Looks like there's some off-axis bright source. The below are frames
00503x015 and 00503x018 W1. Surprisingly, frames 016 and 017 look ok.

Examples of glints are shown below. Challenge is to find the offending culprits
and come up with a geometric model for tagging.

Frames below are a collection from:
00503x016-017   W1 and W2
00503x086-087   W1 and W2
00504x076       W1 and W2
00504x128-129   W1, W2, W3 and W4

From Mike S. (01/03/2010):
You probably have already picked up on this, but when a bright streak
from junk crosses the array you can fairly precisely map out the
geometry of the "dagger" glints from just off chip.   Interestingly,
the dagger is still quite sharp and of its usual shape indicating that
the region producing it is quite small. I've attached an example.

Date: Thu, 31 Dec 2009 18:07:05 -0800 (PST)
From: Ned Wright
To: Peter Eisenhardt, Roc Cutri, Amy Mainzer, Bill Irace
Subject: funny streak

I see a funny streak in frame 00468x192.  I put a screen grab from the
QA system on docushare.  It looks like an out-of-focus donut that is
streaked so it is very nearby.  The time was 2009-364T00:30:01.035 and the
RA and DEC were 174.584 and -30.231.  This is probably not the cover.
The donut looks to be 1.5' in diameter so the object was 0.9 km away.

Also see: 00468x230, 00468x206, 00468x146, and 00467x087

But no out-of-focus streaks in scan 00511x taken 8 hours after cover
eject, so the debris, if that what it is, is clearing out.  Very
interesting satellites streaks though.

        From:     amanda.k.mainzer@jpl.nasa.gov
        Subject:        Detector idiosyncrasies
        Date:   December 31, 2009 12:34:49 PM PST
        To:       fmasci@ipac.caltech.edu, pedro.sevilla@sdl.usu.edu
        Cc:       roc@ipac.caltech.edu, dlp@ipac.caltech.edu

Hi, Frank-

Here is another "feature" for your detector funny page.  On scan 00480x218,
bands W1, W2 and W3 show an anomalous offset over about 1/3 of the array.
On W1 and W2, it follows the direction of the outputs, and W3 has it
squarely through the center of the array.  Very strange.  I can't think of
anything electronic that would produce such an effect as none of the
electronics for these arrays have anything in common until very far down the
chain in the FEB.  

Pedro, do you have any ideas?  I have not seen this anywhere else yet, but I
haven't looked at that many scans.

Amy

Example 1 of droop in w3. Note: grey-scale is inverted.

Frames:
00503x016, 00503x017

At left is the impact from a saturating source. The droop here 
manifests itself as a depression of the whole quadrant,
as well as causing the quadrant to split in the slow read-out 
direction (I think). Note the amplified signal in the adjacent
reference pixels. Also note the donut-like "ghost" at the bottom
and the bad pixel fuzzies (same quadrant). W4 also shows these
"bad fuzzies".

At right is the next frame in the sequence, showing a "bright" latent.
This latent is gone in the subsequent frame.

What's also interesting in this next frame is that the droop reverses
direction, i.e., the signal in the quadrant is elevated. Depending on
the brightness of the offending source, this elevated signal can
continue for the next several frames or more.

Example 2 of droop in w3. Note: grey-scale is inverted.

Frames:
00503x189, 00503x195

At left is the impact of a moderately faint source (but still saturating).
The droop here manifests in a depression of the whole quadrant.

At right is a brighter source, causing the quadrant to split in the
slow read-out direction (I think), maybe at a defective pixel.
Note the amplified signal in the adjacent reference pixels.
This is not seen for the fainter source (at left). Also note
the donut-like "ghost" feature in the lower-right quadrant.

Also of interest are numerous dark latents (the white blobs in  
this inverted color map).

Frame: 00503x016 W4

Banding is the splitting of amplifier quadrants with boundaries
delineated at bad pixels. Also seen are "fuzzy" bad-pixel
regions (e.g., bottom right quadrant).

