AM/FM/CW
Scanning HF/VHF RF
Signal Generator
This low-cost, easy-to-build and user-friendly RF signal generator covers from
100kHz–50MHz and 70–120MHz, and is usable up to 150MHz. It generates
CW (unmodulated), AM and FM signals suitable for a wide range of tests. Its
output level is adjustable between -93dBm and +7dBm and it has an accurate
frequency display. It also includes a scanning function for filter alignment.
I
’ve always wanted a good
AM/FM HF/VHF signal generator. I
have tried to meet that need with a
variety of designs over the years, some
analogue, others using DDS chips.
Recently, I have tried low-cost
fractional-N oscillator chips, includ-
ing the Si5351A. These were only
suitable in specific circumstances,
and did not make for a good general-
purpose test instrument.
Obviously, it’s possible to purchase
an RF signal generator, new or used,
but I couldn’t afford the price of a
good one. Cheap signal generators
lack adequate performance and useful
functions. Those with adequate perfor-
mance are usually too expensive for
most hobbyists, or are unreliable and
difficult and/or expensive to maintain.
I have seen some designs published,
but these are typically simple analogue
LC-based designs with coverage up to
around 150MHz, in a series of five or
six switch-selected bands.
14
Most lack accurate frequency read-
outs or adequate stability. Spurious and
harmonic outputs can also be a problem.
(See the list of references at the end
of this article for three such designs that
I considered and rejected).
Table 1 (overleaf) shows what is avail-
able at the moment. I rejected all of these
options for one reason or another – in-
adequate performance, lack of features,
high price or unreliability.
With few exceptions, the output lev-
els of most of these generators are quite
limited. Those with a variable output
level typically use a simple potentiome-
ter, with little regard to varying output
impedance or accuracy.
Output levels are also often too low
for use in many typical applications.
Modulation, where available, is often
limited. And, finally, some otherwise
useful digital-based designs are now
difficult or impossible to build due to
obsolete parts or unavailable software
or PCB layouts.
Part 1
by Andrew Woodfield
ZL2PD
Basic analogue and digital PLL-based
RF signal generators are available be-
tween about £100 and £150. The ana-
logue generators offer basic CW, AM or
FM modulation. Output level and mod-
ulation depth on the low-cost analogue
generators are typically controlled via
internally mounted trimpots adjusted
through small holes in the panel.
The low-cost digital signal generators
only offer FM and appear aimed at the
two-way radio industry.
These instruments are perfectly func-
tional, but for hobbyists, the features
are too limited. To use them effectively,
you’d need extra equipment such as a
frequency counter, attenuators, amplifi-
ers and a level meter. It’s far easier to have
these features built into the generator.
As Table 1 shows, moving up in the
market significantly increases the price.
Used equipment is available at lower
cost, but many otherwise excellent in-
struments have recognised spare parts or
reliability issues as the equipment ages.
So I needed to come up with my own
design that would tick all the boxes, and
that is just what I have done. See the
table opposite which lists its features
and performance figures.
Design goals
This design represents the outcome of
an extended period of development and
testing over the last few years.
This signal generator provides basic
CW (unmodulated) signals, plus AM
and FM modulation functions, primar-
ily across the high frequency range from
100kHz to 30MHz, with a continuously
variable output level suitable for most
requirements.
This frequency range includes most
common IFs (intermediate frequen-
cies) such as 455kHz, 465kHz, 470kHz,
10.7MHz and 21.4MHz.
Practical Electronics | June | 2020
 Coverage extends to 50MHz, with an-
other range covering 70-120MHz. Cover-
age actually extends up to 150MHz with
some limitations, to permit limited use
in the popular 2m amateur radio band,
as well as parts of the widely used inter-
national 138-174MHz land mobile band.
Key design objectives included low
cost, ease of obtaining parts and ease
of construction. Special parts, such as
chip-based attenuators, for example,
were avoided in favour of the low-cost
combination of slide switches and
standard resistors.
The generator’s RF output is designed
for applications requiring relatively
high RF levels. These include testing
double-balanced diode mixers in high-
performance receivers and for testing
multi-stage passive filters, where stop-
band attenuation measurements require
relatively high signal generator outputs.
Lower RF output levels are also use-
ful, eg, for receiver sensitivity tests.
The minimal useful level is mostly
determined by the limitations of low-
cost shielding and the simple hobbyist
construction methods used.
If an enclosure was carefully milled
from a 25mm thick metal billet with
shielding slots for flexible conductive
inserts, the lower limit could be ex-
tended significantly, but relatively few
hobbyists could achieve this. So I’ve
used simple shielding and a basic DIY
folded aluminium sheet metal box. This
is reflected in the modest lower output
specification limit of around −90dBm.
This is the low-cost AD9850-based DDS signal generator used in this design.
Besides the chip, it has a reference osciallator (the metal can at left) plus a
number of discrete components, including a low-pass filter for the output.
Achieving that performance, how- provide the desired functions and
ever, still requires moderately careful performance required at a modest cost.
enclosure construction. The cheapest digital options include
By using commonly available parts the powerful Silicon Labs Si5351A
and low-cost modules, I have been able device or widely available direct
to keep the overall cost low. I estimate digital synthesis (DDS) modules based
the cost to build this signal generator on chips such as the Analog Devices
currently at around £40. AD985x (see our article on the AD9850
in the September 2018 issue of PE.
Design approach Other digital options include PLL
As shown in Fig.1, a modern signal chips such as the Maxim MAX2870.
generator consists of five functional While it is certainly possible to generate
blocks: the RF oscillator, modulator, sinewaves from both the Si5351A and
RF buffer amplifier, variable attenuator the MAX2870, the additional circuitry
to control the output level, and some required to obtain low harmonic con-
control electronics. The logical imple- tent output signals, coupled with the
mentation of the control electronics challenges of adding modulation, make
is based on a microcontroller. The them less attractive.
final block is the power supply, either AD9850 DDS modules (shown
battery-powered or mains-powered above) are available from ebay and
(or both). AliExpress at reasonable prices.
The oscillator is a key element of any The instrument’s display require-
signal generator. An analogue-based ments are modest, so I decided to use a
wide-range oscillator and modulator common 16×2 character alphanumeric
involving sets of inductors and a tun- LCD. These are easy to read and drive
ing capacitor is impractical and can’t from a micro.

Features and specifications

Specification Comments
Coverage 100kHz-50MHz, 70MHz-120MHz Usable up to 150MHz
Tuning steps 10Hz to 1MHz in decade increments User selected
Accuracy and stability Within 150Hz at 30MHz (typical), 0-40°C, 0-80% humidity Can be enhanced with software calibration
Output level −93dBm to +7dBm (approximate) 50termination
Attenuation steps 0-80dB in 20dB steps (switched) + 0-20dB (variable)
Output socket SMA
Spurious and harmonics Typically better than −30dBc Within specified coverage frequency range
AM 30% modulation @ 1kHz
FM NB (12.5kHz spacing), 1.75kHz deviation @ 1kHz (60%)
WB (25kHz spacing), 3kHz deviation @ 1kHz (60%)
BC (12.5kHz spacing), 50kHz deviation @ 1kHz (60%) Suitable for standard broadcast FM receivers
Scanning Programmable start and stop frequencies 1kHz resolution
10, 20, 50, 100, 200 or 500 steps/sweep Auto step-size calculation
Display 16x2 alphanumeric LCD
Power control Soft on/off switch
Controls Two knobs and eight switches
Power supply 9-12VDC at 250mA
Dimensions 160 x 110 x 25mm (excluding knobs)
160 x 110 x 45mm (including knobs)
Weight ~250g

Practical Electronics | June | 2020 15

 attenuation range. It works well into
mid-VHF frequencies.
The input impedance is main-
tained reasonably close to the desired
50across the adjustment range of the
potentiometer, so the attenuation is pre-
dictable. The output match to 50as the
potentiometer is adjusted is not perfect,
but it’s an acceptable compromise for
this design.

