Designing a Station for the Microwave Bands: Part 2

Glenn Elmore, N6GN

A complete 10 GHz amateur SSB/CW station

Part 1 discussed why the microwave amateur bands may be better than lower frequencies for many applications, though in the past amateurs have viewed them as line-of-sight realms. It described some of the inherent advantages microwaves have for point-to-point communication, even over modern higher power HF. VHF, or UHF stations. These advantages make them very attractive for high volume, high data rate communications like those required for amateur networking.

A local oscillator frequency scheme using common PC boards was presented. It can be used to get a station on all of the amateur microwave bands with a minimum of redundant construction. This scheme uses conventional lower frequency components, readily available microwave oscillators, and only a small amount of additional microwave construction to produce a high quality narrowband station. Part 1 and the rest of this series demonstrate this approach by describing construction of a complete 10 GHz amateur SSB/CW station _the station that holds one end of the current North American 10 GHz DX record of 414 miles.

Spectral Purity

The cornerstone of this station is a spectrally pure and stable 1010 MHz oscillator. Spectral purity, sometimes overlooked more than it should be even on the HF bands, is of particular importance when operating on microwave frequencies. This is because the"contamination" produced by angular (phase or frequency) modulation of a low-frequency reference signal is multiplied right along with the signal itself when harmonic is used in a microwave system. The fact that drift and frequency errors are multiplied is well known to anyone who tries to "net" a pair of fm transceivers on 1200 or even 440 MHz. However, these frequency domain "imperfections" are members of a whole class of impurities given the name "phasenoise". Even a quartz oscillator in a modern HF transceiver exhibits this to some degree. In a well-designed oscillator the "cleanliness" of a signal is related to its operating frequency. On the amateur HF bands these noise characteristics may be so small relative to normal signal-to-noise ratios that they are unobservable, except perhaps as an increase in background noise level down the band from a local "big gun". Some of the early synthesized ham band transceivers exhibited this as noise "humps" a few kHz either side of the carrier frequency on both transmit and receive. Commercial amateur equipment has improved to the point where fundamental overload or other factors usually come into play before the phase noise of the local oscillators is observed. However, as higher frequencies are required and higher harmonic multiples of reference oscillators are used, these unwanted components are multiplied. The relative amplitude of these unwanted signals follows a 20 log N rule, where N is the harmonic number. This means that on the tenth harmonicof a signal, the phase noise sidebands can be expected to increase by 20 log(10) or 20 dB. The 100th harmonic will be 40 dB worse than the fundamental. Consequently, a "clean" signal at 10 MHz, one with say -90 dBC (dB relative to the carrier) noise sidebands or fm spurious signals, might be 60 dB worse at 10 GHz, or -30 dBC. On an S9 signal such noise might be barely audible; however, if the fundamental oscillator was only - 60 dBC the resulting microwave signal might be unusable for communications. Because the 1010 MHz oscillator and its harmonics provide a local oscillator signal for a narrowband station, spectral purity must be maintained. Although a PLL can serve to "clean up" a poorer oscillator at frequencies close in to the carrier, no improvement is made beyond the PLL bandwidth. For this reason the bestavailable oscillator should be used.

fig. 1. The 1010 MHz oscillator circuit uses a coaxial resonator tuned with a voltage variable capacitor and coupling inductor. The oscillator transistor drives the phaselocking circuitry through a buffer amplifier. Two outputs for signal mixing or microwave PLLs are taken from the other end of the half-wave resonator after two stages of amplification.

1 GHz Rreference Oscillator

This oscillator can be used as the LO for a 1296 MHz signal mixer directly, as well as for the microwave harmonic downconverter reference at microwave frequencies. The active device is an inexpensive bipolar transistor. A coaxial resonator is made from PC board and brass tubing. Three separate buffered outputs are provided for phase locking, downconverter reference,and 1296 MHz signal mixer LO. The oscillator is tuned with the same UHF TV tuner diode used in the100 MHz reference along with a short length of wire coupled to the resonator at the low-impedance end.This provides approximately + 3 MHz tuning range around a 1010 MHz center frequency.

Oscillator tuning is somewhat novel; it works in much the same way as "loop modulation" of early radio days. Free running high-power oscillators were used with a carbon microphone connected across a single-turn loop located in the vicinity of a frequency determining inductor. As the operator spoke into the microphone the resistive load across the loop varied,which in turn modulated the loop current. Because this was an induced current, it tended to produce an opposing flux which effectively varied the net tank inductance and frequency modulated the oscillator. The technique worked, but care had to be taken to not to couple too closely or extract so much energy from the tank that one burned up the microphone- not to mention the operator!

