There is an intermittent issue affecting the lower right quadrant (Q2) of MODS1B, where the counts in even and odd channels differ by hundreds of ADU (>~450 ADU).  Because of the readout strategy employed: the detector is read out by four amplifiers, each through two output chains, one for the even and the other for the odd columns, there is normally some vertical striping seen in the raw images, since the bias and gain of the two chains do not match perfectly. However when this issue arises, the striping dominates the raw image. (In pixel and imaging flats where the average count levels are high >15k ADU, however, the effect can go unnoticed in the raw data and appears only in differences or ratios with good images).

This instability is transient and can affect the entire quadrant or only a fraction of rows, as the examples below show.

Figure 1: On the left, a MODS1 Blue 3K x 3K image with the entire quadrant 2 affected. A cut on line 1000 from column 1500 to 1600 (magenta bar in image) illustrates the larger than normal even-odd level difference in quadrant 2. The dark rectangular shadow is of the guide probe. On the right, a raw full frame 8K x 3K MOS spectrum with a fraction of quadrant 2 affected.

This issue first arose in February 2022, but then its frequency tapered off and it was not seen from April through the end of the semester. It came back in mid-October and is intermittent.

On full frame images where the entire Q2 is affected, modsCCDRed will remove the pattern, as it uses the bias levels in the 48-column prescan region. When only a fraction of rows are affected, however, a more customized approach is needed. When only a portion of the detector is read out (3K x 3K, 1K x 1K images or 4K x 3K prism), there is no prescan region to use to remove the effect.

We have not identified the cause or measures to take to stop the behavior, but will update this note if and when we do.