Preliminary
M. Caplinger
January 1995
Users of the Mars Observer Camera (MOC) will interact with the instrument through the MOC Ground Data System (GDS). The GDS is used to define observing programs for the MOC, schedule all the potential image acquisitions, construct uplink command sequences to carry them out, interpret the resulting instrument telemetry, and archive the returned image products for later processing. GDS elements for science analysis will also be developed, but these are not described here.
This document describes the capabilities of the MOC, the characteristics of the Mars Global Surveyor mission that affect the MOC's operation, and how MOC observing programs can be defined. For details on how the GDS then uses the observing programs to control MOC operations, the reader is advised to consult a companion document, the MOC Ground Data System User's Guide.
The MOC consists of three optical systems. Each system has a CCD line-array detector at its focal plane; these detectors are all connected via analog-to-digital electronics to a digital control system containing custom gate arrays, a 32-bit microprocessor, and 11 megapixels of dynamic RAM memory for image storage. A block diagram of the MOC is shown in Figure 1.
Figure 1: MOC block diagram
Functionally, the MOC is two "cameras" in one instrument. The two separate systems are the Narrow Angle, or NA, and the Wide Angle, or WA.
Unlike previous planetary mission instruments such as those on Viking, Voyager, and Galileo, which used vidicons or CCD arrays to take an entire image simultaneously (framing cameras), all three systems in the MOC are line-array or pushbroom systems. The CCDs are 1D detectors consisting of a single line of CCD elements. A 2-D image is built up out of many single readouts of the CCD, and the motion of the spacecraft itself over the planet's surface is used to scan the line array in the downtrack direction. This allows the MOC to have much higher resolution than is possible with framing systems, as well as allowing considerable flexibility in the size of each image.
The NA has an 3.5 meter focal length, f/10 Cassegrain optical system with a total field of view of 7 milliradians (0.44deg.) and a 2048-element CCD detector. From the nominal MGS orbital altitude of 379 km, each pixel is 1.4 meters square. Because the NA is a pushbroom system, the dimensions of an NA image are adjustable in both the downtrack and crosstrack directions. NA image dimensions are limited by the detector size in the crosstrack direction (2048 pixels or about 2.8 kilometers) and by avaliable spacecraft power and image buffer memory size in the downtrack direction. Using the full width of the detector and assuming 2X lossless real-time data compression, the longest possible full-resolution NA image will be about 11,250 lines, or 15.8 kilometers, long. Alternatively, one can sample only a portion of the CCD and trade image width for image length; for example, using only the middle 512 pixels of the CCD, the image will be only 716 meters wide but could be as long as 45,000 lines or 63 kilometers.
The pixels of NA images can also be summed in the instrument to allow larger but lower resolution images to be taken. Summing is always square (the same summing in cross- and downtrack directions) and can range from factors of 2 to 8. With 8x8 summing, each pixel will be 11.2 meters square. At that resolution, the 2048-pixel-wide image is reduced to 256 pixels, but the maximum image length would be 90,000 pixels or about 1000 kilometers.
The above calculations assume that a real-time compressor was being used to compress the image data as it was being acquired. The MOC NA can perform factor of 2 lossless compression in real-time, doubling the maximum size of an image over what would be possible without compression. The other forms of data compression available on the MOC are either lossy (i.e., they involve trading image quality for image size) or they are not real-time (i.e., cannot be applied during an image acquisition, but only later to save downlink data rate) or both. These forms of compression will be the topic of a later section.
The Wide Angle (WA) camera consists of a pair of 11.3 mm focal length, f/6.5 lens systems with 3456-element CCDs. One system has a red (580 to 620 nm) filter, and one has a blue (400 to 450 nm) filter; the systems are otherwise identical. Each WA has a field of view of 140 degrees. At the MGS orbital altitude, the width of the sub-nadir pixel is about 250 meters; because of the greater distance to the limb ("perspective effect") this increases to about 3 km near the limbs. The limb-to-limb view of the planet is about 3000 kilometers wide. Figure 2 shows emission angle and width per pixel (resolution) plotted versus pixel position from the center of the line array.
Figure 2: MOC Wide Angle geometry
As with the NA, the WA image size can be adjusted in both the crosstrack and downtrack directions. Pixels can also be summed in both directions; unlike the NA, this summing need not be square, so that, for example, a limb monitoring observation can be summed in the downtrack direction but not in the crosstrack for maximum vertical resolution.
The WA can acquire images in red, blue, or both colors simultaneously.
