Having deferred the work on the “micro multibander” for some time I finished another small QRP rig (this one for 7MHz) that is suitable for my summer excursions by bike or hiking the local mountains here in the State of Rhineland-Palatinate or the Black Forest that is not that far away on the other side of the Rhine valley.
Besides, this transceiver to be discussed here is some sort of a “remake” of a 20 meter rig I built 3 years before. And this time, the transceiver really fits into a shirt pocket without having to wear “XXXXL”- clothing. ;-):
General circuit description (instead of presenting a block diagram)
The rig uses two mixers NE602 plus one filter as central elements. The signal way is reversed when switching from receive to trasmit mode. This is done by 2 relays and is a well known technique for simple QRP rigs. You will find lots of equivalent ideas on the internet (Example 1, Example 2).
But not to ignore the shortcomings of these designs: They are somehow inferior to my requirements, particularly concercing receiver performance. I prefer to have higher signal gain and an AGC circuit. AGC for me is a must. But these designs can be expanded easily, so I added an AGC controlled interfrequency amplifier with dual gate MOSFET BF998 into the receiver’s signal path enhancing performance significantly.
The frequency generation of the superhet transceiver scheme is simple: Again I use one interfrequency (i. e. 9MHz). The VFO is DDS based on AD9835 operating below the desired radio frequency, which means that it is set to the range of about 2 MHz. Due to this low frequency you could replace the DDS by a VFO if you don’t like the relatively complex work with the software programming and microcontroller stuff). A 2MHz VFO can also be made very stable, so this is an alternative not to be ignoered.
Due to the fact that the schematic is not very difficult to analyze you are kindly requested to refer to it for further talking:
In the center of the schematic you can see the main elements of the circuit: One SSB filter (9MHz), correctly terminated by 2 resistors of 1k each (to ensure proper filter response curve) and two relays with a double set of switches. These relays reverse the way the signal travels through the filter. The advantage of this: You can use the integrated oscillator of the NE612 controlled by a crystal and a tuning capacitor to set the carrier frequency correctly for the lower sideband because the mixer is used as SSB generator and as product detector in common.
A word on chosing the proper relays: An intense examination of the relays’ data sheet is essential. I built a prototype of this transceiver on a breadboard prior to soldering the components to a veroboard. I found that some SMD relays have signifikant coupling capacities between the unused relay contacts (in the range of some Picofarads). So stray coupling was a severe problem. Later I used some second-hand Teledyne RF relays that I had purchased via ebay two years ago (price originally 50€!) for 1€ each. These relays are absolutely superb!
Before we go: In the circuit scheme above I missed out the antenna switch relay because I think every homebrewer knows what to do in this case. 😉 So the receiver’s signal path starts with a band filter for 7MHz consisting of to tuned LC circuits. The coupling is relatively loose. As coils I use the well known coil formers in TOKO style with 5.5mm outside measure.
Coil data for the 7MHz band pass filter (BPF) is 39 turns primary and 9 turns secondary of 0.1 mm enameled wire. The respective capacitor is 33pF. This is a high L to C ratio which gives you excellent LC quality factor. This is mandatory especially when working on the 40 meter band, because of the strong broadcasters starting from 7.200 kHz intermodulation might be a problem when the receiver is connected to a high gain antenna and broadcasters’ signals might overload the first mixer (remember that NE612 has a relatively low IM3!). If you still should have problems coping with too strong out-of-band signals you can reduce the coupler from 4.7pF down to 2.7pF.
In practical terms I could not detect any unwanted signal products even when using an antenna with high rf output voltage. One reasons for this is, that there is no rf preamplifier for the receiver. This avoids overloading the first mixer generally.
The NE612 has two mixer inputs and two outputs. This makes it very suitable for this sort of radio. In receive mode pin 2 of the right NE612 is used as signal input. VFO signal is fed into pin 6. The resulting mixer products are taken out from pin 4. Next the 9MHz filter follows from right to left.
The 9MHz IF signal then is fed into an IF amplifier. This one is equipped with a dual gate MOSFET (BF998), gain is about 15dB when full AGC voltage is applied wich leads to about 6V by the 1:1 volatge divider in the applied to gate 2 of the MOSFET.
The left NE612 is the product detector. I use the internal oscillator with a 9MHz crsytal and a tuning capacitor here. This saves building an extra oscillator and simplifies the rig again.
One AF low pass filter made of 1k resistor, 100uF rf choke and a 0.1 uF capacitor eliminates high frequency remainders generated by the mixing process.
The audio stages are also made simple: One preamplifier (using bipolar transistor in grounded emmitter circuit) and a final stage with LM386 transform the signal to a level that is sufficient to be fed into a small 8 ohm loudspeaker or a set of standrd MP3-player headphones. Because the rig is very small and there was definetely no space for a loudspeaker I use headphones instead.
Keep an eye on the power supply switching of the two audio stages. The problem was to eliminate the switching click and pops to a minimum and to avoid acoustic feedback when unsing a loudspeaker. So the audio preamp is only connected to DC on receive. When switching to transmit the charged capacitors avoid instant cut off supplying some Milliseconds DC to the amp until significantly discharged. The main amplfier on the other hand is connected to permanent DC supply. So it won’t pop when switching from tx to rx an vice versa but can cause feedback. To avoid feedback a transistor is used to cut the speaker/earphone from the power amplifier.
AGC is audio derived. A two stage amplifier provides a DC voltage analog to the audio input sginal strength. First amplifier stage is a common emitter bipolar transistor supplying sufficient audio voltage. This voltage is rectified by a two diode circuit letting only the positive halfways pass. You can use silicon diodes (1N1418) oder Schottky diodes here. An electrolytic capacitor (100uF/6V) provides the time constant respectively DC decay once the signal has disappeared. Output of the DC stage is split. The collector is connected to 12V via a 4.7k resistors causing a voltage drop when the transitor’s conductivity increases. The emitter is fed to the ADC of the microcontroller (pin ADC1) causing a proportional voltage to the voltage of the applied audio signal so that on the OLED an S-meter can be displayed.
