THE "BELTHORN" SSB IF MODULE
|1. Circuit Description||8. DC Switching|
|2. Front End||9. Construction|
|3. Crystal Filter||10. Using the Belthorn SSB Module|
|4. IF Stage||11. On the Air|
|5. Detector/Modulator and CIO||12. And Finally...|
|6. Audio Stages||13. References|
|7. AGC||14. Coil Winding Details|
The idea of a simple, easy to construct SSB IF system is not new. During the 1970's and 80's several designs were published which made use of a new generation of semiconductor IC's and packaged diode ring mixers, ,. Many of these designs used bilateral circuitry where certain stages were used during both during transmit and receive allowing component counts to be kept . Many of these designs were based upon the Plessey SL600 and SL1600 families of IC's which with sensible layout could be assembled "Lego fashion" to produce working IF and audio systems with no requirement for inter-stage matching. These devices are now obsolete and are difficult to find at reasonable prices however, many newer components are now available which with a few design tricks can be used to produce simple designs that perform well.
After constructing several small SSB QRP transceivers I decided to invest some time in developing a simple, modular SSB IF system that could form the basis for many of my future SSB projects. The unit would contain all the circuitry necessary to provide the mixer IF and audio common functions of a single conversion SSB receiver or transceiver. The design criteria was as follows;
The frequency span may seem extreme but with applications from Cave Radio and SCUBA SSB ultrasonic diver voice communications through to HF and VHF Ham Radio, this was the basic design criteria! As in the Plessey and G4CLF designs, versatility was paramount if the module was to be easily adapted to a wide range of ssb applications. By identifying all the stages of a transceiver that are "independent" of the operating frequency and placing them within the module, the familiar configuration shown in Fig 1 was derived.
The development of the circuit followed an empirical approach with lots of bread boarding and experimentation. Before any readers accuse me of plagiarism I must hold my hands up and make it clear that the inspiration for sections of this circuit came from several first rate constructional articles and designs, notably by W7ZOI  and the Plessey designs by G3RZP and G4CLF. With so many good circuit ideas already developed and proven the development of the Belthorn SSB Module became little more than an integration exercise.
The architecture of the IF module is shown in figure 1. Signal flows during transmit and receive are indicated by labelled dashed lines.
The "front end" of the IF module is an SBL-1 diode ring mixer (U1). This gives a potential upper operating frequency limit of 500 MHz. The mixer "DC port" is used for the RF port, this allows operation down to VLF (e.g. for Cave Radio). The two remaining ports are specified for operation above 5 MHz. One is used for the IF port the other for the local oscillator. This arrangement works well, providing that there is sufficient local oscillator drive (at least +10dBm). No degradation in performance is noticeable with local oscillator frequencies down to 2 MHz, a prototype unit used in a 40m transceiver with a 9 MHz IF and a 2 MHz LO works extremely well.
The front end bilateral amplifier was taken straight from the G4CLF design. A J310 (Q1) runs at high current (getting noticeably warm) with gain direction being set by the bias on four signal steering diodes. In the original circuit low capacitance BA 182 switching diodes were used, in this design these are replaced by cheap 1N4148 diodes. The gain of this stage is high enough to provide a good Noise Figure but is not too high so as to cause instability when using 1N4148's diodes for signal path switching.
Whilst a commercial Crystal Filter can be used in this circuit I recommend a "having a go" at building a ladder filter. It can be argued that without adequate test equipment it is difficult to match the performance of a commercial filter, however surprising results can be obtained from carefully constructed simple designs. I've tried building many filters from the various TT articles in Radcom. The most successful and perhaps least scientific approach I've tried is described by Wes Hayward in . The filter shown in the circuit diagram is a six crystal Cohn filter at 10 MHz using stock Farnell Crystals (20 ppm) and 5% 100pF ceramic capacitors. It costs about £7 to build, a fraction of the cost of a commercial filter. I've built two of these, without crystal selection, and swept them on an spectrum analyser - they are superb with >90dB stopband and a 6:60dB shape factor of around 2 providing that they are correctly terminated. With the component values shown this filter design and performance is easily reproduced provided that care is taken to closely match component values. The easiest and quickest way to do this is to use close tolerance parts such as the 20 ppm crystals and 5% or better capacitors. This leaves the terminating impedance as the single most significant performance driver, a low impedance (tens of ohms) will narrow the pass band a high impedance (several hundred ohms) will widen the pass band at the expense of introducing ripple. With the values shown, 220 Ohms should produce good ssb results with a bandwidth of around 2.5kHz. The response of all simple crystal ladder filters is asymmetric with the HF skirt being the steepest. This favours LSB use for best carrier and unwanted sideband suppression. This asymmetry can be reduced by increasing the number of poles (crystals) in the filter. Although I haven't tried it, the filter response plot suggests that satisfactory USB operation will be possible with the filter described by placing the carrier oscillator on the low frequency side of the pass band.
