Radar Modality Handbook

The initial version of the DIRSIG radar capability were created during the developement of 4.2 releases as a stand-alone application that leveraged the DIRSIG core libraries. During the 4.3 development cycle this proof of concept tool was integrated within the main DIRSIG application and internal experimentation continued. At this time, the capability was still an unfunded project being conducted during off hours. During the 4.4 releases a set of minimal demos were created in an effort to obtaining funding for focused development. Closer to the end of the 4.5 release cycle the first version of this document was produced in an effort to help users play with the tool and provide feedback. The 4.6 release brought improvements to the scene and platform descriptions and a major refreshment of the demos.

This application area is currently ranked as experimental due to the lack of some key features and lack of extensive use by RIT and the user community.

The existence of these limitation is primarily a result of funded research priorities. Our past and current research has not placed a priority on resolving this current set of limitations. As researchers, we are always seeking opportunities to to address known limitations if funding is available.

A radar typically emits pulses at a constant pulse repetition frequency (PRF). This might also be described in terms of the pulse repetition interval (PRI) or the pulse repetition time (PRT). The length of a pulse is referred to as the pulse width (PW) in seconds. Each pulse last for a short fraction of the period between pulses, and the duty cycle refers to the fraction of the total time during which the transmitter is emitting energy. For example, if the pulse repetition interval is 100 microseconds and each pulse is 1 microsecond long, then the duty cycle is 1%. Most temporal pulse profiles are approximated as a RECT function (infinite slope on the rise and fall), but are less ideal in reality.

The spectral distribution of the emitted pulse is typically characterized in terms of a central carrier frequency and a spectral bandwidth about that frequency. Most airborne and space-borne imaging real aperture radar (RAR) and synthetic aperture radar (SAR) systems operate in the X Band or the 8.0 - 12.5 GHz (2.4 - 3.75 cm wavelength) range. Like the temporal pulse width, the spectral bandwidth is usually approximated as a RECT function, but the actual frequency envelope is usually less ideal.

This section will eventually describe the receiver in greater detail. For now, the receiver is assumed to have the same angular sensitivity as the transmitter. The key receiver characteristic is the range gate (temporally listening window).

Similar to the approach employed for the LIDAR Modality, we employ a two stage approach for RADAR:

Note that the primary output of DIRSIG is the raw, complex return data. Although it is our plan to distribute a minimal image focusing tool, the user is required to employ their own algorithm to convert the raw, complex return data into imagery (example Matlab code is included in the demos).

Within the radiometry stage, you can run the model in either a single-pass or two-pass mode. In a single-pass radiometry simulation, the transmit and receive computations are computed during a single DIRSIG simulation. The output of the simulation is the complex return. In a two-pass radiometry simulation, the user runs DIRSIG once to transmit the pulses into the scene and then runs a second simulation to collect the pulses onto the receiver. The two simulations are coupled via a storage file (see the twopassmode and nodefilename variables discussed later).

The benefits of the two-pass method is that the coherent hit map created in the first pass and then used in the second pass results in dramatic reduction in the numerical noise. The noise in the single-pass method arises from random sampling that is different from pulse to pulse. This introduces variations in randomly varying distances to stationary objects between pulses, and manifests itself as random variations in coherence lengths.

The following model-based atmospheric models can (should) be discussed the future:

Due to the index of refraction varying vertically and horizontally, the atmosphere can also bend the path of the beam. These effects are not currently accounted for in the SAR calculations, although these calculations have been incorporated into the LIDAR beam propagation.

A specific reflectivity model was introduced for use exclusively with radar simulations. Although most radar collections usually involve a nearly mono-static transmit and receive geometry, this reflectivity model predicts the bi-directional reflectivity of a surface via a 5-parameter formulation:

\( \mathrm{BRDF}( \theta_i, \theta_o, \Delta \phi ) = d + s \cdot \mathrm{MM}( \theta_n, n, k ) \cdot e^{\frac{-\tan^2 \theta}{2 \sigma^2}} \)

where s and d are the specular and diffuse reflectivity coefficients, respectively, sigma is the surface roughness and n and k is the complex index of refraction which drives the computation of the polarized MM (Mueller Matrix) for the specular component.

In DIRSIG 4.6.x and later, the reflectivity model coefficients are stored in the standard material database file associated with the scene.

The excerpt below shows an example MATERIAL_ENTRY from a DIRSIG material database file:

Currently, a radar simulation does not use the same radiometry solver mechanism that the passive and active EO/IR simulations leverage. However, DIRSIG requires that all materials have a radiometry solver assigned to them. The Null radiometry solver allows the material description for a radar simulation to be simplified by specifying a solver that performs no calculations (because it will never be called within the context of a radar simulation).

