Modern QRP Rigs and The Development of The QCX CW Transceiver Kit
Modern QRP Rigs and The Development of The QCX CW Transceiver Kit
Modern QRP Rigs and The Development of The QCX CW Transceiver Kit
Transceiver Kit
Introduction
I present an attempt to describe some of the more modern techniques into QRP radio design. The
majority of QRP radio receivers are based on the same superhet architecture that has dominated
our homebrew, kit and even commercial projects for three decades, with some notable exceptions.
More recent components and designs allow creation of transceivers which are simpler, higher
performance and lower cost, all at the same time!
The three main topics presented here are:
• Escaping three decades of the classic SA602 superhet receiver
• Drift-free, flexible, accurate, easy and cheap: the new generation synthesizer chip
• Introducing microcontrollers to your projects
These topics and various other smaller features that may be new to you, are presented in the
context of the QCX 5W CW transceiver kit designed for QRP Labs in 2017, which is a good
example of how to design a practical, low cost, high performance radio using these techniques.
We’ll stick with CW, and won’t go as far as Digital Signal Processing, Software Defined Radio, etc.
The concepts presented here are modern and high performance and yet won’t be too daunting for
the majority of us!
Let us start with a brief description of the QCX 5W CW transceiver kit, so that we know where the
following discussion is leading to.
This 5W CW transceiver kit, named “QCX” (for QRP Labs CW Transceiver), was designed for the
2017 Youth On The Air (YOTA) summer camp “buildathon” hosted in UK in August 2017 by the
RSGB. It is a mono-band CW transceiver available as a kit for bands 80, 60, 40, 30, 20 or 17m.
The YOTA attendees built the 17m version (after all, the year was 2017… why miss an opportunity
like that!).
The design process was very interesting, due to the constraints. The kit needed to be simple
enough to build in a relatively short time, by a group of young people with potentially a wide range
of construction experience… some with none at all. The cost had to be low to match the limited
budget available. At the same time, I was not interested in designing a toy radio. As far as I was
concerned, it had to be a really high-performance radio packed with useful features and a useful 3
to 5W output power. A rig that would also be of interest to radio amateurs all around the world. I
wanted to create something really special for each young attendee of YOTA 2017 to take home
with them and operate, something that would set a new standard in amateur radio kits.
The result is a low-cost $49 kit using all through-hole parts (except two SMD ICs which are pre-
soldered to the PCB). It has very high performance, which competes well with top-of-the range
commercial radio costing 100x as much. The kit also contains all its own test equipment for
alignment and debugging (DVM, power meter, signal generator, RF power meter), so it can
literally be built without any additional test equipment.
It’s a lot of performance and value for $49 and has been greatly appreciated by the QRP
community. At time of writing (March 2018) over 4,200 of these kits have been shipped.
Direct Conversion
Direct conversion utilizes only a single mixer, straight from RF
to baseband (audio). It eliminates a lot of the problems with
superhets, such as alignment, the birdies, and the crystal filter.
Many people are surprised at the nice clean sound of a direct
conversion receiver, compared to the superhets they are used
to.
The main and very serious disadvantage of a simple direct
conversion receiver, is that the upper and lower sidebands are
both received together at the same time.
The unwanted sideband can be eliminated by some clever mathematics, splitting the signal into
two paths. Either the incoming signal or the oscillator paths must have a 90-degree phase
relationship. It makes no difference whether you apply the 90-degree phase shift to the RF signal,
or the VFO; but the VFO is normally easiest. A 90-degree phase shift applied at the audio output
produces two outputs that when summed, magically reinforce the wanted sideband and cancel the
unwanted one.
Well, the equation is not very easy to understand either, as well as being WRONG… to start with,
the [4:0] part specifies that the phase offset is a 5-bit number whereas in reality it is 7-bits.
The description of Register 165 (“Clk0 Initial Phase Offset”) on page 57 of AN619 is more helpful
and tells us all we really needed to know:
“CLK0_PHOFF[6:0] is an unsigned integer with one LSB equivalent to a time delay of Tvco/4,
where Tvco is the period of the VCO/PLL associated with this output.”
Similarly, there is Register 166, Register 167 for the phase offsets associated with the Clk1 and
Clk2 outputs.
So here we can see that phase offset is a 7-bit value. And very simply, it is the number of quarter-
cycles of the internal VCO that you want to apply as phase-offset to the desired output. No more,
no less.
This example is typical of how difficult it is to make headway with the Si5351A using the SiLabs
documentation. Two documents (datasheet, App note), and three different mentions of the phase
offset; the first is ambiguous, the second is wrong and disagrees with the third, later in the same
document. But the good news is that via head-scratching for some days and lots of experiments, I
could figure it all out, and I’m explaining it here.
The Register 165 explanation also then makes it possible to understand the Si5351A datasheet
document description of the 333ps typical phase offset step! This is because the internal VCO
frequency specification (specified in App Note AN619, not in the datasheet!) is 600-900MHz. If a
“typical” VCO frequency can be taken as mid-range, 750MHz, then the period of a 750MHz cycle
is 1/750,000,000 which is 1.3333 nano seconds. Ha! So, the 333ps is ¼ of a 750MHz cycle 1.333
nano seconds! The mystery is resolved.
Now we understand the phase offset register, it becomes easy. The phase offset is configurable
from 0 to 127 quarter-cycles of the internal VCO. Recall that the internal VCO is divided down to
produce the output frequency in the MultiSynth block. If we choose an even integer divider for the
MultiSynth block and set the phase offset number to be the SAME as the divider, then that’s it! We
get 90-degree phase offset of the output. It’s as easy as that.
For a quadrature (90-degee) LO outputs, just configure one output with 0 phase offset register (in
other words, leave it at the default 0); configure the other output phase offset register to be the
same as the even integer MultiSynth divider.
This register named “PLL Reset” is in fact very poorly named and explained, in my opinion. It has
misled (and continues to mislead) many practitioners to believe that the operation of the Si5351A
is similar to the Si570. In the Si570, the synthesiser can only be moved a relatively small
percentage frequency change, then a “PLL Reset” action is required. People thought (and still
think) that the Si5351A behaves the same way. It doesn’t!
When the PLL Reset command is issued (by setting register 177 bits to 1), it causes an
interruption to the synthesiser output frequency lasting around 10 milliseconds. This creates a
horrible CLICK in the receiver audio. Some Si5351A libraries (and even early QRP Labs code)
issued the reset command at every frequency change. When you tune the receiver, you get this
nasty click-click-click as the frequency is swept. Horrible indeed!
The correct use of this register, which is now implemented in all QRP Labs kits, is to do a PLL
reset only once at the beginning, and thereafter every time the MultiSynth “Divider” and phase
offset parameters are set. The mis-named “PLL Reset” seems to set up the MultiSynth dividers
and initialize the phase offset according to the values in the offset registers. It might have been
more accurate to call it a “MultiSynth Reset” in my opinion.
Once you have done this “Reset” the phase offsets are magically exactly as you asked them to be.
You may now tune the VFO by loading in new PLL feedback dividers (fractional). You can use as
The schematic segment shown is how the audio phase shift is obtained in the QCX transceiver. It
is an active two-path all-pass phase shift network using four operational amplifiers. The circuit is
based on the same phase shift block as the Norcal NC2030 design by Dan Tayloe N7VE and
elsewhere, see http://www.norcalqrp.org/nc2030.htm
Theoretically four poles give four nulls of “perfect” unwanted sideband cancellation. In the real
world, nothing is perfect – there are component tolerances to think about. The unwanted sideband
suppression is maximised when the amplitude of the two paths is equal, and the 90-degree phase
shift is accurate.
To improve the accuracy of the 90-degree phase shift, R17 and R24 multi-turn trimmer
potentiometers allow adjustment of the phase shift at higher and lower audio frequencies
respectively. R27 allows adjustment of the balance between the I and Q channels, to equalise the
amplitude from each path.
The QCX kit includes built in alignment and test equipment, with a signal generator that can inject
a test signal into the receiver input. It makes it easy to perform these adjustments, which will be
described later. The adjustments make it possible to achieve better than 50dB of unwanted
sideband cancellation across the narrow CW bandwidth of interest, as shown by the following
QCX receiver measurement.
There are advantages and disadvantages of the polyphase approach. Theoretically it is less
sensitive to component tolerances than the above 2-path active all-pass method; in other words,
for any given component tolerance level, the polyphase method would provide a more accurate
audio phase shift (and hence better unwanted sideband rejection). But the polyphase method is
hard to adjust, and unless carefully designed can exhibit a non-flat frequency response.
There are three stages of low-pass filtering and one stage of high-pass filtering. The first three
stages retain the 2.5V “midrail” bias all the way through from the input transformer T1. The final
stage IC9A is biased using the 5V supply (avoiding a few extra components to create a real 6V
midrail at half the supply).
Sidetone is generated by the microcontroller and is a squarewave. To my ears, a 700Hz
squarewave sounds horrible. So, to make it sound nice and clean, the sidetone injection is done at
the input to the 200Hz CW filter, which cuts off any of the squarewave harmonics leaving a clean
700Hz sinewave sidetone.
The HI-PER-MITE CW filter also provides a measured 18dB of gain.
This is the measured response of the CW filter, normalized to 0dB at the center frequency.
There is no “real” audio amplifier IC in the QCX receiver. An important revelation for QRPers is
that for headphone use with popular 32-ohm earphones for example, an op-amp is capable of
providing sufficient drive by itself. There is no need for a “proper” audio amplifier IC.
If you wish to drive low-impedance headphones or a loudspeaker, then use an audio amplifier IC –
but I recommend avoiding the LM386! It is another 3-decades old device (at least) which has been
superseded by more recent devices which perform better, as well as being cheaper.
