This circuit was originally outlined in a presentation at the 1996 AMSAT-UK Colloquium held at the University of Surrey and printed in the October 1996 issue of AMSAT-UKs Oscar News.
A recent survey of satellite users world-wide revealed that only 3 percent were active on mode S. With the re-entry of Oscar 13, we will have now sadly seen the end of a mode S transponder that has given years of reliable service and to one of the most appreciated and successful experiments to be carried on an amateur satellite. For this 3 percent, operating mode S meant working at the cutting edge of satellite communications. It is well to remember that the S-Band transmitter on AO-13 was usually run at a power of less than 1 Watt into its small 8 turn helix. When working mode S on AO-13 a typical receive station could consist of a small 0.6 metre dish but with a high performance receive converter having a noise figure of around 0.6dB. With this system it was very easy to receive the satellites signals over distances of 35,000km.
With Phase 3D, the S band transmitter will operate at powers in excess of 40 Watts and with an antenna gain at least 6dB, or four times, greater than Oscar 13.
The receive converter described here, together with a small helix around 18 inches long, will give excellent results with the new satellite.
* System Noise figure of less than 2.0dB
* Conversion gain of 20dB
* Low cost consistent with ease of construction
* One PCB with low component count
* Robust and easy to reproduce
* Easy to align. i.e. No complicated test equipment and less than 15 minutes!
The final result of this process has been to produce a receive converter which I believe meets the design criteria. Testing of a prototype on a Hewlett Packard 8970A Noise Figure meter showed a system noise figure of 1.85dB with a conversion gain of 32dB.
Modern Microwave Components
The last few years have seen a rapid growth in UHF and microwave telecommunications. The expansion of this industry, which is rightly seen as a threat to our amateur allocations, has also led to the development and mass production of some very useful RF devices. The trend in current microwave design is to use components that are internally impedance-matched to 50 Ohms. Our circuit maximises the use of these easy to use and reliable building blocks. Gain is provided by Monolithic Microwave Integrated Circuits (MMICs). The mixer used has Schottky diodes and matching transformers mounted inside its ceramic case and the filter has four connections: input, output and two to ground. A degree in plumbing is definitely not required
The first amplifier used at the antenna input largely determines the overall sensitivity of a receiver. The Low Noise Amplifier (LNA) used here was introduced by Hewlett Packard in 1996. The MGA86563 is a high gain LNA usable within the 500 MHz to 6 GHz frequency range. It is a three stage GaAs FET MMIC constructed using State of the Art PHEMT technology. The device requires a nominal 5 Volt supply at 14mA, but unlike other similar products, can operate from higher voltages thanks to an internal bias regulator.
The converter is constructed on a standard 62 mil (1.58mm) double sided FR4 fibreglass PCB. While this material is not the best choice for use at this frequency, it is more robust and easier to handle than boards using very thin PTFE substrates.
I also decided early on in this project to use surface mount components. I realise that this may discourage some potential constructors but there are two indisputable reasons for doing so. First, I found that the Local Oscillator section could be built using four transistors using Surface Mount Devices (SMD) but needed an extra stage if I used conventional components. Second, there are many opportunities for leaded components to be incorrectly inserted in the PCB. Long wires normally equal a non functioning circuit at S band. With a surface mount component, if its in the right place and soldered correctly, then it will work. As a compromise, I have used large SMD for ease of handling.Circuit Description
The Local Oscillator is based on an excellent design by Sam Jewell G4DDK. Sams original circuit has been adapted in this application for surface mount construction and optimised for a frequency range of approximately 1700 to 2000 MHz. Crystal X1 is a 5th overtone unit with a frequency of 109.333 MHz. Transistor TR1 and TR2 form a Butler oscillator with L1/C1 tuned to the crystal frequency. The output at the collector of the limiting amplifier TR2 contains a high level of harmonics and tuned circuits L2/VC1 and L3/VC2 select the third harmonic at 328 MHz. TR1 and the base of TR2 operate from a 9 Volt supply produced by a 100mA voltage regulator IC5. TR3 is a frequency multiplier which triples the frequency from 328 to 984 MHz. The input is coupled to the base of TR3 by C10. The value of C10 is only 4.7pF which allows enough signal through to drive TR3, but is small enough in value not to degrade the Q factor of L3/VC2, and to provide some rejection of any residual 109 MHz. R9 provides a constant load for the drive and sets the bias point for TR3 ensuring the generation of harmonics.
The output of TR3 is a two stage bandpass filter constructed using microstrip techniques. Z1 and Z2 are inductively coupled, shortened quarter wavelength lines, etched onto the printed circuit board. Each line is tuned by a variable capacitor (the high impedance end) and is AC-coupled to ground at the other (low impedance) end. The transistor is impedance matched into the filter simply by tapping onto Z1 at the appropriate point.
The final stage in the Local Oscillator is a frequency doubler, from 984 to the output frequency of 1968 MHz. The input is coupled to the base of TR4 by C13. The transistor is biased to class B by applying approximately 0.6 Volts to the base produced by potential divider R11, R12. The collector supply is fed to the transistor via stripline Z4 with RF decoupling performed by C14 and C15. During development I found that the standard 1nF chip capacitors were not very effective at decoupling frequencies above 1 GHz. The addition of the porcelain American Technical Ceramics capacitor C14 increased power output by 50 percent. A further increase was achieved by including the quarterwave high impedance transmission line Z3.
The output at 1968 MHz is taken from the collector and applied to another microstrip bandpass filter. A three section filter was selected to allow effective filtering of the local oscillator. Any spurious signals will cause unwanted products to appear at the output of the converter. The filter Z4 Z5 and Z6, provides a clean output of 5mW with unwanted products suppressed by at least 42dB. Each stripline is tuned by a variable capacitor as in the previous stage, but here the type of trimmer chosen is important. The green SKY trimmers (1) specified have a maximum value of 5pF, but more importantly, have a minimum capacitance of not greater than 0.75pF. This will limit the choice to foil or ceramic types which are physically small, or to piston trimmers as marketed by Down East Microwave in New Jersey.
The RF Section
The entire RF section is constructed from modern 50 Ohm block components. The amplifiers, filter and mixer are all internally matched to 50 Ohms and consequently there are no adjustments to be made. The broadband characteristics of the devices used mean that this converter receives all frequencies from 2400 to 2500 MHz and converts them to 432 to 532 MHz.
The input at 2400 MHz is passed via a low loss ATC capacitor C18 and matching inductor L4 to the L.N.A. IC1. Supplying voltage to the LNA requires care as it is only conditionally stable at some frequencies. My final solution was to use RF choke L5 with a series resistor R15 to reduce the Q and ensure stability. Zener diode ZD1 drops the 12 Volt supply to 6.4 Volts which is within the limits of IC1. The LNA output feeds bandpass filter F1. This filter has a bandwidth of 100 MHz and is manufactured from the same type of ceramics technology found in dielectric resonant oscillators. The centre frequency is 2450 MHz providing an ideal response for S band satellite operation. The manufacturers data sheet shows a mid-band insertion loss of 1.16dB with a L.O. rejection of -35dB. Image frequency response is off the graph supplied by Toko but in reality is likely to be limited by coupling between PCB tracks. IC2 is a standard MAR6 MMIC which has a gain of 10dB and a noise figure of 4dB. This feeds a Mini Circuits RMS30 double balanced mixer. The RMS30 is specified to 3 GHz and has internal matching on all ports. The output is taken from the I.F. port to a -3dB resistive attenuator. This was included because the input of the 432 MHz amplifier IC3 may not be 50 Ohms at all frequencies emerging from the mixer.
The Circuit Diagram
The PC Board (108 x 34mm) Trackside.
109 328 984 1968 MHz
IC1 * MGA86563 IC5 * 78L09 TR1 / 2 PMBTH10
IC2 MAR6 - Mini Circuits IC3 RMS30 TR3 / 4 BFR93A
IC4 MAR3 ZD1 5V6 400mW
L1 * 4.5 turn Toko S18 with aluminium core
L2 L3 * 3 turns 22 SWG tinned copper. 4mm inside diameter. 2.5mm off PCB
L4 * 22 SWG tinned copper formed into U shape with I.D. 4mm. 5mm high.
L5 Toko SM inductor 150nH type 32CS
C1, 6, 7 22pF * C4, 8, 9 4.7nF plate ceramic
C2, 5, 11, 16, 17, 19, 20 1nF * C12, 15 1nF
C21, 24, 27, 29, 30 1nF C22, 28 0.1uF
C3, 23, 25 27pF 0805 * C26 10uF Tant
C10, 13 4.7pF * VC5, 6, 7 Sky 0.5 - 5pF
C14, 18 12pF ATC 0.1 inch
* VC1, 2, 3, 4 10pF trimmer 5mm dia - Philips.
R1,3, 6 1k0 R9, 12 2.2k R16 680R
R2 820R R20 220R R17, 19 330R
R4, 5, 8 470R R11 22k R18 18R
R7, 14 21 10R R10, 13, 15 47R
F1 Toko Chip dielectric filter TDF2A-2450-10
X1 Crystal 109.3333 HC18U Series resonant
PCB with etched microstrip elements 108 x 36mm
Construction and Testing
The PCB is double sided with the topside used as a groundplane for all the earth-grounded connections. Some components are mounted on the GP side and these are identified in the parts list. The trackside layout with its component overlay is presented in Fig.2. If you plan to mount the PCB into one of the popular tin plate boxes (1) trim it to size before you fit the components. Any flexing of the PCB after assembly may crack the surface mount devices.
(All through board earth connections on the PCB are made via plated through connections) The components can be fitted in any order, but leave the LNA and L4 until last. The best results are obtained using a minimum quantity of solder. If you use too much, the excess, can be removed with desolder braid.
The LNA is only manufactured in a miniature SM package. I found that it can be soldered successfully if you hold it in the correct position and then solder one of its ground connections. With the device in place, look very carefully at its position. If its not correct then reheat the joint and move the device. When you are totally satisfied that the LNA is in the right place, solder the remaining five leads. It is almost impossible to reposition the device once all six leads have been soldered.
The alignment process is in two stages. You will need an analogue multimeter, a 70cm receiver and a 2.4 GHz signal source - more on this later. Start by pre-setting the variable components as follows. Adjust the core of L1 to be level with the top of the former and then turn the core into the former by another 2 full turns. Set VC1/2 to be 40 percent meshed VC3/4 to 15 percent meshed and VC5, 6 and 7 to be 5 percent meshed. Support the board off the work surface so that the microstrip lines are not detuned and connect 12 Volts. The current should be 60-70mA before tuning and about 130mA when alignment is complete. First, check that the crystal is oscillating by listening for the fourth harmonic near 438 MHz on the 70cm receiver. If no signal can be heard, then adjust L1. With the oscillator running we can now align the three multiplier stages. Set the meter to read 1 Volt full scale and place the probe on the emitter of TR3. With no drive, the Voltage will be zero. Using a trim tool, adjust VC1 and VC2. As you tune the circuits to 328 MHz, TR3 will begin to conduct. Just tune for maximum emitter voltage -its as simple as that! Next, move the meter probe to the emitter of TR4. The voltage should be around 100mV due to the bias resistors on the base. Repeat the tuning process, this time adjusting VC3 and VC4 to give maximum emitter voltage. When correctly tuned, the voltage should increase to over 750mV. One word of caution is necessary here. TR3 is designed as a frequency tripler from 328 to 984 MHz. However, it is possible to tune the striplines to 656 MHz by mistake. Fortunately, this is fairly obvious as the trimmers will be 50 percent meshed at that frequency. The alignment process for the final doubler is a little different as we have now run out of emitters! To adjust the last stage, connect an antenna to the converters input and tour 70cm rig to the I.F. output. At this point, youll need to generate a weak test signal on 2400 MHz. The AMSAT-UK signal source (2) is ideal but your S band converter is very sensitive and will easily pick a harmonic from a VHF / UHF source. Failing that, a few hundred miliwatts of 28 MHz applied to a signal diode and series 50 Ohm resistor will suffice. Place the test source 6 feet away, switch on the receiver and select SSB. As you apply 12 Volts to the converter the noise level from the receiver will increase. Locate the test signal and note the S meter reading. The final three trimmers will all resonate close to minimum capacitance. Adjust each one for maximum S meter reading. This indicates minimum conversion loss in the mixer and completes the alignment.
A tinplate box is available for the converter and is recommended as it has a removable lid and base, giving excellent access. Being tinplate, the PCB can be soldered to the sides of the box along with the RF and power connectors. An alternative solution is to fit the PCB into a diecast aluminium enclosure. This will be much easier to waterproof but do keep the box as small as practical.
The finished converter should be installed as close to the antenna as possible, as coaxial cables have a very high losses at 2.4 GHz. The I.F. output at 70cm can, however, be run back to the shack via long lengths of cable without problem. Because the gain of the converter is 32dB, the coaxial cable can lose up to 10dB without noticeably affecting overall performance. Finally, as most units will be mounted outdoors, local oscillator stability should be mentioned. The L.O. frequency can be measured at all times by listening to the crystals 4th harmonic on 438 MHz. In most cases any drift in the L.O. will be minimal compared to the Doppler shift from the satellite. However, a small 12 V clip-on crystal heater is available (3) and will hold the frequency to better than 1 kHz during cold weather.
1) Tinplate boxes model 7754 and 5pF SKY trimmers are distributed in the UK by Piper Communications. 4, Severn Road, Chilton, Didcot, Oxfordshire. OX11 0PW U.K. Tel. 01235 834328.
2) A Low Cost Signal Source for 2.4 GHz. AMSAT-UK Oscar News No 112 .
3) Crystal Heaters / SKY trimmers.
Microwave Component Service. C/O Ms. P. Suckling. 314a Newton Road, Rushden, Northants.
NN10 0SY U.K.
End of article
A kit of parts is available for this project; It includes a PCB with plated
through holes, the box from piper communications, all components to complete the converter
board, a copy of the article as published by AMSAT in the USA together with some
additional construction notes.
A donation to Amsat-UK is made for each kit sold.
Price £ 70.00 Including recorded delivery to the UK
and small packet air mail to addresses outside the UK.
International Recorded Europe add £2.00
38 Wyndham Crescent
England TW4 5HZ Tel / Fax. +44 (0)20 8572 8615
Notes and comments.
1) To achieve the best noise figure, the wire loop matching inductor for the MGA86563 should be bent from its normal vertical position towards the ground plane. An angle of 20 to 30 degrees from the groundplane works best.
- G4DDK / G0MRF