14

Radar Receivers AESA radiating face

and Digitization (Courtesy of Selex ES.)

T
here is no universally accepted definition of what
constitutes the receiver in a radar system. The most
common usage of the term refers to the unit that
takes the received radar frequency output from the
antenna (usually via a duplexer) and converts the signal into
a form that can be used either for display or for subsequent
digital signal processing.

In modern radars, this simple segmentation of the radar sys-

tem has become increasingly blurred. For example, what
were formerly receiver front ends are increasingly integrated
into phased array antennas, while on the other hand digital
processing, which was previously the preserve of a separate
processing unit, is increasingly being incorporated into the
receiver unit. This chapter deals mostly with receivers in the
classical sense, but important aspects of more recent design
approaches are also described.

There are two basic types of radar receivers: pulsed and

continuous wave (CW). Almost all airborne radars employ
pulsed receivers since most of them time-share the same
antenna for transmission and reception. Pulsed receivers,
which are the focus of this chapter, are more difficult and
complex to build.

Receivers have a wide variety of design requirements depend-

ing on the system type and the operating environment, but two
elements are essential across the board: sensitivity (the ability
to detect small signals, generally limited by thermal noise); and
selectivity (the ability to reject unwanted signals, typically by
frequency filtering).

This chapter describes the important basic principles under-

lying radar receiver design and then demonstrates how these
work together in several examples of complete receiver
systems.
195
196 PART III: Fundamentals of Radar

14.1 Basic Principles

The echo received in an airborne radar is extremely small,
often well below the level of thermal noise, which is typically
only about 10−15 W. The job of the radar receiver is to amplify
these tiny signals and to filter them out from background noise
and clutter—a task that in modern radars employs a great deal
of digital signal processing. This chapter deals mainly with the
analog parts of the receiver from the incoming radio frequency
(RF) signal through to the digitized output, but it does include
some aspects of digital filtering, which is an important part of
modern designs.

In addition to the basic tasks of amplification, filtering, and

digitization, the receiver must avoid any contamination or dis-
tortion of the signal since this may result in loss of sensitivity
or false detections.

14.2 Low-Noise Amplification

The sensitivity of all receivers is limited by thermal noise, and
in radar it is normally thermal noise from the receiver that
dominates (as opposed to external noise sources such as the
Sun). The first part of a radar receiver is a low-noise amplifier
(LNA), which is designed to set the sensitivity of the receiver
at the highest realistic level and is connected, usually via some
form of duplexing device, to the radar antenna.

The LNA is the key element in determining the basic sensitiv-

ity of the receiver, which is characterized by its noise figure,
F, defined by
SNR in
F =
SNR out

where SNR in is the signal-to-noise ratio at the receiver input,

and SNRout is the signal-to-noise ratio at the receiver output.

Even the best LNAs degrade the signal-to-noise ratio, but the
reason for this initial amplification is to ensure that the con-
tribution of any downstream noise sources is minimized. This
can be seen by examining Friis’s formula for cascaded noise
figure: in a cascade of amplifiers with the first having gain G1
and noise figure F1, the second gain G2 and noise figure F2, the
third gain G3 and noise figure F3, and so on, the overall noise
figure F is given by
( F2 − 1) ( F3 − 1) ( F4 − 1) .....
F = F1 + + + +
G1 G1G2 G1G2G3

If G1 is high, this makes all contributions other than F1 neg-

ligible, which is the goal in a well-designed system. The sys-
tem noise figure is largely determined by the first stage in the
receiver chain.

In most modern systems, a semiconductor LNA based on gal-

lium arsenide (GaAs) or gallium nitride (GaN) is employed.
 CHAPTER 14: Radar Receivers and Digitization 197

These components have revolutionized the design of radar

receivers, with common noise figures of 1 dB readily available,
some 10 times better than earlier systems.

Of course, nothing is without a price. It is also vital to avoid

distortion, so it is crucial that the LNA have linearity. A very
high gain device (large G1) will lack linearity. The trade-off
between linearity and noise figure is an important aspect of
receiver design.

In active electronically scanned array (AESA) radar, an LNA

is normally incorporated into each transmit/receive module
of the array, which reduces or eliminates the need for an
LNA at the input of any subsequent receiver. Any given array
has many LNAs—perhaps 1000 or more in a typical airborne
AESA radar. Although at first glance this could result in more
noise entering the radar receiver, this is not in fact the case.
The LNA at a single element will (given sufficient gain) set the
noise figure of the complete system, exactly as in a conven-
tional receiver.

14.3 Filtering
Filtering is vital and appears in many aspects of receiver design,
the most fundamental of which is to make sure that wanted
signals are received with minimal loss while simultaneously
minimizing the amount of noise that appears at the receiver
output. Filtering also plays a vital role in the design of down-
converters and digitizers.

The Matched Filter. Radar receiver designers, unlike designers

of an electronic warfare receiver, have the great advantage that
they know exactly what signal has been transmitted. Therefore,
detecting the echo is immeasurably easier than if the signal is
unknown. The key concept here is the matched filter, which
is designed to match the transmitted signal and maximizes the
signal-to-noise ratio at the receiver output.

Matched filter theory is a major branch of radar engineering

that originated with D. O. North’s 1943 paper. The essence of a
matched filter receiver is that it aims to correlate a known sig-
nal (i.e., the transmitted signal), with an unknown signal (the
received signal) to detect the presence of the known signal in
the unknown signal. The ideal matched filter is a time-reversed
replica of the transmitted signal.