An electret microphone picks the operator’s voice. The signal output level of these microphones is high enough to drive the left NE612 (which serves as balanced modulator in this case) directly. Signal input for the mixer should be 200mV RMS according to data sheet. An electret produces about 0.5 to 1 V pp if spoken with a decent voice in the distance of some centimeters. So you have more than enough audio signal power for the modulator.
BTW: Carrier suppression of the modulator is excellent. I achieved 56dB without doing anything else!
The resulting DSB signal then is fed into the SSB filter, the SSB signal subsequently is directly sent into the right NE612. A band pass filter for 7 MHz eliminates the unwanted mixer products. You should have 400 to 500 mV pp of rf signal here when the transmitter input is fully driven. I recommend a two-tone test generator to check out the linearity of this and the remaining amplifier stages!
Next parts of the transmitter are a band pass filter (same coils and capacitors like th rx bandpass filter), a preamplifier and a driver. The later should put out about 150 mW into a 50 ohm load. They are made more linear by emitter degeneration (4.7 and 2.2 ohm resistors for predriver and driver) and negative feedback. This helps to ensure that transmitter performance is fine when IMD3 products are concerned even if the main IMD3 problems usually occur in the final stage.
To transfer the rf power into the final stage proper impedance matching is mandatory. Input impedance of the final stage is fairly low (<10ohms), therefore a broadband (down)transformer is used. Data is: Core T37-43, primary 12 turns, secondary 4 turns of 0.4 mm enamled wire.
Last stage is a single ended linear amplifier in AB mode equipped with a 2SC1969 rf power transistor by eleflow.com.
BIAS circuit: The combination of the 1k resistor, a silicon diode (1N4002 or equ.) and a capacitor sets up the correct bias. Bias is fed into the cold end of the input transformer. Quiescant current should be around 40mA. A good thermal contact between the diode and the transistor is recommended. As the transistor gets warmer the diode will increase its conductivity so reducing bias current. This will prevent thermal runaway effectively!
To avoid bulky output transformers the PI-filter (7MHz LPF) is part of the tank circuit of the final amplifier transistor. For this power level this is an acceptable and practical solution because the output impedance of the stage is nearly equivalent to 50 Ohms. A certain mismatch is not a severe problem. DC to the final transistor is applied via an rf choke, for exact data please refer to the schematic!
T2 helps to suppress unwanted signals that I encountered when taking the transmitter from the dummy load test environment to a real antenna. I observed unwanted parasetic oscillation in the range of about 1MHz. T2 has a low reactance for this frequency range thus eliminating the oscillations in a reilable way by short circuiting them towards ground.
Powered with 12.5V DC the transmitter will put out slightly more than 5 watts PEP.
AD9835 is a simple but well performing 10-bit DDS chip made by Analog Devices (AD). It is controlled via 3 SPI lines transmitting the frequency data. Maximum output frequency is around 16MHz when the chip is clocked with its maximum clock rate of 50 MHz. Oscillator output voltage is some hundred millivolts peak-to-peak, so you can connect the output directly to pin 6 of the NE612 mixer.
Control signals come from an Arduino Pro Mini board. The microcontroller in this module is, if you are an Arduino user, preinstalled with a bootloader program. I overwrote this small portion of code and use the ATMega168, which is the core of the Arduino, in “native” mode. My software is written in C and transferred via “AVR dude” software using the ISP lines MOSI, MISO, SCK and RESET. These lines are not in the schematic, please refor to ATmega168 data sheet. Alternatively you can use, like shown in the schematic, an ATmega168 controller. So you have to de neccessary wiring on your own.
You will find the source code here. I packed it into an Open Document Text File because of problems I encountered when I tried to store the code into this Blogtext. If you need a compiled HEX-file, please feel free to email me!
Display is a very small OLED with 64 by 32 pixels. The OLED is, to my point of view, a little bit noisy. To suppress any rf traveling on VDD line I use an 82 ohm resistor and a set of bypass capacitors of 100uF and 0.1uF capacity closely connected to the OLED VDD pin to GND.
A low pass filter by the output of the DDS ensures spectral purity and avoids clock oscillator feed through. Remember that if you need another output frequency other than 2 MHz you should redesign the low pass filter.
Tuning is done by a rotary encoder connected to PD5 and PD6 of the microcontroller. I use the pull up resistors internal to the microcontroller, so you won’t see any other things than the mere encoder.
Tunings steps are selected by pushing the encoder knob or another suitable push button. This button is connected to ADC0 in the ATMega168 via a 3.9k resistor. The resulting ADC voltage might be problem because of a certain variation in the values of the pull up resistors that form the second resistor of the voltage divider. There is an outcommented section in the code that will show you the exact ADC value that has to be typed into the code so that key recognition works exactly.
The button once pushed will increase the tuning step by a certain amount of Hz. Steps are 10, 50, 100 (standard step), 250, 500, 1000 and 5000 Hz in and endlessly revolving chain. The step will be reset to 100Hz (standard tuning step) by leaving the tuning knob idle for 2 seconds. That’s all with the controls. Very simple, but sufficient.
The transceiver is constructed on a double sided veroboard with 6 by 8 centimeters area. Components are through hole and SMD where available. The Arduino is mounted to the front panel (another Veroboard carrying the controls etc.) as well as the OLED is. The veroboard is inserted into an aluminium frame connected to the front panel with 4 lateral M2 screws:
Wiring can be made by using the colored lines stripped from old parallel printer cables. These cables have a diameter of precisely 1mm an fit through the holes of the veroboard excactly.
If you connect any external components that are not on the same veroboard use standard 2.54 mm (0.1″) male and female board connectors! This will make it much easier to dismantle and reassemble the rig in case troubleshooting is neccessary.
Use M2 srews instaed of M3 when building very small rigs like this one!
The reverse side of the main arrangement:
Two brass made bends (from the local hardware store and each cut to a length of 8 centimeters) hold the PCB inside the mounting frame. A winding has been cut into the brass to fix the bends with screws in M2.
