UoSAT-12 MINISATELLITE FOR HIGH PERFORMANCE
EARTH OBSERVATION AT LOW COST
Dr. Marc Fouquet
Professor Martin Sweeting
Surrey Satellite Technology Ltd
Centre for Satellite Engineering Research
University of Surrey
Guildford, Surrey GU2 5XH
UNITED KINGDOM
Tel: (44) 1483 259143 Fax: (44) 1483 259503
internet: m.fouquet@ee.surrey.ac.uk
internet: m.sweeting@ee.surrey.ac.uk
WWW: http://www.ee.surrey.ac.uk/EE/CSER/UOSAT/
Surrey Satellite Technology Ltd (SSTL) at the University of Surrey (UK) has pioneered cost-effective satellite engineering techniques for smaller, faster and cheaper satellites to provide affordable access to space. SSTL has designed, built, launched and operated a series of twelve 50kg microsatellites in low Earth orbit which carry a wide range of satellite communications, space science, remote sensing and in-orbit technology demonstration payloads - for both civil and military applications. Eight of SSTL's microsatellites have carried CCD Earth imaging cameras, coupled to powerful onboard processing computers which have demonstrated the capability of this new generation of low-cost, rapid-response, modern small satellites.
SSTL has recently developed an advanced 250kg minisatellite platform to meet the needs of future low-cost Earth observation and remote sensing applications. With enhanced payload mass and volume, attitude stability and communications capabilities, this new platform will support more sophisticated payloads than the minimalist microsatellites. The experimental UoSAT-12 mission, to be launched into LEO in 1997, will carry high resolution and multispectral remote sensing payloads to demonstrate imaging not currently practicable on existing microsatellites.
This paper reviews the design and characteristics of the UoSAT-12 imaging payloads which include multispectral cameras providing 40-metre ground resolution with an 80 km swath width, and a panchromatic imaging instrument providing 10-metre resolution. The trade-off between linear pushbroom sensors and area CCD cameras, and the role of on-board processing for automatic image processing are presented. These techniques enable small, low-cost minisatellites to approach the performance of conventional Earth observation satellites - but at a fraction of the cost.
This pioneering work by Surrey on advanced microsatellites and minisatellites has resulted in imaging missions that can respond rapidly to customer requirements, which are affordable to a far wider community of individual commercial or governmental organisations, and which provide emerging space nations with an independent Earth observation capability.
In principal, there is widespread support for using low-cost spacecraft to perform remote sensing of the Earth. However, in practice there is a general reluctance to abandon conventional design methodology, and as a result, many proposals and studies for small remote sensing spacecraft approach the design as a miniaturised version of a traditional mission. Accordingly, only moderate savings in size and cost are obtained, with the typical design calling for a 400 kg launch mass.
To achieve significantly greater reductions in size (and associated reductions in cost), it is necessary to use a more radical approach to design. At the same time, the designers must accept that there will be additional engineering constraints on very small satellites that do not exist on large conventional spacecraft, imposed by the limited resources of the scaled-down housekeeping systems. Similarly, it may be necessary to sacrifice some of the payload's performance to achieve the desired cost savings. These facts can often be unpalatable to designers, but must nevertheless be addressed if novel implementations are to be achieved.
The design of SSTL's imaging systems differ from conventional remote sensing instruments in several ways. Perhaps the most obvious is that area-array CCD sensors are preferred to conventional scanning designs (either mirror-scanned or pushbroom). The main attribute of an electronic snap-shot camera is that the fixed pixel structure of the CCD ensures image geometry under all circumstances.
Conventional scanning instruments place stringent demands on the satellite's attitude determination and control systems (ADCS) to ensure that the successive scan-lines remain (close to) parallel. Typical requirements are for 0.1° pointing accuracy and 10-3 degrees / sec drift, or better. Achieving this level of stability dictates a large and sophisticated ADCS, at odds with the goal of a small and inexpensive spacecraft.
On the other hand, the snap-shot operation of an electronic camera is immune to any residual drifts in attitude stability. This allows the satellite's ADCS to be substantially reduced in scope, yielding benefits in terms of system size and complexity. As a result of a smaller ADCS, the power budget is vastly reduced, and the on-board control is not so critical to mission survivability. Given the interaction of these and other factors in the overall design, reducing the demands of one or two systems can have knock-on effects, resulting in a vastly simplified and cheaper spacecraft. Thus, by making the imaging payload largely independent of the attitude stability, huge overall savings are made in the cost (financial, technical and logistical) of SSTL spacecraft platforms.
