The heart of this transceiver is an ATmega128 microcontroller (MCU). It controls the vast majority of functions within the radio. E. g.: Frequency generation of the 2 DDS systems, audio tone and AGC decay time, T/R-switching, the presets for transmitter gain on the 6 bands independently, display and panel lights etc. etc.
And, due to usage of a parallel interface for the LCD (8 data lines and 4 control lines) an MCU with sufficient ports had to be used.
First I started with the SPI version of the LCD (ILI9341). This LCD has a high resolution of 240×320 dots. Driven by a relatively slow 8-bit controller like an AVR and the LCD driven in serial mode the performance was inferior.
Next I found that the same LCD is also available with a parallel interface. Then called CP11003. This one uses 12 lines (8 data and 4 control lines minimum), which made it mandatory to use an ATMega128 controller. To enhance speed and performance this one is clocked by a 16 MHz crystal. A touchpad is also integrated, but not used in my application.
Source code in C programming language can be downloaded from Github.
The two DDS oscillators are mounted to the side of the cabinet. They are sited close to the microcontroller board to keep leads short.
Right on the left you can see the small dual-tone oscillator for testing and tuning. Next is the AD9834-equipped local oscillator (LO), centered the AD9951 that serves as the VFO. Right the ATmega128, mounted to a 64 lead breakout board can bee spotted behind the varios cables going to and from this section.
The Dual-Tone Oscillator
This one consists of two simple phase-shift audio oscillators. I have introduced this circuit a longer time ago for testing purposes here in this blog.
The capacitors and resistors in the phase-shifting chain have been chosen to put the two different frequencies to values of about 700Hz and 1900Hz, thus they are not harmonically related. A variable resistors allows the user to set the balance between the two signals so that they are equal in voltage.
Two transistors (a PNP-NPN pair) are switched by Pin PB7 from the microcontroller. There is a respective function in the software that activates the transmitter together with this oscillator for comfortable tuning and testing.
The Local Oscillator (LO)
This one again uses the “good old” AD9834, overclocked to 100MHz. I found that some chips from the “grey market” have problems when being overclocked and therefore produce spurious signals. In case this occurs, it is recommended to step back to the clock frequency of 75 MHz which is high enough for the purpose of the LO.
The oscillator comes with an balun output transformer (will reduce spurs!) and a low-pass filter plus a simple amplifier. The latter basically is not necessary because the LO will only have to drive the inputs of SA602 integrated mixer circuits (200mV RMS) used as SSB generator and rx demodulator. I had another mixer type in mind before, that one needed higher voltage. Thus the coupling to PIN6 of SA602 is only via 5.6pF capacitor to avoid overdriving the mixer and improve signal purity. This will be shown later when we are about to discuss receiver and transmitter circuitry.
Here the AD9951 DDS again comes to operation. This one has got a 14-bit DAC which makes it less prone for spurious signals. The clock rate has been pushed to the limit of 400MHz which, according to datasheet, is the max. clock rate for this DDS module.
You can download a datasheet of a suitable clock oscillator. This device is very small but it can be soldered to a 2 by 2 hole piece of veroboard and then mounted to a piece of headerstrip by soldering wires to the underside of the board:
A voltage divider will reduce the 3.3 V to 1.7V that is acceptable for the clock input of the AD9951 chip.
The DDS circuit is common for frequent readers of this blog:
The low pass filter has been left out because when examining the output signal of the DDS it turned out to contain only very little quantum of harmonics. The max. frequency of this VFO will be 29.7 MHz + 9MHz which equals to 38,7 MHz.
An SSB radio for the HF bands will be presented. Featuring 12 to 20 Watts of output power (depending on DC supply), full DDS frequency generation, covering 6 major frequency bands (1.8, 3.5, 7, 14, 21 and 28 MHz) within the short wave amateur radio spectrum. The rig also features colored LCD and front panel backlight.
In this upcoming series of articles a relatively complex project will be discussed. It is some sort of „remake“ of my last multi-band QRP SSB transceiver that has been entitled the „Gimme Five“-Transceiver and that was finished in 2015. „5“ in that case stands for the 5 major (i. e. „classical“) RF bands: 80m, 40, 20m, 15m and 10m the radio covered. This new project (called the „Midi6“, because it is not a “Micro” or a “Mini” transceiver 😉 ) covers one band more, the range has been extended to 160m.
The basic features of this construction are:
Dual DDS frequency generation (AD9951 as VFO, AD9834 as LO),
Colored LCD (CP11003) with resolution 240×320 pixels,
Single conversion superhet receiver, interfrequency 9 MHz,
5 stage high quality transmitter, Pout=20W (max. at 15V DC) , featuring a microcontroller driven regulated gain stage to ensure absolute constant output on all bands,
Integrated 2-tone oscillator for testing and tuning,
Front panel full backlight.
“Experimental radio” means that there is enough space inside the cabinet to change boards and test new ideas in the same space. Also certain components like the SSB-filter have been made as “plug-in” components to enable quick change of the part. Also the connector between the various transmitter and receiver stages have been done by “jumpers” and header strips so that resistors and capacitors can be changed quickly to experiment with other values.
The radio has been realized with standard veroboards (0.1″ pitch), SMD components and been put into a homemade aluminum cabinet using 2 layer sandwich construction inside the cabinet.
Here a snapshot of the operational transceiver. Cabinet size, by the way, is 7.5 x 16.5 x 19.5 centimeters (2.95 x 6.5 x 7.68 inches). These dimensions are in the range of other multiband QRP transceivers like the Elecraft K2 (larger) or the Icom IC703 (a bit smaller).
A general and good practice in engineering is a steady process of improvement. This article describes the construction of a high performance transmitter/receiver for SSB (voice) communication covering the 14MHz (20 meters) high frequency amateur radio band.