Together with 2 halves of a bent aluminium cabinet covered with “DC-fix” (a German manufacturer of self-adhesive PVC coating) the final rig looks like that:
DK7IH QRP SSB 40m/7MHz pocket size transceiver – final assembly
So, that’s the end of the story so far. Now it’s time for going outdoor and test the rig in field use. 😉
Work is in progress. The recent weeks I finished all the 6 modules that are going to be the receiver:
Band pass filter section
Relay switches for switching the BPFs
RF preamp, RX mixer and IF preamp
IF main amp
Product detector and AF amp section
Mounted together to an aluminium carrier board it looks like this:
On the picture the board is not equipped with the neccessary wiring yet to give the reader more sight on the single circuits. Next I will draw a schematic of each board to point out the used circuitry for those who want to build this or a similar receiver.
First test are promising so far, the receiver is sensitive, has a very low noise figure (due to dual gate MOSFETs in the preamp and the two main IF amp stages) and has shown no problems to cope with high out-of-band broadcaster signals on the 40 meter band which is due to the SBL-3 mixer I have used that has a good IM3 performance..
In my last article I talked about my ideas fo a new transceiver project beyond the QRP level. First pictures of cicuitry were also shown. In the meanwhile the transceiver has been finished, some minor changes had to be made and now it’s time to go to the details.
All construction objectives (compact size, sufficient output power to establish even DX contacts on 40 meters, good stability, good receive performance, rigidness for outdoor use) have been met so far as I can say. I had the rig with me, when I was on vacation on the Island of Jersey (GJ/DK7IH/P) from 12th to 19th of August this year. It was big fun operating the rig. Lots of stations were calling during the two days when I was on 40 meters. ODX was HL1AHS, OM Kun from Seoul. So, this was very nice for 50 watts and a vertical antenna made of a fishing rod.
First, to give an overview, let’s have a look on the completed transceiver. Cabinet size is 7.5 x 16 x 6 centimeters.
As you can see, the rig is very compact in size. The block diagram gives you an overview what is inside. Receiver section is on top, DDS can be found in the center and the transmitter is sited at the bottom of the diagram. As you can see, it’s again not rocket science and SDR-virus could not strike as well. 😉
The next posts will describe the rig in details step by step. Proceed with the receiver’s front end.
On one hand sunspot cycle is on the decline. Conditions on the higher bands tend to deteriorate gradually. On the other hand I wanted a very compact transceiver for outdoor (particularly holiday) activities. Besides this and due to the fact that I prefer monoband operation when on tour, the potentially ideal band had to be found. Based on these prerequisites I made up my band to go to 40 meters for this project.
Some interesting concepts for QRP-transceivers for this band can be found on the web. The “Santerre” and the “ILER40” should be mentioned. These operate with power levels of about 5 watts which I thought should not be enough in most of the situations. My opinion is, that on 40 meters it’s not the best decision to operate on the standard QRP watt level, e. g. like the mentioned before and my other rigs have been designed for. With high probability you’re about to get lost in the naturally higher band noise and QRM that is around on 7 MHz. Thus, 50 to 70 watts possibly should be a more suitable power level to operate on.
From the electronic point of view this transceiver doesn’t involve very much new stuff. The 4-stage transmitter with push-pull driver and final from my multi-band rig with emitter degeneration circuitry in every stage to improve linearity but without negative feedback in this circuit. Some NE612/SA602-mixers in receiver and transmitter, MC1350 as the rx’s if-amplifier, LM 386 as audio amplifier, busines as usual. DDS with AD 9551 for frequency generation. No fuss, no SDR-stuff. “Old school” homebrewing. Not more, but not less.
The challenge this time was to increase package density once more to an achievable maximum. The maximum size I wanted to have was the size of a carton of cancer sticks, also known as “cigarettes”.
Another problem of the project was to get a high power rf amplifier with an output level very much beyond standard QRP and avoiding any unwanted coupling, parasetic oscillations etc. even when components are very densly packed. A clean signal is a must, not of the package size.
The center of the construction is made of the aluminium carrier of the 50 to 70 watts final that is in the center of a three layer arrangement consisting of
rf power amplifier
rf board with SSB generator, tx mixer, pre-amp, pre-driver and driver (push-pull).
The parts of this assembly:
The main frame carrying all the other stuff:
Dissambled mounting frame of QRP transceiver for 40 meters (2016 by DK7IH)
The aluminium plane centered will keep the PA 50 to 70 watts board. This one is equipped with two MRF455 rf power transistors made by Motorola.
Above and below this board the receiver and transmitter boards are mounted to a three layer package.
These 3 boards are stacked and plugged into the front unit that you can see underneath. The front subassembly is formed of
the front panel with user controls, microphone socket and LCD
the DDS system and
the relay switching board.
Both construction groups are joined in an angle of 90° by a set of plugs and sockets so that they can be put apart easily and fast for service or improvements. All boards are made of 5×7 cm “FR4” material Veroboards.
In the center of the right board package you can see the PA amplifer, capable of delivering 50 to 70 watts SSB signal, on top the rx board, at the bottem the transmitter board.
An important issue for a high power transceiver is the transmitter’s final amplifier. Particularly thermal conductivity should be kept in mind. In this case one problem, I thought, might occur because the power transistors are not mounted directly to the rear panel of the transceiver where heat can be lead to the outside easily. Here, the two MRF455s are sited in the center of the sandwich construction holding the 3 main rf boards. But as this transceiver is for voice operation only with its comparatively low duty cycle of 10 to 20%, this was not considered as an unsolvable problem. But thermal aspects must be kept in mind, anyway.
As you can see, the final transmitter stage is based on an aluminium sheet metal of 2 mm thickness. Under the board there is a second layer also of 2 millimiter thickness aluminium. By the rear end it is joined to a solid piece of square shaped aluminium rod that itself is connected to another two thick layers of alumium which are srewed to the rear panel holding a heat sink. For this heat sink I’m currently searching a more massive one.
The thermal test was a longer QSO with Dave, M5AFD . During this longer QSO with transmission times of up to 3 to 4 minutes each the center alu panel got a temperature of 60° to 65° centigrade. Thermal stress? Not worth mentioning yet!
So far I’ve done about 50 QSOs on 40 meters with this rig, gradually improving some things. Later I will publish the detailed scheme as soon as it is finished. By now prospects are good to bring this compact QRO transceiver with me on a holiday trip to Jersey Island planned this summer.
