The detector is read out by four amplifiers. The even and odd columns of each quadrant are read out in two output chains, each with its own gain and bias level.
Figure 1.1 Schematic of the 8-channel readout geometry for the 8192 x 3188 pixel full-frame detector. A 48-column prescan region is shown; for subregions of interest used for imaging and prism modes: e.g 3K x 3K, 1K x 1K and 4K x 3K; there is no prescan. (This figure was taken from the MODS Instrument Manual).
This produces a vertical striping on the raw images, the ‘even-odd’ effect. This can be removed using the overscan (actually, the prescan) regions through the task, modsBias, which is part of the modsCCDRed package. While observing, copy your image to a /scratch directory and run modsBias on it:
The output will be an overscan-subtracted and trimmed image with the suffix, “_ot”, appended, e.g. mods2b.yyyymmmm.nnnn_ot.fits.
Figure 1.2: A zoom on the central region of a raw MODS2 Blue Channel image, to illustrate the “even-odd”. effect. This image also illustrates the quadrant-to-quadrant differences, also resulting from differences in the gain and bias levels of each amplifier. Bias subtraction (or “overscan”-subtraction in the case of the full-frame images) and flat-fielding should remove these differences, albeit not perfectly.
Figure 2.1: A MODS2 Blue channel image with amplifier glow in all four corners. Counts along the green vertical line from top to bottom are plotted to the right. Full-frame images from both the Blue and Red channels of MODS2 show amplifier glow, but in this particular example the glow is relatively bright. Neither Blue nor Red channel images of MODS1 show amplifier glow.
There are two levels of saturation: 1/ counts greater than or equal to 65535 ADU, where the 16-bit ADC converter tops out; and 2/ counts above full well, which is 200,000 e- or (80,000 – 117,000 ADU, depending on the gain), where charge spills over into surrounding pixels.
The four MODS detectors (MODS1 Blue, MODS1 Red, MODS2 Blue and MODS2 Red) respond differently to saturated stars and objects.
When counts just exceed 65535, the profile will be flat-topped and, on MODS1 Blue and MODS2 Red, there will be vertical trails, extending from object down to the middle rows (the last row read out). These trails appear in subsequent images, but now extending from the object to the top and bottom edges of the detector rather than the middle rows. They diminish over time.
Figure 3.1: Saturation effects on a series of MODS2 Red channel images. Stars which peak at >65535 ADU leave trails which extend to the middle rows of the detector, the last ones to be read out (top left; mods2r.20161118.0047.fits). On the subsequent slit image, the trails persist, but this time going from the object to the top or bottom edge (top right; mods2r.20161118.0048.fits).
Figure 3.2 (below): The trails can even persist to the first spectra, but they diminish over time.
Figure 3.3 (above): The montage of deep, 600-sec, MODS1 g_sdss and z_sdss images g (top left and right) and 200-sec g_sdss and r_sdss MODS2 images (bottom left and right) illustrates saturation effects and stray light arcs. The very saturated star in the upper right quadrant of the MODS1 images has B=12.2, V=11.6, R=11.2. The star to its left, within a cyan circle of radius 10” (for scale) has B=14.6, V=14.0, R=14.7 and the circled star in the bottom right quadrant, above the left part of the guide probe shadow has B=15.8, V=15.4 and R=15.4. Cyan arrows point to stray light arcs which, in the MODS1 images, these have count levels ~400 above the background and probably are the source of an additional artifact on the background (in the lower left quadrant as indicated by an arrow). The encircled star on the MODS2 images has B=13.92, V=13.64 and R=13.55 and arrows point to stray light arcs seen on these images as well.
Figure 3.4: is a zoom about the 4 encircled stars in Figure 5a. In the MODS2 Red image (lower right) that is part of a dithered sequence, the persistent trails left by saturated stars are visible.
Heavily saturated stars affect the image along rows as well. If the star is on or near a quadrant boundary, then the rows in both quadrants are affected.
