Steve’s Article about his linear amplifier
Remember, you must be the holder of an Amateur Radio license to operate this equipment on the air!
If you have Ivex Software, the schematics can be downloaded here. I recommend you right-click on this link, and select “Save Target As…”
After much contemplation, and longing to QRO, I determined that I wanted an amplifier. The conclusion to build or buy was decided when I found that there was not much offered meeting the criteria that I set for myself. These limitations were that it should run on standard house current and not require 240 VAC. This limit was set by the problem that I live in a very old neighborhood where many of the houses only have 120 volts AC available. The power limitation was set at a maximum of only 15 amps available to the radio shack. Approximately 500 to 600 Watts PEP output maximum was chosen. Whether to use tubes or solid state was then considered. I decided on using transistors, as that would eliminate having to deal with the high voltages needed to support the anodes. I also didn’t want filaments heating up the room that I use for the shack plus the power consumption required for making tubes glow and is a waste of energy. Transistor amplifiers are broad-banded, eliminating the need to tune the amplifier each time I QSY’ed. Another goal was to have this amplifier automatically change bands, along with the Yeasu FT-757 HF rig I now use. Picture of EB104 board
Then the choice for which part to use had to be made. After searching the Internet, I found that most of my work was done! Communications Concepts[i] has several kits. One of them caught my eye. It is a kit based on the Motorola MRF150 part, and copies the amplifier in EB-104[ii], a Motorola Engineering Bulletin. This design uses four devices in push-pull/parallel. It has individual bias adjustments, and an on-board regulator for the bias.
Some of the other features that I desired are to have complete metering capabilities. That must include, drive power, power out, reflected power, drain current and drain voltage.
This amplifier is to use switching provided by the transceiver. Although there are many amplifiers on the market that can be RF sense keyed, I prefer to have the rig control this action either through push to talk (PTT) or VOX. For my personal choice, it would seem that there would always be some clipping of the first syllable or Morse character if the amplifier had to wait for exciter power to cause the amplifier to initiate. Another reason is the extra stress on the relays trying to make contact while live with 160 or more volts RMS at nearly 4 amperes of RF current. Also, during this time period there would, no doubt, be short periods of time (best case of 15 milli-seconds) where there would be no load on the transceiver, and the power amplifier module, which could destroy an expensive set of transistors in either the amp itself, or the exciter.
Another feature I want is that this accessory be as small and compact as possible. The house where my family lives is small, and to consume a large portion of real estate with a physically large amp doesn’t seem prudent.
Something else I wanted was to be self-contained. It must have the power supply, amplifier and the filters built into the same cabinet.
I began by finding a cabinet. I found one that looks very similar to the LMB CO - 1 series. (See picture of cabinet) It measures almost 10” wide, 12” deep by 6 inches high, not counting the extra 1” for the feet. This is a used cabinet, but LMB does make one that is very similar. The only difference I could see was that this one has a top that opens with a hinged lid. The LMB series allows the whole top to be removed. Chassis Layout
I then went shopping for the power transformer. This component needs to supply 50 volts at peak currents of up to 20 amps peak. I found one that suits this need on an Internet auction. It was designed for a 500-watt stereo amplifier, and has 70 volts RMS center tapped (35 volts RMS per side) at 14 amps continuous. This is close to 900 watts input available. (35 * 1.414 = 49.5 Volts DC, 14 Amps continuous * 1.5 for ICAS service = 21 Amps Peak, 49V * 20A = 1030 Watts Input, at 45% efficiency (Worst case) gives 465 watts output minimum. At 55% efficiency this gives about 560 Watts.)
Then I also found some 33,000 uF capacitors, and a couple of 50 amp diodes. So far, so good. I saved a couple of hundred bucks! Caps, $1 each, new would have cost $20 apiece. Power transformer, $10. New? Over $30. Diodes $5, new are about $15 each. Cabinet, $0, new $50 to $60.
I next constructed the power supply, fitting the transformer into the far left rear of the cabinet. The capacitors squeeze in between the transformer and the front panel. I mounted the power diodes on the left of the cabinet. (See Cabinet picture)
A word about safety! Even though there is only 50 volts on the capacitors, this is enough to give a very bad shock. Also, keep in mind that you are looking at over 1200 watts of stored energy on the caps, and rings can melt at the currents this supply can produce. Surge amps can approach 200 to 300 amperes of current! Welding rods melt with as little as 30 or so amps. I recommend that while working on this power supply you remove all jewelry, including rings and watches.