Other attractions of solid-state sensors are their low power, compactness, robustness and an absence of moving mechanics. SSTL exploits the great advances in terrestrial semiconductor technology to make imaging cameras that fit in the palm of your hand, but without sacrificing reliability, sensitivity or operational performance.
Accordingly, choosing to use electronic cameras rather than scanners has allowed SSTL to deploy imaging payloads on tiny 50 kg microsatellites with success [7]. However, even when the performance of microsatellite ADCS are capable of supporting scanning instruments, it is by no means certain that these conventional architectures will be preferred over cameras. Even the data from LANDSAT, SPOT and IRS must be corrected for minor attitude disturbances - the guaranteed image geometry offered by a camera's area-array CCD makes these post-processing operations unnecessary.
Therefore, by considering the interactions between the platform and payload modules, and re-assessing conventional solutions where necessary, it has been possible for SSTL to develop imaging missions at a tiny fraction of the cost of conventional spacecraft. Whereas conventional satellite design methodology could be summarised by "what systems and features do I need to fulfil my mission specifications", the philosophy adopted with success by SSTL is better described by "what can I do with what I am given."
Adopting an alternative mind-set has enabled SSTL to develop a programme of advanced microsatellites fulfilling a wide range of applications [1,2}: data communications to remote portable terminals [3], communications research, technology demonstration, radiation studies, small space science, and Earth imaging.
Eight SSTL missions have carried Earth imaging cameras providing the first demonstrations of the feasibility of microsatellite remote sensing. Launched in July 1991, SSTL's UoSAT-5 [4] microsatellite represented a milestone in the development of low-cost Earth observation from space. In addition to being the first sub-100 kg satellite to demonstrate reliable, predictable and repeatable Earth observation, it was also the first imaging mission independent of governmental control. Since then, SSTL has confirmed this innovative remote sensing capability with the collaborative microsatellite missions, KITSAT-1 (Aug 1992) and PoSAT-1 (Sept 1993). [5,6,7]
A typical SSTL microsatellite mission costs US$2-3 million, and the basic camera unit is around US$120k. When one considers the low cost of these cameras relative to existing satellite imaging systems, the quality of the imagery is very high - the data from the PoSAT-1 cameras in particular is quite spectacular. And despite such low cost, reliability has not been compromised - all the cameras have worked well, accumulating over two decades of in-orbit heritage. The UoSAT-5 camera continues to provide excellent pictures, more than five years after launch.
Despite the success of the UoSAT-5, KITSAT-1 and PoSAT-1 missions, it is clear that the engineering constraints (mass, volume, power, attitude stability, communications, orbits) encountered on microsatellites limit the scope for certain moe demanding scientific and commercial EO applications. To support a broader range of activities, SSTL has developed a multi-mission minisatellite platform. The first flight of this new spacecraft bus will be the 300 kg UoSAT-12 mission [8], scheduled for launch in 1997.
Amongst the numerous mission objectives (including the in-orbit qualification of the new platform), one of the main goals for UoSAT-12 is the development of enhanced imaging capabilities. The particular attributes of this minisatellite platform for imaging are the improved attitude stability (0.1° pointing, 10-3 degrees / sec drift rates), cold-gas propulsion for station-keeping, and a 500 kbps S-band downlink. Of course, the spacecraft has increased mass and volume capacity, and is capable of accommodating payloads of up to 150 kg.
The rest of this paper reviews the imaging systems under preparation for the UoSAT-12 mission. Predictably, the design of these modules inherits a great deal from the existing microsatellite units, and circuits with flight heritage are re-used where appropriate. Indeed, the electronics retain backwards compatibility with SSTL microsatellites, and will be featured on the forthcoming TMSat mission [7] (scheduled for early 1997). The greater mass and volume available on the UoSAT-12 minisatellite will be exploited by using more sophisticated optics.