The design of a given MOC observing program is strongly influenced by the orbital parameters of the Mars Global Surveyor spacecraft. MGS is in a nearly-circular, sun-synchronous orbit with an inclination of 92.3deg. and a period of about 2 hours. The result of this is that the spacecraft crosses the dayside equator at about the same local time throughout the mission (nominally 2 PM). The dayside passes are made from north to south, and every equator crossing is shifted westward from the previous one by 28.4 degrees of longitude. After 89 orbits or 7 sols, the tracks nearly meet but are separated by ~60 km eastward; this seven-sol period is called a repeat cycle. After 329 orbits or 26 sols, the orbital tracks again nearly coincide, but are shifted westward by 24.7 km. This period is called a mapping cycle. Because of the alternating eastward-westward shifts, the swaths are "filled in" with tracks of closer and closer spacing, until after 19 mapping cycles (or 550 sols), the planet is covered completely by 3.1 km swaths, approximately the field of view of the MOC NA. The 19-mapping-cycle period is called a supercycle. All of these periods are shown graphically in Figure 3.
Figure 3: Mars Global Surveyor orbital track spacing
The nominal orbital repeat scheme above implies considerable navigational and spacecraft control accuracy. It must be noted, however, that the 3.1 km spacing requires accuracy well in excess of that specified by the Mars Global Surveyor Project. In practice the position of the spacecraft will not be so precisely predicted or controlled. For users of the WA, control will probably be good enough to have an excellent idea of the timing of the entire observing program well in advance. On the other hand, timing for programs containing specific NA targets cannot be predicted beyond a few weeks at best. This should not be of great concern to a MOC user, however, as the GDS allows the specification of targets by location only. The GDS can handle all timing issues automatically without user intervention.
Unlike the cameras on previous missions like Viking and Voyager, the MOC is not mounted on a scan platform. It has no independent pointing capability. The center of the field of view of all three optical systems is always pointing directly at the nadir (to the accuracy of the spacecraft's attitude control.) For the NA, this means that the image can cover at most about 1.5 km on either side of the spacecraft ground track. Because the repeat cycle orbit spacing, it is not possible to build up multiple-swath NA mosaics in time less than one supercycle. (Note that there are only 1.22 supercycles during the one martian year of the primary mission.) During each supercycle, there is only one opportunity to image any given point at the equator with the NA (nominally - because of navigation errors there may be multiple opportunities at some targets, and none at others.) Closer to the poles, there may be opportunities to observe a given area more than once.
Because the WA has such a wide field of view, it can see to the limb of the planet, a distance of about 1500 kilometers on either side of the ground track. As a result, it is possible to treat the wide angle as though it could be pointed, by sampling only a portion of the CCD. Naturally, the closer an image is to the limb, the lower its per-pixel resolution and greater the emission angle at each pixel.
Because the NA has such a very narrow field of view relative to the positioning and attitude error of the spacecraft, it would be difficult to determine just where the camera was pointing when the image was acquired. Because of this, we expect that most NA images will be accompanied by a small WA image, called a context image, that will serve to locate the NA image in a wider field of view and allow it to be located on maps or in earlier spacecraft images. To support this, the offset between the NA and WA lines of sight will be precisely measured prior to launch.
The GDS controls MOC operations as follows. First, the user community defines a series of image acquisitions. The simplest such definition is a request to image a particular point on the planet with particular settings of the NA or WA cameras. This has been generalized to a particular area of the planet, a fraction of areal coverage, a range of emission angle, and a number of repetitions. For example, a user can tell the GDS to "take five 512x512 full-resolution WA images in two colors of any area between 130deg. and 140deg. west longitude and 10deg. N and 20deg. N latitude if the emission angle is <=10deg.."
Given a collection of such specifications, the GDS examines the predicted ground track of the spacecraft over some planning interval (typically a week or two) and determines which observations can be made in that period. It then calculates the timing for all possible observations and checks to see if the instrument resources (power, buffer space, downlink data rate, and so on) can accommodate all the observations. If conflicts arise, observations with lower priority are thrown out or modified until the sequence is feasible. This conflict resolution can be done either automatically (by the application of simple rules) or interactively. Once a valid sequence is generated, it is converted into explicit MOC commands and uplinked to the instrument for execution. The GDS keeps track of all commanded acquisitions and informs the requestor when the image has been acquired and is available for examination in the MOC database.
The following parameters can be used to specify a MOC observation:
cameras: NA, WA, both
color: (for WA) red, blue, both
image width (crosstrack)
image height (downtrack)
downtrack summing
crosstrack summing (must be equal to downtrack summing for NA)
latitude range
longitude range
emission angle range
minimum fraction of area covered by image
compression types allowable
amount of compression
image priority
number of acquisitions
time interval range between successive acquisitions
On the other end of the spectrum from the observation of single points of the planet is global monitoring of the entire planet. The MOC automatically acquires a global map image of Mars with the WA at a given resolution (normally 7.5 km/pixel) and a daily map is constructed from this. The GDS schedules global observations as a fixed overhead rather than as a series of timed acquisitions. Users who desire to make other observations of a global nature are urged to make their needs known to the author. In particular, limb observations may fall into this category.
