Monday, February 14, 2011

Simple beacon for 1296 MHz - S53MV design

Not so many beacon articles can be found on the web where everything is well explained and documented regarding the home-brew beacon design for the microwave frequencies. One of the authors who really care about it is Matjaz Vidmar S53MV. All his projects are well described, proven in practice, and all assembled from the cheap and wide spread electronic parts. The most important, they are working !

Between all of his numerous design, the beacon for the 23 cm band is one of the simplest microwave project. The project was published years ago and can be found in the script "Beacon 99" (PDF format) on the following address: .
The design is straight forward, 648 MHz oscillator and 1296 MHz multiplier with small amplifier. All beacon was assembled on only two 0.8mm FR4 PCB. The oscillator PCB is single side and the 1296 MHz doubler PCB is double side with one side used for the ground. If you strictly follow the design and instructions there should be no mistake. The 1296 MHz multiplier is a no-tune design, so plug and pray. The 648 MHz oscillator board consists of the chain of multipliers where each one should be tuned to the exact frequency to obtain the good signal on the output (10dbm).

The multiplier chain is simple :
Oscillator 27 MHz
x2 multiplier 54 MHz
x3 multiplier 162 MHz
x2 multiplier 324 MHz
x2 multiplier 648 MHz

As I already build the 648 MHz oscillator for the 1296 ZIF SSB TRX (in the same script) I decide to use this one for the beacon with the main crystal frequency on 18 MHz. The signal on 648 MHz was stable and the power was higher than 10 dBm. Frequency can be adjusted through the multi turn trimmer feeding the voltage to the varactor diode. With the cheap computer grade 18 MHz crystal the range 1295.900 - 1296.100 MHz was obtained with no problem and stable output power and frequency. Attention should be used when tuning the trimmer capacitors. Each multiplier stage, from the oscillator upward, should be tuned to the maximum power at the same time controlling the frequency of the multiplier stage that we are tuning. That way no mistake can be done and you should end with the correct output frequency.

The 1296 MHz multiplier board is a no tune, so after connecting the power you should have 23 dbm (200mW) output power. Also here, I did not follow the design because that time I did not have all parts ready. So I apply some modifications:
Instead of the 1st ATF35376 (multiplier) a MMIC SNA386 was used, modifying also the bias network. Both gate resistors are removed together with the LED diode with 330 ohm resistor and 10 ohms resistor was replaced with the 240 ohm. Standard MMIC multiplier configuration was used to double the frequency from the 648 to 1296 MHz. There was enough drive power to saturate SNA386 and to generate strong harmonic on 1296 MHz.

Straight after the 4 pole strip line filter the first amplifier ATF35376 was again replaced but now with MMIC ERA-2. Gate resistor was removed and MMIC bias was adapted for the 12 volt supply. 4V7 diode was removed together with 180 ohm resistor and 10 ohm bias resistors replaced with 220 ohm.

Of course, to complete the modification, the last amplifier BFP196 was replaced with the BFG540x transistor amplifying the signal to the +23dbm output with the power supply of 12 volts.

Just a two things to take care when building this multiplier, the grounding MMIC and the output transistor via are just 3.2 mm holes closed from the ground side by thin copper foil. Tho holes are then filled up with the hot melted tinol creating good contact with the copper foil. Four filter stubs are grounded using the 0.6mm silver wire (one wire from the RG-214 center conductor). Bend the wire creating the C shape to allow good connection (2mm) on he top and the ground side of the PCB. Tight the wire with the pliers and solder the wire both sides.

A short troubleshooting list if you run in the problems:
- BFX89 transistors should be mounted as close as possible to the PCB, not more than 1 mm above it
- BFX89 can be replaced with the BFY90 with better results
- BFR91 transistor is soldered directly from the copper side
- 22 nF blocking capacitors a very important, use the good quality ceramic capacitors
- keep all connecting leads as short as possible, a must for the capacitors
- the coils L4 and L5 are identical, pay attention on direction
- use only high quality trimmer capacitors in the multiplier chain
- up to now I build several oscillator boards for the frequencies of 576, 648, 720, 864 MHz and the only problem I had was the trimmer capacitor. After some time, the low quality trimmers are changing their value. The tuning point may be quite sharp on the second and the third stage, so this can cause that your oscillator after some time stop working, more often if your beacon is located outside, on the mast close to the antenna.

The proper shielding is required as per article instructions and the beacon can be mounted close to the antenna to avoid the loss in the coaxial cable. The keying is done by means of switching the power supply to the 648/1296 MHz stage. The oscillator board is always powered to achieve the better frequency stability.

I prefer to have the beacon not exposed to the atmospherics. The frequency stability is also less affected by the temperature if the beacon is held indoor. Of course, the loos in the coaxial cable have to be covered by extra amplifier to have the same power on the antenna. My beacon was built around standard 1H 17" rack together with the power supply and the IK0WRB beacon keyer. The keyed voltage is little bit lower than 12 volts. This result with the lower power output (100mW), but also the final transistor lower temperature. In the rack there is enough place for the extra amplifier and the inter-digital filter if required.