To allow rapid development and construction of its satellites, SSTL employs extensive modularity. Thus the basic structure is defined, with module trays for the power system, telecommand, telemetry, receiver, transmitter, attitude determination and control, and the on-board computers. This leaves a number of standard module trays available for experiments and payloads.
Figure-1:
Schematic of transputer and modular camera connectivity for
SSTL microsatellites
A modular concept is also developed within the imaging system by using electronic building blocks. The three elements include the image sensor, the microcontroller and frame-store, and the transputer image processing unit. This modularity allows different combinations of sensors, optics and processing capabilities to be selected, in response to the requirements of SSTL's customers. This also provides a framework for incremental upgrades without necessitating complete redesign.
Figure-1 shows the configuration of the four modular cameras and the transputer processing units for the FASat-Bravo and TMSat microsatellites. A number of communications channels are available (e.g. via the CAN-bus, and high speed point-to-point serial links). The topology is the same for the UoSAT-12 systems, except that six cameras and four transputers will be used.
Each SSTL camera comprises an image sensor board and a microcontroller / frame-store board. These two printed-circuits fit into a small (130 x 80 x 60 mm) milled box. The extensive use of surface-mount technology, populated on both sides of six and eight layer boards, is needed to achieve this miniaturisation.
Significant effort has been devoted to ensuring that the mechanical and electrical interfaces of these camera are identical, allowing them to be interchanged with ease. Each board measures 120 x 70 mm with common positions for the CCD sensor, connectors and mounting points across the designs. Mechanical adapters allow a wide range of optics to be mated to the standard module box. Figure-2 shows the modular camera fitted with an f/1.4 75 mm lens as flown on FASat-Bravo and TMSat. The modular notion has also been adopted between the microcontroller and the image sensor. All of the command parameters are sent to the camera head over a simple serial link adhering to the Motorola Serial Peripheral Interface standard. This interface is bi-directional to receive telemetry; the SPI link is expandable in either direction to accommodate more commands or status lines. In the opposite direction, a standard set of data and control lines have been defined for dumping digitised imagery into the frame-store memory. In addition to the 8-bit stream of digitised video, the sensor's sequencing logic produces clocks at the pixel, line and frame rates so that the frame-store can be synchronised with the incoming data. Finally it is possible to interconnect the various cameras with 'genlocking' signals to ensure they all capture their imagery at the same instant. Of course, no circuitry is shared between the separate modular cameras, ensuring that any failures are contained to the faulty camera.
Each of the modular cameras has a mass of 0.6 kg when ready for flight, excluding the mass of the optics which can vary widely. The base-line power consumption of the microcontroller circuit is 225 mW (45 mA at 5 V), peaking at 5.6 W (400 mA at 14 V) for the few seconds when the sensor circuits are active during image capture.
An option is to augment the sensor and microcontroller boards with a third board to give the camera fully isolated power and communications interfaces. Switching regulators provide the camera's +5v and +15V rails, and opto-isolators are used on all of the communications channels.
Figure-2: The SSTL modular camera configured for the FASat-Bravo / TMSat Narrow Angle Camera
SSTL currently has designs supporting three different CCD image sensors: the EEV CCD02-06 (384 x 576 pixels), the EEV CCD04-06 (578 x 576 pixels), and the Eastman-Kodak KAI-1001 (1024 x 1024 pixels). The designs for the EEV devices are derived from the well-established systems used on the SSTL microsatellites [4]; Kodak 2048 x 2048 cameras are also under development.
The Kodak sensor camera represents a new SSTL development. Predictably, the front-end of the SSTL modular cameras possess the same basic architecture as any CCD imager. Whereas the EEV devices used a chipset provided by the manufacturer to implement the essential functions of the camera, SSTL has developed these functions in-house for the Kodak CCD. Although requiring more development than employing the EEV chipsets, the circuitry for the Kodak camera has been designed to be flexible and reconfigurable, so that a wide range of new CCD devices can be supported in future. Thus, a programmable FPGA logic array generates all the timing sequences necessary to drive the camera in synchronism. These clocks are fed to the CCD sensor via suitable current and voltage buffers, which can be adjusted over a wide range of operating conditions. The CCD's output signal is presented to the video processing stages for the requisite amplification, sampling, blanking and level shifting. The conditioned video signal is quantised by an analogue-to-digital converter. Although only 8-bits of digitised imagery are collected, 10-bit converters are used to ensure no reduction in the effective resolution (which could be degraded by inadequate linearity, distortion or noise performance).
