Contents
Contents
10.1 The Mechanism of Mixers and Mixing
10.5 A Survey of Common Mixer Types
10.1.1 What is a Mixer?
10.1.2 Putting Multiplication to Work
10.5.1 Gain-Controlled Analog Amplifiers
As Mixers
10.2 Mixers and Amplitude Modulation
10.2.1 Overmodulation
10.2.2 Using AM to Send Morse Code
10.5.3 The Diode Double-Balanced Mixer:
A Basic Building Block
10.2.3 The Many Faces of Amplitude
Modulation
10.2.4 Mixers and AM Demodulation
10.5.2 Switching Mixers
10.5.4 Active Mixers — Transistors as
Switching Elements
10.5.5 The Tayloe Mixer
10.3 Mixers and Angle Modulation
10.5.6 The NE602/SA602/SA612:
A Popular Gilbert Cell Mixer
10.3.1 Angle Modulation Sidebands
10.3.2 Angle Modulators
10.3.3 Mixers and Angle Demodulation
10.5.8 An Experimental High-Performance
Mixer
10.4 Putting Mixers, Modulators and Demodulators
to Work
10.4.1 Dynamic Range: Compression,
Intermodulation and More
10.4.2 Intercept Point
10.5.7 An MC-1496P Balanced Modulator
10.6 References and Bibliography
Chapter
10
Mixers, Modulators and
Demodulators
At base, radio communication involves
translating information into radio form,
letting it travel for a time as a radio signal, and translating it back again. Translating information into radio form entails
the process we call modulation, and
demodulation is its reverse. One way or
another, every transmitter used for radio
communication, from the simplest to
the most complex, includes a means of
modulation; one way or another, every
receiver used for radio communication,
from the simplest to the most complex,
includes a means of demodulation.
Modulation involves varying one or
both of a radio signal’s basic characteristics — amplitude and frequency
(or phase) — to convey information. A
circuit, stage or piece of hardware that
modulates is called a modulator.
Demodulation involves reconstructing the transmitted information from the
changing characteristic(s) of a modulated radio wave. A circuit, stage or piece
of hardware that demodulates is called a
demodulator.
Many radio transmitters, receivers and
transceivers also contain mixers — circuits, stages or pieces of hardware that
combine two or more signals to produce
additional signals at sums of and differences between the original frequencies.
This chapter, by David Newkirk,
W9VES, and Rick Karlquist, N6RK,
examines mixers, modulators and
demodulators. Related information may
be found in the Modulation chapter,
and in the chapters on Receivers and
Transmitters. Quadrature modulation
implemented with an I/Q modulator, one
that uses in-phase (I) and quadrature
(Q) modulating signals to generate the
0° and 90° components of the RF signal,
is covered in the chapters on Modulation and DSP and Software Radio
Design.
Amateur Radio textbooks have traditionally handled mixers separately from modulators
and demodulators, and modulators separately from demodulators. This chapter examines
mixers, modulators and demodulators together because the job they do is essentially the
same. Modulators and demodulators translate information into radio form and back again;
mixers translate one frequency to others and back again. All of these translation processes
can be thought of as forms of frequency translation or frequency shifting — the function
traditionally ascribed to mixers. We’ll therefore begin our investigation by examining what
a mixer is (and isn’t), and what a mixer does.
10.1 The Mechanism of Mixers
and Mixing
10.1.1 What is a Mixer?
Mixer is a traditional radio term for a circuit that shifts one signal’s frequency up or down
by combining it with another signal. The word mixer is also used to refer to a device used to
blend multiple audio inputs together for recording, broadcast or sound reinforcement. These
two mixer types differ in one very important way: A radio mixer makes new frequencies out
of the frequencies put into it, and an audio mixer does not.
MIXING VERSUS ADDING
Radio mixers might be more accurately called “frequency mixers” to distinguish them
from devices such as “microphone mixers,” which are really just signal combiners, summers
or adders. In their most basic, ideal forms, both devices have two inputs and one output.
The combiner simply adds the instantaneous voltages of the two signals together to produce
the output at each point in time (Fig 10.1). The mixer, on the other hand, multiplies the instantaneous voltages of the two signals together to produce its output signal from instant to
instant (Fig 10.2). Comparing the output spectra of the combiner and mixer, we see that the
combiner’s output contains only the frequencies of the two inputs, and nothing else, while the
mixer’s output contains new frequencies. Because it combines one energy with another, this
process is sometimes called heterodyning, from the Greek words for other and power. The
sidebar, “Mixer Math: Mixing as Multiplication,” describes this process mathematically.
The key principle of a radio mixer is that in mixing multiple signal voltages together,
it adds and subtracts their frequencies to produce new frequencies. (In the field of signal
processing, this process, multiplication in the time domain, is recognized as equivalent to the
process of convolution in the frequency domain. Those interested in this alternative approach
to describing the generation of new frequencies through mixing can find more information
about it in the many textbooks available on this subject.)
The difference between the mixer we’ve been describing and any mixer, modulator or
demodulator that you’ll ever use is that it’s ideal. We put in two signals and got just two
signals out. Real mixers, modulators and demodulators, on the other hand, also produce
distortion products that make their output spectra “dirtier” or “less clean,” as well as putting
out some energy at input-signal frequencies and their harmonics. Much of the art and science of making good use of multiplication in mixing, modulation and demodulation goes
Mixers, Modulators and Demodulators 10.1
Fig 10.1 — Adding or summing two sine waves of different frequencies (f1 and f2) combines their amplitudes without affecting
their frequencies. Viewed with an oscilloscope (a real-time graph of amplitude versus time), adding two signals appears as a
simple superimposition of one signal on the other. Viewed with a spectrum analyzer (a real-time graph of signal amplitude versus
frequency), adding two signals just sums their spectra. The signals merely coexist on a single cable or wire. All frequencies that
go into the adder come out of the adder, and no new signals are generated. Drawing B, a block diagram of a summing circuit,
emphasizes the stage’s mathematical operation rather than showing circuit components. Drawing C shows a simple summing
circuit, such as might be used to combine signals from two microphones. In audio work, a circuit like this is often called a mixer —
but it does not perform the same function as an RF mixer.
Fig 10.2 — Multiplying two sine waves of different frequencies produces a new output spectrum. Viewed with an oscilloscope, the
result of multiplying two signals is a composite wave that seems to have little in common with its components. A spectrum-analyzer
view of the same wave reveals why: The original signals disappear entirely and are replaced by two new signals — at the sum and
difference of the original signals’ frequencies. Drawing B diagrams a multiplier, known in radio work as a mixer. The X emphasizes
the stage’s mathematical operation. (The circled X is only one of several symbols you may see used to represent mixers in block
diagrams, as Fig 10.3 explains.) Drawing C shows a very simple multiplier circuit. The diode, D, does the mixing. Because this circuit
does other mathematical functions and adds them to the sum and difference products, its output is more complex than f1 + f2 and
f1 – f2, but these can be extracted from the output by filtering.
10.2 Chapter 10
Mixer Math: Mixing as Multiplication
Since a mixer works by means of multiplication, a bit of math can show us how
they work. To begin with, we need to represent the two signals we’ll mix, A and B,
mathematically. Signal A’s instantaneous amplitude equals
A a sin2πfa t
(1)
in which A is peak amplitude, f is frequency, and t is time. Likewise, B’s instantaneous amplitude equals
A b sin2πfb t
(2)
Since our goal is to show that multiplying two signals generates sum and difference frequencies, we can simplify these signal definitions by assuming that the peak
amplitude of each is 1. The equation for Signal A then becomes
a(t) = A sin (2πfa t)
(3)
and the equation for Signal B becomes
(4)
b(t) = B sin (2πfb t) Each of these equations represents a sine wave and includes a subscript letter to
help us keep track of where the signals go.
Merely combining Signal A and Signal B by letting them travel on the same wire
develops nothing new:
(5)
a(t) + b(t) = A sin (2πfa t) + B sin (2πfb t) As simple as equation 5 may seem, we include it to highlight the fact that multiplying two signals is a quite different story. From trigonometry, we know that multiplying
the sines of two variables can be expanded according to the relationship
sin x sin y = 1 [cos (x − y) − cos (x + y)]
(6)
2
Conveniently, Signals A and B are both sinusoidal, so we can use equation 6 to
determine what happens when we multiply Signal A by Signal B. In our case, x =
2πfat and y = 2πfbt, so plugging them into equation 6 gives us
AB
AB
a(t) × b(t) =
cos ( 2π [ fa − fb ] t ) −
cos ( 2π [ fa + fb ] t )
2
2
(7)
Now we see two momentous results: a sine wave at the frequency difference
between Signal A and Signal B 2π(fa – fb)t, and a sine wave at the frequency sum
of Signal A and Signal B 2π(fa + fb)t. (The products are cosine waves, but since
equivalent sine and cosine waves differ only by a phase shift of 90°, both are called
sine waves by convention.)
This is the basic process by which we translate information into radio form and
translate it back again. If we want to transmit a 1-kHz audio tone by radio, we can
feed it into one of our mixer’s inputs and feed an RF signal — say, 5995 kHz — into
the mixer’s other input. The result is two radio signals: one at 5994 kHz (5995 – 1)
and another at 5996 kHz (5995 + 1). We have achieved modulation.
Converting these two radio signals back to audio is just as straightforward. All we
do is feed them into one input of another mixer, and feed a 5995-kHz signal into the
mixer’s other input. Result: a 1-kHz tone. We have achieved demodulation; we have
communicated by radio.
into minimizing these unwanted multiplication products (or their effects) and making
multipliers do their frequency translations
as efficiently as possible.
10.1.2 Putting Multiplication
to Work
Piecing together a coherent picture of how
multiplication works in radio communica-
tion isn’t made any easier by the fact that
traditional terms applied to a given multiplication approach and its products may vary
with their application. If, for instance, you’re
familiar with standard textbook approaches
to mixers, modulators and demodulators, you
may be wondering why we didn’t begin by
working out the math involved by examining
amplitude modulation, also known as AM.
“Why not tell them about the carrier and
Fig 10.3 — We commonly symbolize
mixers with a circled X (A) out of tradition,
but other standards sometimes prevail (B,
C and D). Although the converter/changer
symbol (D) can conceivably be used to
indicate frequency changing through
mixing, the three-terminal symbols are
arguably better for this job because they
convey the idea of two signal sources
resulting in a new frequency. (IEC
stands for International Electrotechnical
Commission.)
how to get rid of it in a balanced modulator?” A transmitter enthusiast may ask “Why
didn’t you mention sidebands and how we
conserve spectrum space and power by getting rid of one and putting all of our power
into the other?” A student of radio receivers,
on the other hand, expects any discussion of
the same underlying multiplication issues to
touch on the topics of LO feedthrough, mixer
balance (single or double?), image rejection
and so on.
You likely expect this book to spend some
time talking to you about these things, so it
will. But this radio-amateur-oriented discussion of mixers, modulators and demodulators
will take a look at their common underlying mechanism before turning you loose on
practical mixer, modulator and demodulator
circuits. Then you’ll be able to tell the forest from the trees. Fig 10.3 shows the block
symbol for a traditional mixer along with
several IEC symbols for other functions mixers may perform.
It turns out that the mechanism underlying multiplication, mixing, modulation and
demodulation is a pretty straightforward
thing: Any circuit structure that nonlinearly
distorts ac waveforms acts as a multiplier to
some degree.
Mixers, Modulators and Demodulators 10.3
NONLINEAR DISTORTION?
The phrase nonlinear distortion sounds
redundant, but isn’t. Distortion, an externally
imposed change in a waveform, can be linear;
that is, it can occur independently of signal
amplitude. Consider a radio receiver frontend filter that passes only signals between 6
and 8 MHz. It does this by linearly distorting
the single complex waveform corresponding
to the wide RF spectrum present at the radio’s
antenna terminals, reducing the amplitudes
of frequency components below 6 MHz and
above 8 MHz relative to those between 6 and
8 MHz. (Considering multiple signals on a
wire as one complex waveform is just as valid, and sometimes handier, than considering
them as separate signals. In this case, it’s a bit
easier to think of distortion as something that
happens to a waveform rather than something
that happens to separate signals relative to
each other. It would be just as valid — and
certainly more in keeping with the consensus
view — to say merely that the filter attenuates signals at frequencies below 6 MHz and
above 8 MHz.) The filter’s output waveform
certainly differs from its input waveform; the
waveform has been distorted. But because
this distortion occurs independently of signal
level or polarity, the distortion is linear. No
new frequency components are created; only
the amplitude relationships among the wave’s
existing frequency components are altered.
This is amplitude or frequency distortion, and
all filters do it or they wouldn’t be filters.
Phase or delay distortion, also linear,
causes a complex signal’s various component frequencies to be delayed by different
amounts of time, depending on their frequency but independently of their amplitude. No
new frequency components occur, and amplitude relationships among existing frequency
components are not altered. Phase distortion
occurs to some degree in all real filters.
The waveform of a non-sinusoidal signal
can be changed by passing it through a circuit
that has only linear distortion, but only nonlinear distortion can change the waveform of
a simple sine wave. It can also produce an output signal whose output waveform changes as
a function of the input amplitude, something
not possible with linear distortion. Nonlinear
circuits often distort excessively with overly
strong signals, but the distortion can be a
complex function of the input level.
Nonlinear distortion may take the form of
harmonic distortion, in which integer multiples of input frequencies occur, or intermodulation distortion (IMD), in which different
components multiply to make new ones.
Any departure from absolute linearity results in some form of nonlinear distortion,
and this distortion can work for us or against
us. Any amplifier, including a so-called linear amplifier, distorts nonlinearly to some
degree; any device or circuit that distorts
10.4 Chapter 10
Fig 10.4 — Feeding two signals into
one input of a mixer results in the same
output as if f1 and f2 are each first mixed
with f3 in two separate mixers, and the
outputs of these mixers are combined.
nonlinearly can work as a mixer, modulator, demodulator or frequency multiplier. An
amplifier optimized for linear operation will
nonetheless mix, but inefficiently; an amplifier biased for nonlinear amplification may
be practically linear over a given tiny portion of its input-signal range. The trick is to
use careful design and component selection
to maximize nonlinear distortion when we
want it (as in a mixer), and minimize it when
we don’t. Once we’ve decided to maximize
nonlinear distortion, the trick is to minimize
the distortion products we don’t want, and
maximize the products we want.
KEEPING UNWANTED DISTORTION
PRODUCTS DOWN
Ideally, a mixer multiplies the signal at one
of its inputs by the signal at its other input, but
does not multiply a signal at the same input
by itself, or multiple signals at the same input
by themselves or by each other. (Multiplying
a signal by itself — squaring it — generates
harmonic distortion [specifically, secondharmonic distortion] by adding the signal’s
frequency to itself per equation 7. Simultaneously squaring two or more signals generates
simultaneous harmonic and intermodulation
distortion, as we’ll see later when we explore
how a diode demodulates AM.)
Consider what happens when a mixer must
handle signals at two different frequencies
(we’ll call them f1 and f2) applied to its first
input, and a signal at a third frequency (f3)
applied to its other input. Ideally, a mixer
multiplies f1 by f3 and f2 by f3, but does not
multiply f1 and f2 by each other. This produces output at the sum and difference of f1
and f3, and the sum and difference of f2 and
f3, but not the sum and difference of f1 and
f2. Fig 10.4 shows that feeding two signals
into one input of a mixer results in the same
output as if f1 and f2 are each first mixed with
f3 in two separate mixers, and the outputs
of these mixers are combined. This shows
that a mixer, even though constructed with
nonlinearly distorting components, actually
behaves as a linear frequency shifter. Traditionally, we refer to this process as mixing and
to its outputs as mixing products, but we may
also call it frequency conversion, referring to
a device or circuit that does it as a converter,
and to its outputs as conversion products. If
a mixer produces an output frequency that is
higher than the input frequency, it is called an
upconverter; if the output frequency is lower
than the input, a downconverter.
Real mixers, however, at best act only as
reasonably linear frequency shifters, generating some unwanted IMD products — spurious signals, or spurs — as they go. Receivers
are especially sensitive to unwanted mixer
IMD because the signal-level spread over
which they must operate without generating
unwanted IMD is often 90 dB or more, and
includes infinitesimally weak signals in its
span. In a receiver, IMD products so tiny that
you’d never notice them in a transmitted signal can easily obliterate weak signals. This is
why receiver designers apply so much effort
to achieving “high dynamic range.”
The degree to which a given mixer, modulator or demodulator circuit produces unwanted
IMD is often the reason why we use it, or don’t
use it, instead of another circuit that does its
wanted-IMD job as well or even better.
