From: Andrew C. Phillips
Subject: Slit images seen in internal flats
Date: Tue, 15 Oct 2002 14:25:29 -0700

Some of us have noted features in some internal flats that look like "arcs" of light and vary in number and position depending on which slitmask is in place. These have been seen with the 900-line grating. They appear only with internal flats; identical-setup dome flats are clean. I spent some time after the last DEIMOS run investigating these. Here's what I've found:

  1. With the 900-line grating (Slider 4), the arcs move as the grating is tilted,thus, the grating is clearly involed.
  2. Putting in the 1200 line grating (Slider 3) at the _same_ grating tilt (-5.65 degree), the same arcs show up. This means whatever we're dealing with is purely geometrical, ie, it is not specific to a given grating or slider.
  3. The arcs actually look like images of slitlets, and in fact in one case can clearly be identified with slitlets on the mask. Given the large size of the images (~1000 px long) I suspect we are dealing with light passing through the slits and being imaged directly onto the detector.
  4. A slitlet 10.3 mm long produces an image on the detector about 22mm long, ie, a magnification of about 2.
  5. The arcs seem to have a radial distortion to them, that is, straight slits produce slightly curved images (at least those off the X-axis do).
  6. There is usually a fainter image (and in some cases a third, even fainter image) at higher image-Y location. These secondary images may be slightly smaller in size.
My current thinking is that something (probably the flat ring around the camera mouth) is reflecting light onto the grating which then reflects the light in zeroth order into the camera. The path-length from slitmask to camera mouth is about 10 in, from camera to grating is 11.5, so the full path-length here (neglecting transverse distance) is 33 inches. The camera focal length is 15 in, and I assume this approximates the distance to the detector. The ratio 33:15 is very similar to the magnification we see. Thus, my hypothesis seems reasonable and explains observed behaviors 1-4. One constraint here is that the images should be in very bad focus (focus should occur at about 30" image distance), so the beam producing the images must be very slow.

Another check on this has to do with angles: the path between slitlet and camera edge is at an angle of 14 deg wrt DEIMOS Z-axis. If I'm doing all my arithmetic right, this means it enters the camera at an angle of about 2-deg wrt the camera axis, clearly within the acceptance angle of the camera. Assuming (correctly, I believe) that the grating is on-axis wrt the camera axis, this light should strike the grating about 3.8-in off center; since the grating is 8-in total length, this means the light does (barely) hit the grating as well. Thus, this also supports my hypothesis, since the proposed path will indeed throw light into the camera.

I suspect the "curved" appearance has to do with the fact that the reflecting surface is a ring, and since the images are far from focus, their location will depend on what area of the pupil is illuminated. For most straight slitlets, the portion of the pupil illuminated will follow an arc (since the reflecting ring is narrow and circular) and thus the out-ot-focus image produced will also be an arc.

The beams will be slow -- as I recall, the ring is about 1/4" wide, so at 10" distance this is f/20. I'm not sure if this is slow enough to explain the images we see, though. However, the focus will be much better across the slit than along it, and this is indeed observed.

I'm not sure about the other, fainter images seen -- perhaps someone can check if there are other surfaces near the camera mouth that could reflect light along a similar path?

In summary, I believe the "ghosts" seen in internal flats with the 900-line grating are due to reflected light off the steel ring around camera mouth.

Recommendation to Observers

Until the modification is effected, observers should use dome flats only for the 900-line grating, and other gratings that work at a similar tilt. Observers should be aware of the problem and watch for it in their data.