Radio Detection and Ranging (Radar)

Part I

Reza Nasiri Mahalati

EE259 - Principles of Sensing for Autonomy

Stanford University

May 8, 2023
 Outline

Brief history of radar

Radar physics

Radar antennas

Radar range equation

Radar system architecture

Range and velocity estimation in FMCW radar

Range and velocity estimation in PMCW radar

EE259 - Principles of Sensing for Autonomy Brief history of radar 2

 Radio detection and ranging (radar)

▶ Radar is a sensor that uses RF electromagnetic waves for mapping

▶ Can determine the distance (range), angle and radial velocity of
objects relative to the sensor
▶ Radar concept first proposed in the early 1900s by Heinrich Hertz
▶ First experimental radar system was developed in 1935 by Robert
Watson-Watt in the UK
▶ Initially used as early warning systems for the military
▶ First operational system was the Chain Home radar network built by
the British Royal Air Force in 1938 and used during WWII:
– Operated at 20 − 30MHz frequency
– 99 mile maximum range
– 140◦ × 40◦ field of view in azimuth and elevation
– 21 sites along the east coast of the Great Britain
– Provided early warning of direction and size of German aircraft
– Indefinitely postponed the German invasion of Great Britain!

EE259 - Principles of Sensing for Autonomy Brief history of radar 3

 1939: The Chain Home radar network

EE259 - Principles of Sensing for Autonomy Brief history of radar 4

 2023: Modern radar applications

Today modern radar systems with vastly different form factors and
performance are used in a variety of applications beyond military, such as
avionics, meteorology, security, astronomy, agriculture and autonomy

EE259 - Principles of Sensing for Autonomy Brief history of radar 5

 Outline

Brief history of radar

Radar physics

Radar antennas

Radar range equation

Radar system architecture

Range and velocity estimation in FMCW radar

Range and velocity estimation in PMCW radar

EE259 - Principles of Sensing for Autonomy Radar physics 6

 Electromagnetic radiation

Electromagnetic (EM) radiation is waves of electromagnetic fields prop-

agating in space and carrying energy, e.g. RF waves and light.

EE259 - Principles of Sensing for Autonomy Radar physics 7

 Electromagnetic fields
Maxwell’s equations model the behavior of electromagnetic fields:
∇ · E = ρ/ϵ
∇·H =0
∂H
∇ × E = −µ
∂t
∂E
∇×H =J +ϵ
∂t

ρ: electric charge density (C/m3 )

J: electric current density (A/m2 )
ϵ: permittivity of medium (F/m)
µ: permeability of medium (N/A2 )
∂E
Divergence of E: ∇ · E = ∂E y
∂x + ∂y + ∂z
x ∂Ez

Curl of E:   
∂Ey ∂Ey
∇ × E = ∂E ∂Ex ∂Ez ∂Ex


∂y
z
− ∂z x̂ + ∂z − ∂x ŷ + ∂x − ∂y ẑ
EE259 - Principles of Sensing for Autonomy Radar physics 8
 Electromagnetic waves

▶ Since the magnetic field H is divergence-free, a magnetic vector

potential A can be defined such that:
∇ × A = µH
∂V
∇ · A = −µϵ ,
∂t
where V is the electric potential in volts and the second equation is
called the Lorentz condition
▶ Plugging A into Maxwell’s equations and some manipulation gives
the wave equation for vector potential A:
1 ∂2A
∇2 A − = −µJ,
c2 ∂t2
√
where c = 1/ µϵ is the speed of light
▶ Time varying current densities can result in radiated EM fields
▶ Is the same as sound wave equation in homogeneous case (J = 0)
EE259 - Principles of Sensing for Autonomy Radar physics 9
 Electromagnetic waves
▶ Assume source of time varying current contained in a volume V ′

▶ Solution to the wave equation for A is given by:

J(r′ , t − R/c) 3 ′
Z
µ
A(r, t) = d r
4π V ′ R
▶ τ = R/c: time required for effect of J to be felt a distance R away
EE259 - Principles of Sensing for Autonomy Radar physics 10
 Electromagnetic waves

▶ Assume the electric currents at the EM radiation source (i.e. the

antenna) are single frequency (time-harmonic):

J(r′ , t) = J(r′ )ej2πf t ,

where f = c/λ is the excitation frequency

▶ Resulting EM fields are also time-harmonic: A(r, t) = A(r)ej2πf t
▶ Magnetic vector potential wave equation simplifies to:

∇2 A + ∥k∥2 A = −µJ,

where ∥k∥ = 2π/λ is the wavenumber

▶ Solution to the time-harmonic wave equation is given by:

e−j∥k∥R 3 ′
Z
µ
A(r) = J(r′ ) d r
4π V ′ R

EE259 - Principles of Sensing for Autonomy Radar physics 11

 Electromagnetic waves

▶ Once A(r) is solved for, can find E(r) and H(r) fields by plugging
A back into Maxwell’s equations:
1
H= ∇×A
µ
j
E = −jωA − ∇ (∇ · A) ,
ωµϵ
where ω = 2πf
▶ The average power flux density carried by the EM wave is given by
the average Poynting vector:
1
Sav = ℜ {E × H ∗ } (W/m2 )
2
▶ Sav gives the average power per unit area carried by the EM wave

