Coincidence Loss in the Grisms

The coincidence loss in the grism images has two distinct components.

The first component is that of the background, which is locally flat. Here “locally” refers to a scale of 30-60 pixels. The background varies only a little over the nominal grism image, with a significant drop only at the edges of the image. In the clocked grism image, the background varies more, since the reduced aperture cuts off zeroth order throughput at the upper left hand side of the image, and the background is mainly due to zeroth order throughput. The zeroth order is sensitive to the whole instrumental spectral range of 1600-8000A, and reduced according to the effective area of the zeroth order which peaks in the 4000-5000A range for the uv grism.

The effective area of the uv grism is much smaller than that of the visual grism, while the visual grism is not sensitive to wavelenghts below 2500A. As a result, the background photon rates in the uv grism and visual grism are not very dissimilar, and the backgrounds have a similar range in coincidence loss factors.

The second component is the spectra which form linear tracks, about 6-8 (sub)pixels wide, and under an angle to the pixels rows and columns. Their brightness may be less than the background, or much larger. The observed continuum in the brightest spectra starts showing a pattern, only approximately periodic of brightness, at the scale of about 3 physical CCD pixels, corrected for the slant of the spectra over the pixel grid. Since the physical CCD pixels are 8 (sub)pixels on which the photon counts are reported by the instrument, that means a periodicity of 26-30 pixels, depending on the angle of the spectrum relative to the pixels.

Coincidence loss in the background

  • Fordham et al. MIC detector measurements and interpretation
  • UVOT detector model computation results

Coincidence loss in the spectra

  • Measurements of coi-periodicity as function of detector mode to show period - angle relationship. P ~ -24/cos(angle) pix.

Coincidence loss due to nearby pixel

Now pull all the evidence together to propose a model for this.

  • interpretation background coi-loss
  • interpretation coi-periodicity spectral flux
  • application to point sources with a bright background to determine radius coi-area => 13 pixels radius

Coincidence loss correction approach for the grism

  • use background correction consistent with 13 pixel radius circular area.
  • use the nearby pixel correction to the coincidence loss of the spectrum + background and background only.
  • do a running boxcar average to the spectrum for calculating the coincidence loss correction, of -24/cos(angle) pixels.
  • determine the nearby coi-correction scale factor by calibration


  • Fordham et al
  • Breeveld et al