This works regardless of the characteristics of the transmitted

signal: it can be a simple pulse; a frequency-modulated chirp; a
binary code sequence; or anything else, provided it is known.
The principle is most easily understood just by considering a
simple rectangular pulse of width τ. The matched filter for a
simple pulse is one whose frequency response is the Fourier
transform of the pulse. In simple terms, this equates to a filter
with a bandwidth of approximately 1/τ.
198 PART III: Fundamentals of Radar

A typical radar pulse might be 1 µs wide, so the matched fil-

ter needs to have a bandwidth of 1/10−6 Hz, or 1 MHz. This
is sufficient to pass the pulse with little degradation but is
narrow enough to minimize noise at the output. A narrower
bandwidth will reduce the noise but also the signal, whereas a
wider bandwidth will admit too much noise.

An ideal radar receiver would implement this filter right at

the front end. This would ensure that only desired signals
entered the radar receiver and would maximize sensitivity.
Unfortunately, it isn’t that easy in practice: a 1 MHz filter
implemented at the carrier frequency of a typical airborne
radar (10 GHz) is extremely difficult to implement because of
the very high precision (in relative frequency terms) required.
The existing techniques that accomplish this all have very
major drawbacks and are rarely used. So in practice this
narrowband filtering is usually applied later in the receiver,
after the incoming signal has been downconverted to a lower
frequency.

14.4 Downconversion
Almost all radar receivers employ frequency downconver-
sion. Although theoretically unnecessary, this is a solution that
allows filters and amplifiers to be realized in practice.

The basic idea is that the incoming RF signal, which carries

Mixer a modulation with a bandwidth sufficient to represent radar’s
pulses, is converted to a lower carrier frequency known as
RF Input IF Output the intermediate frequency (IF), fIF, but without changing the
modulation bandwidth. The downconversion process is carried
LO
out by a device called a mixer.

Figure 14-1. A simple mixer stage: the RF input and the Local Mixers (Fig. 14-1) are nonlinear devices that effectively multiply
Oscillator (LO) are mixed together to provide an output at an two signals together—in this case the RF signal and a fixed
Intermediate Frequency (IF). reference signal called the local oscillator (LO). We can easily
see how this works mathematically.

We can represent the RF signal at its carrier frequency, fc, by

S = sin(ωct)

where ωc = 2πfc, and t is time. Similarly, the LO signal at fLO is

LO = sin(ωLOt)

Then the output of the mixer, which multiplies the two sig-
nals, is

S ⋅ LO = sin(ω c t ) ⋅ sin(ω LOt )

= 12 (cos((ω c − ω LO )t ) − cos((ω c + ω LO )t ))

The output of the mixer is now two signals: the lower

sideband at a frequency below the original RF signal at
frequency fc − fLO; and the upper sideband above it at fre-
quency fc + fLO. Normally only the lower sideband is desired,
so the upper sideband (which is significantly separated in
 CHAPTER 14: Radar Receivers and Digitization 199

Mixers
M ixers are a critical element of radar receivers.
mathematically,
they can be regarded as a device that multiplies two
signals together, which gives the sum and difference frequen-
RF, so for half the LO period (3/2 × the RF period) we select one
arm of the balanced RF signal, and for the next half of the LO
period we select the other arm of the balanced RF signal. The
cies. In practice, such an ideal device does not exist, but good resultant waveform repeats every 3/2 × the period of the RF
approximations can be built. signal; that is, it is at 2/3 the frequency, which is the desired
mixer output.
The most common type used in radar is the double balanced
mixer, which uses four switching devices (diodes or transistors).
A highly simplified schematic of one type, the diode ring mixer,
is shown in the following figure.

Local
Oscillator

In a practical design, the mixer incorporates baluns (short for

balanced–unbalanced), which are circuits that can take an
unbalanced signal (e.g., RF on a coaxial line) and produce a
balanced output to feed the mixer. This is also normally done
with the LO.

Balanced
RF Signal

The LO signal is fed to the top and bottom points of the diode
ring. In practice, a very high amplitude drive signal (compared
with the RF signal) is used, so the LO signal approximates to
a square wave (due to saturation effects in the drive amplifi-
ers). The result is that half the time one side of the diode ring
is biased “on” and the other side is biased “off.” This situation
RF signal at the left-hand point of the diode ring
continually reverses at the LO frequency.

RF signal at the right and point of the diode ring

The RF signal is fed to the left and right points of the diode
ring in balance; the signal at the left point is 180 degrees out of
phase with the signal at the right point. Under the control of the
LO, one arm of the balanced RF signal is alternately shorted out
by the diodes that are biased on. This gives rise to a complex
waveform in the RF circuit, the mixer output, which includes
a strong component at the desired difference frequency. This
is then filtered out (using other components not shown here)
and fed to the next stage of the receiver.
A simple example is shown where the LO is at one-third of the Resultant signal with LO at 1/3 the RF frequency- the signal is now at 2/3
RF signal. The period of the LO is three times the period of the the RF frequency
200 PART III: Fundamentals of Radar

IF LO RF Upper
(RF-LO) Sideband
(RF+LO)
Frequency

Figure 14-2. These frequencies are present at the input (RF and LO) and the output of the mixer stage.

frequency) is filtered out. The lower sideband is centered

on fIF (Fig. 14-2).

