Camera testing

During the on-sky tests at Geneva it was found that the bright stars exhibited unusual streaks, for example in Fig. 5.6. To determine the cause of this streaking, experiments were run on a single NGTS camera. We initially wanted to determine whether this effect was unique to the different camera used at Geneva.

In this section I discuss the process of analysing one of the NGTS cameras. The streaking behaviour seen with the on-sky data was characterised to determine the cause, and is discussed in Section 5.3.1. Section 5.3.2 describes the process of characterising the dark current. The measured behaviour with temperature is compared to the expected behaviour from the manufacturer, and the long term characteristics were measured.

Figure 5.6: Example raw image from the NGTS instrument at Geneva showing the bright stars streaking. The image has not been corrected for bias, dark or flat field effects.

5.3.1 Streak characterisation

Camera tests were also being undertaken by a team at Leicester University, where a more controllable optical environment was available. Tests other than dark current measurements were performed on the cameras and some strange artefacts were no- ticed during these tests. The camera being tested displayed an unexpected V-shaped feature across the centre of its dark frames, which changed shape depending on what readout settings were used for the exposure. This feature was first discovered in the camera taking the on-sky data, and was not visible with the camera being tested at Warwick. Another artefact that was apparent on both cameras being tested at Leicester and Warwick were cosmic ray signals which caused charge to spread along pixel columns. These cosmic ray hits have a large flux centred on a single pixel as they do not follow the same path as the rest of the flux through the telescope and interact with the CCD directly. They can cause the pixel they hit to saturate, so we initially attributed this charge bleed to saturation effects.

These streaks show an exponential decrease in the number of counts above the saturated pixel, and appear similar to problems with charge transfer efficiency (CTE) [see e.g. Dolphin, 2000]. This lack of efficiency causes some charge to be left behind on each vertical shift of a CCD readout, causing streaks to be left behind

objects. The streak was exponentially decreasing as the amount of charge left behind is a fraction of the original flux. The flux due to CTE effects f(y) along a column from pixel 0 where the peak flux occurs is

f(y) =f0δy (5.4)

wheref0 is the flux at pixel 0,δ is the CTE value,y is the column pixel offset from

pixel 0. To characterise the streaking behaviour, we used the dark frames collected for the analysis described in Section 5.3.2.

To characterise the hot or dead pixels of the CCD, which would interfere with analysing the cosmic rays, a master dark frame was constructed. This master dark frame was then subtracted from each dark frame in turn and we were left with the difference images for each exposure. Each image was then analysed for any pixels which were over a threshold value of 10000 ADU, which combined with a master dark frame value of∼ 2000 ADU gives a total threshold of 12000 ADU. This is much less than the electronic saturation point of the CCD but there should not be any pixels normally above this value without the presence of cosmic rays, so it filters frames with cosmic ray hits well. By removing any hot pixels or other similar defects we could be sure that only cosmic rays will breach this threshold. So that we could visually search for patterns in the saturation behaviour, images of the local regions around each saturated pixel were created. To diagnose the effect the regions were ranked by the brightest pixel value in the region under the assumption that this was the cosmic ray. By ranking by flux we could see that the streaking occurred only after the saturation point of the CCD.

We noticed that the CCD that we were studying had two different saturation levels for the different halves of the CCD. This was first discovered with the camera operating at Geneva when a sky flat was overexposed. The left half saturated at the electronic saturation level of 65535 ADU and the right half saturated at a lower value. Figure 5.7 shows a slice across a sky image taken at Geneva,X= 1024 is the mid-point. This shows that the CCD has two different saturation points, one for each half. The left half (X <1024) saturates electronically whereas for the right half (X >1024) the full well depth of the CCD is reached, at about 56000ADU×G(2.1) = 118000e−. The right half also shows some structure suggesting that the full well depth is not a constant across the entire frame.

To remove the effect of the CCD saturating at different points for the different halves, the cosmic ray regions were grouped into the left half (X≤1024) and right half (X >1024) and considered separately. We found that some cosmic ray hits did

0 500 1000 1500 2000 X 56000 58000 60000 62000 64000 F lu x

Figure 5.7: Horizontal slice across an overexposed flat frame taken from Geneva, averaged across 10 rows.

not exhibit streaks and some did (see Fig. 5.8 for two examples). We inferred the presence of a limiting flux value, which we attributed to the saturation point of that particular pixel.

The camera at Leicester with the V-shaped issue was also undergoing this test, taking lots of long dark exposures and waiting for a cosmic ray to hit. To ascertain whether the V-shape influenced the saturation point, the frames from this camera were analysed. It was found that saturated pixels were much more likely to streak if they were inside the V-shaped feature. This was further corroborated by shining a focussed light onto a region of the CCD that would change status from being inside the V region to outside depending on the shape of the V region (Fig. 5.9). Clearly the saturation only exhibits itself when the illuminated region is inside the V-shaped feature. It was thought that the other cameras where we observed the streaking effect but not the V-shaped region behaved as if the interior of the V-shaped region covered the whole CCD. Based on this result the cameras were returned to Andor and the streaking issue fixed by adjusting the clock voltages.

5.3.2 Dark current measurement

Specifications for the NGTS cameras were provided by the manufacturer, includ- ing a measurement of the dark current at different temperatures. To ensure the

0 50 100 150 ∆X 0 50 100 150 200 250 300 350 ∆ Y 0 50 100 150 ∆X 0 50 100 150 200 250 300 350 ∆ Y

Figure 5.8: Two regions from the dark frame series after the master dark frame has been subtracted. Left: peak flux value 18218 ADU, right: peak flux value 52836 ADU.

Figure 5.9: Illuminated region of the CCD tested at Leicester. The left panels were taken with a vertical shift speed of 76µs, the right panels were taken with a vertical shift speed of 38µs. Upper panels show the entire CCD, lower panels are centred and zoomed on the illuminated region.

performance characteristics of the camera matched the supplied specifications, a comparison of the dark current with temperature to the expected behaviour was performed.

To test the behaviour of the dark current with temperature, exposures were taken at different permutations of exposure time and temperature, with tempera- tures −10, −30 and −50◦_{C and exposure times 2, 20 and 200 seconds, logarith-} mically spacing the exposure times. One frame at −10◦C and 200 seconds was saturated and therefore not used in the analysis. Each of the remaining 8 frames were bias-subtracted and converted to electrons per second. We compared the mea- sured median flux values of these frames at each temperature with the specifications from Andor, shown in Fig. 5.10. A proposed relation was available from e2v1 for the CCD used (model number CCD47-20), and is given by

Qd

Qd0

= 1.14×106T3e−9080/T, (5.5) and is also shown in Fig. 5.10. The Qd0 parameter represents the temperature at

293 K (20◦C). In Fig. 5.10 this parameter was fitted for the black line and gives a best fit value of Qd0 = 73381 near the typical quoted value of 100000 represented

by the lower of the two grey dashed lines in Fig. 5.10. Due to time constraints we did not measure the intermediate dark current values at −40 and −20◦C and the lower temperatures were only achievable by a more expensive cooling.