Amplitude modulation
with the AD9850
A key objective of the signal generator
was to deliver both amplitude (AM) and
frequency modulation (FM), as well as
providing an unmodulated RF signal.
Amplitude modulation with the
AD9850 is well documented. Analog
Table 1: I looked at a range of currently available commercial equipment, both Devices, the chip’s manufacturer, help-
new and used. However, for anything that had better-than-mediocre performance, fully published an application note
that third column definitely caused me some heartache! I estimate the instrument
(AN-423) which describes adding a
described here could be built for not much more than £40, plus case.
small-signal NMOS FET and a few ad-
A rough outline of the design began transistor becomes warm during use, ditional parts to do this. A quick test
to take shape and, adding up proces- but a heatsink is not required. confirmed that it works as described.
sor pins required, the very common The design of the attenuator stage Most signal generators use a 1kHz
ATmega328P 8-bit microcontroller also posed some challenges. Recently, modulation tone, which can be pro-
appeared suitable. While an Arduino PE4302 30dB step attenuator chips have duced in several ways. One approach
was briefly considered, I would need to become popular. While only relatively is to use the ATmega328 to generate a
use practically every pin on the device, new devices, these have recently been 1kHz square wave using one of its inter-
and I wanted to keep the instrument listed by the manufacturer as obsolete. nal timers and then filter this to give a
compact, so I decided to use a stand- The replacement devices, while having 1kHz sinewave. But extensive filtering is
alone ATmega328 processor. improved performance, also come at a required to obtain a suitable tone. That
The RF buffer amplifier requires only substantially increased price. involves quite a few extra parts.
modest gain. It must handle the some- Relay-controlled fixed attenuators can A second, similar approach is to use
what unusual 200output impedance be used, but with an eye on cost and the ATmega328’s counter/timer in its
of the AD9850 module and the following simplicity, I decided to use inexpensive pulse-width modulated (PWM) mode.
50attenuator stages and 50output. slide switches instead. Experience has The resulting waveform is closer to
Another consideration is that the buffer shown these to perform adequately for a sinewave but still requires some
should not be overloaded by the some- this type of application. However, these filtering to remove the 31kHz PWM
times high output swing of the AD9850. limit the attenuator steps to specific at- frequency. Usefully, that filter is far less
Numerous designs published on the tenuation values. Ideally, the generator complex given the much higher clock
Internet suffer from this problem. should have a fully variable output level. frequency compared to the 1kHz tone.
The buffer should also maintain its So I decided to build and test a A third option is to build a discrete
gain across the design frequency range. Serebriakova attenuator as an alterna- 1kHz sinewave oscillator and just use
And the buffer should be able to work tive to a more costly PIN diode-based the ATmega328 to turn it on and off as
into a reasonable range of loads and design. This configuration is shown in required. At first glance, the discrete
survive typical bench treatment. the lower right-hand corner of Fig.4, oscillator approach is attractively
I’ve used MMIC amplifiers such as the the circuit diagram. simple and uses relatively few com-
ERA-series devices from Mini-Circuits It’s a simple passive resistor net- ponents, so I tested this out, using the
to buffer AD9850, AD9851 and AD9854 work that acts as a variable attenuator, circuit shown in Fig.2.
DDS chips in the past. These drive well suited for basic designs like this. It works quite well. The 3.3nF ca-
50loads with good performance. Apparently of Russian origin, the pacitor value can be adjusted to give
However, in testing this signal gen- attenuator network uses a 500linear the required modulation level at the
erator with a wide variety of filters, potentiometer to give a 20dB variable AD9850’s RF output. This works by
amplifiers, receivers, transmitters and
other loads, several MMICs suffered
early deaths. These were probably due
to the very low impedances presented
by some of the test filters.
The search for a more suitable buffer
stage was ultimately concluded with
the inclusion of a traditional single-
stage buffer amplifier using a robust
2N4427 VHF transistor. It is widely
available at low cost, as is its near-
equivalent, the 2N3866. It proved more
than adequately robust over many Fig.1: the basic arrangement of a modulated signal generator with adjustable
months of use. The TO-39 case of the output level. Our design follows this configuration.

16 Practical Electronics | June | 2020


replacing the fixed resistor (RSET) on
pin 12 of the AD9850, typically 3.9k,
with the variable resistance of Q2’s chan-
nel. This resistance sets the AD9850
digital-to-analogue converter (DAC)
current and, subsequently, the AD9850
RF output level.
By varying the gate voltage of the
2N7000 at 1kHz using the voltage
from the collector of audio oscil-
lator Q1, the AD9850 RF output is
amplitude modulated. However, this
analogue tone is not precisely 1kHz.
Its frequency is determined by the
passive components around Q1. To
give a more accurate (and potentially
adjustable) modulation frequency, the
PWM-based approach was used in the
final circuit. See the ‘OUTPUT LEVEL
CONTROL’ section in Fig.4.
Pin 11 (output PD5) of IC1 produces
the 1kHz sinewave as a 31kHz PWM
square wave, or potentially at other fre-
quencies by changing the software. This
is filtered and used to control a current
sink made using standard NPN transis-
tors. An extra 100nF bypass capacitor
was added to pin 12 to the final PCB to
address AD9850 module stability.
The 31kHz pulse-width modulated
1kHz signal is produced by the AT-
mega328 from its 8MHz internal RC
oscillator. The variable DC voltage of
0-5V arriving on the base of Q1 is con-
verted to a variable collector current in
Q1 of 0-700µA, the maximum current
value being set by its 1kemitter resis-
tor. This figure was selected to exceed
the 625µA maximum current sink range
required by the AD9850.
This approach is not perfect. Using
the RSET pin and the standard unbal-
anced RF output from the AD9850
module, the typical approach used in
these low-cost modules, the output
modulation produced is asymmetric.
In practice, however, this does not
matter terribly.
This simple circuit delivers clean-
sounding amplitude modulation with
Practical Electronics | June | 2020
Fig.2: a typical
example of how you
can apply amplitude
modulation to
the output of an
AD9850-based signal
generator module
using discrete
components.
In the end it was
decided to abandon
this idea in favour
of a PWM-based
microcontroller
approach.
the AD9850 and uses fewer components
than the other options. It also allows
other modulation tones to be added in
future if required. Finally, this approach
also adds another important feature –
reasonably accurate linear control of
the AD9850 RF output level.
Note though that this approach re-
quires the removal of that 3.9kresistor
from the module as supplied, and the
addition of a wire to control pin 12 from
Q1 to one of its pads. This change will
be described in more detail later.
Frequency modulation (FM)
Again, there are several options to
produce FM with the AD9850. One ap-
proach would be to externally modu-
late the AD9850’s separate 125MHz
reference crystal oscillator. Frequency
and phase modulation could be both
implemented this way. Unfortunately,
the 125MHz reference oscillator in the
low-cost modules is inside a sealed
metal can.
There is no external voltage tuning
input which might otherwise be pressed
into use to produce FM. It’s possible to
replace the reference oscillator module
with a discrete oscillator to allow for
external modulation, but that takes
some effort. It is also possible to use the
AD9850 internal phase modulation reg-
ister but resolution is too limited (4 bits).
Fig.3: the output of a DDS signal generator module contains the wanted
frequency plus a number of alias frequencies. These are normally filtered out
but it is possible to instead filter out the fundamental frequencies and keep one
of the higher alias frequencies to extend the signal generator’s range.
What is Frequency
Modulation (FM)?
With frequency modulation, the audible
tone of (say) 1kHz results from the carrier
frequency of the signal generator being
instantaneously shifted (or ‘deviated’)
from its nominal frequency in proportion
to the amplitude of the modulating tone.
As the amplitude of the tone increases,
at that 1kHz rate, the carrier frequency of
the generator proportionally increases.
Similarly, as the 1kHz tone’s amplitude
decreases, the carrier frequency is pro-
portionally decreased. It is proportional
because the extent of the carrier frequency
shift, or deviation, depends on the signal
bandwidth required.
For broadcast radio FM, the peak devia-
tion is ±75kHz. The resulting signal fills the
standard FM broadcast channel bandwidth
of 200kHz. Traditional VHF FM two-way
radio transceivers used for amateur radio
or commercial/government mobile radio
use a much smaller ±5kHz deviation, and
these signals occupy 25kHz channels.
More modern so-called ‘narrow-band’
amateur FM transceivers typically use
±2.5kHz deviation, and these use more
densely-packed channels spaced apart
by 12.5kHz.
Another Analog Devices application
note (AN-543) suggests a solution. It
describes a powerful Analog Devices
DSP chip which samples incoming
stereo audio at 48ksamples/sec and
then sends a stream of 40-bit frequency-
setting words serially at very high speed
to the AD9850.
Each of these 40-bit words programs
the AD9850 to a new instantaneous fre-
quency, which is necessary to emulate a
stereo FM signal (including the 19kHz
and 38kHz pilot tones).
With some care and a few lines of
assembly code for speed where neces-
sary, the ATmega328 can modulate
the AD9850’s output frequency in this
manner. Sadly, the resulting modulation
sounds pretty average. The problem is
the time required by the ATmega328 to
17
send the serial string of 40 bits to the
AD9850 each time its frequency has to
be updated for frequency modulation
via the typical 3-wire interface.
The poor result is not surprising. With
the conventional serial load method and
our 8MHz, 8-bit chip, it is (just!) possible
to load four modulation samples per
1kHz cycle into the AD9850. A four-
point sinewave is actually a triangle
wave, which is full of harmonics!
Closer study showed that there is
another way to communicate with the
AD9850 chip. Almost every AD9850/51
based design uses the three-wire serial
bus to send 40-bit control words to the
AD9850 each time the frequency needs
to be updated.
However, the AD9850 can also be
controlled using a parallel interface.
This requires sending five 8-bit words
in quick succession to the chip, along
with some control signals via two or
three additional pins. The only pub-
lished example I could find is based
on a PIC processor.
There is a considerable advantage
in this method. Rather than taking
about 250µs for the ATmega328 to load
each 40-bit word serially, the parallel
approach can reduce this to as little
as 2.5µs.
With the parallel loading method, it
is possible to send 20 samples per 1kHz
cycle without any trouble at all, even
with the (relatively) slow 8MHz clock
in the ATmega328. This is much closer
to a proper sinewave. The difference is
clearly audible in an FM receiver. The
parallel method gives a demodulated
signal that sounds very clear and clean,
just like a sinewave should.
So for FM, the 20-point sampled
waveform is created by calculating the
required AD9850 output frequency
every 50µs and sending that data over
the fast parallel interface.
The FM deviation is controlled by
changing the magnitude of the frequen-
cy changes which occur 20,000 times
per second (20 points × 1kHz).
Selecting narrow band FM (the LCD
shows ‘FM-NB’) on this generator
for 12.5kHz spacing for FM two-way