The method used here doesn't extract much power from the tank, as the load the varicap presents to the loop is mostly reactive. Any such dissipation is undesirable as it acts to lower the operating Q of the resonator. The varicap value and coupling wire inductance are chosen to be below self-resonance for any tuning voltage. This is done to limit the maximum current and control energy loss in the tuning circuit resistances. If the tuning circuit tunes too close to resonance, oscillation may stop. With nominal loop dimensions and the indicated varicap, the 1010 MHz oscillator tunes with a nearly straight frequency/voltage tuning curve. The 5 MHz tuning range is ample to maintain lock once the other loops and coarse tuning are adjusted to center the output frequency.

fig. 2. The 1010 MHz oscillator is built from PC board and copper flashing. The assembly is divided into three compartments-one for the resonator and one each for the oscillator and signal amplifier circuits. Minimum lead length is usedwhen components are soldered directly to the circuit board material.

Two versions of this oscillator have been built. The first used a quarter-wave line allowing physically smaller construction, but requiring a dielectric supportfor the high-impedance end of the line to obtain thelowest "microphonics". The second approach uses a half-wave line and, although longer, is simpler to construct. The quarter-wave version allows tuning versatility by "telescoping" the inner conductor with a length of the next smaller size brass tubing sliding through the center of the fixed tubing. It tuned it continuously from 800 to more than 1200 MHhz.The half-wave version has another advantage. When you place the oscillator transistor with its isolation amplifier on one end of the resonator and signal amplifiers on the other end, the resonator serves to isolate spurious signals which might be present in the downconverter/phase lock circuitry. This "autofiltering" makes it easier to achieve -80 dBC spectralpurity at 1 GHz. Similar performance can be obtained with the quarter-wave version, but more stages of isolation and careful shielding are required.

fig. 3. 1 GHz oscillator.

The BFR91 oscillator transistor is optimized to have maximum negative resistance at 1 GHz with the insertion of approximately 7 nH of inductance in its base lead. This inductor is just the 3/8 inch of lead length between the transistor package and the feedthrough capacitor ground on the end wall. The emitter is coupled into the resonator with a loop, also bypassed in a feedthrough capacitor on the same wall. Base and emitter biasing resistors are connected on the outside.The 1 GHz oscillator schematic is shown in fig. 1. Figure 2 shows the mechanical dimensions and positioning for the resonator, feedthrough capacitors, andcoupling loops. A photo of the completed oscillatoris shown in fig. 3.


The oscillator could be built entirely of PC board, but I chose to make the end walls from copper flashing. This makes it easier to solder the brass tubing after the resonator box has been assembled. The sides,center, ends, and partition should all be punched or drilled before soldering. Holes for the oscillator emitter loop and buffer amplifier input loop are in the center wall. Amplifier transistor emitters and all bypass capacitors can be soldered directly to the board material with virtually no excess lead length. The oscillator emitter lead can protrude right through the center wall hole and be soldered to its coupling loop. An 8-32 brass nut should be soldered to the inside wall of theresonator so that a tuning screw can be inserted later. If possible use 1/8th or 1/16th watt resistors. The physically smaller packages should have less associated inductance. Choose feedthrough capacitors small enough to fit snugly against the brass tubing on the oscillator end. These must be soldered in place since their nuts would otherwise interfere with the brass tubing protruding from the end wall.


Begin check-out without tuning screws and apply12 volts. The oscillator emitter (measured at the outside of its feedthrough capacitor) should sit at about3.5 volts, and the amplifier transistors should have 6 to 10 volts on their collectors. Collector currents ofabout 15 mA for the BFR91 and 40 mA for the BFR96 amplifiers are fine. All three outputs should have a load connected; a 50-ohm resistor may be tacked across an unused output as a temporary load. If a powermeter or other calibrated detector is not available, an inexpensive power detector may be made (fig. 4). Anapproximate calibration curve useful through the VHF range is shown in fig. 5. At 1 GHz the curve may not accurately predict the detected power because of differing construction techniques and component characterisics, but the detector should still be useful for determining relative output powers and adjusting the 1010 MHz circuits. I built the detector right on the cable end of the same type of SMB coax connector I used throughout. You can use it to verify ECL outputs as well as oscillator performance.

fig. 5. A plot of the detector output voltage as a funtion of input power shows a useful range from about 0 to + 16dBm

A 1 GHz frequency counter or a spectrum analyzer is extremely useful for tune-up. If such test equipment is not available, build the 1 GHz harmonic downconverter described in the next section. Use it to convert the 1 GHz signal down to the HF range of a generalcoverage receiver or low frequency counter. If you usea receiver, couple the downconverter lightly or use an attenuator to avoid overload. Overloading can cause confusion because of images and other spurious responses.