EE259 - Principles of Sensing for Autonomy Radar physics 12

 Radiated power from an antenna
▶ For a radar antenna, Sav gives the radiation power density
▶ Average radiated power Prad is calculated by integrating the normal
component of Sav over a closed surface encompassing the antenna:
I I
1
Prad = Sav · n̂ ds = ℜ {E × H ∗ } · n̂ ds (W)
Σ 2 Σ

n̂: unit vector normal to the closed surface Σ

ds: infinitesimal area of the closed surface Σ

EE259 - Principles of Sensing for Autonomy Radar physics 13

 RF wave attenuation in air
▶ Similar to sound, EM waves get attenuated as they propagate in air
▶ Attenuation peaks due to absorption by water and oxygen molecules
▶ Frequencies used in robotics radar: 20 − 20.25GHz and 76 − 81GHz

EE259 - Principles of Sensing for Autonomy Radar physics 14

 Radar frequency bands
Usage of electromagnetic spectrum is highly regulated by FCC and other
international governing bodies. Specific bands are available for radar.

EE259 - Principles of Sensing for Autonomy Radar physics 15

 Radar frequencies in robotic applications

▶ Two main frequency bands currently used:

– 24 − 24.25 GHz within the IEEE K-band
– 76 − 81 GHz within the IEEE W-band
▶ 76 − 81 GHz is the current state of the art due to clear advantages:
– Larger available bandwidth and better range and angular resolution
– Smaller form factor (∼ 9x smaller aperture for same resolution)
– Higher transmit power allowed (55 dBm vs. 20 dBm for 24 GHz)
▶ 120 − 140 GHz radar is an active area of research, although
frequency band not yet open for industrial use

EE259 - Principles of Sensing for Autonomy Radar physics 16

 Radar principle of operation

▶ TX antenna sends an RF signal to the environment

▶ Roundtrip TOF τ is measured by detecting the echo from a target
▶ Target range (distance) is calculated from TOF: R = cτ /2
▶ Can also measure target bearing (DOA) and relative radial velocity
EE259 - Principles of Sensing for Autonomy Radar physics 17
 Outline

Brief history of radar

Radar physics

Radar antennas

Radar range equation

Radar system architecture

Range and velocity estimation in FMCW radar

Range and velocity estimation in PMCW radar

EE259 - Principles of Sensing for Autonomy Radar antennas 18

 Radar antennas
▶ Antennas are the interface between RF waves and electric currents
in radar circuits
▶ They can act as RF wave transmitters, receivers or both
▶ Different types of antennas used in radars:

▶ Microstrip antennas are most common in radars for autonomy

– Made of copper patches on a PCB substrate (e.g Rogers 3003)
– Fed via traces on the PCB
– Low cost, small size and good performance
– Can easily place a large antenna array on a single PCB
EE259 - Principles of Sensing for Autonomy Radar antennas 19
 Microstrip antenna array for 77GHz automotive radar

EE259 - Principles of Sensing for Autonomy Radar antennas 20

 Antenna regions of radiation
▶ Let D be the largest physical dimension of an antenna radiating at
wavelength λ
▶ Space surrounding the antenna is divided into two main regions:
– Near-field: ∥r∥ < 2D2 /λ, radiated field approx’ed as spherical waves
– Far-field: ∥r∥ > 2D2 /λ, radiated field approximated as plane waves

EE259 - Principles of Sensing for Autonomy Radar antennas 21

 Far-field radiation
▶ Spherical waves locally well approximated by plane waves in far-field
▶ Assuming antenna placed at the origin:
e−j∥k∥∥r∥
Eff (r) ≈ E(θ, ϕ) êE
∥r∥
e−j∥k∥∥r∥ 1
Hff (r) ≈ H(θ, ϕ) êH = r̂ × Eff (r)
∥r∥ η
H : orthogonal unit vectors in (θ̂, ϕ̂) plane, êH = r̂ × êE
êE , êp
η = µ/ϵ: impedance of medium (377Ω in free space)

EE259 - Principles of Sensing for Autonomy Radar antennas 22

 Far-field radiation power density
▶ E and H fields in the far-field are orthogonal
▶ The far-field radiation power density is given by:
1 1
Sav = ℜ {Eff × Hff∗ } = ℜ {Eff × r̂ × Eff∗ }
2 2η
1  ∥Eff ∥2 ∥E(θ, ϕ)∥2
= ℜ ∥Eff ∥2 r̂ = r̂ = r̂ (W/m2 )
2η 2η 2η∥r∥2

EE259 - Principles of Sensing for Autonomy Radar antennas 23

 Far-field radiation

E and H fields have almost zero radial components in the far-field and
are orthogonal to each other:
1
Hff (r) = r̂ × Eff (r),
η
where η is the impedance of the medium:
r
∥Eff (r)∥ ∥E(θ, ϕ)∥ µ
η= = = ≈ 377 Ω (in free space)
∥Hff (r)∥ ∥H(θ, ϕ)∥ ϵ

The far-field radiation power density is given by:

∥Eff ∥2 ∥E(θ, ϕ)∥2

Sav = r̂ = r̂ (W/m2 )
2η 2η∥r∥2

Almost all radars are designed for targets at the far-field.

