The base of the power device sits through a cut-out in the floor of the diecast box and is bolted directly onto a small extruded aluminium heatsink. An alternative would have the base of the power device sitting on the floor of the diecast box. This is not recommended for two reasons, both concerned with providing an effective path to conduct heat from the FET. Firstly the floor of the diecast box is not particularly smooth, which results in a poor thermal path. Secondly, having the floor of the diecast box in the thermal path introduces more mechanical interfaces and hence more thermal resistance. Another advantage of the chosen constructional technique is that it correctly aligns the device leads with the top face of the circuit board.
Using the specified heatsink will require the use of forced air cooling (a fan). If you plan not to use a fan, a much bigger heatsink will be required, and the amplifier should be mounted with the heatsink fins vertical to maximise cooling by natural convection.
The circuit board consists of a piece of fibre glass PCB (printed circuit board) material clad with 1oz Cu (copper) each side. I used Wainwright to form the circuit nodes - this is basically self-adhesive bits of tinned single sided PCB material, cut to size with a hefty pair of side-cutters. An easy alternative is to use pieces of 1.6mm thick single sided PCB material, cut to size and then tinned. These are glued onto the ground plane with a cyanoacrylate type adhesive (e.g super-glue or Tak-pak FEC 537-044). This method of construction results in the top side of the PCB being an excellent ground plane. The only exception to this are the two pads for the gate and drain of the FET. These were created by carefully scoring the top layer of copper with a sharp scalpel, and then removing the slivers of copper with the assistance of a fine point soldering iron tip and the scalpel. Running the iron tip along the isolated piece of copper loosens the glue sufficiently for the Cu to be peeled off with the scalpel. The gate pad thus created is clearly visible in the photograph of the prototype
Having made the aperture in the PCB for the base of the power device to sit through, I wrapped copper tape through the slot to join the upper and lower ground planes. This was done in two places, underneath the source tabs. The copper tape was then soldered top and bottom.
See photograph for suggested component positions. The vertical screen to the right of the enclosure is a piece of double sided PCB material, soldered to the top ground plane on both sides. This is an attempt to improve the final harmonic rejection, by reducing coupling between the inductors that form the output match and the inductors making up the LPF. To do these kind of soldering jobs a 60W or greater soldering iron will be required - preferably a temperature controlled one. This iron will be too over the top for the smaller components so a smaller iron will be required as well.
As mentioned below, the LPF inductors are soldered directly to the tabs of the metal clad capacitors.
Suggested Rough and Ready Construction ProcedureCut out a piece of double sided PCB material for the main board (approx. 100 x 85mm) Create the aperture for the FET, using a selection of drills and files. Use the FET as a template, if required, but don't blow it up with static. Make sure you'll end up with the drain on the right side. Drill six holes in the PCB, these are to hold the PCB to the diecast box Place the PCB in the box and use the holes in the PCB to drill through the box Temporarily screw the PCB to the box Work out where the heatsink is going to go, underneath the box The device should end up towards the centre of the heatsink. Either drill some more holes through the whole lot, and re-use some of the existing PCB/box holes and extend these down through the heatsink. Temporarily screw the heatsink to the PCB/box assembly. When you look into the top of the box you should now see a piece of heatsink revealed, the same size as the base of the FET. Rig yourself up some static protection (if you've got an old blown-up device or a bipolar device in the same package you won't have to bother with this) and drop the device into the aperture in the board. Use the FET to give you give the centre positions of its' mounting holes Take everything to bits again. Make two holes in the heatsink for the FET Drill the holes in the two ends of the box for the RF connectors and the feedthrough capacitor Tin the PCB, top and bottom, with a big iron. Use just enough solder to get a smooth finish but not too much to create raised areas of solder, especially on the bottom, as these will prevent the PCB sitting flat against the box floor. Create the two islands for the FET gate and drain, as detailed in the above paragraph Solder copper tape between top and bottom faces of the PCB underneath where the source tabs will be Create the PCB islands, tin them, stick them on the PCB using the photograph as a guide Create and fit the screen between the amplifier and the LPF areas Fit all the remaining PCB components, with the exception of the FET Fit the PCB to the box and the heatsink Fit the and connect and the RF connectors and the feed-through capacitor Taking anti-static precautions again, apply the thinnest continuous film possible of heat transfer paste to the base of the FET. This can be conveniently done with a wooden cocktail stick Bend up the last 2mm of each of the FET's leads. This will make it much easier to remove, should the need arise Screw the FET to the heatsink. Too loose and the device will over-heat, too tight and you will distort the flange of the device and once again it will overheat. If you've got a torque screwdriver, look up the recommended torque and use it. If you've understood the instructions correctly, the tabs of the device will be fractionally above the PCB Solder the FET in with the big iron, first the sources, then the drain, finally the gate. You may have to disconnect L4 and L5 while you are fitting the FET, but don't disconnect R3 as this provides static protection for the device.
