80 Watt power amplifier

Senin, 02 Agustus 2010
This is a pretty conventional design, using bipolar transistors in a tuned class C circuit. Thanks to the use of two stages, the amplifier can be driven to full power with less than 1 watt driving power, so that a large gain margin results in this transmitter.

Bipolar VHF power transistors have a severe affinity for low frequency self-oscillation. To obtain stability in this amplifier, I employed several techniques, such as placing the resonances of base and collector chokes far apart, damping the chokes with resistors, using RC combinations for absorption of unwanted frequencies, using feedtrough capacitors for bypassing on the board, etc. It took some tweaking, but the amplifier ended up unconditionally stable.

The impedance matching network between the two transistors calls for such a low inductance, that it would be impractical to make it with actual wire. So I used a micro stripline etched on the PCB. Also, the power and SWR sensor at the output was made with micro striplines.

Click on the schematic to get a full resolution version which also includes detail about the micro striplines and other parts.

This amplifier has a low pass filter at the output, resulting in a signal clean enough to be directly connected to an antenna. The SWR meter was placed before the filter, in order to clean out the harmonics produced by its diodes. In any case, while the signal is clean enough to easily satisfy usual legal and technical requirements, this transmitter should not be used at a multi-transmitter site without further narrowband filtering! This is so because any other strong signals on nearby frequencies would be picked up by the antenna and coupled to the power transistor, which would mix it up with the own signal, creating a wide array of intermodulation products, some of which would be re-radiated! This is a common and very big problem in many multitransmitter sites. In such places, NOT EVEN ONE transmitter should be allowed on the air without narrowband filtering! Such filtering is easily accomplished by means of a single tuned cavity, which can be constructed from copper tubing or sheet.

Here is the PCB layout, including the microstrips. The board is 20cm long and is double-sided, with the backside being a continuous groundplane except for two small pads at the driver transistor base and collector. I cut out these pads with a knife, rather than making a whole computer drawing for that!

You will have to drill and cut out the openings for the transistors. The power transistor is mounted from above, while the driver transistor, due to its small height, is mounted under the board. Both transistors are mounted after soldering copper foils into the PCB openings, to join the upper and lower groundplanes, and the driver transistor also has such copper straps connecting the base and collector pads to the upper side of the board. Here you can see how the transistors are soldered to the board, and the spacers I used to give it the correct height. I first mounted the board and transistors to the heatsink, then soldered the output transistor in lace, then tack soldered the drive transistor's emitter leads from above, through the opening, then again removed the board and soldered the driver transistor fully. In this way the proper mechanical fit is assured. MAke sure the transistor mounting surfaces are flat! My power transistor came with a slightly rounded surface, so I first had to sand it flat! This is critical for good heat transfer. Of course, use good thermal grease when finally mounting the amplifier to the heatsink. You can see that that there are also a few more places where things connect through the board for best grounding. Of course, the shield around the board also joins the two ground planes.
And here is the parts overlay, as usual without parts identification!

This is how the complete power amplifier looks from above. You can see the striplines, how the feedtrough caps (used as collector decoupling caps) are installed, etc. Note the copper clad mica capacitors in the low pass filter at the upper right. But let's better look in detail at some interesting areas:
Here you can see both transistors and the matching network between them. I couldn't find trimmers that would stand the amount of RF current present in this circuit! Every factory made trimmer I found would melt down! So I made my own mica compression trimmers, using brass and copper sheet, brass base plate, brass compression washer, and mica sheets originally intended for TO-247 capsule mounting. All connections in the trimmers are soldered, not just riveted like in many factory made trimmers. That solved the problem, but even these trimmers get warm in use! Note how the trimmers at both the input and the output of the power transistor have their ground connections very close to the emitter leads.