EE259 - Principles of Sensing for Autonomy Radar antennas 24

 Antenna radiation intensity

▶ Radiation intensity (a.k.a radiation pattern or beam pattern)

U (θ, ϕ) is the power radiated from an antenna per unit solid angle
▶ Is a far-field parameter closely related to the radiation power density:

∥E(θ, ϕ)∥2
U (θ, ϕ) = ∥r∥2 ∥Sav ∥ = (W/sr)
2η
▶ Total radiated power can be found by integrating the radiation
intensity over a full 4π solid angle:
I Z 2π Z π
Prad = U (θ, ϕ) dΩ = U (θ, ϕ) sin θ dθ dϕ (W)
Σ 0 0

Σ: close surface encompassing the antenna

dΩ = sin θ dθ dϕ: infinitesimal element of solid angle

EE259 - Principles of Sensing for Autonomy Radar antennas 25

 Antenna radiation intensity

EE259 - Principles of Sensing for Autonomy Radar antennas 26

 Isotropic antenna
▶ Isotropic antenna has constant radiation intensity for all directions
▶ Perfectly isotropic antennas are not physically realizable

I I
Prad = U0 dΩ = U0 dΩ = 4πU0
Σ Σ
Prad
U0 =
4π
EE259 - Principles of Sensing for Autonomy Radar antennas 27
 Horn antenna
U (θ, ϕ) sometimes described as separate 2D patterns in E- and H-planes

EE259 - Principles of Sensing for Autonomy Radar antennas 28

 Linear dipole antenna

U (θ, ϕ) is a function of the antenna geometry and the wavelength λ

EE259 - Principles of Sensing for Autonomy Radar antennas 29

 Antenna directivity
Directivity: ratio of radiation intensity of the antenna to that of an
isotropic antenna with the same total radiated power:
U (θ, ϕ) 4πU (θ, ϕ)
D(θ, ϕ) = = (dimensionless)
U0 Prad
4πUmax (θ, ϕ)
Dmax = max {D(θ, ϕ)} = (dimensionless)
Prad

EE259 - Principles of Sensing for Autonomy Radar antennas 30

 Antenna radiation efficiency and gain

▶ Radiation efficiency: ratio of total radiated power to power input

to the antenna:
Prad
erad = (dimensionless)
Pin

▶ Gain: ratio of radiation intensity of the antenna to radiation

intensity obtained if antenna input power was radiated isotropically:
4πU (θ, ϕ)
G(θ, ϕ) = = erad D(θ, ϕ) (dimensionless)
Pin
EE259 - Principles of Sensing for Autonomy Radar antennas 31
 Example: 77GHz microstrip antenna, λ ≈ 3.9mm

EE259 - Principles of Sensing for Autonomy Radar antennas 32

 Antenna effective area

▶ Antennas are also used to receive echo EM fields from targets

▶ Antenna effective area describes its power capturing characteristic:

Ps
Aeff = (m2 )
∥Sin ∥

Ps : power delivered to the receiver from antenna terminals (W)

Sin : power flux density of wave incident on antenna (W/m2 )
▶ For an incident plane wave from direction (θ, ϕ), it can be sown that:

λ2
Aeff = G(θ, ϕ)
4π

EE259 - Principles of Sensing for Autonomy Radar antennas 33

 Outline

Brief history of radar

Radar physics

Radar antennas

Radar range equation

Radar system architecture

Range and velocity estimation in FMCW radar

Range and velocity estimation in PMCW radar

EE259 - Principles of Sensing for Autonomy Radar range equation 34

 Radar range equation
▶ Can find max theoretical range of a radar using antenna parameters
▶ Assume monostatic radar with antenna location r′ and target at r
▶ Target bearing (DOA) in antenna frame of reference: (θt , ϕt )
▶ Far away targets scatter incident radar waves isotropically

EE259 - Principles of Sensing for Autonomy Radar range equation 35

 Radar range equation

▶ Magnitude of the radiation power density incident on target is:

U (θt , ϕt ) PTX G(θt , ϕt ) 2

Wi = ∥Sav (R, θt , ϕt )∥ = = (W/m )
R2 4πR2
R = ∥r − r′ ∥: the target range (m)
PTX = Pin : input power to TX antenna by transmitter circuit (W)
▶ Target isotropically scatters back part of the incident power:

Ps = σWi

Ps : Total power echoed back by target (W)

σ: target radar cross section (RCS) (m2 )
▶ RCS is a function of target’s geometry, material and wavelength,
e.g. average RCS of an adult male is 1m2 in the K-band

EE259 - Principles of Sensing for Autonomy Radar range equation 36

 Radar range equation
▶ Since target reradiates Ps isotropically, echo power density received
by the antenna is:
Ps PTX G(θt , ϕt )σ 2
Wr = = (W/m )
4πR2 (4π)2 R4
▶ The echo power received at the antenna terminals is:
PTX Aeff G(θt , ϕt )σ PTX G2 (θt , ϕt )λ2 σ
PRX = Aeff Wr = ≈ (W)
(4π)2 R4 (4π)3 R4
▶ Finally, we include losses. Let L ≤ 1 denote all losses, including
atmospheric attenuation and internal losses of the radar hardware:
PTX G2 (θt , ϕt )λ2 σL
PRX = (W)
(4π)3 R4
▶ Therefore, the target range is given by:
1
PTX G2 (θt , ϕt )λ2 σL 4