The transmitter is described via a decomposition of the major characteristics:

The receiver is described via a decomposition of the major characteristics:

The following options are specific to a radar simulation:

The following is an example of an <instrument> section of a .platform file compatible with DIRSIG 4.6.x:

This example employs the "basic radar" (non-polarized) instrument type. The transmitter emits pulses at a 10 kHz rate with 1.0e+09 Watts (1 GW) continuous power. The pulses have a 9.6 GHz carrier wavelength (note that the 0.001 spectral bandwidth is ignored at this time, but the <width> is necessary). The pulses are linear frequency modulated with a 2.506e+13 Hz/sec chirp rate. The beam shape is ideal (uniform) with a 0.01 x 0.01 radian (or 10 x 10 mrad) angular width. The transmitted pulse is split up into 250,000 energy packets which are followed for a maximum of 3 bounces. The pulse-to-pulse coherence hit map employs nodes that will be no closer than 0.5 meters. The receiver is listening for pulse returns between 1.335e-04 and 1.495e-04 seconds after a pulse is fired. The returned signal is digitized with 2.0e-08 second sampling (50 MHz), which results in 801 (inclusive) range bins. The output phase history filename will be sar.img and the the carrier frequency is left mixed in the data (because stripcarrier is set to 0). The dimensions of the output phase history file will be 801 x N pixels, where N is the number of pulses shot. This will be determined by the duration of the collection task and the 10 kHz pulse rate defined here (for example, a 1 second collection will result in 10000 pulses and the image will be 801 x 10000).

In contrast to most EO/IR imaging systems, most radar systems are not oriented to point directly down (nadir) at the scene. Instead, they employ some sort of across-track look angle. This is is handled by orienting the platform, electronic beam steering (via phasing) or a combination of the two. The DIRSIG model does not have have an interface to model a phased array in order to steer the beam at this time. Therefore, the user must rely on platform orientation or platform-relative pointing to direct the beam.

The DIRSIG platform coordinate system has the +Y axis as the nominal along-track axis. Therefore, across-track pointing is rotation about the along-track or Y axis.

There are multiple ways to define the across-track pointing angle:

Although the final option works just as well as the others, it is less favored because it conceptually implies that the aircraft is flying with a constant roll.

The transmit simulation can be run using the following command-line syntax:

The radar.mode option is what enables the two-pass mode. In this case, the option was assigned the transmit value to trigger the transmit mode calculations.

The primary output of the transmit simulation is the hit_nodes.dat file, which is a special file that records transmit energy packet hit points within the scene. These are then used in the receive simulation to compute coherent return intensities.

The receive simulation can be run using the following command-line syntax:

The output of the receive simulation is a complex phase history, which is described below in detail.

The output of a DIRSIG radar simulation is the complex signal received by the antenna. This is stored into an output file that is organized as a single band "image" (complete with an ENVI header to visualize it):

For more information on the structure of the output binary data file, look at the DIRSIG image file format document.

The complex phase history produced by a DIRSIG SAR simulation must be "focused" into an image. For traditional imaging SAR applications, the focusing algorithm reconciles the returns from all ranges across all pulses to produce an 2D intensity map for the area that was mapped. The details of SAR image focusing algorithms are beyond the scope of this manual but many detailed resources are available. At this time it is expected that most users will be using their own focusing algorithms with the complex phase histories generated by the DIRSIG model. Those algorithms will need access to information that is contained in some of the DIRSIG input/output files:

Some of the DIRSIG demos include a Matlab focusing algorithm based on the work of Gorham and Moore. The example below shows an array of corner (trihedral) reflectors and an animation of the focused intensity image as the algorithm incorporates new intensity vs. range information from each pulse.

This section outlines the short-term, technical tasks required to get this capability out of the "experimental" phase.

Once the current modeling capabilities have been cleaned up and stabilized, the next step involves the addition of an advanced, data-driven mechanism that would allow users to model advanced systems. The primary goal of this effort would be providing a mechanism to drive the transmitter on a pulse by pulse basis. This would allow the user to describe the beam shape, carrier frequency, chirp segments, etc. on a per-pulse basis. This would allow the end user to model nearly arbitrary modulation schemes including step chirp, binary-phase coding, poly-phase coding, etc. Furthermore, allowing for multiple transmitters and per-transmitter pulse descriptions like those previously described would allow phased array systems (beam steering) to be directly modeled (currently the user can have only a single transmitter and beam steering must be accomplished with platform orientation or platform-relative pointing).