AD9850-based CW/AM/FM HF/VHF Signal Generator

Fig.4: along with the 16×2 LCD module, the ATmega328P microcontroller (IC1) drives the AD9850 signal generator
module using an 8-bit parallel bus plus three control lines. This allows it to modulate the output frequency at 20kHz,
which results in clean 1kHz frequency modulation. Amplitude modulation is applied using PWM from pin 11 of
IC1, which is filtered and then controls a current sink comprising transistors Q1 and Q2. The resulting current flow
controls the signal generator output level. The output signal is buffered by transistor Q3 and then passes four switched
20dB attenuators and then a 0-20dB variable attenuator (VR2) which gives a 100dB overall output range. Q4 and Q5
form a ‘soft power’ switch for the circuit, which is controlled by pushbutton switch S3.

18 Practical Electronics | June | 2020

radios produces ±1.5kHz FM; select-
ing wideband FM, for older 25kHz-
channel-spaced two-way radios, gives
±3kHz FM (‘FM-WB’), while selecting
broadcast FM produces ±50kHz FM
signals (‘FM-BC’).
Frequency scanning
A further feature was added for test-
ing and aligning filters. For example,
while designing this Signal Generator,
I was also building a 9-band HF trans-
ceiver. Its receiver front end features
nine sets of coupled tuned circuits,
each requiring careful alignment, with
three or four adjustments per set.
In the scanning mode, the generator
briefly produces a signal on a series
Practical Electronics | June | 2020
of discrete frequency steps across a
defined range. For the transceiver ex-
ample, the Signal Generator could be
programmed to produce signals across
each of the nine bands used for the
bandpass filters being tested.
By monitoring the amplitude of the
resulting output from each filter on an
oscilloscope, it is possible to quickly
align each filter while seeing the im-
pact of every change. This forms, in ef-
fect, a ‘poor man’s spectrum analyser’.
This saves considerable time and effort
over manual alignment methods.
The start and stop frequencies can
be set anywhere across the range of
the signal generator. Since filters are
generally fairly broad, a 1kHz step size
for setting the start and stop frequency
is acceptable.
I decided to add a SCAN pushbutton
to the design, to enable this mode. As I
had run out of pins on the ATmega328,
I used two diodes (D1 and D2) so that
pressing this button is effectively
equivalent to pressing the two existing
buttons (MODE and STEP) simultane-
ously. The micro can detect this as a
press of the SCAN button – see Fig.4.
Expanded frequency coverage
Typical AD9850 modules are fitted
with a 125MHz reference oscillator.
DDS oscillators deliver clean sine out-
puts up to about 30% of the reference
frequency; in this case, say 40MHz.
19
 NAVIGATING THE MENUS
Starting frequency and mode
Press ‘MODE’ to select next mode (AM)
Next press selects narrowband FM
Twice more selects broadband FM
(wideband FM not shown)
Once more selects SCAN mode
MODE button
Pressing SCAN selects ‘start’ frequency
(Adjust with “tune/step”)
Pressing SCAN again selects End;
then Steps
Pressing SCAN again starts Scanning
SCAN button
MODE button
Increasing but acceptable levels of
aliasing products are present in the
output spectrum up to 45% of the
reference frequency, say 50MHz.
Beyond this, as the output frequency
approaches the Fourier limit of about
60MHz, spurious products render the
output unusable.
Cheap modules are usually supplied
with an onboard elliptical low-pass fil-
ter with a cutoff frequency of 70MHz to
maximise the output frequency range.
20
In fact, these modules have three out-
puts. The first is the filtered output as
described. It appears on my module
on the pin labelled ‘SINB’.
An adjacent pin, ‘SINA’, might appear
to be similar. However, this signal comes
directly from the AD9850 DAC. It is a
180° phase-shifted (inverted) version
of the signal at SINB but without any
additional low-pass filtering.
The third available output comes
from an internal comparator in the
AD9850. It produces a square wave ver-
sion of the output. This is output level
dependent, the duty cycle being set by
adjusting a miniature trimpot on the
module. If it is adjusted for a good 50%
duty cycle output at a lower frequency
setting, it tends to be less accurate at
higher frequencies.
There is little difficulty in obtaining
reasonably clean filtered signal genera-
tor outputs up to 50MHz from the fil-
tered (SINB) pin. Some testing showed
that output was acceptable down to
100kHz. That’s useful for covering re-
ceiver intermediate frequencies (IF) and
IF filters between 455kHz and 470kHz,
for example.
Looking more closely at the module,
the second SINA output looked poten-
tially useful too. Because this output
is not filtered, the full set of DDS alias
frequencies are available here.
In one example, illustrated in Fig.3,
the ‘wanted’ output (labelled Fout) is at
30MHz. As the user increases this fre-
quency, tuning towards 35MHz for ex-
ample, this output frequency increases,
shown by the blue arrow.
At the same time, the AD9850 (like
all DDS chips) also produces ‘alias’
frequencies. These are shown in orange.
The nearest is at 95MHz, ie, the clock
frequency of the DDS (125MHz) minus
30MHz. It decreases in frequency as the
user tunes from 30 to 35MHz, ending up
at 90MHz (ie, 125-35MHz).
There are many other alias frequencies
which are produced simultaneously, the
next nearest being at 155MHz (the clock
frequency of 125MHz plus 30MHz),
with others at 220MHz, 280MHz and so
on, theoretically continuing forever. The
direction these alias outputs tune can be
seen by the direction of the arrows, some
rising while others reduce in frequency
as the primary frequency is increased.
The amplitude of all of these signals
follows a strict mathematical relation-
ship, called the ‘sine x upon x’ curve.
That’s shown in green on the figure.
There’s about a 10dB level difference
between the 30MHz output and the
95MHz alias signal, for example.
That’s the reason for the substantial
onboard filter on the AD9850 module.
It’s a low-pass filter designed to cut off at
70MHz, so the majority of these aliased
products do not appear at the SINB out-
put. However, since there is no similar
low pass filter on the SINA output, these
alias signals are all usefully present, in
full, at this pin.
As the user continues to tune the
AD9850’s output upwards in frequency,
the ‘wanted’ and first ‘alias’ output ulti-
mately coincide and pass each other at
Fout = 62.5MHz.
A few tests using this SINA pin sug-
gested that the usually unwanted alias
frequencies above 65MHz could be
obtained from the module using an ex-
ternal high-pass filter (HPF). That would
allow the signal generator to provide
useful outputs from, say, about 70MHz
up to about 120MHz. With additional
filtering, still higher aliasing products
could be filtered out and amplified.
This permits the generator to pro-
duce signals across the 2m amateur
band or across part of the 138-174MHz
land mobile bands. As it turns out,
useful outputs across these bands
could be obtained just from using a
single HPF, and the maximum tun-
ing frequency for the signal generator
was therefore set at 150MHz. Those
wanting other bands or fewer aliasing
outputs can modify the HPF to suit
individual requirements.
Detailed circuit description
The final circuit arrangement is shown
in Fig.4. While it may appear complex
at first glance, this design uses remark-
ably few components given the range
of modulation modes and coverage it
provides. Some of the complexity is
hidden in the software for IC1.
To enable the frequency modulation
described above, the AD9850’s 8-bit
data port (pins D0-D7) is connected
to micro IC1’s PORTB digital outputs
(PB0-PB7). The three 10kseries resis-
tors have been added so that IC1 can
be reprogrammed in-circuit (via ICSP
header CON3) while IC1 is still con-
nected to MOD1.
MOD1 is also connected to 5V
power (VCC) and GND, plus the
slave select (SS) and reset (RST) pins,
which go to digital I/Os PC4 and PD4
on IC1 respectively.
Its two output signals are fed to the
HPF and switch S4, while the square
wave output goes to CON4, although
the signal which appears there is of
limited use, as its duty cycle varies
with frequency.
With switch S4 in the position
shown, the lower frequency (100kHz-
50MHz) signals pass through S4a, the
100nF coupling capacitor and S4b
directly on to the buffer amplifier (the
base of transistor Q3).
For higher-frequency signals, S4 is
moved to the alternative position where
Practical Electronics | June | 2020
the buffer amplifier is fed from the HPF
output, which receives its input from
the unfiltered DDS output pin.
The HPF is a standard seven-pole
Chebyshev filter. Elliptical filters pro-
vide a faster pass-to-stop band cut-off,
but the resulting spurious and harmonic
rejection is less effective compared
with the Chebyshev type. The filter
was optimised to suit standard leaded
components and home-made inductors.
For best performance, the coupling
between the coils must be minimised.
The PCB layout provides for small tin
plate shields to be fitted between filter
stages, a simple and effective solution.
The alternative HPF shown could
potentially shift the 70-150MHz upper
output range to 125-187.5MHz with
appropriate software changes.
RF buffer amplifier
As noted earlier, the buffer amplifier is
a robust discrete design, based on NPN
transistor Q3. This is a well-known sin-
gle-transistor broadband arrangement
providing about 15dB gain along with
good dynamic range. Gain is necessary
to provide the required maximum out-
put level for the signal generator and
to compensate for the insertion loss of
the Serebriakova attenuator.
Alternative discrete buffers seen in
other AD9850/51 based designs lack
sufficient gain across the output range
and/or frequently overload with the
typically higher module output levels
present below 10MHz.
By contrast, this buffer amplifier’s
gain is relatively flat and only reduces
above 50MHz. This is acceptable given
the application and circuit simplicity.
If you find the 2N4427 transistor
difficult to source, you may be able to
find a 2N3866 instead, although the
gain may reduce by several decibels.
The output of the amplifier is taken
from the centre tap of autotransformer
T1 and coupled to the output attenuator
by a 100nF capacitor.
The attenuator consists of four iden-
tical 0/20dB switched attenuators, fol-
lowed by the aforementioned 0-20dB
Serebriakova attenuator, giving an
overall range of 0-100dB. This allows
