How a Solid-State Pulse-Doppler Marine Radar Works: The Kelvin Hughes SharpEye

Fabio Romano

Preliminary paper · background reading for the thesis A Modular .NET Gateway for Naval Radar Simulation

1. Introduction

A radar is, at heart, a clock. It sends out a radio pulse, waits for an echo, and turns the delay into a distance. Almost everything that makes a modern radar interesting is about getting a clean echo out of a noisy sea, and then moving the result to a display. The Kelvin Hughes SharpEye, a family of X-band and S-band radars for ships, is a good example, because it differs from the classic marine radar in two respects at once: how the signal is generated and how it is processed.

The paper has three aims. It explains the physical and signal-processing principles behind the SharpEye (sections 2 and 3). It describes how the radar's output reaches the outside world as network data (section 4). And it points out what this means for a software component that has to speak to the radar (section 5). The last point is what the thesis is about.

1.1 Sources and their limits

The sources are the manufacturer's current product description [1], trade-press articles written at the product's launch [2][3], general references on radar signal processing [4][5], and documents on the ASTERIX radar-video format [6][7][8]. Installation handbooks of the SharpEye are circulated on third-party sites, but I could not read them, so nothing here rests on them.

Public material on the SharpEye is largely written by the manufacturer and its distributors. It describes the approach but gives no waveform parameters, pulse repetition frequency, rotation rate or filter design. Where a statement is a general property of radar rather than a SharpEye specification, the text says so. The SharpEye is also a family (the Mk11 and Mk5, among others [10]), and the figures quoted below come from the Mk11 page [1]. They may differ from those of the unit used for the thesis.

2. Radar Fundamentals

2.1 Range, azimuth and the polar picture

The distance R to a target follows from the round-trip time t of the echo and the speed of light c, because the wave travels there and back [5]:

R = c · t / 2

A marine radar adds direction by turning the antenna. At each antenna position (an azimuth) the receiver records the echo strength over time, and so over range. The result is a polar data set: for every azimuth, a row of samples at increasing range. A display converts this to the familiar plan position indicator (PPI) image, with the ship at the centre. This polar structure matters later, because it is also the shape of the data the radar sends over the network.

2.2 Two kinds of resolution

Two echoes can be told apart in range only if they are separated by more than the radar's range resolution. For a simple, unmodulated pulse of duration T this is cT/2, so a shorter pulse sees finer detail. A shorter pulse carries less energy, however, and weaker echoes are harder to detect against noise [4]. This conflict between resolution and detection is the problem that pulse compression resolves (section 3.2).

2.3 The Doppler effect and coherence

A target that moves toward or away from the radar shifts the frequency of its echo. For a carrier frequency f₀ and radial velocity v, the shift is [5]:

f_D = 2 · f₀ · v / c

A pulsed radar rarely measures this shift directly. It observes that, from one pulse to the next, the phase of the echo from a moving target changes by ΔΘ = 4πvΔt/λ, where Δt is the time between pulses and λ the wavelength [5]. Measuring a phase change presupposes that the transmitted pulses have a known and stable phase. A radar with this property is called coherent, and this property is what distinguishes the SharpEye from the magnetron radar.

3. The SharpEye Design

The manufacturer describes the SharpEye as a solid-state, pulse-Doppler, coherent radar built with solid-state electronics throughout [1]. Three ideas in this description carry most of its technical weight.

3.1 A solid-state transmitter instead of a magnetron

Conventional marine radars generate their pulse with a magnetron, a vacuum tube that delivers high peak power in very short pulses. A magnetron is simple and powerful, but its output phase is not controlled from one pulse to the next, so the radar is not coherent and cannot use pulse-to-pulse phase information [5]. A solid-state amplifier driven by a stable oscillator has the opposite profile: a low peak power, but a precisely known waveform and phase.

The published power figures reflect this. The Mk11 is quoted at 300 W for the X-band version and 200 W for the S-band one [1]. A typical marine magnetron peaks at about 30 kW, a figure quoted in a reader comment on a trade article [2]. A radar with a hundred times less peak power still needs enough energy per echo to detect small targets. The signal processing described next supplies the difference.