Image Rejection. An important issue in the design of this

type of receiver is the image frequency response. An input
at the so-called image frequency, fi, will also result in an
output at fIF, where f i = fc – 2fIF (for the lower sideband case).
This is because a signal at the image frequency mixes to a
frequency of f i – fLO = –fIF, which is indistinguishable from
a signal at fIF. Thus, any signal at this image frequency will
appear in the IF band at the mixer output, resulting in inter-
ference (Fig. 14-3).
To combat this, it is important to filter the RF signal prior to the
mixer to minimize any input at the image frequency. Typically
the signal is filtered to a bandwidth approximately equal to
the IF. The design of such filters is itself a complex matter. It is
important to minimize any frequency-dependent amplitude or
phase distortion of the signal within the passband; otherwise,
high bandwidth signals will be significantly degraded.
Intermodulation. A further issue is intermodulation between dif-
ferent signals at the input. As far as the mixer is concerned, any
two signals at different frequencies will be multiplied together to
give intermodulation products, regardless of at which port of the
mixer the signals are present. Thus, if two frequencies (e.g., the
desired echo and a jamming signal) are present at the input and
are separated by the IF, they will intermodulate to produce an
output at the IF (quite apart from the mixing products produced
by intermodulation with the local oscillator). Filtering prior to
the mixer can also be used to eliminate this problem, provided

–IF 0 IF Image LO RF
(Img-LO) (RF-LO) Freq

Frequency

Figure 14-3. Image frequency signals at the image frequency can mix down and be indistinguishable from the desired IF output.
 CHAPTER 14: Radar Receivers and Digitization 201

the bandwidth of the input filter is made sufficiently narrow. Use

of a high IF makes filtering to reject such signals easier.
A simple rule of thumb in the design of a downconverter is
always to filter the input signal to a bandwidth less than the
subsequent IF before the mixer. It is easiest to use an example
to see how this works. Let’s imagine we have a radar operating
on a carrier frequency of 10 GHz and we want to downconvert
to an IF of 3 GHz, with a total signal bandwidth of 1 GHz.
We use a local oscillator of 7 GHz, which produces sidebands
centered on 3 GHz and 17 GHz. We then place a 1 GHz band- Sideband
width filter, centered on the IF (2.5 GHz~3.5 GHz), to remove RF Filter Mixer Filter

the upper sideband. The image frequency is 4 GHz, and a jam-

ming signal on 7 GHz could produce a damaging output on RF IF
the IF. However, if we filter the input signal to a bandwidth
of 2 GHz (9~11 GHz) before the mixer, this eliminates these
LO
signals before they can do any damage: any possible mixing
products will lie outside the IF bandwidth (Fig. 14-4). Figure 14-4. The mixer stage is shown here with RF and IF filtering.

Homodyne Receivers. These receivers, also called direct con-

version or zero IF receivers, are the simplest type of down-
converting receiver, where fLO is made equal to fc. This type
of receiver converts directly down to baseband (zero IF).
Although this is very simple, this type of receiver cannot follow
the previously outlined basic principles; it would require a zero
bandwidth filter at the input to the mixer. This type of receiver
may be found in very simple radars, but it is unsuitable for any
system that must be resistant to jamming and interference.

Heterodyne Receivers. Also called superheterodyne or superhet

receivers, in these downconverting receivers fLO is significantly dif-
ferent from fc. This type of design can incorporate the necessary
filtering needed to observe the aforementioned design principles.

Multiple-Stage Downconversion. It is often necessary to down-

convert to a low IF to carry out narrow-band filtering or ana-
log-to-digital conversion. If the downconversion is done in a
single stage, this makes the receiver vulnerable to jamming and
interference as it becomes close to a homodyne receiver. The
solution to this is to downconvert in multiple stages, with filter-
ing and amplification at each stage, in which the same prin-
ciple is observed: before mixing, filter to less-than-subsequent
IF. This approach facilitates practical filter designs because it
avoids the need for very narrow fractional bandwidths (i.e., fil-
ter bandwidth is a very small fraction of the carrier frequency),
which are difficult to implement.
A typical X-band radar receiver includes several stages of
downconversion: initially to an ultra high frequency (UHF) IF
(a few GHz); then to a very ultra high frequency (VHF) IF
(a few hundred MHz); then possibly to baseband.

14.5 Dynamic Range

Dynamic range is a simple but often misunderstood concept.
The key issue is a receiver’s ability to pass large and small sig-
nals simultaneously so that the signals are not distorted.
202 PART III: Fundamentals of Radar

All amplifiers are nonlinear; that is, the output is not simply a
multiple of the input at all times. Therefore, if signal is applied
to an amplifier, it will not simply be made bigger; it will also
be distorted. The bigger the input signal, the greater the distor-
tion; only small signals can be amplified with very little distor-
tion. This is bad news, because a distorted signal can easily
give rise to false detections. The trick here is to make sure
the amplifier is linear enough to deal with the biggest signal
it is likely to encounter, and to do this means increasing the
bias currents to the transistors in the amplifier, which in turn
increases power consumption. This is the price that must be
paid if a high dynamic range is necessary.

Other tricks sometimes work. In a radar that measures range

Max Gain at without ambiguity, it is possible to vary the gain of the receiver
Gain Long Range throughout the pulse repetition interval (PRI). This is known as
Low Gain at
Short Range sensitivity time control (STC), or swept gain. It works because
strong signals usually originate close to the receiver, whereas
weak signals are usually at long range. STC gives just the illu-
0 Range
sion of high dynamic range because it cannot pass large and
small signals simultaneously; they have to be separated in time.
Figure 14-5. Sensitivity time control is a technique where the Nevertheless, it is a very useful compromise and is used in
receiver gain is varied as a function of time (or range).
many systems (Fig. 14-5).

In radars where range is ambiguous, STC cannot be used. Such

radars as a rule also employ high levels of Doppler process-
ing. Therefore, the demand for high dynamic range has to be
fundamentally addressed; otherwise the small target signals
will be lost in spurious signals generated from the large clutter
signals. This entails high bias currents and high power con-
sumption in the receiver’s amplifiers.

