Author’s Name Name of the Paper Session

DYNAMIC POSITIONING CONFERENCE

November 15-16, 2005

Sensors 1

RadaScan: A Local Reference, High Resolution Radar,

Dynamic Positioning Sensor

Dr. Dominic Pearce

Guidance Navigation (Leicester, UK)

Return to Session Directory

D. Pearce, Guidance Navigation RadaScan: A local reference, radar DP sensor

ABSTRACT:
This paper details the concept and processing principles behind RadaScan, a local reference radar sensor,
capable of relaying accurate navigation to dynamic positioning systems. RadaScan represents a next
generation in Sensor technology. It is a swept radar system capable of tracking multiple targets and
providing relative position and heading feedback at 3Hz to a host system.

An overview of the design principles of the sensor and target system is presented. It considers the
technology behind the sensor; a FMCW microwave radar system illuminating a retro-reflective target
capable of introducing a unique identification code by modulation of the radar carrier. This design allows
the radar to easily detect its targets even within a cluttered offshore working environment, dominated by
large installation structures.

A description of the digital signal processing flow that interprets the raw radar return to an accurate
position and heading estimate is presented. Consideration is given to the DSP hardware requirements to
achieve the data throughput whilst working within cost and space constraints. The use of FPGA technology
for front end real-time radar signal processing is demonstrated to be an optimum solution for the DSP needs
of the system. Using an FPGA, continuous radar data processing rates of 50MB/s are achieved across 4
input channels. This data throughput allows the system to process incoming targets at a rate of 10Hz.

The paper further considers the algorithms developed for automatic detection of targets using Constant
False Alarm Rate (CFAR) principles applied in the frequency domain, and the use of target tracking filters,
originally developed for ballistic missile radar trackers, to control hardware feedback loops.

The paper concludes by presenting position data from the sensor, demonstrating the accuracy and overall
specification that can be achieved for a local reference position and heading fix using RadaScan.

Dynamic Positioning Conference November 15-16, 2005 Page 1

D. Pearce, Guidance Navigation RadaScan: A local reference, radar DP sensor

Introduction
RadaScan represents a next generation in local reference sensor technology. It has been specifically
designed to overcome some of the shortcomings of traditional local position reference systems. Being a
microwave radar sensor, it is insensitive to the harsh environmental conditions often experienced in an
offshore environment, being largely unaffected by heavy rain, fog, or other prevailing sea conditions.
RadaScan is a compact ship mounted swept microwave radar system which can interrogate radar retro-
reflective targets 1. It can provide an accurate local position reference to a ship’s dynamic positioning
system from multiples of these targets located on a nearby structure, such as an offshore installation. The
sensor is capable of finding and predicatively tracking targets out to 750m and beyond. The system can be
used as a standalone ‘black box’ sensor, or driven from a graphical user interface identical in appearance
and usability to Guidance Navigation’s well established CyScan user console. The main sensor is a
frequency modulated constant wave 2, 3 (FMCW) mono-pulse 4 radar operating over a 100MHz bandwidth
within the maritime radiolocation band centred at 9.25GHz. The targets are totally passive, requiring a
battery to drive only the modulation electronics that impart a unique code on the reflected radar return, with
no implicit amplification of the microwave signal.

Radar Description
A picture of the complete RadaScan sensor is shown in Figure 1:

Figure 1: The RadaScan sensor with radome cover removed

The radar front end schematic is shown in Figure 2. The transmitter consists of a ramp generator, the rate of
which is driven by the DSP system. This ramp signal is used to frequency modulate the microwave source.
The output of the source is passed through an attenuator and then through a power amplifier before being
split to feed the two squints of a mono-pulse transmit horn, illuminating a large antenna. The same horn is
used to receive the two squints returning from the radar target. To improve the isolation between transmit
and receive channels, the target imparts a polarisation change to the incoming radar signal. Using this
method, it has been possible to achieve 40dB isolation between receive and transmit chains. The two
received squints pass into a hybrid T where they are summed and differenced (the summation = sigma (Σ),
the difference = delta (∆)). Both channels pass through low noise RF amplifiers before being mixed with
the transmit signal, to reduce the signal to IF or base band. The signals then pass through target specific
filters to remove background clutter before being amplified in preparation for analogue to digital
conversion by the processing system.

Dynamic Positioning Conference November 15-16, 2005 Page 2

D. Pearce, Guidance Navigation RadaScan: A local reference, radar DP sensor

TX 1

FM Ramp
Transmitter Modulator

TX 2

Target BaseBand
RX 1
LNA Mixer Filters Amplifiers

Σ Hybrid
T

∆ Target BaseBand
RX 2 LNA Mixer
Filters Amplifiers

Figure 2: Radar schematic.