Each image sensor has an accompanying microcontroller board, making up a modular camera. The controller's design supports the three existing SSTL sensor units, and should be able to accommodate other CCDs with little or no rework.
The camera behaves as a slave terminal, accepting instructions from one of the on-board computers (i.e. the transputers - the hierarchy of the various OBCs, transputers and cameras is shown in figure-3). The microcontroller executes operations in accordance to its firmware. In most instances, these are well-defined sequences such as loading capture parameters (manual or automatic exposure, ADC references, full or partial resolution, etc.), activating the sensor, capturing a scene, and transferring the data. However, a full suite of low-level instructions are available for remote operation (either from an OBC / transputer or even the ground), useful for diagnostics or non-standard operations.
The other task of the microcontroller is to disseminate the imagery captured in the frame-store. The frame-store is a 4 MByte bank of static RAM - enough to buffer four scenes from the Kodak KAI-1001 sensor, twelve scenes from the EEV CCD04-06, and eighteen scenes from the EEV CCD02-06. The frame-store is dual-ported, allowing the sensor unit to write and the microcontroller to read the digitised imagery.
Figure-3: Schematic of on-board data handling hierarchy
The principal external communication routes for the microcontroller are via the two interfaces to the spacecraft's CAN-bus (the industry-standard Controller Area Network). SSTL's microsatellites feature two CAN-busses for redundancy, and the UoSAT-12 minisatellite has three. This is the primary route for all telemetry, telecommand and inter-module communications on UoSAT-12. There is no centralised telemetry and telecommand unit on SSTL's minisatellite - each module is equipped with a controller and a CAN node, implementing an entirely distributed control system.
In addition to the CAN, the camera microcontroller has an interface to a 20 Mbps serial link to the transputer image processing units. The primary purpose of this link is the rapid transferral of the digitised imagery away from the camera, but also provides a redundant pathway for all functions normally implemented on the CAN. In practice, the microcontroller is too slow to drive this link at the full speed of 20 Mbps. Instead for bulk data transfers, it launches a hardware unit which steps through the frame-store byte-by-byte, transferring the image at just under the full bandwidth. At this speed, a 1 MByte scene from the Kodak KAI-1001 sensor can be transferred in under a second - fast enough to sustain continuous imaging for all but the highest resolution applications.
One of the constraints imposed on a payload by a small satellite platform is the capacity of the downlink relative to conventional remote sensing missions (limited by the available power and the antenna gain). Recognising this potential restriction on the ability to exploit SSTL's Earth observation missions, the imaging systems include high performance processor modules to provide autonomous image assessment and compression. These processing units are based around the Inmos T805 transputer, and are integral to the operation of SSTL's imaging modules, resulting in significant improvements in productivity.
The transputer image processing modules provide dedicated computing support for the cameras, relieving the main OBC of these tasks. The transputers manage all aspects of the imaging modules including scheduling, image capture, analysis, pre-processing and compression. In addition to these Earth imaging tasks, the transputers also support the GPS receiver and star field camera modules (when flown).
As with the cameras, the transputer processors are derived from the system first flown on UoSAT-5 [4]. The current design for SSTL's microsatellites uses 25 MHz, 32-bit devices, making them amongst the most powerful computers flown in space. Each processor has access to 4 MBytes of fully error-protected static memory, being upgraded to 32 MBytes for UoSAT-12. The transputers collect imagery from the camera microcontrollers over the 20 Mbps links. They also possess interfaces to the CAN-busses. Each transputer occupies a 300 x 150 mm board (i.e. half a standard SSTL module tray - see figure-4). The microsatellites are generally equipped with two such processors; the UoSAT-12 minisatellite will have five.
As with all computer modules (not microcontrollers) flown on SSTL's satellites, the transputers do not hold their flight code in ROM. Only a bootloader is hard coded, and all applications software is uploaded in orbit. This has permitted the remote sensing experiments to evolve in flight, with the image capture, processing and compression software being upgraded as required.