OTHER MIXER OUTPUTS
In addition to desired sum-and-difference
products and unwanted IMD products, real
mixers also put out some energy at their input
frequencies. Some mixer implementations
may suppress these outputs — that is, reduce
one or both of their input signals by a factor
of 100 to 1,000,000, or 20 to 60 dB. This is
good because it helps keep input signals at
the desired mixer-output sum or difference
frequency from showing up at the IF terminal — an effect reflected in a receiver’s IF
rejection specification. Some mixer types,
especially those used in the vacuum-tube era,
suppress their input-signal outputs very little
or not at all.
Input-signal suppression is part of an
overall picture called port-to-port isolation.
Mixer input and output connections are traditionally called ports. By tradition, the port
to which we apply the shifting signal is the
local-oscillator (LO) port. By convention,
the signal or signals to be frequency-shifted
are applied to the RF (radio frequency) port,
and the frequency-shifted (product) signal
or signals emerge at the IF (intermediate
frequency) port. This illustrates the function
of a mixer in a receiver: Since it is often
impractical to achieve the desired gain and filtering at the incoming signal’s frequency (at
RF), a mixer is used to translate the incoming
RF signal to an intermediate frequency (the
IF), where gain and filtering can be applied.
The IF maybe be either lower or higher than
the incoming RF signal. In a transmitter, the
modulated signal may be created at an IF,
and then translated in frequency by a mixer
to the operating frequency.
Some mixers are bilateral; that is, their RF
and IF ports can be interchanged, depending
on the application. Diode-based mixers are
usually bilateral. Many mixers are not bilateral (unilateral); the popular SA602/612
Gilbert cell IC mixer is an example of this.
It’s generally a good idea to keep a mixer’s
input signals from appearing at its output
port because they represent energy that we’d
rather not pass on to subsequent circuitry. It
therefore follows that it’s usually a good idea
to keep a mixer’s LO-port energy from appearing at its RF port, or its RF-port energy
from making it through to the IF port. But
there are some notable exceptions.
10.2 Mixers and Amplitude Modulation
Now that we’ve just discussed what a fine
thing it is to have a mixer that doesn’t let its
input signals through to its output port, we
can explore a mixing approach that outputs
one of its input signals so strongly that the
fed-through signal’s amplitude at least equals
the combined amplitudes of the system’s sum
and difference products! This system fullcarrier, amplitude modulation, is the oldest
means of translating information into radio
form and back again. It’s a frequency-shifting system in which the original unmodulated signal, traditionally called the carrier,
emerges from the mixer along with the sum
and difference products, traditionally called
sidebands. The sidebar, “Mixer Math: Amplitude Modulation,” describes this process
mathematically.
10.2.1 Overmodulation
Since the information we transmit using
AM shows up entirely as energy in its sidebands, it follows that the more energetic we
make the sidebands, the more information
energy will be available for an AM receiver
to “recover” when it demodulates the signal. Even in an ideal modulator, there’s a
practical limit to how strong we can make
an AM signal’s sidebands relative to its carrier, however. Beyond that limit, we severely
distort the waveform we want to translate
into radio form.
We reach AM’s distortion-free modulation
limit when the sum of the sidebands and carrier at the modulator output just reaches zero
at the modulating wave-form’s most negative
peak (Fig 10.5). We call this condition 100%
modulation, and it occurs when m in equation
8 equals 1. (We enumerate modulation percentage in values from 0 to 100%. The lower
the number, the less information energy is in
the sidebands. You may also see modulation
enumerated in terms of a modulation factor
from 0 to 1, which directly equals m; a modulation factor of 1 is the same as 100% modulation.) Equation 9 shows that each sideband’s
voltage is half that of the carrier. Power varies
as the square of voltage, so the power in each
sideband of a 100%-modulated signal is therefore (1⁄2)2 times, or 1⁄4, that of the carrier. A
transmitter capable of 100% modulation when
operating at a carrier power of 100 W therefore
puts out a 150-W signal at 100% modulation,
50 W of which is attributable to the sidebands.
(The peak envelope power [PEP] output of a
double-sideband, full-carrier AM transmitter
at 100% modulation is four times its carrier
PEP. This is why our solid-state, “100-W”
MF/HF transceivers are usually rated for no
more than about 25 W carrier output at 100%
amplitude modulation.)
Mixer Math: Amplitude Modulation
We can easily make the carrier pop out of our mixer along with the sidebands
merely by building enough dc level shift into the information we want to mix so that
its waveform never goes negative. Back at equations 1 and 2, we decided to keep
our mixer math relatively simple by setting the peak voltage of our mixer’s input
signals directly equal to their sine values. Each input signal’s peak voltage therefore
varies between +1 and –1, so all we need to do to keep our modulating- signal term
(provided with a subscript m to reflect its role as the modulating or information waveform) from going negative is add 1 to it. Identifying the carrier term with a subscript
c, we can write
AM signal = (1 + m sin 2πfmt) sin 2πfc t
(8)
Notice that the modulation (2πfmt) term has company in the form of a coefficient,
m. This variable expresses the modulating signal’s varying amplitude — variations
that ultimately result in amplitude modulation. Expanding equation 8 according to
equation 6 gives us
1
1
(9)
AM signal = sin 2πfc t + m cos (2πfc − 2πfm )t − m cos (2πfc + 2πfm )t
2
2
The modulator’s output now includes the carrier (sin 2πfct) in addition to sum and
difference products that vary in strength according to m. According to the conventions of talking about modulation, we call the sum product, which comes out at a
frequency higher than that of the carrier, the upper sideband (USB), and the difference product, which comes out a frequency lower than that of the carrier, the lower
sideband (LSB). We have achieved amplitude modulation.
Why We Call It Amplitude Modulation
We call the modulation process described in equation 8 amplitude modulation
because the complex waveform consisting of the sum of the sidebands and carrier
varies with the information signal’s magnitude (m). Concepts long used to illustrate
AM’s mechanism may mislead us into thinking that the carrier varies in strength with
modulation, but careful study of equation 9 shows that this doesn’t happen. The carrier, sin 2πfct, goes into the modulator — we’re in the modulation business now, so
it’s fitting to use the term modulator instead of mixer — as a sinusoid with an unvarying maximum value of |1|. The modulator multiplies the carrier by the dc level (+1)
that we added to the information signal (m sin 2πfmt). Multiplying sin 2πfct by
1 merely returns sin 2πfct. We have proven that the carrier’s amplitude does not vary
as a result of amplitude modulation — a fact that makes sense when we realize
that in simple AM receivers the carrier serves as an AGC control signal and (during
“diode” detection) provides, at the correct power level and frequency, the “LO” signal
necessary to heterodyne the sidebands back to baseband.
Mixers, Modulators and Demodulators 10.5
One-hundred-percent negative modulation is a brick-wall limit because an amplitude modulator can’t reduce its output to
less than zero. Trying to increase negative
modulation beyond the 100% point results
in over-modulation (Fig 10.6), in which the
modulation envelope no longer mirrors the
shape of the modulating wave (Fig 10.6A).
A negatively overmodulated wave contains
more energy than it did at 100% modulation,
but some of the added energy now exists as
harmonics of the modulating waveform (Fig
10.6B). This distortion makes the modulated
signal take up more spectrum space than it
needs. In voice operation, overmodulation
commonly happens only on syllabic peaks,
making the distortion products sound like
transient noise we refer to as splatter.
MODULATION LINEARITY
If we increase an amplitude modulator’s
modulating-signal input by a given percentage, we expect a proportional modulation
increase in the modulated signal. We expect good modulation linearity. Suboptimal
amplitude modulator design may not allow
this, however. Above some modulation percentage, a modulator may fail to increase
modulation in proportion to an increase in its
input signal (Fig 10.7). Distortion, and thus
an unnecessarily wide signal, results.
Fig 10.6 — Negative-going overmodulation
of an AM transmitter results in a modulation envelope (A) that doesn’t faithfully
mirror the modulating waveform. This
distortion creates additional sideband
components that broaden the transmitted
signal (B). Positive-going modulation
beyond 100% is used by some AM broadcasters in conjunction with negative-peak
limiting to increase “talk power” without
causing negative overmodulation.
Fig 10.5 — Graphed in terms of
amplitude versus time (A), the envelope
of a properly modulated AM signal
exactly mirrors the shape of its
modulating waveform, which is a sine
wave in this example. This AM signal is
modulated as fully as it can be — 100%
— because its envelope just hits zero on
the modulating wave’s negative peaks.
Graphing the same AM signal in terms of
amplitude versus frequency (B) reveals
its three spectral components: Carrier,
upper sideband and lower sideband. B
shows sidebands as single-frequency
components because the modulating
waveform is a sine wave. With a complex
modulating waveform, the modulator’s
sum and difference products really do
show up as bands on either side of the
carrier (C).
10.6 Chapter 10
Fig 10.7 — An ideal AM transmitter
exhibits a straight-line relationship (A)
between its instantaneous envelope
amplitude and the instantaneous
amplitude of its modulating signal.
Distortion, and thus an unnecessarily
wide signal, results if the transmitter
cannot respond linearly across the
modulating signal’s full amplitude range.
10.2.2 Using AM to Send
Morse Code
Fig 10.8A closely resembles what we see
when a properly adjusted CW transmitter
sends a string of dots. Keying a carrier on and
off produces a wave that varies in amplitude
and has double (upper and lower) sidebands
that vary in spectral composition according
to the duration and envelope shape of the
on-off transitions. The emission mode we call
CW is therefore a form of AM. The concepts
of modulation percentage and overmodulation are usually not applied to generating an
on-off-keyed Morse signal, however. This is
related to how we copy CW by ear, and the
fact that, in CW radio communication, we
usually don’t translate the received signal all
the way back into its original pre-modulator
(baseband) form, as a closer look at the process reveals.
In CW transmission, we usually open and
close a keying line to make dc transitions
that turn the transmitted carrier on and off.
See Fig 10.8B. CW reception usually does
not entirely reverse this process, however.
Instead of demodulating a CW signal all the
way back to its baseband self — a shifting dc
level — we want the presences and absences
of its carrier to create long and short audio
tones. Because the carrier is RF and not AF,
we must mix it with a locally generated RF
signal — from a beat-frequency oscillator
(A)
HBK0409
(B)
Fig 10.8 — Telegraphy by on-off-keying
a carrier is also a form of AM, called CW
(short for continuous wave) for reasons
of tradition. Waveshaping in a CW
transmitter often causes a CW signal’s RF
envelope (lower trace in the amplitudeversus-time display at A) to contain
less harmonic energy than the abrupt
transitions of its key closure waveform
(upper trace in A) suggest should be the
case. B, an amplitude-versus-frequency
display, shows that even a properly
shaped CW signal has many sideband
components.
(BFO) — that’s close enough in frequency to
produce a difference signal at AF (this BFO
can, of course, also be inserted at an IF stage).
What goes into our transmitter as shifting dc
comes out of our receiver as thump-delimited
tone bursts of dot and dash duration. We have
achieved CW communication.
It so happens that we always need to hear
one or more harmonics of the fundamental keying waveform for the code to sound
sufficiently crisp. If the transmitted signal
will be subject to fading caused by varying
propagation — a safe assumption for any
long-distance radio communication — we
can harden our keying by making the transmitter’s output rise and fall more quickly.
This puts more energy into keying sidebands
and makes the signal more copiable in the
presence of fading — in particular, selective
fading, which linearly distorts a modulated
signal’s complex waveform and randomly
changes the sidebands’ strength and phase
relative to the carrier and each other. The
appropriate keying hardness also depends
on the keying speed. The faster the keying
in WPM, the faster the on-off times — the
harder the keying — must be for the signal
to remain ear- and machine-readable through
noise and fading.
Instead of thinking of this process in terms
of modulation percentage, we just ensure that
a CW transmitter produces sufficient keyingsideband energy for solid reception. Practical
CW transmitters ususally do not do their keying with a modulator stage as such. Instead,
one or more stages are turned on and off to
modulate the carrier with Morse, with rise and
fall times set by R and C values associated
with the stages’ keying and/or power supply
lines. A transmitter’s CW waveshaping is
therefore usually hardwired to values appropriate for reasonably high-speed sending (35
to 55 WPM or so) in the presence of fading.
However, some transceivers allow the user to
vary keying hardness at will as a menu option.
Rise and fall times of 1 to 5 ms are common;
5-ms rise and fall times equate to a keying
speed of 36 WPM in the presence of fading
and 60 WPM if fading is absent.
The faster a CW transmitter’s output changes between zero and maximum, the more
bandwidth its carrier and sidebands occupy.
See Fig 10.8B. Making a CW signal’s keying too hard is therefore spectrum-wasteful
and unneighborly because it makes the signal
wider than it needs to be. Keying sidebands
that are stronger and wider than necessary
are traditionally called clicks because of what
they sound like on the air. A more ­detailed
discussion of keying waveforms appears in
the Transmitters chapter of this Handbook.
10.2.3 The Many Faces of
Amplitude Modulation
We’ve so far examined mixers, multipliers and modulators that produce complex
output signals of two types. One, the action
of which equation 7 expresses, produces only
the frequency sum and frequency difference
between its input signals. The other, the amplitude modulator characterized by equations
8 and 9, produces carrier output in addition
to the frequency sum of and frequency difference between its input signals. Exploring the
AM process led us to a discussion of on-offkeyed CW, which is also a form of AM.
Amplitude modulation is nothing more and
nothing less than varying an output signal’s
amplitude according to a varying voltage or
current. All of the output signal types mentioned above are forms of amplitude modulation, and there are others. Their names and
applications depend on whether the resulting
signal contains a carrier or not, and both sidebands or not. Here’s a brief overview of AMsignal types, what they’re called, and some of
the jobs you may find them doing:
• Double-sideband (DSB), full-carrier
AM is often called just AM, and often what’s
meant when radio folk talk about just AM.
(When the subject is broadcasting, AM can
also refer to broadcasters operating in the
535- to 1705-kHz region, generically called
the AM band or the broadcast band or medium
wave. These broadcasters used only doublesideband, full-carrier AM for many years,
but many now use combinations of amplitude
modulation and angle modulation, explored
later in this chapter, to transmit stereophonic
sound in analog and digital form.) Equations
8 and 9 express what goes on in generating
this signal type. What we call CW — Morse
code done by turning a carrier on and off — is
a form of DSB, full-carrier AM.
• Double-sideband, suppressed-carrier
AM is what comes out of a circuit that does
what equation 7 expresses — a sum (upper
sideband), a difference (lower sideband) and
no carrier. We didn’t call its sum and difference outputs upper and lower sidebands
earlier in equation 7’s neighborhood, but
we’d do so in a transmitting application.
In a transmitter, we call a circuit that suppresses the carrier while generating upper and
lower sidebands a balanced modulator, and
we quantify its carrier suppression, which
is always less than infinite. In a receiver, we
call such a circuit a balanced mixer, which
may be single-balanced (if it lets either its RF
signal or its LO [carrier] signal through to its
output) or double-balanced (if it suppresses
both its input signal and LO/carrier in its output), and we quantify its LO suppression and
port-to-port isolation, which are always less
than infinite. (Mixers [and amplifiers] that
afford no balance whatsoever are sometimes
said to be single-ended.) Sometimes, DSB
suppressed-carrier AM is called just DSB.
• Vestigial sideband (VSB), full-carrier
AM is like the DSB variety with one sideband
partially filtered away for bandwidth reduction. Some amateur television stations use
VSB AM, and it was used by commercial
television systems that transmitted analog
AM video in the pre-digital TV days.
• Single-sideband, suppressed-carrier AM
is what you get when you generate a DSB,
suppressed carrier AM signal and suppress
one sideband with filtering or phasing. We
usually call this signal type just single sideband (SSB) or, as appropriate, upper sideband
(USB) or lower sideband (LSB). In a modulator or demodulator system, the unwanted
sideband — that is, the sum or difference signal we don’t want — may be called just that,
or it may be called the opposite sideband, and
we refer to a system’s sideband rejection as
a measure of how well the opposite sideband
is suppressed. In receiver mixers not used for
demodulation and transmitter mixers not used
for modulation, the unwanted sum or difference signal, or the input signal that produces
the unwanted sum or difference, is the image,
and we refer to a system’s image rejection.
Mixers, Modulators and Demodulators 10.7
A pair of mixers specially configured to suppress either the sum or the difference output
is an image-reject mixer (IRM). In receiver
demodulators, the unwanted sum or difference signal may just be called the opposite
sideband, or it may be called the audio image.