I was surprised with the final frequency stability and the shape/sound of the signal. On the video, you can notice the second harmonic frequency received on the FT-817d accompanied with the W1GHZ 23cm transverter (high side injection). Keyed signal can not be noticed due to 50 ohm termination, and the frequency stability is not affected by the keying of the final stage. At the same time notice the power output on the power meter. The beacon was left on the bench for a couple of days connected to the dummy load. No significant change of frequency was noted nor the lower output power.

After the initial beacon connected to the dummy load test, there was time to perform the "live" test running the beacon connected to the real antenna. The first antenna that I have handy and ready was the "Cheap WA5VJB" style yagi antenna. With some 10 dBi gain and indoor, connected to the beacon with a peace of RG-214 cable the antenna was pointed to the nearest mountain (1400 m ASL) from my home location. The idea was to use the already well known reflection point on the top of the mountain so the beacon can be heard in the near Italian provinces. The signal was good and stable, with no significant frequency drift. After several initial reports the beacon was switched off, ready to be mounted on the remote location.

The beacon is planned to be mounted indoor, so quite long run of the coaxial cable will result the low power on the antenna with just a little bit more than 100 mW of the beacon output power. For that reason I add another 10 db amplifier stage at the end easy producing more than 1 watt of output power. The amplifier was mounted in the robust CNC machined aluminum housing attached to the bottom of the rack.

The final result can be heard on the following video clip. Of course, 9A4QV/b signal is the strongest one between the two others.

Thursday, February 3, 2011

W1GHZ 1296 MHz Hairpin filter

Building the Rover transverter for the 1296 MHz I experience the lower power output than expected, so I start looking what can cause this 10 dB lower output than predicted. Not only on the TX side, I notice the lower signal also on the RX side comparing the received signal with the other transverter design with the similar conversion gain. Among the all other parts affecting the signal loss, the filter is the only one that was designed and not fabricated. Paul, the author already made a good job with all relevant measurements on the filter, already published, but I wanted  some other verifications to get some more data to be able to use the same filter in some other projects.

The filter was duplicated from the W1GHZ 1296 MHz transverter (3 pole) on the standard 1.6mm FR-4 laminate. I prefer to use the "toner transfer" technique with my flat iron. Iron already became the standard tool on my workbench :-) A pair of SMA connectors are added to be able to measure and compare the filter more easily. Of course using the SMA connectors bring some additional losses to overall measurements, but I can live with that. The filter itself in the transverter configuration will have even lower I.L. 

IMPORTANT !!! The following measurements are done with the reference signal of -12.72 dBm !!!

So the reference signal was not 0 dBm, but -12.72 dBm and it is necessary to correct all the readings that you see on the following spectrum analyser screen shots. I start with the center frequency (C.F.) 1296 MHz, span 500 MHz to see quickly if my flat iron was successful or not :-) Quick look on the curve and everything looks as predicted. The filter is centered around the 1296 Mhz with the I.L of -4.53 dB (17.25dBm -12.72dBm). Looking the screen the BW - 3dB is 100 MHz.

The next step was to measure the filter response 144 MHz from the center frequency. At 1152 Mhz the signal was 40.28 dB down as per screen shot. I forget to measure the signal on 1440 MHz but we can see that the skirt is not so good as on the lower side, so the signal is only 23.28 dB down on the 1440 MHz. Of course "Right side up" mixing approach should work better than using the higher L.O. frequency in this case.

As I plan to use the same transverter but with the I.F. 432 MHz I measure the filter response on the 864  MHz with signal  62.28 dB down. The center frequency is 1296 MHz and the span is 900 MHz, 10dB/div. Despite the mixing products problem using the 432 MHz for the I.F. transverter is working quite good with this filter.

Using the same filter with the I.F. of 50 MHz to work on 1296 MHz is almost impossible. The signal on 1246 MHz is down only 11.28 dB. This measurement was just to compare this filter skirt with some other filters measured before. So here is the screenshot.

When measured the filter bandpass and the loss, I was still curious about the filter return loss / SWR so additional test was performed using the network analyser. This measurement gave me quite good idea about  the characteristics where the minimum obtained R.L. was 16.29 dB or the SWR of 1:1.36. This way I was quite sure that the mixer RF port is terminated with the impedance close to 50 ohms.

At the end, I tried to tune the filter for even better RL, what was possible but on the other side affecting the pass-band characteristic. This filter is designed as a no tune project and I was happy with the above presented results. They correlate much with all already published by Paul, W1GHZ. The small differences are result of  maybe different FR-4 used, slightly different dielectric constant, or the fact that my filter was not covered - galvanised. Non uniform Er can also affect the results where FR-4 has quite loss on microwave frequencies. Of course, the "Right side up" version of the transverter is using the 4-section hair pin filter with little bit different characteristics, but why not believing to Paul who already made all relevant measurements.

To summarise:

Freq:   1296   1152   1440   1246   864     MHz
Loss:   4.53   40.28   23.28  11.28  62.28  dB