Schematic
Parts List
Reference Description FEC Part No. Quantity C1, C2, C4 5.5 - 50p miniature ceramic trimmer (green) 148-161 3 C3 100p ceramic disc 50V NP0 dielectric 896-457 1 C5, C6, C7 100n multilayer ceramic 50V X7R dielectric 146-227 3 C8 100u 35V electrolytic radial capacitor 667-419 1 C9 500p metal clad capacitor 500V 1 C10 1n ceramic lead through capacitor capacitor 149-150 1 C11 16 - 100p mica compression trimmer capacitor (Arco 424) 1 C12 25 - 150p mica compression trimmer capacitor (Arco 423 or Sprague GMA30300) 1 C13 300p metal clad capacitor 500V 1 C14, C17 25p metal clad capacitor 500V 2 C15, C16 50p metal clad capacitor 500V 2 L1 64nH inductor - 4 turns 18 SWG tinned Cu wire on 6.5mm dia. former, turns length 8mm 1 L2 25nH inductor - 2 turns 18 SWG tinned Cu wire on 6.5mm dia. former, turns length 4mm 1 L3 6 hole ferrite bead threaded with 2.5 turns 22 SWG tinned Cu Wire to form wideband choke 219-850 1 L4 210nH inductor - 8 turns 18 SWG enamelled Cu wire on 6.5mm dia. former, turns length 12mm 1 L5 21nH inductor - 3 turns 18 SWG tinned Cu wire on 4mm dia. former, turns length 10mm 1 L6 41nH inductor - 4 turns 22 SWG tinned Cu wire on 4mm dia. former, turns length 6mm 1 L7 2 ferrite beads threaded onto lead of C10 242-500 2 L8, L10 100nH inductor - 5 turns 18 SWG tinned Cu wire on 6.5mm dia. former, turns length 8mm 2 L9 115nH inductor - 6 turns 18 SWG tinned Cu wire on 6.5mm dia. former, turns length 12mm 1 R1 10K cermet potentiometer 0.5W 108-566 1 R2 1K8 metal film resistor 0.5W 333-864 1 R3 33R metal film resistor 0.5W 333-440 1 D1, D2 BZX79C5V6 400mW Zener Diode 931-779 2 TR1 MRF171A (Motorola) 1 SK1 BNC bulkhead socket 583-509 1 SK2 N type panel socket, square flange 310-025 1 Diecast Box 29830PSL 38 x 120 x 95mm 301-530 1 Heatsink 16 x 60 x 89mm 3.4°C/W (Redpoint Thermalloy 3.5Y1) 170-088 1 Double sided Cu clad PCB material 1.6mm thick A/R Copper Tape or Foil 152-659 A/R M3 nut, bolt, crinkly washer set 16 Non-Silicone Heat Transfer Paste 317-950 A/R
Photograph of Prototype Amplifier
Notes
Farnell Part Numbers are for guide only - other equivalent parts can be substituted. Metal clad capacitors are either Semco MCM series, Unelco J101 series, Underwood, or Arco MCJ-101 series available from, amongst other places, RF Parts. MRF171A available from BFI (UK), Richardson or RF Parts (US) Arco or Sprague trimmers are available from Communication Concepts (US) 18 SWG (standard wire gauge) is approximately 1.2mm diameter 22 SWG (standard wire gauge) is approximately 0.7mm diameter To make the inductors - wind the required number of turns round an appropriately sized former, initially use one wire diameter spacing between each turn. Then pull the turns apart to get the length required in the parts list table. Finally check the value using a network analyser and adjust accordingly. The exception to the above spacing rule is L4, which is close wound. Copper foil is available from craft shops (used in stained glass making) A/R = as required
Low Pass Filter TestingAny RF power amplifier must be followed by a low pass filter (LPF) to reduce the harmonics to an acceptable level. What this level is in a unlicensed application is a moot point, but as the output power is increased, more attention must be be paid to the harmonic suppression. For example, a 3rd harmonic of -30dBc on a 1W unit is 1uW, which is unlikely to cause any bother, whilst -30dBc 3rd harmonic suppression on a 1KW output results in a 1W of power at the third harmonic which is potentially problematic. So for the absolute level of harmonic radiation in the second example to be the same as the first, we now need to suppress the third harmonic by 60dBc.