Figure 3.5: This MODS2 Red image (top) illustrates how saturation on a quadrant boundary can affect entire rows.
Figure 3.6: These two MODS1 red 300-sec z_sdss images how how a heavily saturated (R=8.4) star in the upper left quadrant affects the rows in that quadrant (left). When it falls on the quadrant boundary (right), it affects rows in both quadrants.
Readout Artifacts in binned images
With the current generation of MODS CCD controllers, binning produces electronic image artifacts. These artifacts are particularly noticeable in the 2×2 binned spectra of bright targets, and, for that reason, LBTO does not support 2×2 binning. These may have a similar origin to the saturation effects, where, in this case it is the accumulated charge along the row, the fast readout axis, that when it passes a certain threshold affects the readout. As the spectral trace is closely aligned with the row, it is easy to surpass this threshold for bright targets when binned along the dispersion axis.
For programs requiring a high cadence series of observations, binning along only the spatial, but not the dispersion axis (1×2 binning, ccdbin 1 2), is recommended as a compromise which will reduce overheads but not produce electronic artifacts. Read+write overheads are reduced by a factor one-third for 1×2 binned data, and a factor one-half for 2×2 binning as compared to unbinned (1×1) data.
The images below illustrate these electronics artifacts on MODS1 Red and Blue channel images. As the MODS2 detectors use the same types of controllers, we expect similar artifacts on 2×2 binned MODS2 images, however, we have not examined these as carefully.
Figure 4.1: The dark bands on the 2×2 binned MODS1 Red channel spectra (the top 2 panels: the blue end of the red channel spectrum is at top, and the red end of the red spectrum, in the middle) are binning artifacts. These dark bands are highlighted by the green rectangles and appear as sharp dips around lines ~860 and 990 in the column cuts. They run parallel to the rows, across the entire spectrum and into the prescan regions, even. On 2×2 binned MODS1 Blue channel spectra (bottom), there are similar bands, but in this case they are brighter than the background (to see these, you may need to click on the image to enlarge it).
Figure 4.2: Artifacts on 2×2 binned MODS1 Red channel spectra. The horizontal segments appear random in length, position and counts over the background, with the exception that they are centered on the vertical quadrant boundary.
IR laser spots
The MODS2 Blue and Red spectroscopic images show vertical sprays of dots which, at first glance, look quite a lot like nebular emission. But upon further examination, they can be seen to be evenly distributed along the dispersion axis. They may be difficult to spot in raw images, but once the even-odd effect is removed with modsBias, they are easier to see. Further investigation has identified the origin in the 1.55 micron IR laser used by the flexure compensation system. Even with the very low near-IR low quantum efficiency of the blue and red channel CCDs, the IR laser can be detected in MODS2 images. In MODS1, there is no sign of emission from the IR laser. Though the problem has been identified, it has not yet been resolved.
Figure 5.1: A MODS2 Red channel raw 2D spectrum. The green arrows point to the quasi-regularly spaced vertical sprays of dots which are due to the IR laser.
Figure 5.2: A series of darks, taken with the IR laser on and enabled and the MODS2 Blue channel configured both in direct-grating (top) and dual-grating mode (middle). Both show the vertical spray of dots. There is no emission when the laser is off. The quasi-regular repetition along the x-axis has its origin in the multiple orders of the spectrum reflected from the bypass grating. Why the emission in each order is seen as a vertical spray and not a single dot is unclear.
Dichroic ghosts show up ~145 pixels below the source position and are stronger on the Blue channel than on the Red channel images.
Figure 6.1 shows these ghosts below a few extremely bright lines in a MODS1 Blue channel Xenon comparison lamp spectrum. And Figure 6.2 shows these ghosts at the blue (right) end of a MODS1 Red channel spectrum of the spectrophotometric standard star, Hz44 (on raw Red channel spectra, wavelengths increase from right to left).
