Odd even column effect
The odd-even column effect is the difference in the flux level between odd and even columns in the Hawaii array. The cause of this effect is unknown. It was found that reducing the read speed of the array can reduce the severity of the odd-even column effect. When the read speed is reduced the minimum DIT is increased from 1.77 to 3.55s, which has a significant impact on overheads. The variation in the severity of the effect, and in the minimum possible DIT are summarized below. For service mode observations the allowed minimum DIT is always 3.55s.
| Date | Severity | Minimum DIT |
|---|---|---|
| Jan 27 2004 - present | Disappeared, read speed increased | 1.77 |
| October 25 2003 - Jan 27 2004 | Disappeared for read speed 12 (still present for read speed 6, so read speed kept at 12) | 3.55 |
| August 6 2003 - October 25 2003 | Re-appeared. Read speed decreased on August 8. | 3.55 |
| April 1 2003 - August 5 2003 | Disappeared after intervention, read speed increased | 1.77 |
| November 21 2002 - March 31 2003 | Re-appeared, read speed decreased | 3.55 |
| March 14th 2002 - November 21 2002 | Disappeared after instrument intervention. Read speed increased. | 1.77 |
| December 25th 2001 - March 14th 2002 | Effect reappeared | 3.55 |
| October 21st - December 24th 2001 | Disappeared after instrument intervention | 3.55 |
| August 24th - October 20th 2001 | Effect reappeared | 3.55 |
| July 24th - August 23rd 2001 | Read speed reduced. Effect <1% for flux range 0-20000 ADU | 3.55 |
| March 31st - July 24th 2001 | Noticeable for low flux levels, became much stronger around May 11 | 1.77 |
| Before March 30th 2001 | Noticeable for flux levels > 10,000 ADU | 1.77 |
The dramatic change that occurred during May 2001 is visible in oddeven5.ps and oddeven1.ps. The first plot shows how bias frames are affected. The lower axis is the number of days since April 1st. The second plot plots the amplitude of the effect as a function of flux for the 11th and 12th of May. The amplitude also depends on the quadrant, so the effect is plotted for each quadrant individually.
Since the effect is a function of flux, it cannot be removed by normal sky subtraction techniques, particularly when the background is varying greatly as can be the case when observing with the H filter or at the beginning of the night. In imaging, the effect can be removed very well by Fourier filtering techniques and the photometry, at the level of 1/100th of a magnitude, is unaffected. See the data reduction guide for more details.