I built the power supply (See Schematic) on the left side of the cabinet, keeping as much of the components crunched as far to the side as possible. I saved some room by not using the customary wrap around capacitor mounts with feet, but rather made home-made ‘L’ brackets to hold them against the cabinet side. This method saves almost ¾ of an inch in both horizontal directions. As can be seen in the photo, this is a very tight fit, and both of the 3-inch diameter 33,000 uF capacitors take up even more room than the transformer. The 50 amp diodes are mounted just above the transformer. All wiring is done with 12-gage wire. I recommend stranded wire here. All of the wire ends are terminated in ring lugs, soldered on where the wires connect to screw terminals, such as the power diodes and capacitor bank.
In full power operation, the transistors will be operating with close to 1200 watts PEP input power, and with 500 to 600 watts out will have to dissipate at least another 500 watts as heat. This must be removed from the active devices. I had two choices. I could mount the heat sink on the outside of the case, and allow convection cooling, or I could mount the heat sink inside the cabinet. In this picture, the heat sink and fan are mounted together in the long metallic object just to the right of center in this picture. The fan is located on the right side between the cabinet wall and the heat sink box. In this picture, the copper heat spreader is not yet installed. and install a blower or fan. My choice is to mount the heat sink and the RF amplifier module vertically and attach a small muffin fan to blow cooling air through the cooler. Although this adds extra sheet-metal and machine work, it keeps the hot surface away from accidental damage or obstruction and keeps fingers away from potentially hot surfaces where one could get a burn. Remember, a hotplate consumes nearly the same amount of energy. (See Picture 2)
I found mine on an online auction, and cost me less than $20 bucks. (New ones run around $50 to $60) The one I bought was 10 inches wide, 1-3/4 inches high and 13 inches long. I cut mine to fit so that it would stand on its side and length-wise so that each end can be bolted to the front and rear panels using 4-40 screws. I also attached the device to the floor of the cabinet with the same size screws. This was done to keep the cabinet rigid, and help keep vibration noise down. My heat-sink was used, and did have a few holes drilled and tapped in it. I decided that it would be prudent to re-use these holes to minimize drilling extra holes, so a template was made by taping a sheet of paper to the device side surface and tracing the location of these holes. This will be used to punch the holes for the copper heat spreader described below.
Once the heat sink was cut and trimmed and tested for fit, I mounted a slab of copper .325 inches thick[iii], using screws and put heat-sink compound between the two parts. Aluminum is actually a poor conductor of heat compared to copper, so the copper sheet acts as a heat spreader increasing the effectiveness of the heat sink. I estimate that the heat sink alone would only be able to dissipate about 150 to 200 watts, while with the copper and forced air cooling, it should easily handle the required thermal requirements with some safety. My typical operating style is only about 30%.
A hole was cut at one end of this assembly to mount the fan. The idea is that the fan will blow through this hole and the sheet-metal cover will duct the air around the fins, circulate past the band-pass filter and then past the power transformer, carrying the heat out the rear of the cabinet. I suggest that the fan be mounted in the front end of the heat sink to minimize re-circulation of the hot air, though this brings the fan closer to the operator. The fan noise isn’t so high in level as to be objectionable and is somewhat attenuated by the cabinet. I chose not to temperature control the fan since the noise level is low to begin with. If you find that the noise level is objectionable, you could add a thermostat to the fan, and only turn it on when the heat rises above a preset level. I wanted to minimize the amount of construction, and decided that the fan running anytime the amplifier had power on was something I could live with.
Low-pass filter assembly
Solid-state amplifiers have several problems that have to be handled correctly if the amplifier is to be legally used on the air. First, it must be stable and not become a power oscillator. Transistor amps also are inherently rich in harmonics. This is because of the non-linear junction that all transistors have. Diplexer Schematic
Part of the problem is solved with the design used here. Push-pull amplifiers are not prone to developing even harmonics if the devices are reasonably matched for gain. In this amp, it was found that second and fourth order harmonics are down in excess of 40 dB from the output, so anything added by the filter is gratis. However, the third harmonics are only down a measly 10 dB to 13 from the main signal. With potentially 600 watts PEP, this implies that on the order of 50 to 60 watts of harmonic energy is available to radiate. Obviously this is a problem that has to be taken care of at the amplifier. The requirements are that for 500 watts and less, the harmonic energy must be below –40dB, and for 600 watts the requirement becomes slightly more stringent. I chose the goal of –46 dB as the maximum harmonic content allowable post low-pass filter with a 50 ohm load.