With an applied fixed tuning voltage of 6 volts, insert the tuning screw and set the frequency to approximately 1010 MHz. Adjust the emitter loop slightly to assure oscillation while varying the tuning voltage over the 2-to-10 volt varicap tuning range. Reduce coupling by decreasing the area of the loop and positioning it further from the brass tubing. Use the minimum coupling to maintain output so you can avoid unnecessarily loading the resonator and degrading phase noise.This coupling is somewhat dependent on resonator loading by both the tuning circuit and buffer amplifier input loops. Adjust the buffer amplifier loop (made from the coupling capacitor lead) for minimum coupling consistent with maximum power out of the power splitter. Adjust the emitter loop to maintain output over the whole tuning range. Some iteration between these two adjustments may be necessary to arrive at the best settings. If you find that the oscillator dies at the high end of the tuning range, or just above 10volts, you may need to lower the tuning circuit resonant frequency. Do this by lengthening the tuning inductor slightly. The values shown in the drawing should provide a good starting place and should work without modification. Extreme emitter loop overcoupling can cause mode hoping, the output switching rapidly between two frequencies. This is not a problem if the above adjustment procedure is followed. Reduce coupling if you observe spurious sidebands on the unlocked oscillator or find low-frequency scillations on the bias feedthrough capacitors.

The output amplifier on the signal side is followedby a power splitter made from two 2-inch lengths of semi-rigid coax. This is a simple way to provide two outputs. If only one 1010 MHz source is required, it may be omitted and the single BFR91 buffer amplifier used to provide + 10 dBm for a signal mixer. The two stage amplifier with a BFR96 in the output and the power divider can easily provide two + 13 dBm (20milliwatt) sources.

Once the loops are positioned for proper power output, all that remains is to readjust the tuning screw so the oscillator "free runs" right at the desired frequency. If you adhere to the dimensions for the halfwave version, the oscillator should run at about 1025MHz with no tuning screw and only the 4.5 volts from the resistive divider on the tuning input. It should tune down mechanically to 1000 MHz without a significantchange in output power. When you obtain the proper frequency, secure the tuning screw locking nut.Verify that approximately + and - 2 MHz tuning is produced with 10 volts and 2 volts on the tuning input, respectively.

fig. 6. This harmonic downconverter produces an IF output which is the difference between N multiplied by the downconverter fundamental and the phaselocked oscillator input, where N is an even number. With inexpensive diodes and amateur construction techniques, IF output power of -30 dBm is readily available.

PLL Harmonic Downconverters

The downconverters themselves are similar, although implementation at 10 GHz is somewhat different from that at 1 GHz. Anti-parallel diodes are used with a diplexer arrangement to couple signals in and out. The downconverter block diagram is shown in fig. 6.

The anti-parallel diode pair is effectively an even harmonic mixer. Its simplicity and built-in protection from overload and static damage make it attractive for this application. Depending on harmonic number and phase-locked oscillator frequency, -30 to -40 dB conversion efficiencies are obtainable even with "hamshack" construction_i.e., discrete components or microstrip circuits cut out of Teflon, epoxy PC board material with a small hobby knife. The high-pass filter couples the reference fundamental into the diodes; the low-pass filter couples the IF out. The oscillator can be connected directly to the diode pair through a small capacitor.

