Problem Summary. One 5-pixel readout of the 160µm array displayed significantly deviant behavior during CTA and Observatory Level thermal-vac testing. This readout may not provide useful data during flight. We need to assess the impact of this on the sensitivity of 160µm images, and potentially adjust our dither patterns and/or other operations. The failures seen to date leave some hope that the entire 160 µm array may work on orbit, but there is a high probability that it will not work.
This page is primarily concerned with photometry mode. Scan map mode is addressed here.

Findings and Summary of Current Status (preliminary). Details of sensitivity and coverage analyses for the Photometry and Scan AOTs are presented below. Here we summarize those findings, based on an assumption that that readout 3 is found to be unuseable once we reach orbit.

• Fractional pixel offsets with the spacecraft result in gaps in coverage within the mosaics from the photometry AOT. The gaps in coverage do not lie within the overlap region of the mosaic, but may nonetheless be very undesireable. Half-pixel sampling can be recovered in cases where it is of interest by using a cluster AOT, or potentially by implementing a more complex frame table for MIPS, with several dither positions rather than just 2.
• Complete mosaic coverage is given if we use a 5 pixel cross-scan nod, and a 0 pixel in-scan nod.
• Coverage in the overlap region between the 2 halves of the AOT is very minimal if we use the current, 6-DCE definition, being only 5 pixels high except for one 5 pixel wide block where it is 7 pixels high.
• The stim illumination pattern results in a strong sensitivity gradient across the array. This problem is exacerbated by the loss of readout 3. There are a couple of ways we might deal with this:
• Turn the stim up by a factor of 2, and use only the right half of the array (go to a 10 x 2 array). This has the following benefits and costs:
• More even sensitivity across the overlap region of the mosaic (just a factor of 2 sensitivity contrast to contend with).
• Imaging from left half of array will be compromised to an unkown extent by saturating stim flashes. Depending on the severity of this problem, the useable FOV of the mosaic may be limited to as small as 15 x 7 pixels, with a 5 x 7-pixel wide overlap region.
• De-facto creation of a new AOT, requiring additional calibration observations and changes to SPOT/AIRE.
• Use the full array and nominal stim setting. The benefits and costs are:
• Imaging from right half of array will be compromised by poor repeatability unless we can improve our calibration strategy.
• Decent sized FOV in the overlap region (15 x 5 pixels), with a 25 x 7 total FOV.
• No de-facto new AOT.
• It may be that the best solution lies somewhere between the 1/2 and Full Array solutions.
• We have very little knowledge of the effects of the flight stim illumination gradient on repeatability. This is due largely to the fact that very little good data was obtained on those arrays during CTA and LBTC testing (temperatures were unstable and there were excess noise sources during CTA, and during both test periods exposures were taken a few stims at a time, and so suffer from the "first stim bad/Stansberry effect".
• We have no data from characterization test that directly address the questions raised here, because the illumination pattern is relatively uniform, and we have not explored the effects of saturating stim flashes.

Recommendations (strawman).
Today we need to:

• Decide whether we prefer the 1/2 Array or Full Array approach to the Photometry AOT, and whether the preference is provisional or very secure.
• Decide what nod pattern we prefer.
• Pick where we want the target to fall on the array.
• Decide if further exploration of the impact of saturating stim flashes and strong stim illumination gradients is warranted, and what the scope of any exploration should be (e.g. using spatial information in the stim flashes as well as temporal information).
• Prepare draft statement for SSC use in communicating with the GTO and Legacy scientists.

In the near-term we should pursue the following actions:

• Develop a frame table to implement a nod of 5 pixels cross-scan by 0 pixels in-scan, with our prefered target placement, and be ready to uplink it early in IOC.
• Consider repercussions for IOC activities.
• Pursue further characterization and analysis, if appropriate.

In the long-term:

• Prepare some materials for potential updates to the SOM.
• Determine whether the AOT should be changed to image a larger FOV.
• Consider how and when changes might be made.

Coverage Impact

The advertised MIPS 160 µm sensitivities and integration times per pixel assume single DCE coverage in each 1/2 cycle of the AOT, or double coverage in the overlap region of the mosaic for the full AOT. As currently defined, the AOT delivers more coverage than this in some areas. The filled FOV resulting from 1/2 of the AOT is 5' x 1.87' (20 x 7 pixels). If the scan direction is vertical, the top 2 and bottom 2 rows of the FOV are single coverage, while the central 3 rows are doubly covered (see figures below). Because of this we can consider any portion of mosaics with double coverage as not being impacted by the loss of readout 3. This conclusion is diluted somewhat by the desire to image sources on different parts of the array. I.E. multiple coverage just by the same pixel is not as favorable as multiple coverage with different pixels. If we discover that readout 3 is indeed unusable after launch, this second consideration will have to be weighed against other considerations such as providing images without gaps and reasonable coverage of the sky surrounding the source of interest.