A receiver capable of rejecting the opposite
sideband or audio image is said to be capable
of single-signal reception.
• Single-sideband, full-carrier AM is akin
to full-carrier DSB with one sideband missing. Commercial and military communi­
cators may call it AM equivalent (AME) or
compatible AM (CAM) — compatible because
it can be usefully demodulated in AM and SSB
receivers and because it occupies about the
same amount of spectrum space as SSB.)
• Independent sideband (ISB) AM consists
of an upper sideband and a lower sideband
containing different information (a carrier
of some level may also be present). Radio
amateurs sometimes use ISB to transmit simultaneous slow-scan-television and voice
information; international broadcasters
sometimes use it for point-to-point audio
feeds as a backup to satellite links.
10.2.4 Mixers and AM
Demodulation
Translating information from radio form
back into its original form — demodulation — is also traditionally called detection.
If the information signal we want to detect
consists merely of a baseband signal frequency-shifted into the radio realm, almost any
low-distortion frequency-shifter that works
according to equation 7 can do the job acceptably well.
Sometimes we recover a radio signal’s
information by shifting the signal right back
to its original form with no intermediate frequency shifts. This process is called direct
conversion. More commonly, we first convert
a received signal to an intermediate frequency
so we can amplify, filter and level-control it
prior to detection. This is superheterodyne
reception, and most modern radio receivers
work in this way. Whatever the receiver type,
however, the received signal ultimately makes
its way to one last mixer or demodulator that
completes the final translation of information back into audio, video, or into a signal
form suitable for device control or computer
processing. In this last translation, the incoming signal is converted back to recoveredinformation form by mixing it with one last
RF signal. In heterodyne or product detection,
that final frequency-shifting signal comes
from a BFO. The incoming-signal energy
goes into one mixer input port, BFO energy
goes into the other, and audio (or whatever
form the desired information takes) results.
If the incoming signal is full-carrier AM
and we don’t need to hear the carrier as a tone,
10.8 Chapter 10
Fig 10.9 — Radio’s simplest demodulator, the diode rectifier (A), demodulates an
AM signal by acting as a switch that multiplies the carrier and sidebands to produce
frequency sums and differences, two of which sum into a replica of the original
modulation (B). Modern receivers often use an emitter follower to provide lowimpedance drive for their diode detectors (C).
we can modify this process somewhat, if we
want. We can use the carrier itself to provide
the heterodyning energy in a process called
envelope detection.
ENVELOPE DETECTION AND FULLCARRIER AM
Fig 10.5 graphically represents how a fullcarrier AM signal’s modulation envelope
corresponds to the shape of the modulating
wave. If we can derive from the modulated
signal a voltage that varies according to the
modulation envelope, we will have successfully recovered the information present in the
sidebands. This process is called envelope
detection, and we can achieve it by doing
nothing more complicated than half-waverectifying the modulated signal with a diode
(Fig 10.9).
That a diode demodulates an AM signal by allowing its carrier to multiply with
its sidebands may jar those long accustomed to seeing diode detection ascribed
merely to “rectification.” But a diode is certainly nonlinear. It passes current in only one
direction, and its output voltage is (within
limits) proportional to the square of its input voltage. These nonlinearities allow it to
multiply.
Exploring this mathematically is tedious
with full-carrier AM because the process
squares three summed components (carrier,
lower sideband and upper sideband). Rather
than fill the better part of a page with algebra,
we’ll instead characterize the outcome verbally: In “just rectifying” a DSB, full-carrier
AM signal, a diode detector produces
• Direct current (the result of rectifying
the carrier);
• A second harmonic of the carrier;
• A second harmonic of the lower sideband;
• A second harmonic of the upper sideband;
• Two difference-frequency outputs (upper
sideband minus carrier, carrier minus lower
sideband), each of which is equivalent to the
modulating wave-form’s frequency, and both
of which sum to produce the recovered information signal; and
• A second harmonic of the modulating
waveform (the frequency difference between
the two sidebands).
Three of these products are RF. Low-pass
filtering, sometimes little more than a simple
RC network, can remove the RF products
from the detector output. A capacitor in series
with the detector output line can block the
carrier-derived dc component. That done,
only two signals remain: the recovered modulation and, at a lower level, its second harmonic — in other words, second-harmonic
distortion of the desired information signal.
10.3 Mixers and Angle Modulation
Amplitude modulation served as our first
means of translating information into radio
form because it could be implemented as simply as turning an electric noise generator on and
off. (A spark transmitter consisted of little more
than this.) By the 1930s, we had begun experimenting with translating information into radio
form and back again by modulating a radio
wave’s angular velocity (frequency or phase)
instead of its overall amplitude. The result of
this process is frequency modulation (FM) or
phase modulation (PM), both of which are often grouped under the name angle modulation
because of their underlying principle.
A change in a carrier’s frequency or phase
for the purpose of modulation is called deviation. An FM signal deviates according to
the amplitude of its modulating waveform,
independently of the modulating waveform’s
frequency; the higher the modulating wave’s
amplitude, the greater the deviation. A PM
signal deviates according to the amplitude
and frequency of its modulating waveform;
the higher the modulating wave’s amplitude
and/or frequency, the greater the deviation.
See the sidebar, “Mixer Math: Angle Modulation” for a numerical description of these
processes.
10.3.1 Angle Modulation
Sidebands
Although angle modulation produces un-
Mixer Math: Angle Modulation
An angle-modulated signal can be mathematically represented as
fc (t) = cos (2πfc t + m sin (2πfmt))
= cos (2πfc t) cos (m sin (2πfmt)) − sin (2πfc t) sin (m sin (2πfm t))
(10)
In equation 10, we see the carrier frequency (2πfct) and modulating signal (sin
2πfmt) as in the equation for AM (equation 8). We again see the modulating signal
associated with a coefficient, m, which relates to degree of modulation. (In the
AM equation, m is the modulation factor; in the angle-modulation equation, m is
the modulation index and, for FM, equals the deviation divided by the modulating
frequency.) We see that angle-modulation occurs as the cosine of the sum of the
carrier frequency (2πfct) and the modulating signal (sin 2πfmt) times the modulation
index (m). In its expanded form, we see the appearance of sidebands above and
below the carrier frequency.
Angle modulation is a multiplicative process, so, like AM, it creates sidebands on
both sides of the carrier. Unlike AM, however, angle modulation creates an infinite
number of sidebands on either side of the carrier! This occurs as a direct result of
modulating the carrier’s angular velocity, to which its frequency and phase directly
relate. If we continuously vary a wave’s angular velocity according to another
periodic wave’s cyclical amplitude variations, the rate at which the modulated wave
repeats its cycle — its frequency — passes through an infinite number of values.
(How many individual amplitude points are there in one cycle of the modulating
wave? An infinite number. How many corresponding discrete frequency or phase
values does the corresponding angle-modulated wave pass through as the modulating signal completes a cycle? An infinite number!) In AM, the carrier frequency stays
at one value, so AM produces two sidebands — the sum of its carrier’s unchanging
frequency value and the modulating frequency, and the difference between the carrier’s unchanging frequency value and the modulating frequency. In angle modulation, the modulating wave shifts the frequency or phase of the carrier through an
infinite number of different frequency or phase values, resulting in an infinite number
of sum and difference products.
Mixers, Modulators and Demodulators 10.9
(A mathematical tool called Bessel functions
helps determine the relative strength of the
carrier and sidebands according to modulation index. The Modulation chapter includes
a graph to illustrate this relationship.) Selectivity in transmitter and receiver circuitry
further modify this relationship, especially
for sidebands far away from the carrier.
10.3.2 Angle Modulators
Fig 10.12 — A series reactance modulator
acts as a variable shunt around a
reactance — in this case, a 47-pF
capacitor — through which the carrier
passes.
Fig 10.10 — Angle-modulation produces
a carrier and an infinite number of upper
and lower sidebands spaced from the
average (“resting,” unmodulated) carrier
frequency by integer multiples of the
modulating frequency. (This drawing is
a simplification because it only shows
relatively strong, close-in sideband
pairs; space constraints prevent us
from extending it to infinity.) The relative
amplitudes of the sideband pairs and
carrier vary with modulation index, m.
countable sum and difference products, most
of them are vanishingly weak in practical
systems. They emerge from the modulator
spaced from the average (“resting,” unmodulated) carrier frequency by integer multiples
of the modulating frequency (Fig 10.10). The
strength of the sidebands relative to the carrier, and the strength and phase of the carrier
itself, vary with the degree of modulation
— the modulation index. (The overall amplitude of an angle-modulated signal does
not change with modulation, however; when
energy goes out of the carrier, it shows up in
the sidebands, and vice versa.) In practice,
we operate angle-modulated transmitters at
modulation indexes that make all but a few
of their infinite sidebands small in amplitude.
Fig 10.11 — One or more tuning diodes can serve as the variable reactance in a
reactance modulator. This HF reactance modulator circuit uses two diodes in series
to ensure that the tuned circuit’s RF-voltage swing cannot bias the diodes into
conduction. D1 and D2 are “30-volt” tuning diodes that exhibit a capacitance of 22 pF
at a bias voltage of 4. The bias control sets the point on the diode’s voltage-versuscapacitance characteristic around which the modulating waveform swings.
10.10 Chapter 10
If you vary a reactance in or associated
with an oscillator’s frequency-determining
element(s), you vary the oscillator’s frequency. If you vary the tuning of a tuned circuit
through which a signal passes, you vary the
signal’s phase. A circuit that does this is
called a reactance modulator, and can be little
more than a tuning diode or two connected to
a tuned circuit in an oscillator or amplifier
(Fig 10.11). Varying a reactance through
which the signal passes (Fig 10.12) is another
way of doing the same thing.
The difference between FM and PM depends solely on how, and not how much,
deviation occurs. A modulator that causes
deviation in proportion to the modulating
wave’s amplitude and frequency is a phase
modulator. A modulator that causes deviation
only in proportion to the modulating signal’s
amplitude is a frequency modulator.
INCREASING DEVIATION BY
FREQUENCY MULTIPLICATION
Maintaining modulation linearity is just
as important in angle modulation as it is in
AM, because unwanted distortion is always
our enemy. A given angle-modulator circuit
can frequency- or phase-shift a carrier only
so much before the shift stops occurring in
strict proportion to the amplitude (or, in PM,
the amplitude and frequency) of the modulating signal.
If we want more deviation than an angle
modulator can linearly achieve, we can operate the modulator at a suitable sub-harmonic
— submultiple — of the desired frequency,
and process the modulated signal through a
series of frequency multipliers to bring it up
to the desired frequency. The deviation also
increases by the overall multiplication factor,
relieving the modulator of having to do it all
directly. A given FM or PM radio design may
achieve its final output frequency through a
combination of mixing (frequency shift, no
deviation change) and frequency multiplication (frequency shift and deviation change).
THE TRUTH ABOUT “TRUE FM”
Something we covered a bit earlier bears
closer study:
“An FM signal deviates according to the
amplitude of its modulating waveform, independently of the modulating wave-form’s
frequency; the higher the modulating wave’s
amplitude, the greater the deviation. A PM
signal deviates according to the amplitude
and frequency of its modulating waveform;
the higher the modulating wave’s amplitude
and/or frequency, the greater the deviation.”
The practical upshot of this excerpt is that
we can use a phase modulator to generate FM.
All we need to do is run a PM transmitter’s
modulating signal through a low-pass filter
that (ideally) halves the signal’s amplitude for
each doubling of frequency (a reduction of
“6 dB per octave,” as we sometimes see such
responses characterized) to compensate for
its phase modulator’s “more deviation with
higher frequencies” characteristic. The result
is an FM, not PM, signal. FM achieved with a
phase modulator is sometimes called indirect
FM as opposed to the direct FM we get from
a frequency modulator.
We sometimes see claims that one piece
of gear is better than another solely because
it generates “true FM” as opposed to indirect
FM. We can debunk such claims by keeping
in mind that direct and indirect FM sound
exactly alike in a receiver when done correctly.
CONVEYING DC LEVELS WITH
ANGLE MODULATION
Depending on the nature of the modulation source, there is a practical difference
between a frequency modulator and a phase
modulator. Answering two questions can tell
us whether this difference matters: Does our
modulating signal contain a dc level or not?
If so, do we need to accurately preserve that
dc level through our radio communication
link for successful communication? If both
answers are yes, we must choose our hardware and/or information-encoding approach
carefully, because a frequency modulator
can convey dc-level shifts in its modulating
waveform, while a phase modulator, which
responds only to instantaneous changes in
frequency and phase, cannot.
Consider what happens when we want to
frequency-modulate a phase-locked-loopsynthesized transmitted signal. Fig 10.13
shows the block diagram of a PLL frequency
modulator. Normally, we modulate a PLL’s
VCO because it’s the easy thing to do. As
long as our modulating frequency results in
frequency excursions too fast for the PLL to
follow and correct — that is, as long as our
modulating frequency is outside the PLL’s
loop bandwidth — we achieve the FM we
seek. Trying to modulate a dc level by pushing
the VCO to a particular frequency and holding it there fails, however, because a PLL’s
loop response includes dc. The loop, therefore, detects the modulation’s dc component
as a correctable error and “fixes” it. FMing
a PLL’s VCO therefore can’t buy us the dc
response “true FM” is supposed to allow.
Fig 10.13 — Frequency modulation, PLL-style.
Fig 10.14 — Frequency-sweeping a constant-amplitude signal and passing it through
a low-pass filter results in an output signal that varies in amplitude with frequency.
This is the principle behind the angle-demodulation process called frequency
discrimination.
We can dc-modulate a PLL modulator, but
we must do so by modulating the frequency
of the loop reference. The PLL then adjusts
the VCO to adapt to the changed reference,
and our dc level gets through. In this case, the
modulating frequency must be within the loop
bandwidth — which dc certainly is — or the
VCO won’t be corrected to track the shift.
10.3.3 Mixers and Angle
Demodulation
With the awesome prospect of generating
an infinite number of sidebands still fresh in
our minds, we may be a bit disappointed to
learn that we commonly demodulate angle
modulation by doing little more than turning
it into AM and then envelope- or product-
detecting it! But this is what happens in many
of our FM receivers and transceivers, and we
can get a handle on this process by realizing
that a form of angle-modulation-to-AM conversion begins quite early in an angle-modulated signal’s life because of linear distortion
of the modulation by amplitude-linear circuitry
— something that happens to angle-modulated
signals, it turns out, in any linear circuit that
doesn’t have an amplitude-versus-frequency
response that’s utterly flat out to infinity.
Think of what happens, for example, when
we sweep a constant-amplitude signal up in
frequency — say, from 1 kHz to 8 kHz —
and pass it through a 6-dB-per-octave filter (Fig 10.14). The filter’s rolloff causes
the output signal’s amplitude to decrease
as frequency increases. Now imagine that
Mixers, Modulators and Demodulators 10.11
we linearly sweep our constant-amplitude
signal back and forth between 1 kHz and
8 kHz at a constant rate of 3 kHz per second.
The filter’s output amplitude now varies
cyclically over time as the input signal’s
frequency varies cyclically over time. Right
before our eyes, a frequency change turns
into an amplitude change. The process of
converting angle modulation to amplitude
modulation has begun.
This is what happens whenever an anglemodulated signal passes through circuitry
with an amplitude-versus-frequency response
that isn’t flat out to infinity. As the signal deviates across the frequency- response curves of
whatever circuitry passes it, its angle modulation is, to some degree, converted to AM — a
form of crosstalk between the two modulation types, if we wish to look at it that way.
(Variations in system phase linearity also
cause distortion and FM-to-AM conversion,
because the sidebands do not have the proper
phase relationship with respect to each other
and with respect to the carrier.)
All we need to do to put this effect to practical use is develop a circuit that does this
frequency-to-amplitude conversion linearly
across the frequency span of the modulated
signal’s deviation. Then we envelope-demodulate the resulting AM, and we have achieved
angle demodulation.
Fig 10.15 shows such a circuit — a discriminator — and the sort of amplitude-versus-frequency response we expect from it.
It’s actually possible to use an AM receiver
to recover understandable audio from a narrow angle-modulated signal by “off-tuning”
the signal so its deviation rides up and down
on one side of the receiver’s IF selectivity
curve. This slope detection process served as
an early, suboptimal form of frequency discrimination in receivers not designed for FM
It is always worth trying as a last-resort-class
means of receiving narrowband FM with an
AM receiver.