The output matching network uses the same kind of trimmers. The one that appears in the low middle of the picture is the one that takes the most current, more than 15 amperes of RF! In continuous service, and at VHF where the skin depth is very small, this is a large current. The same goes for the tank "coil", which is made from a strip of 0.5mm copper sheet bent in "U" shape. Despite its good thermal connection to the board, it gets hot enough to become impossible to touch! Of course, anyway you shouldn't touch it while the transmitter is on, because in addition to a heat burn you would get an even nastier RF burn!
A similar problem happened with the capacitors for the output low pass filter. I tried to use RF-rated dipped silver mica capacitors, as shown in the photo above in its upper right corner, but they got so hot that they started smelling! Surely their silver electrodes are too thin. They wouldn't have lasted long in this service. I didn't have any better RF capacitors on hand, and instead of ordering heavy duty metal clad mica capacitors at several dollars each, I decided to make my own. Here is one example, shown alongside a TO-92 transistor for size comparison. I used 0.5mm copper sheet for the external electrode, 0.1mm copper foil for the inside one, and mica cut from TO-247 insulators.
Here is close-up edge-on look at one of my copper clad mica capacitors, held in the jaws of a wooden clothes clip for the photo!
Since the thickness of those mica insulators for semiconductor mounting varies a great deal, making these capacitors is a cut and try process. I measured the mica's thickness as best I could, calculated the surface necessary for the capacitors, built them, and then measured them, using a test coil and a grid dip meter. I wrote the value on each, and continued making capacitors until I had some of values close enough for my low pass filter. The rest I kept in stock for other projects!

It's fun to notice that copper clad mica capacitors built in this way perform just as good as factory made ones, that you can make any value you need, and that they cost about 1% as much as the nice shiny branded ones!

In the low pass filter, these copper clad mica capacitors get barely warm. Since they are well soldered flat to the board, I don't know if they conduct their loss heat into the board, or if they are only warmed up by the filter coils! Because these coils surely do get warm in use, despite being wound from very thick wire.

For the tests I mounted the amplifier board on a rather large heat sink. It consists of a 10 * 20 cm copper plate of 6mm thickness, to which I soldered 20 fins, made of 0.5mm copper sheet, measuring also 10 * 20cm each, having L-shaped soldering edges. I made this heat sink some months before for investigation purposes (see my thermal design page), and since it was lying around, I used it. But with the total power dissipation of this amplifier being something like 50 watts, a much smaller heat sink would be good enough, if a small fan is used. Still, a copper heat spreader is a good idea, because the power transistor is used at its maximum rating.

This photo shows the transmitter being tested on my admittedly not very tidy workbench! You can see the exciter in the lower left, and the amplifier with its overly big heatsink standing on aluminium comb supports to avoid bending the thin fins. There is my Aiwa power and SWR meter, and a large oil-can dummy load to safely swallow the 80 watts (actually that dummy load can take a kilowatt for a few minutes). An analog multimeter is showing the current, and the rest are boxes of parts, tools, etc. The audio board ended up outside the photo, along with the digital multimeter, frequency counter, oscilloscope, etc. It was quite a mess, but worked very well!

I ran several tests on the transmitter. One endurance test consisted in running at 80 watts output for one week nonstop. No problems were noticed. Other tests included temperature shifting, vibration (to check for microphonics), varying the supply voltages, etc. The transmitter seems to be very well behaved in every regard.

Then the qualitative tests were done. The stereo separation, measured through my homemade FM receiver, came out as 52db. That's better than most. The signal/noise ratio was beyond my measuring capabilities, which top out at 82dB! That's better than almost anything one can hear from commercial stations! The distortion was also too low to be measured, a result of the careful balancing of the residual varactor nonlinearity with the effect of series capacitance.

Then the ear test came! I hooked up my CD player, the transmitter, FM receiver, amplifier and speakers, so that I could switch the sound forth and back between the original signal from the CD, and the signal going through the transmitter, a few meters of air (the radiation from the low pass filter coils is much more than enough for this distance), and the receiver. I played a CD by Roby Lakatos, the King of Gypsy fiddlers, which I like a lot and which is great for testing because of its crisp, clean and full sound. I was quite impressed by the fact that I could switch forth and back between the original and the transmitted signal, without detecting a difference by ear! So I'm happy to tell that this transmitter preserves the full audible quality of a first-rate CD signal! The less than perfect stereo separation is no issue at all, because no listener, even in critical mode, can discern between 50dB separation, and perfect separation!