3.2 Pulse compression

The energy of a pulse is its power multiplied by its duration. If peak power is limited, as it is in a solid-state radar, the way to send more energy is to send a longer pulse. A longer plain pulse would ruin the range resolution (section 2.2). Pulse compression avoids this by modulating the long pulse, for instance with a linear frequency sweep (a chirp) over a bandwidth Δf, and then passing the echo through a matched filter, which correlates it with a copy of the transmitted waveform [4]. After this step the echo becomes a short peak whose width is about 1/Δf, so that:

range resolution = c / (2 · Δf)

The resolution now depends on the bandwidth rather than on the pulse length. The signal-to-noise ratio improves by roughly the time-bandwidth product T·Δf, the compression ratio [4]. The price is range sidelobes, ripples around the compressed peak that could be mistaken for weak targets. They are suppressed by applying a window function, at the cost of a slightly lower peak [4].

The manufacturer states that the SharpEye uses enhanced pulse compression, which gives range discrimination over the whole instrumented range [1], and a launch article describes a low-power, patented pulse sequence [3]. The actual waveform (a chirp, a phase code or something else) is not published. Commenters have noted that pulse compression has long been used in military radar and became affordable for marine use only with modern digital signal processors [2].

3.3 Pulse-Doppler processing and clutter rejection

Sea, rain and land return echoes that can be much stronger than a small boat, and they obscure the picture. A noncoherent radar can only reduce this clutter by lowering its gain, which also removes weak real targets. A coherent radar can use motion. Kelvin Hughes describes its method as extracting the relative motion of targets by measuring the phase of the received echoes relative to the phase of the transmission [2]. This is the pulse-to-pulse phase change of section 2.3, evaluated for each range cell.

The echoes are sorted by their Doppler shift into velocity bands, which separates the wanted targets from clutter [3]. In general, clutter such as rain and slowly moving waves has Doppler content different from that of a vessel under way, so filters can remove it without lowering the gain. The manufacturer states that this removes clutter without degrading the picture and that targets with a radar cross section of 0.5 m² can be detected in cluttered environments [1]. Because the radar itself moves with the ship, the velocities are relative.

Pulse-Doppler processing has intrinsic limits, which are standard radar theory rather than SharpEye data. A pulsed radar samples the phase only once per pulse, so velocities above a certain value are ambiguous: they appear as a different, lower speed. The speeds at which a target looks stationary to the filter, and is therefore lost, are the blind speeds. They are usually avoided by varying the pulse repetition frequency [5]. The SharpEye's handling of these effects is not public.

3.4 Tactical features and hardware

The Mk11 page lists several operating modes that follow from the programmable waveform: selectable frequency, low-power operation, a low-probability-of-intercept mode that reduces how easily the platform is detected, sector transmission and blanking, which limit the angles in which the radar transmits [1]. The antenna is a combined turning unit with a direct-drive motor, optionally paired with a second antenna for IFF (identification friend or foe). The housing is carbon fibre [1]. Early versions worked with existing Kelvin Hughes S-band antennas [2].

4. Data Interfaces

4.1 From analogue video to a network stream

Traditionally a radar passed its video to the display as an analogue signal, accompanied by a trigger marking range zero and pulses that indicate the antenna's rotation. A digital radar then had to convert its signal to analogue only for the display to digitise it again, adding noise and cost [6]. The SharpEye avoids this: the manufacturer states that it outputs radar video and control over Gigabit Ethernet, with full built-in test (BITE) and status reporting, and an open architecture for connection to combat management systems and cameras [1].

4.2 ASTERIX and CAT-240

ASTERIX is a family of message standards from European air traffic control. Each family member, called a category (CAT), defines one kind of message. Category 240 covers radar video. Its specification dates from 2009 and has been adopted by a number of radar manufacturers [6][8]. A 2013 paper by a vendor of radar interfaces lists the SharpEye series among them [9]. Nothing in the standard ties it to air traffic: it is now used for naval, security and vessel traffic radars [6].

