2012-06-24 

- Changed calculation to assume that we do A B A' B' dither pattern.
  Previously we assumed that the sky noise was derived from a single
  exposure, i.e. only local sky subtraction from that exposure.  In
  practice we find that subtracting A - B gives better results, but
  with a sqrt(2) penalty.  If you do A B A' B' then you can recover a
  factor of sqrt(4/3) back. So the S/N penalty is sqrt(2 * 3 / 4) =
  sqrt(1.5) = 1.22.  But we'll do slightly worse than that because our
  dithers between A and A' (0.4") are not quite large enough so that
  the apertures are completely disjoint.  So we will degrade S/N by
  ~1.3 relative to non-dithered case.

2012-04-17: 

 - Fixed error in throughput calculation; for typical slits the error
   in the calculation was ~10%

- Changed aperture size to be 1.4 x FWHM, which is optimal for a
  gaussian