At 1 GHz, packaged diodes and discrete capacitors and inductors can be used. Lead length should be kept to a minimum, but otherwise the circuit is extremely simple to build. The diodes generate considerable energy at odd harmonics of 100 MHz. However, the isolation of the 1010 MHz oscillator resonator, mention the buffer amplifiers, keeps this energy from showing up in the signal output. These sidebands are for the most part amplitude, not frequency modulated, and don't get "amplified" when higher harmonics of the 1010 MHz signal are used as a reference signal in the 10 GHz downconverter.

fig. 7. The 1 GHz harmonic downconverter and IF amplifier circuit produce approximately -10 dBm IF output when drivenwith 100 MHz ECL levels and the 1010 MHz oscillator. Leads should be kept short and good VHF practice followed, otherwise no special precautions need to be taken.

fig. 8. Completed 1 GHz downconverter and common PLL board

The PLL IF signal from the downconverter is approximately 30 dB below the reference or locked oscillator levels. This conversion loss is made up for in the bipolar amplifier and the two ECL line receivers on the phase-lock circuit.With 10 to 13 dBm reference drive, IF output doesn't change dramatcally for 0to 10 dBm oscillator input.Around -30 dBm PLL IF power is typical for bothconverters_plenty to drive the last ECL line receiver before the phase comparator well into saturation. The IF output may actually drop if the oscillator input level is increased too far. The 1 GHz harmonic downconverter schematic diagram is shown in fig. 7. Figure 8 is a photo of the completed 1 GHz downconverterand common PLL board.

The 100 MHz reference signal is bandpass filtered and amplified from the 0 dBm ECL levels. The filtering makes sure that any low level, low-frequency digital signals which might be present on the 100 MHz ECL output don't "ride" straight through to the PLL IF amplifiers. Diode drive of 10 to 20 milliwatts is adequate.

The 10 GHz downconverter is functionally the same as the l GHz version. Here, however, a pair (or half a quad) of diodes in a small package is used to avoid parasitic inductance and capacitance associated with the larger discrete diodes. Many of the filter elements are made using microstrip techniques instead of lumped components. Chip capacitors are used to minimize parasitic inductance.

Because most of the 10 GHz oscillator power is needed for converting the VHF signal to and from 10GHz, a hybrid coupler is used to extract only enough to make the PLL downconverter operate. This hybrid has one of its input ports terminated with a discrete resistor. This termination needn't be very good at 10GHz, as the object of the coupler is simply to extract a sample of the energy (10 dB or so down) and its directivity isn't particularly important. Use as physically small a resistor as possible with 0lead length. All of the high-impedance lines may be made from some small diameter wire and soldered across the wider traces. Number 38 wire should be fine for this.

The signal mixer is shown with the 10 GHz downconverter and can be built on the same board at the same time. This makes it possible to get on the band as soon as the 10 GHz oscillator is locked and a VHF IF is available. The 10 GHz harmonic downconverter is shown in figs. 9A and 9B.

fig. 9. The 10-Ghz and 1010 MHz downconverters arefunctionally identical. At 10 GHz. however, microstrip components replace discrete components. A hole is provided in the 1/32 inch Teflon board material under diode ring, D1, to allow shorting the diode leads to the backside ground. Radial transmission lines on these same leads help assure a low-impedance ground connection.

Fig 10
shows a 2:1 layout of the combined 10 GHz downconverter/signal mixer.

fig. 10. The downconverter/signal mixer assembly is the only microwave circuit that needs to be constructed. Thetraces may be made using a small hobby knife by the"cut and peel" technique. Tolerances are not extremelycritical although a microscope can be a great aid. Moredetail of the downconverter and signal mixer portionsis given in figs. 9 and 19, respectively.

fig. 11
shows a schematic of the microwave board.

fig. 11. The microwave board mainly uses distributed elements. Impedance of the microstrip transmission lines is controlled by width variation. The high-impedance lines can be made from bare No. 30 wire soldered right to the traces; thisis easier than trying to cut or etch 0.005 inch wide traces.

Locking to 1010 MHz

The 1010 MHz common PLL circuit is nearly identical to that of the 100 MHz reference oscillator, only the loop filter values are different. For this loop, the phase comparator VCO input comes from the filtered and amplified output of the 1 GHz harmonic downconverter. A 35 MHz low-pass filter follows the downconverter; the PLL IF is first amplified by a two-stage controlled-gain amplifier. l used this configuration instead of another ECL line receiver for two reasons: it allowed variation of the stage gain by changing a single resistor value, and the bipolar amplifier has lower bandwidth than the ECL line receiver. The rest of the PLL circuitry is identical to the 100 MHz phase lock except for the loop filter component values. The bandwidth of this loop is set to approximately 50 kHz.

