Note: the temporary C4 CCD is documented elsewhere.



The pixel values in the images from C0 and C1 are the signal read at the end of the exposure. They are not divided by the exposure time à la ESO.

Software Gain

The read-noise determined from raw bias images is about 9 raw DN for C0 and 13 raw DN for C1. This oversampling of the noise leads to relatively poor compression, with bzip2-compressed sizes being about 40% of the original sizes. This is a problem since the Internet link to the OAN/SPM is quite slow.

Therefore, we divide the raw images by "software gain" factors of 4 on the mountain prior to compression and transfer to the archives. Effectively, we discard the lower two bits of each pixel value. This process reduces the bzip2-compressed sizes to about 27% of the original sizes. Since we are still sampling the images at better than 1/2 of the read noise, we lose no information (Price-Whelan & Hogg 2010).

If you have doubts as to whether an image has been divided by the software gain factor, look for a SOFTGAIN record in the header. If one is present, it specifies the software gain that has been applied.

One consequence of this is that the images effectively saturate at 16383 DN, as if they were taken with a 14-bit ADC.

Henceforth, when we refer to "DN", we mean "DN in archive images" (i.e., raw DN divided by the software gain factor).

Bit Frequencies

The figures show the relative frequency of 1 bits in all of the flat field images taken on 2012 September 16. The frequencies are close to 0.5 in the first five bits, which suggests that the ADC is working well.

C0 bit frequencies. C1 bit frequencies.


The CCDs are cooled by recirculating ethylene glycol solution. The performance of the system depends on the temperature of the coolant which in turn depends on the ambient temperature. In summer we can cool to -30 C, but in winter to -40 C. The table shows the history of the CCD temperature set-points.

2012 September 1 to 2012 October 11-35 C-35 C
2012 October 13 to present-40 C-40 C

In July and August 2012 we had a number of problems with cooling. C1 had a blockage, which prevented it reaching its operating temperature. Furthermore, C0 had it set point below the minimum temperature it could attain, resulting in an instable temperature.

Gain and Read Noise

We used data from 2012 September 16 to determine the gain and read noise. We used the IRAF findgain tasks with six independent pairs of biases and flats for each detector.

Gain (electrons/DN)4.20±0.254.66±0.02
Read Noise (DN)3.35±0.012.41±0.00
Read Noise (electrons)14.1±0.811.3±0.1

We also determined the read noise on 35 nights between 2012 September 2 y 2012 November 8 using bias images and the IRAF imcombine task to estimate the population standard deviation between pixels. The results were 3.28±0.05 DN for C0 and 2.35±0.04 DN for C2. These values are 3%-4% smaller than those determined by findgain, presumably the result of better rejection of radiation events. Furthermore, the small variation shows that, at least over this limited period, the read noise was stable and was the same both at -35 C and at -40 C.

Bias Level

The bias level depends on the CCD temperature. In the operating range of -30 C to -40 C, the dependence is about 0.4 DN/C for C0 and about 0.3 DN/C for C1.

C0 bias temperature dependence. C1 bias temperature dependence.

The table shows the bias level at -35 C and -40 C. We determined the bias level at -35 C on 23 nights between 2012 September 2 y 2012 October 11 and at -40 C on 23 nights between 2012 October 13 and 2012 December 13.

Bias Level at -35 C506.11 ± 0.33637.38 ± 0.22
Bias Level at -40 C502.56 ± 0.22635.46 ± 0.27

Dark Current

The dark current is shown in the follow table and figures. These values were calculated using data from 2012 September 24 and 2012 October 13. In this temperature interval, a temperature change of 5 C for C0 and 7 C for C1 results in a halving of the dark current.

Dark current (DN/s) at -30 C0.8740.074
Dark current (DN/s) at -35 C0.4370.040
Dark current (DN/s) at -38 C0.2810.033
Dark current (DN/s) at -40 C0.2040.025

C0 dark temperature dependence. C1 dark temperature dependence.

The following images show the dark current at -35 C in C0 (left) and C1 (right) from 0 to 5 times the median dark current (9.35 e/s for C0 and 1.05 e/s for C1).

