Scott Glenn

1
Development of a Triple Nested HF Radar Test Bed
for Current Mapping and Ship Detection
Scott Glenn Josh Kohut Hugh Roarty Rutgers
University Don Barrick Pete Lilliboe Pam
Kung CODAR Ocean Sensors And Many Others .
2
The Operational Research Observatory
Rutgers University Coastal Ocean Observation
Lab The COOLRoom Operations Center
CODAR Network
Cable
Glider Fleet
X-Band
L-Band
3
CODAR Ocean Sensors, Ltd.The Leader in HF
Surface-Wave Radar Ocean Monitoring

• Company principals have been continuously
  involved in HFSWR for 35 years
• Patented CODAR Hallmarks
• Compact antenna system --small footprint offers
  unobtrusive coastal presence
• Unmanned, real-time operation
• Low power -- both radiated and required input
  supply
• GPS-synchronized multiple-radar and multi-static
  operation at same frequency
• Currents mapped to 200 km with Long-Range
  backscatter SeaSondes
• Bistatic augmentation by CODAR demonstrates
  coverage extension to 330 km
• Over 150 CODAR SeaSondes manufactured and sold --
  85 of all HFSWRs
• Systems logged over 4 million operating hours to
  date
• Recent "Dual-Use" objectives being pursued to
  examine ship detection/tracking
• Robust, multiple-look at same target defies
  evasion
• HFSWR is CODAR's only product -- provides us
  unparalleled focus of direction

4
5 MHz
CODAR System Antennas
Receive Antenna
Transmit Antenna
25 MHz and 13 MHz
5
Typical CODAR Remote Site Setup
Transmitter Receiver
6
Sites 7 Long-range (5 MHz) URI, UConn, Mote
Marine, UMaine, USF 1 Medium Range (13 MHz) 2
Standard Range (25 MHz)
7
Ocean.US Surface Current Mapping Initiative 2003
8
Nested CODAR Total Vector Coverage
5 MHz Transmit Buoy
25 MHz Transmit Buoy
9
Bragg Peaks from a Moving Transmitter (4.66 MHz)
10
Bragg Peaks from a Moving Transmitter (4.66 MHz)
11
Past CODAR Vessel Tracking Grants

• Sponsor Office of Naval Research
• Year 2002-2004 Duration 24
  months
• Purpose Initial Large Vessel Tracking Tests for
  a Dual-use HF Radar
• Amount 1,500,000 Status Complete
• Sponsor Counter Narco-Terrorism
• Year 2003 Duration 9
  months
• Purpose Augmentation of ONR Grant for Small
  Vessel Tracking Tests
• Amount 350,000 Status Complete
• Sponsor Norwegian Battle Lab and Experimentation
  (NOBLE)
• Year 2005 Duration 2.5
  months
• Purpose Phase 1 Performance Study
• Dual-use Long-Range CODARs for
  the Northern Coast of Norway
• Amount 90,000 Status Complete

12
CODAR Vessel Tracking Research Team

• Rutgers University
• Scott Glenn, Josh Kohut, Hugh Roarty
• CODAR Test Bed Operations
• CODAR Ocean Sensors
• Don Barrick, Pete Lilliboe, Pam Kung
• CODAR Hardware Development
• Frequency Domain Detections Peak
  Picking
• Applied Mathematics, Inc.
• Bill Browning
• Tracking Application of Submarine
  Trackers to HF Radar
• NorthWest Research Associates
• LJ Nickisch, Sergey Fridman, Mark
  Hausman
• Time Domain Detections SIFTER
• Stanford Research Institute
• Jim Barnum, Eric Simpson

13
Early Objectives

• Test new transmission waveforms
• Multistatic SeaSonde data from the network will
  be collected, however, this initial set of tests
  will focus on analyzing the backscatter data only
• Test different detection algorithms
• Test the sensitivity of the MUSIC direction
  finding algorithm to ideal versus measured beam
  patterns for hard targets
• Define the parameters to optimize dual-use ship
  tracking and current mapping with a Codar
  SeaSonde

14
HF RADAR NETWORK
180 km
RUTGERS
15
Ship Tracking Throughout Rutgers HF Radar Network
16
60 Transits of Oleander in 2004
17
5 MHz Loveladies, NJ
New Jersey Installations Used for Oleander Tests
5 MHz Sandy Hook ,NJ
18
Detection Algorithm

• Simultaneous multiple sliding window
• FFTs in Doppler processing
• Two types of background calculation ---
• median and IIR
• 3D background (Time, Range and
• Doppler) varying with sea echoes
• Thresholding of peaks --- local SNR of
• monopole or at least one of the two
• dipole antennas have to be above the
• threshold
• MUSIC algorithm used to determine
• bearing
• Bearing precision determined by SNR
• (1/sqrt(SNR))

19
Oleander DetectionsFebruary 28, 2005 Outbound
128 point FFT 6 dB Threshold
256 point FFT 7 dB Threshold
512 point FFT 8 dB Threshold
20
Pepper Plots

• All targets detected 9 dB above background

21
Ship Tracking Algorithm

• A Kalman Filter provides a recursive solution to
  the least squares problem.
• Assumptions include linear target motion and
  normally distributed measurement errors.
• Tracker inputs are time radar transmitter and
  receiver positions range, bearing, and range
  rate and range, bearing, and range rate
  uncertainties.
• Tracker outputs are target position velocity
  and estimates of position and velocity
  uncertainties (covariance matrix).
• Target Maneuver Test a statistical test is used
  to estimate whether a combination of two straight
  tracks fit the data better than a single straight
  track.

