Technology Demonstration of Ka-band Digitally-Beam formed Radar for Ice Topography Mapping

GLISTIN (Glacier and Land Ice Surface Topography Interferometer) is a spaceborne interferometric synthetic aperture radar for topographic mapping of ice sheets and glaciers. GLISTIN will collect ice topography measurements over a wide swath with sub-seasonal repeat intervals using a Ka-Band digitally-beamformed antenna. This paper will give an overview of the system design and key technology demonstrations including a Im x Im digitally-beamformed Ka-band waveguide slot antenna with integrated digital receivers. We will also detail the experimental scenario that we will use to demonstrate both the beamforming and interferometric performance of this system.


INTRODUCTION
This paper describes technology being developed to support a novel Ka-band (35 GHz) radar for mapping the surface topography of glaciers and ice sheets. During our three year program (funded by NASA ESTO Instrument Incubator Program), we are developing instrument and mission designs and are demonstrating a Im x Im Ka-band digitally beamformed (DBF) slotted-waveguide array antenna. This DBF array is composed of slotted-waveguide receiving elements, 16 Ka-band down-converters and digital receivers. The use of digital beamforming in the elevation (cross-track) direction yields a wide-swath with a sufficient number of looks but does not require nearly as much power as a wide-beam antenna would.

GLISTIN INSTRUMENT DESIGN
The proposed instrument, the "Glacier and Land Ice Surface Topography Interferometer" (GLISTIN) and depicted in Figure 1, will map the surface topography of glaciers and ice sheets at high spatial resolution, high vertical accuracy, independent of cloud cover, with a swath-width of 70 km. The system is a single-pass, single platform interferometric synthetic aperture radar (InSAR) with an 8.6 mm wavelength, which minimizes snow penetration yet remains relatively impervious to atmospheric attenuation. In contrast to lidars, the instrument will be insensitive to clouds, provide significant swath-widths, cover the poles submonthly, and provide inherently variable spatial resolution: high spatial resolution for meter-scale vertical precision on glaciers and coastal regions and coarse spatial resolution for decimeter accuracy on featureless ice sheet interiors. Previous attempts at using millimeter-wave InSAR faced fundamental problems, including limited swath widths and high transmitted power requirements.
Our approach overcomes those two major limitations by applying digital beamforming techniques to standard InSAR. Both Ka-band antennas and digital beam forming have been used in various ground-based and airborne applications, however, no antenna systems exist that are substantially similar to the proposed system. To date, no civilian spaceborne imaging InSAR system has utilized Kaband. There has also never been, to our knowledge, any digital beam forming radar flown in space. This technology has no alternatives when high resolution and swath is required other than the use of extremely high power transmitters that are impractical from both a technological and power consumption standpoint The use of DBF achieves greater than an order of magnitude savings in power (1 kW as compared to 14 kW). Our system also results in a substantial mass savings when compared with a lower frequency system. For example, for equivalent accuracy at 13 GHz (proposed Wide Swath Ocean Altimeter frequency and the highest interferometric radar frequency to date) a boom of nearly 24m would be required as opposed to the 8m of our design.