R= (m)
(4π)3 PRX
EE259 - Principles of Sensing for Autonomy Radar range equation 37
 Radar range equation

Radar receiver sensitivity is the minimum echo power that the radar
receiver can detect. For a target with RCS σ and DOA (θt , ϕt ) and a
min
radar with receiver sensitivity PRX , the maximum detectable range is
given by the radar range equation:
 41
PTX G2 (θt , ϕt )λ2 σL

Rmax = min
(m)
(4π)3 PRX

Note that the maximum range depends on the fourth root of the trans-
mitted power PTX , e.g. doubling the transmit power will only increase
the maximum range by 19%.

EE259 - Principles of Sensing for Autonomy Radar range equation 38

 Outline

Brief history of radar

Radar physics

Radar antennas

Radar range equation

Radar system architecture

Range and velocity estimation in FMCW radar

Range and velocity estimation in PMCW radar

EE259 - Principles of Sensing for Autonomy Radar system architecture 39

 Radar system block diagram

▶ Older systems did all the processing in analog domain

▶ Bistatic systems use separate TX/RX antennas (no circulator)
▶ Advanced systems use arrays of antennas for improved performance
▶ As usual, complex exponentials (phasors) used for math convenience
EE259 - Principles of Sensing for Autonomy Radar system architecture 40
 Types of radar
Two main types of radar according to the waveforms transmitted:
▶ Pulse radar
– a.k.a pulse-Doppler radar or pulse-echo radar
– Similar to pulse-echo sonar, waveform is amplitude modulated and
repetitively transmits short pulses
– Simplest pulses are sinusoidal waveform with rectangular envelopes
– Advanced pulse waveforms can improve sensitivity and resolution
– Ranging via simple thresholding or matched filtering
– Velocity estimation via measuring the Doppler shift
– DOA estimation via physical scanning or active beamforming
▶ Continuous wave (CW) radar
– Continuously transmits a waveform
– Waveform either frequency or phase modulated (FMCW or PMCW)
– Better resolution than pulse radar, but needs more advanced DSP
– FMCW radar ranging uses frequency chirps, similar to CTFM sonar
– PMCW radar ranging uses PRN codes, similar to GPS
– Velocity estimation via measuring doppler shift
– DOA estimation via physical or digital beamforming (DSP)
EE259 - Principles of Sensing for Autonomy Radar system architecture 41
 Radar waveforms
Typically same waveform is transmitted repeatedly at a fixed rate (PRI)

EE259 - Principles of Sensing for Autonomy Radar system architecture 42

 Types of radar

Two types of radar according to number of antennas:

▶ Single-input single-output (SISO)
– Uses a single TX antenna and a single RX antenna
– DOA estimation requires physically scanning the antenna
– Requires highly directive antennas to achieve good DOA resolution
– Example: airport surveillance radar (ASR) with parabolic antenna
▶ Multiple-input multiple-output (MIMO)
– Uses an array of TX antennas and/or an array of RX antennas
– DOA estimation using beamforming (no physical scanning)
– Antennas can be almost isotropic if array is large
– Example: Automotive radars
Two types of radar according to antenna configuration:
▶ Monostatic: TX/RX antennas are the same or colocated
▶ Bistatic: TX/RX antennas are separated

EE259 - Principles of Sensing for Autonomy Radar system architecture 43

 Radars for modern autonomous systems

Most suitable types of radar for modern autonomous systems are MIMO
FMCW and PMCW radars (a.k.a imaging radars). These radars offer
high range, DOA and velocity resolution. They have a small form factor
with no moving parts and have low power consumption.
Typical performance parameters of a state of the art MIMO FMCW
radar system:
▶ Maximum range 400m. range resolution 10 − 50cm
▶ Field of view 150◦ × 50◦ , DOA resolution 0.5 − 1◦
▶ Maximum velocity ±90mph, velocity resolution 0.5mph
▶ Measurement rate 10 − 50Hz
▶ Power consumption 20 − 30W

EE259 - Principles of Sensing for Autonomy Radar system architecture 44

 A MIMO FMCW radar transceiver
Fully integrated 3 × 4 MIMO FMCW radar trancseiver IC for 76 − 81GHz

EE259 - Principles of Sensing for Autonomy Radar system architecture 45

 Outline

Brief history of radar

Radar physics

Radar antennas

Radar range equation

Radar system architecture

Range and velocity estimation in FMCW radar

Range and velocity estimation in PMCW radar

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 46
 FMCW radar block diagram

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 47
 Doppler effect

▶ Assume radar transmits an

unmodulated CW carrier:

st (t) = ej2πfc t

▶ Received signal would be:

sr (t) = αst (t − τ ) = αst (t − 2R/c)

= αej2πfc (t−2R/c)