Once the 1010 MHz oscillator is built and adjusted, you are ready to lock it up. Use one of the common PLL boards with the loop filter component values inpart 1, Table 2. Set the jumper wires on the phase comparator input for the " + " configuration. If the PLL board is working properly (remember that you can test it ahead of time by using it to lock up the 100 MHz oscillator), the loop should close and "pull in" the 1010 MHz oscillator exactly on frequency. This lock can occur if the 100 MHz loop is locked or free running, and the output frequency should be exactly 10.1 times the 100 MHz crystal oscillator frequency. Make sure you use the 10 MHz reference to lock at 1010MHz and the 20 MHz reference if you are trying to lock to 1020 MHz.

Troubleshoot any problems by checking the PLL board and the 1010 MHz oscillator independently of each other. As long as the oscillator tunes over the correct range and the PLL board is working, there should be no difficulty in achieving and maintaining lock. Once this is done, you have an LO for use in a1296/2304 station or as a reference oscillator for locking your 10 GHz oscillator.

10 GHz oscillator selection and locking

The 10 GHz oscillator is locked in the same manner as the 1010 MHz reference. The tuning Circuit may depend on the type of oscillator available. Generally, only enough tuning range to overcome drift and instability is used. If too much tuning range is provided, the microwave oscillator might get on the wrong side of the downconverter reference frequency harmonic, giving an IF with the wrong tuning sense.If this happens, the PLL amplifier tries to tune the oscillator in the wrong direction to acquire phase lock nd the loop will remain saturated and unlocked. Fora 20 MHz PLL IF, 30 MHz of total electronic tuning range should be adequate, and this combined with about a 10-volt swing out of the loop amplifier suggests a 3 MHz/volt tuning sensitivity. If the microwave oscillator is unstable or drifts (necessitating a greater tuning range), an ECL divide-by-2 or divide-by-4 could be inserted right at the phase comparator input. Of course, this would produce a different locked output frequency, and all other IF and oscillator frequencies in the system might have to be reselected. The loop filter component values would also have to berecomputed.

fig. 12. Measured tuning curve for a M/A-ComGunnplexer'M. The curve is fairly draight indicating nearly constant tuning sensitivity, particularly in the 410 voltregion.

fig. 13. A three-terminal regulator provides clean bias for the Gunnplexer. A Zener diode and two resistors provide an offset and scale the tuning to the approximately 3 MHz/volt sensitivity required for the common phaselock board.

Selection of the 10 GHz oscillator depends upon what is available and within your budget. The M/AComm Gunnplexers work extremely well and require very little additional circuitry. If you have one of these as part of a wideband station, you may want to use the 10,220 MHz locking scheme. If there is already some broadband 10 GHz activity in your area and you don't want to give it up entirely, this approach will allow switching between modes. The Gunnplexer canbe operated with its internal diode mixer for operation on 10220/10250 wideband duplex, or phase locked to10220 and used with a 148 MHz SSB transceiver for10368 MHz narrowband weak signal work. The wideband station can also be run phase locked with modulation of the 20 MHz reference signal in the 1020 MHz loop phase. (This should end any local discussionsabout who is or is not on the right frequency!)

The M/A-Comm Gunnplexers have electronic tuning and need only level shifting and scaling of the tuning voltage. A typical tuning curve for a GunnPlexeris shown in fig. 12. Driving the tuning input directly from the loop amplifier provides too much tuning range and could allow "latch-up" on the wrong side of the IF, as mentioned before. It is a simple matter to scale the tuning input to reduce the approximately7 MHz/volt sensitivity down to about 3. A circuit providing this scaling, as well as a regulated 10-volt bias supply, is shown in fig. 13. This circuit will maintain proper output and tuning even when the power supply voltage drops slightly below 12 volts. A low dropout regulator may be substitued for the LM317K for particularly low inputs. This is of concern primarily when mountain topping with discharged batteries as the only power source! The phase-locked Gunnplexer produces an excellent 10080 MHz signal (fig.14).

