The Axino mini-FM transmitterSeptember 2016
This small project arose from my earlier (2014) attempt at making an FM transmitter and which, for the reasons of expediency, was based around a MAXPRO3000+ transmitter board from pcs-electronics. You may read the full technical description of that exercise via this link.
The MAXPRO3000 had some issues; namely RF harmonic levels too high, deviation sensitivity varied considerably with RF frequency and the modulator became unstable with low audio frequencies. The current project is an exercise to see if a simple transmitter could be built without having similar problems.
The technique is hardly ground-breaking and is entirely conventional. A modern approach would most likely comprise a DDS and a lot of programming, but I considered that too steep a learning curve at this stage, so stayed with the simple VCO and single loop synthesizer approach.
The VCO itself is the Mini-Circuits POS-150. This covers 75MHz to 150MHz and produces about +9dBm RF output. To minimise pulling, the VCO feeds a fixed 3dB attenuator and is then followed by a variable attenuator, so that the nett module RF level can be varied from -6dBm to +6dBm. An NEC divide by 10 prescaler supplies the synthesizer board. An adjustable regulator supplies a stable and low noise 12.6Vdc.
The synthesizer board (below) is based around the venerable Motorola MC145151P chip. A bank of 11 DIP switches set the main divider ratio. Another bank of 3 switches sets the R divider. Step size on the FM band is 100kHz and the prescaler is ÷10, making the reference frequency 10kHz. Using a 10.24MHz reference crystal, the R divider is set to 1024.
An op-amp loop filter and audio buffer feed the VCO.
A mute facility is important so that the transmitter can be immediately shut down if the synthesizer loop is out of lock. The LD pin of the MC145151P is high when the loop is locked but low going pulses appear if it becomes unlocked. A comparator stage is arranged to go low for an unlocked condition. The mute line disables the regulator in the subsequent power amplifier.
An unlocked PLL is not the only reason to mute the transmitter. The VCO could produce a locked output outside the FM band by either mis-setting of the divider switches, or by a fault in the synthesizer. Of particular importance is to avoid producing output in the aviation frequencies which are directly above the FM band. This VCO is capable of operating up to 150MHz. To avoid these issues, a window comparator is set so that if the VCO tuning voltage goes either above that which produces 108MHz or below 87.6MHz, the mute line is also pulled low.
1 watt power stage
This is a straight-forward two stage amplifier capable of up to 1 watt if the supply rail is above 16V. A low-dropout regulator feeds the driver stage plus the bias supply to the final stage. The regulator is shut down if the mute line is pulled low. The PA gain drops by 30dB, so although the RF output does not completely stop when the system is muted, it will be producing only 1mW or less.
1 watt is the maximum EIRP allowed under the N.Z LPFM rules. Most broadcasters use a dipole or ground-plane whip, either of which has around 2dBi gain. So, in most cases, the transmitter power can be a maximum of 600mW, assuming a short coax run. The power of this transmitter may be manually set down to around 120mW using the attenuator on the VCO board.
A low pass filter at the output assures that the harmonics are at least 56dB below the output carrier power. In addition, rudimentary coupled lines with diode detectors are added to provide rough indications of both forward and reflected power.
The occupied spectrum mask on the RH spectrum analyser pic is from the N.Z LPFM regulations. For this, the stereo encoder (section follows) is in use. The transmitter/P.A meets all the LPFM performance requirements. The modulation sensitivity varies only 0.25dB between low end of the FM band and the high end. In these respects it is much better than the MAXPRO3000+ used previously. The synthesizer locks quickly and the 10kHz sidebands are at least 60dB below carrier. Since there is no ALC for the RF output, power does start about 1dB low from cold and takes 10-15 minutes to stabilise. Modulation bandwidth (3dB) of the MCL POS-150 VCO is stated as being 100kHz, which is only just enough. Measurements indicated that the response with 53kHz modulation applied is around 0.7dB down. This can be expected to slightly compromise stereo separation at the higher audio frequencies.
A simple metering display was built using a 1.8" TFT and an Arduino Pro mini. A small 5V dc-dc converter provides the 5V rail. There are no control functions; only monitoring.
There are four analog metering points; forward power and reflected power plus deviation and supply bus. Power indications are fed from the diode detectors on the P.A board. Deviation is derived from a peak detector, fed from the MPX input to the modulator board. In addition there is a signal from the P.A mute line which facilitates an enabled/muted indicator and there is a connection from the N10 DIP switch of the synthesizer. This produces a high band/low band signal to alter the power scaling accordingly and provide a small indicator on the display. In theory all 11 synth divider switches could have been brought out and decoded into an actual frequency readout, but a paper chart adjacent to the DIP switches saves a lot of wiring and programming. The NZ LPFM regulations limit possible frequencies to a group of nine at the very bottom end of the band and eleven at the very top, so it was decided to simply signal low or high band operation by use of the N10 DIP switch.