Various modules that have proven high performance, liability and ruggedness in recent constructions will be combined to form a radio with outstanding receiver performance, an ultra linear transmitter with output range 15 to 20 watts and a top audio sound quality both on transmit and receive.
Key features are:
Dual DDS frequency generation with AD9834 (Local oscillator) and AD9951 (VFO),
Microcontroller (MCU): ATMega644P by ATMEL,
Single conversion superhet receiver with 9MHz interfrequency (IF) and preamplifier, mixer and IF amplifier equipped with Dual-Gate-MOSFETs,
Audio-derived automatic gain control (AGC),
Transmitter with MC1496 as double sideband (DSB) modulator and NE602 as transmit mixer,
power transimitter with 4 stages, final stage in push-pull mode.
Another version of this radio has been built before. But this one was equipped with a variable frequency oscillator (VFO) because of nostalgia reasons. Unfortunately a VFO lacks certain features (frequency stability above all) which can be overcome by using digital frequency synthesis without losing performance. Usage of a high performance DDS systems is a prerequisite to achievement and a possible solution.
Most building blocks of that respective radio have been redesigned except the VFO section that turned out as not being able to deliver the projected frequency stability to a 100% degree. Frequency instability occurred because of the flatness of the former cabinet that brought the aluminum case too close to the VFO tuned LC circuit. Aluminum has a huge tendency to expand under the influence of heat so the rig was very temperature sensitive. That undeniable fault lead to a complete reconstruction using the old RX and TX modules and building a new set of frequency generators.
Parts of the old cabinet were reused but because of the fact that the whole rig got increased vertical expansion, the cabinet was “stretched” with two lateral strips of Aluminum.
Also a full electronic transmit/receive switch with p-channel power MOSFETs has been designed to avoid usage of a DC switch relay and get a “smooth” switching.
Another objective of this radio was to get out the absolute best performing circuits of the recent projects and to build a real high-performance radio. Hence this transceiver is also some sort of an improvement of the “Old school SSB TRX” as well. The circuit empirically turned out to be very good for communication in the 14MHz band. Because of this frequent readers of this website might detect certain similarities. 😉
The Receiver section
The design objectives were:
Low noise (achieved by using Dual-Gate-MOSFETs with the receiver to a large extent)
High dynamic range (achieved by using a Dual-Gate-MOSFET as receive mixer)
High AGC range (achieved by taking RF preamp and IF amp into the AGC chain)
Good audio quality (achieved by using a TBA820M as integrated AF amplifier circuit and a 5 cm loudspeaker)
RF preamplifier and receive mixer
The radio frequency preamplifier has been designed primarily to improve the receiver’s noise figure. Delivering additional gain only is relevant in second order.
Preselection is performed with only one tuned circuit int G1 line. The center frequency of this circuit is 14.180MHz. In the output section of the stage an another identical LC circuit has been installed. This turned out to be sufficient because there is no immediate need of higher preselection. The subsequently placed mixer, that is also equipped with a Dual-Gate-MOSFET has very good high level processing qualities. No interfrequency feedthrough could be observed with various antennas. No IMD occured even when signals were very strong. Testing out in the field with partable antenna very far from man-made noise sources the receiver was very quiet and even very weak stations could be received and read with Q5.
To get most of gain swing from AGC the preamplifier is controlled by a DC voltage between 0 and 12V supplied by the AGC control stage to be described later. This voltage is halved by a 1:1 resistor voltage divider because maximum gain of the Dual-Gate-MOSFET occurs with about 6V DC applied to G2.
Clipping diodes that are sometimes used to prevent high voltage entering the 1st stage have not been installed because they are prone to produce unwanted IMD products if signal levels from the antenna are too high and undesired mixing takes place there.
To prevent self-oscillation in the preamplifier, the tuned circuit LC1 and LC2 are connected together in a special way. G1 is connected to the tuned section of LC1. This section has high impedance, thus it should be connected to a load which also has high impedance. The coupling section of the coil with low impedance is connected to the 50Ω antenna. There are not two tuned parts of the LC circuits together in one stage.
The output of the Dual-Gate-MOSFET (low impedance) is connected to the coupling winding, the high impedance tuned part is going the high impedance of G1 of the mixer. The impedance ratio between the two coils is 16:4 due to the winding ratio of 4:1 of the coil set.
The sensitivity and noise figure of the whole receiver is determined by these two stages. Measurements showed that the minimum discernable signal is about 0.1µV which is very good for a short wave receiver.
SSB-Filter, IF amplifier, Demodulator, AF amp and AGC
The following stages are some sort of best practice combination of circuits that have proven to perform very well in the recent projects.
SSB-Filter and relay
The SSB filter is switched with a special rf relay by Teledyne® ensuring excellent isolation of relay ports with very low capacities in the unswitched signal path. Here the usage of shielded cable is mandatory for connecting the relay/filter section to the transmitter (see later text!). A clamp diode has been installed to eliminate high voltage peaks due to self-induction when the relay is switched. This will prevent the MOSFETs in the switching unit from excessive voltage and possible destruction.
A proven and reliable circuit can be found here as well. One stage delivers IF gain of about 12dB which is sufficient because the mixer following as a demodulator (NE612) also propduces some dB of gain. Too much gain in this section only contributes to high noise in the speaker later and is not desirable.
The Dual-Gate-MOSFET in this stage is also integrated to the AGC chain. Together with the RF preamp installed in the front and also being part of AGC control end we will get some 20 to 25 dB of gain swing when AGC is fully driven. This turned out to be enough, only in some rare cases I found that the manual gain control (also included in this recevier) needs to be used in addition when AGC is not able to cope with excessive signal levels.