▶ If radar and target are moving

relative to each other, range R
becomes time-varying and we get:

sr (t) = αej(2πfc t+ψ(t)) ,

where ψ(t) = −4πfc R(t)/c

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 48
 Doppler effect
▶ Instantaneous frequency of sr (t) given by the derivative of its phase:
1 d 2vr
fr = (2πfc t + ψ(t)) = fc − fc ,
2π dt c
where vr = Ṙ is the relative radial velocity between radar and target
▶ Define the Doppler shift as:
2vr 2vr
fd = −
fc = − (Hz)
c λ
▶ If radar antenna and target are located at r′ and r respectively:
2vr 2Ṙ 2(ṙ − ṙ′ ) · r̂
fd = − fc = − fc = − fc ,
c c c
where r̂ is the unit vector pointing from antenna to the target
▶ The instantaneous frequency of sr (t) would be fr = fc + fd , hence:
sr (t) = αst (t − τ )ej2πfd t
RX signal is attenuated by α, delayed by τ and Doppler shifted by fd .
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 49
 FMCW radar ranging

▶ Assuming SISO radar to simplify notation, MIMO ranging is identical

▶ Transmitted and received signals are linear frequency chirps:
st (t) = ej2π(fc +bt/2)t , 0≤t≤T
j2πfd t
sr (t) = αst (t − τ )e , τ ≤t≤T +τ
b: slew rate (Hz/s)
T : chirp duration (s)
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 50
 FMCW radar ranging

Receiver downconverter output (a.k.a. IF signal) is:

sIF (t) = µs∗r (t)st (t)

= µα exp(−j2π[(fc + b(t − τ )/2)(t − τ ) + fd t]) exp(j2π(fc + bt/2)t)
= µα exp (j2π(bτ − fd )t + j2π(fc − bτ /2)τ )
= µα exp (j(2π∆f t + φ(τ ))) ,

∆f = ft − fd : frequency of IF signal
ft = bτ = 2bR/c: frequency shift proportional to target range
fd = −2Ṙfc /c: frequency shift proportional to target radial velocity
φ(τ ) = 2π(fc − bτ /2)τ : phase of IF signal

▶ Need to estimate IF signal frequency ∆f for ranging (same as

CTFM sonar)

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 51
 FMCW radar ranging

▶ Range and velocity (Doppler) measurements are inherently coupled:

∆f = ft − fd
ft c
R=
2b
fd c fd λ
vr = Ṙ = − =−
2fc 2
▶ In practice, this is often of little consequence since fd ≪ ft
▶ Example: 77GHz radar, b = 30MHz/µs, R = 20m, Ṙ = 45m/s:

ft ≈ 4.0MHz, fd ≈ −23.1kHz (fd /ft < 0.01)

▶ Hence, we assume ∆f ≈ ft and estimate range from the frequency

of IF signal:
∆f c
R=
2b
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 52
 FMCW radar multi target ranging
▶ If antenna radiation patterns are wide, can receive echos of multiple
targets from a single transmitted chirp, for K targets we get:
K
X
sIF (t) = µ αi exp (j(2π∆fi t + φ(τi )))
i=1

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 53
 FMCW radar ranging algorithm
▶ The ranging algorithm is a Fourier transform followed by peak search
▶ Implemented on the digitized IF signal sIF [n] with sampling rate fs

▶ Find target ranges from peak frequency estimates:

∆fˆi c k̂i fs c
R̂i = = , (fs : ADC sampling rate)
2b 2N b
▶ Unbiased range estimator, easy to implement, efficient
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 54
 FMCW radar maximum range vs resolution tradeoff

▶ Minimum range discrimination corresponds to one FFT bin:

fs 1
∆fmin = = , (N = T fs )
N T
▶ The resulting range resolution is:

∆fmin c c c
∆Rmin = = =
2b 2bT 2B
▶ To improve resolution need to increase chirp bandwidth B = bT
▶ Maximum range unambiguously detectable by FFT algorithm is:

∆fmax c fs c
Rmax = = , ∆f ∈ [0, fs )
2b 2b
▶ To increase maximum range need to increase fs or reduce b
Keeping all other parameters constant and increasing the slew rate b
will increase range resolution but decrease maximum range.
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 55
 FMCW radar velocity estimation
▶ The phase of the IF signal, φ(τ ), contains Doppler information
▶ Assume two chirps transmitted and echoed from the same target

▶ Let target have a radial velocity vr relative to antenna

▶ IF signals from the two chirps will give the same frequency estimate
∆fˆ since R changes very little within chirp duration T (10s of µs)
▶ Since bτ ≪ fc (MHz vs. GHz), can approximate IF signal phases:

φ1 = 2π(fc − bτ1 /2)τ1 ≈ 2πfc τ1 , φ2 ≈ 2πfc τ2

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 56
 FMCW radar velocity estimation

▶ Phase difference between the two IF signals is proportional to fd :

4πfc ∆R 2fc ∆R
∆φ = φ2 − φ1 ≈ 2πfc ∆τ = = 2πT ≈ −2πfd T
c cT
▶ vr = Ṙ ≈ ∆R/T is assumed small enough that it doesn’t affect the
frequency estimate of consecutive IF signals, but it affects the phase
▶ Can estimate fd and radial velocity vr from IF phase difference:
∆φ fd λ λ∆φ
fˆd = − ⇒ v̂r = − =
2πT 2 4πT
▶ For vr estimate to be unambiguous, need ∥∆φ∥ < π, so the
maximum unambiguous velocity would be:
λ∥∆φ∥max λ
∥vr ∥max = =
4πT 4T
To increase max detectable velocity, need to reduce chirp duration T .
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 57
 FMCW radar velocity algorithm
▶ vr estimation using ∆φ of two consecutive IF signals is not robust
▶ Solution is to use a frame of M consecutive chirps (e.g. a hundred)