Figure-4: The SSTL dual Transputer module prepared for FASat-Bravo / TMSat
A number of approaches are used on SSTL's spacecraft to improve the 'information per byte' of imagery on the downlink. Implemented by the transputers, the techniques used are:
The combined effect of these procedures is dramatic, providing over a 400% increase in daily yield on PoSAT-1. Specialist compression / candidate selection routines are used for the star sensor, and algorithms suited to compressing multispectral imagery are under development. Clearly, while the judicious use of on-board image analysis, processing and compression is of particular benefit to microsatellites, these techniques would enhance the performance of any remote sensing system.
Five different variations (six cameras) of the SSTL modular camera will be used on the UoSAT-12 minisatellite, providing a good demonstration of the versatility of the modular system.
In some instances, the specifications of these instruments are not finalised and cannot be reported here, even though the intended launch date is a year away. This is typical of SSTL's short development cycle, and of the engineering flexibility that is needed within an organisation committed to fast-response missions. Indeed, it is imperative that mission time-scales be kept short to ensure the projects remain within very tight budgets.
The Wide Angle Camera (WAC) is essentially the same instrument as flown on the UoSAT-5 and PoSAT-1 microsatellites, but benefiting from the new packaging of the modular camera.
An extremely wide angle lens (4.8 mm focal length) is used to offer continental coverage. The ground resolution from the standard 800 km, Sun-synchronous, orbit favoured for remote sensing is 2 km per pixel. The camera is fitted with an optical filter giving near-IR sensitivity, providing strong contrast between land, sea and cloud.
While meteorological imaging remains a viable application of microsatellites, the higher resolution imagers on SSTL spacecraft have generated the most interest to date. Therefore, the WAC often acts as a 'spotter' camera to assist in locating the scenery from the other cameras. Sometimes this can be difficult because of the narrow field of view of these cameras.
The SSTL High-resolution Camera (SHC) is designed to have a 10 metre spatial resolution from UoSAT-12's intended orbit of 650 km. The camera uses the newly developed circuits for the Kodak KAI-1001 CCD, whose architecture is especially well-suited to high resolution imaging of moving scenes.
An f/4 600 mm refractive objective is used to achieve this high resolution. The mass of the lens is 6 kg, in addition to the 0.6 kg of the standard modular camera; supporting sway-braces and other mechanical items will bring the total system mass to close to 10 kg. The mass, and particularly the length, of the optics precludes the deployment of this configuration on one of SSTL's microsatellites, so it is an ideal payload for the larger capacity of UoSAT-12.
SSTL expects this camera to provide imagery approaching that of the SPOT panchromatic mode, albeit with a reduced field of view. Given its low cost, it is certain that the radiometric and geometric calibration of the SHC on UoSAT-12 will be inferior to that of the SPOT instruments, but nevertheless it should be a useful remote sensing tool. The validation of these expectations will only be available once the system is in orbit, when expert users can compare the imagery side-by-side. Nevertheless, SSTL feels that one of the greatest assets of a low-cost, fast-response mission like UoSAT-12 is the ability to get payloads into orbit. This way results can be obtained under real operating conditions, rather than extrapolating from performance observed during ground-based simulations - and at a fraction of the cost.
Despite the inclusion of a microlens structure, the sensitivity of the KAI-1001 sensor is not as good as can be expected from a typical full-frame or linear CCD array. This is caused by having to sacrifice some of the device's photosensitive surface to implement its electronic shuttering and integration control features. For this reason, it will not be possible to give the SHC a narrow spectral sensitivity and maintain a good signal-to-noise performance, given the short integration times needed to combat motion blur. This is a common difficulty for high resolution instruments, e.g. SPOT PS mode (10 m) or LANDSAT-7 / ETM+ band 8 (15 m).
Accordingly the SHC will have wideband spectral characteristics, most probably in the middle of the visible spectrum (green and red). This presents a compromise between using shorter wavelengths that are more sensitive to atmospheric haze, and longer wavelengths where lower quantum efficiency implies a wider passband, subject to greater chromatic effects. The exact characteristics of the optical filter will depend on the results of various tests and field trials. Fortunately, filters can be substituted late in the system's integration, requiring only minimal refocusing.
