Flat-field periodic structure¶
(29 July 2013, Gary Bernstein & Andres Plazas)
If you take a median of each column in a DECam flat and plot the results vs x, you will see a periodic series of spikes at the part-per-thousand level. The period is 27.33 pixels (perhaps 27.31?). Below is the row-flat for CCD 53 in g band, after
- shifting the right half to match the gain on the left half
- removing an overall slope
- folding on an 82-pixel period by taking the median pixels i + 82*j for j=1,...24
A sawtooth-like pattern with amplitude 0.002 peak-to-peak is apparent. A smaller 2-pixel period at ~0.0001 pixels also looks to be present.
Presumably this is a variation in pixel size. Such effects have been detected in other CCDs. No periodicity is apparent in the y direction.
This is the same signal seen in the lab by Tom Diehl: see docdb-3085
The effect on measurements¶
This pattern is present in the flat fields but is not a QE variation (probably). Dividing by the flat field would therefore be expected to generate a spurious fluctuation in the measured fluxes of objects. But the amplitude is 2 mmag p-p, 0.5 mmag RMS, well below the current floor in photometric accuracy. And of course dithering will reduce this further.
The pixel-size variation also induces an astrometric error if it is ignored. A pixel-size sawtooth with 0.002 pixel amplitude and 27.3 pixel period will generate an astrometric error up to 7 mas, but the RMS will be several times smaller, just 1-2 mas, so also currently in the noise.
The asymmetry of the pixel-size variation will generate a spurious shear on objects, with a peak-to-peak variation of 0.002. This is actually not that small as far as weak-lensing shear goes, even if the RMS is smaller. However it will be suppressed by (a) the finite extent of galaxies smearing out the sawtooth a bit, and (b) dithering pointings moving each galaxy around the sawtooth. So I believe we are safe in ignoring this. Plus the period of 27.33 pixels is only ~7 arcsec, smaller than any signal we'll be looking for.
However if we discover at some point that the Universe has periodic 7" ripples of dark matter oriented north-south, we might want to keep this in mind.