▶ Let sIF [m, n] be the discrete IF signal from mth chirp in the frame
▶ If fd constant within frame, all consecutive IF signals have same ∆φ:
φ0 = φ(τ0 ) = 2πfc τ0 , φm = φ(τm ) = φ0 + m∆φ = φ0 − 2πfd mT
▶ For every m, sIF [m, n] has frequency ∆f and phase φ(τm ):
sIF [m, n] = µαej2π∆f nTs ejφ0 e−j2πfd mT , 0 ≤ n < N, 0 ≤ m < M
where Ts = 1/fs is the ADC sampling interval
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 58
 FMCW radar velocity algorithm
▶ More generally, with echos from K different targets a signal frame is:
K
X
sIF [m, n] = µ αi ej2π∆fi nTs ejφ0i e−j2πfdi mT
i=1

φ0i : Phase of IF signal from ith target at the start of the frame
fdi = −2vri /λ: Doppler shift of the ith target
▶ Target velocities are assumed almost constant within a frame
▶ 1D N -point FFT of sIF [m, n] along n-axis (range-FFT) gives:
K
X
FFTn {sIF [m, n]} = µ αi FFT{ej2π∆fi nTs }ejφ0i e−j2πfdi mT
i=1
K
X
= R̃i [k]e−j2πfdi mT ,
i=1

ˆ
where ∥R̃i [k]∥ = ∥µαi ejφ0i FFT{ej2π∆fi nTs }∥ peaks at k̂i = N ∆ffsi
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 59
 FMCW radar velocity algorithm
▶ A second 1D M -point FFT along m-axis (Doppler-FFT) gives:
SIF [l, k] = FFTm {FFTn {sIF [m, n]}}
K
X K
X
= R̃i [k]FFT{e−j2πfdi mT } = R̃i [k]D̃i [l],
i=1 i=1

where D̃i [l] = FFT{e−j2πfdi mT }

▶ ∥D̃i [l]∥ has a peak at ˆli , proportional to Doppler shift of ith target:
ˆli ˆli ˆli
fˆdi = − fs′ = − =− ,
M MT TF
fs′ = 1/T : effective sampling rate along Doppler m-axis
TF = M T : Frame duration (s)
▶ Hence, the relative radial velocity of the ith target is estimated as:

fˆdi λ ˆli λ ˆli λ

v̂ri = − = = (m/s)
2 2M T 2TF
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 60
 FMCW radar range-velocity map

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 61
 FMCW radar range-velocity estimation algorithm
Range and velocity can be jointly estimated using a 2D M × N FFT on
the raw data matrix, followed by a 2D peak search
K
X
SIF [l, k] = R̃i [k]D̃i [l] = FFTm,n {sIF [m, n]}
i=1

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 62
 FMCW radar maximum velocity and resolution tradeoff

▶ Minimum Doppler shift discrimination corresponds to one FFT bin:

fs′ 1 1
∆fdmin = = =
M MT TF
▶ The resulting velocity resolution is:
∆fdmin λ λ
∆vrmin = =
2 2TF
▶ To improve velocity resolution need to increase frame duration TF
▶ Maximum velocity unambiguously detectable by FFT algorithm is:

∥fdmax ∥λ (f ′ /2)λ λ
∥vrmax ∥ = = s = , fd ∈ [−fs′ /2, fs′ /2)
2 2 4T
▶ To increase maximum velocity need to reduce chirp duration T
Keeping all other parameters constant and decreasing the chirp dura-
tion T will decrease velocity resolution but increase maximum velocity.
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 63
 Realistic FMCW chirp waveform

▶ In reality, chirp waveforms include nonlinarities and distortions

▶ ADC is only turned on during most linear portion of the chirps
▶ An idle time Tid added between chirps to remove blind time
▶ Effective chirp duration becomes Teff = T + Tid , and fs′ = 1/Teff

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 64
 FMCW radar chirp and frame configuration
▶ Modern integrated FMCW transceivers (e.g. AWR2243) have user
configurable chirp and frame parameters, and ADC sampling rates
▶ Different configs for different use cases (e.g. short vs. long range)

Table: Example SISO configurations for short, medium and long rang

Parameter Symbol Units Long Range Medium Range Short Range

Carrier frequency fc GHz 77 77 77

Carrier wavelength λ mm 3.89 3.89 3.89
Chirp slew rate b MHz/µs 10 12 15
Chirp idle time Tid µs 8 10 12
Chirp duration T µs 30 45 50
Chirp bandwidth B MHz 300 540 750
ADC sampling rate fs Ms/s 16.67 11.11 5.00
Samples per chirp N – 500 500 250
Chirps per frame M – 256 128 128
Frame duration TF ms 9.728 7.04 7.94
Frame rate F Hz 102.8 142.0 125.9
Maximum range Rmax m 250 138 62
Range resolution ∆Rmin m 0.5 0.28 0.2
Maximum velocity vrmax mi/hr ±57.24 ±39.55 ±35.08
Velocity resolution ∆vrmin mi/hr 0.45 0.62 0.54

