The exciter has the functions of providing a stable, low noise, frequency-selectable RF signal, modulate it with the multiplex signal provided by the audio board, and amplify it to a controllable output power sufficient to drive the power amplifier. My exciter uses a PLL frequency synthesizer, which covers the FM band in 100kHz steps. The VCO covers only a few MHz without readjustment, resulting in low noise. Modulation is performed independently of frequency control, and with special consideration for low noise. The output power is controllable from zero to 4 watts. A PLL unlock detector is included, to shut down the transmitter in the event of a malfunction.
The hearth of the exciter is a Colpitts VCO. It is powered from a local 9V regulator, and has the frequency controlled by two back-to-back varactors, resulting in minimal loading and thus ultra low phase noise. A sample of the VCO signal is divided down by a prescaler IC and applied to a PLL chip, which gets its reference from a custom made quartz crystal and divides it down to 6250 Hz. The frequency is set in binary fashion by a ten-way dip switch, which controls the main programmable divider. If the PLL is unlocked, Q1 switches on an output that should be used to disable the power amplifier. The phase detector output of the PLL chip is filtered and level-shifted by an op amp, to be injected into the frequency control varactors of the VCO.
The modulation signal is applied to a separate varactor, which is biased to run in a reasonably linear range, and being separate from the frequency control circuit, it's not affected by the PLL voltage. All signal and control voltage coupling is done through chokes, instead of inductors, to get lower noise. The bandwidth of the modulation input is wide enough not only for stereo, but also to allow later addition of a utility subcarrier (SCA) signal.
The output of the VCO goes through an emitter follower buffer stage, then through a broadly tuned class A amplifier, followed by a class B driver and a class C power amplifier, which use medium-Q tuned impedance matching networks. These last two stages are powered from a separate input, so that the output power can be controlled from zero to 4 W by adjusting this voltage from zero to 15V. The intention is using this feature for automatic drive control of the final stages, and protection of the transmitter.
Note that the output of this module does not have enough harmonic filtering to connect it directly to an antenna. If you want to use this exciter as a stand-alone low power transmitter, you should add a low pass filter.The exciter is built on a double sided PCB, which has its top side copper left mostly undisturbed as a ground plane. The copper is removed only around non-grounded pins. The ground connections are soldered on the top side, so it's not necessary to have plated-through holes.
This drawing shows the two sides of the PCB, so that you can print it and fold it in the middle to see how the two parts align. You will have to invert the image to print it for making the board, so that the ink gets in contact with the copper.
This PCB is fitted with soldered shields all around and between stages, on both sides of the board. They are best installed before populating it.This image shows the parts layout. Again, you will have to find out which part is which, using the schematic. It should be quite easy. Be careful, because there is one component on the schematic that is NOT included in the board design! It was added later, during debugging, and soldered under the board! To make things more interesting and challenge you a bit, I will NOT tell you which part that is! You will find out when you end up having one part left over after assembling the board! :-) The drawings of the coils are a reasonably close match to their actual sizes.
And this is how the assembled exciter looks! You might notice the machined aluminium part that encloses the output transistor. I made it on my hobby lathe. It's a rather sophisticated way for connecting the TO-5-cased transistor to an external heat sink! A simpler bracket will work as well. My original idea was to stand this module on edge on a chassis or against a cabinet wall, to use that as heat sink. Anyway, the circuit is so efficient that the transistor barely needs an additional heatsink at all! I did all testing without adding anything more than what's shown here.
Many of the parts came from junked equipment. That includes the trimmers and the dipped chokes. But compatible parts are available new. The crystal was made by JAN Crystals. To order it, specify a frequency of 6.4000 MHz, fundamental mode, parallel resonant, 30pF load capacitance, HC-49 holder, with standard temperature, stability and tolerance ratings.
The output is connected via a BNC socket. All other connections go through feedtrough capacitors. The shield is completed by push-on covers, made of the same material used for the shield walls shown here. It's nothing else than coffee tin cans, cut open and flattened! Some chocolates and cookies also come in suitable cans!
Alignment of this circuit is not difficult. First you set all trimmers to mid range and program the frequency. For this task, you simply add the switch weights: The least significant switch produces 100kHz, the second adds 200kHz, the next 400kHz, and so on, until the eighth, which adds 12.8 MHz. The ninth actually connects to two inputs of the PLL chip, so it adds 76.8 MHz, with the tenth switch adding 102.4MHz. To calculate switch settings for a given frequency, you simply decompose it into its binary components, and set the proper switches. Note that a switch that is ON is NOT adding its frequency contribution! For example, if you want to transmit on 96.5 MHz, you would set switches 9, 8, 7, 3, and 1 to OFF, the others to ON. The full range of frequencies you can set in the synthesizer covers the entire FM broadcast band and quite a bit more, but the rest of the circuit was designed only for the broadcast band.
Now you should connect a 15V power supply to the main power input only, with a voltmeter at the output of U3, and a frequency counter at the collector of Q4. If you get the correct frequency, you are in big luck and should go and play lottery! Usually the VCO will be out of capture range. If the voltmeter reads around 14V, it means the frequency is too low. If it reads close to zero, it means the frequency is too high. The frequency counter should agree with this. You need to adjust the VCO center frequency to bring it into range. For this task you have two adjustment points: One is C20, the other is bending L4! Usually the trimmer alone does not give enough range, so feel free to bend the coil. When you have adjusted the VCO roughly right, the PLL will lock in, and you will get a stable output frequency, very close to the one you want. Adjust L4 and C20 so that the voltmeter reads roughly 9V. Such a relatively high varactor voltage is convenient for best noise performance, because it keeps the varactors from entering conduction at the RF peaks . Ideally you should adjust the coil so that the trimmer is near center range with the voltage at 9V. This gives you easiest correction later.
Now you can set the reference crystal to the precise frequency, by adjusting C12 so that the frequency on the counter is exactly the correct one.
Let's go to the power stages: Connect an RF power meter and a 50 ohm dummy load to the output, and apply a few volts to the variable voltage input. Adjust C28, C32, C37 and C38 for highest power. If you run out of range in any trimmer, correct that by bending the coils connected to it: L5, L7, L11, L10. Now increase the voltage and retouch these trimmers. You should get 4 to 5 watts output at 15V of supply voltage.
To avoid microphonic noises, after completing the adjustment you should seal the oscillator coil, and maybe too the other air wound coils, with bees wax or some other suitable material. Slight readjustment of the trimmers might be needed after that.
Now you can connect the audio board to the exciter. Apply a 1kHz signal to the audio board (both channels is best), strong enough to drive the board into moderate limiting, and adjust R68 on the audio port to get +/- 75kHz deviation. If you don't have a deviation meter, you can get close by hooking a scope to the audio output of an FM receiver, tuning it to several local stations, note the audio levels produced by them, and then tune to your transmitter and set its deviation to match that level. But this system is very imprecise. It's best to get or make a real deviation meter.
If you ever want to change the frequency, you have to reprogram the dip switches and then retouch all trimmers, and possibly the coils, except for C12, which should only require retouching after several years, when the crystal has aged.http://ludens.cl/Electron/fmtx/fmtx.html