The next step in improving my homemade QRP multiband transceiver was to reconstruct the DDS VFO. This was not urgently neccessary but after some months of continously operating the rig I was not 100% satisfied with the spurious performance of the AD9850. The AD9850 is a DDS device with only a 10-bit digital-analog-converter (DAC). These ones tend to put out still a quite high number of spurious emissions aka “birdies”.”Birdies” then are detected in the receiver causing unpleasant beep tones.
The AD9951 DDS module (some general information)
Analog Devices (AD) also offers more professional DDS-chips with a better performing 14-bit DAC. The AD9951 is such a device. It is offered for about 25 USD by mouser.com and other vendors. It is also used in commercial ham band transceivers thus we can deduce that performance is improved compared to the cheaper DDS devices made by AD.
The AD9951 needs multi supply voltages, i. e. 3.3V and 1.8V. Digital inputs are 5V-compatible if 3.3V as D_VDD_IO input voltage is applied. The device can be clocked up to 400Mhz. If you use a clock generator with lower frequency, an internal clock multiplier can be used. But this deteriorates phase noise to a certain degree. For my VFO which works in the range only from 13 to 20 MHz I use a simple standard 5V 120MHz clock generator and I do not use the internal clock multiplier. To make the clock oscillator’s output 1.8V compatible a simple voltage divider has been applied.
Note that performance concerning phase noise of the DDS also depends on the voltage of the clock generator. The lower it is, the more the phase noise performance will deteriorate. To calculate exact configuration of your voltage divider keep in mind that input impedance (1.5 kΩ according to datasheet) and input capacitance (3pF) are paralleled. Input capacitance causes reactance depending on the input frequency given by the following equation:
As usual for me the DDS output consists of a balanced-to-unbalanced broadband transformer followed by an rf amplifier. For the latter one I use the MAV-11 broadband amplifier made by Minicurcuits.
This the circuit of my improved DDS device:
The first impression when I connected the AD9951-DDS to my HP 8558B spectrum analyzer was that the signal looked different compared to a AD9850 generated signal. I’m very sorry, but so far I’ve got no photos but I’m planning an article that will deal with DDS comparisons which will have photos of the spectrographic analysis of the various signals. But it was visible on the first sight, that the signal looked much cleaner than a one produced by a DDS with 10-bit-DAC.
OK, there were much lower sidetones to the main signal on the spectrum analyzer. Measurements are one side of the medal, but how would the DDS perform in the receiver? It approved to have improved when I installed the new DDS into the transceiver and listened to the bands. “Birdies” have vanished to nearly 100%, only some very weak spots are discernable. And the receiver noise also seems to have lowered by a certain degree. But I don’t have measured that so far. Measurements are still to be done. OK, let’s check it out the following days (20 meter is very quiet today!) and then I will be able to say if and how the 14-bit DAC.
So, if you want to get the best performance (low number of spurs, low phase noise) out of the AD9951, here are some basic hints:
Stay away far from the Nyquist-Frequency of the chip! Basically this is one third (33%) of the clock rate. I recommend to lower this down to one fifth. So, if you use a clock rate of 100 MHz, don’t let the DDS produce more than 20 MHz!
Don’t use the internal clock multiplier! Use a clock generator with the highest possible frequency!
Use a clock generator that will put out 1.8 Volts pp!
Keep ground leads on your pcb as wide and short as possible!
Decouple AVDD, DVDD and DVDD_I/O effectively!
Set DAC_RSET to a value that DAC current stays lower than 10mA. 3.9k is a recommended value.
The DDS is mounted to a small piece of veroboard using a 48-lead breakout-board for TQFP48 ICs:
Flexible wiring is used to connect the board to the microcontroller. Shielded cable is mandatory for connecting the rf feed to tx and rx mixers.
Underneath you’ll find some code snippets to set the frequency of the AD9951 device. The code has been copied 1 by 1 from my SSB transceiver’s software. Thus modification for your purposes might be neccessary.
After inspiring discussion with a reader of my blog I’ve changed the routines to optimize code concerning performance. The major objective was to avoid intense use of floating point functions because they are slow. Instead I used bitshift operations widely. But there is one floatingpoint calculation left:
fword = (unsigned long) frequency * 35.790648;
The floatingpoint constant 35.790… results from a division of 0xFFFFFFFF by f_clock which is given by the equation of the tuning word (see datasheet of AD9951). This could be converted to a bitshift operation too, if you use a programmable clock oscillator tuned to 134,217,727 Hz. Then the multiplication factor is 32 which can be easily achieved with another bitshift operation.
Thanks for reading!
Setting the AD9951’s frequency output
void set_frequency(unsigned long frequency)
// FQ_UD: PD0 (green)
// DATA: PD1 (white)
// CLK: PD2 (blue)
// RESET: PD3 (pink)
unsigned long fword;
int t1, shiftbyte = 24, resultbyte;
unsigned long comparebyte = 0xFF000000;
//Calculate frequency word
//Clock rate = 120002500
//0xFFFFFFFF / 120002500 = 35.790....
fword = (unsigned long) frequency * 35.790648;
//Initiate transfer to DDS
PORTD &= ~(1); //FQ_UD lo
//Send instruction bit to set fequency by frequency tuning word
//Calculate and transfer the 4 bytes of the tuning word
//Start with msb
for(t1 = 0; t1 < 4; t1++)
resultbyte = (fword & comparebyte) >> shiftbyte;
comparebyte >>= 8;
shiftbyte -= 8;
//End transfer sequence
PORTD |= 1; //FQ_UD hi
//Send one byte to DDS
void spi_send_byte(int sbyte)
// PORT usage
// FQ_UD: PD0 (green)
// DATA: PD1 (white)
// CLK: PD2 (blue)
// RESET: PD3 (pink)
int t1, x = 0x80;
for(t1 = 0; t1 < 8; t1++)
PORTD &= ~(4); //Bit PB2 set to 0
//Set respective bit to 0 or 1
if(sbyte & x)
PORTD |= 2; //SDATA Bit PB1 set to 1
PORTD &= ~(2); //SDATA Bit PB1 set to 0
//SCLK line set to 1 // = set clock line to RISING edge to store bit in frequency word
PORTD |= 4; //Bit PB2 set to 1
x >>= 1; //Shift bit to divide x by 2
Resetting the AD9951-chip:
(A reset must be performed once immediately after your program was started and before you
transmit the first instruction to the AD9951 DDS chip)
// FQ_UD: PD0 (green)
// DATA: PD1 (white)
// CLK: PD2 (blue)
// RESET: PD3 (pink)
PORTD |= 0x08; //Bit PD3 set
_delay_ms(1); //Hold reset line hi for at least 20ns.