Compared to a MC1350 IC equipped IF amplifier I found that gain control is much smoother because the V->dB function is very much less precipitously with the Dual-Gate-MOSFET than it is with the MC1350.
NE612 is built-in here. The main advantage of this IC is that it requires only a few components and it has got an additional gain of about 12dB or more.
In VDD line you will find a 5.6V Zener to bring 12..14V supply voltage down to about 6V. There are also two capacitors. The 0.1uF is for bleeding off rf energy from or to the supply rail, the same is the purpose of the 10uF cap for audio frequencies or low frequency noise present on VDD line. This noise sometimes originates from the digital components in the radio and should be eliminated at all reasonable points in the circuit. Also it will help to prevent the high gain amplifier chain from self-oscillating in the audio frequency range.
Audio frequency amplifier section
Two ICs are used here. The first is an operational amplifier (uA741) with a 150kΩ resistor as part of negative feedback circuit. This value is comparatively low. If (in rare cases) higher gain should be needed it can be replaced by e. g. 330kΩ or even more.
The main audio amp is the TBA820M, an integrated audio amplifier in 8 pin DIL case. It is an interesting alternative for LM386 because tendency for self-ocillation is much lower within the TBA820M. But it requires some more components. TBA820M can be switched with the load (speaker) to VVD or GND. I use a headphone jack in the radio, that is grounded, hence I prefer the latter version.
A “good” loudspeaker with 5cm of diameter was found by ordering a larger series of different speakers from Chinese vendors via ebay. The differences in sound quality are breath-taking. So it is worthwhile spending some money and order a larger variety of speakers and install the very best one.
This is a circuit I have used many times and it has proven to work very reliable. If you wish to have different settings concerning attack and decay time then another cap can be added via a switch to GND in parallel to the 47uF cap. Another 100uF for example will give a few extra fractions of a second in attack/decay time.
A 20kΩ variable resistor is used for manual gain setting. The AGC voltage that is near to VDD (12V or more) is divided and so AGC and manual gain control can be combined. At least until the point where noch AGCing will take place because the resulting voltage is <3V.
The “AGC thres.” variable resistor shown in the schematic will determine the point where AGC becomes active. I usually set it that way that solely band noise does not affect the AGC. Stronger stations (coming with S5 or 6 with a commercial transceiver) should give first minor influence on the AGC voltage. That is the point where amp gain should start dropping gradually. Strong stations must set AGC voltage to nearly 0 V.
The Transmitter section
The transmitter generally consists of two parts:
The SSB generator and the TX mixer, and
the Power Amplifier.
The full schematic of the two parts together:
Starting from the left we see the microphone amp. A nostalgic but still available operational amplifier integrated circuit (741) is used here. The amp has high gain (about 30dB) to make a dynamic microphone connectable. There is no DC feeding for an electret microphone. If you should wish to use one then the negative feedback resistors should be lowered to about 47kΩ and the audio level should be carefully observed to avoid excessive driving. DC must also be applied for htis type of microphone!
Double sideband generator
The MC1496 (still available as NOS in 14 pin DIL case or fresh from the market in SMD package by ON Semiconductors) offers high carrier suppression of about 50 to 60 dB. Therefore a network of 2 x 10kΩ and a 50kΩ variable resistor has been installed. The crucial point: To make full usage of this network the carrier offset must be set properly. If you should notice that there is no point within the full swing of the 50kΩ variable resistor then the carrier frequency should be readjusted.
A balanced output transformer has been installed to improve carrier suppression and to enhance output voltage.
SSB filter coupling out
The usage of shielded cable is mandatory here to avoid transfer of rf stray energy into the DSB and SSB line!
This stage also is equipped with an NE612 doubly balanced mixer due to reasons of circuit simplicity.
14MHz Band pass filter
This filter also needs observation. I use the TOKO style coil formers familiar from other projects. The winding data can be found in the schematic. The coil formers must have the ferrite caps and metal shield cans on to avoid incoupling of rf energy from the subsequent power stages. The filter should be placed away from the higher power stages to avoid self-oscillation inside the transmitter section.
RF amplifier power stages
The amplifier presented here has been tested in 2 different radios so far and has proven to be very stable, very linear and very rugged against antenna mismatch. The power levels are about 10 db gain per stage. From the second stage on the output impedance is 50Ω. This makes it easier to measure power levels with a 50Ω standard dummy load.
The 2 watt driver stage uses a PI-filter instead of a broadband transformer. This is because I intended to save some space on the veroboard and for a monoband transmitter this is a practical solution. If you should find out that there is a mismatch that results in losing gain, then the capacitors can slightly be modified because the L-network has impedance transforming capabilities. By knowing input versus output impedance and calculating a “Q”-factor subsequently L and C can be computed to get a defined step-down impedance (Link for further information). This is a useful method and, in case of low pass filter like applied here, there is also a filter for harmonics.
Driver and PA power amp are biased for AB-mode, all other stages operate in A-mode to ensure best linearity. Strategies using emitter degeneration and negative feedback are inherent in preamp and predriver stage.
All transistors apart from preamp stage require usage of heat sinks.
Impedance matching is either not done (stage 1 to 2), by transformer (stage 2 to 3) or by L-network (stage 3 to 4). Whereas from stage 3 to 4 also there is a transformer applied to split the signal symmetrically to the two bases of the final transistors.
If there should be a tendency for self-oscillation within this stage the input transformer secondary winding can be center tapped and put to GND via a 0.1 capacitor.
Power out depends on DC power voltage and is about 20 watts when run on 13.5 V DC power supply an the amplifier terminated to a 50Ω load.