Some means of tuning must be provided if anoscillator without an electronic tuning input is used.The Gunn oscillators in automatic door openers canbe made to work by using bias voltage "frequency pushing". These are very similar to Gunnplexers except for their lack of electronic tuning and a mixer diode. The tuning deficiency can be overcome by using the bias/tuning circuit in fig. 15. Here a three terminal regulator sets the bias and tunes the oscillator for phase locking. To pick the nominal bias point, plot a frequency versus bias voltage curve for your particular oscillator, this will vary from oscillator to oscillator. Usually a range of bias can be found (often just on one side of maximum power output) that provides a fairly straight tuning curve or nearly constant tuning sensitivity. A plot of a typical bias-tuned oscillator is shown in fig. 16. The tuning resistor valuesare selected to tune over a 24 MHz range with 2 to10 volts on the tune input. The nominal operating point, with 6 volts applied to the tuning input, is set at the center of this range. If 24 MHz tuning is not possible, use the maximum available and recalulate the PLL component values for the different tuning sensitivity.

fig. 14. Signal produced by the phaselocked Gunnplexer.

fig. 15. An adjustable three-terminal positive regulator can both bias and tune a surplus Gunn oscillator via "frequency pushing". the oscillator's frequency/bias voltage dependency. The table shows alternate component values for different "push-tuning" sensitivities. Part of the voltage setting and tuning resistance is bypassed to reduce noise on the regulator output.

fig. 16. This is a tuning curve of a surplus Solfan oscillator of the type used in burglar alarm motion detectors and automatic door openers. Both output power and frequency are dependent upon bias voltage. By plotting a similar curve and selecting a useful portion of the tuning curve, you can find component values for biasing and tuning almost any similar oscillator. In this case, a bias of 8.25 volts +-1.25 volts will tune the output over approximately a +-12 MHz tuning range.

The three terminal regulators work in this application because they have several hundred kHz of bandwidth and can follow a 50 kHz bandwidth error signalwithout adding much additional phase shift. This is necessary for the loop to remain stable. The regulators do add some noise to the oscillator output when used in this configuration; reduction of this is the reason for splitting up and bypassing part of the voltage setting and tuning resistances. This technique is not the ultimate in low phase noise performance, but the -90 dBC noise sidebands obtainable (1 Hzbandwidth) are more than adequate for amateur useand will probably never be observed unless signal strengths are 30 or more dB above S9. The Gunnplexers, with their built-in tuning, will probably be at least 8 to 10 dB cleaner than this. Although I have not tried them, many of the oscillators in automotive radar detectors should work well. Another source of suitable oscillators is the type used for police radar guns.The NEC ND751AAM for 10 GHz (ND610AAM for 24GHz) has similar characteristics. Any of these 10 GHz sources should have adequate drive power for the signal mixer described next.

The oscillators' output and antenna connections need to be in coax in order to use the downconverter and mixer. Although waveguide is well behaved and very low in loss, coax is versatile and convenient. I have used coax throughout the 10 GHz station, both at 1 and 10 GHz. Miniature SMB "snap on" connectors work well at 1 GHz and below, even when used with poor quality lossier coax cable. In the microwave region, 0.086-inch semi-rigid cable is a pleasure to work with; the cable and corresponding SMA connectors are fairly easy to find. To cut the cable to length, first score the outer conductor with a sharp knife; then grab each side of the score mark with a pair of needlenose pliers and break. The Teflon dielectric can be trimmed away and the cable end slid into the connector or soldered directly to the circuit, depending upon the application.

If your oscillator is similar to the door-opener type, it probably has a waveguide output and will require a waveguide-to-coax adaptor. These are often available as surplus but if you don't have or can't get one, it is easy to build an acceptable substitute. The version shown in fig. 17 made from a short length of commercial waveguide works very well, although you'llneed metal-working equipment. If your shack doesn't include much more than a soldering iron, hacksaw,and file, the second version made from PC board in fig. 18 is for you.

fig. 17. A waveguide/coax adapter which provides verylow loss and better than a 1.2:1 VSWR can be made froma short length of standard WR90 waveguide. A standardSMA connector is threaded into a hole in the broad wallof the guide located 0.200 inch from the shorted end. Aflange made from a suitably stiff piece of brass or copper is soldered to the other end.

fig. 18. If metal-working equipment necessary to fabricate the adapter shown in fig. this not available. an acceptable substitute can be made from standard 1/1~inchPC board and brass shim stock. A channel is first madefrom the board material and then the brass top/short issoldered. Continuity from the inside of this homebrewwaveguide to the front of the flange is provided by stickycopper foil tape.