you to adjust the output from about
−93dBm to +7dBm.
As mentioned earlier, this range
is limited by shielding effectiveness
and RF signal leakage across the atten-
uator sections.
Better shielding between sections
is likely to allow another 20dB fixed
attenuator to be added, significantly
improving its utility for small signal
work. Further improvements would
likely require considerable additional
design efforts around the power supply
and control sections.
Practical Electronics | June | 2020
User interface
IC1 updates the 16×2 LCD using a typi-
cal 4-bit interface. The lower four bits
of PORTC on IC1 (pins 23-26) drive the
four upper LCD data pins, while pins
12 and 13 (digital outputs PD6 and
PD7) drive the RS and EN control lines
of the LCD.
Backlight brightness is fixed using a
1kresistor, with the backlight powered
whenever the device is on, and trimpot
VR1 provides contrast adjustment.
The Grey code pulses from the rotary
encoder are sensed using IC1’s PD2
and PD3 digital inputs (pins 4 and 5),
while presses of the encoder’s integral
pushbutton and the SCAN and MODE
pushbuttons (S1 and S2) are sensed
using digital inputs PD0 and PD1 (pins
2 and 3).
These have internal pull-ups enabled
so that they are held high when no but-
tons are being pressed.
As mentioned earlier, diodes D1 and
D2 have been added to allow presses of
three buttons to be sensed using the two
available pins.
Jumper JP1 and ICSP header CON3
have been provided to allow IC1 to be
re-programmed in situ. Removing JP1
prevents the programmer from trying
to power the RF circuitry. CON3 has
the standard Atmel 6-pin program-
mer pinout.
Power switching
The external power supply, nominally
12V DC, directly powers the output
buffer. The buffer can operate down to
9V although harmonic distortion at full
output increases by about 6dB at 9V
compared to 12V.
The 12V supply is also regulated
down to 5V by REG1 for the AD9850
module and the ATmega328 processor.
The AD9850 module is current-hungry,
so REG1 requires a heatsink.
Dissipation losses would be reduced
by using a switchmode regulator, but
this can introduce switching noise
inside the signal generator, and could
potentially modulate the output buffer
output signal.
As it turns out, the metal signal gen-
erator case forms an effective heatsink
for REG1, and this avoids the need for
additional hardware.
The signal generator will continue to
operate with a supply voltage down to
6V; however, its performance degrades
significantly below 9V. By 6V, the maxi-
mum output falls by 10dB and harmon-
ics are only suppressed by 10dB due to
the reduced dynamic range in the buffer
stage. So, operation at 6V is possible but
not recommended.
A ‘soft switch’ circuit has been added
to allow the use of a momentary push-
button (S3) as a power switch.
The circuitry to provide this function
is shown at the upper right of Fig.4. It
was initially described by Zetex in their
February 1996 Design Note 27, for use
as a relay driver.
However, several problems were en-
countered with that design, including
some curious component choices and
overheating. A minor redesign and the
use of a higher-gain switching transistor
solved them all.
When the supply is initially connect-
ed, the voltage appears on the emitter of
Q4 and the 1µF capacitor charges via the
three series resistors (2.7k, 1kand
270k). However, Q4 cannot turn on
until momentary switch S3 is pressed
and no current is drawn from the supply.
When S3 is pressed, current is sup-
plied to the base of Q5, which switches
it on, and it in turn sinks current from
the base of PNP transistor Q4, switch-
ing it on also and bringing up its col-
lector voltage.
Current can then flow from Q4’s
collector to Q5’s base via the two 1k
series resistors, so Q5 remains on and
so does Q4.
However, the 1µF capacitor discharg-
es because Q5’s collector is now being
pulled low, to 0V. So if S3 is pressed
again, Q5’s base goes low, switching it
off, and in turn switching off Q4, so the
circuit is back in the initial off-state.
Part Two, next month
Next month’s article will have the parts
list, details of PCB assembly, case con-
struction, programming IC1 and how
to use the RF Signal Generator. We’ll
also have performance data, including
spectrum plots.
Reproduced by arrangement with
SILICON CHIP magazine 2020.
www.siliconchip.com.au
References
1. Gary McClellan, Programma-II
synthesised signal generator, Radio-
Electronics magazine, Aug & Sept
1981 (300kHz to 30MHz CW/AM
signal generator, 10kHz tuning steps,
10-300mV output)
2. G. Baars, PE1GIC, DDS RF Signal
Generator, Elektor, October 2003
(50Hz to 70MHz, CW/AM/FM, 1Hz to
1MHz tuning steps, 0 to -127dBm out)
3. Ian Pogson, Solid state modulated RF
test oscillator, Electronics Australia,
May 1979 (455kHz to 30MHz in four
ranges, approximately 100mV output)
4. http://lea.hamradio.si/~s53mv/dds/
theory.html
5. www.picmicrolab.com/ad9850-
pic16f-interface-parallel-data-load/
21
AM/FM/CW
Scanning HF/VHF RF
Signal Generator
We introduced this RF signal generator last month. It is an ideal entry-level
test instrument for anyone into radio: capable, yet low in cost and quite easy
to build. None of the parts are too hard to come by, either. . . Now let’s get into
building it – and getting it up and running. We also have some performance
plots and instructions on how to use it.
T
he Signal Generator is built
on one double-sided PCB cod-
ed 04106191, measuring 152.5
× 102mm, available from the PE PCB
Service. Refer to the PCB overlay dia-
gram, Fig.5.
Most of the top (component-side)
surface has been retained as a ground
plane for added shielding. No SMD
parts are used in the construction of
the Signal Generator, making it rela-
tively easy to build.
Start by fitting all the resistors where
shown. It’s best to check each one with
a DMM set to measure ohms before fit-
ting them, as the colour bands can be
hard to distinguish). Don’t forget the
47Ω resistor hiding under S4.
Then mount diodes D1 and D2, en-
suring they are oriented as shown.
Next, mount IC1’s socket with its
notched end facing the top of the board.
Now fit the ceramic and MKT ca-
pacitors, which are not polarised.
Don’t get the different values mixed
up though. There’s also one of these
under S4. Follow with trimpot VR1
and plastic package transistors Q1,
Q2, Q4 and Q5. Q4 is a different type
than the other three.
Next, solder 6-pin header CON3 and
two-way headers CON4 and JP1 to the
board, followed by the power socket
(CON1) and then the electrolytic ca-
pacitors. These are polarised: the long-
er lead must go to the pad marked with
a ‘+’ on the PCB. The stripe on the can
indicates the negative side.
Fit the three pushbutton switches,
with the flat side oriented as shown
in Fig.5, ensuring they are pushed
down fully onto the board before sol-
dering their pins. S3 is red while S1
and S2 are black. You now have almost
enough components mounted to test
the power supply.
28
It is recommended that you attach
REG1 to the case for heatsinking, but
we haven’t built the case yet. Any-
way, the easiest way to do this is to
cut the three regulator leads short,
then solder 25mm lengths of medium-
duty hookup wire to the stubs, using
some small diameter heatshrink tub-
ing to insulate the solder joints and
the lead stubs.
You can then solder these three
leads to the regulator pads on the PCB,
ensuring that it is soldered the right
way around – ie, so that if you hold it
up above the board with the wires not
crossing over, the tab is facing away
from the board, as shown in Fig.5.
Early testing
Now you can apply 12V power to DC
input connector CON1 and make some
checks. Unfortunately, there is no
power-on indicator LED at this stage
(there will be when MOD1 is fitted),
so the simplest check is to measure
the voltage at the right-hand pin of JP1
relative to a ground point such as the
mounting screw hole in the middle of
the board. At this stage, there should
be little to no voltage.
Now briefly press power switch S3,
and you should measure close to 5V
on the right-hand pin of JP1. Press S3
again and that voltage should drop
away to almost zero. That confirms
that the power supply section is work-
ing correctly.
Modifying the AD9850 module
Minor modifications are required to
the AD9850 module before mounting it
on the PCB. Three SMD resistors need
to be removed and a thin wire soldered
to one of the free pads. These changes
are shown in Fig.6 and the accompa-
nying photo of the modified module.
Part 2
by Andrew Woodfield
ZL2PD
The module I used is, I believe, the
most-common version, but there ap-
pear to be other versions available that
use the same circuit but a different lay-
out. So if your module does not look
exactly the same as mine, don’t panic!
You can use a DMM set on continu-
ity mode to identify the resistors con-
nected to pins 3, 4 and 12 of the IC and
then remove them.
You can do this by heating the ends
of the resistors alternately with a sol-
dering iron while holding the body
of the resistor with tweezers. Once
enough heat has been applied, you can
lift it right off the board. If you have
a hot-air rework station that makes it
even easier.
It’s then just a matter of soldering a
100mm length of light-duty hookup
wire, or Kynar (wire wrap wire) to the
now-empty pad which connects to pin
12 of the IC, as identified in the pho-
to. This will be soldered to the main
board later.
Winding coils L1-L3
The three inductors, L1-L3, are wound
with 0.8mm diameter (26 gauge) enam-
elled copper wire. These are air-cored,
meaning the coils are first wound
around a suitably sized former, then
the former is removed.
The coil diameters should all be
3mm, so a 3mm drill bit shaft or 3mm
diameter metal tube would be suitable.
The coil is then self-supporting when
mounted on the PCB.
L1 and L3 need to be 160nH while
L2 is 150nH. To achieve this, wind 11
turns for each coil, but then stretch L2
so that it is around 1mm longer than
the other two. That reduces its induct-
ance to the required value. (You could,
of course, use an inductance meter to
verify the coils if you have one).
Practical Electronics | July | 2020
 Fig.5: use this
overlay diagram
as a guide
to building
the Signal
Generator. We’ve
shown both LCD
screens in place
here, (Jaycar
QP5516 and
Altronics Z7013;
one on top of the
other) but you
would only fit
one or the other.
Edge connector
CON2’s middle
pin is soldered
on the underside
of the board. VR2
can be a standard
16mm pot
mounted through
the board, with
the body on the
underside, or
a 9mm vertical
PCB-mounting
type.