A further negotiation is automatic gain control (AGC), which

gives only an illusion of greater dynamic range. However, this
approach is useful because the level of signals that the radar
has to deal with can vary greatly with time, and AGC gives a
way of compensating for this.

In absolute terms, a radar would ideally have a dynamic range

of perhaps 200 dB (1020:1) to cope with the full range of signals
it is likely to encounter in all circumstances. This is not practi-
cal, and 100 dB is a more realistic design (although even that
is not easy to achieve). A dynamic range of 100 dB, plus an
additional 100 dB of AGC, is the sort of compromise typically
found in a high-performance radar receiver.

An AGC system measures the radar’s operating environment

and adjusts the gain accordingly so that the instantaneous
dynamic range of the receiver is optimally used. This may be
done in a variety of ways, the simplest of which is to provide
the operator with a gain control. Most modern systems, how-
ever, are automatic and use the measurements of the signal
background to adjust the receiver. These control loops need to
be carefully designed because they can be exploited by jam-
ming systems to reduce radar sensitivity.
 CHAPTER 14: Radar Receivers and Digitization 203

14.6 Spurious Signals and Spectral Purity

Doppler radar uses spectral analysis to separate small targets –40
from large clutter signals. An important issue in this type of –50
radar is spectral purity. The transmitted signal and the local
–60
oscillator signals are generally designed to be pure tones at
–70
specific frequencies, but in practice they possess components at

dBc / Hz
other frequencies due to practical limitations in the way they are –80

generated. These unwanted components include both discrete –90

spurious frequencies (or spurs) and broadband phase noise. –100

A typical phase noise plot is shown in Figure 14-6. The plot –110

shows (as a function of frequency on a logarithmic scale) the –120

10 100 1000 10000 100000 1000000
power spectral density of the noise measured in dB below Hz
carrier (dBc) in a 1 Hz bandwidth. An ideal, pure tone would
Figure 14-6. This typical phase noise plot shows phase noise
consist simply of a line at zero on the frequency scale, but no spectral density as a function of frequency offset from the carrier.
practical oscillator is as good as this. In the figure, at a fre-
quency 1000 Hz away from the carrier, the power spectral den-
sity is about –98 dBc/Hz. This means that in a typical detection
filter bandwidth of 100 Hz the noise from the reference oscil-
lator will be 78 dB below the carrier (–98 + 10 log10(100) = –78).
This limits the radar’s ability to detect small signals at low
Doppler offsets. It’s important to note that filtering to narrower
bandwidths always helps to reduce the effects of phase noise.
Spurs fall into two classes: signal-related spurs (SRS); and
non-signal-related spurs (NRS). The former may increase or
decrease in amplitude with the presence of a signal and are
normally internally generated, whereas the latter tend to be at
a constant level and are caused by nonlinearities in the receiver
design, which give rise to intermodulation products.
Spurs differ from phase noise in that they are inherently nar-
row band, so changing the radar’s filtering bandwidth makes
little, if any, difference to their level, unlike for phase noise.
They are damaging because they can appear as false targets;
phase noise results in an increase in the radar’s noise level,
which reduces its sensitivity.
Control of spurs and phase noise is an issue principally for
the radar’s signal generation system, but maintaining adequate
dynamic range is also critical. It is also vital to eliminate other
sources of spurious signals, such as breakthrough from power
supplies and interference due to poor screening from other
signals within the system. A radar receiver typically uses a
great deal of internal screening to isolate sensitive parts of the
receiver, so most receivers internally look like a series of sepa-
rate metal boxes (Fig. 14-7).

14.7 Digitization
Early radar receivers provided an analog output, called the video
output, direct to the display, which had been downconverted
to the frequency needed to drive the display. This approach
is now obsolete, and all radars employ some form of analog- Figure 14-7. This typical airborne radar receiver shows separately
to-digital (A/D) conversion at the receiver output. Digitized screened modules.
204 PART III: Fundamentals of Radar

output greatly facilitates subsequent signal processing and dis-

play processing.
An A/D converter provides a sequence of digital values that
represent the output voltage of the radar receiver at discrete
time intervals. Normally, the output voltage is sampled on a
regular basis, and the sample value is held in a sample-and-
hold (S/H) circuit while it is converted to digital form. An S/H
typically consists of a capacitor to hold the voltage and an
electronic switch to disconnect the input. The digital represen-
tation of the signal, in binary form, can then be processed in a
digital signal processor or used to drive a digital display.

The earliest A/D converter designs sampled the radar output

after downconversion to a low IF in what was called the video
output because in analog radars this was the output that drove
the display. The sampling rate chosen was similar to the radar’s
range gate, so the sequence of samples represented the radar
output at various ranges. This approach required all the radar
pulse matched filtering to be carried out in analog form prior
to A/D conversion. Although this method was effective for
radars with a small range of pulse sizes, complex, multifunc-
tion radars demanded a range of filters that could easily drive
up the complexity of the receiver.

More recent designs, enabled by modern A/D conversion tech-

nology, employ much higher sampling rates and carry out digi-
tization at a higher IF. This allows more of the radar’s filtering
to be performed digitally, which is greatly beneficial since the
digital processing is readily reconfigurable and stable.

In theory, this idea could be extended all the way up to the

RF, with direct digitization of the incoming signal. The high
dynamic range required in radar is what mainly prevents this
from happening: A/D conversion technology simply cannot
support the required performance.