From D. Padgett, 12/30/2009:

Median W3 and W4 images were made from 90 images off the galactic 
plane for the first cover off scan (00467x) and latest scan (00483x)

Comparison of these median images show the appearance of dark latent 
images, new hot pixels, and the decay of initial transient effects 
around bad pixel clusters

From left to right below: image of CR at zscale stretch; same image
showing just the saturated core (all ~32756); the next frame in the
sequence shows a conspicuous latent - same shape as the saturated region!;
the next frame after this shows a hint of a latent - bust mostly in the noise.

If only saturated regions cause w1,w2 latents, then they should be easy to
tag/avoid by looking at the previous frame's mask.

example below: frames 00198x004,006,007 w1.

I hope they're transient and disappear with the next anneal! 
These litter all W3, W4 frames in scan 00571x and a few scans 
preceding this (see also 9 above). See animation below for 
what caused this mess.

Shown below is an animation of the culprit source leaving
one of these latent fuzzies in W3. The red circle in the fifth
frame encompasses pixels which are all saturated (region diameter
~ 50 pixels), The green circle in frames 6 and 7 shows the
bright latent. The offending source had flux densities of
148 and 780 Jy at 12 and 25 microns respectively according 
to IRAS. Perhaps repeated scanning over this same source 
caused the mess shown above. 

Those birds sure like our W3, W4 arrays!

  From:   fajardo@ipac.caltech.edu
  Date: January 12, 2010 4:02:23 PM PST
  To:   fmasci@ipac.caltech.edu

Here is an example of "fuzzy trail" or "bird dropping" latents in W3,
but in a semi-regular, repeated pattern. They must be from repeated
cross-stepping over a bright parent source. Note also the various columns
of repeated patterns of "point sources." This example comes from scan
00804a, in frames 001 through 036. Davy and I were looking at isolated
cases of the "bird dropping" latents. They are present in scans such as
00788a or 00789a, pretty much throughout the whole scan. They consist of
a single trail, very faint; the attached example is a zoomed view of the
lower right of the array.

These are seen in both W3 and W4. The example below shows the
offending star (left) and latent (right) from W4 frames:
00570x002,003.

An example of droop in W3 (like in 6 & 7 above) occurring in two quadrants 
simultaneously. Both the initial-hit frame (00570x225) and the subsequent 
frame (00570x226) are shown.

In W3, W4 frames: 00630x060 - 063. This is either some weird scatter,
or some "dark object" that progressively got closer and brighter.

Indeed! Here's what Debbie Padgett found: 

	From: 	  dlp@ipac.caltech.edu
	Subject: 	Data weirdness from yesterday
	Date: 	January 6, 2010 11:12:41 AM PST
	To: 	  fmasci, roc

Hi Frank,
  The big scattered light feature that I pointed out yesterday is
a diffraction spike from the Moon at a distance of 65 deg!
The angle of the bright feature points back at the Moon.
It's hitting us pretty hard every time we go around that part
of the orbit. It's a big saturation event in 678x at a distance
of 40 degrees. Roc, you were right about the "cleaning",
but I think it is just replacing the individual latents with
one big one.

Yikes,
Debbie

Here it is again in a later scan (frameset 00752a108):

In framesset 00620l011. This appears to be a diffracted/scattered 
glint from a bright source.

Here's a level-1b W2 frame (00631x070) which at first, we thought we had 
a new set of hot pixels. However, we believe it's a cosmic ray blast, 
possibly from a secondary shower.

   
Here's a similar blast in W4 (01020a109). Thankfully, not related to the 
one above.

This is frameset 00619a003 (W1-W4). How can this be?  
Not sure how the trail lengths in W1 and W2 can be different,
and not sure why the length in W3,W4 is shorter than W1,W2.

I think this is because the pixels in each array do not start
integrating at the same time and the read-out per amplifier
output is serial, i.e., regions from the previous exposure have
not yet finished reading out. If an object is moving fast enough,
it will be partially detected. Furthermore, W1 and W2 are
reading out in perpendicular directions. 