This is a spectroscopical analysis of the fully driven transmitter (f=14.200kHz, Pout = 20.1 watts, VDD=13.0V) and the remaining carrier.
Harmonics are filtered very effectively . This is achieved by using a push-pull final stage driven in AB mode. Some authors say this is useful to eliminate odd number harmonics. On the other hand there are two sections of low pass filtering (one between driver and PA, one following PA). The figure of the output spectrum between and 50 MHz below:
The Dual DDS Oscillator System
The DDS has got the following features:
VFO: AD9951 + amplifier,
LO: AD9834 + amplifier,
LCD: NOKIA 5110,
Tuning: Optical rotary encoder by Bourns,
User interface: 4 keys to control the digital settings,
Analog inputs: User keys (ADC0), VDD (ADC1), S-Value (ADC2), TX PWR (ADC3), PA Temp. (ADC4).
The control lines for DDS1 (AD9951) and DDS2 (Ad9834) are as follows:
The colors are the colors used for the cables in my radio.
The LCD is connected likewise:
The NOKIA5110 LCD has been designed for VDD=3.3V. Please use 10kΩ resistors in the control lines which are not in the schematic! 3.3V are derived more or less closely by switching 2 Si-Diodes in series which results in a voltage drop of about 1.4V. Hence the LCD gets 3.6V DC from the 5V supply chain which is no problem for the module. One big advantage of the Nokia LCD should not be forgotten: It is very quiet and does not produce any discernable digital noise. Thus it is my favourite meanwhile for receivers on the RF bands.
For both DDS modules coupling out the rf is done with symmetrical circuits using trifilar broadband transformers. 10 turns on a FT37-43 core are a good choice. This will enhance gain and reduce spurs.
DDS2 is clocked to 110MHz, but keep in mind, that AD9834 is specified for 75 MHz max. clock rate only. I found out that modules from the “grey market” sometimes fail and produce lousy signals when overclocked. You can see that on a scope when extra peaks appear or with the spectrum analyzer when spurious signal are frequent. I recommend buying with Mouser or anther trusty vendor for example or reduce clock rate in case of problems in signal quality.
Power consumption is not excessive because both DDS modules are for low power application, unlike the AD9850 or AD9835, that draw much higher current. Power rate is 300mA when in receive mode with LCD backlight on.
The C-code for the software has about 2600 lines source code and can be downloaded here.
A standard CB DC supply cable is used here. Unfortunatel the plugs equipped with a cable and fuse holder are widely availabe but the sockets have to be stripped from old CB trasnceivers.
On the air the transceiver performs great. Audio is clear and powerful what the QSO partners often tell me. The receiver is fun to listen to, sounding soft AND precise. Maybe I will do a YouTube video the next weeks to prove it! 😉
All my rigs are for portable, hiking, bicycle trips and travel to foreign countries. I use Aluminum as a basis for the hardware to keep the radio lightweight. With this radio a ground plane made of 0.8mm Aluminum sheet metal has been used that one has been enforced with a lateral additional ground plane carrying the DDS system (see pictures in this article, please). Thus the base frame is pretty rigid and not prone for bending.
Front an rear panel are made from 0,.8mm Al sheets (rear) and 0.5 mm Al sheet (front).
The various subassemblies (DDS, receiver, transmitter) are split into different modules and are seperatelay fixed with bolts and washers mounted to special spacer bolts for screws of 2mm diameter. This ensures better grounding instead of using larger veroboards. Connections are made from flexible stranded hook-up wire and shielded cable for rf and audio signals.
On the undersides of the single boards copper foil is used for lines with GND portential.
In this paper we will discuss a single sideband amateur radio transmitter/receiver for the 40 meter band that has been designed to ensure good performance characteristics with reasonable number of parts (no “overkill” in component use), particularly concerning the receiver. Circuit simplicity and over-average performance were to be combined.
The background: Some years ago I had built the ancestor of this transceiver and afterwards posted an incomplete series of articles (starting here). The transmitter was considered to be quite OK (I could even work a station from South Korea when operating as GJ/DK7IH some years ago) but the receiver was weak.
The shortcomings originated from the rf preamplifier I used together with the 1st mixer, an NE602. The latter had severe problems to cope with the high signal levels on the 40 meter band from out-of-band broadcast stations transmitting on the 41m band (f>7200kHz) or from very strong amateur stations transmitting in-band. This is caused by the technical specs of this Gilbert cell mixer. NE602 has been designed for mobile phone applications and not for shortwave radios. Its IMD 3 is only -15dBm whereas it is able to detect weak signals (-119dBm with an S/N ratio of 12 dB) according to datasheet. Due to this NE602 was excluded from being used at least in the receiver.
Another point was that the rig was too small and too densely packed to be called “service friendly”. Thus I dismantled the radio some times afterwards and had in mind rebuilding it with another receiver and a little bit more space inside.
The project has had to meet certain requirements that I would like to point out first:
Frequency generation: Dual-DDS-System: AD9835 as local oscillator and AD9834 as VFO. ATMega644A as MCU (Download source code here)
Receiver: Single conversion superhet, 9 MHz interfrequency with commercial filter (supplied by http://box73.de) shared by transmitter and receiver and relay switched, “NE 602-free zone” ;-), 4 dual gate MOSFETs in rf preamp, rx mixer, if amplifier and product detector, audio stages with BC547 as preamp and LM386 as main audio amplifier.
Edit: I found that there was strong signal of self-reception around 7.100kHz which was not a spurious signal from one of the DDS. It has been a mixing product of one or two oscillators together with a signal from the microcontroller. So I changed the interfrequency to 10.7MHz which cured the problem. I tried to calculate the issue but was not succcesful because I do not know all the frequencies in the microcontroller. I think it is most probable that it is a harmonic of the PWM signal I use for controlling the LED front lights.