Nyquist Criterion. The basic principle of sampled-data systems

was established by Shannon and Nyquist, who proved that the
minimum sampling frequency, fs, required to represent a sig-
nal with maximum frequency, f, is given by fs = 2f. Sampling
of higher frequencies leads to aliasing, the phenomenon that
occurs when the sampled signal appears erroneously as a
lower frequency. In practical systems, then, it is important to
include a low-pass filter, called the anti-aliasing filter, prior to
the S/H and A/D conversion.

Effective Number of Bits (ENOB). In radar, dynamic range is

all-important. An A/D conversion with n bits can represent 2n
voltage values and in principle can represent a dynamic range
of approximately 6n dB. However, practical converters are
imperfect and achieve a lower dynamic range than the number
of bits might suggest. A converter’s ENOB is that over which
the converter meets a certain specified level of performance
in terms of linearity and freedom from spurs. It is difficult to
generalize, but the ENOB is usually two to three bits less than
the nominal number of bits.
 CHAPTER 14: Radar Receivers and Digitization 205

Maximum Input Frequency. Another important characteristic

of an A/D conversion, maximum input frequency must be cho-
sen to match the Nyquist criterion for the radar signal. Higher
frequency converters always have a lower ENOB. This is for
a wide variety of reasons, including the effects of time jitter
(which becomes more dominant at higher frequencies) and the
difficulty of building converters fast enough to match the sam-
pling rate. In practical radar receiver designs, the downconver-
sion architecture and the A/D converter designs are the subject
of a complex trade-off to achieve the required dynamic range
and noise performance.

Noise. The digital representation of the signal introduces an

additional noise component, known as quantization noise. In
power terms, it can be shown that this noise has a power of
Q2/12, where Q is the voltage quantization interval. To pre-
serve the noise figure of the receiver, it is essential that the
front-end thermal noise of the system is accurately represented
at the A/D converter output. If we design a receiver such that
front-end thermal noise has a mean voltage of 2Q, the quanti-
zation noise will be about 16 dB below the thermal noise. This
is about right to ensure that the back-end noise is negligible
compared with the thermal noise floor.

Types of A/D Converters. A/D converters have a variety of

designs, two of the most common of which are the direct con-
version, or flash; and the successive approximation. The flash
is probably the simplest in concept; it includes a large array of
voltage sources with each one equal to the value of every pos-
sible digital output. The input signal is compared with each of
these possible voltages in parallel via a large array of compara-
tors. This allows the digital representation to be obtained very
rapidly. This design is used for the fastest A/D converters. The
limitation, however, is that the complexity of the comparator
network restricts the design to about 8 bits, and an 8-bit system
requires 28 comparators, where each extra bit doubles the com-
plexity and power consumption. By contrast, successive approx-
imation uses only a single comparator to sequentially compare
the input voltage with a range of voltages that is successively
narrowed. At each successive step, the converter compares the
input voltage to the output of an internal digital to analog con-
verter that has been set to the midpoint of the expected voltage
range. The measured error is then used to set a smaller range
for the next step. This design exploits the fact that it is easier
to build an accurate digital-to-analog (D/A) converter and that
by continually narrowing the voltage range higher resolution
and accuracy is obtained. The disadvantage is that the time this
method takes hampers the maximum input frequency.

Digitization in Noncoherent Radar. Noncoherent radars use

just the amplitude of the received radar signal and not the
phase information. Thus, the digitizer only must convert the
voltage output of the radar receiver for subsequent processing.

Normally, the radar receiver is linear, so the digital output is

directly proportional to the RF input signal. However, in some
206 PART III: Fundamentals of Radar

radar designs logarithmic amplifiers are used prior to the A/D

conversion to compress the dynamic range on the display. The
same function can be achieved using an A/D converter with
nonlinear quantization steps, but this is much less common.
Digitization in Coherent Radar. Coherent radars use both the
amplitude and phase of the received radar signal in subse-
quent Doppler processing. To represent this complex quantity
digitally, either the phase angle, φ , and an amplitude value, A,
must be calculated, or, much more commonly, two Cartesian
I quantities known as the in-phase (I) and quadrature (Q) values
ADC
are used. These are simply related by
LO
Clock I = A cos(φ)
90°
Q = A sin(φ)
ADC
Q Digitizing the radar output to provide I and Q values can be
conveniently combined with the final downconversion stage.
Figure 14-8. A receiver digitization architecture using analog I/Q
downconversion and baseband digitization. In this case, the IF signal is split into two equal parts and fed
to two parallel mixers. The LO signal is fed to both, but in
one case with a 90° phase shift (Fig. 14-8). Two separate A/D
converters digitize the output of the two mixers simultane-
ously, providing the I and Q values. This approach requires
only low-frequency A/D converters, which makes it easier to
I achieve the required dynamic range. However, it is a complex
analog design because the two signal paths must be precisely
matched in gain, phase, and delay. This is seldom possible in
0,1,0,–1 practice, and complex calibration techniques using injected
ADC
1,0,–1,0 test signals are usually employed to calculate appropriate
corrections.