CAT-240 carries the polar structure of section 2.1 directly. Each return is a header followed by a data block. As documented in one manufacturer's implementation [7], the header gives the start and end azimuth of the block as 16-bit angles, the starting range bin, the duration of a range cell, the number of bits per cell and a compression flag, and the size of the data. The cell duration fixes the range scale, since a cell of duration τ covers cτ/2 of range. A timestamp and a manufacturer-specific field follow.

Three practical properties of the format deserve attention.

• High message rate. One message holds one azimuth, so the rate equals the azimuths per rotation multiplied by the rotation rate. One implementation gives 400 azimuths at 4 rotations per second, which is 1,600 messages per second [7]. A general example in the literature has 1,000 messages per second of 4,000 samples each [6]. These are illustrations of the format, not SharpEye figures.
• Transport. The standard does not prescribe one, and in practice the messages are sent as UDP packets, unicast or multicast [7]. UDP gives no delivery or ordering guarantee, which is acceptable for a picture refreshed on every rotation.
• Compression is optional and unspecified. The standard allows compressed video without saying how. Implementations either send raw data or use a common scheme such as ZLIB. Uncompressed video has a constant bandwidth, whereas the compressed size grows and shrinks with clutter, so a network that works in calm conditions may overload in heavy clutter [6].

4.3 What the public documents do not say

CAT-240 describes the video that the radar sends. The messages that report the state of the antenna, and the messages with which a display controls it, are not covered by the public sources, and I found no public description of them for the SharpEye. A system that has to reproduce the dialogue between a radar and its antenna therefore depends on the manufacturer's documentation and on observing the real traffic.

5. Implications for Software Integration

Several consequences follow for a program that has to exchange data with such a radar, such as the gateway in the thesis.

• Throughput and ordering. Video arrives as a continuous stream of hundreds or thousands of small messages per second, in rotation order. A receiver has to keep up and preserve the order, which favours queues that are separate from the processing of each message.
• Different delivery needs. Video tolerates the occasional loss and suits UDP. State and control messages are infrequent, but a lost control command is a real fault, which is a reason to carry them over a reliable transport.
• Polar geometry on both sides. A simulation that produces radar-like returns, for instance by casting rays from the antenna and recording collisions, has to express them as azimuth and range, because that is what the radar format represents.
• The documentation may not match the device. Because the public material is partial and written for specific models, the reliable reference is the traffic of the actual equipment.

The thesis develops these points for a concrete case. It follows the same order: the equipment and its messages first, then the design of a gateway that is configurable enough for the set of converted messages to change.

Continue with the thesis

References

1. [1] Hensoldt. Kelvin Hughes Mk11 SharpEye: combined navigation and naval surveillance radar. Product page. hensoldt.net
2. [2] Panbo. SharpEye radar: look, ma, no magnetron. Trade article and reader comments. panbo.com
3. [3] Maritime Executive, 12 October 2006. Article on the launch of the SharpEye radar. maritime-executive.com
4. [4] Wikipedia. Pulse compression. en.wikipedia.org
5. [5] Wikipedia. Pulse-Doppler radar. en.wikipedia.org
6. [6] D. G. Johnson, Cambridge Pixel. ASTERIX CAT-240 for Radar Video Distribution: Theory and Practice. White paper, March 2022. cambridgepixel.com
7. [7] Navtech Radar. ASTERIX CAT-240 Protocol for Video Data. Documentation of one manufacturer's implementation. navtechradar.atlassian.net
8. [8] EUROCONTROL. CAT240: Specification for Surveillance Data Exchange, ASTERIX. eurocontrol.int
9. [9] Cambridge Pixel. Paper presented at Radar India, 2013 (cited from its listing in search results; I could not open the document). cambridgepixel.com
10. [10] Hensoldt. Kelvin Hughes Mk5 SharpEye: radar technology for smaller vessels. Product page. hensoldt.net