Science Motivation
The Greenland and Antarctic ice sheets together hold enough ice to raise global sea level by 80 m. The annual exchange of mass on the ice sheets is equivalent to 8mm/yr of sea level, so that any fluctuation in that level of exchange is significant on a global scale. Recent airborne laser altimetry campaigns, satellite radar altimetry, and ICESat altimetry in Greenland and Antarctica have revealed glacier thinning rates ranging from a few cm/yr in the interior to meters or tens of meters per year at the coast, along channels occupied by outlet glaciers. Most interior changes are explained in terms of fluctuations in snowfall, whereas large coastal changes are caused by glacier ice dynamics. While coastal changes dominate in Greenland and West Antarctica, changes in interior accumulation have a significant impact on total mass balance in East Antarctica; it is therefore important to monitor both interior and coastal regions. In order to obtain meaningful results on ice sheets based on existing observations and interpretation of the results, we estimate that surface elevation needs to be measured with a sub-1O cm accuracy on a 1 km scale in the interior, and a few tens of cm at a spatial resolution of 100m at the coast, where the kilometer-scale dimensions of glaciers demand finer resolution. If these measurement objectives are achievedspecifically, better than Im at 100 m resolution on glacier ice along the coast, less than 10 cm at 1 km resolution in the interiorone will be able to improve current estimates of ice sheet mass balance obtained from other altimetry techniques significantly. Compared with satellite radar altimetry, this instrumentation will be able to resolve coastal changes more accurately (errors could be larger than 100% at present).
Spaceborne Instrument design A Ka-band single-pass InSAR is ideally suited to meet the requirements for glacier and ice sheets. Single-pass imaging radar interferometry has unique capabilities in providing independent of solar illumination. A Ka-band (35GHz) center frequency maximizes the single-pass interferometric accuracy (proportional to the wavelength), reduces snow penetration (when compared with lower frequencies), and remains relatively impervious to atmospheric attenuation. Figure 1 illustrates the measurement configuration whereby two antennasdisplaced in the cross-track directionview the same region on the ground. The interferometric combination of data received on the two antennas allows one to resolve the path-length difference from the illuminated area to the antennas to a fraction of a wavelength. From the interferometric phase the height of the target area can be estimated. Therefore, an InSAR system is capable of providing not only the position of each image point in along-track and slant range as with a traditional SAR, but also the height of that point through interferometry. The height precision is a function of the baseline length, the system frequency and the radar signal to noise ratio [1]. While the millimeter-wave center frequency maximizes the interferometric accuracy from a given baseline length, the high frequency also creates a fundamental problem of swath coverage versus signal-tonoise ratio. While the length of SAR antennas is typically fixed by mass and stowage or deployment constraints, the width is constrained by the desired illuminated swath width. As the across-track beamwidthwhich sets the swath sizeis proportional to the wavelength, a fixed swath size equates to a smaller antenna as the frequency is increased. This loss of antenna size reduces the two-way antenna gain to the second power, drastically reducing the signal-to-noise ratio of the SAR system. To overcome this fundamental constraint of high frequency SAR systems, we will use digital beamforming (DBF) techniques to synthesize multiple simultaneous receive beams in elevation while maintaining a broad transmit illumination. Through this technique we will preserve a high antenna gain on receive, reduce the required transmit power, and thus enable high frequency SARs, and high precision InSAR from a single spacecraft.
As previously noted, a digitally beamformed receive array was selected in order to achieve the required swath and looks while minimizing the required transmitter power. There are several possible approaches achieving a wide measurement swath. The pro and cons can be summarized as follows: 1. Wide Beam: -A wide beam (covering the full swath) is used on both transmit and receive. Pros: Simplicity Cons: Reduced antenna gain requires substantially higher transmitter power to produce adequate SNR fine resolution, topographic mapping over a wide swath 2. Narrow Beam ScanSAR: Of the above configurations, only digital beamforming provides the required measurement performance and swath using feasible transmit power levels.
Following from the science requirements for accuracy, and spatial and temporal sampling we have derived the measurement system parameters. The system parameters are summarized in Table 1. We have chosen a 0.063 m x 4.0 m transmit antenna to illuminate 7 degrees in elevation. At a boresight incidence angle of 22 degrees, this results in a ground swath of 70 km. On receive, the full swath is synthesized as 16 simultaneous subswaths in elevation using DBF over the full receive antenna (1 x 4 m comprised of an array of sixteen 0.063 m x 4 m antenna "sticks" in elevation). One fundamental consideration of using an elevation array is the trade between the size, or number of "sticks" and the ability to steer the array due to the presence of grating lobes. In order to avoid grating lobes the array must be critically sampled spatially (X/2). In the case of GLISTIN, the inter-stick spacing is -7.42 resulting in grating lobes at +/-7 degrees. When off-boresight beams are formed, the grating-lobe levels become significant, as they are no longer rejected sufficiently by the element pattern of the sticks. However, because beam formation occurs only in elevation, we are easily able to range-gate out the ambiguous returns. By simple geometry, the local slope of the surface could cause the grating-lobe to occur at the same range as the desired viewing angle. However, in this scenario the scene would not only need to have a slope of 18 degrees (quite possible for glaciers) but also to extend for 42km for a total height of 14km, a geometry which does not exist on Earth. The peak transmit power is 1.5 kW, which is within the realm of currently available technology. The pulse repetition frequency is 4 kHz to satisfy critical sampling requirements. The bandwidth is kept relatively low (40MHz) to minimize data-rates yet satisfy glacial resolution goals. Despite this, the data-rate still presents a substantial challenge.

GROUND-BASED DEMONSTRATION
The overall objective of this program is to develop and demonstrate the key technology of the GLISTIN concept including the end-to-end measurement technique and associated processing. A block diagram of the key development, the digital antenna array, is shown in Figure  2. The antenna will be integrated into a simple radar system, mounted on a positioner at a site overlooking the JPL facility, and used to collect an interferometric image. The demonstrated antenna will be a Im long (1/4 of the spaceborne length) full-height array (tm, 16 "sticks").
Dedicated digital receivers will be integrated with each antenna stick and a 0.5m interferometric baseline will be synthesized by interfering the beamformed returns from the upper and lower halves of the array. An interferometric image will be produced through a combination of digital beamforming for the fine-scale coupled with elevation scanning on a coarse scale using the positioner. Azimuth scanning will be achieved by rotating the positioner in that axis.
The two key sub-elements of the demonstration are the antenna aperture and the integrated digital receiver. These efforts are described in subsequent sections. Additional hardware such as data acquisition computer are included in the demonstration but are not considered a focus of this technology risk reduction program.

WAVEGUIDE SLOT ARRAY DEVELOPMENT
The development of a Im x 4m aperture at Ka-band is a challenging task. The first step is to decide which antenna technology will be used. This decision is largely influenced by mass, loss, profileall of which must be very lowand compatibility with the system's digital beamforming network. A waveguide slot array was selected as the best candidate to meet these requirements.
It is envisioned that the GLISTIN Im x 4m spaceborne aperture will be subdivided into four Im x Im waveguide slot array panels for deployment, with each panel consisting of 160 slots by 160 slots. Existing technologies, however, do not permit fabrication of a complete Im x Im array panel in a single step. So, each panel must be comprised of several smaller subarrays that are mounted on an external support frame. The full Im x 4m aperture would then be connected through separate power divider networks attached to the back of the array.
Given the complexities involved in fabricating the full Im x 4m aperture, the development path must be selected carefully in order to identify potential problems early in the project. In order to test the digital beamforming radar on Earth, a Im x Im array panel was chosen for the demonstration aperture. This section discusses and presents the preliminary stages in development of the Im x Im waveguide slot array panel for the ground-based demonstration.

Theory
The GLISTIN demonstration antenna is a large slot array with 160 x 160 elements (tm x lm). Yee has shown that the design and analysis of large slot arrays may employ an infinite array model with excellent accuracy [2]. In such a model, the external mutual coupling between radiating elements is modeled by considering each slot to be embedded in an infinite array. Thus the external mutual coupling is computed in terms of Floquet wave summation [3] instead of element by element mutual coupling integral as is done in conventional slot arrays [4]. The accuracy of the infinite array model becomes better as the array size becomes bigger and its validity has been established even for a small 7x7 array [2]. Based on Scharstein's work, [5] we believe that the use of the infinite array model for external mutual coupling employed in the GLISTIN antenna may introduce errors in slot voltages only in the outermost two rows and two columns with minimal error in the input reflection coefficient. Several computer programs were developed to solve the pertinent coupled integral equations for the slot aperture electric field by the method of moments. For the infinite array mutual coupling problem, our computer program was validated by comparing our results with those of Ansoft High Frequency Structure Simulator (HFSS).
In the design of the slot array we use half-height radiating waveguides to minimize the higher order mode coupling effects in the junction regions. Full height guides are used for feed and input waveguides. The slot conductances in the radiating waveguide and the normalized resistances in the feed waveguides have been chosen to maximize the return loss bandwidth as much as possible, even though the infinite array slot admittance has an extremely narrow band response. All of the moment method computations assumed square ended slots. Since the manufacturing process introduces round ended slots a correction is made for slot lengths assuming 'equal area' criterion. The smallest subarray unit in the 160x160 element array is a lOxIO slot array. Moment method and HFSS simulations of a full lOxIO subarray in an infinite array coupling environment are shown in Figure 3. Both these simulations include all intermediate coupling slots and waveguides (i.e. the reflection coefficient shown is at the input waveguide to one lOxIO subarray). As can be seen, there is good agreement between the two methods. There is even better agreement when only one mode is included in the MoM analysis, which is what the HFSS simulation included. The slight shift lower in resonant frequency for the MoM calculation is due to the inclusion of higher order modes.
Input to 10x1O subarray with regular slot offsets In order to compute the radiation patterns we first ran the moment method computer program for an infinite array of sub-arrays. Subsequently, we used the pattern multiplication principle to obtain the pattern of the entire array. The unit cell pattern is computed by summing the radiation from each radiating slot in the unit cell with the appropriate excitation computed by the moment method. The array factor is multiplied by the pattern of the unit cell. The moment method program was subsequently modified to account for scanning in the H-plane, as required by the system beamforming network. Thus there is a linearly progressive phase from sub-array to sub-array along the Hplane. It was interesting to note that the unit cell pattern didn't show much deviation between the scanned case and the broadside radiation. Therefore the scanned patterns can be obtained by multiplying the array factor for the particular scan direction by the pattern of the unit cell in the unscanned case. Radiation patterns were computed over a frequency range of 35.16 to 36.06 GHz. The patterns were found to be excellent with very little degradation over this frequency range where the return loss is better than 10 dB.

Design and Fabrication
As mentioned above, the smallest subarray unit is lOxIO elements, and four of these are combined to make the next subarray level up: a lOx40 array. This subarray was selected as the prototype to validate both the model and fabrication techniques, and will be discussed in this subsection. The results of this prototype cycle will be used to scale up to a larger 40x80 subarray tile, which will then be used to fill the final Im x Im aperture of the demonstration antenna.
There are two main components in the design: the radiating and coupling slots (already introduced in the theory section), and a four-way equal-split corporate feed network that feeds each lOxIO subarray in the larger lOx40 prototype array. The feed network was designed using standard WR-28 waveguide dimensions to facilitate calibration and measurement. The main components of the feed network are H-plane Tee septum power dividers, Hplane bends, equal-length waveguide runs to each subarray, and an E-plane Tee at the input to the 1 Ox40 subarray. All components in the feed network are tuned to the required operating frequency of 35.66GHz.
An additional feature introduced into the feed network was an E-plane phase "trombone", whose main function is to equalize the phase to each lOxIO subarray with shims, and required only if fabrication tolerances create phase imbalances. However, by adding mounting flanges at the trombones, it is not only possible to measure the performance of the feed networkthe equality of amplitude and phase to each subarraybut also to measure the individual performance of each lOxIO subarray. A standard WR-28 waveguide calibration kit can easily be used to achieve this. An aluminum cover with two E-plane bends is used to close off the trombone for full lOx40 subarray operation.
In conjunction with the electrical design, it is necessary to make the part compatible with the fabrication process, which in this case is aluminum dip-brazing. In order for this process to go smoothly and reliably, several features must be included in the mechanical design of the components. Working closely with the vendor, these details were finalized and a 3-D solid model was created. Shown in Fig. 4, this solid model was also used to program the CNC machines.
The dip-brazing process involves first machining the various layers of the array in aluminum and then fusing them together by dipping them in a 1000°F salt bath for several seconds. After a thorough cleaning to remove excess salt from the interior, and final machining to complete mounting fixtures, the part is ready for testing. The final brazed part, shown in its feed network test configuration, is given in Figure 5.

Measurements
Overall, the results were very good for a first iteration. The input match for the full 10x40 subarray, given in Figure 6, shows the resonant frequency is about 1% below the design frequency of 35.66GHz, and within simulation and fabrication tolerance errors. The power divider and individual lOxIO subarray matches are shown in Figure 7. As can be seen, the lOxIO subarrays are also tuned slightly low, but the power divider provides more than 30dB return loss at the design frequency of 35.66GHz.  The E-and H-plane patterns at 35.66GHz are shown in Figures 8 and 9, respectively. Agreement with array theory is very good. The measured E-plane beamwidth of 1.730 agrees very well with the 1.70°predicted by array theory. Similary, the measured H-plane beamwidth of 6.98°agrees very well with the 6.87°predicted by array theory, and also meets the requirement for a beamwidth greater than 6.8°.
Additional measurements, not shown here, indicate that the patterns begin to degrade at frequencies below about 34.8 GHz. Note that the return loss at this frequency for the full 10x40 subarray is about 20dB, so it is clear that return loss alone is not a good indicator of performance. Patterns and return loss must be simultaneously tuned to the required operating frequency.
A key conclusion from the measurements is that any adjustments made to retune the input match must not have a negative impact on the patterns at 35.66GHz, which are already close to satisfying the requirements for the Im x Im demonstration antenna. Recent moment method calculations suggest that it may be possible to adjust all the slot lengths by 1% to compensate for the missed resonant frequency. While this is the simplest and most appealing of all possible solutions, it also does not take into account fabrication repeatability.    MHz. Bandpass sampling is utilized to reduce the overall data rate, which is a recurring theme in this system. The 10-bit ADC selected provides more than 7 effective bits 0x4O subarray.
for an input at L-band. These 10-bits are passed through a 1 to 4 demultiplexer and finally into an FPGA. Initially the FPGA will serve as a buffer and a means to communicate onto the data bus, but may eventually be used for real-time 'ERS processing of data. The front panel data port (FPDP) standard was selected for transferring data from the receiver. This standard is simple in implementation, provides high-speed capabilities and provides the ability to ndent receivers and bus together many devices.
y. We are currently iver with 80 MHz Multi-Channel Configuration im DBF antenna. ..
Iere modifd fo anthen To accomplish digital beamforming, this system requires vere modified for the multiple receivers working in unison.
For the fundamental design demonstration there will be 16 receivers all communicating ment ofo thespabot e data to a centralized data acquisition system. The largeted in the Table 2. scale control scheme for this system is depicted in A single FPGA is used to communicate timing and configuration information to each of the 16 receivers. This FPGA is also used to aggregate the data from each receiver, which communicates its data onto a parallel FPDP bus. These aggregated data are then communicated to a data acquisition (DAQ) system through a serial FPDP (sFPDP) interface. sFPDP was selected for transmission of data to the DAQ as it overcomes some distance limitations associated with parallel FPDP. The DAQ stores the data onto a hard-disk array for post-processing.

Spaceborne Implementation
The aforementioned hardware is intended to be utilized in a ground-based demonstration. However, careful consideration was given to all major design decisions and their potential impact on a spaceborne implementation. The key digital components: ADC, DMUX and FPGA were all selected due to the availability of space-qualified equivalents. A flight configuration will require the antenna to be mounted via a boom to the spacecraft. Similar to the demonstration configuration, the receive elements must be mounted directly onto this antenna, so care has been taken to minimize power consumption, size and mass of the receive elements to facilitate this configuration. A few shortcuts were taken to simplify the demonstration system, such as the use of spinning disks for data storage and the use of a slower pulse repetition frequency. These decisions were made to accommodate the large volume of high-speed data generated by the system. The flight data rate problem is further exaggerated by the larger receive window required. To address this problem, scalability and data reduction techniques have been considered. In the demonstration configuration, sFPDP is used to transfer data from the entire receive array to a centralized storage. This optical standard was selected for simplicity, but also due to the scalability of optical media. While the sFPDP protocol can transfer up to 2.5 Gbps, many other optical standards exist that are capable of operating at much higher data rates. Additionally, in year three of this program, the potential for utilizing the FPGAs onboard each receive element for realtime processing of radar data will be examined. This technique can reduce the data rate and the overall volume of data that must be stored between downlinks thereby eliminating the reliance on spinning disks. While a considerable amount of work remains to shape this system into a flight capable equivalent, a path exists. This design will continue to mature as the demonstration system is refined throughout the life of this program.

CONCLUSIONS
We have described the design of a novel interferometric radar system for measurement of ice topography. In order to reduce technology risks associated with this ambitious instrument we are developing a ground-based demonstration of the key technology challenge: a Im x Im Ka-band DBF antenna.
We have developed tools for analyzing large slotted waveguide arrays and have shown results that are in good agreement with commercially available codes (HFSS) for test cases. These tools were used to design a 10 x 40 element prototype subarray that is part of the 160 x 160 element array being developed for the demonstration. Measured results show good agreement with simulation albeit with some discrepancy in frequency of the return loss minima. Further iteration will be used to correct this issue.
Designs and breadboards have been developed for the array digital receivers. The next iteration in the digital receiver design is focused upon reducing its' size so that 16 receivers can be accommodated on the back of the Im x Im aperture.
In the next two years of our three year program we will assemble and test all the components of the Im x Im DBF array and demonstrate the performance of the array using antenna range measurements and interferometric imaging of nearby terrain and structures. In parallel with this technology demonstration we will continue to develop the system and mission designs and improve the fidelity of our design, analysis and simulation tools.

ACKNOWLEDGEMENT
This research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.