The filter has to present a load impedance that the amp was designed to operate into. In our case, this is 50 ohms. Since the filter has to be designed to resist passing the harmonics that is not going to happen for those byproducts. Harmonics in a typical low pass filter circuit that are higher in frequency than the cutoff of the filter are either absorbed by the filter components, or are reflected back into the amplifier where it adds to the heat dissipation problem of the transistors and also contributes to other byproducts of the amplifier such as inter-modulation distortion.
Since my goal was to build a linear and not spend hour upon hour of design work, I searched previous articles and publications and found an excellent design by William Sabin, W0IYH[iv] where he passes the lower desired signal to the antenna port using a low-pass filter, and sends the higher frequency by-products via a high-pass filter into a termination. (See his article) This arrangement of low-pass/high pass filters is called a diplexer. This keeps the load to the amplifier very close to ideal at all times, adding to the stability of the amp. It also adds additional attenuation to any harmonics simply because the harmonics are absorbed and not allowed to bounce back and forth between the transistors and the filter. It also reduces the mismatch presented to the transistors. One other aspect is that since the harmonics are absorbed, there is less energy to mix within the active devices used in this amp, thereby lowering inter-modulation causing distortion products. I made a few component changes to his design. First, I used larger toroid forms. I used capacitors rated for higher current and voltage than was used in the original design. I moved the dump resistor from the board, and mounted it to the bottom of the chassis. I also chose one that can handle the full power of the harmonics Mine is rated at 50 Watts.
Go to this page to see the coil data I used, and here to see the parts list. These are starting places, and you will probably have to remove a turn, and stretch or compress the turns to tune each coil. Use the method shown below. It works very well, and isn’t affected by stray capacitances or load much. It should give you very satisfactory results.
The board layout was done with a program called “IVEX WinBoard[v]. It is available at many parts suppliers. The program details are available at http://www.ivex.com. You can also download a demo version here, but it has very limited capabilities. The version that I use can be purchased for $29.95. It allows you to print full size templates. You also are able to produce Gerber files, if you have a vendor that can make the boards for you. One advantage of this is that the vendor will silk-screen the board for you if you want. If you have Ivex Software, the schematics can be downloaded here. I recommend you right-click on this link, and select “Save Target As…”
The inductors for the low-pass section of each filter are wound with two parallel 18 gage wire. My calculations showed that the filters had to carry three and a half amps at the operating frequency. This called for a minimum of 16 gage wire. The first attempt to wind a coil on the toroid form using 16 gage proved to be a formidable task. I discovered that I could not bend the wire around the form without fear of breaking the toroid form. I came upon the idea to wind a double winding of 18 Gage. This gage is much easier to wind, and even winding 2 strings at the same time proved to be much easier. (Picture of 80 meter coil)
The 10/12 meter, 17/15 meter and the 20 meter bands were wound on T130-0 forms, and the 40/30 meter, 80/75 meter, and 160 meter bands were wound on T130-6 cores. The reason for the choice of material was decided on because of the huge re-circulating currents expected at this power level. Using higher permeability cores would cause excessive heating in the form. Also, temperature stability is a must, as large excursions in the tuning of the coils would degrade the filter response curves. I felt that this would exacerbate the losses in the core, and might cause more heating, causing a vicious cycle. The few extra turns needed on the form caused no problem other than adding a few extra inches of wire. (See pictures of complete coil set)
MAKING the Boards
EI9GQ[vi] did a wonderful article about making your own printed circuit boards. It is located on his web page. The following is a quote of how he does it;
To make smart looking PCB's, all
you need is: A computer, a laser printer, copper clad board, etching
solution, a clothes iron and some Epson glossy photo paper. You can buy
special film for making PCB's, but I have found that the Epson paper gives
better results. I use Epson photo quality glossy paper for inkjet printers.
I have used this method to make
PCB's for both DIL and surface
My laser printer is an old Apple
LaserWriter II NTX (300dpi.) If you don't have access to a laser printer, use
an inkjet printer to print the layout on ordinary paper, then copy the image
to the Epson paper, using a photo copier. I haven't tried this method but it
When building the filter boards, you will find that there are areas on the ground planes that need to be drilled and a small wire fed to both sides, then soldered. This is to give continuity to both sides of the board. This should add to VHF and UHF stability of the amp. If you have the boards made at a circuit shop, and the holes are plated through you will not have this chore. Do not forget to include these. It is important to maintain ground continuity on both layers. This ensures the minimization of ground loops. It also helps to insure shielding.
