previous page   back to table of contents   next page

IV. WISE Data Processing

6. Quality Assurance (QA)

a. Steps in QA

ii. ScanFrame QA

ii.2. Frame-level QA checks

Automated QA consisted of a set of metrics judged against pre-assigned thresholds. The QA web interface contained other tabular and graphical info with which QA scientists could further judge the quality of the data. The automated frame-level metrics are described in the first subsection below, followed by the ancillary tabular and graphical data in the second subsection.

ii.2.1. Frame-level metrics.

Factors qc, qi, and qn were determined per-frameset and were the basis for the Frame Quality (FQ) score for each frameset. FQ was the minimum of the qc, qi, and qn factors multiplied by ten. FQ was an integer of value 0 (low quality), 5, or 10 (high quality).

qc: This was the quality of completion. If the status value from the ScanFrame pipeline was equal to 0 (indicating a problem-free completion), then qc=1.0. Otherwise, qc=0.0 because the frame failed processing.

qi: This was the quality of image size and shape. This factor was the smaller of two subfactors qi_syn and qi_rat. The subfactor qi_syn was a check to look for possible image smearing using the NoisePix metric. The NoisePix value was checked for the shortest wavelength band (W1), and qi_syn was computed as follows:

The second subfactor, qi_rat, was a check for possible image smearing using the ellipticity of detected sources. The value of qi_rat = 1.0 if rbovera > 0.6; otherwise, qi_rat=0.0.

qn: This was the quality of the frame as judged by its background and noise characteristics. This factor was the smallest of three subfactors qn_bkg, qn_sgm, and qn_bdp. (The qn_rfp and qn_moc subfactors, used in first-pass processing as a way of monitoring possible cases of unanticipated electronic noise, were dropped as they were not needed for second-pass processing.) These were computed as follows:

ii.2.2. Frame-level checks

In the following discussion, frame 019 of scan 01432a is used to illustrate the kinds of tabular and graphical material used by the QA scientists to assess quality at the frame level.

Arrays sometimes exhibit channels or quadrants offset in background or noise characteristics from their neighbors. Depending upon the size of the discrepancy these offsets can have detrimental effects on data quality. Figure 1 shows how any isolated amplifier channel with signal differing markedly from the rest can be detected. In W1, an anomalous channel would manifest as a set of columns with anomalous signal; in W2 it would as a set of rows; in W3 and W4 it would as a pronounced asymmetry between the lower and upper rows, and the left and right columns.

Checks such as those in Figure 1 monitored the overall distribution of pixel values to look for discrepancies. The histogram of pixel values could reveal anomalously high or low values that would skew the otherwise approximately symmetrical distributions. The passage of a bright satellite, the effects of moonlight, or the presence of a very bright astrophysical source were examples that would skew pixels to the bright end of these histograms.

Figure 1 - The histogram of pixel values in the frame, in DN units, for each band.

The plots of Figure 2 were used to search for possible smearing or streaking of point sources. If streaking or smearing was prevalent in the frame, the average difference in aperture versus profile-fit magnitudes for well detected (but unsaturated) sources would differ significantly from zero due to the mismatch between the PSF shape used for the profile fit and the true PSF shape for the streaked sources. In such cases, the aperture magnitude measurement is far less affected.

Figure 2 - Plots of the difference between profile-fitting (WPRO) and standard-aperture photometry as a function of profile-fit magnitude for all stars with valid photometry.

Figure 3 show how qn_bkg was able to detect scattered moonlight. In this case, framesets from another scan are chosen because the example scan, 01432a, does not contain any scattered moonlight.

Figure 3 - In scan 02101a, framesets 117-146 were affected by diffraction spikes and other optical artifacts from the Moon. The median signal near frames 120 and 140 -- corresponding to the diffraction spikes themselves -- exceeded the threshold of 3000 in W2, 10000 in W3, and 8000 in W4.

Figure 4 shows for scan 01997b how the qn_bkg factor can detect effects from an anneal in W3 and W4:

Figure 4 - Median or mode of the pixel values in each frame as a function of frameset number for scan 01997b. Framesets 001-010 were measured within 11 minutes of an anneal and the median signal in W4 for these exceeds the threshold of 8000, causing the qn_bkg factor to fail. The decaying mean or median signal in W3, W4 as a function of frame number is characteristic of an anneal recovery.

Last update: 2012 January 24

previous page   back to table of contents   next page