ghosts in ULTRACAM (Vik Dhillon & Mark Stevenson, February 2001)





The dichroics in ULTRACAM will produce ghost images, as shown in the diagram below.



The u' ghost is not too bad (0.0006%), but it can be seen that the g' ghost will be significantly worse (0.1764%). A similar thing applies to the r' arm, which I have not drawn here. We contacted Tully about this, who modelled the effect using Zemax. Here is the resulting ghost spot diagram and here is his accompanying email message describing the effect. Tully found that the ghosts are indeed as strong as indicated above, but they are formed in a collimated beam. This means that they will fall on top of the main image at the CCD and will be in focus, so we will not notice them. This is not quite true of the extreme corners of the field (see the spot diagrams). There, the ghosts will be severely chromatically aberrated and spread over a disc ~8 times the radius of the main image (and approximately centred on it). Because of this, it is unlikely we will notice the ghost. The edges of the field (not the corners) will, however, be fine.

Another problem is polarisation. The dichroics are designed to work for random incident polarisation. If the light is 100% linearly polarised in a plane perpendicular to the long axis of the dichroic, then the dichroic cut point will move by ~??nm compared to incident light polarised in the orthogonal direction. This has been confirmed to us by Dr Helmut Kestler of CVI Technical Optics, who are manufacturing our dichroics. This means that if we are observing polarised objects, the dichroic cut will move around, giving us variable flux in the bandpasses. There is not much we can do about this when observing polarised objects apart from:

  1. using a narrower band filter which lies completely above/below the (moving) cut-point.
  2. putting a rotating half-wave plate in the beam. Note that the short exposure times we will be using means that the HWP will have to be spinning very rapidly.
  3. avoid observing magnetic systems! It is a possible that some sort of crude correction might be feasible, where one looks at the anti-correlation in the flux between two adjacent pass-band (on the assumption that the instrinsic flux variations in the different wave-bands are the same).