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 65
 Remarks on FMCW range-velocity estimation

▶ Range and velocity estimation are spectral estimation problems

▶ Spectral estimation via FFT is very efficient computationally
▶ More complex spectral estimation algorithms offer better resolution
▶ MIMO range-velocity processing is same as SISO (more on this later)
▶ Actual maximum range for a target of certain RCS is minimum of
Rmax from radar range equation and the Rmax from range-FFT

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in FMCW radar 66
 Outline

Brief history of radar

Radar physics

Radar antennas

Radar range equation

Radar system architecture

Range and velocity estimation in FMCW radar

Range and velocity estimation in PMCW radar

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 67
 PMCW radar block diagram

▶ Principle of operation same as GPS, x[n] is the digital ranging code

▶ Different types of codes used, e.g. MLS, Barker, Gold, etc.
▶ Far less PMCW systems are commercially deployed than FMCW,
hence the system architecture and DSP are not fully standardized
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 68
 PMCW radar ranging

▶ Assuming SISO radar to simplify notation, MIMO ranging is identical

▶ TX repeatedly transmits the same BPSK modulated ranging code
▶ RX signal is attenuated, delayed and Doppler shifted:

st (t) = ej2πfc t s(t) = ej2πfc t ejπx(t) , 0 ≤ t ≤ Tc

j2πfd t
sr (t) = αst (t − τ )e , τ ≤ t ≤ Tc + τ

Tc : codeword duration (s)

x(t) ∈ {0, 1}: ranging code (assume MLS sequence with M chips)
s(t) = ejπx(t) ∈ {+1, −1}: BPSK modulated ranging code
▶ Receiver downconverter output (a.k.a. IF signal) is:

sIF (t) = µsr (t)e−j2πfc t = µαe−j2πfc τ ejπx(t−τ ) ej2πfd t

= µ̃αs(t − τ )ej2πfd t ,

where µ̃ = µe−j2πfc τ
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 69
 PMCW radar ranging

▶ Assume no Doppler shift for now: fd = 0

▶ Reminder: MMSE estimation of τ is by correlation of sIF (t) and s(t)

PMCW ranging is done by calculating the cyclic correlation of the IF

signal and a replica of the transmitted ranging code. The correlator
output will show a peak at the roundtrip delay τ .
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 70
 PMCW radar ranging

▶ Correlations are done in the digital domain

▶ sIF [n] is the digitized IF signal with N samples per chip
▶ s[n] is the digitized BPSK ranging code with N samples per chip
▶ ADC sampling rate is fs = 1/Ts = M N/Tc
▶ Discrete cyclic correlation is given by:
N −1
MX
1
RssIF [n] = sIF [k]s[k − n]mod M N = µ̃αRss [n − n̂τ ],
MN
k=0

where n̂τ = ⌊fs τ ⌋ is the correlation peak index, and Rss [n] is the
ranging code autocorrelation function
▶ Target range estimate is given by:

τ̂ c n̂τ c
R̂ = =
2 2fs

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 71
 PMCW radar multi target ranging
▶ If antenna gain patterns are wide, can receive echos of multiple
targets from a single transmitted code, for P targets we get:
P
X
sIF (t) = µ̃i αi s(t − τi ), (∀i : fdi = 0 for now)
i=1
▶ Since the correlator is a linear filter, output will have P peaks at τi ’s

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 72
 PMCW radar maximum range and resolution

▶ Minimum range discrimination corresponds to one correlation bin:

c cTc
∆Rmin = Rn+1 − Rn = =
2fs 2M N
▶ Most systems downsample the IF signal to 1 sample/chip before
correlation (N = 1):
cTc
∆Rmin =
2M
▶ Need to reduce Tc /M (chip duration) to improve range resolution
▶ Maximum range corresponds to delay equal to codeword duration:
cτmax cTc
Rmax = =
2 2

Need longer codewords (Tc ) to increase maximum range Rmax . For

fixed Rmax , need to increase the number of chips in the codeword (i.e.
faster code rate fcode ) to improve range resolution.
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 73
 PMCW radar velocity estimation

▶ When fdi ̸= 0, can estimate target velocities from Doppler shifts

▶ Unlike GPS, typical Doppler shifts are small enough that ej2πfd t is
almost constant within many consecutive codewords
▶ Example: 77GHz radar, Rmax = 250m, Tc = 1.67µs, vr = 45m/s:

fd ≈ −23.1kHz, 2πfd Tc ≈ 0.24rad

Phase shift due to Doppler within one codeword is 0.24/2π ≈ 4%

▶ No need to demodulate sIF [n] with fd before every correlation
▶ Similar to FMCW, need longer observation windows to estimate fd

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 74
 PMCW radar velocity estimation

▶ Consider a frame of IF signal consisting of K consecutive subframes

▶ Each subframe consists of L consecutive codewords
▶ Subframe duration is Tb = LTc and frame duration is TF = KTb
▶ Doppler signal in subframe k is almost constant at ej2πfd kTb
▶ Let sIF [k, n] be the discrete IF signal from the k th subframe:

sIF [k, n] ≈ µ̃αej2πfd kTb s[n − n̂τ ], 0 ≤ n < LM N, 0 ≤ k < K

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 75
 PMCW radar velocity estimation

