Operational Calibration of the Imagers and Sounders
on the GOES-8 and -9 Satellites, Page 12

Michael Weinreb, Michael Jamieson, Nancy Fulton, Yen Chen, Joy Xie Johnson,
James Bremer, Carl Smith, and Jeanette Baucom

6.2 Relativization

The signal from any scene is the instrument's output while viewing the scene, minus its output when viewing space. Until the spring of 1996, however, the data in the visible channels of the GOES-8 and GOES-9 sensors were transmitted to users in absolute counts, i.e., scene counts without the space counts subtracted out. In this form, the data were consistent with the visible-channel data from other NOAA satellites past and present.

The process of differencing the signals from the scene and space is termed "relativization," because its outcome is a signal relative to space. Relativization removes effects of the zero-signal background, long-term drifts, and detector-to-detector offset differences. For the imagers, relativizing the signals is usually not important under normal conditions. The outputs of the photodiode detectors are relatively stable in time and consistent among detectors. The space clamps continually set the count level from space at approximately 29 counts, thereby preventing significant drifts caused by temperature changes. And there are no detectable drifts from 1/f noise in the visible channels. For a single detector, the rms variation in space-clamp levels ("clamp noise") is less than one count (out of 1024), and the variability in the space-clamp level among detectors is of the order of one count.

However, some relatively rare performance anomalies in the imagers have provided a reason to relativize. Before the spring of 1996, the visible imagery from both imagers experienced occasional incidents of severe striping, which consisted of positive or negative offsets, between 10 and 30 counts in magnitude, in the outputs of one of the eight detectors. The cause is believed to be noise spikes in the space clamps of the offending detector. Furthermore, for a few days around the equinoxes, the space clamp level of physical detector 6 of the GOES-9 imager's visible-channel array was observed to increase by approximately five counts for a few hours of the day, which also caused image striping. Relativization should prevent such anomalies from causing stripes in the images. Relativization would also be of benefit for the sounders. Because they do not clamp on space, their outputs show diurnal and seasonal drifts and detector-to-detector variations, tens of counts (out of 8192) in magnitude.

Table 4. Dates relativization was first enabled

Therefore, in the spring of 1996 (see Table 4), relativization of the visible-channel imagery from the imagers and sounders was made operational. The data are relativized before being normalized. Relativization requires two steps. First, the mean count value from the most recent space look is subtracted from the pixel counts. (For the imager, the space-look data are from the post-clamp following the preceding space look.) Second, a constant count level, X₀, is added back in. Without the second step, relativization would have two undesirable consequences. First, when space itself is the target, the distribution of the data would be approximately Gaussian with a mean of zero. Half of the distribution would have count values less than zero and would be lost, since GVAR (the GOES VARiable format data stream, which users receive) will not accommodate negative integers. Second, the overall brightness of the image would change significantly between the "relativization off" and "relativization on" states.

For all eight visible-channel detectors of all imagers, the value of X₀ is 29. For all four visible-channel detectors of all sounders, the value of X₀ is 920. Since these are the nominal values for the space-count levels in the absence of relativization, there is hardly any change in overall image brightness between the "relativization off" and "relativization on" states.

At first glance, it may seem that the two steps of relativization would cancel each other out. This is not the case, however, because the mean space value and X₀ are not necessarily equal. The mean space-count level varies from space look to space look, as it is affected by noise (imager) or noise and drift (sounder), whereas X₀ is invariant.

Next  |  Previous  |  Table of Contents

Contact Michael P. Weinreb at michael.weinreb@noaa.gov
Latest Revision: July 9, 1997