9.8 Removal of Spectroscopy Calibration Lamp Signature

The DFOSC spectroscopy calibration lamp was accidentally left on while taking 8 of the photometry science images, i.e. with the DFOSC in imaging mode. These images are listed in Table .

 

Table: Images affected by the spectroscopy calibration lamp.

Image	Filter	Field
d1549	GR	00
d1550	JB	00
d1552	GR	14
d1553	JB	14
d1555	GR	15
d1556	JB	15
d1558	GR	13
d1559	GR	13

 

It was not possible to see the lamp signature in the Johnson B images - the lamp must have been very red - so only the 5 Gunn r images needed to have the lamp signature removed.

By coincidence, an image, dfsc1205, had been taken earlier, also with the lamp on. This image was binned $4 \times 4$. This image was expanded to full size using blkavg, had bias subtracted, had some cosmic-ray-events removed, and was trimmed. It is shown in Figure  (left).

 

However, the lamp signature in the science images did not look exactly like d1205. The bright small ring in the middle right hand side of the image matched well if it was shifted, but the arcs in the right hand side was symmetric in the science images, whereas it was asymmetric in d1205.

The lamp correction image was constructed as follows:

• The ring was cut out of d1205, and shifted in x and y, and scaled in intensity.
• The symmetric arc was taken from the science image d1555 of field 15, where the arc is free of galaxies in the image. Unfortunately, field 15 is next to the central field, field 00, so at the position of the arc, there was a contribution from (especially) the very bright galaxy R256/NGC3309 A model of this was made by simply taking the corresponding part from the field 00 image d1278 (since the fields overlap), translating it to the corresponding location in d1555, and subtracting the d1278 background level to get only the galaxy contribution. The shift was computed using 3 stars in common to the two images.

  The arc could now be cut out of d1555. The d1555 background was subtracted to get only the arc contribution. A number ($\sim$ 50) of weak objects was removed by replacing the affected pixels by a linear interpolation of the neighbor values, using fixpix. The image was then median filtered using a $11 \times 11$ box, using the task median.

• The low frequency variation was fitted to d1555 with the ring and the arc subtracted, using imsurfit. The objects (stars and galaxies) in d1555 was excluded from the fit by specifying the parts of the image not affected by galaxies, and by doing a sigma-rejection for the rest (setting low=3, up=2, and niter=3). The function used was a $2 \times 2$ piece cubic spline.

The three parts were put together to the image lamp_5mGR, which is shown in Figure  (right). This image has a maximum value of 46 and a minimum value of -7. It was subtracted from the 5 affected Gunn r science images mentioned in Table  - no scaling was necessary. The correction worked remarkably well.