UoSAT-12 will also be flying a pair of identical Multispectral Cameras (MSC), again making use of the Kodak KAI-1001 implementation of the SSTL modular camera. These cameras employ f/4 180 mm objectives to achieve 35 metre spatial resolution. Whereas the SHC will emulate the SPOT Panchromatic mode, the MSCs are inspired by the visible and near-IR capabilities of the LANDSAT Thematic Mapper.
To provide multispectral capabilities, mechanical wheels will rotate optical filters in front of each sensor. Within reason, there is no particular limitation on the number of filter bands that can be accommodated, and up to ten slots should be available on each camera. The exact passbands have yet to be chosen, but will certainly include the standard remote sensing spectra (similar to those of the LANDSAT7 / ETM+ instrument, i.e. blue: 0.45-0.52µm, green: 0.52-0.60 µm, red: 0.63-0.69 µm, near-IR: 0.77-0.90 µm, and panchromatic 0.51-0.90 µm). A violet-UV filter (0.35-0.42 µm) will also be present for evaluating air quality / smog concentration. SSTL is actively seeking suggestions for additional, or even alternative, bands that would fulfil specialist applications which are not well served by existing systems, but which could be demonstrated on this mission.
To generate a multispectral image set, successive image frames must be gathered through different filters. The filters employed for a given image will be dictated by ground control, falling back on a default sequence if this is not specified. The standard mode of operation will be to capture images in four spectral bands, filling the microcontroller's frame-store. Alternatively, the camera can be programmed to capture the imagery at reduced resolution, allowing eight spectral channels to be recorded, but at a reduced resolution of 40 x 80 or 80 x 80 metres (achieved using charge-amalgamation or 'pixel binning' in the CCD's read-out register).
The camera produces image frames every 60 ms, even though the integration time is much shorter, around 2 ms. The satellite's orbital velocity will displace the scene by 400 metres, or ten image lines, between frames. This will be a constant offset, which will be removed on-board by the transputers. Given that satellite imagery has strong correlations in the spectral dimension, this re-alignment is important if high compression ratios are to be achieved. In theory, attitude drift will result in misalignments between the successive image frames, as experienced with pushbroom scanners. In practice, given the rapid collection rate, these drifts will be negligible at under 0.1 of a pixel displacement across the multispectral set.
In addition to providing redundancy (the imagers are identical but completely independent), two multispectral cameras will be flown to increase the field of view - one camera will be nadir pointing, and the other will be mounted at a 3 degree offset across-track. This will increase the coverage swath to 2048 pixels (70 or 80 km, depending on orbit altitude). Because of the high speed links to the transputer units, it is possible to image contiguous scenes to produce long swathes along the ground-track. While there will be small misalignments between the successive scenes due to attitude drift, each 1024 x 1024 pixel image will have guaranteed registration. Identifying common ground points to splice the images together is a routine manual post-processing operation.
Although the MSCs will first be flown on the UoSAT-12 minisatellite, all aspects of their design have been carefully considered to ensure that similar instruments can be flown on future microsatellites. The mass of the entire camera, including lens, filter wheel and motor, electronics, and mechanical housing are expected to be under 3 kg.
In keeping with the low-cost tradition of SSTL's research programme, commercial off-the-shelf optics are employed for the SHC and MSCs. Despite being top-grade optical systems, targeted at specialist professional applications, the cost and availability of these objectives are vastly more attractive than for equivalent custom-designed systems.
In some ways, the design of commercial lenses is not optimised for Earth observation. Their intended applications call for optical performance to be maintained over a wide range of focus distances and apertures, whereas a satellite imager operates at a fixed focus and aperture. However these effects are minimised due to the image height of a CCD being much smaller than a 35 mm photographic image, and the pixel dimensions being very coarse compared to the grain of film. Preliminary trials with representative optics indicate an image patch (the physical manifestation of the theoretical point-spread-function) of around ten microns for both SHC and MSC systems - the same dimensions as the pixel spacing.
Of course, there is more to producing a high resolution imager than placing a large lens in front of the existing CCD camera. SSTL is undertaking a test campaign to characterise the performance of the objectives under both laboratory and field conditions. While these are routine procedures in conventional remote sensing and astronomical programmes, lengthy and expensive campaigns are, in many ways, contrary to the principles of low-cost spacecraft engineering. Part of the test campaign for the UoSAT-12 instruments is to evaluate the results from rapid field trials against conventional MTF test-bench measurements. Initial results indicate higher-than-expected correlations between these approaches (within 10%).
