The Space-Based Visible Program: Appendix II Calibration and Testing
Introduction
Introduction
The process of calibration and end-to-end system testing was instrumental to the success of the SBV sensor. The calibration effort resulted in a detailed understanding of the performance of the SBV sensor and allowed the conversion of SBV sensor observations into accurate measurements. The end-to-end testing established the confidence that the SBV flight hardware would function as expected after launch, and was key to the development of the ground-based command and data processing capability in the SPOCC.
Calibration
Most of the calibration efforts went into the telescope and focal plane. Measurements made early in the development process verified that the electronics had insignificant effects on system calibration compared with calibration-source stabilities and equipment-setup repeatability. An engineering model of the electronics was interchanged with the flight electronics with no detectable changes. Figure 1 shows a cutaway drawing of the telescope and a field-of-regard view in spacecraft coordinates [1]. The telescope has a distorted field of regard, a tradeoff made against focus quality and design time. The distortion requirement was driven by the signal processor; to maintain a high probability of detection any straight line in space projected onto the focal plane would have no more than 0.2-pixel deviation from a straight line over a 100-pixel length. Focus quality and radiometric calibration were also critical measurements.
FIGURE 1. The SBV telescope and focal-plane projection of the four CCD imagers. Because of the off-axis design of the SBV sensor, a considerable amount of well-defined distortion exists within the images. This distortion is removed mathematically on the ground in the data-reduction processing pipeline. Reprinted from the Johns Hopkins APL Technical Digest by permission [1].
Distortions were mapped by projecting a helium-neon laser spot sequentially to 256 positions per CCD imager and then calculating thirty-two coefficients (sixteen per axis) for a polynomial fit. Each CCD imager had its own set of coefficients. In all cases the errors were well below the 0.2-pixel data-analysis error budget.
Flat-field uniformity was measured by placing a large integrating sphere in front of the telescope aperture. Three separate flat-field functions were noted: (1) as predicted by the optical design, the steradiancy of pixels varied somewhat as a function of field position; (2) there was a slight misalignment of a light shield over the focal plane; and (3) the CCD imager had an inherent sensitivity pattern. Over the field of regard, the flat-field nonuniformity was about 12%, mostly due to the predictable steradiancy change. All effects were repeatable to better than 1%.
Radiometric calibration accuracy was defined as the residual errors after compensating for the stable flat-field patterns and reference radiometer accuracy. Data were taken with the same large integrating sphere used for flat-field measurements. Data errors were dominated by equipment calibration accuracy, setup repeatability, and drift, rather than inherent sensor instabilities. The calibration variations had a repeatability of about 2.5% and are estimated to have absolute errors of less than 10%, which is well within the required 20%. Additional effort was not expended to refine the radiometric calibration process once the program requirements had been exceeded.
Dark current and noise tend to be related in CCD camera systems. At a focal-plane temperature of -40oC, the nominal operating temperature, the imagers had a dark current of eighteen electrons per second per pixel, well within the allowed limit of one hundred electrons per second. Noise, with the cover closed, measured to be 1.1 digital numbers root-mean squared (rms) at a 0.4-sec integration time, resulting in a noise of 6.9 electrons rms. Noise was measured both as a temporal variation of a single pixel and as an area average in a single frame. Both methods gave the same results. Focal-plane calibrations conducted prior to integration to the telescope showed that the SBV CCDs were capable of transferring charge packets as low as ten electrons with no apparent loss. The charge transfer remained unchanged down to operating temperatures of -55oC, where the dark current was less than two electrons per pixel. The focal planes were screened for pocket density (pockets are electron traps in the imager that cause a variety of temporal and spatial imaging distortions). There were no pockets noted greater than one hundred electrons on any imager, and none were detectable on CCD 3, the boresite imager.
The last calibration parameter was the bidirectional reflectance distribution function (BRDF), a measure of the scattering mechanisms inside the telescope. Figure 2 shows the BRDF measurement history of the telescope over nearly two years, where the sensitivity of the SBV sensor is plotted over time at two different tangent angles above a fully illuminated earth disk. The telescope was delivered to the vendor meeting its specification; the BRDF, however, seems to be slowly degrading with time, probably because of minor particle buildup on the mirrors. The sensitivity is calculated from measured BRDF data and plotted in Figure 2 as the minimum-diameter specular sphere detectable with a signal-to-noise ratio of six at altitudes of both 100 km and 350 km above the earth tangent point. Smaller diameters indicate better detection sensitivity, which corresponds to better rejection of the sunlight reflected from the earth. The rolloff is slower than was expected, allowing the SBV sensor to meet its performance requirements, even with some additional future contamination. The telescope has not been cleaned since delivery; we believe that the good level of cleanliness is due to strict adherence to contamination control procedures and to design decisions made with contamination control in mind. In fact, the contamination control was so successful that the SBV sensor arrived on orbit with the same BRDF as measured on the ground prior to launch. This lack of contamination is a significant achievement, since the launch vibrations could have easily redistributed any contaminants in the telescope to the mirrors.
FIGURE 2. Measurement history of the bidirectional reflectance distribution function (BRDF). The BRDF, a measure of scattering mechanisms within the telescope, is a key metric in the determination of the sensitivity of the SBV sensor. The red line indicates the original design goal of detecting a 68-cm sphere at a tangent height of 100 km, while the dashed blue line represents the goal of detecting a 38-cm sphere at a tangent height of 350 km. The actual performance of the SBV sensor is indicated by the solid black squares for a 100-km height and by the open black squares for a 350-km height.
End-to-End Testing
With a system as complex as the SBV sensor, there is always the worry that a critical parameter has been missed and the SBV sensor will not work properly on orbit. Ideally, prior to launch, the sensor should be taken to an observatory where it could operate by looking at real scenes containing clutter (stars) and moving objects (satellites). However, the risk of contaminating the telescope forced us to reject this idea. To simulate a real observation, Lincoln Laboratory assembled an optical scene generator to project images into the telescope. Figure 3 shows a block diagram of the end-to-end test setup [1].
FIGURE 3. End-to-end test block diagram. The SBV sensor was designed, integrated, tested, and calibrated at Lincoln Laboratory, prior to its integration on the MSX satellite at the Johns Hopkins University Applied Physics Laboratory. Reprinted from the Johns Hopkins APL Technical Digest by permission [1].
A series of complex framesets were generated by a VAX workstation and loaded into a Tektronix high-resolution video monitor. These scenes contained stationary objects simulating stars and moving objects simulating satellites. The scenes simulated variable star brightnesses and several different satellite motions. The monitor images were projected through a collimator into the SBV telescope. After alignment to a CCD, the framesets were synchronized to the SBV camera's timing by using test port signals. The resulting images were collected by the camera and sent to the signal processor. The results were then checked against predictions.
The flight hardware detected all objects and sent correct message types to the ground data-analysis system. Just one problem was found; the metric positions of the simulated stars were in error. This problem was traced to distortions in the scene generator, both static and dynamic. The test was repeated for the remaining three CCDs in the SBV focal plane, and no other problems were found.
The data acquired during the end-to-end tests were input to the SPOCC data-reduction process to verify proper function and to validate that the expected accuracy could be achieved. With calibration completed, the remaining tasks were to integrate the SBV sensor into the MSX spacecraft at the Applied Physics Laboratory at Johns Hopkins University, and run through the spacecraft-level testing. The electronics were integrated successfully in March 1993 and the telescope was integrated successfully in May 1994.
Return to SBV Table of Contents