Having low VSWR to the amplifier transistors is important. In this respect, the accuracy of the inductor values should be reasonably close to the predicted values. I used a very simple method to measure the inductors. I found a capacitor of known accuracy (I used a 2000 pF that measures 1986 pF on a bridge with a .1% accuracy for the lower value inductors and an 820 pF that measures 798 pF for the higher values), and connected it in series with the inductor I was making. Picture of test setup) I then bridged the output of a signal generator with these, and tuned the signal generator frequency to where I saw a dip in the output. Tune the frequency for a dip measured on the indicating device. I used an oscilloscope for my display. If you have an RF volt meter, this might be a better choice, as it could give better resolution. The dip indicates resonance. This method requires an accurate way to measure frequency. My signal generator, an HP-8640B, has a built in counter, but if yours doesn’t you can bridge the circuit with one as well. Stray capacitance doesn’t seem to affect the results very much. The formula for resonance is “=sqrt((25330.3/LuH X CpF” This is easy to use in a spread sheet program such as Microsoft Excel or works. That way, you won’t have to keep re-entering the variables in a calculator, and re-calculate your
findings. Simply plug in the Values and the spreadsheet program will do all the work. See the example below. Where the answer 2.154138 is, is where the formula resides. Your value of test capacitor will go where I have 1985, and the desired inductor you are trying to tune goes where the inductor value is shown. If the test setup produces a frequency lower than what your target shows in the program, then either spread the turns out on the form, or remove a turn to tune the coil. You want to get within 1% of that produced with the program. This is only a rough tuning, and when the filter is built, you will want to fine tune the inductors for lowest SWR with the filter terminated with accurate loads both on the output of the low pass filter and the high pass network. I started off with finding the self-resonance of my test fixture. This showed that I had a stray inductance of approximately .056 uH which has to be taken into consideration while pre-tuning the inductors!
Another method available would be to read the article written by Bill Carver[vii] which gives a description of a simple L/C tester. If you don’t have an accurate method of measuring inductors and capacitors, you might consider building this unit. It will give you the ability to construct inductors to within 1% or 2% of the required values. This is necessary to achieve a return loss of better than –22 dB, or a VSWR of less than 1.17:1 to the power transistors.
All of the toroid inductors wound on size T130 forms use two 18 Gage wires wound together as one wire. I chose to use this method as 16 gage is very stiff, and difficult to wind on the form. 18-gage wire is much easier to work with, and with two parallel windings, gives ample current carrying capacity.
Each individual filter is constructed on its own board; there is provision for 6 boards covering all of the MF/HF frequencies except 60 meters. Since the current FCC regulations only allow 50 watts PEP-ERP, there seemed no logical reason to include this band. These filters mount to a “mother board” which contains the relays to switch in the proper filter for the desired band of operation. Using this construction technique allows compressing the real estate so that a more dense construction can be achieved. The filter modules mount perpendicular to the motherboard. If fewer bands are needed or wanted, then just delete those band modules.
The filter selection is controlled by 18 relays. The first group of relays steers the RF output of the amplifier into the proper filter for the band of operation. The next relay directs the output of the harmonic filter to the dump load, a 50-watt non-inductive resistor. The final relay on the motherboard sends the output of the low-pass filter to the antenna changeover relays.
Control for the band switching comes from the HF radio. Many HF radios have BCD band data output that can be used to control band switching data for this linear. If your radio does not have this capability, then a six-position switch can be used. The BCD to decade decoder IC decodes the BCD data. The output of this circuit drives the MOSFET switches that turn on the proper filter. They also drive the band indicator LED’s mounted on the front panel. Because the rig gives the data out as BCD for each band, and the filter sections in many instances cover multiple bands, there are diodes that further decode this information. For example, the 6th filter covers the 12 and 10-meter bands. The 5th filter covers the bands 15 and 18 meters, etc. It would not be prudent to build filters for each band when one filter can easily cover the needs for multiple bands. The cost of building the filters, plus the real estate taken up by the extra-unneeded filters is wasteful. It would require twice the cost to build for nine bands as oppose to only six!
A separate board that contains 2 single pole double throw relays controls antenna changeover. In the un-energized position, the RF in jack is connected directly to the output connector, with an RF sensor mounted in between the output relay, and the output connector.
