One of the major challenges for the GOES I-M series was to develop a navigation system that would improve earth location and registration accuracy. That involved maintaining a stable imaging platform in a three-axes satellite environment and predicting image shifts due to orbital motion as well as instrument motion and thermal distortion. More details are available in Kelly (1989).
Image navigation refers to accurate earth location (latitude/longitude) of each pixel within an image. This requires precise knowledge of the spacecraft's position in the orbit and attitude within that orbit at the time of image acquisition. The spacecraft's position is determined as a function of time (orbit) as well as the orientation of the imager and sounder reference optical axis (roll, pitch, and yaw) with respect to the reference orbital axis (attitude). Three types of observations are obtained to determine orbit and attitude: stars, landmarks, and range. Those observations are processed on the ground in the Sensor Processing System (SPS) and Product Monitor (PM) and are then transmitted to the Orbit and Attitude Tracking System (OATS), where the actual calculation is performed (Kelly 1989).
Star measurements are obtained by pointing the instrument just east of a star's predicted position and waiting while the star crosses the instrument field of view; a typical detectable star diameter is 23 µrad, and its apparent movement is 72 µrad s-1. Separate star sensing capability exists in the imager and the sounder. The star's actual position is processed in the SPS and passed on to the OATS. It is anticipated that three star measurements with good geometric separation will be obtained every half-hour without any significant impact on imaging operations. This should provide sufficient input data to determine daily attitude profiles.
Landmark measurements are made by locating geographic features of known latitude/longitude in the imager data. The imager line/ pixel associated with a latitude/ longitude is then sent to the OATS for input to the orbit determination process. This landmarking function is performed semiautomatically using cross-correlation techniques in the PM. While imager visible data is the primary source for landmarks, they can also be obtained from imager IR and sounder visible data. Landmarks are obtained in an off-line manner and therefore have little impact on operations. One landmark will be obtained per hour; more may be processed depending on operational needs. Range is estimated in the SPS by measuring the elapsed time between the uplink and downlink signal of the retransmitted data. The SPS formulates a range measurement for OATS input. A range measurement is performed every half-hour.
TABLE 6. Image navigation and registration performance requirements (indicated to 3 ~) . Requirements are relaxed near midnight to accommodate the impact of solar heating on the instruments.
Imager
IMC and MMC continuously correct imager and sounder pointing for deterministic orbit and attitude effects and for the effect on satellite attitude induced by the scan and slew motions of the instrument themselves. These two corrections together are intended to produce images registered within the performance requirements summarized in Table 6.
The IMC models, and then removes, orbit and attitude motion from each image. Coefficients that describe the orbit and attitude contribution to pixel shifts are generated in the OATS and uplinked to the spacecraft. These coefficients are applied in an orbit and attitude model in the onboard attitude and orbit control electronics (AOCE), which computes correction signals at a rate of 64 per second. These signals are applied to the azimuth and elevation servomotors of the imager and sounder, and compensate for the predicted orbit and attitude motion. For example, if perturbations to optical axis motion due to orbit and attitude motion were predicted to be in the southeast direction, correction signals would be generated in the northwest direction. This produces an image with no apparent motion. Through this process each subsequent image is registered to the previous image within error tolerances (Table 6).
In addition to IMC, a second correction signal is applied to the instrument to enhance registration accuracy. The GOES I-M imager and sounder scan mirrors operate independently. While one instrument is scanning, the other could be slewing for star sensing, a space look, or performing blackbody calibrations. The scan and slew motion of one instrument affects the spacecraft attitude and hence the pointing of the other scanning instrument in a predictable fashion. This second correction, MMC, is automatically generated by separate control logic in the AOCE. The compensation signal is generated at the onset of scanner motion in one instrument and applied to the scanner servomotor of the other instrument. These corrections are applied continuously and are independent of ground operation.
The IMC system references all images to a perfect GOES projection. This projection (defined by satellite subpoint) is input to the IMC coefficient generation process in the OATS. The OATS then computes coefficients that produce images earth located and registered to this standard grid over the period that the coefficients are in effect. The "fixed gridding" allows gridding information to be generated once in a 24-h period. The IMC process fits the images to this standard grid. NOAA plans to use this capability so that all images are referenced to a standard grid centered over the equator at the nominal satellite subpoint. IMC biases all images to this "perfect GOES projection"; the grid is generated once per day at the implementation of a new IMC set on board the spacecraft. Plans are to maintain this projection within the navigation specifications except when the satellite is being moved from one station to another or when the satellite inclination increases beyond 2 degrees near the end of the mission.