Microcontrollers
Throughout the text so far, I keep referring to “the microcontroller”. A microcontroller is a single IC
which integrates a microprocessor, some memory, and some peripherals all on one chip. It’s like a
miniature computer-in-a-chip. We have GOT to have a microcontroller, because if nothing else, we
need something to control the Si5351A synthesizer chip (or DDS or Si570, if you prefer those). For
those people who have not used microcontrollers, and perhaps secretly or openly FEAR them…
you really should give it a try! Today there are easier and lower cost ways to get started than ever
before. There is a wealth of information and advice out there.
Anyone can do it! My Grandfather introduced me to his ZX Spectrum 8-bit home-computer in the
1980’s and continued with subsequent machines until his passing (aged 84). These days, my
daughter (age 4) takes a table to her pre-school for elementary coding classes. If you’re aged 4 to
84, I know you can do it. And I bet outside that range too.
LCD Module
These 16-character 2-row LCD modules
(known as 1602) are available very
inexpensively. If you want a frequency display
on your radio, it is no longer necessary to wire-
wrap a bunch of discrete logic chips and 7-
segment LED displays. The ubiquitous
HD44780 controller standard is implemented
on practically every LCD module you can find.
In the QCX transceiver kit, the microcontroller is already telling the Si5351A what frequency to
produce, by setting its register configuration. It is
a simple matter to write this information to the
screen at the same time. This picture shows the
information which can be shown on the QCX
transceiver display.
These display modules can operate in 4-bit or 8-
bit mode. The 4-bit mode is preferable because it
uses fewer processor I/O pins. Display modules
with an I2C interface are also available, albeit slightly more expensively. These can generally be
wired in parallel with other I2C devices such as the Si5351A because in general each device will
have a different 7-bit I2C address. Therefore, they require NO additional I/O pins at all!
Sidetone generation
All modern microcontrollers contain a number of timers, which can also be used as Pulse Width
Modulation (PWM) generators to certain output pins of the microcontroller. Both the frequency and
the pulse width can be varied by the microcontroller.
This presents a simple and ideal way to implement sidetone in a modern transceiver design. The
actual frequency can be configured at the whim of the operator. The sidetone volume can also be
adjusted, simply by altering the pulse width.
Any regular stream of pulses (rectangular waveform) can be expressed as the sum of an infinite
series of sinewaves, at the fundamental frequency and all the harmonics (with varying
amplitudes). In the QCX transceiver the sidetone generation is injected into the audio signal path
BEFORE the 700Hz CW filter. This neatly removes all of the harmonic content of the rectangular
waveform, leaving only the fundamental sinewave. This is pleasant to listen to, but it also provides
a sidetone volume adjustment at the command of the microcontroller. Maximum volume occurs at
50% duty cycle. Reducing the duty cycle reduces the amplitude of the fundamental sinewave, and
therefore the volume of the sidetone. In the QCX transceiver the sidetone volume is adjustable in
100 levels from 0 to 99. The pulse width must be calculated from the chosen volume via a
logarithmic function because of the way the human ear perceives sound.
CW Decoder
A CW decoder is a nice feature to have, for some people. Others hate it. If you fall into the former
category, there are a large number of CW decoder examples available for the Arduino platform.
Many of these are variations on a project by Hjalmar OZ1JHM http://skovholm.com/cwdecoder
and indeed this project was my inspiration for the CW decoder in the QCX transceiver kit. In the
end, I discovered some problems with this decoder (which I may call bugs), and several
2) RF Power meter: Addition of a diode detector (D4, C42) and scaling potential divider (R56,
R58) creates a simple RF power meter measuring 0 to 5W. The firmware measures peak
voltage, but must account for the diode voltage drop, and must then calculate the RF Power
assuming a 50-ohm dummy load. Pin 2 of the header can be connected to the RF power
output of the transmitter to make a reasonable measurement of the transmitter power.
3) Signal generator: In this function the Si5351A is just operated on its own as a free-standing
signal generator, covering the entire range 3.5kHz to 200MHz. The Clk0 and Clk1 outputs
are operated in quadrature (90-degree phase shift) while this is possible, i.e. while the
frequency is above 3.2MHz. No buffering is included, just the raw Si5351A outputs. They
are quite robust but not infinitely so. A buffer amplifier would be advisable too perhaps. The
last thing you want to be doing is replacing that tiny 10-pin SMD
chip (I speak from painful experience).
Above is the block diagram of the QCX CW transceiver. The full schematic is in the 141-page
manual on the QCX web page http://qrp-labs.com/qcx . The QCX CW transceiver kit costs $49
from QRP Labs.
Conclusion
I hope this article has generated some ideas for you to use in your own homebrew projects.
Microcontrollers can be incredibly useful and you won’t know what you’re missing until you try it,
then you’ll never be without one again. The Quadrature Sampling Detector (a.k.a. “Tayloe
Detector”) offers high performance at low cost, and an escape from the classic SA602 superhet
receiver. Modern synthesizer chips such as the Si5351A make building a flexible, accurate, drift-
free and cheap oscillator 60dB easier than ever before.
Happy homebrewing and happy QRP!