▶ More generally, with echos from P different targets we get:

P
X
sIF [k, n] = µ̃i αi ej2πfdi kTb s[n − n̂τi ], 0 ≤ n < LM N
i=1

where µ̃i = µe−j2πfc τi , and n̂τi = ⌊fs τi ⌋

▶ We can average the L IF codewords within one subframe to get:
P
X
s̄IF [k, n] = µ̃i αi ej2πfdi kTb s[n − n̂τi ], 0 ≤ n < MN
i=1

▶ Note that s̄IF [k, n] contains the exact same signal as sIF [k, n] but
will have higher SNR due to noise averaging

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 76
 PMCW radar velocity estimation

▶ The cyclic correlation of s̄IF [k, n] and s[n] will be:

N −1
MX
1
Rss̄IF [k, n] = s̄IF [k, l]s[l − n]M N
MN
l=0
P N −1
MX
1 X
= ej2πfdi kTb µ̃i αi s[l − n̂τi ]s[l − n]M N
MN i=1 l=0
P
X
i
= R̃ss [n]ej2πfdi kTb ,
i=1

i
where R̃ss [n] = µ̃i αi Rss [n − n̂τi ] has a peak at n = n̂τi

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 77
 PMCW radar velocity estimation
▶ Taking a 1D K-point FFT of Rss̄IF [k, n] along k-axis gives:
SIF [l, n] = FFTk {Rss̄IF [k, n]}
P
X P
X
i
= R̃ss [n]FFT{ej2πfdi kTb } = i
R̃ss [n]D̃i [l],
i=1 i=1
j2πfdi kTb
where D̃i [l] = FFT{e }
▶ ∥D̃i [l]∥ has a peak at ˆli , proportional to Doppler shift of ith target:
ˆli ˆli ˆli
fˆdi = fs′ = = ,
K KTb TF
fs′ = 1/Tb : effective sampling rate along Doppler k-axis
TF = KTb : Frame duration (s)
▶ The relative radial velocity of the ith target is estimated as:

fˆdi λ ˆli λ
v̂ri = − =−
2 2TF
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 78
 PMCW radar range-velocity map

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 79
 PMCW radar range-velocity estimation algorithm

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 80
 PMCW radar maximum velocity and resolution tradeoff

▶ Minimum Doppler shift discrimination corresponds to one FFT bin:

fs′ 1 1
∆fdmin = = =
K KTb TF
▶ The resulting velocity resolution is:
∆fdmin λ λ
∆vrmin = =
2 2TF
▶ To improve velocity resolution need to increase frame duration TF
▶ Maximum velocity unambiguously detectable by FFT algorithm is:

∥fdmax ∥λ (f ′ /2)λ λ
∥vrmax ∥ = = s = , fd ∈ [−fs′ /2, fs′ /2)
2 2 4Tb
▶ To increase maximum velocity need to reduce subframe duration Tb
Keeping all other parameters constant and decreasing the subframe
duration Tb will decrease velocity resolution but increase max velocity.
EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 81
 PMCW radar code and frame configuration
Similar to FMCW radar, the code and frame parameters of a PMCW
radar need to be configured differently for different use cases

Table: Example SISO configurations for short, medium and long rang

Parameter Symbol Units Long Range Medium Range Short Range

Carrier frequency fc GHz 77 77 77

Carrier wavelength λ mm 3.89 3.89 3.89
Code rate fcode chip/s 306.39 555.05 616.51
Codeword length M chip 511 511 255
Codeword duration Tc µs 1.67 0.92 0.41
ADC sampling rate fs Ms/s 306.39 555.05 616.51
Samples per chip N – 1 1 1
Codes per subframe L – 22 60 150
Subframes per frame K – 256 128 128
Subframe duration Tb µs 36.74 55.20 61.50
Frame duration TF ms 9.4116 7.0705 7.9415
Frame rate F Hz 106.2 141.4 125.9
Maximum range Rmax m 250 138 62
Range resolution ∆Rmin m 0.49 0.27 0.24
Maximum velocity vrmax mi/hr ±59.21 ±39.39 ±35.36
Velocity resolution ∆vrmin mi/hr 0.45 0.61 0.55

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 82
 FMCW and PMCW radar systems comparison
▶ PMCW and FMCW systems can have the same performance:
– Equal subframe and chirp duration (Tb ≈ T ) for same ∥vrmax ∥
– Equal subframes and chirps per frame (K ≈ M ) for same ∆vrmin
– Tc = N/B for same Rmax
– Chips per code equal to samples per chirp (M ≈ N ) for same ∆Rmin
– 10 − 100 times higher ADC sampling rate for the PMCW system
▶ Advantages of PMCW radar systems:
– Simpler analog circuitry (e.g. waveform generator)
– Advanced DSP and better interference rejection since fully digital
– Sharper range peaks due to matched filtering (vs. FFT for FMCW)
▶ Advantages of FMCW radar systems:
– Cheaper and lower power due to lower ADC sampling rate
– Less DSP resource requirement (processing done partially in analog)

EE259 - Principles of Sensing for Autonomy Range and velocity estimation in PMCW radar 83