If you want to achieve the alterna-
tive inductor values mentioned last
month, reduce the number of turns to
six, then stretch L2 by around 0.5mm.
Now remove the enamel at each end
of the remaining wire on each coil.
Some enamel coatings vaporise while
being tinned, but most must be scraped
off with a sharp knife.
Take care if you use the latter ap-
proach, especially to avoid cutting
yourself. You can verify that you’ve
scraped off the insulation properly by
tinning the wire ends and then check-
ing that the solder has adhered.
But note that you don’t want a lot of
excess lead length on these coils; just
enough to make it through the mount-
ing holes on the PCB and be soldered
on the underside.
So cut the wire ends to length before
stripping the enamel.
Don’t stretch or compress the coils
to fit the pads on the PCB as this will
affect their inductance; just use a short
length of extra wire at one or both ends
to reach the mounting pads.
Winding the transformer
T1 is wound on a 7mm-long ferrite
balun core. Begin with 400mm of
0.315mm-diameter (28 gauge) enam-
elled copper wire.
Fold the wire in half so the two cut
ends meet, then twist the two wires to-
gether to produce a twisted wire simi-
lar to that shown in Fig.7.
It can have anywhere from one to five
twists every 20mm; this isn’t critical.
Twisting the wire simply makes wind-
ing the wire onto the core a little easier.
Wind four turns of the twisted wire
onto the core and trim the ends of the
‘bifilar’ wires, so you have four short
lengths of wire each about 20mm
long appearing at one pig-nose end
of the core.
Tin these four ends. Use a multime-
ter to identify the start and end of the
two coils.
The start of one coil and the end
of the other (shown as ‘AS’ and ‘BF’
in the diagram) go to the two central
mounting pads for T1 (either together
into one pad, or separately into each),
while the other two wires go to the
mounting pads at either end.
It doesn’t matter which goes to
which, as the coil is symmetrical.
Again, cut the leads to leave just a
minimal amount and then strip the
enamel off and tin them before solder-
ing them to the board.
This should allow you to mount the
balun close to the board, so it won’t rat-
tle around after the wires are soldered.
Proceeding with construction
Now fit metal-can transistor Q3 close to
the PCB, leaving about 1mm between
the bottom of the device and the up-
per PCB surface. Don’t install it firmly
down on the PCB because the metal
case of the transistor is internally con-
nected to the collector terminal of Q3.
Also, before you solder it in place,
check the metal case is not touching
any adjacent component leads.
Next, fit your modified AD9850 DDS
module by soldering two 10-pin head-
ers to the PCB, then soldering the mod-
ule to the pins on top of these headers.
The wire you connected to that
module earlier connects to the lead
of transistor Q1 which is closest to
MOD1. RevB PCBs have a dedicated
pad for this wire.
Otherwise, solder it directly to Q1’s
lead, on the top side of the PCB. Ei-
ther way, trim the wire to length be-
fore stripping and soldering it. This
wire should ideally be routed under
the module for neatness. If you keep
it short, it won’t move around later.
Next, fit output socket CON2. As
it’s an edge connector, push it onto
the edge of the PCB, with the central
pin sliding over the central pad on the
bottom side.
Solder that central pin, plus the
posts on either side, on both the top
and the bottom sides of the PCB.
As this is a fairly substantial chunk
of metal being soldered to copper
planes, you will need a hot iron and
be generous with the solder.
Then install mini slide switches S5-
S9. The board is designed with slots
to suit their lugs, so you can solder
them right down onto the PCB. Again,
be generous with the solder to ensure
good joints.
Next, mount the LCD. There are three
possible headers to suit different LCD
module styles, although Jaycar QP5516
or Altronics Z7018 are the best fit.
For the Jaycar LCD, solder an 8×2-
pin DIL header to the row of pins near-
est the left edge of the PCB, then at-
tach the four short tapped spacers to
the corner mounting holes from the
bottom of the board, using 5mm ma-
chine screws.