PORTD &= ~(0x08); //Bit PD3 erase
This is just to illustrate the technical description of my 5 band QRP SSB transceiver with a little more material. I took some pictures of the final assembly of the rig. This is how it looks when put into the cabinet:
The front panel labelling is made with a text processor painting the text on a piece of grey paper. After this has been printed the labels are cut out and covered with adhesive tape. The tape is slightly larger than the piece of paper so that it can be used to fix the label to the front panel.
The cabinet is bent of 2 halves of 0.5 mm aluminium sheet metal joint by bolt connectors. As I mentioned before, to keep the rig neat in size I’ve used a sandwich construction that you can see underneath. The transmitter board is on top,
the switching board with relays and output low pass filters is centered and the receiver board is sited at the bottom:
To make the construction more rigid I’ve inserted 4 threaded bars from the front panel to the rear plate:
The heat sink is from an old PC where it was used to cool the processor.
The front is a separate unit which holds the microcontroller, the colored display and the DDS unit. This VFO can be seen on the right on the next photo:
I owned an old ATLAS 215 RF transceiver once. This unit was destroyed because by a momentary laps of reason. I put 12 V DC to it with wrong polarity. Sh..! Since this very sad day I’ve always wanted a new 5 bander rig. OK, I could have bought one. Money is not the big problem. But, as a passionate homebrewer, I thought to myself: “Why not building a 5 band trx instead of the monoband rigs that I’ve built so far?“. This decision was made in summer 2015. Now, in late autoumn, this rig s nearly completely finished.
So, let’s tell the whole story…
Building a multi band rig is, as you might conceive, much more complicated than setting up a monobander. Seems sensible. So, first I went through literature to avoid fulminant project failure which is more probable when making serious mistakes that can easily be avoided by careful planning.
What to read? First, “Solid state design for the radio amateur” by Hayward and DeMaw is still a very useful book. Even when sometimes people mail me and say “Oh boy, why are you always bulding all those nostalgia radios with old style components like the ‘741’? Boy, today, SDR is the standard!“. My simple answer then is: “No, it is NOT!“.
To learn more about multi band QRP transceivers I also browsed the web for similar projects. There are some, but often there is no full cicuitry given. Here are two sites that might be interesting: M0DGQ and VK3EPW.
So, after some days of conceiving, I finally had the basic concept in mind. Mixing the modern times (DDS and microcontrollers) with the old style of QRP homebrewing. And what about “SDR”? Sorry, not on the agenda. Based on these prerequisites, here are the main ideas for this radio:
DDS controlled by microcontroller
Colored graphic LCD
5 ham bands (80, 40, 20, 15 and 10 meters) to be covered
10 Watts out, transmitter equipped with bipolar transistors in push-pull mode (lots of CB-types still on my stock or replaced e. g. by ELEFLOW types)
Single conversion superhet receiver with AGC and high dynamic range (diode ring mixer). Must be able to deal with high signal levels on the lower ham bands.
Compact size to make the radio fit for portable, outdoor and holiday use (flight!)
Using “The usual suspects” like NE602, MC1496, LM386, MC1350 to be involved.
To avoid any fuss in recevier and ssb generation design I have again picked up the single-conversion principle that has successfully been applied in my former transceivers which uses only one interfrequency (IF). One of the reasons for this, besides circuit simplicity, is the fact, that a single conversion superhet produces less “birdies” than a double conversion one. Birdies are a particular problem when you use DDS frequency generation. Every DDS produces spurii that can be received with more or less intense signal strength in your own receiver. So I was not keen on having more than the inevitable number of self-reception signals by adding a second interfrequency.
I love to build rigs that are compact and neat. I don’t like the bulky boxes only capable for home use. To make this fairly complex radio not too large, I found that sandwich design would be the best way to save space and make the transceiver friendly for building and service.
Based on the “sandwich-idea”, the rig consists of three different layers:
Switch board with final low pass filters (centered between the two others)
All boards are connected from the laterals by a homemade plug system so that the single Verobaords can be removed and reinstalled into and from the assembly quickly. Here’s an overview picture showing this construction method:
In the middle of the sandwich you can recognize the switching unit from where the plugs are lead to the receiver and transmitter board. The idea behind this on one hand makes the transceiver very compact. On the other hand it is very service friendly. It takes only 1 or 2 minutes for example getting the transmitter board out for changing a component. Pull the side connectors, remove 4 board screws + 3 screws at the rear-mounted heat sink and take out the board. Remounting the board takes some seconds more. But there is no need to disconnect and reconnect an endless number of cables each time you perform a modification. All lines are connected by the side-plugs. And that’s it! One shortcoming should be mentioned: When I built the rig there was a little bit more effort concerning wiring. But it pays!
As board No. 4 the DDS-VFO together with the microcontroller and the LCD have been put behind the front panel.
VFO and LO frequency considerations
In this transceiver the interfrequency (IF) is 10 MHz which is based on the fact that respective crystals can cheaply be purchased in larger quantities to build appropriate ladder filters with. I bought about 50 pieces for around 10 Euros. But other crystal frequencies are also possible. For example computer crystals of 9,832 kHz, standard 9,000 kHz xtals or others in the same frequency region can be used. Taking your IF between 8 and 12 MHz might be the best choice in my opinion.
Hint: Before choosing the final IF and purchasing crytals or ready-made filters it is highly recommended to do some basic calculations to ensure that none of the harmonics either of the VFO or the LO falls into the desired receiving band going together with a certain set of frequencies generated by the two oscillators.
So watch out where the frequencies of VFO and LO are located in the radio frequency spectrum! A spreadsheet software is extremely useful for this purpose, because you can change the IF quickly and see at first glance what you will get out when the DDS is about to generate the needed frequency to bring you to the desired band. Also keep harmonics in mind!
And also keep in mind that the image frequency on every band should not fall into shortwave broadcast bands (e. g. the 49m-band!) where strong commercial stations may appear. OK, sometimes this can’t be avoided but it’s worth thinking of it before you start to build.