The very long focal lengths of the UoSAT-12 cameras, principally the SHC, are much more sensitive to errors in focusing than their predecessors. There is some concern that the intense vibration regime of potential launchers for UoSAT-12 (notably silo-launched vehicles), will cause shifts in the positioning and alignment of the sensor and optics. Shifts as small as 20-30 µm will result in serious loss of optical performance. As a result, SSTL is developing strategies for refocusing the optics once in orbit. Unfortunately, the high cost and long-lead times of the titanium, invar or carbon-fibre structures often used for space-bound optical systems are incompatible with the budgets and time-scales of this project.
The attributes which make solid-state cameras attractive for Earth observation apply equally to imaging stars. UoSAT-12 will carry two star cameras acting as precision attitude pointing sensors. One camera will have wide sky coverage (20° x 15°) to ensure that enough stars will always be in the field of view to perform the pattern matching, and the other will cover a smaller region (10° x 7°) but at finer resolution. The optics for these star cameras are very bright, f/0.85 25 mm and f/1.0 50 mm respectively.
The star cameras represent an alternative application of the standard SSTL modular camera, in this case employing the EEV CCD02-06 sensor. However, these instruments need increased sensitivity to operate under much lower lighting conditions than the usual visible cameras, and several features have been included into the design specifically for this. Techniques such as long integration times (up to a second), reduced output noise bandwidth, a non-linear transfer function, and low-noise amplification are needed to obtain usable signals from the faint stars.
The UoSAT-12 star cameras are enhanced from the experimental module flown on the PoSAT-1 microsatellite10. This practical experience has had significant influence on the design of the current system, by revealing many subtle operational considerations that would never have been observed in ground-testing. It is particularly interesting to note the trade-off between the cost of the imager and the amount of on-board computing that is available. In other words, low cost sensors can be used provided there is adequate processing power to model and compensate for the camera's imperfections.
At the time of writing, the structural model of the UoSAT-12 minisatellite has successfully passed its qualification vibration testing, which was particularly rigourous to ensure compatibility with all possible launch vehicles, including converted ICBMs.
The imaging systems currently being developed for UoSAT-12 will represent milestones in the development of low-cost remote sensing of the Earth. It is expected that the imagery from the SHC and MSC will be able to fulfil applications currently exclusive to large, conventional spacecraft such as SPOT and LANDSAT, despite costing a tiny fraction of these traditional missions. Nevertheless, terrestrial solid-state imaging continues to progress, and SSTL is actively considering these technologies to further develop its remote sensing capabilities.
The first objective in developing SSTL's imaging systems is to accommodate CCDs with larger pixel counts - initially with 2048 x 2048 pixels, but eventually even larger. This image size will be comparable to that from the LANDSAT MSS, but captured by a snap-shot camera rather than a whiskbroom scanner. To achieve wider coverage, it will of course be possible to mount cameras side-by-side, as with the two UoSAT-12 MSCs.
Whereas all the CCDs currently employed by SSTL have all-electronic operation, it appears inevitable that these larger sensors will have full-frame architectures and will need some form of mechanical shuttering. While this departs from SSTL's traditional avoidance of active mechanics, it will present a wider range of options to customers. Given the low financial risks but moderate payload capacity of a microsatellite mission, this solution represents an attractive trade-off. Conversely, on a minisatellite where the payload real-estate is not so restricted, the choice may be to deploy multiple all-electronic cameras to improve the overall reliability.
The on-going emergence of area-array CCDs with infra-red sensitivities will be exploited by SSTL. The first stage will be to accommodate devices covering the 1-2 µm wavelengths, providing crucial information on vegetation stress. These devices will most probably use platinum-silicide as the sensing medium, and will require thermo-electric cooling. Otherwise, the electronic control of these CCDs is very similar to those currently employed, and much of the circuitry can be re-used.
Of course, the existing electronics can be adapted just as easily to support linear CCD arrays for pushbroom imaging, should this become desirable.