Antenna properties.
The radar antenna and horn arrangement controls the beam shape of the emitted radar wave-front. The
antenna also imparts considerable gain to the system. The antenna has been designed specifically to
optimise the radar beam shape for its intended application, specifically a wide elevation (or E-plane)
pattern to allow the system to cope with the pitch and roll of a vessel, and to sight targets high above the
radar at close range. Similarly, a tight azimuth (or H-plane) pattern to ensure good bearing accuracy. The
measured results for the reflector patterns are shown in figures 3a & 3b:

Figure 3a: Reflector elevation gain plot. Figure 3b: Reflector azimuth gain plot

The beam shape can be defined in terms of its angle between half power points (-3dB). In elevation, the
beam is 15 degrees wide, and in azimuth, 2.4 degrees.

Retro-reflective Target Properties

The target has been uniquely developed in conjunction with the radar sensor to ensure that the system as a
whole is easy to use and setup. The targets are true retro-reflectors, i.e., a signal received is reflected back

Dynamic Positioning Conference November 15-16, 2005 Page 3

D. Pearce, Guidance Navigation RadaScan: A local reference, radar DP sensor

to its point of origin. At the same time, the target flips the polarisation of the reflection by 90 degrees whilst
imparting an identifying modulation code. This code is used by the radar processing system to not only
identify the target, but to reject the background clutter normally encountered by any marine radar system.
This includes the signature of the large installation where the target is placed, and clutter from other
sources, such as other vessels.

For ease of installation, two factors have been important in their design:
Firstly, the targets are passive, i.e. they do not amplify the incoming radar signal. They only require an in-
built battery to drive the modulation electronics that impart the targets identification code. This battery has
a 2 years continuous use lifespan. The passive nature of the target means that no fixed installation or power
source is required. Targets can be placed and removed as and when required.
Secondly, the targets have a very wide viewing or acceptance angle. The measured near field response of
the target is illustrated in figure 4.

Radiation Patterns for 9.25GHz Sidebands

-70
-90 -75 -60 -45 -30 -15 0 15 30 45 60 75 90

-75

-80
Recieved Power (dB)

1.75MHz
AzimuthElevation
T1
1.75MHz
AzimuthAzimuth
T2
2.5MHz Elevation
Elevation T1
-85
Elevation
2.5MHz T2
Azimuth
Azi Drop
Azimuth Dropoff
Elev Drop
Elevation Dropoff
-90

-95

-100
Angle from Boresight (Degrees)

Figure 4: Near field target reflection response, T1 = target id 1, T2 = target id 2, Azi Drop = azimuth half
power points, Elev Drop = elevation half power points (3dB down).

The acceptance or viewing angle of the target can be defined using the angle between the half power
response points either side of boresight ( in the same way as defined for the antenna beam shape). In
elevation the target acceptance angle is 47 degrees, whilst in azimuth, 90 degrees. The view angle however
depends on the range. For instance at close range, the viewing angle is much greater since the system has
considerable spare receiver gain, and the antenna has a wide elevation pattern. In trials, at 50m range, the
radar was able to view up to 85 degrees from boresight in azimuth (170 degrees total viewing angle). At
25m range, the radar was able to view the target at 35 degrees from boresight in elevation (70 degrees total
viewing angle). For instance, in a typical DP working environment, this would allow a vessel to ‘look up’
at targets on an installation, some 17.5m above the position of the radar on the vessel.

Dynamic Positioning Conference November 15-16, 2005 Page 4

D. Pearce, Guidance Navigation RadaScan: A local reference, radar DP sensor

Radar Digital Signal Processing Hardware

A single platform provides all the radar processing hardware requirements, as well as the client server
interface. The system is diagrammatically described in figure 5.

128/256 Compact
RS232 / MB Flash &
UART 422 /485 soldered IDE

Geode
VGA
USB SC2200 Display
266 MHz 586

32-bit
PCI Bus
Ethernet PMC+
RJ45
J4 to FPGA

Lowcost
AD9244
SharcFIN
(ADC)
4-ch (MS2)
14 Flags /Irqs
ext
Links LCHH
4-ch Sports (21161) Flash
DAC 2M x 8
AD5324 32 32 8 (BMS/MS3)
8-ch
ext
Xilinx Exp. Bus 32
Direct FPGA
FPGA I/O 32
SBSRAM SDRAM
1-wire 512kW 64 MB
32 32
Thermometer (MS1) (MS0)

PLL
HH Bus
Interface
32-bit
Isolation
DIO

Figure 5: RadaScan processing hardware schematic

The board is split in half:

In the ‘DSP’ section, the board provides four analogue to digital converters, sampling incoming radar
signals with 14-bit resolution at 65 MWords/s, controlled by, and linked directly to a Xilinx Spartan 3
FPGA. The FPGA provides a platform for control of external radar hardware via isolated 32-bit digital I/O
and an 8 channel DAC. The FPGA also interfaces directly to four Analog Devices 21161 DSP processors
via high speed link ports. The DSP processor cluster shares access to a large block of fast access SBSRAM
and a further block of SDRAM. A block of Flash memory is provided storing the runtime image of the
FPGA and DSP executables.
The ‘Embedded PC’ section comprises a Geode ‘thin client’ processor, a compact flash hard drive
providing windows XP embedded, 256MB memory, dual serial and Ethernet ports. The geode has access to
the DSP processors internal memory, flash, SBSRAM, FPGA registers and SDRAM via its PCI interface.
The embedded PC hosts the RadaScan network server allowing multiple users to connect to the unit
simultaneously. The DSP cluster alone provides 1.2GFLOPS of processing power at 32bit resolution. All
this hardware is provided in a small area board, compliant with the microATX standard.

Dynamic Positioning Conference November 15-16, 2005 Page 5

D. Pearce, Guidance Navigation RadaScan: A local reference, radar DP sensor

Radar Processing Flow

The front end digital signal processing executed by the FPGA is shown in Figure 6. The FPGA architecture
and parallel processing capability lends itself very well to high speed, large volume data processing. The
FPGA performs target filtering and digital down conversion duties. It can process up to 50 Mbytes /second
in real time, allowing the RadaScan system to interpret targets at rates greater than 10Hz. This system
redundancy means the sensor can easily maintain a 3Hz rotation speed, and target refresh rate, processing
multiple targets per revolution.
Legend
Sigma T1
Delta T1
Target de- Sigma T2
Target modulation Delta T2
Filters
ADC

Target Triggers

Phase
sensitive 2- Stage
detector Downsample

Encoder

DSP
Encoder CLUSTER
Processor High Speed Link Ports
Marker & Interrupts

Figure 6: FPGA DSP schematic for front end radar processing.

Data delivered to the DSP cluster is used to determine the range and bearing to the target, and to control
radar hardware using feedback from the target range and bearing trackers.
The range is determined in the frequency domain using an adaptive threshold CFAR 5 (Constant False
Alarm Rate) algorithm to identify the target signal in the spectrum above a certain signal to noise ratio
threshold. An 8192 pt FFT divides the spectrum in to a series of range bins which are tested sequentially
for a target ‘hit’. Once a target is detected, the CFAR window is narrowed to maximise the signal to noise
ratio, and is continuously manoeuvred by the target range tracker, following the targets range signature.
The range tracker also controls the source ramp rate to ensure that the target range always falls within the
radars optimum detection window or range bin.
The bearing is determined by a process model fit to the measured target sigma envelope. This method
provides accuracy far exceeding a traditional radar system.

Target Tracking
Both range and bearing are tracked using fading memory polynomial (FMP) trackers 6. These are one step
predictors providing feedback to control the radar hardware. The trackers are optimally parameterised to
track targets in an environment with the velocity and acceleration characteristics of a typical offshore DP
equipped vessel. The range tracker controls the source ramp rate and CFAR detection window, whilst the
bearing tracker controls the radar’s data acquisition window for that target.
Figures 7a & b display range and bearing plots for a target at ~187m.

Dynamic Positioning Conference November 15-16, 2005 Page 6

D. Pearce, Guidance Navigation RadaScan: A local reference, radar DP sensor

Figure 7a: Range Plot Figure 7b: Bearing Plot

Sensor Accuracy & System Specification

The results from extensive trials have been used to define the following system specification:
Range Accuracy: better than 0.1% of range.
Operating Range: 10 to 750m
Angular repeatability: STD 0.3 mrads @ 200, 1.0 mrads @ 500m
Beam Shape: elevation: ± 7.5 degrees @ -3dB gain
azimuth: ± 1.2 degrees @ -3dB gain
Max Elevation at Close Range: +35 degrees @ 25m.

References
1. Briggs, J.N. Target Detection by Marine Radar, IEE Radar, Sonar Navigation series 16.
2. Skolnik, M.I. Introduction to Radar Systems, Third Edition.
3. Komarov, I.V. & Smolskiy, S.M. Fundamentals of Short-Range FM Radar
4. Sherman, S.M, Monopulse Principles & Techniques.
5. Nitzberg, R. Radar Signal Processing & Adaptive Systems
6. Morrison M. Introduction to Sequential Smoothing & Prediction.

Dynamic Positioning Conference November 15-16, 2005 Page 7