In this design I made the decision to implement a 7 pole Chebyshev low pass filter. A Chebyshev was chosen as the phase and amplitude ripple within the passband was not critical, and the Chebyshev gives a better stop band attenuation than compared to say, a Butterworth. The design stopband was chosen to 113MHz, giving a 5MHz implementation margin from the highest desired passband frequency at 108MHz and the start of the stopband at 113MHz. The next critical design parameter was the passband ripple. For a single frequency design it is normal practice to choose a large passband ripple, for example 1dB, and tune the peak of the last passband maxima to the desired output frequency. This gives the best stopband attenuation because greater passband ripple results in more rapid stopband attenuation. A seven pole filter has 7 reactive elements, in this design four capacitors and three inductors. The more poles, the better the stopband attenuation, at the expense of increased complexity and more passband insertion loss. An odd number of poles is required as both the input and output impedance was designed to be 50R.
As this design is wideband, this constrains the passband ripple to a level such that the passband return loss does not become to horrible. Using the excellent Faisyn shareware filter design utility (available from FaiSyn RF Design Software Home Page) allows these trade-offs to be easily investigated, and I settled for a passband ripple of 0.02dB. This program also calculates the filter values for you, and outputs a netlist in a format suitable for inputting into the most popular linear circuit simulators. With 7 poles, the choice was available to use 4 capacitors and 3 inductors or 3 capacitors and 4 inductors. I chose the former, on the grounds that it results in one less component to wind. The capacitor values given from the faisyn program were examined to check that they were close to a preferred value, which they were. If they had fallen between preferred values, the options would include paralleling two capacitors together, which unnecessarily ups the component count, or subtly tweaking the stopband frequency and passband ripple to get a more desirable set of values.
To implement the filter, I decided to use standard size metal clad capacitors made by Unelco or Semco. The inductors were made from 18 SWG (standard wire gauge) tinned copper wire. In my experience there is little to be gained from using silver plated copper wire. The inductors were formed round of the centre of a standard RS or Farnell tweaking tool (FEC 145-507) - this has a diameter of 0.25 inch, 6.35mm. Otherwise use the appropriately sized drill bit. The outer two inductors were wound clockwise, the inner one was wound counter-clockwise. This is an attempt to reduce the mutual inductive coupling between the inductors, this tending to degrade the stopband attenuation. For the same reason the inductors are arranged at 90° to each other, rather than all in a straight line. The inductors are soldered directly to the tabs of the metal clad capacitors. This keeps losses to a minimum. A carefully constructed filter of this type can exhibit a passband insertion loss of better than 0.2dB. Here are the test results for the prototype unit.
No comments:
Post a Comment