As said before, my rig uses a DDS system for obtaining versatile VFO functionality (equipped with an AD9850 made by Analog Devices). To get “a feeling” for this chip I first used one of those AD9850 modules available from China on ebay for e few bucks. The problem: They are more or less crap. at least for RF use. I don’t know what chips they install there, but the board was cheaper than buying a single AD9850 on the free market. No more questions. These boards produce a large number of spurii (aka “birdies”). A much greater number than I ever encountered with my last rigs using an AD9835 DDS chip that is nearly spurii-free. So I decided to buy some surplus original AD9850 chips and a breakout board to which I soldered the DDS chip and tried my own DDS board. It worked from the start. But there are also some spurii so I changed the AD9850 DDS to an AD9834 which is similar to the AD9835 from my other rigs. The AD9834 can be overclocked to about 100 MHz which makes him a candidate to substitude the AD9850 in the DDS. In the end I decided to use a more professional DDS by Analog Devices and came to the AD9951 that is also used in Kenwood’s TS590 transceiver. This one, as you might gues, worked best. Here is the description of this optimized DDS vfo.
Realizing the Transceiver
The DDS-VFO, the microcontroller and the colored LCD
For a long time I have owned some colored LCD cellphone displays from a dive computer project that I had completed some years before. These boards are equipped with an ATMega128 micorocontroller which is capable to deliver enough capacity and calculation power even for extended software. So why not using this surplus material?
First, here’s the scheme of this unit:
Here’s an older picture of the practical realization of this DDS-VFO (running an older version of the software) with the LCD built in behind the front panel of my transceiver:
Why is using a DDS really cool?
What always was a problem in “pre-DDS”-times was to switch the VFO to the desired band. With the old 80/20-meter transceiver concept this was not a big problem, the 5-MHz-VFO did the job on either band. But it things were more complicated when you wanted to cover the other ham bands with only one VFO. Some basic concepts were switiching the coils and/or capacitors or bulding a superheterodyne VFO that mixed a basic VFO signal to the desired frequency. Lots of birdies included, sometimes at least.
With an appropriate DDS no superheterodyne or LC-switching circuitry is required to get the wanted VFO signal for each band. The AD9850 is capable of being clocked with 125 MHz maximum rate. The AD9951 can be clocked to 500 MHz even if 400MHz are nominal by the manufacturer. Due to physical reasons (Nyquist frequency) the highest frequency you can get out of such a DDS is about one third of the clock oscillator frequency what in the first case means about 40 MHz. So you can easily recognize that you can get all the VFO sigs from the one DDS you need to drive an rf transceiver without that VFO fuss well known in the older times. In my rig I achieved to stay very much below the margin of 40 MHz because output voltage due to the frequency calaculation you can see underneath. Keep in mind: Signal quality of the DDS decreases significantly the closer you get to its maximum frequency. Among other reasons this is due to the fact that a low-pass filter is switched to the output of the DDS. I recommend to stay at least 5 to 10 times lower than Nyquist freqeuncy if possible.
For the two lower bands (80 and 40 meters) the operating frequency is calculated by the following scheme:
f = fif – fVFO (I)
For the upper three bands (20, 15 and 10 meters) another equation is used:
f = fif + fVFO (II)
Two effects are desired by using this scheme:
First, there is no requirement to switch the local oscillator because it is always on the “right” sideband. The sideband relay in my rig is without current except from those times when you want to use the “wrong” sideband. Second, the maximum frequency the DDS has to produce is about 19 MHz (for the 10-meter band). This will result in sufficient output voltage out of the DDS chip to drive the TX mixer (SA602) properly. For the receiver I use a diode ring mixer, which requires about 7 dBm injection level. Therefore an additional amplifier is required. The DDS can not drive the diode balanced mixer without amplification! In my AD9951 DDS an additional amplifier has been installed, please look there.
Now let’s look at the microcontroller unit. This time I used a pre-manufactured circuit board that contains a colored LCD together with an ATMega128 µC to do all the digital work. The board is an “D072” produced by German based manufacturer display3000.com. The board has nearly all the ports of the installed ATMega128 to be accessed by the developer. Only some lines of PORT B are reserved for the LCD SPI signals. But due to the fact that the ATMega128 has got plenty of available ports that is not a problem at all.
The cellphone display can be run in 2 different modes. Either you can use 256 or 65536 colors. Because of performance reasons I decided that 256 colors are enough because only one byte per pixel needs to be transferred in this mode. In high color mode on the other hand each pixel contains 2 bytes of data which makes data transfer a bit slower. This mode is mainly for showing photos on the 2.1″ display which I didn’t intend to do.
All neccessary auxiliary moduls are also present on the D072-board including voltage regulators, charge pump and positive/negative power supply for the LCD, the MAX232 chip used for serial communication purposes and so on. Even a PWM modul is installed to make the display light controlable by software. All what you have got to do is to connect the various ports of the microcontroller to your application. These are the DDS chip, some sensors for measurement issues using the AT’s ADC, switches and so on. This is the reason why I drew only the relevant lines for my transceiver to the schematic above.
As I mentioned above, DDS is a nice thing. I only had to find a DDS chip that could cover the frequency range from 3 to about 20 MHz used for my rig. I first chose the AD9850 by Analog Devices.
But some experiments were neccessary to optimize the VFO circuit. First, to increase output voltage, I used a symmetric output with a trifilar wound transformer.
Balanced output to my experience is recommended when using Analog’s DDS chips that supply 2 output ports. Use a simple output transformer (e. g. on an Amidon FT37-43-core, trifilar wound). The transformer is followed by an amplifier. See the full scheme for this one! A low pass filter in between these two sections improves waveform and eliminates harmonics. The signal then is fed into the TX and RX mixers simultaneously. The DDS now puts out 2 Volts p-p on from 3 to 20 MHz giving a pure sine wave with flat signal levelling over the full frequency span.
Hint: If you own a spectrum analyzer you can write a simple test program to produce sine waves over the whole frequency an check the flatness of your DDS this way.