Oleander Constant Course and Speed Tracker
Solution Using CODAR Detections from 23 November
2002
22
Tx
Rx
23
RESULTS
SMALL BOATS

• SeaTow 41 41 length
• SeaTow 25 25 length

24
SeaTow 41
Loveladies 5 MHz
13 nm
Tuckerton 5 MHz
25
SeaTow 41
Boat Speed 25 knots
26
SeaTow 25
27
SeaTow 25
28
ADCP and CODAR Radial Velocity Comparisons
50 45 40 35 30 25 20 15 10 5 0
Ideal Patterns Measured Patterns
A D C P
RMS Difference (cm/s)
20 40 60 80 100 120 140 160
180 200 220
Bearing (degrees)
29
Pattern Measurement
30
25 MHz System in Clear Environment
31
Simulated Bearing Error with Distorted, Measured
Antenna Patterns Compared with Real Ship Bearing
Data

• Simulated points follow power law
• Compared to
• Ship detections at lower SNRs may differ
  because
• Noise peaks are mis-identified as ships
• "Noise" near peak contains ship signal,
  i.e., it is too high

32
5 13 MHz Receive Antennas in Cluttered
Environment
33
Simulated Bearing Error with Added Bias from
Improper Antenna Pattern Compared with Real Ship
Bearing Data

• For Simulation
• Measured distorted pattern inputted
• Ideal pattern used to recover echo
• Expected power-law fit is offset by 16
• Ship is also offset when inappropriate ideal
  pattern is used

34
Present Focus

• Hardware Improvements
• 13 MHz Codar SeaSonde, December 2004
• 13 MHz Codar Superdirective System, July 2005
• Software Improvements
• SIFTER Algorithm

35
Sandy Hook Facility March 2004Continuous
Operation Since July 2001
36

• Sandy Hook
• Test Bed
• Present Day
• 25 MHz system installed March 2004
• 5 MHz Transmit Antenna moved 1 wavelength back
  from shoreline
• 13 MHz System Installed December 2004

37
25 MHz Tx/Rx
5 MHz Tx
13 MHz Rx
5 MHz Rx
13 MHz Tx
38
(No Transcript)
39
Low Pass
Hook
40
(No Transcript)
41

• Sandy Hook
• Test Bed
• Future
• 13 MHz SuperDirective System Installation July
  2005

42
Does a High-Directivity HF Receive Array Require
a Long Expanse of Half-Wavelength Spaced Elements?

• 4.5 MHz Design
• 5-Element Arrays
• 20-m High Mast
• 19 dB Directivity
• Compare to
• 32-element linear monopole array
• 1 km length at 1/2 wavelength spacing
• 20 dB directive gain

• No!

43
Super-Directivity Design Principles Sacrifice
Unneeded Efficiency for Compact Fractional
Wavelength Spacing

• Pattern depends on number of elements
• Odd number in circle guarantees pattern
  invariance with angle
• Pattern shape/beamwidth is independent of
  frequency
• Efficiency (noise factor) depends on array radius
• External HF noise level sets minimum efficiency
  so that
• Better efficiency does not increase SNR

44
SuperDirective System
25-MHz Heptagonal Array Built and Tested at CODAR
45
SuperDirective Beam Patterns 360 degree coverage

• Blue curve is theoretical pattern for 7-element
  array
• Red results from use of measured transponder
  pattern

46
13-MHz Heptagonal Array Being Built and Tested at
CODAR

• 23-foot (7-m) high mast
• 9-foot (3-m) arms
• 8-foot (2.7-m) dipoles
• 2 masts 21 dB directivity over ground
• -32 dB efficiency

47
SuperDirective System
13-MHz Heptagonal Array Being Built and Tested at
CODAR for deployment at Rutgers

• 23-foot (7-m) high mast
• 2 masts 21 dB directivity over ground
• -32 dB efficiency
• The gain of a directional antenna with the
  footprint of an omni directional antenna

48
Continued testing of 13 MHz

• Utilize AIS signal to ground truth multiple ship
  tracks
• New and Different Targets

49
Automatic Identification System (AIS)

• Required on vessels of
• 300 gross tonnage or greater,
• Length Over All (LOA) over 20 meters, or
• carries more than 50 passengers for hire

50
(No Transcript)
51
GPS Track of RV Hatteras April 5-21, 2005
52
GPS Track of Oceanus April 7-25, 2005
53
Transits of Oleander in 2005
54