The capability to maintain one GOES projection significantly simplifies user earth location. Since the gridding information is generated only once per day, the user also needs to generate earth location information only once per day, at the implementation of a new IMC set. This earth location information is accurate within specified error tolerances (see Table 6) from image to image for the ensuing 24-h period.
After ground processing of the spacecraft signal, visible channel data are transmitted to users as digital count values with 10 bits of information for the imager and 13 bits for the sounder. The data stream contains calibration coefficients for users wishing to convert count values to radiances and albedos. Those coefficients are determined before launch with an integrating sphere whose calibration is traceable to the National Institute of Standards and Technology. After launch, the validity of that calibration is uncertain since it is possible that the sensor gains may change.
As with the present GOES instruments, the visible data from the imager are normalized to compensate for differences in gain among the eight channels (Weinreb et al. 1989). Sounder visible data are normalized separately. The data are normalized in the SPS at Wallops in real time with 10-bit conversion tables for the imager and 13-bit conversion tables for the sounder. This is done by relating the raw radiance outputs, in digital counts, to normalized ones. Striping should be less of a problem for GOES I-M than it is with the present GOES because of the finer quantization of the GOES I-M intensity scales.
The silicon photodiode detectors used in GOES-I have been proven to be more stable over time than the photomultiplier tubes used in the GOES-VAS; thus, a single set of normalization tables should be valid for months at a time. The normalization tables are generated off-line at the National Environmental Satellite Data and Information Service (NESDIS) in Suitland, Maryland, by matching of empirical distribution functions (Weinreb et al. 1989). The basic idea is that with a large ensemble of measurements the distribution of intensity measurements in every channel should be the same. One channel is designated as a reference channel, and the outputs of other channels are modified, so their intensity distributions are the same as that of the reference channel. The reference channel is chosen so that its observations fill as much of the range of digital counts as possible without clipping at either the low or high ends. Even more important, the reference channel should have a stable gain that does not change rapidly with time.
The infrared channels are calibrated in real time from data acquired in flight when the sensors view space and onboard warm blackbodies. The onboard blackbodies are external to the entire optical trains of both the imager and the sounder and fill their aperture; this procedure allows a full-system calibration rather than the partial calibration of the GOES-VAS, where the radiation from the blackbody calibration source bypasses the telescope. The calibration equation, which relates sensor output x (in digital counts) to scene radiance R, is R = qx2 + mx + b. The coefficients m and b are the slope and intercept, respectively. The quadratic term (with coefficient q) corrects for possible nonlinearities in sensor response, which may occur with the channels using mercury cadmium tellurium (HgCdTe) detectors.
The current GOES satellites, which spin at 100 rpm, experience small diurnal temperature excursions. However, since the GOES I-M satellites are three-axes stabilized, temperatures within the sensors vary by tens of degrees Kelvin over a 24-h period. Therefore, the coefficients in the calibration equation must be updated frequently. To accomplish this, both the imager and the sounder view space routinely (see Table 1). The sounder views its blackbody every 20 min, while the imager normally views its blackbody every 10 min unless doing so interrupts an image in process. At the most, the imager operates 30 min between blackbody viewsÑfor example, when it makes a full-disk image. Drift in detector response [often referred to as "1/f noise" (Bak et al. 1987)] dictates that the space and blackbody calibration looks occur frequently enough to keep changes in calibration within the specified noise levels. The ground system processing interpolates between calibration events, so that different calibration coefficients are determined for each data sample to account for the linear portion of the detector drift.
Data acquired when the imager and sounder view space and the onboard blackbody determine the slope and intercept but not the coefficient of the quadratic term in the calibration equation. That coefficient is determined in thermal vacuum tests before launch by exposing the imager and sounder to a laboratory blackbody at temperatures between 180 and 320 K; the blackbody calibration is traceable to the National Institute of Standards and Technology. This calibration procedure is repeated several times with the imager and sounder held at different temperature "plateaus," whose range exceeds the operating temperatures expected in flight. Thus, the quadratic coefficient for the calibration of each spectral band is characterized as a function of instrument temperature.
Calibration slopes and intercepts are computed in the SPS computer at the Wallops CDA in real time. Slopes are recomputed immediately after the imager or sounder views its onboard blackbody. The temperature of the blackbody is determined as an average of the readings from the eight thermistors. The radiance of the blackbody is calculated from the convolution of the Planck function over the spectral response of the instrument.