The main thing that makes a multiband transceiver much more complex than a monobander is the neccessity to switch frequency related filters in certain parts of the circuit even if broadband designs are widely applied. There are three sections in which band switching must take place in my rig:
In the receiver section:
At the front end.
In the transmitter section:
At the output of the transmit mixer before first amplifier, and
after the final stage where the low-pass filters for each band are selected.
At the low power end of the chains (front end and tx mixer) where signal levels don’t exceed 1 or 2 volts p-p several ways of switching are theoretically possible:
Diodes as switching elements
Method 1 usually involves stray capacities, so signal seperation is far from being overwhelming. Effective decoupling is a big problem with this on the other hand cheap and handy method.
Alternative 2 ensures very much a higher level of signal seperation but depends on the availability of the respective ICs needed like the 74HC4052 (a multiplexer IC) for example
Because of the fact that from a former model railraod project I still had a large bunch of 5V coil voltage OMRON relays, I decided to use these. Most relays normally tend to have good RF decoupling capabilities because capacities of the unswitched contact pairs are in the 1..2 pF region or even lower. That is absolutely OK for rf designs! Check the data sheet for exact data before buying a certain relay model.
In my transceiver the 3 relays sets for each band (receiver front end, TX mixer output and low-pass filter after final amplifier stage) are connected in series and are driven directly by 12 volts by a simple switching transistor directly controlled by a port line from the microcontroller using a small transistor driver stage (see circuit of microcontroller board). Each relay gets about 4 volts of coil voltage (then drawing about 40 mA) which is more than enough because my OMRON relays start reliable switching from a coil voltage of about 3.3 V. So the transceiver theoretically can be operated down to 10 Volts DC without any danger that the band relays might not switch correctly.
Band switch drivers
The drivers for each band that drive 3 series switched relays per band are controlled by one seperate port of the controller (Ports PE3 to PE7). PE3 is for 80 meters, PE4 for 40 and so on. BC547 transistors are enough because the 5V Omron PCB relays draw about 35 mA. Max. collector current for a BC547 is 100mA.
Measurement section for “environmental” data
A lot of values are detected with sensors in this rig. Several voltage dividers are visible on the upper left part of the scheme. They measure:
Power supply voltage
Temperature of the TX final stage heat sink
RX signal level derived from the AGC
The signals are fed into the various PORTS of the ADC of the µC. Check for the software code that will be soon postet for details.
I have to admit, I underestimated the diffculties that I was going to face when constructing a 3 to 30 MHz broadband QRP amplifier with a power output level of about 10 watts on all the five bands covered. I read a lot of stuff in advance, particularly the ARRL library, to get aquainted with the basics of designing linear broadband amplifiers for rf.
To make clear what I’m talking about, here’s the final and full circuit after hours and hours of testing, improving, testing, conceiving, improving…. 😉
The mere power amplifier strip has, as you can see, got 4 stages:
Final amp (push-pull)
The design is capable of delivering about 20 watts of rf power but I wanted to run it within a well defined distance from its limits to ensure top signal quality. On the other hand I had kept in mind that a broadband design always offers significantly lower overall gain than a monoband amplifier. So there should always be a certain amount of gain in reserve.
You may ask why I used 2 push-pull stages. The answer is: Due to the fact that they are operated in class AB-mode I wanted to keep linearity as high as possible. Class AB amplifiers have got a relatively low standing current (bias) and therefore a higher degree of efficiency than class A. The transistor gets only a low quiescent base current, so linearity lacks compared to class A. But it does not get far as hot as his class A brother.
The solution to the bias-problem is the push-pull circuit. Each transistor of the pair now only amplifies one of the half-waves of the duty cycle. These are seperated by the input transformer and are put together again by the output transformer:
So, in contrast to a class A amp which has to be biased for amplfiying the positive half-wave AND the negative share of the waveform and thus needs a much higher bias current, the push-pull design can be operated with lower bias because it has to amplify only one part of the wave cycle which is now split by polarity into two transistors.
And, because a push-pull amp offers built-in even-harmonics suppression, the production of overtones is limited by this circuit without the need for filters.
After the amp was finisehd I did a lot of research using my old HP8558B spectrum analyzer and found that harmonic suppression is excellent as well as third order products are if you don’t use the full power capabilities of the amplifier strip.
How to make an amp broadband
The main problem with transistor amplifiers (mainly those designed for rf purposes) is the gain-vs-frequency problem. When increasing the operating frequency of such an amplifier you’re about to lose 3 dbs of gain per octave. This means, you get high amplification rates on 80 meters and a very much lower one on 10 meters. Not very nice if you want to achieve a comparable amount of gain on all the bands the transceiver is going to cover.
Next topic: I use CB transistors with a relatively high transition frequency (ft) of about 150 MHz or even higher. This means on 80 and 40 meters you can expect tremendous gain. A strong tendency towards self-oscillation and instabilties is one outcome of this situation. But on the other hand for 15 and 10 meters this high ft is a must!
So, a lot of potential problems ware waiting for the ambitious radio constructor. But, no need to get desperate: There are various cures for the various problems. To make the amp stable and clean I applied the following techniques:
Amplifier stages are coupled by relatively low valued capacitors,
My first mistake before bypassing correctly was to try to bring monoband concepts to a broadband amplifier. Adequate emitter bypassing was one of these. First I used large capacitors in the range of 0.1µF. Much too much, as I found out later. OK, a high valued emitter bypass capacitor hands you back a lot of gain but unfortunately not equally distributed over the frequency span. It’s much better to reduce the value of the emitter bypass capacitors down to some nanofarads. You may ask: “Why is this the case?” Simple answer: The resistance of a capacitor for alternating current is given by
XC = 1/(2*Pi*f*C)
Low value capacitors thus prefer higher frequencies and attenuate the lower ones.
2. The same is true for the emitters bypasses:
The bypass capacitors lets ac flow unresitedly over the paralleled resistor.
3. In addition, negative feedback is applied to the 3 of the 4 stages to reduce gain when the operating frequency is low:
A certain amount of rf energy floats back to the base but is 180° out of phase thus compensating input energy.
4. Emitter degeneration also helps to get gain constant AND lowers distortion:
The unbypassed capacitors causes a voltage drop when emitter-collector current rises. Therefore the voltage between base and emitter lowers (base is biased to a fixed value), the gain decreases.