QUADRATURE DETECTION
It’s also possible to demodulate an anglemodulated signal merely by multiplying it
with a time-delayed copy of itself in a doublebalanced mixer as shown in Fig 10.16; the
sidebar, “Mixer Math: Quadrature Demodulation,” explains the process numerically.
An ideal quadrature detector puts out 0 V
dc when no modulation is present (with the
carrier at fc). The output of a real quadrature detector may include a small dc offset
that requires compensation. If we need the
detector’s response all the way down to dc,
we’ve got it; if not, we can put a suitable
10.12 Chapter 10
Fig 10.15 — A discriminator (A) converts an angle-modulated signal’s deviation into an
amplitude variation (B) and envelope-detects the resulting AM signal. For undistorted
demodulation, the discriminator’s amplitude-versus-frequency characteristic must be
linear across the input signal’s deviation. A crystal discriminator uses a crystal as part
of its frequency-selective circuitry.
Fig 10.16 — In quadrature detection, an angle-modulated signal multiplies with a
time-delayed copy of itself to produce a dc voltage that varies with the amplitude
and polarity of its phase or frequency excursions away from the carrier frequency. A
practical quadrature detector can be as simple as a 0° power splitter (that is, a power
splitter with in-phase outputs), a diode double-balanced mixer, a length of coaxial cable
1
⁄4-λ (electrical) long at the carrier frequency, and a bit of low-pass filtering to remove
the detector output’s RF components. IC quadrature detectors achieve their time delay
with one or more resistor-loaded tuned circuits (Fig 10.17).
blocking capacitor in the output line for aconly coupling.
Quadrature detection is more common
than frequency discrimination nowadays
because it doesn’t require a special discrimi-
nator transformer or resonator, and because
the necessary balanced-detector circuitry can
easily be implemented in IC structures along
with limiters and other receiver circuitry. The
NXP Semiconductor SA604A FM IF IC is
Mixer Math: Quadrature Demodulation
Demodulating an angle-modulated signal merely by multiplying it with a timedelayed copy of itself in a double-balanced mixer results in quadrature demodulation
(Fig 10.16). To illustrate this mathematically, for simplicity’s sake, we’ll represent the
mixer’s RF input signal as just a sine wave with an amplitude, A:
A sin (2πft) (11)
and its time-delayed twin, fed to the mixer’s LO input, as a sine wave with an
amplitude, A, and a time delay of d:
A sin [2 π f (t + d)] (12)
Setting this special mixing arrangement into motion, we see
A sin (2πft) × A sin (2πft + d)
=
A2
A2
A2
cos (2πfd) −
cos (2πfd) cos (2 × 2πft) +
sin (2πfd) sin (2 × 2πft)
2
2
2
(13)
Two of the three outputs — the second and third terms — emerge at twice the input
frequency; in practice, we’re not interested in these, and filter them out. The remaining term — the one we’re after — varies in amplitude and sign according to how far
and in what direction the carrier shifts away from its resting or center frequency (at
which the time delay, d, causes the mixer’s RF and LO inputs to be exactly 90° out of
phase — in quadrature — with each other). We can examine this effect by replacing f
in equations 11 and 12 with the sum term fc + fs, where fc is the center frequency and
fs is the frequency shift. A 90° time delay is the same as a quarter cycle of fc, so we
can restate d as
1
4fc The first term of the detector’s output then becomes
d=
(14)
A2
cos ( 2π(fc + fs )d)
2
=

A2
1 
cos  2π(fc + fs )

2
4fc 

=
 π πf 
A2
cos  + s 
2
 2 2fc  (15)
When fs is zero (that is, when the carrier is at its center frequency), this reduces to
A2
 π
cos   = 0
2
 2
(16)
As the input signal shifts higher in frequency than fc, the detector puts out a positive dc voltage that increases with the shift. When the input signal shifts lower in
frequency than fc, the detector puts out a negative dc voltage that increases with the
shift. The detector therefore recovers the input signal’s frequency or phase modulation
as an amplitude-varying dc voltage that shifts in sign as fs varies around fc — in other
words, as ac. We have demodulated FM by means of quadrature detection.
one example of this; Fig 10.17 shows another, the Freescale Semiconductor (formerly
Motorola) MC3359 (equivalent, NTE860).
PLL ANGLE DEMODULATION
Back at Fig 10.14, we saw how a phaselocked loop can be used as an angle modu­lator.
A PLL also makes a fine angle demodulator.
Applying an angle-modulated signal to a PLL
keeps its phase detector and VCO hustling to
maintain loop lock through the input signal’s
angle variations. The loop’s error voltage
therefore tracks the input signal’s modulation,
and its variations mirror the modulation signal. Turning the loop’s varying dc error voltage into audio is just a blocking capacitor
away.
Although we can’t convey a dc level by
directly modulating the VCO in a PLL angle
modulator, a PLL demodulator can respond
down to dc quite nicely. A constant frequency
offset from fc (a dc component) simply causes
a PLL demodulator to swing its VCO over
to the new input frequency, resulting in a
proportional dc offset on the VCO controlvoltage line. Another way of looking at the
difference between a PLL angle modulator
and a PLL angle demodulator is that a PLL
demodulator works with a varying reference
signal (the input signal), while a PLL angle
modulator generally doesn’t.
AMPLITUDE LIMITING REQUIRED
By now, it’s almost household knowledge
that FM radio communication systems are
superior to AM in their ability to suppress
and ignore static, manmade electrical noise
and (through a characteristic called capture
effect) co-channel signals sufficiently weaker
than the desired signal. AM-noise immunity
and capture effect are not intrinsic to angle
modulation, however; they must be designed
into the angle-modulation receiver in the form
of signal amplitude limiting.
If we note the progress of A from the left
to the right side of the equal sign in equation 13, we realize that the amplitude of a
quadrature detector’s input signal affects the
amplitude of a quadrature detector’s three
output signals. A quadrature detector therefore responds to AM, and so does a frequency
discriminator. To achieve FM’s storied noise
immunity, then, these angle demodulators
must be preceded by limiter circuitry that
removes all amplitude variations from the
incoming signal.
Mixers, Modulators and Demodulators 10.13
Putting Mixer Math to
Work through DSP
Fig 10.17 — The Freescale MC3359/NTE680 is one of many FM subsystem ICs that
include limiter and quadrature-detection circuitry. The TIME DELAY coil is adjusted for
minimum recovered-audio distortion.
10.4 Putting Mixers, Modulators and
Demodulators to Work
Variations on relatively few mixer types
provide mixing, modulation, and demodulation functions in the majority of Amateur Radio receivers, transmitters, and transceivers.
Comparing their behaviors and specifications
begins with a grounding in key mixer characteristics related to dynamic range — the
span of signal strengths a circuit can handle
without generating false signals that can interfere with communication.
10.4.1 Dynamic
Range: Compression,
Intermodulation and More
The output of a linear stage — including
a mixer, which we want to act as a linear
frequency shifter — tracks its input signal
decibel by decibel, every 1-dB change in its
10.14 Chapter 10
input signal(s) corresponding to an identical
1-dB output change. This is the stage’s firstorder response.
Because no device is perfectly linear, however, two or more signals applied to it generate
sum and difference frequencies. These IMD
products occur at frequencies and amplitudes
that depend on the order of the IMD response
as follows:
• Second-order IMD products change
2 dB for every decibel of input-signal change
(this figure assumes that the IMD comes from
equal-level input signals), and appear at frequencies that result from the simple addition
and subtraction of input-signal frequencies.
For example, assuming that its input bandwidth is sufficient to pass them, an amplifier
subjected to signals at 6 and 8 MHz will produce second-order IMD products at 2 MHz
In describing the idealized behavior of
mixing, modulation and demodulation circuits and processes through mathematics, this chapter’s Mixer Math sidebars
reflect a profound underlying reality: If
what a mixer, modulator or demodulator
does can be described with math, then all
a mixer, modulator or demodulator does
is math. This is true not only of mixers,
modulators and demodulators, but of all
electrical and electronic circuitry, analog
and digital. An electrical or electronic
circuit is merely a math machine — a
machine that, for much of the history
of human use of radio and electronics,
has necessarily been hardwired to be
capable of solving only one equation or
set or class of equations. When you build
an electronic circuit, interconnecting its
resistors, capacitors, inductors, active
devices and other components, you
build out the signal-processing equation it will solve, atom by mathematical
atom. Adjusting the controls of such a
purpose-built math machine, every switch
you throw or control modifies the equation it solves, and therefore the work it
performs.
No matter how painstakingly and elegantly engineered and constructed such
a practical realization of math may be,
however, it is only so operationally flexible. If you want to significantly alter the
equation(s) it solves, you must change its
circuitry by replacing and/or reconnecting
components and/or by adjusting whatever
controls you may have included to tweak
its operation on the fly.
The Computer-Aided Circuit Design
chapter of this Handbook also characterizes electrical and electronic circuits as
math machines. It does so for the purpose
of exploring the use of personal computers — which are generic, arbitrarily and
readily programmable and reprogrammable math machines — as engineering tools capable of simulating, through
mathematical modeling, the electrical and
electronic behavior of real-world circuits.
The next logical expansion of the use of
electronic computers in radio is to move
beyond just simulating the mathematical
signal-processing functions of circuits to
actually performing those functions in real
time through mathematics applied in real
time —as physical circuitry has done for
decades and still does, but better, and
more flexibly, repeatably, and modifiably.
Such real-time applied-math techniques,
well-established and increasingly widely
applied as digital signal processing (DSP)
and fundamental to the design, evolution,
and growing power of software-defined
radios (SDRs), are likely already in use,
or will soon be, in a radio you own. You
can learn more about DSP and its techniques, applications and limitations in this
Handbook’s DSP and Software Radio
Design chapter.
(8–6) and 14 MHz (8+6).
• Third-order IMD products change 3 dB
for every decibel of input-signal change (this
also assumes equal-level input signals), and
appear at frequencies corresponding to the
sums and differences of twice one signal’s
frequency plus or minus the frequency of
another. Assuming that its input bandwidth is
sufficient to pass them, an amplifier subjected
to signals at 14.02 MHz (f1) and 14.04 MHz
(f2) produces third-order IMD products at
14.00 (2f1 – f2), 14.06 (2f2 – f1), 42.08 (2f1 +
f2) and 42.10 (2f2 + f1) MHz. The subtractive
products (the 14.02 and 14.04-MHz products
in this example) are close to the desired signal and can cause significant interference.
Thus, third-order-IMD performance is of
great importance in receiver mixers and RF
amplifiers.
It can be seen that the IMD order determines how rapidly IMD products change
level per unit change of input level. Nth- order
IMD products therefore change by n dB for
every decibel of input-level change.
IMD products at orders higher than three
can and do occur in communication systems,
but their magnitudes are usually much smaller
than the lower-order products. The secondand third-order products are most important
in receiver front ends. In transmitters, thirdand higher-odd-order products are important
because they widen the transmitted signal.
10.4.2 Intercept Point
The second type of dynamic range concerns
the receiver’s intercept point, sometimes simply referred to as intercept. Intercept point
is typically measured (or calculated by ex-
trapolation of the test results) by applying two
or three signals to the antenna input, tuning
the receiver to count the number of resulting
spurious responses, and measuring their level
relative to the input signal.
Because a device’s IMD products increase
more rapidly than its desired output as the
input level rises, it might seem that steadily
increasing the level of multiple signals applied to an amplifier would eventually result
in equal desired-signal and IMD levels at the
amplifier output. Real devices are incapable
of doing this, however. At some point, every
device overloads, and changes in its output
level no longer equally track changes at its
input. The device is then said to be operating in compression; the point at which its
first-order response deviates from linearity by
1 dB is its 1-dB compression point. Pushing
the process to its limit ultimately leads to saturation, at which point input-signal increases
no longer increase the output level.
The power level at which a device’s second-order IMD products would equal its firstorder output (a point that must be extrapolated
because most likely the device is in compression by this point) is its second-order intercept point. Likewise, its third-order intercept
point is the power level at which third-order
responses would equal the desired signal.
Fig 10.18 graphs these relationships.
Input filtering can improve second-order intercept point; device nonlinearities
­determine the third, fifth and higher-oddnumber intercept points. In preamplifiers
and active mixers, third-order intercept
point is directly related to dc input power;
in passive­ switching mixers, to the localoscillator power applied.
Fig 10.18 — A linear stage’s output tracks
its input decibel by decibel on a 1:1 slope
— its first-order response. Second-order
intermodulation distortion (IMD) products
produced by two equal-level input signals
(“tones”) rise on a 2:1 slope — 2 dB for
­every 1 dB of input increase. Third-order
IMD products likewise increase 3 dB for
­every 1 dB of increase in two equal tones.
For each IMD order n, there is a corresponding intercept point IPn at which the
stage’s first-order and nth order products
are equal in amplitude. The first-order
output of real amplifiers and mixers falls
off (the device overloads and goes into
compression) before IMD products can
intercept it, but intercept point is nonetheless a useful, valid concept for comparing
radio system performance. The higher an
amplifier or mixer’s intercept point, the
stronger the input signals it can handle
without overloading. The input and output
powers shown are for purposes of example; every device exhibits its own particular
IMD profile. (After W. Hayward, Introduction
to Radio Frequency Design, Fig 6.17)
Mixers, Modulators and Demodulators 10.15
Testing and Calculating Intermodulation Distortion in Receivers
Second and third-order IMD can be measured using the
setup of Fig 10.A1. The outputs of two signal generators are
combined in a 3-dB hybrid coupler. Such couplers are available from various companies, and can be homemade. The
3-dB coupler should have low loss and should itself produce
negligible IMD. The signal generators are adjusted to provide
a known signal level at the output of the 3-dB coupler, say,
–20 dBm for each of the two signals. This combined signal is
then fed through a calibrated variable attenuator to the device
under test. The shielding of the cables used in this system is
important: At least 90 dB of isolation should exist between the
high-level signal at the input of the attenuator and the low-level
signal delivered to the receiver.
The measurement procedure is simple: adjust the variable
attenuator to produce a signal of known level at the frequency
of the expected IMD product (f1 ± f2 for second-order, 2f1 – f2
or 2f2 – f1 for third-order IMD).
To do this, of course, you have to figure out what equivalent input signal level at the receiver’s operating frequency
corresponds to the level of the IMD product you are seeing.
There are several ways of doing this. One way — the way used
by the ARRL Lab in their receiver tests — uses the minimum
discernible signal. This is defined as the signal level that produces a 3-dB increase in the receiver audio output power. That
is, you measure the receiver output level with no input signal,
then insert a signal at the operating frequency and adjust the
level of this input signal until the output power is 3 dB greater
than the no-signal power. Then, when doing the IMD measurement, you adjust the attenuator of Fig 10.A1 to cause a 3-dB
increase in receiver output. The level of the IMD product is
then the same as the MDS level you measured.
There are several things I dislike about doing the measurement this way. The problem is that you have to measure noise
power. This can be difficult. First, you need an RMS voltmeter
or audio power meter to do it at all. Second, the measurement
varies with time (it’s noise!), making it difficult to nail down a
number. And third, there is the question of the audio response
of the receiver; its noise output may not be flat across the output spectrum. So I prefer to measure, instead of MDS, a higher
reference level. I use the receiver’s S meter as a reference.
I first determine the input signal level it takes to get an S1
reading. Then, in the IMD measurement, I adjust the attenuator to again give an S1 reading. The level of the IMD product
signal is now equal to the level I measured at S1. Note that
this technique gives a different IMD level value than the MDS
technique. That’s OK, though. What we are trying to determine
is the difference between the level of the signals applied to the
receiver input and the level of the IMD product. Our calculations will give the same result whether we measure the IMD
product at the MDS level, the S1 level or some other level.
An easy way to make the reference measurement is with the
setup of Fig 10.A1. You’ll have to switch in a lot of attenuation
(make sure you have an attenuator with enough range), but
doing it this way keeps all of the possible variations in the measurement fairly constant. And this way, the difference between
the reference level and the input level needed to produce the
desired IMD product signal level is simply the difference in
attenuator settings between the reference and IMD measurements.
Calculating Intercept Points
Once we know the levels of the signals applied to the
receiver input and the level of the IMD product, we can easily
calculate the intercept point using the following equation:
IPn =
n × PA − PIMn
n−1
(A)
Here, n is the order, PA is the receiver input power (of one of
the input signals), PIMn is the power of the IMD product signal,
Fig 10.A1 — Test setup for
measurement of IMD performance.
Both signal generators should be
types such as HP 608, HP 8640, or
Rohde & Schwarz SMDU, with phasenoise performance of –140 dBc/Hz
or better at 20 kHz from the signal
frequency.