Clock Q An alternative way of achieving the same result is through sam-

pling at the IF and using digital downconversion (Fig. 14-9).
Figure 14-9. A receiver digitization architecture using IF
digitization and digital I/Q downconversion achieves the same In this architecture a single, much higher frequency A/D con-
function as the analog design of Figure 14-8. verter is employed. Its output is then multiplied by the digital
equivalent of the LO, which can be just a series of zeroes and
plus or minus ones (which avoids the need for high-speed digital
multipliers). This method uses the Hilbert transform. The two
resultant digital data streams are then digitally low-pass filtered,
just as in an analog system, to produce the I and Q outputs.
This design is physically simpler than the analog method, and
it avoids the need for precisely balanced circuits or complex
calibration. However, it needs a much higher speed A/D con-
verter, which will inevitably have a lower dynamic range. The
choice between analog and digital downconversion for coher-
ent systems is a complex trade-off, but today A/D convert-
ers have improved to the point that a digital architecture can
deliver the levels of performance required.
An example of a modern IF digitizer is shown in Figure 14-10.
It consists of a single double-sided printed circuit board with
surface-mounted components. The smaller dark components
are the A/D converters, and the large silver components are
Figure 14-10. Pictured here is an actual IF digitizer unit with digital field-programmable gate arrays (FPGAs), which implement the
downconversion. (Courtesy of Selex ES.) digital downconversion and digital filtering.
 CHAPTER 14: Radar Receivers and Digitization 207

14.8 Radar Receiver Architectures

Different receiver architectures have developed as airborne
radars have grown in complexity and functionality and as
the need to reject jamming and interference have increased.
Current designs all use some or all of the fundamental building
blocks described in Section 14.1.
Early airborne radars were noncoherent: they relied solely on
the amplitude of the radar echo, and any phase information was
discarded. The receivers in these simple radars amplified and
downconverted the radar echo to a low video IF, which was then
rectified before the amplitude was displayed to the operator. A
detector diode carried out the rectification step, which discarded
the phase information. Although this type of radar was very suc-
cessful and has remained in use (in one form or another) until
today, its major drawback (for an airborne radar) is that it can
easily lose a target in the very large ground clutter echo.
To address this drawback, coherent radars were developed,
where phase information is not discarded and Doppler filtering
can be employed to separate the target from clutter. In early
systems this was carried out using a large bank of analog fil-
ters, each individually tuned to a different Doppler frequency.
Unfortunately, this design was largely incompatible with a
pulsed radar design because the analog filters could operate
correctly only on continuous wave (CW) signals.
Interrupted CW radars provided a partial solution to this prob-
lem by employing a 50/50 transmit/receive duty cycle at the
expense of significant signal loss, which allowed for the use
of an analog filter bank. This technique was implemented suc-
cessfully in many airborne radars in the 1960s and 1970s. Still,
it suffered from a significant problem with eclipsing: because
the receiver was turned off for half the time, target echoes
could easily be lost. In addition, the analog filters were very
prone to drifting off-tune.
A/C converters provided a major breakthrough in solving the
issue of target loss due to large ground clutter echo. They allowed
the Doppler filter bank to be mechanized using a digital fast
Fourier transform (FFT), with a separate one for each range gate.
This avoided all the major drawbacks of the earlier design: the
signal losses were directly eliminated, and the digital filter bank
was reproducible and stable (unlike its analog predecessor).
Progress since then has been more incremental. Major inno-
vations include higher speed digitization, which allows more
filtering to be performed digitally. This is more stable, repro-
ducible, and easily reconfigured. In addition, multiple parallel
receiver channels support advanced spatial processing tech-
niques for clutter and jamming rejection.

14.9 Pulsed Noncoherent Receivers

The pulsed noncoherent receiver is the simplest type used in air-
borne radars. A typical block diagram is shown in Figure 14-11.
208 PART III: Fundamentals of Radar

Sideband Pulse
RF Filter Filter Filter
Detector
DISPLAY
RF
1st IF 2nd IF

LO1 LO2

Figure 14-11. This block diagram shows a noncoherent pulse receiver.

It shows the main mixing and filtering steps, but for simplicity
the necessary intermediate amplifiers are omitted. The output
of the radar antenna is fed to the receiver via an RF filter that
limits the band of the input signal to prevent unwanted mix-
ing products and may be preceded by an LNA. Following the
initial downconversion to a UHF IF, the unwanted mixing side-
bands are filtered out, and a second-stage downconversion to
video (typically less than 100 MHz) is carried out. At this point
a fixed analog pulse matched filter is used before the detector
stage. The pulse matched filter doubles as a filter to remove
unwanted sidebands.
This type of radar commonly only has one pulse width, so
only one pulse matched filter is required. If the radar has more
than one pulse width, different filters are needed with a mech-
anism for switching between them.
This type of receiver commonly has a fairly poor noise figure,
perhaps 10 dB or more. An LNA at the front end will improve
matters significantly, and in a well-designed receiver a noise
figure of 3 dB is achievable. However, this will require a lot
of attention to detail it throughout the signal path, ensuring
that losses in filters and mixers are compensated by suitable
distributed amplifiers.
Earlier noncoherent receivers would simply feed the detected
video to an analog display. This has now been superseded by
the insertion of an A/D converter at the receiver output. The
digitized video signal can then be conveniently displayed on
any suitable digital device as well as put in a form that permits
further signal processing.

14.10 Pulsed Coherent Receiver with Baseband

Digitization
The majority of modern airborne radar receivers are coher-
ent to support advanced modes of operation including clutter
rejection and synthetic aperture radar. A typical block diagram
is shown in Figure 14-12.

It shows the main mixing and filtering steps, but for simplicity
the necessary intermediate amplifiers are omitted. At the front
end, the LNA sets the noise figure of the system by providing
sufficient gain that the noise contribution of subsequent stages
is minimized. The signal is then downconverted in two stages,
with intermediate filtering. As in the noncoherent receiver,
 CHAPTER 14: Radar Receivers and Digitization 209

Anti-Alias
Filter
I
ADC
Sideband Pulse
LNA RF Filter Filter Filter
LO3
Clock
RF 90°
1st IF 2nd IF

ADC
LO1 LO2
Q

Figure 14-12. This block diagram shows the elements of a coherent pulse receiver with baseband digitization.

pulse matched filtering is usually carried out at a VHF. Here this

is the second IF signal, which is split into two, with one replica
being downconverted with a 90° phase shifted LO before anti-
alias filtering and A/D conversion. The I and Q data streams are
then made available to downstream digital signal processing.