But, as I said before, all these measures cost a certain amount of gain. That’s why I use a 4 stage amplifier rather than a one with only three stages. The effort, by the way, paid. The waveform of the two-tone test is absolutely top on all the 5 bands. But it was a hard way to go there!
Crucial mistakes and their correction
When I tested the first version of the amplifier, I was deeply disappointed. The signals were bad, distortion was a severe problem. Also I had lots of parasetics particularly on the higher bands and self-oscillation occured as soon as I started to increase the drive coming from the tx mixer. One of the reasons for this was improper shielding of the band filter section that follows the TX mixer. I will talk about this in a later post.
By the way, don’t ask me, if the amp was “flat”. No, it wasn’t. Definetely NOT! That was the point when I began to hate this project. 😉 But then (instead of giving up) I started thinking of what I could have done wrong. Step by step I got closer to my goal…
Construction methods for the test procedures: “Plug and pray!”
To enable me to change the “critical” components like emitter bypass capacitors very fast without taking the soldering iron into my hand I use SIP socket strips:
Advantage: You can run endless tests without being endangered to seriously damage the solder pads of your Veroboard! The needed numbers of pins are cut from the strip with a coping saw or a mini cut-off wheel. Once soldered they remain in the board till the end of days. Disadvantage: If your component has thin wires, these will fall out off the socket. But for 98% of my components this works fine!
Components prone for being optimized by this method are:
Emitter bypassing capacitors
Negative feedback resistors and/or capacitors (whatever you have!)
Optimize your rf transformers!
RF amplifiers in high-power stages (i. e. above the Milliwatt level) usually use transformers for coupling rf energy from one stage to the next. They hereby can serve as impedance transformers, because input impedances of trasistor stages are often in the range of only some ohms whereas the output impedances of the previous stages are 4 to 10 times higher. I own a large number of my favourite Amidon toroid cores so that I can produce a variety of test transformers that can be soldered quickly to soldering nails. Again this is to avoid the repeated soldering process ruining my Veroboards:
By methods like these and 2 months of steady improvement I finally got the transmitter working the way I wanted it to have.
Measurement results at the completed unit:
Output power in two-tone-test: 10 Watts at 11.9 V DC power supply
Carrier suppression: greater than 50dB
IMD3 products: 36dB (measured at 14,200 kHz)
Spectral analysis of output signal (f=14,200 MHz):
Now it’s time to discuss the receiver section of my 5 band QRP SSB transceiver. The main objectives for the receiver were:
Must be able to deal with high signal levels particularly on the 80- and 40-meter-band.
Must be able to seperate strong out of band signals (broadcaster etc.).
Must be able to seperate strong in band neighbourhood signals.
Must have high dynamic range.
Here’s a brief description of the various stages of the receiver board:
5 relay switched 3 pole band pass filters make up the front end,
followed by dual gate MOSFET-preamplifier and a
diode ring mixer equipped with a diplexer,
IF preamp with bipolar transistor,
IF main amp with IC (good ol’ MC1350),
product detector with 2 diodes,
AF preamp with bipolar transistor,
AF final with LM386.
As usual, here’s the circuit first:
Some words concerning the various sections of the circuit:
The front end
On the left you can see 5 relay switched band pass filters. To ensure maximum out-of-band signal suppression I chose 3 pole filters. The effort is a little bit higher, I have to concede, but it’s absolutely worth. No intermodulation or other interference by strong broadcasters close to ham bands (particularly on 40 meters!) occurs.
The filter coils are wound on TOKO style coil formers of this kind. The relays are OMRON G6A-234P pcb relays. They are designed for 5 V coil voltage. Due to the fact that my rig needs 3 sections of relay switched circuitry (rx front end, band filter past the tx mixer and low pass filter past the exciter) I switched the corresponding relays for one band in series. They then are driven by 12 volts controlled by the ATMega128 driving my DDS system. The wiring of the relays is a little bit more complicated as if I had them switched in parallel but in the end this was a nice way of recycling a larger bunch of these relays that I still had on stock from a former model railroad project. 😉
Past the front end filters next stage is the well-known dual gate MOSFET rf preamp controlled by AGC.
The amplifier terminates broadband style (toroid transformer) putting its rf energy into a diode ring mixer which by definition is a balanced mixer circuit. The DDS VFO signal is injected on the other side of the mixer. Please note that you need a preamplfier if you run that mixer type by a DDS because the outputlevel of a DDS (if not amplified) does not suffice the 7 dbm a diode ring mixer needs for proper operation. Therefore I’ve included a small signal amp with a bipolar transistor.
To minimze spurs in your receiver it is of maximum improtance that this VFO amp works 100% linear. Keep an eye on not overdriving the amplifier! Any signal level beyond linear operation condition produces spurious emmissions. If available check the output with a spectrum analyzer! Reduce input voltage by inserting a voltage divider made of resistors (not capacitors!) if input level is too high!
A diode ring mixer also needs to be accurately terminated to 50 Ohms for optimized performance. Thus I’ve added a diplexer after the mixer which ensures an adequate termination of the mixer on the IF frequency.
IF and AF stages
Next steps are an IF preamplifier (which is connected to a manual gain control potentiometer sited in the front panel), a ladder filter with about 2.4 kHz width and the IF main amp equipped with MC1350 by Motorola. The IF amplifier IC is connected to the AGC strip at the end of the receiver section.
The succeeding diode based product detector is fed by the amplified IF and by the carrier oscillator which is also sited on the receiver circuit board.
An audio preamp and the LM 386 as the power amp do the final job of amplifiing of the audio signal together to loudspeaker level.
The audio-derived AGC circuit is the same like in my hand-held transceiver. Two dc outputs are available. One delivers increasing voltage when signal strength is rising, the other decreases voltage under the same condition. The first one is about to control the MC1350, the later one is for the dual gate MOSFET that can be found in the receiver’s front end.
Here, for the final, is an overview of the receiver board in my 5 band QRP SSB transceiver. Thanks for reading!
First QSOs went very fine on 20 meters where my antenna is tuned to. Let’s see what the rig will show the next weeks, I’ll keep you informed. Please watch later posts on this blog that will show adaptions and modifications of this rig. 😉