10.5 A Survey of Common Mixer Types
Depending on the application, frequencymixer implementations may vary from the
extremely simple to the complex. For example, a simple half-wave rectifier (a signal diode, such as a 1N34 [germanium] or a
1N914 [silicon]) can do the job (as we saw
at Fig 10.9 in our discussion of full-carrier
AM demodulation); this is an example of a
switching mixer, in which mixing occurs as
10.16 Chapter 10
one signal — in this case, the carrier, which in
effect turns the diode on and off as its polarity reverses — interrupts the transmission of
another (in demodulation of full-carrier AM,
the sideband[s]).
A switch can be thought of as an amplifier
toggled between two gain states, off and on,
by a control signal. It turns out that a binary
amplifier is not necessary; any device that
can be gain-varied in accordance with the
amplitude of a control signal can serve as a
frequency mixer.
10.5.1 Gain-Controlled
Analog Amplifiers As Mixers
Fig 10.19 shows two analog-amplifier
mixing arrangements commonly encoun-
and IPn is the nth-order intercept point. All powers should be
in dBm. For second and third-order IMD, equation A results in
the equations:
IP2 =
IP3 =
2 × PA − PIM2
2 −1
(C)
You can measure higher-order intercept points, too.
Example Measurements
To get a feel for this process, it’s useful to consider some
actual measured values.
The first example is a Rohde & Schwarz model EK085
receiver with digital preselection. For measuring second-order
IMD, signals at 6.00 and 8.01 MHz, at –20 dBm each, were
applied at the input of the attenuator. The difference in attenuator settings between the reference measurement and the level
needed to produce the desired IMD product signal level was
found to be 125 dB. The calculation of the second-order IP is
then:
IP2 =
For IP3, we set the signal generators for 0 dBm at the attenuator input, using frequencies of 14.00 and 14.01 MHz. The
difference in attenuator settings between the reference and IMD
measurements was 80 dB, so:
=
3 (0 dBm) − (0 dBm − 80 dB)
3 −1
−30 dBm + 10 dBm + 80 dB
2
=
−20 dBm + 80 dB
2
= +30 dBm
Synthesizer Requirements
To be able to make use of high third-order intercept points at
these close-in spacings requires a low-noise LO synthesizer. You
can estimate the required noise performance of the synthesizer
for a given IP3 value. First, calculate the value of receiver input
power that would cause the IMD product to just come out of the
noise floor, by solving equation A for PA, then take the difference
between the calculated value of PA and the noise floor to find the
dynamic range. Doing so gives the equation:
ID3 =
2
(IP3 + Pmin )
3
(D)
Pmin = –174 dBm + 10 log (BW) + NF
= –174 dBm + 10 log (2400) + 8
= –132 dBm
The synthesizer noise should not exceed the noise floor when
an input signal is present that just causes an IMD product signal
at the noise floor level. This will be accomplished if the synthesizer noise is less than:
ID + 10 log (BW) = 114.7 dB + 10 log (2400)
0 dBm + 80 dB
= +40 dBm
2
= 148.5 dBc/Hz
We also measured the IP3 of a Yaesu FT-1000D at the same
frequencies, using attenuator-input levels of –10 dBm. A difference in attenuator readings of 80 dB resulted in the calculation:
tered in vacuum-tube-based vintage radio
gear. In one (Fig 10.19A), the RF and LO
signals are applied to the control grid of a
pentode amplifier; in the second, the RF and
LO signals are applied to different grids of a
tube specifically designed to act as a mixer.
RF, LO, and IF (RF and LO sum and difference frequencies) energy is present at the
output of both. In both circuits, a double-
=
Where ID3 is the third-order IMD dynamic range in dB and
Pmin is the noise floor in dBm. Knowing the receiver bandwidth,
BW (2400 Hz in this case) and noise figure, NF (8 dB) allows us
to calculate the noise floor, Pmin:
2 (−20 dBm) − (−20 dBm − 125 dB)
2 −1
= −40 dBm + 20 dBm + 125 dB = +105 dB
IP3 =
3 (−10 dBm) − (−10 dBm − 80 dB)
3 −1
(B)
3 × PA − PIM3
3 −1
IP3 =
in the passband of the receiver. Such synthesizers hardly exist.
— Dr Ulrich L. Rohde, N1UL
tuned transformer selects the desired output
frequency and provides some reduction of
unwanted components. The pentagrid mixer
circuit (Fig 10.19A) provides better isolation
between its RF and LO inputs because of its
separate RF and LO-injection grids.
Fig 10.20 shows a simple JFET front end
that consists of an AGCed amplifier (2N4393)
and 2N3819 or 2N5454 mixer with source LO
injection. As in the tube circuits of Fig 10.19,
RF, LO, and IF spectral components are all
present in the output of this simple circuit.
Mixers based on FET cascode amplifiers (Fig 10.21) can provide excellent performance, and were common (in dual-gate
MOSFET form) across at least two generations of Amateur Radio gear. As with the
circuits presented in Figs 10.19 and 10.20,
Mixers, Modulators and Demodulators 10.17
RF, LO, and IF spectral components are present in the output of these mixers, and filtering
is necessary to select the desired component
and reduce unwanted ones.
10.5.2 Switching Mixers
Most modern radio mixers act more like
fast analog switches than analog multipliers.
In using a mixer as a fast switching device,
we apply a square wave to its LO input with a
square wave rather than a sine wave, and feed
sine waves, audio, or other complex signals to
the mixer’s RF input. The RF port serves as
the mixer’s “linear” input, and therefore must
preferably exhibit low intermodulation and
harmonic distortion. Feeding a ±1-V square
wave into the LO input alternately multiplies
the linear input by +1 or –1. Multiplying the
RF-port signal by +1 just transfers it to the
output with no change. Multiplying the RFport signal by –1 does the same thing, except
that the signal inverts (flips 180° in phase).
The LO port need not exhibit low intermodulation and harmonic distortion; all it has to
do is preserve the fast rise and fall times of
the switching signal.
Fig 10.19 — Vacuum-tube amplifiers as frequency mixers. At A, RF and LO signals are
applied to the control grid of a 6EJ7 pentode; at B, a RF and LO signals are applied to
grids 1 and 3, respectively, of a 6BA7 pentagrid mixer. The 6BA7 and its relatives and
similars (6A8, 6K8, 6SA7, 6SB7, and 6BE6, among others), designed specifically for
frequency-conversion use, can be configured to generate their own LO signal; such a
self-oscillating tube mixer is commonly called a converter.
REVERSING-SWITCH MIXERS
We can multiply a signal by a square wave
without using an analog multiplier at all. All
we need is a pair of balun transformers and
four diodes (Fig 10.22A).
With no LO energy applied to the circuit,
none of its diodes conduct. RF-port energy (1)
Fig 10.20 — This simple JFET receiver front end for 7 MHz features an AGCed RF amplifier (Q1, a 2N4393) and gain-controlledamplifier mixer (Q2, a 2N3819 or 2N5485). LO voltage applied to the source of Q2 varies the stage’s gain to produce sum and
difference outputs. T2 selects the product (4 MHz). Any JFET with a pinchoff voltage of –1 V is suitable for Q1; the point labeled AGC
can be grounded to operate the RF stage at full gain. The inductors consist of no. 22 enameled wire wound on plastic bobbins used
to hold the lower thread in “New Home” brand sewing machines; if toroidal equivalents are substituted, the turns ratios in T1 and T2
should be preserved. This circuit was described in “Build the No-Excuses QRP Transceiver,” December 2000 QST.
10.18 Chapter 10
Fig 10.21 — Cascode FET amplifiers used as mixers, as presented by Hayward, Campbell and Larkin in Experimental Methods in RF
Design. RF is applied to the gate of one device, and LO and a small positive bias are applied to the gate of the other; commercial
applications of the MOSFET version commonly include positive bias on the RF-port gate. A dual-gate MOSFET is shown at A;
suitable devices include the 3N187, 3N211, 40673 and BF998. (Internally, a dual-gate MOSFET consists of two single-gate MOSFETs in
cascode.) Applying AGC rather than this circuit’s combination of LO and source-derived bias to gate 2 of the MOSFET would turn this
arrangement into one of the AGCed IF amplifiers used in commercial ham gear for over 40 years. Two JFETs in cascode are shown at
B; suitable devices include the 2N4416, 2N3819, 2N5400-series parts, J310/U310 and MPF102, among others. R1 loads the circuit to
eliminate oscillation at the RF-port resonance; it may be increased in value or eliminated altogether if the circuit is stable without it.
Ratios accompanying the transformers at B convey the number of turns. Per EMRFD, the JFET circuit shown, biased for 3.4 mA at
12 V, operates with a conversion gain of 8 dB; a noise figure of 10 dB; and a third-order intercept point of +5 dBm.
can’t make it to the LO port because there’s
no direct connection between the secondaries
of T1 and T2, and (2) doesn’t produce IF output because T2’s secondary balance results
in energy cancellation at its center tap, and
because no complete IF-energy circuit exists
through T2’s secondary with both of its ends
disconnected from ground.
Applying a square wave to the LO port
biases the diodes so that, 50% of the time,
D1 and D2 are on and D3 and D4 are reversebiased off. This unbalances T2’s secondary by leaving its upper wire floating and
connecting its lower wire to ground through
T1’s secondary and center tap. With T2’s secondary unbalanced, RF-port energy emerges
from the IF port.
The other 50% of the time, D3 and D4 are
on and D1 and D2 are reverse-biased off.
This unbalances T2’s secondary by leaving
its lower wire floating, and connects its upper
wire to ground through T1’s secondary and
center tap. With T2’s secondary unbalanced,
RF-port energy again emerges from the IF
port — shifted 180° relative to the first case
because T2’s active secondary wires are now,
in effect, transposed relative to its primary.
A reversing switch mixer’s output spec-
trum is the same as the output spectrum of a
multiplier fed with a square wave. This can
be analyzed by thinking of the square wave in
terms of its Fourier series equivalent, which
consists of the sum of sine waves at the square
wave frequency and all of its odd harmonics.
The amplitude of the equivalent series’ fundamental sine wave is 4/π times (2.1 dB greater
than) the amplitude of the square wave. The
amplitude of each harmonic is inversely proportional to its harmonic number, so the third
harmonic is only 1⁄3 as strong as the fundamental (9.5 dB below the fundamental), the 5th
harmonic is only 1⁄5 as strong (14 dB below
the fundamental) and so on. The input signal
mixes with each harmonic separately from
the others, as if each harmonic were driving
its own separate mixer, just as we illustrated
with two sine waves in Fig 10.4. Normally,
the harmonic outputs are so widely removed
from the desired output frequency that they
are easily filtered out, so a reversing-switch
mixer is just as good as a sine-wave-driven
analog multiplier for most practical purposes,
and usually better — for radio purposes — in
terms of dynamic range and noise.
An additional difference between multiplier and switching mixers is that the signal
flow in a switching mixer is reversible (that
is, bilateral). It really only has one dedicated
input (the LO input). The other terminals
can be thought of as I/O (input/output) ports,
since either one can be the input as long as
the other is the output.
CONVERSION LOSS IN SWITCHING
MIXERS
Fig 10.22B shows a perfect multiplier
mixer. That is, the output is the product of
the input signal and the LO. The LO is a
perfect square wave. Its peak amplitude is
±1.0 V and its frequency is 8 MHz. Fig 10.22C
shows the output waveform (the product of
two inputs) for an input signal whose value is
0 dBm and whose frequency is 2 MHz. Notice
that for each transition of the square-wave
LO, the sine-wave output waveform polarity
reverses. There are 16 transitions during the
interval shown, at each zero-crossing point of
the output waveform. Fig 10.22D shows the
mixer output spectrum. The principle components are at 6 MHz and 10 MHz, which are
the sum and difference of the signal and LO
frequencies. The amplitude of each of these
is –3.9 dBm. Numerous other pairs of output
frequencies occur that are also spaced 4 MHz
Mixers, Modulators and Demodulators 10.19
apart and centered at 24 MHz, 40 MHz and
56 MHz and higher odd harmonics of 8 MHz.
The ones shown are at –13.5 dBm, –17.9 dBm
and –20.9 dBm. Because the mixer is lossless,
the sum of all of the outputs must be exactly
equal to the value of the input signal. As explained previously, this output spectrum can
also be understood in terms of each of the
odd-harmonic components of the squarewave LO operating independently.
If the mixer switched losslessly, such as
in Fig 10.22A, with diodes that are perfect
switches, the results would be mathematically identical to the above example. The
diodes would commutate the input signal
exactly as shown in Fig 10.22C.
Now consider the perfect multiplier mixer
of Fig 10.22B with an LO that is a perfect
sine wave with a peak amplitude of ±1.0 V. In
this case the dashed lines of Fig 10.22D show
that only two output frequencies are present,
at 6 MHz and 10 MHz (see also Fig 10.2).
Each component now has a –6 dBm level.
The product of the 0 dBm sine-wave input at
one frequency and the ±1.0V sine-wave LO
at another frequency (see equation 6 in this
chapter) is the –3 dBm total output.
These examples illustrate the difference
­between the square-wave LO and the sinewave LO, for a perfect multiplier. For the
same peak value of both LO waves, the
square-wave LO delivers 2.1 dB more output
at 6 MHz and 10 MHz than the sinewave LO. An actual diode mixer such as
Fig 10.22A behaves more like a switching
mixer. Its sine-wave LO waveform is considerably flattened by interaction between the
diodes and the LO generator, so that it looks
somewhat like a square wave. The diodes
have nonlinearities, junction voltages, capa­
citances, resistances and imperfect parameter
matching. (See the RF Techniques chapter.)
Also, “re-mixing” of a diode mixer’s output
with the LO and the input is a complicated
possibility. The practical end result is that
diode double-balanced mixers have a conversion loss, from input to each of the two major
output frequencies, in the neighborhood of
5 to 6 dB. (Conversion loss is discussed in a
later section.)
10.5.3 The Diode DoubleBalanced Mixer: A Basic
Building Block
Fig 10.22 — Part A shows a general-purpose diode reversing-switch mixer. This mixer
uses a square-wave LO and a sine-wave input signal. The text describes its action.
Part B is an ideal multiplier mixer. The square-wave LO and a sine-wave input signal
produce the output waveform shown in part C. The solid lines of part D show the
output spectrum with the square-wave LO. The dashed lines show the output spectrum
with a sine-wave LO.
10.20 Chapter 10
The diode double-balanced mixer (DBM)
is standard in many commercial, military and
amateur applications because of its excellent balance and high dynamic range. DBMs
can serve as mixers (including image-reject
types), modulators (including single- and
double-sideband, phase, biphase, and quadrature-phase types) and demodulators, limiters,
attenuators, switches, phase detectors and
frequency doublers. In some of these applica-
tions, they work in conjunction with power
dividers, combiners and hybrids.
THE BASIC DBM CIRCUIT
We have already seen the basic diode DBM
circuit (Fig 10.22A). In its simplest form, a
DBM contains two or more unbalanced-tobalanced transformers and a Schottky-diode
ring consisting of 4 × n diodes, where n is the
number of diodes in each leg of the ring. Each
leg commonly consists of up to four diodes.
As we’ve seen, the degree to which a mixer
is balanced depends on whether either, neither or both of its input signals (RF and LO)
emerge from the IF port along with mixing
products. An unbalanced mixer suppresses
neither its RF nor its LO; both are present
at its IF port. A single-balanced mixer suppresses its RF or LO, but not both. A doublebalanced mixer suppresses its RF and LO
inputs. Diode and transformer uniformity in
the Fig 10.22 circuit results in equal LO potentials at the center taps of T1 and T2. The
LO potential at T1’s secondary center tap is
zero (ground); therefore, the LO potential at
the IF port is zero.
Balance in T2’s secondary likewise results
in an RF null at the IF port. The RF potential
between the IF port and ground is therefore
zero — except when the DBM’s switching
diodes operate, of course!
The Fig 10.22 circuit normally also affords
high RF-IF isolation because its balanced
diode switching precludes direct connections
between T1 and T2. A diode DBM can be
used as a current-controlled switch or attenuator by applying dc to its IF port, albeit with
some distortion. This causes opposing diodes
(D2 and D4, for instance) to conduct to a degree that depends on the current magnitude,
connecting T1 to T2.