Here's another found by Ned: 

     From:   wright.astro.UCLA.EDU   
     Subject:        Re: funny streak
     Date:   January 8, 2010 9:46:30 PM PST

I am looking at 00752a102 and 103.  Frame 102 has a streak in W1 and W2
but nothing in W3 and W4, then there is a faint streak in W3 and W4 on
frame 103, apparently in the same position as the blue streak in 102.
Weird.

Top row: frames: 00628x283,284; Bottom row: 00629x166,167 (all W3). 
Here's what happens when a bright source hits a quadrant boundary.
Droop is seen to affect the adjacent quadrants, and a rebound (with
latent) is seen in the subsequent frames (at right). What's also 
interesting is that the quadrant splitting (below left) is not
symmetric.

A scattered glint (I think) with an artistic charm.
Frame: 00629x018-w1

  From:   cgelino.ipac.caltech.edu
  Subject: Re: another glint
  Date: January 7, 2010 2:29:11 PM PST
  To:   fmasci.ipac.caltech.edu

Check out 00724a055.  W3 has a glint and a ghost from a bright star in the
previous frame, droop in the same quadrant as the glint either from the
ghost or a bright star sitting there (complete with it's own ghost),
scattered light from the moon, and a bright latent from the star that's
causing the glint (there is probably "droop rebound" in that same quadrant,
but it is hard to see with the scaling)!  For good measure, we've thrown
in a satellite trail.

A cosmic ray with concentric "secondary showers" we presume.
This is level-1b frame 00752a244 W1. The green are NaNs that have
replaced bad and saturated pixels. These CR events also occur in W2. 
Haven't seen them in W3, W4 yet.

The first below (00748a002-w2-int-1b) has a cosmic ray shaped like a
tadpole. The second (00750a232-w3-int-1b) appears to be a cascade of
some sort. 

There's a bright star in 00772a163 (W1) with a strange halo.

A 16-spiked diffraction pattern in 00784a239-w3? Appears to be a 
record holder.

There's a gradient in W1, W2 of 00904a045 that doesn't look astrophysical.
Nothing in the preceding or following frames.

        From: 	fajardo.ipac.caltech.edu
	Subject:  example of criss-cross pattern in W2
	Date: 	January 20, 2010 1:00:05 PM PST

   Here is an example of what I mentioned yesterday at the ICS meeting. It is a 
"criss-cross" pattern in W2. Normally, you would expect to see horizontal bands 
in W2, corrresponding to the read-out channels. This example shows some of such 
horizontal bands. But it also shows vertical banding, somewhat reminiscent of 
the read-out channels in W1.
   The effect will probably be amiliorated with better flat-fielding.
   Yesterday it was asked whether the effect was seen only on the JPEGs in the 
QA pages and not on the FITS images themselves.
   Here I attach both the JPEG of the frame from webQA, and a screen grab of 
the FITS image. Both show the effect.
   Thanks!
   							Cheers,
   							Sergio

Some weird reflection seen in W2 of frame 01053a101.

Interesting sharp angle in this satellite trail. There are clearly 
two satellites, maybe three.  Looks like there's also a ghost latent
due to a satellite that passed by during scan-mirror flyback.
This is frame 01046a046-w3.

Really interesting spinning space junk on this frame (01501a240 mainly w3).
Rotation rate is about 0.2s (estimated by Mike Skrutskie).