Transmitter: 4 stages, 3 of them in push-pull mode, Siemens made mixer IC S042P (really old fashioned, but still available) as DSB generator and TX mixer, rf amplifiers (2N2219A) after filter and tx mixer.
Design: Really “cool” with blue backlight. Sandwich built, not the size of a “micro transceiver”, but handy for travelling.
The Block Diagram
The diagram can be derived from the old project, it is nearly the same:
The basic outline of the radio is standard and should not be further discussed.
Dual DDS (VFO and Local Oscillator (LO))
This time I wanted to use 2 digital oscillators. The reason was just to have fun. 😉 Here is the schematic:
The source code has got about 2200 lines. With the GNU C compiler this leads to a HEX-file of about 43kB. Because of this the controller had to have a little bit of more memory. A “644” is a good choice here. It is clocked internally to 8 MHz clock rate. Radio and user data (user operated keys, S-Meter, TX PWR meter, temperature sensors attached to final transistors) is lead to the analog-digital-converter (ADC) of the MCU. Rotary encoder (optical) is fed into digital inputs. Integration of an RTC is projected but not done yet.
Here an AD9834 is used. It is overclocked with 110MHz clock rate. For my receiver with a DDS chip purchased from Mouser this works without any abnormality. With a a chip from the “free market” (ebay) I found that there were strange clicks in the signal. So, I do not really recommend overclocking under any circumstance and/or not to such a high degree.
This DDS is is not terminated with a low pass filter. Due to the high clock rate there is no clock oscillator feedthrough which is supported by the design of the following amplifier having an audio frequency transistor in the last stage (BC547 and later BCY59) that limits high frequency components due to its early gain decay in the frequency spectrum. The two stage amplifier has been designed for excellent linearity to prevent impurities in output spectrum.
The first peak showing the 16MHz signal and the next peak is the first harmonic about 30dB below. Other peaks are from local sources (PC, Printer).
The sine wave also looks quite OK:
This one contains an AD9835 synthesizer clocked to 50 MHz. An LPF here is mandatory. A simple but linear amplifier brings the signal up to 3Vpp which is OK for driving the dual gate MOSFET in the receiver. For the transmitter mixers this amount of voltage is too high, small capacitors reduce the voltage to an acceptable value.
From another project that I once had built and that is not more in use, a dive computer, I had a 4 lines/20 characters text display that is fairly large. This was to be designated as the LCD for this transceiver.
Building a receiver for the 7MHz amateur band is challenging. On one hand the circuit should be very sensitive for weak signal reception, particularly during day when the band conditions are low due to solar radiation and density of the D-layer. This means the receiver should have a higher gain whereas noise figure does not play a predominant role due to band characteristics with high atmospheric noise on 7MHz.
Next request is high dynamic range to eliminate the spurious signals that occur when front end stages are loaded with high input signal levels.
And last but not least AGC control range should be as wide as possible to cope with weak and very strong signals without the request to intervene by adapting manual gain control. For this a preamp also benefits.
Active mixers like the NE602 show low performance under these conditions. Some high-current mixers like the SL6440 exist, but there are alternatives. On one hand the classical diode ring mixer might come into perspective, otherwise Dual-Gate MOSFETs are well known as having a fairly good ability to cope with high signal levels and so don’t tend to deteriorating the receiver’s performance severely. Besides they offer some gain and low noise figure (which has not been the main objective in this case) and the circuit is very compact and therefore it was the best choice for a receiver that had been intended to be constructed onto a board of 6 x 8 centimeters.
After these thoughts the following circuit turned out to be the right onset for a receiver inside the projected rig.
Circuit explanation (Receiver)
On the left we start with a 2 pole LC band pass filter for 7 MHz. The coils are wound on TOKO style coil formers (5.5mm size), winding data and parallel capacitors are given in the drawing. The coupling capacitor (2.7pF) between the two LC circuits is very small for such a low frequency. This makes the filter response curve sharper but leads to a slight weakening of the signal coming through the filter. But as the whole receiver has plenty of gain and a very good noise figure, this is the reason why some weakening of the input signal is acceptable.
Next is the preamplifier for the received band. It is connected to the AGC chain. You can expect some 25 to 30dB gain swing by driving up gate 2 of the dual gate MOSFET from 0 V to 6V. A 1:1 voltage divider decrease the 0..12V AGC voltage to 0..6 V where th3N205 MOSFET is close to amplify with maximum gain. Exceeding 6 to 7 volts does not result in significant more gain swing, so I usually drive the MOSFET from 0 to 6.5 volts UG2 (with 13 Volts of supplied voltage.
The coupling when going from the preamplifier to the receiver mixer is in broadband style. The 3N205 has a very high gain and tends to self-oscillate. A second LC circuit makes the device more prone to going self-resonant and hence produce unwanted signals.
This mixer is very simple and needs only a few components. Both signals are fed into the gates of the dual gate MOSFET. Rf goes to gate 1 whereas gate 2 (the AGC input) is fed with the oscillator signal). Gate voltage depends on the voltage drop at the source resistor and therefore is stabilized. The oscillator signal should be in the range of 2 to 3 volts rf (pp) for a dual gate MOSFET. Lower values will deteriorate the performance of the mixer, e. g. its dynamic range. This signal switches the semiconductor and a superposition of the two signals occurs thus leading to the production of sum and difference of the original frequencies. These signals are fed into…
The SSB filter
which is a commercial one (Supplier box.73.de). The reason why I don’t ladder filters anymore is that I found it extremely difficult (not to say impossible) to get a symmetric filter response curve thus making the lower and upper sideband of the receiver sounding different even when the carrier frequency has been adjusted very thoroughly.