TRIPLE-BALANCED MIXERS
The triple-balanced mixer shown in Fig
10.23 (sometimes called a “double doublebalanced mixer”) is an extension of the single-diode-ring mixer. The diode rings are fed
by power-splitting baluns at the RF and LO
ports. An additional balun is added at the IF
output. The circuit’s primary advantage is
that the IF output signal is balanced and isolated from the RF and LO ports over a large
bandwidth — commercial mixer IF ranges
of 0.5 to 10 GHz are typical.
It has higher signal-handling capability
and dynamic range (a 1-dB compression
point within 3 to 4 dB below LO signal levels) and lower intermodulation levels (by 10
dB or more) than a single-ring mixer. The
triple-balanced mixer is used when a very
wide IF range is required.
Adding the balancing transformer in the IF
output path increases IF-to-LO and IF-to-RF
isolation. This makes the conversion process
much less sensitive to IF impedance mis-
Testing Mixer Performance
In order to make proper tests on mixers using signal generators, a hybrid coupler
with at least 40 dB of isolation between the two input ports and an attenuator are
required. The test set-up provided by DeMaw in QST, Jan 1981, shown in Fig 10.B1
is ideal for this. Two signal generators operating near 14 MHz are combined in the
hybrid coupler, then isolated from the mixer under test (MUT) by a variable attenuator. The LO is supplied by a VFO covering 5.0-5.5 MHz and applied to the MUT
through another variable attenuator. The output is isolated with another attenuator,
amplified and applied to a spectrum analyzer for analysis.
Attenuation should be sufficient to provide isolation (minimum of 6 to 10 dB required) and to result in signal levels to the mixer under test (MUT) appropriate for the
required testing and as suitable for the particular mixer device.
The 2N5109 amplifier shown may not be sufficient for extremely high intercept
point tests as this stage may no longer be transparent (operate linearly) at high signal
levels. For stability tests, it is recommended to have a reactive network at the output
of the mixer for the sole purpose of checking whether the mixer can become unstable.
The two 14 MHz oscillators must have extremely low harmonic content and very
low noise sidebands. A convenient oscillator circuit is provided in Fig 10.B2, based
on a 1975 Electronic Design article. — Dr Ulrich L. Rohde, N1UL
DeMaw, W1FB, and Collins, ADØW, “Modern Receiver Mixers for High Dynamic Range,”
QST, Jan 1981, p 19.
U. Rohde, “Crystal Oscillator Provides Low Noise,” Electronic Design, Oct 11, 1975.
Fig 10.B1 — The equipment setup for measuring mixer performance at HF.
Fig 10.B2 — A low-noise VXO circuit for driving the LO port of the mixer under
test in Figure 10.B1.
Mixers, Modulators and Demodulators 10.21
number of turns used with a given core. Increasing a transformer’s core size and number
of turns improves its low-frequency response.
Cores may be stacked to meet low-frequency
performance specs.
Inexpensive mixers operating up to 2 GHz
most commonly use twisted trifilar (threewire) windings made of a wire size between
#36 and #32. The number of twists per unit
length of wire determines a winding’s characteristic impedance. Twisted wires are
analogous to transmission lines. The transmission-line effect predominates at the higher end of a transformer’s frequency range.
Fig 10.23 — The triple-balanced mixer uses a pair of diode rings and adds an additional
balancing transformer to the IF port.
matches. Since the IF port is isolated from the
RF and LO ports, the three frequency ranges
(RF, LO and IF) can overlap. A disadvantage
of IF transformer coupling is that a dc (or
low-frequency) IF output is not available, so
the triple-balanced mixer cannot be used for
direct-conversion receivers.
DIODE DBM COMPONENTS
Commercially manufactured diode DBMs
generally consist of a supporting base, a
diode ring, two or more ferrite-core transformers commonly wound with two or three
twisted-pair wires, encapsulating material,
an enclosure.
Diodes
Hot-carrier (Schottky) diodes are the devices of choice for diode-DBM rings because
of their low ON resistance, although hambuilt DBMs for non-critical MF/HF use commonly use switching diodes like the 1N914
or 1N4148. The forward voltage drop, Vf,
across each diode in the ring determines the
mixer’s optimum local-oscillator drive level.
Depending on the forward voltage drop of
each of its diodes and the number of diodes
in each ring leg, a diode DBM will often be
specified by the optimum LO drive level in
dBm (typical values are 0, 3, 7, 10, 13, 17,
23 or 27). As a rule of thumb, the LO signal
must be 20 dB stronger than the RF and IF
signals for proper operation. This ensures
that the LO signal, rather than the RF or IF
signals, switches the mixer’s diodes on and
off — a critical factor in minimizing IMD
and maximizing dynamic range.
10.22 Chapter 10
Transformers
From the DBM schematic shown in Fig
10.22, it’s clear that the LO and RF transformers are unbalanced on the input side and
balanced on the diode side. The diode ends
of the balanced ports are 180° out of phase
throughout the frequency range of interest.
This property causes signal cancellations
that result in higher port-to-port isolation.
Fig 10.24A plots LO-RF and LO-IF isolation versus frequency for Synergy Microwave’s CLP-403 DBM, which is specified for
+7 dBm LO drive level. Isolations on the
order of 70 dB occur at the lower end of the
band as a direct result of the balance among
the four diode-ring legs and the RF phasing
of the balanced ports.
As we learned in our discussion of generic switching mixers, transformer efficiency
plays an important role in determining a mixer’s conversion loss and drive-level requirement. Core loss, copper loss and impedance
mismatch all contribute to transformer losses.
Ferrite in toroidal, bead, balun (multi-hole)
or rod form can serve as DBM transformer
cores. Radio amateurs commonly use FairRite Mix 43 ferrite (µ = 950).
RF transformers combine lumped and
distributed capacitance and inductance. The
interwinding capacitance and characteristic
impedance of a transformer’s twisted wires
sets the transformer’s high-frequency response. The core’s µ and size, and the number
of winding turns, determine the transformer’s
lower frequency limit. Covering a specific
frequency range requires a compromise in the
DIODE DBMS IN PRACTICE
Important DBM specifications include
conversion loss and amplitude flatness across
the required IF bandwidth; variation of conversion loss with input frequency; variation
of conversion loss with LO drive, 1-dB compression point; LO-RF, LO-IF and RF-IF isolation; intermodulation products; noise figure
(usually within 1 dB of conversion loss); port
SWR; and dc offset, which is directly related
to isolation among the RF, LO and IF ports.
Most of these parameters also apply to other
mixer types.
Conversion Loss
Fig 10.24B shows conversion loss versus
intermediate frequency in a typical DBM.
The curves show conversion loss for two
fixed RF-port signals, one at 100 kHz and
Fig 10.24 — The port-to-port isolation of
a diode DBM depends on how well its diodes match and how well its transformers
are balanced. (A) shows LO-IF and LO-RF
isolation versus frequency and (B) shows
conversion loss for a typical diode DBM,
the Synergy Microwave CLP-403 mixer. In
(B), LO driver level is +7 dBm.
Fig 10.25— Simulated diode-DBM output spectrum with four LO harmonics evaluated.
Note that the desired output products (the highest two products, RF – LO and RF + LO)
emerge at a level 5 to 6 dB below the mixer’s RF input (–40 dBm). This indicates a mixer
conversion loss of 5 to 6 dB. (Serenade SV8.5 simulation.)
the another at 500 MHz, while varying the
LO frequency from 100 kHz to 500 MHz.
Fig 10.25 graphs a diode DBM’s simulated output spectrum. Note that the RF input
(900 MHz) is –40 dBm and the desired IF
output (51 MHz, the frequency difference
between the RF and LO signals) is –46 dBm,
implying a conversion loss of 6 dB. Very
nearly the same value (5 dB) applies to the
sum of both signals (RF + LO). We minimize
a diode DBM’s conversion loss, noise figure
and intermodulation by keeping its LO drive
high enough to switch its diodes on fully and
rapidly. Increasing a mixer’s LO level beyond
that sufficient to turn its switching devices
all the way on merely makes them dissipate
more LO power without further improving
performance.
Insufficient LO drive results in increased
noise figure and conversion loss. IMD also
increases because RF-port signals have a
greater chance to control the mixer diodes
when the LO level is too low.
APPLYING DIODE DBMS
At first glance, applying a diode DBM
is easy: We feed the signal(s) we want to
frequency-shift (at or below the maximum
level called for in the mixer’s specifications,
such as –10 dBm for the Mini-Circuits SBL1 and TUF-3, and Synergy Microwave S-1,
popular 7 dBm LO power parts) to the DBM’s
RF port, feed the frequency-shifting signal (at
the proper level) to the LO port, and extract
the sum and difference products from the
mixer’s IF port.
There’s more to it than that, however, because diode DBMs (along with most other
modern mixer types) are termination-sensitive. That is, their ports — particularly their
IF (output) ports — must be resistively terminated with the proper impedance (commonly
50 Ω, resistive). A wideband, resistive output
termination is particularly critical if a mixer
is to achieve its maximum dynamic range in
receiving applications. Such a load can be
achieved by:
• Terminating the mixer in a 50-Ω resistor
or attenuator pad (a technique usually avoided
in receiving applications because it directly
degrades system noise figure);
• Terminating the mixer with a low-noise,
high-dynamic-range post-mixer amplifier designed to exhibit a wideband resistive input
impedance; or
• Terminating the mixer in a diplexer, a
frequency-sensitive signal splitter that appears as a two-terminal resistive load at its
input while resistively dissipating unwanted
outputs and passing desired outputs through
to subsequent circuitry.
Termination-insensitive mixers are available, but this label can be misleading. Some
termination-insensitive mixers are nothing
more than a termination-sensitive mixer
packaged with an integral post-mixer amplifier. True termination-insensitive mixers
are less common and considerably more
elaborate. Amateur builders will more likely
use one of the many excellent terminationsensitive mixers available in connection with
a diplexer, post-mixer amplifier or both.
Fig 10.26 shows one diplexer implementation. In this approach, L1 and C1 form a
series-tuned circuit, resonant at the desired
IF, that presents low impedance between the
diplexer’s input and output terminals at the
IF. The high-impedance parallel-tuned circuit
formed by L2 and C2 also resonates at the
desired IF, keeping desired energy out of the
diplexer’s 50-Ω load resistor, R1.
The preceding example is called a bandpass diplexer. Fig 10.27 shows another type:
a high-pass/low-pass diplexer in which each
inductor and capacitor has a reactance of
70.7 Ω at the 3-dB cutoff frequency. It can
be used after a “difference” mixer (a mixer
in which the IF is the difference between the
signal frequency and LO) if the desired IF
and its image frequency are far enough apart
so that the image power is “dumped” into the
network’s 51-Ω resistor. (For a “summing”
mixer — a mixer in which the IF is the sum
of the desired signal and LO — interchange
the 50-Ω idler load resistor and the diplexer’s “50-Ω Amplifier” connection.) Richard
Weinreich, KØUVU, and R. W. Carroll described this circuit in November 1968 QST as
one of several absorptive TVI filters.
Fig 10.28 shows a BJT post-mixer amplifier design made popular by Wes Hayward,
W7ZOI, and John Lawson, K5IRK. RF
feedback (via the 1-kΩ resistor) and emitter
degeneration (the ac-coupled 5.6-Ω emitter
resistor) work together to keep the stage’s
input impedance near 50 Ω and uniformly resistive across a wide bandwidth. Performance
comparable to the Fig 10.28 circuit can be
obtained at MF and HF by using paralleled
2N3904s as shown in Fig 10.29.
AMPLITUDE MODULATION
WITH A DBM
We can generate DSB, suppressed-carrier
AM with a DBM by feeding the carrier to
its RF port and the modulating signal to the
IF port. This is a classical balanced modulator, and the result — sidebands at radio frequencies corresponding to the carrier signal
plus audio and the RF signal minus audio
— emerges from the DBM’s LO port. If we
also want to transmit some carrier along with
the sidebands, we can dc-bias the IF port (with
a current of 10 to 20 mA) to upset the mixer’s
balance and keep its diodes from turning all
the way off. (This technique is sometimes
used for generating CW with a balanced
modulator otherwise intended to generate
DSB as part of an SSB-generation process.)
Fig 10.30 shows a more elegant approach
to generating full-carrier AM with a DBM.
As we saw earlier when considering the
many faces of AM, two DBMs, used in con-
Mixers, Modulators and Demodulators 10.23
Fig 10.26 — A diplexer resistively terminates energy at unwanted
frequencies while passing energy at desired frequencies. This
band-pass diplexer (A) uses a series-tuned circuit as a selective
pass element, while a high-C parallel-tuned circuit keeps the
network’s terminating resistor R1 from dissipating desiredfrequency energy. Computer simulation of the diplexer’s response
with ARRL Radio Designer 1.0 characterizes the diplexer’s
insertion loss and good input match from 8.8 to 9.2 MHz (B) and
from 1 to 100 MHz (C); and the real and imaginary components
of the diplexer’s input impedance from 8.8 to 9.2 MHz with
a 50-Ω load at the diplexer’s output terminal (D). The high-C,
low-L nature of the L2-C2 circuit requires that C2 be minimally
inductive; a 10,000-pF chip capacitor is recommended. This
diplexer was described by Ulrich L. Rohde and T. T. N. Bucher in
Communications Receivers: Principles and Design.
10.24 Chapter 10
junction with carrier and audio phasing, can
be used to generate SSB, suppressed-carrier
AM. Likewise, two DBMs can be used with
RF and LO phasing as an image-reject mixer.
PHASE DETECTION WITH A DBM
As we saw in our exploration of quadrature
detection, applying two signals of equal frequency to a DBM’s LO and RF ports produces
an IF-port dc output proportional to the cosine
of the signals’ phase difference (Fig 10.31).
This assumes that the DBM has a dc-coupled
IF port, of course. If it doesn’t — and some
DBMs don’t — phase-detector operation is
out. Any dc output offset introduces error
into this process, so critical phase-detection
applications use low-offset DBMs optimized
for this service.
BIPHASE-SHIFT KEYING
MODULATION WITH A DBM
Back in our discussion of square-wave
mixing, we saw how multiplying a switching mixer’s linear input with a square wave
causes a 180° phase shift during the negative
part of the square wave’s cycle. As Fig 10.32
shows, we can use this effect to produce biphase-shift keying (BPSK), a digital system
that conveys data by means of carrier phase
reversals. A related system, quadrature
phase-shift keying (QPSK) uses two DBMs
and phasing to convey data by phase-shifting
a carrier in 90° increments.
10.5.4 Active Mixers —
Transistors as Switching
Elements
We’ve covered diode DBMs in depth because their home-buildability, high performance and suitability for direct connection
into 50-W systems makes them attractive
to Amateur Radio builders. The abundant
availability of high-quality manufactured diode mixers at reasonable prices makes them
Fig 10.27 — All of the inductors and capacitors in this
high-pass/low-pass diplexer (A) exhibit a reactance
of 70.7 Ω at its tuned circuits’ 3-dB cutoff frequency
(the geometric mean of the IF and IF image). B and C
show ARRL Radio Designer simulations of this circuit
configured for use in a receiver that converts 7 MHz to
3.984 MHz using a 10.984-MHz LO. The IF image is at
17.984 MHz, giving a 3-dB cutoff frequency of
8.465 MHz. The inductor values used in the simulation
were therefore 1.33 µH (Q = 200 at 25.2 MHz); the
capacitors, 265 pF (Q = 1000). This drawing shows idler
load and “50-Ω Amplifier” connections suitable for a
receiver in which the IF image falls at a frequency above
the desired IF. For applications in which the IF image
falls below the desired IF, interchange the 50-Ω idler load
resistor and the diplexer’s “50-Ω Amplifier” connection so
the idler load terminates the diplexer low-pass filter and
the 50-Ω amplifier terminates the high-pass filter.
Mixers, Modulators and Demodulators 10.25
Fig 10.28 — The post-mixer amplifier
from Hayward and Lawson’s Progressive
Communications Receiver (November
1981 QST). This amplifier’s gain, including
the 6-dB loss of the attenuator pad, is
about 16 dB; its noise figure, 4 to 5 dB;
its output intercept, 30 dBm. The 6-dB
attenuator is essential if a crystal filter
follows the amplifier; the pad isolates
the amplifier from the filter’s highly
reactive input impedance. This circuit’s
input match to 50 Ω below 4 MHz can be
improved by replacing 0.01-µF capacitors
C1, C2 and C3 with low-inductance 0.1-µF
units (chip capacitors are preferable).
Q1 — TO-39 CATV-type bipolar transistor,
fT = 1 GHz or greater. 2N3866, 2N5109,
2SC1252, 2SC1365 or MRF586 suitable.
Use a small heat sink on this transistor.