A small time delay is added, and is used to enable the transceiver after the time delay to prevent the transceiver from generating any RF until the changeover occurs in the amp.
The keyer circuit consists of a pair of inexpensive relays that switch the input/output through the amplifier board. (Schematic) Q2 senses the key line from the transceiver and inverts that signal, and sends it to Q4, an emitter follower that amplifies the current to drive Q1 and Q3. Q1 goes low, enabling the relays. Q3 enables the transceiver transmitter section only after a small time delay to prevent the exciter from producing any RF until the relays in the amp have had a chance to switch. It typically takes 15 to 20 micro-seconds for the relays to energize fully, so 60 milli-seconds should be adequate. If the time delay is to short for your relays, then increase the value of R4 accordingly. The transceiver will not be enabled until the voltage on Q3 reaches 5 – 6 volts.
Diode D1 enables fast turn off of Q3, while D2 drives the FET and C1 through R4. Although this circuit is a little slow for QSK at high speeds, it does protect the exciter and the power amplifier from seeing an open while the relays are switching. It also prevents the contacts from seeing any RF power while transitioning.
If you wanted a lamp or LED to indicate that the amplifier was keyed, the circuit for Q3 could be duplicated on your board to do this.
The meter is a 1 ½” square faced meter that was picked up surplus for next to nothing. I am using one that has a 0 to 100 Micro-Amp scale. Simpson manufactured it. You can use any sensitivity you can find, just change the resistor values for the scale you use. Mine is built for 10,000 ohms per volt. If you find a meter with 1 Ma. Full scale, use 1,000 ohms per volt. If you should be so lucky as to find a 50 Micro-Amp movement, then use 20,000 ohms per volt to figure out what resistors to substitute here. Meter Circuit
The scale was removed, and a new one was made using a really nifty program called meter129[viii]. See the article printed in the October issue of QST. Read and follow all the links at the web page shown in the footnote, as several updates have been added since the article was published and several needed utilities are also available to help you get your printer to interface with the program. This program helps you make a really professional meter face with the appropriate scales. It has the ability to make linear, logarithmic and anti-logarithmic scales. You can even draw up to four scales on one meter like I did. You can choose to draw the arcs or not. The meter face that I made came out as good as a professionally manufactured meter, with exactly the legions that were needed. Mine reads 0 to 600 watts on the top scale to represent power out, 0 through 25 amps indicating drain current for the next scale, 0 - 60 volts showing the supply voltage, and 0 to 40 watts of drive power.
The meter scales I drew are referenced to two arcs. The tick marks are then drawn to indicate from both the upward and downward directions from the arc. The top arc is used for the power out, and amp scales, and the lower arc references the voltage and drive scales. The software intelligently knows where to draw the tick marks, and place the arcs and scale readings. You only have to enter the data via windows that clearly show you what will happen to the meter scale.
The meter is switched using a six position double circuit rotary switch bought at Radio Shack. This part cost me $3.98. I wired it up to provide readings of power out and reflected power by connection to the RF power sensor connected to the antenna relay board module. Current is read off a home-made series shunt. This is constructed of a short piece of brass tubing that is approximately 6” in length. I used a lab grade power supply and a 1 ohm resistor as a load. I set the lab power supply to 12 volts, and set the current limiting to 10 amps to calibrate the point where the panel meter read 10 amps. At this point, I soldered the meter lead. If needed either you can move the tap wire, or a notch can be filed in the center of this to raise the resistance until it provides the proper reading to the meter. Be careful not to file the tubing to small, or it may act like a fuse!
Drain voltage is read directly from the power supply through a 600K resistor, and drive watts is read from the power sensor on the amplifier input. The remaining position can be set up to read other monitored points if desired. I chose at this time not to. Maybe some day I might add another point of interest. Some suggestions might be heat sink temperature or bias voltage. Maybe even AC mains voltage. Since you are reading this with the desire to build your own amplifier, feel free to personalize your creation!
My knob also is a Radio Shack item. There are four of them in a package. I do recommend filing a small flat spot on the shaft so that the knob cannot slip on the switch. I did this to prevent the knob from slipping on the switch shaft.
More to come soon!
[i] Communications Concepts web page. Check here for parts
[ii] Motorola Engineering Bulletin EB104, by Helge Grangeburg. 1983 Available for reading and
[iv] QEX magazine, July/ August, 1999 pg 20
[vii] The LC Tester, Communications Quarterly, Winter 1993, a copy of the schematic can be found