Practical Electronics | July | 2020 29

 REMOVE THESE
SMD RESISTORS

Fig.6: these three SMD resistors must be removed from the AD9850 DDS module. One
of the pads which connected to the now-gone 3.9kresistor makes a handy connection
point for the extra wire needed to connect pin 12 of the IC (RSET) to the collector of
transistor Q1 on the main board, for output level control. See also the close-up photo CONNECT THE RSET
at right. (PIN 12) WIRE HERE

You can then slip the LCD over the
pin header and attach it using four more
5mm machine screws, then solder the
header pins to the top of the LCD.
The procedure for the other LCDs
is similar except some LCDs may re-
quire short jumper wires to connect
to the PCB.
The final two components to fit are
rotary encoder RE1 and potentiome-
ter VR2. Mounting RE1 is easy; make
sure it’s perpendicular to the PCB and
pushed all the way down before sol-
dering its pins. Solder its five pins and
two mounting lugs; you will need a
hot iron for the latter, and be gener-
ous with the solder.
For VR2, we’ve provided two dif-
ferent options. The prototype used a
16mm potentiometer with its body on
the underside of the PCB and its shaft
passing up through a hole.
Mounting it in this way is a bit fid-
dly, but there are two benefits: this is a
standard part that’s easier to get, and its
shaft will line up perfectly with push-
buttons S1/S2 and the access hole for
trimpot VR1 (if provided).
Alternatively, if you can get your
hands on a 9mm PCB-mounting right-
angle potentiometer, it will be dead
easy to mount to the PCB, as it’s fitted
similarly to RE1. However, due to the
location of the hole for the 16mm pot’s
shaft, its shaft will sit around 3.5mm
lower than S1/S2 and VR1. (This is
hardly a tragedy – the choice is yours!)
Now plug in the ATmega328 mi-
crocontroller (IC1), making sure its
pin 1 is oriented correctly, towards
the upper-left corner of the board. (If
you haven’t already programmed it or
purchased a programmed chip, see the
panel below detailing the program-
ming instructions.)
Further testing
Later, we will be attaching REG1 to the
metal case. However, since we haven’t
built it yet, for further testing tempo-
rarily attach a flag heatsink or attach
it to a spare sheet of metal using a ma-
chine screw and nut.
You can now apply 12V power to
CON1, press S3 and check that you
can control the output frequency and
amplitude (see the operating instruc-
tions below).
Power the unit down before finish-
ing construction.

Two inter-coil screens, show in red on the overlay) must be The modified AD9850 module in situ on the main PCB.
fitted between the coils as shown here. These can be cut from The three SMD resistors are all removed and the yellow wire
a scrap of tinplate (eg, a food tin). This photo also shows the is soldered to the appropriate pad – the one marked R6.
mounting of the 7805 regulator on the case heatsink. (make sure it is the pad closest to the AD9850 IC).