Other crystal filters may be used in this design providing that the circuit is modified to present the correct terminating impedance. Transformer coupling of the filter was used to allow this to be easily done by adjustment of turns ratios. For example, to install a 600 Ohm filter T3 and T4 turns ratios are set to 1:1 to present a unity impedance transform of the R2 and R6 terminating resistors.
This was where the fun started! I wanted to use readily available parts so that this circuit could be easily replicated by anyone with a bit of constructional experience. The prohibitive cost of the SL1612, (my first choice from past designs), is a classic sign of obsolescence. After checking it was confirmed that the device is no longer in production, this drove me to look for an alternative. I calculated that I needed about +30dB gain and an AGC range of around 70dB in this design to give good performance. To complicate matters this had to be bilateral. I had a frustrating time trying to source suitable parts and was surprised at how scarce and limited old faithful's such as dual gate mosfets have become. I noticed that many of the American designs used Motorola MC1350P's and after a bit of looking around found that they are available in the UK through various suppliers including Farnell and JAB Electronics in Birmingham. The MC1350P in the configuration shown (U2) provides about 45 dB of gain and 65 dB of AGC range.
On transmit the MC1350P IF amplifier is bypassed by a simple diode switch connecting the 50 Ohm balanced modulator port to the crystal filter. Matching is provided by T4A. On receive no provision has been made to reverse bias the diodes to improve isolation. With a higher IF gains this would most certainly be necessary but with the gain distribution in this design the "off" resistance of the diodes provided adequate isolation.
The appeal of using one stage to perform two functions is realised nicely here. A second SBL-1 (U3) is used as a product detector on receive and a balanced modulator on transmit. The 50 Ohm RF port is matched into the IF stage by T3. The DC port is used for the audio input and output, R11 provides a 50 Ohm termination.
Several single transistor carrier oscillator designs were tried to keep the component count down. All of these failed to provide sufficient mixer drive characterised by high conversion loss giving poor receive performance and low transmit output level. Additionally, these simple oscillators had tuned output circuits which, when adjusted, interacted with the crystal trimmer setting, shifting oscillator frequency and making start up unreliable. Reluctantly a more conventional approach was adopted which uses a FET Collpits oscillator followed by a simple FET buffer. This slightly more complicated configuration was well worth the compromise on component count giving more than adequate mixer drive, reliable oscillator start up and smooth tuning.
A standard low noise preamplifier design employing an NE5534N (U4) is used to amplify the detected audio by +30dB to drive U5 the LM386N AF Power amplifier and AGC system. AF roll off is provided by C46 to reduce the characteristic wideband demodulated noise produced this type of IF in the absence of "tail end" filtering. No further audio filtering is used other than the low pass LC arrangement at the modulator DC port. A six pole connector (SK3) is used to connect the AF gain control to the circuit board. The pins are configured to provide access to the audio pre amplifier output, the input of the AF power and AGC stages and provide ground connection and +12 volts on receive. This feature allows a proposed optional piggy back audio filter unit to be added (yet to be developed) without modification of the main PCB. Whilst not entirely necessary a variable bandwidth audio filter can be useful in heavy QRM.
During transmit a second NE5534N (U6) with variable gain is used to inject low level audio into the modulator. J1 selects high/low microphone terminating impedance. No switching is needed between the transmit and receive audio stages as the low impedance of the modulator prevents any interaction. I use a cheap dynamic microphone insert from Maplin's mounted inside an old ptt housing. Works a treat and produces beautiful audio.