Voltage from the diode RF detectors has a square law relationship to power, so a plot was made of power vs voltage and an equation established, which is calculated by the processor in order to display power. A different set of factors applies between low band and high band operation and the processor takes account of these. A small dot on the top RH corner is red for low band and blue for when a high band channel is in use. The forward power readout is reasonably accurate, but reflected power is more problematical. This is due to the poor directivity of my RF probes, which means that much of the reflected signal voltage is actually from the forward power. Significant 'reflected' voltage exists even for a very good 50 ohm load. Although an offset and a scale factor are applied, the change of voltage is small over a large range of 'good' loads; i.e when the reflected power is small, so because of this it was elected to not display actual reflected power, but simply scale reflected power into good, moderate and poor categories and call it antenna merit. This is calculated from reflected/forward, so at least gives a guide to the quality of the antenna.
To display deviation, I had initially wanted a bargraph along the bottom, which would refresh quickly enough to capture fast peaks. Alas my Arduino programming skills can only be classed as rudimentary, and I could not create a slow enough refresh to display variable character fields clearly while simultaneously providing a rapid rewrite of the bar. So, I have reverted to numerical display of deviation which updates 5 times a second, plus an red 'ovr' flag if deviation exceeds 75kHz.
Aareff Stereo Coder
A stereo encoder was now necessary, but it was decided to buy a kit. I purchased the Aareff SCM kit. We will leave out the fact that it took over 10 weeks to arrive, taking considerable dialogue with both the supplier and NZ Post, and not even mention that the day after it did arrive; the packaging somewhat beaten up, NZ Post told me it was in Moscow!
The kit has a nicely made pcb; double-side, ground-plane on top, with plated-through holes. There were no assembly instruction, so these were requested and emailed from Aareff. All listed electronic components were present and correct. The board clearly has provision for RCA pcb sockets, which were not supplied; instead the I/O is also on board headers, which were supplied. The jumpers necessary for selection of pre-emphasis curve were not supplied. All the fixed resistors appear to be carbon-film. I prefer metal-film types for their much lower temperature coefficient and would use carbon-film types now only for LED dropping resistors and other non-critical tasks. Aareff also use nasty skeleton carbon trimpots at the audio input; one per channel, which seems odd, since any incorrect setting between left and right will result in a significant loss of separation.
Assembly was easy. An low-dropout regulator was added since my raw dc rail is 16.6V and the board requires 13.8V.
Power-up and testing
The dc drain was 22mA inclusive of the added regulator. The board worked first time and appeared to meet the basic specs given by the manufacturer. The essential architecture is of the switching type, where left and right channels are alternately switched to the output at 38kHz, using a cmos bilateral switch. The pilot is summed with the M+S channel following the switch. All frequencies are derived from a 4.864MHz crystal using a CD4060 chip. The circuit is designed for audio input levels of 0.775V rms and with slight adjustments of the input level trimpots, produce an MPX output of 0.82V when the pilot ratio is correct. Neither the MPX output level, nor the pilot level are adjustable, so your modulator sensitivity must meet this requirement. It was a happy coincidence that my modulator needed precisely this level.
The 38kHz rejection was 57dB while separation was 54dB at low frequencies worsening to 17dB at 15kHz.
Performance limitations and modification
While the Aareff coder functions adequately well and meets essential specifications, it is not that precise. It does not incorporate audio low pass filters, neither does it have a 19kHz notch filters. These functions would be needed externally ahead of the coder board. Audio distortion was initially higher than expected. I measured THD of 1kHz at 0.26%; not too bad you would think, but 3rd harmonic was highest, making distortion subjectively worse than the figures would indicate. I noted that prior to the 4016 cmos switch, distortion was much lower. I simply replaced the 4016 with a 4066 chip, and the THD immediately reduced to 0.035% at 1kHz.
This pic shows the MPX output after substituting the cmos bilateral switch. Input is L = R = 1kHz at 0.775V rms per channel. The spectrum illustrates the now low THD, correct pilot level, good 38kHz rejection and separation. Another check by oscilloscope (not shown) confirmed the pilot phase is correct.
Following the cmos switch is an op-amp low-pass filter/buffer which reduces the 114kHz (3*38kHz) component. This it does to the extent of 10dB, which actually proved to be sufficient. However, the filter rolls off 53kHz by 2.4dB and also has a phase shift of 23 degrees at 53kHz compared to low frequencies. This is more than enough to compromise high frequency separation. Ideally a much better MPX buffer and filter are required. The response at 53kHz should be no more than 0.1dB down and with less than 3 degrees of phase shift in order to not compromise separation too greatly.
The pre-emphasis network does not produce the full boost required, which could make the sound slightly soft. The circuit has a fixed low pass section in order to avoid boosting supersonic frequencies. The nett effect is to roll off 15kHz by 3dB before pre-emphasis is applied. The 50us curve should be +13.6dB at 15kHz and is actually +10.6dB. I accept this philosophy as being a fair compromise.
No attempt has been made to further modify the Aareff coder other than to substitute the cmos analog switch i.c. The system is working sufficiently well for the application. I would like to add analogue low pass filters prior to the coder but that will have to wait until a low component count sub-system is found for the task.
Make a comment via the contact page