T1 — Broadband ferrite transformer, ≈42
µH per winding: 10 bifilar turns of #28
enameled wire on an FT 37-43 core.
Fig 10.29 — At MF and HF, paralleled
2N3904 BJTs can provide performance
comparable to that of the Fig 10.28 circuit
with sufficient attention paid to device
standing current, here set at ≈30 mΑ for
the pair. The value of decoupling resistor
R1 is critical in that small changes in its
value cause a relatively large change in the
2N3904s’ bias point. This circuit is part of
the EZ-90 Receiver, described by Hayward,
Campbell and Larkin in Experimental
Methods in RF Design.
Fig 10.30 — Generating full-carrier AM
with a diode DBM. A practical modulator
using this technique is described in
Experimental Methods in RF Design.
Fig 10.31 — A phase detector’s dc output
is the cosine of the phase difference
between its input and reference signals.
10.26 Chapter 10
Fig 10.32 — Mixing a carrier with a square
wave generates biphase-shift keying
(BPSK), in which the carrier phase is
shifted 180° for data transmission. In
practice, as in this drawing, the carrier
and data signals are phase-coherent so
the mixer switches only at carrier zero
crossings.
excellent candidates for home construction
projects. Although diode DBMs are common in telecommunications as a whole,
their conversion loss and relatively high LO
power requirement have usually driven the
manufacturers of high-performance MF/HF
Amateur Radio receivers and transceivers to
other solutions. Those solutions have generally involved single- or double-balanced FET
mixers — MOSFETs in the late 1970s and
early 1980s, JFETs from the early 1980s to
date. A comprehensive paper that explores
the differences between various forms of
active mixers, “Performance Capabilities of
Active Mixers,” by Ulrich Rohde, N1UL,
is included on the CD-ROM accompanying
this Handbook.
Many of the JFET designs are variations
of a single-balanced mixer circuit introduced
to QST readers in 1970! Fig 10.33 shows the
circuit as it was presented by William Sabin in
“The Solid-State Receiver,” QST, July 1970.
Two 2N4416 JFETs operate in a commonsource configuration, with push-pull RF input
and parallel LO drive. Fig 10.34 shows a
similar circuit as implemented in the ICOM
IC-765 transceiver. In this version, the JFETs
(2SK125s) operate in common-gate, with the
LO applied across a 220-W resistor between
the gates and ground.
Current state of the art for active mixers
in the HF through GHz range replaces discrete device designs with integrated designs
such as the Analog Devices AD8342 (www.
analog.com). Using an IC greatly improves
matching of the active devices, improving
circuit balance. The AD8342 has a conversion
gain of 3.7 dB, a noise figure of 12.2 dB, and
an input IP3 of 22.7 dBm. The device operates
with a single-voltage power supply and is
well-suited to interface with digital hardware,
such as for SDR applications. Reference circuits for applications at HF and VHF/UHF
are provided in the device’s datasheet.
Fig 10.33 — Two 2N4416 JFETs provide high dynamic range in this mixer circuit from
Sabin, QST, July 1970. L1, C1 and C2 form the input tuned circuit; L2, C3 and C4 tune
the mixer output to the IF. The trifilar input and output transformers are broadband
transmission-line types.
Fig 10.34 — The ICOM IC-765’s single-balanced 2SK125 mixer achieves a high
dynamic range (per QST Product Review, an IP3 of 10.5 dBm at 14 MHz with preamp
off). The first receive mixer in many commercial Amateur Radio transceiver designs
of the 1980s and 1990s used a pair of 2SK125s or similar JFETs in much this way.
Mixers, Modulators and Demodulators 10.27
10.5.5 The Tayloe Mixer
[The following description of the Tayloe
Product Detector (a.k.a. — the Tayloe Mixer)
is adapted from the July 2002 QEX article,
“Software-Defined Radio For the Masses, Part
1” by Gerald Youngblood AC5OG. — Ed.]
The beauty of the Tayloe detector (see references by Tayloe) is found in both its design
elegance and its exceptional performance. In
its simplest form, you can build a complete
quadrature downconverter with only three or
four ICs (less the local oscillator) at a cost
of less than $10.
Fig 10.35 illustrates a single-balanced version of the Tayloe detector. It can be visualized as a four-position rotary switch revolving
at a rate equal to the carrier frequency. The
50-Ω antenna impedance is connected to the
rotor and each of the four switch positions is
connected to a sampling capacitor. Since the
switch rotor is turning at exactly the RF carrier frequency, each capacitor will track the
carrier’s amplitude for exactly one-quarter of
the cycle and will then hold its value for the
remainder of the cycle. The rotating switch
will therefore sample the signal at 0°, 90°,
180° and 270°, respectively.
As shown in Fig 10.36, the 50-Ω impedance of the antenna and the sampling capacitors form an R-C low-pass filter during the
period when each respective switch is turned
on. Therefore, each sample represents the
integral or average voltage of the signal during its respective one-quarter cycle. When the
switch is off, each sampling capacitor will
hold its value until the next revolution. If the
RF carrier and the rotating frequency were
exactly in phase, the output of each capacitor
will be a dc level equal to the average value
of the sample.
If we differentially sum outputs of the 0°
and 180° sampling capacitors with an op
amp (see Fig 10.35), the output would be
a dc voltage equal to two times the value of
the individually sampled values when the
switch rotation frequency equals the carrier
frequency. Imagine, 6 dB of noise-free gain!
The same would be true for the 90° and 270°
capacitors as well. The 0°/180° summation
forms the I channel and the 90°/270° summation forms the Q channel of a quadrature
downconversion. (See the Modulation chapter for more information on I/Q modulation.)
As we shift the frequency of the carrier
away from the sampling frequency, the values
of the inverting phases will no longer be dc
levels. The output frequency will vary according to the “beat” or difference frequency
between the carrier and the switch-rotation
frequency to provide an accurate representation of all the signal components converted
to baseband.
Fig 10.37 is the schematic for a simple,
single-balanced Tayloe detector. It consists
10.28 Chapter 10
Fig 10.35 — Tayloe detector: The switch rotates at the carrier frequency so that
each capacitor samples the signal once each revolution. The 0° and 180° capacitors
differentially sum to provide the in-phase (I) signal and the 90° and 270° capacitors sum
to provide the quadrature (Q) signal.
G=
Rf
4R ant
(17)
For example, with a feedback resistance,
Rf, of 3.3 kΩ and antenna impedance, Rant, of
50 Ω, the resulting gain of the input stage is:
Fig 10.36 — Track-and-hold sampling
circuit: Each of the four sampling
capacitors in the Tayloe detector form
an RC track-and-hold circuit. When the
switch is on, the capacitor will charge to
the average value of the carrier during its
respective one-quarter cycle. During the
remaining three-quarters cycle, it will hold
its charge. The local-oscillator frequency
is equal to the carrier frequency so that
the output will be at baseband.
of a PI5V331, 1:4 FET demultiplexer (an
analog switch) that switches the signal to each
of the four sampling capacitors. The 74AC74
dual flip-flop is connected as a divide-by-four
Johnson counter to provide the two-phase
clock to the demultiplexer chip. The outputs
of the sampling capacitors are differentially
summed through the two LT1115 ultra-lownoise op amps to form the I and Q outputs,
respectively.
Note that the impedance of the antenna
forms the input resistance for the op-amp
gain as shown in Equation 17. This impedance may vary significantly with the actual
antenna. In a practical receiver, a buffer amplifier should be used to stabilize and control
the impedance presented to the mixer.
Since the duty cycle of each switch is 25%,
the effective resistance in the RC network is
the antenna impedance multiplied by four in
the op-amp gain formula:
G=
3300
= 16.5
4 × 50
The Tayloe detector may also be analyzed
as a digital commutating filter (see reference
by Kossor). This means that it operates as a
very-high-Q tracking filter, where Equation
18 determines the bandwidth and n is the
number of sampling capacitors, Rant is the
antenna impedance and Cs is the value of the
individual sampling capacitors. Equation 19
determines the Qdet of the filter, where fc is the
center frequency and BWdet is the bandwidth
of the filter.
BWdet =
Q det =
1
π n R ant Cs
fc
BWdet
(18)
(19)
By example, if we assume the sampling
capacitor to be 0.27 µF and the antenna impedance to be 50 Ω, then BW and Q at an
operating frequency of 14.001 MHz are computed as follows:
1
= 5895 Hz
BWdet =
π × 4 × 50 × (2.7 × 10 −7 )
Q det =
14.001 × 10 −6
= 2375 5895
(20)
The real payoff in the Tayloe detector is
its performance. It has been stated that the
Fig 10.37 — Single balanced Tayloe detector.
Fig 10.38 — Alias summing on Tayloe
detector output: Since the Tayloe detector
samples the signal, the sum frequency
(fc + fs) and its image (–fc – fs) are located
at the first alias frequency. The alias
signals sum with the baseband signals
to eliminate the mixing product loss
associated with traditional mixers. In a
typical mixer, the sum frequency energy
is lost through filtering thereby increasing
the noise figure of the device.
ideal commutating mixer has a minimum
conversion loss (which equates to noise figure — see the RF Techniques chapter) of
3.9 dB. Typical high-level diode mixers have
a conversion loss of 6-7 dB and noise figures
1 dB higher than the loss. The Tayloe detector
has less than 1 dB of conversion loss, remarkably. How can this be? The reason is that it
is not really a mixer but a sampling detector
in the form of a quadrature track-and-hold
circuit. This means that the design adheres to
discrete-time sampling theory, which, while
similar to mixing, has its own unique characteristics. Because a track and hold actually
holds the signal value between samples, the
signal output never goes to zero. (See the
DSP and Software Radio Design chapter
for more on sampling theory.)
This is where aliasing can actually be used
to our benefit. Since each switch and capaci-
tor in the Tayloe detector actually samples
the RF signal once each cycle, it will respond
to alias frequencies as well as those within
the Nyquist frequency range. In a traditional
direct-conversion receiver, the local-oscillator frequency is set to the carrier frequency
so that the difference frequency, or IF, is at
0 Hz and the sum frequency is at two times
the carrier frequency. We normally remove
the sum frequency through low-pass filter-
Mixers, Modulators and Demodulators 10.29
ing, resulting in conversion loss and a corresponding increase in noise figure. In the
Tayloe detector, the sum frequency resides
at the first alias frequency as shown in Fig
10.38. Remember that an alias is a real signal
and will appear in the output as if it were a
baseband signal. Therefore, the alias adds to
the baseband signal for a theoretically lossless detector. In real life, there is a slight loss,
usually less than 1 dB, due to the resistance of
the switch and aperture loss due to imperfect
switching times.
10.5.6 The NE602/SA602/
SA612: A Popular Gilbert
Cell Mixer
Fig 10.39 — The SA602/612’s equivalent circuit reveals its Gilbert-cell heritage.
Introduced as the Philips NE602 in the
mid-1980s, the NXP SA602/SA612 mixeroscillator IC has become greatly popular with
amateur experimenters for receive mixers,
transmit mixers and balanced modulators. The
SA602/612’s mixer is a Gilbert cell multiplier.
Fig 10.39 shows its equivalent circuit. A Gilbert cell consists of two differential transistor
pairs whose bias current is controlled by one
of the input signals. The other signal drives the
differential pairs’ bases, but only after being
“predistorted” in a diode circuit. (This circuit distorts the signal equally and oppositely
Fig 10.40 — The SA602/612’s inputs and outputs can be single- or double-ended (balanced). The balanced configurations minimize
second-order IMD and harmonic distortion, and unwanted envelope detection in direct-conversion service. CT tunes its inductor
to resonance; CB is a bypass or dc-blocking capacitor. The arrangements pictured don’t show all the possible input/output
configurations; for instance, a center-tapped broadband transformer can be used to achieve a balanced, untuned input or output.
10.30 Chapter 10
to the inherent distortion of the differential
pair.) The resulting output signal is an accurate multiplication of the input voltages.
Fig 10.41 — An NPN transistor at the output of an SA602/612 mixer provides power
gain and low-impedance drive for a 4.914-MHz crystal filter. A low-reactance coupling
capacitor can be added between the emitter and the circuitry it drives if dc blocking is
necessary. (Circuit from the Elecraft KX1 transceiver courtesy of Wayne Burdick, N6KR)
Fig 10.42 — SA602/612 product detector AGC from the Elecraft KX1 transceiver.
Designed by Wayne Burdick, N6KR, this circuit first appeared in the Wilderness Radio
SST transceiver with an LED used at D1 for simultaneous signal indication and
rectification. The selectivity provided by the crystal filter preceding the detector works
to mitigate the effects of increasing detector distortion with gain reduction, (Circuit
courtesy of Wayne Burdick, N6KR, and Bob Dyer, K6KK)
SA602/612A VARIANTS
The SA602/612 began life as the NE602/
SA602. SA-prefixed 602/612 parts are specified for use over a wider temperature range
than their NE-prefixed equivalents. Parts
without the A suffix have a slightly lower
IP3 specification than their A counterparts.
The pinout-identical NE612A and SA612A
cost less than their 602 equivalents as a result of wider tolerances. All variants of this
popular part should work satisfactorily in
most “NE602” experimenter projects. The
same mixer/oscillator topology, modified for
slightly higher dynamic range at the expense
of somewhat less mixer gain, is also available in the mixer/oscillator/FM IF chips NE/
SA605 (input IP3 typically –10 dBm) and
NE/SA615 (input IP3 typically –13 dBm).
SA602/612 USAGE NOTES
The SA602/612’s typical current drain is
2.4 mA; its supply voltage range is 4.5 to
8.0 V. Its inputs (RF) and outputs (IF) can
be single- or double-ended (balanced) according to design requirements (Fig 10.40).
The equivalent ac impedance of each input is
approximately 1.5 kΩ in parallel with 3 pF;
each output’s resistance is 1.5 kΩ. Fig 10.41
shows the use of an NPN transistor at the
SA602/612 output to obtain low-impedance
drive for a crystal filter; Fig 10.42 shows
how AGC can be applied to an SA602/612.
The SA602/612 mixer can typically handle
signals up to 500 MHz. At 45 MHz, its noise
figure is typically 5.0 dB; its typical con­
version gain, 18 dB. Note that in contrast
to the diode-based mixers described earlier,
which have conversion loss, most Gilbertcell mixers have conversion gain. Considering the SA602/612’s low current drain, its
input IP3 (measured at 45 MHz with 60-kHz
­spacing) is usefully good at –15 dBm. Factoring in the mixer’s conversion gain results
in an equivalent output IP3 of about 3 dBm.
The SA602/612’s on-board oscillator can
operate up to 200 MHz in LC and crystalcontrolled configurations (Fig 10.43 shows
three possibilities). Alternatively, energy
from an external LO can be applied to the
chip’s pin 6 via a dc blocking capacitor. At
least 200 mV P-P of external LO drive is
required for proper mixer operation.
The SA602/612 was intended to be used
as the second mixer in double-conversion
FM cellular radios, in which the first IF is
typically 45 MHz, and the second IF is typically 455 kHz. Such a receiver’s second
mixer can be relatively weak in terms of
dynamic range because of the adjacentsignal protection afforded by the high selectivity of the first-IF filter preceding it. When
Mixers, Modulators and Demodulators 10.31
Fig 10.43 — Three SA602/612 oscillator configurations:
crystal overtone (A); crystal fundamental (B); and
LC-controlled (C). T1 in C is a Mouser 10.7-MHz IF
transformer, green core, 7:1 turns ratio, part no.
42IF123-RC.
Fig 10.44 — A 7-MHz direct-conversion receiver based on the NE602/SA602/612. Equipped with a stage or two of audio filtering and
a means of muting during transmit periods, such a receiver is entirely sufficient for basic Amateur Radio communication at MF and
HF. L1 and L2 are 1.2 mH. This receiver is described in greater detail in Experimental Methods in RF Design.
10.32 Chapter 10
Fig 10.45 — Speech amplifier and 9-MHz balanced modulator using an
MC-1496P analog multiplier IC. The transformer consists of 10 bifilar turns
of no. 28 enameled wire on an FT37-43 ferrite toroidal core with a 3-turn
output link. The pin numbers shown are those of the DIP version of the
1496; builders using other package variants should consult manufacturer
data to obtain the correct pinout.
used as a first mixer, the SA602/612 can
provide a two-tone third-order dynamic
range between 80 and 90 dB, but this figure
is greatly diminished if a preamplifier is used
ahead of the SA602/612 to improve the system’s noise figure.