Although apparently quite simple and elegant, this architecture

is comparatively difficult to implement because it requires sev-
eral different LO signals, complex matched circuitry, and (in
general) active calibration schemes to achieve the desired level
of performance.

14.11 Pulsed Coherent Receiver with IF

Digitization
More modern coherent receivers employ digital downconver-
sion. A typical block diagram is shown in Figure 14-13.

As before, the LNA at the front end sets the noise figure of the
system by providing sufficient gain that the noise contribution
of subsequent stages is minimized. The signal is then downcon-
verted to a high IF (typically several GHz) in a single stage. After
this the signal is anti-alias filtered before digitization with a high-
speed A/D converter. The formation of the I and Q signals and
all subsequent filtering is carried out digitally, either in custom
digital circuitry, an FPGA, or a programmable digital processor.
FPGAs are a preferred approach as they can provide very high
throughput digital processing for a relatively low-power budget,
but different systems may employ different approaches.

Sideband Anti-Alias
LNA RF Filter Filter Filter
0,1,0, –1
ADC
RF
1,0, –1,0
1st IF

LO1
Clock Q

Figure 14-13. This block diagram shows the coherent pulse receiver with IF digitization.
210 PART III: Fundamentals of Radar

This design greatly simplifies the required analog circuitry and

avoids the need for much complex and difficult analog design.
Pulse matched filtering is now carried out digitally (e.g., within
the dotted boundary shown in Fig. 14-13) and can easily be
reconfigured to support a wide range of pulse waveforms,
including both unmodulated pulses and coded pulses (for
pulse compression).

The problem with this type of design is the need for a very
high-speed A/D converter. According to Nyquist’s criterion,
the sampling rate must exceed twice the maximum input fre-
quency. Thus, with an IF of 2 GHz the A/D converter would
appear to run at 4 GHz or more. This is challenging if we need
a good dynamic range, but fortunately there is a clever trick
that allows the use of slower A/D converters.

Sub-Nyquist Sampling. Nyquist’s criterion applies to repetitive

signals but makes no particular assumption about their struc-
ture other than their frequency content. In the case of a pulsed
radar receiver, the signal has certain special characteristics that
we can take advantage of so that in fact we can sample slower
than Nyquist’s criterion would suggest.

The radar signal consists of a signal with finite bandwidth dic-

tated by the radar’s range resolution. If the resolution is 150 m,
the required signal bandwidth is nominally only 1 MHz. This
signal is modulated on the RF carrier and is replicated at IF
following downconversion. We can thus imagine the IF signal
as being a pure carrier frequency with a limited narrowband
modulation superimposed upon it. In principle, it is necessary
to sample this signal at only twice the pulse bandwidth, not
twice the IF. Thus, a lower speed A/D converter can be used.

In practice this is not so simple, and how well it works depends

(once again) on the quality of filtering. To use sub-Nyquist sam-
pling, the anti-alias filter becomes a band-pass filter, centered on
the IF, rather than a low-pass filter as is normally envisaged. If
this were all that were required, things would be relatively sim-
ple, but in fact complex detailed circuit characteristics of the A/D
converter are often the limiting factor. Practical A/D converters
have a maximum input frequency beyond which the bandwidth
of their internal circuitry is exceeded. Therefore, even if a suit-
ably filtered IF signal is put in, the low-pass characteristic of the
A/D converter will suppress the wanted signal. As ever, practi-
cal designs are a complex trade-off of various factors, but sub-
Nyquist sampling is an important and effective technique.

An example of a modern receiver of this type is shown in

Figure 14-14.

14.12 Multichannel Receivers

Most airborne radars employ several parallel receiving channels.
Figure 14-14. Pictured here is an actual coherent pulse receiver
with IF digitization. The downconverter and filtering components Early designs employed these for monopulse tracking, where
can be seen on the upper side, with the digital components on the the signal was compared in two precisely matched receivers and
underside. (Courtesy of Selex ES.) the difference was used to calculate the angular tracking error.
 CHAPTER 14: Radar Receivers and Digitization 211

Such techniques are still widely used, but it is less common to

find precisely matched analog receivers simply due to the sheer
difficulty of achieving the required performance. More modern
designs use active calibration techniques to measure and cor-
rect errors and are therefore much more stable and accurate.

Monopulse Receivers. These are the most common type and

use two or sometimes three parallel receivers. One is dedicated
to the main antenna sum channel output and the others to the
azimuth and elevation difference channel outputs. In a two-
channel design the azimuth and elevation difference signals
are time-multiplexed. Monopulse receivers are normally just
straight replicas of single channel receivers.

Guard Channel Receivers. In some radars a guard channel is

incorporated to resolve between antenna mainlobe and sid-
elobe returns. The requirement here again is for good matching
between the main and guard channel receivers in time, ampli-
tude, and frequency so that an accurate comparison can be car-
ried out and unwanted signals rejected. The most sophisticated
guard systems carry out cancellation where an amplitude and
phase shifted version of the guard channel output is added to the
sum channel to cancel out an unwanted signal, such as a jammer.