The developments proposed will strive to emulate the performance of existing remote sensing missions, but at a tiny fraction of the cost. However, simply replicating the performance of existing spacecraft will not realise the full potential of inexpensive imaging micro- and minisatellites. To justify the huge cost of traditional Earth observation missions, the data must serve a wide user-base. However, this inevitably results in a dilution of the specifications for spatial resolution, swath width, spectral coverage, temporal updates and orbit phasing for any specific interest group. On the other hand, the low cost of small satellites makes it possible for individual corporations or government departments to procure a spacecraft tailored to their specific needs.
To have these capabilities ready off-the-shelf, SSTL will be investing in the development of high sensitivity imagers, capable of delivering good signal-to-noise performances when sensing through narrow bandpass optical filters. Ultimately, the design will evolve to a hyperspectral system, with the user selecting the bands of interest for any given scene. To reduce the downlink loading, only those spectral bands actually required will be collected.
The successful in-orbit demonstrations of low cost Earth imaging from microsatellites such as UoSAT-5, KITSAT-1 and PoSAT-1 have played an important part in changing the attitude of the established space industry towards small satellites for remote sensing. Graduating from these demonstration missions, Surrey Satellite Technology Ltd. is now developing a range of operational imaging systems to meet the needs of commercial users.
Recognising that SSTL's 50 kg microsatellite platform will not be adequate to support all Earth observation payloads, a minisatellite platform has been developed to accommodate more demanding payloads. The first flight of this platform, UoSAT-12, will carry numerous communications, technology demonstration and imaging payloads, including low-cost star cameras, a 10 metre resolution camera, and two multispectral 40 metre resolution cameras.
The rate of progress in terrestrial electronic imaging technology is phenomenal, meaning that by the end of the decade small spacecraft will be capable of capturing imagery similar to that of SPOT and LANDSAT. The short cycle times of SSTL's missions provides an ideal opportunity to capitalise on these technological developments.
Finally, although SSTL has a very proactive research and development programme in Earth observation, its corporate philosophy and workforce embrace a positive approach to meeting customers' needs. The on-going R&D, including the designs of the UoSAT-12 payloads, are in direct response to numerous commercial requests. All aspects of the company encourage flexibility and adaptability in providing inexpensive solutions to spacecraft engineering. The company welcomes all input regarding the opportunity, the suitability (or current short-comings) in executing useful tasks from these small satellites.
[1] Sweeting M.N: UoSAT microsatellite missions; Electronics & Communication Engineering Journal, IEE, June 1992.
[2] Sweeting M N: Microsatellite & Minisatellite Programmes at the University of Surrey for Effective Technology Transfer & Training in Satellite Engineering; Proc. of Int. Symp. on Satellite Communications and Remote Sensing, 20-22 Sept 1995, Xi'an China.
[3] Da Silva Curiel R.A. & Ward J.W: Operational Store and Forward Payload for Low-Earth Orbit Microsatellite Data Collection and File Transfer Communications; Proc. of Int. Symp. on Satellite Communications and Remote Sensing, 20-22 Sept 1995, Xi'an China.
[4] Fouquet M: The UoSAT-5 Earth Imaging System - in-orbit results; 2nd conference on Small Satellite Technologies and Applications, SPIE symposium on Aerospace Sensing, Orlando Florida, 20-22 April 1992.
[5] Fouquet M. & Brewer A: The Role of Microsatellites for Earth Observation - Eight years of orbital experience at the University of Surrey; Conference on Small Satellites for Remote Sensing, The Space Congress, 24-25 May 1995, Bremen Germany.
[6] Fouquet M. & Sweeting M.N: Remote sensing using the University of Surrey's microsatellites; Proc. of Int. Symp. on Satellite Communications and Remote Sensing, 20-22 September 1995, Xi'an China.
[7] Fouquet M. & Sweeting M.N: Earth Observation using Low Cost SSTL Microsatellites; Symposium on Earth Observation Sensors & Technology, 47th IAF Congress, 7-11 October 1996, Beijing China.
[8] Please consult SSTL World Wide Web sites for information.
[9] Rabbani M. & Jones P.W: Digital image compression techniques, SPIE Optical Engineering Press, 1991, ISBN 0-8194-06481.
[10] Leitmann M, et al: PoSAT-1 Star Imaging System - In-flight performance; 4th conference on Small Satellite Technologies and Applications; SPIE/Europto symposium on Satellite
Remote Sensing, Rome Italy, 29-30 Sept 1994.