01457b271 - appears to be something bright off the "bottom"
of the frames.  Maybe a very bright trail just off the FOV?  
Nothing visible in the next frame that would do it, so probably
not astrophysical. 
Roc

From:   fajardo.ipac.caltech.edu
Subject: example of band W1 slightly smeared PSF
Date: January 27, 2010 5:17:28 PM PST
To:   fmasci.ipac.caltech.edu
Cc:   fajardo.ipac.caltech.edu

Hi Frank,
    Here is the example of a W1 bright star that is "elongated," or more 
precisely, that has a bump to its left. Many bright stars exhibit this 
phenomenon. It is probably an optical effect.
    This case is from a QuickLook scan: 10022e109. The contour levels in the 
left panel are 3.e2, 4.e2, 1.e3, 2.e3, 1.e4, 1.7e4. The lowest level, where the 
asymmetry is inferred, is about 1.7% of the highest unsaturated signal level.
    The axial ratio in this image is 0.86, and the angle w.r.t. the x-axis is 2 
degrees.
    The star is at the top center of the frame---I have arbitrarily divided 
frames in 9 cells, to investigate the spatial dependance of this phenomenon. It 
seems to be severe away from the center of the array.
    Thanks!

    Cheers,
    Sergio

The below are w3, w4 from framesets 01568a133-137. A menagerie of wild patterns.
We're scanning close to the moon again and it aint going away.

An oddity in framesets: 01564a132-139. From Davy: 
There's a vertical stripe along the lefthand edge, most prominent in 
W2 but also present in W3 and W4, which appears in some of the 
moon-affected framesets in this scan. Must be moon-induced somehow, 
but I don't know how.

From Frank:
That W2 stripe is a puzzling one. It's not perfectly vertical and has a 
deviation of ~8 pixels. The stripes in W3,W4 are at a different location, 
about 7 W3 pixels off but do appear vertical. If these were all moon-glint 
related, it would have to originate way before the beamsplitters. Can't be 
a satellite because the stripes aren't in the same location, our 
band-to-band offsets can't possibly be greater than 5 pixels, and the 
lack of any w1 emission given the strength in w2 defies simple blackbody 
physics.

From Sergio:
Hi Davy,
    In the frames that you mentioned, the separation with the Moon is about 22 
degrees. The position angle of the Moon is 278-280 degrees (east of North). On 
the other hand, the position angle of the Y-axis of the frame is about 16 
degrees (east of North). Consequently, the vector from the frame to the Moon 
points nearly downwards (by "nearly" I mean about 5 degrees, so the Moon is 
mostly down and slightly to the left).
    The same goemetry applies to the case Frank mentioned: 01565b frames 
158-163.
    By these remarks I don't mean to completely explain the feature, but just 
add info. in case you want to think about it further.
    Thanks!

Shown below is frameset 01564a135. W1,W2 (top), W3,W4 (bottom).

As the title says. Not as strong in early survey scans. Most prominent in 
the top-right quadrant. The variation is <~ 0.6% peak-to-peak on a scale of 
~10 pixels. The phase (stripe locations) vary frame-to-frame and appear
to be an additive effect. On consultation with SDL (MMR mtg 04/15/2010), it
was suggested this could be a consequence of the gradual warm-up of the W3 
optics and/or electronics for these scans since the S/C is tilted more towards
the Sun than normal. The below is frame 10039a127-w3.

Could these be zodiacal dust bands? 

Seen in:
10039a121-123
01664a118-121
01662a069-070

This is consistent with their ecliptic latitude range: ~0.06 - 5.72 deg. 

A later note from Chris:

From:   cgelino.ipac.caltech.edu
Subject: Re: bestiary updated
Date: May 3, 2010 1:48:53 PM PDT
To:   fmasci.ipac.caltech.edu

Frank,
A note on Idiosyncracy #35 (sporadic horizontal banding in W4).  I do not 
believe that it is zodiacal dust bands.  I just came across a set of 
images (00765b152-168), that shows this banding.  The banding appears to 
be stationary for a few frames, disappears for a few frames, and then 
comes back for a few more frames (stationary across those few frames). 
It should be moving between frames if it is something in the sky. 
Curiously, these frames are all negative elat like your example.
Chris

A mystery!

Appears to be a strong CR with some sort of charge bleed. Haven't seen this
before. Frame below is 10072b112-w1-int-1b.jpg.