30 Practical Electronics | July | 2020


Fig.7: autotransformer T1 is easy to
make, with just four bifilar turns wound
on the small ferrite balun core. AF
and BS are interchangeable and are
connected together on the PCB.
Fitting the shields
You will notice several holes around
the buffer, attenuator, output and band
select/HPF sections of the board. There
are also lines on the PCB ‘silkscreen’
between these holes. This is where
shield plates can be fitted.
However, you do not need to fit
shields in most of these areas; the only
ones that are critical are those between
the three high-pass filter sections (be-
tween L1 and L2, and L2 and L3).
So you only really need to cut two
shield pieces and mount them using
four posts in the holes provided. These
are shown in red on the PCB overlay
diagram, Fig.5.
Each shield piece should be around
8mm high and cut from 0.5mm tin
plate, or recycled tin cans (a fruit or
Milo tin lid is ideal).
The strips are then mounted to the
board using component leads off-cuts
soldered into the holes shown in red.
This is simple yet effective.
You could fit shields in the other
locations, but testing has shown that
it makes virtually no difference to
the device’s performance so I don’t
feel that it’s worth the time and ef-
fort to do so.
Making the enclosure
I couldn’t find a suitable ready-made box
for the Signal Generator, so I came up
with a relatively easy way to make one.
It’s a simple folded metal box and
works well, resulting in a unit that is
light but robust, compact and effective-
ly shielded. Dimensioned drawings of
the metalwork are available on the July
2020 page of the PE website – they’re
a little too large to publish here! The
two panels are cut and folded from
0.8mm-thick aluminium sheets. The
top cover and base may each be cut
from a small 300 × 250mm sheet, mak-
ing it relatively inexpensive to build.
Practical Electronics | July | 2020
Parts list – HF/VHF RF SIGNAL GENERATOR
1 double-sided PCB, coded 04106191, 152.5 x 102mm (from the PE PCB Service)
1 AD985x-based DDS module (MOD1)
1 PCB-mount barrel power socket (CON1)
1 SMA edge-mount socket (CON2)
1 2x3 pin header (CON3)
2 2-way pin headers (CON4)
1 jumper shunt/shorting block (JP1)
1 16x2 alphanumeric LCD with backlight (LCD1)
[eg, Jaycar QP5521 or Altronics Z7018]
1 500mm length of 0.8mm diameter enamelled copper wire (for winding L1-L3)
1 400mm length of 0.315mm diameter enamelled copper wire (for winding T1)
1 7mm ferrite balun core (for T1) [Jaycar LF1222, Altronics L5235]
1 PCB-mount vertical rotary encoder with integral switch (RE1) [Jaycar SR1230] –
IMPORTANT see below
1 28-pin narrow DIL socket (for IC1)
2 10-pin headers (for mounting MOD1)
1 16-pin SIL or 8 x 2 DIL header (for LCD)
4 6.3mm-long M3 tapped nylon spacers (for LCD)
8 5mm M3 panhead machine screws (for LCD)
2 black PCB-mount momentary pushbuttons (S1,S2)
[eg Jaycar SP0721, Altronics S1096]
1 red PCB-mount momentary pushbuttons (S3) [Jaycar SP0720, Altronics S1095]
5 DPDT mini slide switches (S4-S8) [Jaycar SS0852, Altronics S2010/S2020]
1 9mm-diameter knob to suit VR2
1 28-34mm-diameter knob to suit RE1
1 0.5mm-thick tin plate or cleaned tin-plated steel cans
2 0.8mm-thick aluminium sheets, 300 x 250mm
1 adhesive panel label, 157 x 107mm
4 small self-adhesive rubber feet
Hookup wire, misc. enclosure hardware
Semiconductors
1 ATmega328P microcontroller programmed with 0410619A.hex, DIP-28 (IC1) – a
programmed IC is available from: www.siliconchip.com.au/Shop/9/5056
1 7805 5V 1A linear regulator, TO-220 (REG1)
3 BC548 NPN transistors, TO-92 (Q1,Q2,Q5)
1 2N4427 NPN RF transistor, TO-39 (Q3)
1 BC327 PNP transistor, TO-92 (Q4)
2 1N4148 small signal diodes (D1,D2)
Capacitors
2 10µF 50V electrolytic 1 1µF 50V electrolytic
11 100nF 63V MKT 1 10nF 63V MKT 1 1nF 63V MKT or 50V ceramic
2 15pF 50V C0G/NP0 ceramic 2 10pF 50V C0G/NP0 ceramic
Resistors (all 0.25W 1% metal film)
2 470k 1 270k 5 10kΩ 1 3.9k 1 2.7kΩ 5 1k
1 820 1 390 5 220 8 56 2 47 2 27
1 10k mini horizontal trimpot (VR1)
1 500 9mm vertical PCB-mount or 16mm standard potentiometer (VR2)
Encoders: we have discovered that some rotary encoders look identical but work
differently, resulting in erratic operation. The V14 firmware addresses this; by default, it
works with pulse-type encoders. You can identify these by testing continuity across the
two internal switches; if they are both always open when the encoder is at rest, it is a
pulse type. With the level type, one or both switches may be closed at rest, depending
on the encoder’s rotation. If you have a level-type encoder and V14 software, solder a
100kΩ resistor from pin 28 of the Atmel chip to ground, on the underside of the PCB.
That will change the software to work with level-type encoders.
This grade of aluminium is light from standard drills, a metal nibbling
enough to be cut and folded easily with tool is ideal for cutting out the rectan-
hand tools, but heavy enough to form gular holes.
a sturdy box for the Signal Generator. Final finishing during fitting can be
Several holes need to be drilled and completed with a fine file.
cut into the panel for the controls, slide The completed PCB is mounted using
switches, regulator and the LCD. Aside spacers and 3mm machine screws. It’s
31
 Programming the ATmega328 micro
You can purchase a pre-programmed mi-
crocontroller (see parts list). Alternatively,
to program AVR family microprocessors,
you need a programmer such as the USBasp
(see www.fischl.de/usbasp/ for details
and drivers). This can be purchased
online from many suppliers for
a few pounds.
Suitable free software is
available for Windows, Linux
and iOS online. This description
focuses on the Windows version.
You need to install the USBasp
drivers and download programming
software. For Windows, this includes:
eXtreme Burner
http://extremeelectronics.co.in/avr-
tutorials/gui-software-for-usbasp-based-
usb-avr-programmers/
AVRDUDESS
http://blog.zakkemble.net/avrdudess-a-
gui-for-avrdude/)
Khazama
http://khazama.com/project/programmer
Plug it in and complete the installation of
the USBasp programmer into your PC. If
you have the option of 3.3V or 5V program-
ming levels, select 5V.
Launch
the pro-
gramming software
you downloaded earli-
er and set the target device
to ‘ATmega328’ or ‘Atmega 328P’,
depending on your chip. Both may be
used. Now download the HEX file for this
project from the July 2020 page of the PE
website (if you don’t already have it) and
select it as the file to be used to program
the chip in your software.
Make sure JP1 has not been fitted to
your signal generator board; if it has, re-
move it now. Note that since some of the
ATmega328 pins connect to the AD9850
module, the AD9850 module’s power LED
will still light up and flash while the pro-
grammer is connected and running, de-
spite having removed JP1 and therefore
cut the power supply to the module. This
is of no concern.
Plug the six-pin connector from the
USBasp programmer into CON3 on the
signal generator PCB, making sure that pin
1 on the programmer cable lines up with
the pin 1 indicator on the PCB.
Now select ‘Write FLASH buffer to
chip’ or ‘Write – Flash’ to program the
ATmega328 with the HEX file. The LEDs
on the USBasp will blink furiously for a
minute or two while the HEX file is loaded
into the ATmega328. A bar graph may be
displayed in some cases on the PC screen,
to show progress.
You then have to program the
ATmega328 internal ‘fuses’. These
configure the operating characteristics of
the ATmega328 to suit the software being
run on the device.
For this step, insert the following set-
tings into the relevant Fuse page/section
of the programming software, then click
on ‘Write’ to send the data to the fuses:
Low byte: 0xE2
High byte: 0xD9
Extended byte: 0xFF
Lock byte: 0xFF
Since the processor and display are pow-
ered via the programmer, once program-
ming is complete, the display will briefly
show the start-up message and then the
initial signal generator screen. At this
point, you can unplug the programming
cable from CON3 and place a shunt on JP1.

best to line it up with the holes in the
lid to figure out exactly where it will sit
in the case before marking and drilling
out the three mounting holes in the base.
Alternatively, as in the prototype,
the Signal Generator PCB can be held
onto the front panel using the rotary
encoder nut, although it would prob-
ably be better to attach using at least
one tapped spacer too.
Small self-tapping screws are used
to hold the cover to the base of the
box. Once you’ve cut and bent the
sheets, rivet or screw the 7805 regu-
lator (REG1) onto the metal cover just
before the final step of screwing the
cover to the base.
The front panel artwork is shown
in Fig.8. It can be printed and covered
with self-adhesive plastic film.
Trim the front panel artwork to cut
out the holes for the various controls
and display and test-fit onto the com-
pleted metalwork.
The most reliable method to fix the
artwork in place is to spray the rear
side of the artwork with adhesive spray
obtainable from most art shops. While
tacky, press the panel artwork into
place. Remove the rotary encoder nut

Fig.9: the CW (carrier wave, ie, unmodulated) output at
10MHz/−28dBm with a span of about 9-37MHz, selected
to include the first two harmonics. This shows the second
harmonic (20MHz) at around −40dB and the third (30MHz)
at around −47dB.
Fig.10: analysis of the AM output at 10MHz/−12dBm with
a 20kHz span (ie, 9.99-10.01MHz). The 1kHz sidebands are
visible either side of the carrier, as are the 1kHz modulation
tone distortion products at ±2kHz (−21dB below the 1kHz
fundamental) and ±3kHz (−26dB below the fundamental)
indicating acceptable audio distortion levels. The
modulation depth is the industry test standard, 30%.

32 Practical Electronics | July | 2020

before attaching the
front panel, then re-
attach it on top.