Downlink data rate is a precious commodity for the MOC. The minimum data rate is only 700 bits/second, and at that rate it takes 13 1/2 hours to transmit an uncompressed 2048-square NA image.
To alleviate this problem, the MOC can use a wide variety of data compression techniques to reduce the storage requirements of images. The most important of these is a real-time (implemented in hardware), predictive, lossless compressor that can reduce the byte size of an NA image by a factor of about 2 on average. We expect that in normal circumstances, this compressor will always be used[*] , and its use is invisible in the final image. With some limitations, the same lossless compression scheme can be applied to WA images.
It is theoretically impossible to compress MOC image data losslessly by much more than a factor of 2. Often, however, a factor of 2 reduction is not enough, as the user would like to take more or larger images than would be possible under instrument constraints. To address this, the MOC provides several lossy compressors that can compress data by factors of 5 to 10 times or more. Some of these are use lossy predictive techniques and have the advantage of high speed - it should be possible to compress NA images 3-4 times in real-time (with hardware) via predictive coding. The other compressors use transform-based methods and provide better compression, but are implemented in software and are therefore much slower - they take tens of minutes to compress a normal-sized NA image. Therefore, they can be used only to increase the number of images taken, and not to increase the size of a single image.
In effect, lossy compression "throws away" image data in an attempt to save space. The compressors go to great lengths to retain as much visual information as possible, but still, the use of lossy compression must be carefully weighed against the impact to the science the compressed image will be used to address. Figure 4 shows an image compressed by a variety of methods.
Figure 4a: air photo of Lake Vanda shoreline, Antarctica, MOC NA resolution
Figure 4b: 10x DCT compression (requant 12)
Figure 4c: 20x DCT compression (requant 26)
Figure 4d: 4x predictive lossy compression
As mentioned before, the GDS is capable of scheduling MOC operations automatically, given a list of observation specifications. Typically, the MOC must make compromises, as not all observations that could be made can be made simultaneously. The GDS applies a number of "heuristics", or simple intuitive rules, to a sequence of images that contains conflicts. These rules are subject to the following constraints:
* lower priority images are degraded or deleted before higher-priority ones;
* images can be compressed (losing more data) only to some user-specified limit; beyond that limit, they are deleted instead;
* NA images take priority over WA images, all other things being equal;
* images can be compressed less than specified only if that improves throughput;
* images can be changed in size or summed further if the user permits.
Remember that the primary source of conflicts among image acquisitions is buffer space. Consider the following simplified example of conflict resolution: if image 2 follows image 1 in time, and there is not enough space to store both images in the buffer at once, and the downlink data rate is low enough that the part of image 1 that could be transmitted before image 2 must be taken still doesn't allow image 2 to fit, then something must give. If image 1 is the lower-priority image, the GDS tries the following:
* Image 1 could be compressed in real-time, making it smaller immediately.
* Image 1 could be compressed by software, making it smaller eventually.
* Image 1 could be decreased in size, either crosstrack or downtrack.
* Image 1 could be further summed.
If image 2 is lower priority, the GDS tries the same things on it. Remember that the user can specify that some of these operations should not be considered: for example, the user may have no use for a reduced-resolution (summed) image, so the system will not try that modification to save the image.
If, after simulating the results of all such operations, the conflict still persists, then the lower-priority image is removed from consideration and goes back into the list of pending acquisitions, to be considered on a later pass if the opportunity arises.
This conflict resolution process can be run entirely automatically. It can only perform as well as the specifications permit, however. The best way of scheduling a sequence of observations will be to use the interactive tools of the GDS to examine conflicts and modify acquisitions by hand. Though time-consuming, this allows the user to make reasonable science tradeoffs that are clearly beyond the capacity of the GDS to make automatically. The tools to expedite this kind of sequencing are still being designed and will be described in a later version of this document.
The GDS attempts to automate MOC operations as much as possible, freeing science users to concentrate on the science of their observing programs, not the scheduling of those programs. At the same time, the GDS will allow dedicated users total insight into, and control over, the operations of the GDS related to their observing programs.
June 1989 - initial version.
August 1989 - revised following review and updated to 379-km-altitude orbital parameters. Distributed to science team (without compression figures.)
January 1995 - Added Figure 4 in digital form, revised for MGS and converted for WWW access.
Return to MSSS Home Page