• Developed by Mission Research Corporation (MRC)
• Originally developed for ROTHR (Relocatable Over
  The Horizon Radar)
• SIFTER rejects peaks that do not move in a
  consistent way
• SIFTER finds smoothest distribution of
  scatterers that reproduces HFSWR or ROTHR
  measurements
• Targets appear as localized peaks

55
SIFTER Signal Inversion For Target Extraction
and Registration

• New signal processing algorithm to detect weak
  signals in noise and clutter
• Radar returns can be expressed as convolution of
  propagation and radar signal processing with
  distribution of scatterers
• SIFTER uses Tikhonov regularization with
  Newton-Kontorovich method to solve inverse
  problem for distribution of scatterers
  (scattering surface)
• SIFTER incorporates Equation of Continuity to
  enhance scatterers that evolve in time with
  target-like motion
• Clutter is modeled with autoregressive technique
  and removed

56
SIFTER CODAR Signal Model
Time-domain complex signal model
Time-domain complex signal model
Time evolution update equations
AR coefficients from Burgs algorithm applied to
range-guard-banded IQ data
57
SIFTER Results
Cross Spectra after SIFTER
Cross Spectra before SIFTER
58
SIFTER Peaks, Sandy Hook January 17, 2004
Oleander track
59
Norwegian Navy/NOBLE Program for Ship
Surveillance and Environmental Monitoring around
Nordkap

• Current map coverage from proposed Long-Range
  SeaSonde Sites

Russian Border
60
Vessels of Interest
"Moscow" Oil Tanker Length 243 m Height 70
m RCS 46 dBsm 5 dBsm
"Oleander" Type Length 119 m Height 30 m RCS
36 dBsm 4 dBsm
Fishing Trawler 14-m high mast RCS 28 dBsm
61
Standard Long-Range SeaSondeSNR Performance with
Range
62
SuperDirective-Augmented Long-RangeSeaSonde SNR
Performance with Range

• 50 w gt 1000 w
• 13 dB Tx power gain
• 15 dB SuperDirective
• Rx antenna increase

63
HF Surface Wave Radar Performance against Ships
Question How Do Environmental Factors Affect
Performance?

• Graphical Result Is Normalized to "Standard"
  Sea Conditions
• Water salinity is 33 PSU
• Water temperature is 14 C
• Sea-state waveheight is zero (smooth spherical
  sea surface)
• Noise is C.C.I.R. predictions 52 dB above
  internal at 5 MHz
• Curves Show Effect of Typical Environmental
  Changes on Range
• How to Read Curves
• Assume target is visible at range given on
  horizontal axis
• Read from vertical axis range decrease for
  same target visibility

64
Example for Long-Range SeaSonde at 5 MHz Frequency
65
Vessel Masking by Clutter

• FFT/CIT time 128 s
• Vessel RCS
• Moscow 46 dBsm
• Oleander 36 dBsm
• Trawler 28 dBsm
• Standard Long-Range SeaSonde Parameters

66
Backscatter Coverage KJOL
67
Backscatter Coverage SKAG
68
Bistatic Coverage KJOL gt SKAG
69
Findings

• Compact CODAR HF Radars in standard
  configurations can detect both large and small
  vessels while simultaneously mapping surface
  currents and measuring surface waves.
• Peak Picking Detection Algorithms have been
  demonstrated for vessels traveling o utside the
  Bragg Peak in the Doppler Spectrum.
• Range rate is most accurate, then range,
  then bearing.
• Bearing accuracy is inversely proportional
  to the square-root of signal to noise ratio.
• Measured antenna patterns reduce bearing
  bias
• Peak Picking depends on the vessel peak exceeding
  a highly variable background by a specified
  threshold a wide range of sensitivities remain
  to be explored.
• Detection characteristics are consistent with
  expectations derived from the radar equation,
  providing a means to estimate performance
  enhancements associated with system
  modifications.
• The most significant system modification
  super-directive receiver antennas have
  undergone preliminary tests to measure beam
  pattern measurements indicate that the
  configuration can generate and steer a beam.
• Theoretical studies indicate that long-range
  CODARs provide the required coverage to act as a
  dual-use gap filler for Norwegian Security,
  Safety and Environmental needs.
• Association step is currently person-in-the-loop.
  Association in a multi-ship environment remains
  a challenge to be explored.
• Existing submarine tracking experience is
  directly applicable to HF surface vessel
  tracking. Detections from multiple sites
  provide a significant advantage.

70
Future Work

• Thoroughly Test Performance of the Existing
  Monostatic Superdirective System with both Peak
  Picking and SIFTER Detections in a Multi-ship
  Environment
• Construct and Demonstrate the Relocatable
  Monostatic Superdirective System
• Construct and Demonstrate the Multi-static
  Superdirective Testbed
• Develop and Test Multi-static Detection
  Algorithms
• Initiate Association and Tracking Development
  with Existing Datasets in a Multi-ship
  Environment
• Develop Interactive Detection, Association and
  Tracking Package