NOTE: I've included the pre-first-stim DCE in these coverage maps, resulting in a coverage factor of 3 for the rows above and below the FOV center. This is probably not right, as in subsequent cycles of the AOT the coverage in those rows would be 2, not 3. I can change that later if need be.

The next two figures show the coverage in 1/2 cycles of the current 6 DCE AOT, and in the final mosaic. Two possibilities are shown: a 5 pixel X-scan nod between 1/2 cycles, and a 10 pixel X-scan nod. The numbers on the plots give the number of DCEs of coverage for a particular region. In the upper two panels of each plot (FOV 1 & FOV 2) the # of DCEs is per 1/2 AOT cycle. In the lower panel of each plot (Mosaic) the # of DCEs is per AOT cycle. The nominal number of DCEs in the mosaics is 2, for the purposes of considering impact on needed integration time.

A 5 pixel cross-scan nod provides a mosaic without any gaps in coverage. The 5 pixel nod mosaic also provides at least the nominal coverage (2 DCEs) over a fairly large, contiguous area. This nod also provides that double coverage using different pixels of the array within the overlap region of the mosaic. The minimum dimension of the area needed for photometry is about 5 x  pixels - enough to contain the entire core of the Airy function out to the first minimum. The middle-upper block of 15  x 5 pixels in this mosaic satisfies this criterion, although only barely in the in-scan direction. We need to decide where to place the source within the overlap region of FOVs 1 and 2, but that decision will also depend on how we choose to run the stimulator (see Sensitivity section below).

A 10 pixel cross-scan nod (pictured at right) provides a larger overall mosaiced FOV, but which has a 2 x 10 block with zero coverage. The cross-scan dimension of the overlapping region drops from 15 pixels, with the 5 pixel nod, to 10 pixels, giving somewhat less robust imaging of the sky surrounding the target. We need to decide where to place the source within the overlap region of FOVs 1 and 2, but that decision will also depend on how we choose to run the stimulator (see Sensitivity section below).

The figure at right illustrates the coverage with the spacecraft nod defined in the current Photometry AOT, 6.5 pixels cross-scan, and 0.5 pixels in-scan. The coverage pattern is yet more complex than for the mosaics with integer pixel shifts, and includes a region of zero coverage at the lower left. Given that the pixel scale at 160 µms samples the PSF quite well, we may wish to use integer-pixel nods rather than 1/2 pixel nods in order to produce mosaics without any gaps in coverage.

Stim Illumination and Sensitivity

The sensitivity of imaging with the 160 µm array will be affected by the health of readout 3 and by the illumination pattern of the stims. While the intrinsic signal to noise of an image will be roughly constant over the array, the repeatability of measurements where the stim illumination is optimal will be significantly better than where the stim is dimmer. The desired stim setting is roughly 10⁴ DN/sec, and should deliver about 1% repeatability or slightly better. However, the illumination pattern of the stim varies by a factor of about 4 from the bright end (at left in these illustrations) to the dim end (at right). If we allow some headroom for responsivity increases (say x2) and for flat field (say 40% variation), and want to avoid having the stim saturate any pixels in a 2 second flash, the brightest we can set the stim is about 10⁴ DN/sec at the left end of the array. At the right end, the brightness will then only be 2500 DN/s, considerably less than optimum. The effect of this on sensitivity in the cross-scan direction is illustrated below. NOTE: The sensitivity plots below all have the same, somewhat arbitrary, scaling. I.E. for a given target the figures reflect the relative sensitivity that would be produced by observing the target with the different observing schemes.

I've approximated the effect of the illumination pattern using the results of Karl Misselt's work on repeatability on the 160µm array. He finds that as stim brightness decreases below about 10⁴ DN/sec, repeatability worsens pretty quickly, approaching 4% at 2500 DN/sec. I've assumed that the dependence of repeatability on stim brightness is roughly linear below 10⁴ DN/sec. This approximation is illustrated by the solid line at right. The corresponding repeatability related signal-to-noise ratio, 1/sigma, is shown by the dashed line. I've assumed that the two rows in the array have the same illumination level as a function of pixel #, and the same repeatability (except for readout 3, where the SNR is 0).