The filter is used for the SSB transmitter as well. To ensure maximum signal separation between the two branches (tx and rx) and between filter input and output I again us a high quality rf relay made by Teledyne. When choosing a relay intercontact capacitance is crucial. It should (if possible) be < 1 pF.
Don’t forget a clamp diode to VDD over the relay coil to eliminate high voltage voltage peaks generated by self inductance when the coil is switched off. Voltages up to 100 Volts can occur. This might damage the transmit-receive section of this transceiver that is equipped with semiconductors only and does not use a relay.
This circuit is the same like that of the rf preamp. It also is part of the AGC chain, thus delivering another 25 to 30 dBs of gain swing so that overall gain swing is around 50 to 60dB. In practical research over a long period of observation I found that with an antenna delivering high signal voltage (Delta loop) it was not possible to overdrive the receiver to a level where signal distortion was audible.
A tuned circuit is also placed here to increase gain. Tuned amplifiers usually have higher gain than broadband ones. It is highly recommended to ground the metal cans of the coil to prevent any self-oscillation. But as I found out, this amplifier is not very prone to go to self-oscillation state.
Here again a dual gate MOSFET is used. The circuit is nearly the same like the RX mixer except from the output section. We can see a low pass filter here, consisting of 2 Cs (0.1uF) and a resistor (1k). You can use a radio frequency choke instead, 1mH is recommended.
This section consists of two parts, a preamp (with bipolar BC547) and a final amplifier (LM386 IC). It is well-known that this IC tends to oscillate. One measure to prevent this is to keep leads short, switch a low-pass filter (capacitor 100uF and R=33Ω) into the VDD line and to reduce the gain capacitor between pins 1 and 8 to a degree where self-oscillations terminate.
A switching transistor cuts off the audio line by short circuiting it when on transmit. This eliminates any noise when switching. The rx/tx switch now is 100% “click free”. A very pleasant way of operation. 😉
This is another re-use of a circuit I have frequently used before. It is desired to reduce its output voltage down to 0 volts when a more or less strong af signals appear at the input. The agc voltage is derived from the audio signal of the receiver. Some say that this is not the best choice because you need more time (an af cycle last much longer as an rf cycle) for the waveform to generate the regulating DC voltage.
Nonetheless I have never observed popping or unpleasant noise from incoming very strong signals. The agc response rate is so fast that you won’t notice that it just has regulated even when a strong signal comes in. Only with very, very strong signals a slight “plopp” sound is observable but it is not unpleasant.
A second capacitor can be switched in parallel to the 33uF one. This can either be done by a transistor switch (like shown in the schematic) that in this case is controlled by an output PIN of the MCU. An alternative that I found later is to use the MCU pin directly to switch the cap. When not using the additional cap you must switch the pin as an input so that there is no positive voltage from the pin to the circuit. When you intend to ground the transistor (agc in “slow” position) then the pin mus be set as output by defining the DDR-register respectively AND the pin must be set to 0. So you can get rid of the switching transistor.
Another possibility would be to derive the agc from the interfrequency signal. The problem that occurs in this case is that you have to decouple the local oscillator (bfo) very carefully from the place where agc circuit is placed. Otherwise you are at risk to detect the bfo signal by the agc which leads to reduced response range in the agc. In addition this receiver uses a higher rf voltage level for the mixers (2 to 3 Vpp each). By this the amount of stray energy is higher inside the circuit and thus this rf energy might be detected very early by the agc.
In the emitter line there is a resistor (68Ω) which produces a voltage drop when the transistor is driven. This is fed into the ADC of the microcontroller driving the S-meter display part.
First the circuit:
This amplifier is a simple common-emitter circuit with the directly grounded emitter of the BC547 transistor. This circuit is linear only for low input voltages but suitable for the connected dynamic microphone since this does not produce more than some millivolts of audio energy. Bias comes from the 390kΩ resistor. At the input you find a 2.2nF capacitor from base to GND which helps to prevent coupling in rf energy from the transmitter to the audio stage and thus leading to an impure signal.
The DSB generator + amplifier
The amplified microphone signal is used to produce a double-sideband signal. The ic I use here is an antique but still available part by German manufacturer Siemens, the S042P. It includes a so-called “Gilbert-cell” mixer and an oscillator but the latter is not used here (DatasheetApplication note (in German)).
The S042P mixer needs some more components compared to the well-known NE602 integrated circuit but fewer ones than the MC1496. It is designed for 12V usage, thus no voltage regulation is required.The ic can be applied in balanced mode or non-symmetrical. To save components I use the unbalanced circuit alternative. A slight loss in output power is acceptable in this case, there are amplifiers post each mixer in this transmitter.
Ic gain is about 16.5 dB, DC current is about 3 mA.
A crucial point is the signal level of the local oscillator. S042P needs only some hundred millivolts of oscillator voltage. To prevent overdriving I experimented with different values of the coupling capacitor. 5.6pF seemed best because the LO produces some volts peak-to-peak.
Following there is an amplifier that is a standard circuit and has been tuned for maximum linearity in order to reduce distortion to a minimum (which is also true for the following stages). You can see the well understood 2 master ways of achieving max. linearity in an amplifier stage:
Negative feedback between collector and base (i)
Emmitter degeneration (II)
i) The first measure goes along with the 2.7kΩ resistor between collector and base of the transistor. This resistor provides positive dc bias voltage to the base and leads 90° out-of-phase ac voltage to the transistor’s input. This reduces gain and therefore distortion. But due to the fact that the whole transmitter strip has plenty of gain, this loss in gain is not a serious problem.
ii) The 10Ω resistor in the emmitter line is not bypassed by a capacitor. This stabilizes the circuit. When the current through transistor increases the emmitter voltage will rise (according to Ohm’s law) and the voltage between collector and emmitter drops. This reduces voltage difference between base and emmitter and hence also reduces gain.