Having got the working breadboard design this far I couldn't resist the challenge of an AGC system! All the books tell you that an RF AGC scheme is the best. This is true but to get an RF derived AGC system working you need to ensure minimal carrier leakage - both out of the demodulator and through the IF, and develop about +120dB of RF/IF gain to drive the detector diodes at a reasonable signal threshold. Not an easy thing to achieve in a simple design, particularly with stage re-use and bilateral amplifiers. So AF AGC was the only solution. I based the design upon one of Wes Hayward's  which employs full wave detection and is DC coupled. This approach cuts down the number of components needed but has the disadvantage of being susceptible to switching transients causing full AGC to be applied on TX to RX with a resulting annoying delay in the receiver gain recovering. This problem is fixed by a simple discharge circuit which operates during transmit to receive switching. The circuit is built around an LM324 quad op amp package (U7). Two amplifiers, U7a and U7b, are configured as a an audio pre-amplifier and full wave rectifier. The gain is provided by U7a which operates as a non inverting amplifier with gain set by VR3. The DC output bias of the AGC system is set by VR4, this is used to adjust the AGC threshold; the point at which an increasing AGC voltage reduces receiver gain. The output of U7a drives U7b a unity gain inverting amplifier. The outputs of both these amplifiers, which are equal in amplitude but in antiphase, are rectified by D7 and D8 to obtain full wave rectification. The output of the rectifiers charge the AGC capacitor C34 with R54 setting the attack time and R35 the decay time. The design left me with a spare Op Amp (U7d) which I originally had earmarked for a full hang circuit. I used this as a monostable to discharge the AGC capacitor on TX to RX switching. The result is an Audio AGC system with one of the best performances I've come across. With a few extra parts it should be possible to add full hang AGC to U7(D) and incorporate the monostable action.
The T/R switching provides TX and RX rails which apply Vcc to make a stage active and Ground when inactive. This ensures correct operation of the bilateral stages. The DC switching has been deliberately left off the printed circuit board as this will usually form part of the host transceiver's power system.
If you are building this circuit as a "one off" I strongly recommend the use of "Ugly" construction. In this, the circuit is assembled above a piece of copper laminate board with ground connections being made direct to the copper and all other connection being made by component leads or short lengths of tinned copper wire. Some components, such as the diode ring mixers, are best glued to the board on their backs leaving their pins upright (known by the American Hams as "dead bug" construction!) With care and little practice this technique allows the construction of compact and very rigid circuits that are easy to modify - a bonus for experimenters.
A few words on ferrite cores. Many projects have been abandoned because the constructor could not find the precise ferrite components described in the design. In certain high power or very low frequency applications core material and dimensions are important, however for broadband low power HF and general IF signal use generally there is only one critical factor the "initial permeability" ui. This, combined with number of winding turns, sets the winding reactance of the transformer at the frequency it is to be used at. For correct operation this should be at least 4 times the source or terminating impedance of the winding at the lowest frequency of operation. For example a winding being driven from a 50 Ohm source should display a reactance of at least 200 Ohms. The ferrite transformer cores used in my prototypes were made from ferrite beads and small toroid cores that were ready to hand. These are typically 850ui or 43 type material and are ideal for use from LF to at least 30MHz. The type of cores to use in the broadband sections are 0.5" diameter toriods which offer ample magnetic volume for the applications and are easy to wind. If you get really stuck, as I mentioned earlier, 850ui ferrite beads make good "balun" cores. Glue two of these side by side to resemble a pair of binoculars and wind the transformer by threading the wire round through both holes. About three turns are adequate for 50 Ohm use at 10MHz, more turns will be needed for higher impedance's or lower frequencies. Note that the ferrite bead approach is only good for fine wires, up to about 0.28mm, any larger diameter will restrict the number of turns.
For those wishing to make their own PCB's a layout is provided. All ground connections are made to the top earth plane. Note that to save space, the layout has been designed to accommodate vertically mounted radial electrolytic capacitors. Pads shown without holes are intended to be drilled through to the earthplane and soldered to ground with short lengths of wire such as discarded component lead trimmings.