When the SA602/612 is used as a second
mixer, the sum of the gains preceding it should
not exceed about 10 dB. An SA602/612 can
serve as low-distortion (THD <1%) product
detector if overload is avoided through the use
of AGC and appropriate attenuation between
the ‘602/612 and the IF strip that drives it.
The SA602/612 is generally not a good
choice for VHF and higher-frequency mixers because of its input noise and diminishing IMD performance at high frequencies.
There are applications, however, where 6-dB
noise figure and 60- to 70-dB dynamic range
performance is adequate. If your target specifications exceed these numbers, you should
consider other mixers at VHF and up.
Fig 10.44 shows the schematic of a complete 7-MHz direct-conversion receiver
based on the SA602/612 and the widely used
LM386 AF power amplifier IC. Such simple
product-detector-based receivers sometimes
suffer from incidental envelope detection,
which causes audio from strong, full-carrierAM shortwave or mediumwave broadcast
stations to be audible regardless of where
the receiver LO is tuned. RF attenuation and/
or band-limiting the receiver input with a
double- or triple-tuned-circuit filter can usually reduce this effect to inaudibility.
10.5.7 An MC1496P Balanced
Modulator
Although it predates the SA602/612,
Freescale’s MC1496 Gilbert cell multiplier
remains a viable option for product detection
and balanced modulator service. Fig 10.45
shows an MC1496-based balanced modulator that is capable of a carrier suppression
greater than 50 dB. Per its description in
­Hayward, Campbell, and Larkin’s Experimental Methods in RF Design, its output with
audio drive should be kept to about –20 dBm
with this circuit. LO drive should be 200 to
500 mV P-P.
10.5.8 An Experimental
High-Performance Mixer
Examining the state of the art we find that
the best receive-mixer dynamic ranges are
achieved with quads of RF MOSFETs operating as passive switches, with no drain
voltage applied. The best of these techniques
involves following a receiver’s first mixer
with a diplexer and low-loss roofing crystal
filter, rather than terminating the mixer in a
strong wideband amplifier.
Colin Horrabin’s (G3SBI) experimenta-
Mixers, Modulators and Demodulators 10.33
Fig 10.46 — A shows the operation of
the H-mode switching mixer developed
by Colin Horrabin, G3SBI. B shows
the actual mixer circuit implemented
with the Linear Integrated Systems
SD5000 DMOS FET quad switch IC and
a 74AC74 flip-flop.
tions with variations of an original highperformance mixer circuit by Jacob
­Mahkinson, N6NWP, led to the development of a new mixer configuration, called
an H-mode mixer. This name comes from
the signal path through the circuit. (See
Fig 10.46A.) Horrabin is a professional scientist/engineer at the Science and Engineering Research Council’s Darebury Laboratory,
which has supported his investigative work
on the H-mode switched-FET mixer, and
consequently holds intellectual title to the
new mixer. This does not prevent readers
from taking the development further or using
the information presented here.
Inputs A and B are complementary squarewave inputs derived from the sine-wave local
oscillator at twice the required square-wave
frequency. If A is on, then FETs F1 and F3
are on and F2 and F4 are off. The direction
of the RF signal across T1 is given by the E
arrows. When B is on, FETs F2 and F4 are on
and F1 and F3 are off. The direction of the
RF signal across T1 reverses, as shown by the
F arrows.
This is still the action of a switching mixer,
but now the source terminal of each FET
switch is grounded, so that the RF signal
switched by the FET cannot modulate the gate
voltage. In this configuration the transformers
are important: T1 is a Mini-Circuits type T4-1
and T2 is a pair of these same transformers
with their primaries connected in parallel.
10.34 Chapter 10
10.6 References and Bibliography
The following mixer references include
books, journals and other publications as well
as websites. They are arranged by publication date.
BOOKS
E. E. Bucher, The Wireless Experimenter’s
Manual (New York: Wireless Press, Inc.,
1920).
F. E. Terman, Radio Engineers’ Handbook
(New York: McGraw-Hill, 1943).
R. W. Landee, D. C. Davis and A. P.
Albrecht, Electronic Designer’s
Handbook, 2nd ed, Section 22 (New
York: McGraw-Hill, 1957).
A. A. M. Saleh, Theory of Resistive Mixers
(Cambridge, MA: MIT Press, 1971).
L.G. Giacoletto, Electronics Designers’
Handbook, 2nd ed (New York: McGrawHill 1977).
P. Horowitz, W. Hill, The Art of Electronics,
2nd ed (New York: Cambridge University
Press, 1989).
RF/IF Designer’s Handbook (Brooklyn,
NY: Scientific Components, 1992).
RF Communications Handbook (Philips
Components-Signetics, 1992).
W. Hayward, Introduction to Radio
Frequency Design (Newington, CT:
ARRL, 1994).
L. E. Larson, RF and Microwave Circuit
Design for Wireless Communication,
(Artech House, 1996).
S. A. Maas, The RF and Microwave Circuit
Design Cookbook (Artech House, 1998).
U. L Rohde, D. P. Newkirk, RF/Microwave
Circuit Design for Wireless Applications
(New York: John Wiley & Sons, 2000).
(Large number of useful references in the
mixer chapter.)
U. L. Rohde, J. Whitaker, Communications
Receivers: DSP, Software Radios, and
Design, 3rd ed (New York: McGraw-Hill,
2001. (Large number of useful references
in the mixer chapter.)
H. I. Fujishiro, Y. Ogawa, T. Hamada and
T. Kimura, Electronic Letters, Vol 37,
Iss 7, Mar 2001, pp 435-436. See IET
Digital Library (www.ietdl.org).
S. A. Maas, Nonlinear Microwave and RF
Circuits, 2nd ed (Artech House, 2003).
S. A. Maas, Noise in Linear and Nonlinear
Circuits (Artech House, 2005).
G. D. Vendelin, A. M. Pavio and U. L.
Rohde, Microwave Circuit Design Using
Linear and Nonlinear Techniques (New
York: John Wiley & Sons, 2005). (Large
number of useful references in the mixer
chapter.)
Synergy Microwave Corporation Product
Handbook (Paterson, NJ: Synergy
Microwave Corporation, 2007).
JOURNALS AND OTHER
PUBLICATIONS
R. Weinreich and R. W. Carroll, “Absorptive
Filter for TV Harmonics,” QST, Nov
1968, pp 20-25.
B. Gilbert, “A Precise Four-Quadrant
Multiplier with Sub Nanosecond
Response,” IEEE Journ Solid-State
Circuits, Vol 3, 1968, pp 364-373.
W. Sabin, “The Solid-State Receiver,” QST,
Jul 1970, pp 35-43.
G. Begemann, A. Hecht, “The Conversion
Gain and Stability of MESFET Gate
Mixers” 9th EMC, 1979, pp 316-319.
P. Will, “Broadband Double Balanced
Mixer Having Improved Termination
Insensitivity Characteristics”, AdamsRussell Co, Waltham, MA, US Patent No
4,224,572, Sep 23, 1980.
C. Tsironis, R. Meierer, R. Stahlmann,
“Dual-Gate MESFET Mixers,” IEEE
Trans MTT, Vol 32, No 3, 04-1984.
B. Henderson and J. Cook, “Image-Reject
and Single-Sideband Mixers”, WatkinsJohnson Tech Note, Vol 12 No 3, 1985.
H. Statz, P. Newman, I. Smith, R. Pucel
and H. Haus, IEEE Trans Electron
Devices 34 (1987) 160.
J. Dillon, “The Neophyte Receiver,” QST,
Feb 1988, pp 14-18.
B. Henderson, “Mixers in Microwave
Systems (Part 1)”, Watkins-Johnson
Tech Note, Vol 17 No 1, 1990,
Watkins-Johnson.
B. Henderson, “Mixers in Microwave
Systems (Part 2)”, Watkins-Johnson
Tech Note, Vol 17 No 2, 1990,
Watkins-Johnson.
R. Zavrel, “Using the NE602,” Technical
Correspondence, QST, May 1990, pp
38-39.
D. Kazdan “What’s a Mixer?” QST, Aug
1992, pp 39-42.
L. Richey, “W1AW at the Flick of a
Switch,” QST, Feb 1993, pp 56-57.
J. Makhinson, “High Dynamic Range MF/
HF Receiver Front End,” QST, Feb 1993,
pp 23-28. Also see Feedback, QST, Jun
1993, p 73.
J. Vermusvuori, “A Synchronous Detector
for AM Transmisions,” QST,
pp 28-33. Uses SA602/612 mixers as
phase-locked and quasi-synchronous
product detectors.
P. Hawker, Ed., “Super-Linear HF Receiver
Front Ends,” Technical Topics, Radio
Communication, Sep 1993, pp 54-56.
S. Joshi, “Taking the Mystery Out of
Double-Balanced Mixers,” QST, Dec
1993, pp 32-36.
U. L. Rohde, “Key Components of Modern
Receiver Design,” Part 1, QST, May
1994, pp 29-32; Part 2, QST, Jun 1994,
pp 27-31; Part 3, QST, Jul 1994, 42-45.
U. L. Rohde, “Testing and Calculating
Intermodulation Distortion in Receivers,”
QEX, Jul 1994, pp 3-4.
B. Gilbert, “Demystifying the Mixer,” selfpublished monograph, 1994.
S. Raman, F. Rucky, and G. M. Rebeiz,
“A High Performance W-band Uniplanar
Subharmonic Mixer,” IEEE Trans MTT,
Vol 45, pp 955-962, Jun 1997.
M. J. Roberts, S. Iezekiel and C. M.
Snowdan, “A Compact Subharmonically
Pumped MMIC Self Oscillating Mixer
for 77 GHz Applications,” IEEE MTT-S
Digest, 1998, pp 1435-1438.
M. Kossor, WA2EBY, “A Digital
Commutating Filter,” QEX, May/Jun
1999, pp 3-8.
K. S. Ang, A. H. Baree, S. Nam and I.
D. Robertson, “A Millimeter-Wave
Monolithic Sub-Harmonically Pumped
Resistive Mixer,” Proc APMC, Vol 2,
1999, pp 222-225.
C. H. Lee, S. Han, J. Laskar, “GaAs
MESFET Dual-Gate Mixer with Active
Filter Design for Ku-Band Application,”
IEEE MTT-S Digest, 1999, pp 841-844.
H. Darabi and A. A. Abidi, “Noise in
RF-CMOS Mixers: A Simple Physical
Model,” IEEE JSSC, Vol 35, Jan 2000,
pp 15-25.
P. S. Tsenes, G. E. Stratakos, N. K.
Uzunoglu, M. Lagadas, G. Deligeorgis,
“An X-band MMIC Down-Converter,”
WOCSDICE 2000, pp X.7-X.8
M. T. Terrovitis and R. G. Meyer,
“Intermodulation Distortion in CurrentCommutating CMOS Mixers,” IEEE
Journ of SSC, Vol 35, No 10, October
2000.
M. W. Chapman and S. Raman, “A
60 GHz Uniplanar MMIC 4X
Subharmonic Mixer,” IEEE MTT-S Int
Microwave Symp Dig, 2001, pp 95–98,
2001.
M. Sironen, Y. Qian and T. Itoh, “A
Subharmonic Self-Oscillating Mixer with
Integrated Antenna for 60 GHz Wireless
Applications,” IEEE Trans MTT, 2001,
pp 442–450
D. Tayloe, N7VE, “Letters to the Editor,
Notes on ‘Ideal’ Commutating Mixers
(Nov/Dec 1999),” QEX, March/April
2001, p 61.
K. Kanaya, K. Kawakami, T. Hisaka, T.
Ishikawa and S. Sakmoto, “A
94 GHz High Performance Quadruple
Subharmonic Mixer MMIC,” IEEE
MTT-S Int Microwave Symp Dig, 2002,
pp 1249–1252.
Mixers, Modulators and Demodulators 10.35
J. Kim, Y. Kwon, “Intermodulation Analysis
of Dual-Gate FET Mixers,” IEEE Trans
MTT, Vol 50, No 6, Jun 2002.
B. Matinpour, N. Lal, J. Laskar, R. E.
Leoni and C. S. Whelan, “A Ka Band
Subharmonic Down-Converter in a GaAs
Metamorphic HEMT Process,” IEEE
MTT-S Int Microwave Symp Dig, 2001,
pp 1337–1339.
T. S. Kang, S. D. Lee, B. H. Lee, S. D. Kim,
H. C. Park, H. M. Park and J. K. Rhee,
Journ Korean Phys Soc 41, 2002, 533.
D. An, B. H. Lee, Y. S. Chae, H. C. Park,
H. M. Park and J. K. Rhee, Journ Korean
Phys Soc 41, 2002, 1013.
D. Metzger, “Build the ‘No Excuses’
Transceiver,” QST, Dec 2002, pp 28–34.
W. S. Sul, S. D. Kim, H. M. Park and J.
K. Rhee, Jpn Journ Appl Phys, Vol 42,
2003, pp 7189-7193 (see jjap.ipap.jp/
link?JJAP/42/7189/).
H. Morkner, S. Kumar and M. Vice, “An
18-45 GHz Double-Balanced Mixer with
Integrated LO Amplifier and Unique
Suspended Broadside-Coupled Balun,”
2003 Gallium Arsenide Integrated Circuit
Symp, Nov 2003, pp 267-270.
D. Tayloe, N7VE, “A Low-noise, Highperformance Zero IF Quadrature
Detector/Preamplifier”, March 2003, RF
Design.
D. Tayloe, N7VE, US Patent 6,230,000,
“A Product Detector and Method
Therefore”.
10.36 Chapter 10
M.-F, Lei, P.-S, Wu, T.-W. Huang, H. Wang,
“Design and Analysis of a Miniature
W-band MMIC Sub Harmonically
Pumped Resistive Mixer,” 2004 IEEE
MTT-S Int Microwave Symp, Vol 1, Jun
2004, pp 235-238.
Ping-Chun Yeh, Wei-Cheng Liu and
Hwann-Kaeo Chiou, “Compact 28-GHz
Subharmonically Pumped Resistive
Mixer MMIC Using a Lumped-Element
High-Pass/Band-Pass Balun,” IEEE
Microwave and Wireless Component
Letters, Vol 15, No 2, Feb 2005.
J. Browne, “Wideband Mixers Hit High
Intercept Points,” Microwave & RF
Journal, Sep 2005, pp 98-104.
Chien-Chang H. and Wei-Ting C., “MultiBand/Multi-Mode Current Folded UpConvert Mixer Design,” Proc APMC
2006, Japan.
X. Wang, M. Chen and O. B. Lubecke,
“0.25 um CMOS Resistive Sub threshold
Mixer,” Proc APMC 2006, Japan.
Ro-Min Weng, Jing-Chyi Wang and
Hung-Che Wei, “A 1V 2.4 GHz Down
Conversion Folded Mixer”, IEEE
APCCAS, Dec 4-7, 2006, Singapore,
pp 1476-1478.
B. M. Motlagh, S. E. Gunnarsson, M.
Ferndahl and H. Zirath, “Fully Integrated
60-GHz Single-Ended Resistive Mixer
in 90-nm CMOS Technology,” IEEE
Microwave and Wireless Component
Letters, Vol 16, No 1, Jan 2006.
F. Ellinger, “26.5-30-GHz Resistive Mixer
in 90-nm VLSI SOI CMOS Technology
With High Linearity for WLAN” IEEE
Trans MTT, Vol 53, No 8, Aug 2005.
U. L. Rohde and A. K. Poddar, “High
Intercept Point Broadband, Cost Effective
and Power Efficient Passive Reflection
FET DBM,” EuMIC Symposium, 10-15
Sep 2006, UK.
U. L. Rohde and A. K. Poddar, “A
Unified Method of Designing UltraWideband, Power-Efficient, and High
IP3 Reconfigurable Passive FET Mixers,”
IEEE/ICUWB, Sep 24-27, 2006, MA,
USA.
U. L. Rohde and A. K. Poddar, “Low Cost,
Power-Efficient Reconfigurable Passive
FET Mixers”, 20th IEEE CCECE 2007,
22-26 Apr 2007, British Columbia,
Canada.
INTERESTING WEB SITES FOR
MIXERS
www.analog.com/en/content/0,2886,770%
255F848%255F67465,00.html
www.maximic.com/solutions/base_
stations/index.mvp?CMP=3758&gclid=
CIzn1J_lm40CFSgRGgodD2Go6g
www.pulsarmicrowave.com/products/
mixers/mixers.htm
www.synergymwave.com/products/
mixers/?LocTitle=Mixers
news.thomasnet.com/fullstory/507272
www.wj.com/products/
HighInterceptFETMixers.aspx
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement