> But range resolution is much easier to improve; it is just a matter of
> measuring time -- and time is by far the easiest physical quantity to
> measure. You get it almost for free;
That is not exactly correct. For most if not all radars, the range resolution is determined by the effective pulse duration, usually turned upside down and expressed as bandwidth because of the use of various pulse compression techniques.
> But the orbital motion of the satellite only yields one dimension of
> improved aperture. A one-dimensional picture may be "imaging" in some
> sense, but not in the sense of producing a picture that a human can
> look at. Or can a second dimension be extracted somehow?
Indeed it can, and even the third dimension. In the case of ISAR (it's kind of a special case of more general SAR processing), what usually happens is that scatterers at different distances from the effective axis of a rotating target have different radial velocities relative to the radar. (Rotation doesn't have to be rotation in inertial space, just changing aspect angle like you see as a car drives by.) Position is thus encoded by velocity, velocity is encoded by Doppler, and number-crunching allows the positions of the scatters on the body to be backed out. There can be ambiguous and degenerate situations which mess up the image if the real or effective rotation axis happens to have an unfortunate orientation during the period of the observation. Googling for "Inverse Synthetic Aperture Radar" will discover a number of papers that explain how the trick is done.
As it happens, there's an excellent nontechnical paper that explains this a bit and, even better, deals with the very systems of relevance to this thread. Unfortunately, it doesn't seem to be on the Web, but is mentioned at http://www.ll.mit.edu/news/journal/12_2.html. I've excerpted some of it below.
Wideband Radar for Ballistic Missile Defense and Range-Doppler Imaging of Satellites William W. Camp, Joseph T. Mayhan, and Robert M. O'Donnell Lincoln Laboratory Journal Volume 12, Number 2, 2000 [EXCERPTS]
Lincoln Laboratory led the nation in the development of high- power wideband radar with a unique capability for resolving target scattering centers and producing three-dimensional images of individual targets. The Laboratory fielded the first wideband radar, called ALCOR, in 1970 at Kwajalein Atoll. Since 1970 the Laboratory has developed and fielded several other wideband radars for use in ballistic-missile-defense research and space-object identification. In parallel with these radar systems, the Laboratory has developed high- capacity, high-speed signal and data processing techniques and algorithms that permit generation of target images and derivation of other target features in near real time. It has also pioneered new ways to realize improved resolution and scatterer-feature identification in wideband radars by the development and application of advanced signal processing techniques. Through the analysis of dynamic target images and other wideband observables, we can acquire knowledge of target form, structure, materials, motion, mass distribution, identifying features, and function. Such capability is of great benefit in ballistic missile decoy discrimination and in space-object identification...
The distinguishing characteristic of a wideband radar is its fine range resolution, which is inversely proportional to the operating bandwidth. Such a radar system has a range resolution that is a fraction of the linear dimensions of its intended targets... With range resolution fine enough to encompass a target in a significant number of resolution cells, it becomes possible to distinguish individual scattering centers, which occur at regions of physical discontinuity. A ballistic missile warhead, for example, exhibits radar reflections from the nose, body joints, and base, as well as other points of discontinuity such as antenna ports. To observe radar reflections from smaller discontinuities, the radar must be able to operate at short wavelengths, since discontinuities much smaller than a wavelength will in general produce low-intensity reflected signals from the target. In addition, a short wavelength is desirable for observing curved surfaces, because when the wavelength is short compared to the radius of curvature, the radar reflection is dominated by specular reflection, thus allowing a finer determination of the size and shape of corresponding surfaces....
In the above scenario, the radar produces a one-dimensional range profile of the target. However, if the target is rotating about an axis that has a component perpendicular to the radar line of sight, such that some scattering centers are moving toward the radar with respect to others that are moving away from it, we can construct a one-dimensional cross-range profile for each range cell through Doppler processing of the radar returns. The range and cross-range profiles can then be combined to produce a two-dimensional range-Doppler image of the complete body. We can analyze this image to yield body size, body shape, the position and nature of scatering centers, the presence of internal reflections, the rotation rates, and the rotation axes for the object. In addition, these images can provide valuable information on the nature of the materials used in constructing the body, and information about antennas, apertures, and interior structure of such an object.
Three-dimensional images can be generated from the two- dimensional images by using a technique called extended coherent processing. With this technique a series of range- Doppler images are collected over a time period when the target presents different look angles to the radar. The series of range-Doppler images is then coherently processed and referenced to a particular look angle. The resulting three-dimensional images produce even greater detail of target features than the two-dimensional range-Doppler images. The image of the damaged Skylab orbiting laboratory shown in Figure 1 is an example of this kind of processing.
[Caption for Figure 1]: Simulated radar image (actual radar images of satellites remain classified) of the NASA Skylab orbiting laoratory, with a damaged solar panel on one side and a partially deployed solar panel on the other.
ALCOR... was the first high-power, long-range, wideband field radar system. Lincoln Laboratory was the prime contractor for ALCOR;... It became operational at Kwajalein Atoll in 1970, and was probably the first wideband radar in the world to reach that status, although research in this field was under way elsewhere during the 1960s, mostly notably at Rome Air Development center... It was located next to the TRADEX radar on Roi-Namur Island in the Kwajalein Atoll...
ALCOR operates at C-band (5672 MHZ) with a signal bandwidth of 512 MHZ that yields a range resolution of 0.5 m. (The ALCOR signal was heavily weighted to produce low range sidelobes with the concurrent broadening of the resolution.) Its wide- bandwidth waveform is a 10 microsecond pulse lineraly swept over the 512-MHz frequency range. High signal-to-noise ratio of 23 dB per pulse on a one-square-meter target at a range of a thousand kilometers is achieved with a high-power transmitter ( 3 MW) peak and 6 kW average) and a forty-foot- diameter antenna. Cross-range resolution comparable to range resolution is achievable with Doppler processing for targets rotating at least 3 degrees in the observation time. The pulse-repetition frenquncy of this waveform is two hundred pulses per second...
In 1972 the TRADEX UHF radar was replaced by an S-band system... while retaining its L-band capability. Although TRADEX S-band was the sedcond wideband radar system to be brought on line by Lincoln Laboratory, the approach taken to achieve fine range resolution was substantially different from that taken in ALCOR and later wideband systems. The S-band wideband waveform grew out out of research on frequency-jumped pulses conducted by the Laboratory in the 1960s. The new set of S-band waveforms includes one with a signal bandwidth of 250 MHz. This bandwidth is achieved by transmitting a string, or burst, of 3 microsecond pulses; each pulse has a different center frequency such that the bandwidth spanned by the burst is 250 MHz. The spacing and number of pulses in a burst is variable and the maximum repetition rate is one hundred bursts per second. Coherent integration of a burst yields a range resolution of one meter...
The Haystack Long-Range Imaging Radar
From its earliest days of operation ALCOR was called on to track and image satellites, both domestic and foreign. ALCOR's historic imaging of the booster rocket body of China's first orbiting satellite in 1970 was followed by similar success in imaging the USSR's Salyut-1 space station in 1971. The demonstration of such imaging capability led to the establishment of the Space Object Identification program at Lincoln Laboratory. One of the many successes of this Program was the imaging of NASA's troubled Skylab orbiting laboratory shortly after its launch in 1973, as shown in Figure 1. Telemetry data from Skylab showed that something was wrong with the deployment of the solar panels. ALCOR images showed that one solar panel was missing and the other panel was only partially deployed, and there was no evidence of the presence of the micrometeorite shield. This information was extremely useful to NASA in determining how to recover from this mishap and successfully continue the Skylab mission.
As the Space Object Identification program continued, the capability to image satellites out to deep-space ranges soon became an important requirement. The limited sensitivity of ALCOR allowed the observation of satellites only out to intermediate altitudes. At the same time, researchers wanted to improve range resolution, extend range-window coverage, and achieve higher pulse-repetition frequencies in order to eliminate cross-range ambiguities in the images of rapidly rotating space objects. Concurrent advances in signal processing technology gave promise that such improvements in radar techniques could readily be accomodated. After exploring a number of options, researchers determined that a high- performance long-range radar imaging capability could be provided by a new radar system added to Haystack, the Lincoln Laboratory-built radio-astronomy, communication and radar research facility located in Tyngsboro, Massachusetts. The development of this new radar capability for haystack was sponsored by ARPA. After the facility was completed in 1978, operations were supported by the U.S. Air Force.
The Haystack system has a number of features that rendered this option extremely attractive. It has a large diameter (120 ft) antenna needed to achieve deep-space ranges... The Haystack antenna surface tolerance allows efficient operation up to 50 GHz, thus readily supporting operating at X-band (10 GHz) with a bandwidth of 1024 MHz, and a resulting range resolution of 0.25 m... [T]he new system [is] known as the Long Range Imaging Radar, or LRIR.
The LRIR, which was completed in 1978, is capable of detecting, tracking, and imaging satellites out to synchronous-orbit altitudes, approximately 40,000 km. The range resolution of 0.25 m is matched by a cross-range resolution of 0.25 m for targets that rotate at least 3.44 degrees during the Doppler processing interval. The wideband waveform is 256 microseconds long and the bandwidth of 1024 is generated by linear frequency modulation. The pulse-repetition frequency is 1200 pulses per second...
The Haystack Auxiliary Radar
When the LRIR became operational in 1978, the Haystack facility was being operated by a university consortium know as the Northeast Radio Observatory Corporation, or NEROC. The primary mission of Haystack was radio astronomy. Lincoln Laboratory contracted with NEROC to use the facility for satellite tracking and imaging for a thousand hours per year. Eventually, U.S. Space Command and NASA recognized that this arrangement for sharing Haystack was too restrictive to satisfy their needs. The restriction was especially limiting when there was an immediate need to assess new or unexpected foreign launches, or when space debris needed to be catalogued.
The Haystack Auxiliary Radar (HAX) system... was conceived to eliminate this restriction on observation time, and simultaneously further improve range resolution and pulse- repetition frequency. HAX operates at Ku-band with its own forty-foot antenna, transmitter, RF hardware and receiver, but it shares the LRIR control and signal and data processing systems with Haystack. HAX, which began operation in 1993, is the first radar to have a signal bandwidth of 2000 MHz, which improves the range resolution to 0.12 m. HAX represents a signiicant advance in radar imaging capability, producing finer and sharper images of satellites than the Haystack LRIR. It is also extremely useful in producing detailed information for NASA on the locations, orbits and characteristics of space debris.
The Millimeter Wave Radar
The Millimeter Wave Radar, or MMW, was built at Kwajalein by Lincoln Laboratory (with significant contributions by the University of Massachusetts, RCA, and Raytheon) to extend the general imaging and tracking capabilities of ALCOR and to develop millimeter-wavelength signatures of ballistic missile components. The MMW... became operational at Ku-band (35 GHZ) in 1983, and W-band (95.48 GHz) in 1985, sharing a paraboloidal antenna with a diameter of forty-five feet. Both systems initially featured wideband waveforms of 1000-MHz spread generated by linear FM, and achieved 0.28-m range resolution. The transmitted pulse width is 50 microseconds at a maximum pulse-repitition raate of 2000 pulses per second. The initial peak power at Ka-band was 60 kW and at W-band was 1.6 kW... [Several upgrades have been made.]
[Descriptions of Cobra Dane, Cobra Judy and Cobra Gemini omitted.]