Another approach is to optimize the stim for use on the right 1/2 of the array, allowing it to saturate on the left side where readout 3 is. This has the advantage that the repeatability related sensitivity contrast across 1/2 the array is only 1/2 as strong, providing more even coverage. I've approximated the repeatability for a case such as this as shown in the figure at right. The stim is set to provide a signal of 10⁴ DN/sec at pixel 11, the first pixel on the right 1/2. The repeatability worsens towards the right edge because of the illumination pattern, with the same fall-off as in the figure above. The repeatability in this scenario would worsen to the left of pixel 11 because of the effects of the saturating stim flashes (latents, short stim ramps, possibility of current-starvation of the readouts, ...). I made up a function to qualitatively reflect the poorer performance we might expect with such an overly bright stim on the left half, with the illumination pattern again playing a role, and causing performance to degrade significantly at the left edge.

Sensitivity Maps: Coverage and Stim Illumination Effects

The effect of the illumination pattern induced sensitivity gradients and uneven coverage due to a failure of readout 3 on the sensitivity of the images and mosaics from the 160 µm Photometry AOT are shown below. Examples are shown for both the Normal Stim / Full Array case, and the Bright Stim / Right-Half Array case. We need to consider both the relative merits of these approaches, and where within the maps to center sources for the Photometry AOT.

	160 µm Sensitivity Map, Normal Stim/Full Array, 5 pixel Nod
	Sensitivity maps resulting from setting the stim to give about 104 DN/sec at the left edge of the array. TOP and MIDDLE: Relative sensitivity within the image produced by 1/2 of the currently defined 160 µm AOT, which uses 6 DCEs. The effects of both the stim illumination pattern and loss of coverage by readout 3 are visible. BOTTOM: Relative sensitivity within the mosaic produced by combining the two maps above.
	Contours of signal to noise ratio pictured in the mosaic just above. The pixels are oblong in this figure (taller than they are wide). The nominal SNR in this plot that we need to achieve our advertised sensitivity is about 1.1. If we choose this mode of operation we need to decide where within the overlap region of the two FOVs to center the target.

	160 µm Sensitivity Map, Bright Stim/Half Array, 5 pixel Nod
	Sensitivity maps resulting from setting the stim to give about 104 DN/sec at the middle of the array. TOP and MIDDLE: Relative sensitivity within the image produced by 1/2 of the currently defined 160 µm AOT, which uses 6 DCEs. The effects of both the stim illumination pattern is seen on the right half. The left half shows the hypothetical effect of the saturating stim flashes, which saturate most at the left edge, and the loss of coverage by readout 3. BOTTOM: Relative sensitivity within the mosaic produced by combining the two maps above.
	Contours of signal to noise ratio pictured in the mosaic just above. The pixels are oblong in this figure (taller than they are wide). The nominal SNR in this plot that we need to achieve our advertised sensitivity is about 1.1. If we choose this option there is relatively little choice where to center the target - it would be centered up in the 5 pixel wide overlap region in order to avoid the region where the stims are saturating.

---

The BIG PICTURE: An AOT with 8 DCEs Instead of 6.

The above solutions are all impacted by the restricted FOV of the current 6 DCE AOT. Expanding the AOT to 8 DCEs gives a FOV that provides reasonable allowance for slightly extended sources and for characterization of the sky surrounding the target. The figures below illustrate such an expanded AOT, utilizing the nominal stim setting to avoid any saturation of stim flashes, in terms of coverage and sensitivity.

	160 µm Sensitivity Map, Normal Stim/Full Array, 5 pixel Nod 8 DCE version
	Sensitivity maps resulting from setting the stim to give about 104 DN/sec at the left edge of the array. TOP and MIDDLE: Relative sensitivity within the image produced by 1/2 of a hypothetical 160 µm AOT, which uses 8 DCEs. The effects of both the stim illumination pattern and coverage degradation due to the loss of readout 3 are visible. BOTTOM: Relative sensitivity within the mosaic produced by combining the two maps above.
	Contours of signal to noise ratio pictured in the mosaic just above. The pixels are oblong in this figure (taller than they are wide). The nominal SNR in this plot that we need to achieve our advertised sensitivity is about 1.1. If we choose this mode of operation we need to decide where within the overlap region of the two FOVs to center the target.

---

The REALLY BIG PICTURE: Large Field Mode

The figure below summarizes the coverage in the Large Field AOT as currently defined. Good coverage would be provided by simply doing 2 cycles with a 5 pixel offset between them (times 2 to give enough stim flashes for calibration). The AOT is taken in 2 halves, with the scan mirror chopping between Sky 1 and Object one, then a large nod by the spacecraft, and then chopping between Sky 2 and Object 2.