The coupling to the next stage is done by a capacitor of 0.1uF. This causes some impedance mismatch. But that is as well not a big problem because the gain reduction here helps to prevent the whole transmitter from unwanted oscillations by diminishing overall gain.
Here the second S042P is used. The 9 MHz SSB signal is coupled to pin 13 of the ic, a DC connection is established to pin 11. These two pins represent the base connectors for the two current control transistors and should be bridged by a DC resistor in this circuit.
The 150Ω resistor from pin 10 and pin 12 to GND defines the gain of the mixer. Here you can use down to 150Ω but should have a resistor towards VDD to limit current and avoid excessive heating of the device. In this case another 150Ω is used.
VFO signal is coupled symmetrically to pins 7 and 8 via a small trifilar toroid. See schematic for details and please note that center tap is not used here. This is in contrast to the output transformer where the tap is used to feed supply voltage into the mixer.
Another 7 MHz band pass filter terminates the mixer, data for coils and capacitors is in the schematic.
This amplifier has got 4 stages and except from the first one all are in push-pull mode. The power distribution for these 4 stages is as follows:
The first of the 4 power stages is the same as the post dsb generator amplifier so there is not more to add concerning this stage. Rf energy is taken out via a transformer with a primary and a tapped secondary winding. This is to provide the balanced structure necessary for the following push-pull stage.
This is the first push-pull stage. Its bias is derived from a voltage divider connected to the tap of the input transformer.
Please note: In contrary to the schematic I have installed 2 devices of the 2SC1973 type because the signal turned out to be much purer with these ones on the spectrum analyzer.
A tapped output transformer feeds the amplified rf energy to next board. Output impedance is 50Ω. The coupling to next stage then is done via a shielded cable of (nearly) the same impedance.
This one has an input transformer also center tapped. The tap goes to a bias network consisting of a current limiting resistor (1kΩ), two diodes forming the lower part of a voltage divider and some capacitors as part of a low pass filter to avoid coupling in of radio frequency (rf) energy. The two diodes must be thermally connected to the cases of the transistors. In case these heat up, the diode increases its conductivity thus reducing its resistance. The bias voltage drops and heating is stopped. So, thermal runaway is prevented.
For these two stages (predriver and driver) DC is fed through low pass filter (RFC and 2 caps 0.1uF) to prevent coupling of rf energy via the VDD line.
This stage receives input from a balanced structure without a center fed transformer. Instead bias current is linked in via a network of radio frequency chokes and two resistors of 5.1Ω each.
Bias is provided by a current regulating transistor and should be set to about 100mA.
The MRF455 transistors are mounted directly to the aluminium structure of the sheet metal carrying the whole transceiver boards. When mounting them to the Veroboard I did not solder them directly. I used 1.6mm screws and washers to press the brass connectors to the copper strips of the amplifier board:
With this I could have been able to remove the precious transistors without having to unsolder them when the device might have turned out to be a failure. But it was not, thank God!
The output transformer is the one I have used in my old 14MHz PA and the ancestor of this radio. It is from an old ATLAS 215 transceiver and I hope that this will be the final place for the transformer.
Two temperature sensors (KTY-81-210) have been installed to measure the temperature of each transistor. They are connected to the microcontroller via voltage dividers (see schematic, please!)
Low Pass Filter and Power Measurement Unit
For the low pass filter I use 2 toroids T50-2. These might appear small but from one source (that I have forgotten) I remember to have found that for 50 watts of power this core is still suffice. Metal powder cores can stand much more power compared with same sized ferrite toroids.
The power measurement unit consists of a network that starts with a resistor of 12kΩ to ensure a significant voltage drop in signal level, then two rectifier diodes (1N1418 or equivalent) follow, some low pass filtering eliminating the last rf energy and the resulting direct current voltage is fed to a variable resistor to set an adequate voltage level for the ADC in the microcontroller.
The rf output made out of a two-tone audio signal measured at the antenna connector:
The spectroscopical analysis shows the signal on the f -> V figure:
A very simple circuit. Two PNP power transistors are used but they don’t have that much to do. They are only designed for switching the low-power parts of the radio. The high current to the drivers and final amplifiers is permanently present in the collector lines but the bias lines are tx/rx-switched and go to 0V during receive periods. This reduces requirements for the power rating of the switch board.
When pushing the PTT the base of the lower transistor is pulled to GND. So it becomes conductive and TX DC is applied. Via the diode the upper transistor loses its negative voltage and becomes non-conductive.
One interesting thing was the blue backlight to illuminate the front panel controls. It is made using SMD LEDs that are soldered to small pieces of Veroboard and fixed with 2-component glue to transparent light-scattering plastic bought from a local supplier for architects and designers. This material is used for making models of houses and stuff like that. As light distributor this material is excellent. The LEDs are powered by a linear transistor connected to the pulse width modulation (PWM) output of the microcontroller so that light intensity is adjustable.
Hint: When programming the PWM functions it might occur that PWM frequency is audible in the receiver. If something like that occurs another frequency can be selected without changing the performance as soon as it is high enough that human eyes aren’t able to recognize a flickering.
The covers used for the labels and the LCD shield are made from 2mm acrylic and fixed with screws of 1.6 respective 2mm diameter.