Multichannel Receivers. These ideas may be extended up

to systems with several tens or even hundreds of channels.
Adaptive sidelobe cancellation is a generalized form of guard
channel cancellation that can deal with many interfering sig-
nals simultaneously, and this can be extended to include can-
cellation of clutter signals. Achieving good balance across a
receiver array is important, because any imbalances, while
they can be nulled out, reduce the ability of the system to deal
with unwanted external signals.

The ultimate in multichannel receivers is to have an indi-

vidual receiver for each antenna element in an electronically
steered array. This can offer enormous flexibility, although at
the expense of a very large amount of digital signal process-
ing (and cost). This approach currently remains impractical for
most airborne radars, although some low-frequency radars, with
a limited number of antenna elements, have already successfully
gone this way. It is likely to become more common in the future.

14.13 Specialized Receivers

This section focuses on two specialized types of radar receivers.

Frequency-Modulated CW Receivers. Most of the discussion in

this chapter has concerned pulsed receivers since these are
by far the most common type in airborne radar, principally
because most airborne radars time share the same antenna for
transmission and reception. However, some radars, typically
very small ones, use separate transmit and receive antennas
and thus are able to use CW transmissions. In practice, nearly
all such radars use frequency-modulated CW (FMCW) where
the modulation frequency is used to measure range.
212 PART III: Fundamentals of Radar

Designs for this type of receiver differ little from the coherent
pulse designs already discussed. The main difference is that
the filtering is now defined by the FM bandwidth rather than
the pulse bandwidth. In other respects they are very similar.
Stretch Receivers. Stretch, or deramp-on-receive, receivers are a
specialized type commonly used in very high range resolution
radars such as synthetic aperture mapping radars. The architec-
ture of the stretch receiver is made so that a relatively narrow
band receiver can provide a much wider actual bandwidth.
How is this trick achieved? The idea is to exchange time and
frequency. A stretch radar transmits a linear FM pulse, where
the carrier frequency is linearly ramped up over a range that
defines the total bandwidth of a transmission. So, for example,
an X-band radar might ramp the carrier frequency from 8 to
10 GHz over a 100 µs period, providing a total bandwidth of
2 GHz. So far so good, but if our receiver has a bandwidth of
only 100 MHz, how do we cope?
The solution is that, at the expected time of arrival of the
returned signal, the first LO is then ramped at precisely the
same rate as the transmitted signal. The echo and the LO ramp
in parallel, and the difference frequency remains constant (for
echoes from the same range). Varying the range will thus cause
a different frequency to emerge from the first mixer. If this
is done correctly, the bandwidth of the signal emerging from
the mixer will be kept within the overall bandwidth of the
receiver so that very fine range resolution can be achieved by
frequency analyzing the signal output. The limitation is that
this works over a limited range swath.
This trick of exchanging time for frequency is simple enough
to implement in the receiver. It is necessary (in a coherent
receiver) only to ensure that the front end has enough band-
width to pass the full signal and to be able to ramp the first
LO appropriately. Most of the complexity falls on the signal
generation and timing; the receiver is almost unchanged.

14.14 Summary
Radar receiver designs are a complex trade-off. They must
achieve very high dynamic range, low noise figure, and high
spectral purity within a compact space and with performance
in many cases limited by fundamental physical limits. Simple,
elegant designs that digitize at the radar carrier frequency are
impractical for all but the simplest radars and are likely to
remain so.

Practical receiver designs all employ some form of frequency

downconversion, and the filtering design is usually the cru-
cial factor in determining receiver performance. Filtering
must be carried out at RF, IF, for anti-aliasing and for matched
filtering.

A/D conversion technology is a critical enabler, and mod-

ern designs allow more and more filtering to be performed
 CHAPTER 14: Radar Receivers and Digitization 213

digitally, which is much more stable and repeatable. The pen-

alty is that this places very stringent requirements on the A/D
converter.

Further Reading
H. T. Friis, “Noise Figures of Radio Receivers,” Proceedings of
the IRE, pp. 419–422, July 1944.
D. O. North, “An Analysis of the Factors which Determine
Signal/Noise Discrimination in Pulsed Carrier Systems,”
Proceedings of the IEEE, Vol. 51, No. 7, pp. 1016–1027, July
1963.
P. P. Vaidyanathan, “Generalizations of the Sampling Theorem:
Seven Decades after Nyquist,” IEEE Transactions Circuits
Systems I, Vol. 48, No. 9, September 2001.
H. Nyquist, “Certain Topics in Telegraph Transmission Theory,”
Transactions of the AIEE, Vol. 47, pp. 617–644, January
1928 (reprinted in Proceedings of the IEEE, Vol. 90, No. 2,
pp. 280–305, February 2002).
J. B. Tsui, Digital Techniques for Wideband Receivers, SciTech-
IET, 2004.
M. I. Skolnik (ed.), “Radar Receivers,” chapter 6 in Radar
Handbook, 3rd ed., McGraw Hill, 2008.
J. B. Tsui, Special Design Topics in Digital Wideband Receivers,
Artech House, 2009.

Test your understanding

1. A receiver has a front-end LNA with
a gain of 15 dB and a noise figure of
1.5 dB. The second stage of the receiver
has a gain of 20 dB and a noise figure of
10 dB, and the third stage has a gain of
30 dB and a noise figure of 15 dB. What
is the overall noise figure of the receiver?
2. An airborne radar operating on 10 GHz
has a first IF of 2.5 GHz. What is (a) the
frequency of the first local oscillator and
(b) the image frequency?
3. A coherent airborne radar employs
downconversion to baseband (I and Q)
and baseband digitization. The two A/D
converters each have a sampling rate of
100 MHz. What is the maximum signal
bandwidth the receiver can deal with?