After selecting your 10 GHz oscillator, build the appropriate tuning circuit. If a means of measuring 10 GHz frequency (a 10 GHz counter or spectrum analyzer with 1 MHz frequency resolution) is not available, you may use the 10 GHz downconverter and locked 1010 MHz source to determine oscillator operating frequency. An old general coverage receiver with poor selectivity is great for this, because an unstable signal is easy to hear as it drifts past. Take the same precautions mentioned earlier to avoid overload. Whenyou do hear a signal, verify that it is on the correct side of the 10iO MHz reference harmonic at 10.1 GHz by making sure that the IF signal is tuning lower as you tune the microwave oscillator higher. Use mechanical tuning for this because (unless you have a M/AComm or other "known" oscillator) you can't be sure what the sense of the electronic tuning is. If the tuning characteristics are unknown, use your general coverage receiver or low-frequency counter on the PLL IF to plot a tuning curve as a function of the oscillator bias voltage. A few oscillators will "mode" and jump frequency, particularly when not properly matched and operated at extremely low or high biases. A nominal operating point between 6 and 10 volts is appropriate for most oscillators I have tried. Select the nominal bias voltage as the center of a reasonably straight 24 MHz range near the maximum power bias point, or in a mode-free region. Select the tuning scaling resistors from fig. 15 based on the change in bias necessary to produce 24 MHz frequency change; this gives 3 MHz/volt sensitivity at the tuning input. The sense of this tuning may be either positive or negative, depending on your particular oscillator. For the motion detector oscillator plotted in fig. 16 I chose a nominal bias point of 8.25 volts. Tuning resistor values were selected for the required volt change (approximately 2.5). These values cause the 2 to 10 volts from the loop amplifier to tune the oscillator over a 24 MHz total range. Resistor and capacitor values for some different oscillator sensitivities are shown in the table.

When you are confident that the oscillator is tuning correctly, preset it to 10080 MHz with 6 volts on the bias circuit tuning input. Do this by coarse tuning for a 20 MHz PLL IF on the correct side of 10100 MHz. If the PLL board is functioning, locking should now be no more difficult than locking the 1010 MHz oscillator. Remember to select the proper wire jumpers based on "high side') LO and the tuning direction of your particular oscillator.

As for the 1010 MHz oscillator, troubleshoot any problems by separating the PLL components and testing them individually. Make sure that the PLL board works on a lower frequency loop. Verify that there are suitable 1-volt peak-to-peak ECL levels on both phasec omparator inputs. Also make sure that the 10 GHz oscillator is tuning properly. Be sure that there is no large (bigger than 1000 pF) bypass capacitor across its bias input; this could limit the fm bandwidth.

Signal Mixers

Commercial mixers that give good performance up through 2304 MHz are available at reasonable prices.Simple "rat race" mixers can be made on Teflon PC board for all bands up to and including 10 GHz; they don't work as well at 24 GHz and above because of packaged diode size and parasitics. A diode mixer with less than 7-dB conversion loss at 10368 MHz (with10080 MHz local oscillator injection) can be cut out of apiece of circuit board. This by itself (no amplifier,preamplifier, or transmit/receive switch) can give S9 signals between similar stations with 4-foot dishes separated by 10 miles!

The 10 GHz signal mixer uses the same diode ring and board material as the 10 GHz harmonic downconverter. Building it on the same piece of board material eliminates two connectors and some coax along with their associated losses. A balun is used to match the mixer diode's IF impedance to 50 ohms. You can make this balun from two toroidal cores, or use aVHF TV 300-to-75 ohm balun. Conversion loss of under 10 dB should be possible over a range of local oscillator powers. Low barrier diodes are indicated in the parts list, but medium and high barrier may be substituted if sufficient 10080 MHz oscillator power is available. Higher drive levels make higher IF levels possible on transmit, and therefore higher 10368 MHz transmit power. To avoid serious distortion, IF power should generally be kept at least 10 dB below the available local oscillator power. A close-up of the 10 GHz signal mixer is shown in fig. 19.

Fig 19. 10 GHz Signal Mixer (Caution, there are errors in this drawing see Part 3

Build the signal mixer as part of the downconverterassembly, and you can be on the air as soon as the 10 GHz oscillator is locked and you have a suitable SSB IF transceiver. Just hook it through a bandpass filter to your antenna!

Part 3 will discuss the following:

a 260 MHz locked oscillator along with amplifiers and switching for the 280-290 MHz IF transverter and a two-stage, 16-dB gain, 2.5-dB noise figure 10 GHz amplifier that can be used on transmit and receive. Two such stations connected to modest size antennas should improve your DX possibilities and could help you break the current world 10 GHz DX record!