3D-printed knobs Reproduced by arrangement with

SILICON CHIP magazine 2020.
Suitable knobs may be
www.siliconchip.com.au
available from normal -20dB -20dB -20dB -20dB RF OUT
suppliers. However,
I designed the knobs 0-20dB
for my Signal Genera-
tor using DesignSpark
Mechanical and 3D-
printed them from
grey PLA filament. MODE SCAN
My knob STL files
can also be down- BAND
loaded from the July 0-50MHz 70-120MHz
2020 page of the PE
website for those
wishing to print their
own knobs. They TUNE STEP
press into place and POWER
hold securely.
It’s useful to add
SILICON ZL2PD HF/VHF AM/FM/CW Scanning Signal Generator
four self-adhesive CHIP siliconchip.com.au DC IN
rubber feet to the
rear of the metal en-
closure. This prevents Fig.8: download this panel label from the July 2020 page of the PE website (as a PDF). Just print
any sharp corners of and laminate it, cut out the display and switch holes, then cut it to size and glue it to the outside
top of the case.
the aluminium box
from scratching the
bench and helps to keep the oscillator The display shows the current out- each click as RE1 is rotated. When you
in one place on the workshop bench. put frequency and operating mode; push this button, the underline below
the Signal Generator always starts at the LCD frequency display moves to
Using the Signal Generator 10.000MHz in CW (unmodulated) mode. indicate the current step setting.
Briefly press power switch S3 to turn The display also features a frequen- The Band switch (S4) selects be-
the Signal Generator on. The display cy ‘dial’ which covers a 1MHz span tween the two output frequency rang-
will show a start-up message, then af- with 100kHz markers. As you rotate es, 0-50MHz (left) and 70-120MHz
ter a short delay, the normal screen. RE1 (‘TUNE’), the output frequency (right), while S5-S8 at the top, in com-
If you cannot see any text on the dis- changes and the cursor on this scale bination with VR2 at right, set the out-
play, adjust VR1. This sets the LCD con- shifts across the ‘dial’. put amplitude.
trast. You can see examples of the vari- Pushing RE1’s knob in (the ‘STEP’ The Band switch must be in the
ous possible displays in the first article pushbutton) changes the increments in correct position for the currently selected
in this series, published last month. which the frequency is adjusted with frequency to get the expected output

Fig.11: a ‘narrow band’ 1.75kHz frequency-modulated
signal with a 10MHz carrier and a 20kHz span. The iconic
equi-spaced 1kHz sidebands of a standard FM signal are
clearly visible.
Fig.12: ‘wideband’ or broadcast radio-style FM, again with
the carrier at 10MHz, this time captured with a 500kHz
frequency span. This clearly illustrates that most of the
signal energy falls within the 200kHz channel bandwidth
permitted for broadcast FM signals.

Practical Electronics | July | 2020 33

amplitude. The HPF is very effective at minimising energy
from aliasing below 70MHz, so the output level can be lower
than expected by over 60dB if the incorrect selection is made.
But no damage will occur as a result of an incorrect setting.
While the upper range is described as 70-120MHz, tun-
ing and operation are maintained up to 150MHz, although
output levels fall significantly above 120MHz.
The maximum output of +7dBm is with S5-S8 all in the
up position and VR2 fully clockwise. For each 20dB of at-
tenuation you need, switch one of S5-S8 into the down po-
sition (it doesn’t matter which). Then for fine attenuation
adjustments, rotate VR2.
For example, if you want −30dBm, set any one of S5-S8
down (+7dBm − 20dB = −13dBm) and then VR2 should
be set quite low, to give an additional 17dB of attenua-
tion. (Note: standard DDS amplitude rolloff impact above
30MHz – see Fig.3 in part 1.)
The Signal Generator mode is selected with brief presses
of the Mode key (S2).
This selects between CW, AM, FM-NB (±1.5kHz devia-
tion), FM-WB (±3kHz deviation), FM-BC (±50kHz devia-
tion), or SCAN mode.
Pressing the Mode key again will select the initial CW
(unmodulated) mode again, and the standard display screen.
Frequency scanning mode
If the SCAN mode is selected, the display changes to show
the currently saved Start and End frequencies for the scan,
and the number of steps selected. At power-on, this is set
to 200 steps. If this is the first time after power has been
applied, the default frequency settings (starting at 1MHz
and ending at 30MHz) are shown. Otherwise, the last used
settings will be displayed.
Pressing the Scan button again allows each parameter to
be selected for adjustment.
Use the TUNE and STEP controls to set the Start and
End frequencies in turn; here, the STEP button selects the
tuning step as usual.
When the scan Steps parameter is selected with the SCAN
button, the TUNE control has no effect, but pressing the
STEP button allows the number of steps to be selected (10,
20, 50, 100, 200 or 500 per scan).
Finally, pressing SCAN again saves the selected values
and starts the scan. The display now reports SCAN instead
of the number of steps.
The scanning frequency increment is calculated by the
Fig.13: measured performance of the high-pass filter
comprising inductors L1-L3 and four small ceramic
capacitors. As you can see, the response is pretty much
flat from 70MHz to 400MHz, but signals from 0-40MHz are
attenuated by 60dB. The transition is smooth and quick, at
around 75dB/octave, or 2dB/MHz.
34
processor using the entered values. The scanning speed is
surprisingly fast.
Scanning may be interrupted and restarted using the
SCAN key. When stopped, the Start and End frequencies,
as well as the number of scan steps, can be adjusted again,
and the scan restarted.
To exit the scan mode, press the MODE key. This also
stops the scan and resets the Signal Generator to the last
scanned frequency, and CW mode.
At each stage, the output can be checked with a suitable
oscilloscope or with other RF test instruments.
Performance
Typical output signals from the Signal Generator are shown
in Figs.9-12. These were captured using a Siglent 3GHz
spectrum analyser. See the figure captions for details.
Fig.13 demonstrates how effective the high-pass filter is,
despite being made from a self-wound air-cored inductors.
This shows that the filter provides 60dB of attenuation for
signals below 40MHz with a virtually flat passband from
70MHz up. The filter roll-off is quite steep at around 75dB/
octave (the span from 40MHz to 70MHz is about 0.8 octaves).
Future possibilities
It is possible to add further features to the software. With
the supplied software, less than 30% of IC1’s program
memory is used.
For example, RF output levelling would be possible, by
using the pin 11 PWM output which drives the RSET pin
of the AD9850 module (currently used to provide AM)
to offset the sinX/X roll-off for frequencies up to about
50MHz, at the cost of a reduced maximum output level at
lower frequencies.
Extended frequency coverage also appears possible
through the use of alternative high-pass filters and/or by
replacing the AD9850 module with one based on the pin-
compatible AD9851.
Some minor additional software changes would be re-
quired to permit the AD9851 to be used. The AD9851
can be clocked at up to 180MHz, which may allow the
generator to operate up to 100MHz in a single range, and
possibly up to 300MHz with a modified HPF. Suitable
AD9851 modules are available from the same sources as
the AD9850-based module.
Adding other modulation modes such as FSK and BPSK
is also feasible, but adding QPSK, for example, may be be-
yond the reach of this design.
Moving to an even more advanced DDS device, such as
one based on the more modern AD99xx series chips, could
be done. However, this would substantially increase the
overall cost and complexity of the device.
It is also possible to replace the basic passive output
variable attenuator network with a more elegant PIN-
diode-based system.
This involves using components that are more difficult
to obtain, but sufficient space has been left in this area of
the PCB for such an addition.
Finally, you could consider adding a numeric keypad
on the front panel to permit the direct entry of frequen-
cies, tuning step and scan settings, plus you could add a
settings memory for frequently used configuration.
However, this would likely require a processor change,
or potentially even an additional microcontroller for han-
dling keypad entry, to obtain the necessary spare I/O pins.
Having said all that, the design as presented is a good
compromise between low complexity and cost, while still
having a useful frequency range and a good set of features.
It makes a great entry-level RF signal generator – a ‘must’
for anyone interested in radio at any level!
Practical Electronics | July | 2020