In the prototypes all connections to the board were made through connectors to allow the module to be easily transferred between transceiver projects. Mutli pole PCB latching connectors from Maplin's were used for all DC and AF connections. The local oscillator and RF ports of the mixer were presented on miniature 50 Ohm SMB co-axial connectors. Where it is intended to permanently install the module in a project some cost may be saved by replacing all connectors with direct wiring.
The order of construction for this, or any other project, should be chosen so as to allow testing at various stages completion. Start with the audio stages followed by the product detector, IF, Crystal filter, bilateral amplifier and front end circuitry. This allows "listening" tests to be made as additional stages are added. Using this approach and a bit of cunning the only test equipment needed to get the module working should be a multi-meter and a grid dip oscillator to provide and RF signal source.
The module is easily applied to most SSB receiver or transceiver applications. The configuration shown in Fig 1 shows the peripheral stages and hardware needed to make a simple transceiver. A few design considerations must be made when applying the module to a specific transceiver design;
If a crystal ladder filter is used the inherent asymmetric performance of the filter favours LSB operation for best filter response. The circuit board layout has been designed to accommodate a single crystal for LSB operation however this does not limit the module to LSB use only, the local oscillator frequency must be chosen to provide the correct sideband; LO+IF=USB, LO-IF=LSB. Alternatively, sideband selection switching could be realised by using two crystals for USB and LSB however with a ladder filter unwanted sideband rejection in USB mode would suffer slight degradation but would probably be acceptable for most uses.
The local oscillator can be a VFO for the lower HF bands, a heterodyned VFO for the upper HF bands or a VXO for the higher HF bands, and with multipliers for 50 MHz and 144 MHz. A slightly more elaborate LO could make use of a synthesiser such as the MC145151P with interpolation between channels being provided by "pulling" the reference oscillator crystal.
On the higher HF bands an RF stage may be needed to improve noise figure and sensitivity. This should provide only enough gain to cause the receiver noise to increase when the transceiver is connected to a resonant antenna. Any more gain will simply degrade the excellent strong signal handling performance of the front end. Whilst the dynamic range will be reduced by the use of an RF amplifier there is compensation in that unwanted local oscillator radiation from mixer leakage will be substantially reduced.
During it's development I've built three variants of this design each of which I've used on 40m with a 2MHz VFO and 5W PA. My only antenna is a half sized G5RV configured as an "inverted V" from the house chimney. Results have been most satisfying, on receive the strong signal performance is excellent and the AGC is very effective. On transmit audio reports are very complimentary with many stations expressing surprise that they are working a QRP station and, (most satisfyingly), that it's home brew! Best Dx so far is Kursk, about 300 miles south of Moscow.
Whether your a seasoned designer or a newcomer to home construction I hope you find this article interesting. I would be most interested to hear from anyone building this design and in particular suggestions or experiences in simplifying it further. My thanks to Bob Edwards G4BBY for researching the references and discouraging me from pursuing my more "wacky" ideas and to Glen Holt G8NOF, for persuading me to write this project up to share with others.
 "An SL600 Series SSB Transceiver" B Comer G3ZVC Radio Communications September 1974
 "PW Helford Transceiver" J Bryant G4CLF Practical Wireless January 1980.
 "Bilateral SSB" V. Aumala OH2CD Radio Communication March 1973
 "A QRP SSB/CW Transceiver for 14MHz" Wes Hayward QST December 1989*
 "Portability" G3TXQ Radio Communication February 1989.
 "Designing and Building Simple Crystal Filters" Wes Hayward W7ZOI QST July 1987*
* Both articles featured in "QRP Classics" published by the ARRL.
14. Coil Winding Details
T1 Pri 3t Sec 10t wound on small ferrite toroid.
T2 1:10 Pri 2t Sec 20t wound on small ferrite toroid.
T3 Pri 10t Sec 7t wound on small ferrite toroid.
T4 Pri 10t Sec 10 turns wound on a small ferrite toroid.
T5 MC 1350 winding -15+15 turns CT T50-2 to resonate at IF with VC1
U3 winding - 3 turns, Q2 base winding - 4 turns