a 100% duty cycle for an extended period of time.  A 1500 watt point source (or even a 750 watt point source) is extremely hard to cool.  To design and build a hobby price point cooling system that can handle this kind of power in such a small space was thought to be beyond the capabilities of this design team (and perhaps impossible).  We decided instead to look for parts in the 100 - 300 watt range, such that each individual RF amplification module would put out perhaps 300 - 750 watts of heat.  This would allow us to use "obvious" and inexpensive heat spreading techniques to extract the heat from the transistors and dump it into either water or air.

Second, the mechanics of splitting and combining the RF going into and out of the amplifier modules.  While two and four way splitter/combiners are reasonably easy to design and build (or even buy, as we choose to), three and five way splitter/combiners are not.  That means that our design needed to be based on amplifier modules putting out either 375 watts or 750 watts per module, which in turn argued for push/pull pairs of parts in the 190 or 375 watt ranges.

Third, IMD and linearity.  Because the market for linear HF power transistors in limited, most "modern" designs are targetted at the medical imaging, wireless microwave communications, or radar markets.  These parts generally are not designed for class AB service in our frequency range.  Many modern plastic packaged parts are otherwise very attractive, and it may in time turn out that one of them is suitable for our application, but at least initially we wanted a part that was known to perform well in class AB communications service at HF.

Finally, power supply considerations.  It rapidly became clear that if we wanted to hit our design price point of $2000 - $2500 complete, we weren't going to be building a power supply from new parts.  The supply alone would cost at least $500, and we didn't have the budget for it.  Luckily, 48 volt surplus telephone central office equipment supplies are readily available and often quite cheap - this argued in favor of using transistors in the 50 volt class.

Putting those four design criteria together quickly led to a solution.  The obvious choice for a proven class AB power transistor in our desired output and frequency range is the Motorola MRF-150.  A 50 volt part, the MRF-150 has poor efficiency (45% worst case) and high cost (around $50 in matched quads) - but is one of the only high power transistors designed for and routinely used in class AB SSB amplifier designs.  The MRF-150 is available dual (MRF-154/7) die packages - but these are both more expensive and harder to cool than their single die brethern.

This device has been around for 30+ years, and has been used in dozens of designs for all kinds of communications amplifiers.  In addition, a basic board design was created by Helge Granberg, K7ES, when he was Motorola's applications engineer for RF amplifiers.  This board design, colloquially known by its application note designation, is called the EB-104 (Engineering Bulletin #104).  The EB-104 board, and the parts required to build it, are available in pseudo-kit from from Communications Concepts.

When used in this board design, four MRF-150's (two push/pull pairs) are capable of 600 watts output, enough in theory for us to get by with just three boards.  However, given both the lack of three way splitter/combiners and the superior linearity and reliability offered by these parts when operating at less than their rated power output - we decided to use four EB-104 boards each putting out 375 watts (for a total of 16 MRF-150 transistors).  This solution isn't low cost, but should meet all of our other design criteria nicely.

In fact, as has been proven by our prototype in its qualification testing, four EB-104 boards is an excellent solution to out RF amplication needs.

Notes on project Gamma - The RF Deck

Introduction and commentary

One would think, if one didn't know anything about solid state amplifiers (and I don't, or at least didn't), that this would be the hard part. In a tube amp, the RF deck is the easy part. There are only a few components (tube, socket, input choke), they are fairly easy to understand, and quite robust (both mechanically and electrically). But with a solid state amplifier, this just isn't true. Transistors are very finicky. Designing a new solid state RF deck is hard. Luckily, it turns out that you don't have to design a new RF deck! A very capable RF engineer at Motorola named Helge Granberg did the heavy lifting about 20 years ago, and wrote it all down for us in a series of application notes and engineering bulletins. Even if you don't want to use Motorola parts (or, to be more precise MA-COM parts - since they now have the rights to the old Motorola transistors), you can still use the bias circuitry, board layout and input/output circuitry from those old Motorola documents.

So the question on the RF deck boils down to a fairly easy set of questions:

  1. Do you want to use Motorola parts or not?
  2. If no, which parts are you going to use, and how are you going to get the boards designed?
  3. If yes, which part ? The obvious choices are the MRF-150 or the MRF-154.

My sense after some discussion and thought is that the answer to the first question is "yes". Even though the MRF-150/151/154/157 design is an old one, it still offers the best performance in several dimensions. First, and perhaps most importantly, these parts have the best IMD characteristics and linearity of any commonly available HF amplifier. Second, the board designs are done, and tested, and have been proven in battle. Third, 48 volt DC power supplies are available used a good prices, which isn's true for the higher voltages required by some alternative parts.

On the third question then, the real issue is this: do you use MRF-150's (150 watts each, requiring either 12 or 16 transistors for our project), or MRF-154/157's (600 watts each, requiring 4 for our project)? Despite the obvious benefits of using the MRF-154's (only two boards and many fewer parts required, and simpler splitter/combiner), the collective wisdom of the solid state crowd seems to be in favor of the smaller parts. The issue that everyone talks about is heat. Apparently, the MRF-154 has major heat management issues. You are making a lot more heat in a small space - getting that heat into a heat sink under contest conditions is apparently quite difficult. I think that if you are willing to go with a water cooled design, using the MRF-154 would be ok - but if you want an air cooled design, sticking with the MRF-150's makes more sense.

There is also a secondary issue of cost. Because the MRF-154's are much less frequently used, they are much more expensive. MRF-150's can be had today for $188 for a matched quad (see below). Advanced Power Technology is about to releast their plug compatible VRF-150 (parts should be available in Q1, 2005), and they are quoting $38 each in quantity 5000.

MRF-154's on the other hand, are hard to find and priced at over $850 per pair. So even if our design uses 16 MRF-150's (we may get away with 12, but that is a subject for later), and even if we pay the current price of $188 per quad, our power transistors will cost $752. Four MRF-154's will cost us the princely sum of $1,700 - and if we blow one up we will be sad indeed.

Notes and commentary

Communications Concepts makes and sells boards and parts kits based on various Motorola application notes. It appears that either the board based on the EB104 (600 watts, 4 x MRF150 ) or the board based on AR347 (1000 watts, 2 x MRF154) would be the best bet.

A potentially interesting source for MRF-150's is:

-----Original Message-----
From: David Gilden [mailto:David.Gilden@rfmw.com]
Sent: Tuesday, January 04, 2005 8:31 AM

We can supply the MRF150 in any fashion you wish.

Singles units are $41.00 each

Matched Pairs are $88.00 each ( 2 transistors)

Matched quads are $188.00 each (4 transistors)

David Gilden

RFMW, Ltd/RF Power Exchange
I had a nice chat with Joel Levine at RFMW. He urges against the MRF-154/7 on heat management grounds. He also suggested parts from IxysRF , which I am looking in to right now. Other potentially interesting transistors include:

- The 2SC2879 , which is used by lots of folks for 100 watt PA's (e.g. in the Elecraft K2/100). This part operates at 12 volts, and is quite a bit worse on linearity/IMD than the Motorola parts are. Also, the idea of 12 volts at 4000 watts is a bit scary.

- The Advanced Power Technology parts. These are plastic packaged (and thus lower cost) RF transistors. Most interesting I think is the ARF463A/B, which are 125 VDC parts rated at 100 watts output.

Notes: Sand Labs makes RF boards (and possible power supplies). http://www.sandlabs.com/pa.htmlunder products/RF

Enter supporting content here