R4K Quadrant Noise

Wednesday, May 4, 2016 11:09 AM

Problem(s) Encountered:

The upper-left quadrant (as oriented when opened in ds9) of R4K has developed a rather extreme increase in noise, as though the readnoise jumped by an order of magnitude.

To illustrate this, I subtracted the overscan of a bias frame with proc4k.py and plotted a random row in the upper half (row 610 of a rroi4x1k image).  In the right side (corresponding to the upper-right quadrant), the mean is 0.14 with an RMS of 1.9 counts, very much what we expect.  On the left side, the mean is -0.29 with an RMS of 13 counts.

See here for the dramatic increase in noise in the left half compared to the right:  http://www.phy.ohiou.edu/~chornock/science/bad-bias.tiff

Or get the FITS file here: http://www.phy.ohiou.edu/~chornock/science/n1.0002b.fits

This electronic noise appears to be additive, not multiplicative, and it is sufficiently close to random that averaging N biases lowers the RMS by sqrt(N) without revealing any clear pattern.  However, the amplitude is sufficiently large to make that quadrant of the detector useless for faint object spectroscopy.

I looked back at my biases over the past year and found that this was not present last fall. In March, that quadrant was noisier than the others, but at a lower level than it currently is.  So clearly there is a bad trend.

I was doing long-slit observations of single objects, so I put objects in the bottom half to avoid this bad quadrant.  However, the lower-left quadrant has its own issues.  There is obvious pattern noise after subtracting the overscan which is not stable from frame-to-frame, but which you just have to live with.


No solution really exists for this yet, but this is some information pertaining to tests performed the next day:

Post Amplifiers for each quadrant (one per quandrant) were swapped around to see if the issues would follow PAs.  It did not appear to.  Interestingly though, when they were put back in their original configuration, the problem appeared to improve by roughly a factor of 2. Unfortunately, through the night, the problem appeared to once again worsen.