   From:   fajardo.ipac.caltech.edu
   Subject: curious feature in a frame with a bright trail
   Date: March 30, 2010 5:04:43 PM PDT
   To:   davy.ipac.caltech.edu, fmasci.ipac.caltech.edu

    There is a gigantic trail in W3 and W4, which causes some uncorrected 
split-quadrant droop in W3. There is also a "source" next to the trail, with 
what appears as an off-shoot (has a cometary appearance - a glint?).
    There is no known solar system or astronomical object at the 
position of the source. It does not appear to be a latent, since the 
previous frame does not show a feature at the corresponding pixels.
    It could be the analog of a "smoke ring" ghost, perhaps from a hot spot in 
the trail, off the frame just to the right.

    Attached is frame: 03054a047-w3-int-1a.jpg

Here's another corner glint (frame 01109b003-w3).

03073a172-w3 appears to have a very bright satellite. It has flyback and 
what appears to be two ghosts. Even the ghosts leave latents in 
subsequent frames.

Is this related to #36? There also appears to be another "satellite latent".

Frameset 03069a146 (w3 shown below) has two satellite trails, one of which 
is very bright and has mirror-flyback effects. Notice also the weird 
horizontal streak we sometimes see when bright things go through the FOV.

This blotchy pattern appears to occur for 15-30 frames at the start or end of
scans acquired after 03080a (~Mar 25). It is strongest in scans 03260a,
03257b, 03256a, 03258a. See also menagerie #26. Weird.

Following deployment of v3.5 on ops, dynacal includes an algorithm to
subtract a running-sky (zero-biased) median to correct for temporal variations
in the bias structure over each array. Consequently, it also does a pretty good
job at subtracting long-term latents when the sky-offset is computed close
to the frame of interest. Since the long-term latents are gain-related,
this subtraction mechanism is not quite right. Therefore for scans with 
a strong background gradient, the end points of scans may get more "sky"
subtracted than normal, leading to "holy" L1b frames (see below). There
is a plan to ameliorate this using a running flat in v4.0. The example
below is frame 03293a001-w3.

Here's a satellite trail that bisected a stationary-split associated with 
a bad pixel cluster in the bottom right quadrant of W4. These splits are  
corrected in the pipeline (along with the droop-induced splits) by using
signals on either side of the split. The signal from the satellite
has messed up the correction. This has occurred twice in the mission so
far: frames 03196a205-w3 and 03173a135-w3. The frame below is the former.

Hi Frank,
I have an odd frameset for you.  Frameset 03389b235, all bands.  There's a 
vertical black stripe in the same spot in bands 3 and 4.  The stripe is on 
the left edge in band 2, then is horizontal in band 1.  There seems to be 
a hint of an inter-frame satellite there, but I'm not sure what could cause 
this structure.  The scan does have moon artifacts, but there's nothing 
weird going on immediately before or after the frame.  The horizontal 
W1 points to an issue in the electronics.
-Doug

Follow-up from Frank:
Something like this (or the inverse rather) is seen in menagerie #5.

We hit Mars (or it hit us) and here's the damage. A very impressive display
nonetheless. Mars' dad (Jupiter) is lurking out there somewhere. The below 
are frames W3 and W4 from frameset 03881a129.

Here are latents from satellite trails that wrecked havoc with the
droop correction (similar to #45). These are frames 03561b034-w3 and
03697b267-w3.

Here's an interesting CR event (or events) that lead to a latent in the
subsequent frame. These are frames: 03581b223-w3, 03581b224-w3.

The culprit here is the strong latent (from hitting Mars) that occurs
exactly over the stationary banding split in the bottom right quadrant
of W4. This is frame: 03730a047-w4.

This is frame 04252a267-w3 and is ~ 3 deg from the SEP.
Moon-glow is unlikely.

Frame 03770a245-w3 (first below) has an elliptical dark hole, with vertical 
bleed, and leaves a latent in the next frame (03770a246-w3). Also seen in w4. 
There's nothing in the preceeding frames.

Seen in w3,w4 of frameset 00964a070. Weird moon (or other bright object)
glow perhaps.

Just like #49. Below we have w3,w4 of frameset 01349a017.