C0 dark image (0 to 9.35 e/s). C1 dark image (0 to 1.05 e/s).

These data were taken just after the CCDs had cooled to -35 C. On the night of 2012 September 22, we measured the dark current at -35 C over an interval of 14 hours. The CCDs reached -35 C at 22:15, and we began to take darks at 22:41. Both CCDs show the dark current dropping, varying with time. For C0, it initially rises and then falls, but only varies by 2%. For C1, it falls monotonically by more than 40% from 0.050 DN/s to 0.033 DN/s.

C0 dark settling. C1 dark settling.

The images below show the differences between the dark current in C0 (left) and C1 (right) between the start and end of the night of 2012 September 22. The scales are -0.006 to +0.030 DN/s from C0 and -0.01 to +0.05 DN/s for C1. In the case of C1, the change is very smooth over the CCD, being dominated by a ramp from about 0.016 DN/s at the bottom of the CCD to 0.006 DN/s at the bottom of the CCD.

C0 dark difference from -0.006 to +0.030 DN/s. C1 dark difference from -0.01 to +0.05 DN/s.

Flat Fields

The images below show twilight flat field images from 2012 September 16 in (top row and left to right) SDSS u, r, and i and (second row and left-to-right) Bessell U, B, and V. We do not currently have flats in SDSS g or Ha.

C0 flat in u (SDSS u) from 0.8 to 1.2. C0 flat in r (SDSS r) from 0.8 to 1.2. C1 flat in i (SDSS i) from 0.8 to 1.2.

C0 flat in BU (Bessell U) from 0.8 to 1.2. C0 flat in BB (Bessell B) from 0.8 to 1.2. C0 flat in BV (Bessell V) from 0.8 to 1.2.

The table shows the RMS variation in the flats.

FilterRMS Variation

The bluer filters have much stronger variations. The bright peak in the lower left of the u and U flats is around 1.28 and the dark trench in the upper right is around 0.82. These features are weaker in B and V and almost imperceptible in r.

The larger doughnuts are probably caused by dust on the filter. The smaller doughnuts by dust on the detector window.

Shutter Error and Jitter

We measured the shutter error and jitter on the night 2012 September 24. We took ten sets of four images of the illuminated dome in r and i. Each set of four images had exposure times of 1, 2, 2, and 1 second. By using an "ABBA" pattern, to first order we remove variations in the dome light intensity. For each set we determined the shutter error (i.e., the additional exposure time above the commanded exposure time). We smoothed these images with a 21x21 median filter and then estimated the mean shutter error and the shutter jitter (i.e., how much the additional exposure time varies). The table shows the minimum and maximum values of the shutter error and the shutter jitter. The images show the pattern of the shutter error for C0 (left) and C1 (right) from 0 to 100 ms.

Shutter error (ms)23-920-48
Shutter jitter (ms)11

C0 shutter error from 0 to 100 ms. C1 shutter error from 0 to 100 ms.

The shutter errors are important in short exposures; in a 2 second exposure they are about 5% in C0 and 2.5% in C1. However, shutter jitter is negligible for exposures of 1 second or more.


The following table shows the zero-points for the instrument. We defined the zero-point to be the DN/s for a star of magnitude 0 (AB magnitude for ugri and Landolt magnitude for UBV) at the zenith.

2012 Sep 2 to 2012 Sep 178.29e71.23e99.08e86.38e86.55e76.16e87.30e8
2012 Oct 1 to 2012 Oct 21.52e81.95e91.34e99.28e81.18e81.05e91.07e9

The effect of recoating the mirrors at the end of 2012 September is evident in these zero-points.

Bright Limit

The following table shows the magnitude (AB magnitude for ugri and Landolt magnitude for UBV) at which a 1 second exposure should give a peak signal of about 12k DN (which is 3/4 of the ADC dynamic range) for 0.75 arcsec FWHM image quality.

FilterBright Limit

To Do

  • Outline reduction.
  • Measure linearity with dome flats.
  • Measure linearity with stars (long/short exposures).
  • Monitor flats with time.
Last modified 2 years ago Last modified on May 21, 2015 3:25:51 AM

Attachments (20)