Figure 6.1: MODS1 Blue channel MOS prism flat field taken with the continuum QTH lamp. The dichroic ghosts are circled in green (red circles are left over from a previous frame). This image also illustrates the red leak through the dichroic, which is enhanced by the non-linear dispersion of the prism, where the number of angstroms per pixel increases towards the red. From left to right, the intensity of the spectrum through each slit increases, then falls sharply at the dichroic notch, however it picks back up as the lamp spectrum is concentrated more and more. The three-part banded structure to the red of the dichroic notch arises from the few percent wiggles in the dichroic transmission curve.
Figure 6.2: This MODS1 Red channel spectrum of the bright spectrophotometric standard, Hz44, illustrates the blue leak through the dichroic and the dichroic ghost that is 145 pixels below the target spectrum.
Light from bright stars (V<~5, but this depends on the image depth) within an (1.9-4 deg) annulus about the target grazes the outer edge of the undersized adaptive secondary mirrors on both SX and DX sides and produces diffuse arcs on both MODS1 and MODS2 images. The outer edge of the arc is determined by a stop within MODS and the inner edge by the rim of the adaptive secondary. The arcs move as the instrument rotates to follow the sky. When the slightly larger rigid secondary is mounted, these arcs are minimized or not seen.
Installation of a baffle around the adaptive secondary would be needed to eliminate these stray light arcs. The baffle would have to be temporary/deployable because emission from it would hamper near- and mid-IR observations at wavelengths longer than K band.
For now, observers can offset the target away from the arc, though this will not insure it will remain free from contamination throughout the observation.
PIs and observers can search for bright stars within a 1.9-4 deg annulus of the field center using a web-based vizier search or, if the cdsclient package has been installed, using the vizquery command.
vizquery -mime=text -source=V/50 -c=Jhhmmss.sss+ddmmss.ss -c.rm=114-240 -out.add=’_r _raj2000 _decj2000′ -out.max=unlimited -oc.form=d Vmag=0..5
will query the Bright Star Catalog (Hoffleit et al. 1991, I/50) for stars with V=0-5 in an annulus 114-240 arcmin about the coordinates hhmmss.ss and ddmmss.ss at center.
A particularly egregious example of these stray light arcs is shown below, for a field that was only 2 degrees from the V=-0.05 mag RGB star, Arcturus. Figure 6.1 (below) illustrates how the the arc differs between the blue and red channels and its location changes as the instrument rotates.
Figure 7.1: The top panel shows how the arcs in the initial MODS1 Blue acquisition image, taken at PA= 60 deg, disappear when the instrument is rotated by 180 degrees, to PA = -120. However, the effect of the arc shows up prominently on the series of MODS1 Red channel spectra, and images taken immediately afterwards show that, for the MODS1 Red channel, the arc contaminates the light which is entering the slit.
Gradient in MODS2 internal imaging flats
The internal calibration unit in MODS2, unlike that in MODS1, does not provide an even illumination. The counts drop by 20% from the center to the edge of the 3K x 3K imaging field of view. If a uniform imaging flat is needed, for extended objects, e.g., then it is recommended to obtain twilight sky flats with MODS2.
Figure 8.1: Imaging 3K x 3K flats (sdss g) for MODS1B (left) and MODS2B (right). In the MODS2B flat, the illumination falls off by ~20% from the center to corner of the image, but in the MODS1B flat, the illumination is approximately flat (in fact it is a little brighter at the bottom right).
Central rows noisy (sequencer fault)
Every once in a while, an image appears in which an equal number of rows above and below the central row are noisy, and the noise pattern is symmetric about the central column (see Figure 9.1 below). This is indicative of a sequencer fault.
We have experienced this problem at one time or another on all of the MODS channels. It is typically intermittent, affecting one image but not the subsequent one. Please report images like this. The sequencer board may just need to be reseated, or fibers checked, but unfortunately, it the sequencer may instead need to be replaced. Work is being done towards using a different CCD control architecture, and this would eventually eliminate this problem.
Figure 9.1: MODS2 Red channel image which shows the noise pattern symptomatic of a sequencer fault.