CrIS noise and its effect on the quality of measurements have been calculated and tracked by SDL through several methods. First, spectra collected in 112 contiguous samples from deep space (DS) and internal calibration target (ICT) views are used to calculate system noise. These are related to Earth scene (ES) geometries for easy comparison against ES measurements by accounting for CrIS optical transmission and field of view (FOV) properties, as well as detector responsivity. The NEdN is calculated as the standard deviation of the sample set, while the Allan deviation is the average difference between successive values. Significant differences between these two values indicate drifting signal levels over time scales less than the dataset length (15 min) and greater than the difference between successive measurements (8 sec). An example of such a drift is the change detector irradiance in ICT view as the ICT temperature changes—variations in ICT temperature up to 0.1–0.2 K over 15 minutes are routine, causing significant differences between NEdN and Allan deviation. Therefore, the measured radiances from the ICT view are temperature corrected prior to the NEdN calculation.
Principle component analysis (PCA) can also be performed on the DS and ICT total NEdN results to determine the amount of noise coming from sources with random noise and spectrally correlated noise. Random noise strongly dominates the signal noise for CrIS under typical conditions; significant contributions from spectrally correlated noise is a good indicator of sensor vibration that may affect signal quality. The PCA method may also be used to calculate the random noise, but not total noise, from 2,000–3,000 ES spectra. The imaginary component of the signal has proven to be more sensitive to the various instrument mechanisms that produce noise. PCA analysis and the NEdN and Allan deviation are applied to both the real and imaginary components in an on-going basis.
In addition, the raw CrIS system responsivity has been tracked over the satellite lifetime. Responsivity values over time are compared against those recorded at the first of the space flight, providing a measure of sensor sensitivity degradation throughout on-orbit operations. Specific spectral windows are examined each time to monitor for the build-up of spectrally relevant contaminants, such as ice, on system optics.
Zavyalov et al. (J. Geophys. Res. Atmos., 2013, 118: 13, 108-13, 120) presented the NEdN and PCA results of CrIS modeling, pre-launch testing, and trending during the first year on orbit. In summary, ground testing of the system in thermal vacuum (TVAC) yielded spectral NEdN lower than the specifications in the system requirement document for all bands and FOVs, except mid-wave (MW) FOV number 7 (FOV7). CrIS measures radiances in long-wave (LW), MW, and short-wave (SW) infrared bands using nine FOVs in a 3x3 array configuration. In addition, system temperature was found to influence noise levels during TVAC. Modeled CrIS noise overpredicted NEdN for LW and SW bands by a factor of two in comparison with NEdN calculated from on-ground testing, while MW predicted NEdN was close to or slightly under measured NEdN.
Zavyalov et al. (2013) reported on-orbit noise performance maintained NEdN levels observed in pre-launch testing, meeting requirement specifications and remaining constant through August 2013. PCA analyses demonstrated that random noise contribution comprised essentially all of the total measured noise. Such consistent and low-noise levels, in conjunction with the results of the PCA analyses, contributed to the decision not to engage the vibration isolation system available on S-NPP.
Continued S-NPP CrIS noise characterization at SDL has yielded results fairly consistent with those reported by Zavyalov et al. (2013) for the first year—CrIS noise levels remain low and below specification levels and the instrument is still performing well. Calculation of the Allan deviation on ES-projected system noise from DS and ICT views has been implemented, with calculations performed for the full on-orbit lifetime. DS and ICT NEdN and Allan deviation, as well as raw system response, have been calculated every sixth day since June 2012 as frequent, routine checks on system noise. Current spectral DS and ICT NEdN levels continue to meet requirement document specifications, with the same exception of MW FOV7. Linear fits to both NEdN and Allan deviation datasets over the five years on-orbit reveal slight, steady increases in both real and imaginary signal components. All noise levels have increased at a rate of about 0.5% per year in each band. Continued PCA analyses reveal contributions to total noise are still dominated by random noise. Relative system responsivity remains within 0.5% of the initial responsivity for LW and MW bands; responsivity in the SW band has degraded by up to 5% below initial levels. There is no indication of contaminant buildup.
The S-NPP CrIS sensor continues to have low-noise levels, with little contribution from spectrally correlated noise. Lessons learned from S-NPP noise characterization activities have been and will continue to be applied to the CrIS instrument on the follow-on JPSS sensor, J1.