The two push-buttons right in top position consist of two bars of acrylic (4.2mm diameter) and are having mechanical contact to small spring-loaded switches behind the front panel:
Directly under these acrylic bars there are two LEDs shining into these rods and because of total reflection inside the tubing the optic conductor is sending the light to the front side when the LEDs are powered on. That is how it looks at night:
This is a sandwich construction again. On the first side there is the DDS board (left), the receiver (center) TX mixer and preamplifier (right) and the SSB generator (back). Also there is a 5 lead connector holding the 5 ISP lines (MOSI, MISO, CLK, RESET and GND). This makes firmware updates easy because you don’t have to open the case when you want to update software.
The other side holds the TX low pass filter plus power measurement unit (left), the power amplifier (center) and the predriver and driver (right). In the back you can see the rx/tx switch board:
“On the air”
Again big fun this transceiver! During the ARRL DX contest last weekend I could work some statesiders. With Delta Loop and 50 watts, fairly OK. Working Europe all day is no problem with 50 watts.
During the first QSOs I had reports that the audio sounded clear but somehow “narrow”. I had used an electret mike that time and could not use a dynamic one because the preamplifier following the microphone did not have enough gain. Then, to solve this problem, I decided to do a full reconstruction of the SSB generator board. The one then had used had an AN612 mixer integrated circuit (from an old CB radio). This one was dismantled and replaced by the S042P board. The change took me 3 hours to develop and solder but it paid. I use a Motorola dynamic microphone now that has a very rich and clean sound. I monitored it on a web based SDR receiver, made a recording and found it to be OK.
OK, dear fellow hams, that’s the story so far, some supplements will sure be made, so stay tuned!
This is a software project for building a VFO with the 75MHz clocked AD9834 synthesizer chip by Analog Devices. Due to the Nyquyst theoreme with a maximum clock rate of 75 MHz a frequency of 37.5 MHz can be achieved. When overclocking the chip to 100 MHz (which has been succesfully tested in many cases) the maximum output frequency theoretically rises up to 50 MHz. But when coming close to these boundaries signal quality deteriorates severely. So it is recommended not to produce higher output frequencies than 25 MHz (75 MHz clock) or respectively 33 MHz (100MHz clock).
The AD9834 is a low power (20mW power consumption when VDD=3.3V) DDS module. It can handle up to 5.5 V as VDD (2.3V min.), so 5V single supply use makes circuitry simple. It comes in a 20 lead TSSOP case, breakout boards are available.
SPI signal structure
Programming a desired frequency into the DDS chip is performed by a 3 line communication, a serial peripheral interface (SPI). These three lines are called
SCLK (the clock signal)
FSYNC (the signal that determines the end of the transfer of a single word (16 bits)
SDATA (the frequency or control information packed in 16 bit word)
The timing diagram found in AD’s datasheet gives the precise structure of the signal communication:
FSYNC is high when the first word (16 bits) is going to be transferred. SCLK is also high in this moment. Then FSYNC is set low, 16 bits subsequently are transmitted via the SDATA line. After one bit has been transmitted, SCLK is set low for 10 ns minimum and then goes high again for the next bit. Transfer starts with MSB (D15).
After 16 bits have been transmitted, FSYNC is set high again, showing the DDS chip that the word has been completely transferred. As soon a FSYNC is low again the DDS is ready for the transmission of the next 16 bits.
The DDS chip “language”
The chip has two frequency registers (FREQ0 and FREQ1). These registers contain 28 bits of frequency information. They can be addressed individually and are divided into two 14-bit sections each (MSB and LSB). In addition it is possible to load the LSB independently from the MSB if only a minor frequency change is required.
Frequency registers are selected by the first two MS-bits (DB15 and DB14) of a 16 bit structure sent to the DDS. “01” determines a frequency load for FREQ0, “10” loads the FREQ1 register.
Besides the frequency information some controls must be sent to the DDS chip. A control is also 16 bits wide. A control is initiated by a “00” starting sequence for DB15 and DB14.
So we can distinguish the purpose of a word by its first two bits:
“01” + 14 following bits loads the FREQ0 register,
“10” + 14 following bits loads the FREQ1 register,
“00” + 14 following bits transfers a control word.
One important control bit under the looking glass
There are many features to control the AD9834 chip. We want to limit this to the absolute basics. The most important bit is DB13. If you set this to “1” the chip is informed that the frequency information for the register to be loaded next will come in two consecutive 16 bit words addressing the respective frequency register with the 14 + 14 bits of frequency information.
So the first step is to transfer the “00” signaling a control code and next the “1” signaling that the user wants to write two 16 bit words for changing the frequency. The rest of the 16 bits of this control can be left “0”. This results in
is the first word to be transmitted.
Frequency calculation and transfer
The frequency data of the waveform the user wants the chip to put out is determined by 28 bits, a so called “frequency word”. The formula is
frequency word = 2^28 / fclk * f
frequency word: a floating point number that will later be converted into a long integer containing the frequency information for the chip,
fclk: The master clock rate of the clock oscillator connected to the DDS [Hz],
f: The frequency the user wants to be generated [Hz].
With a clock rate of 75 MHz a user frequency of 1 MHz would be calculated as a frequency word of
By leaving only the integer part of the number we get 3579139 which now is the frequency word that must be transferred to the chip.
Converted to binary this number figures out as
This is now split into two parts, 14 bits each:
Now we must tell the DDS in which of the two frequency registers we want to store this. Therefore we add the 2-digit-code for the desired frequency register in front of the respective number. In this example the destination is FREQ0, so we add “01”. The result are two words of 16 bits each:
Together with the control that allows us to write the 2 words consecutive into the chip we get a complete sequence of
0010000000000000 011110100000011 010000110110100
because the correct order is CONTROL first, then LSB, and MSB last. In HEX this is 0x2000, 0x3D03, 0x21B4.
Code examples in C for the AVR family follow. First the declarations, then some defines in advance so that you can adapt the code easily to your layout: