This page provides technical information about the detector arrays, and knowledge of these details is not needed to prepare your observations.
Detector Array characteristics
The WFCAM IR detector arrays are Rockwell Hawaii 2™ 2048×2048 pixel HgTeCd PACE devices. In operation, the arrays are cooled to 75K and each array is read out in its 8 channel per quadrant mode through a 32 channel SDSU controller.
Table 2.1 summarises the properties of the 4 WFCAM arrays.
These arrays have been found to show inter-pixel capacitance, meaning that adjacent pixels are not independent. With about 20% of the incident on any given pixel appearing in adjacent pixels. This has 2 effects – slight degradation of the image quality, and correlated noise. One effect of the latter is that a simple gain measurement assuming Poisson statistics overestimates the gain by about 20%. This effect has been included in the numbers quoted below.
Cosmetically, the array for camera number 3 is the cleanest and also shows good QE. Thus, if your observations target a single object, we reccomend that you place it on camera 3. To do this, you should specify a telescope pointing centre 795 arcseconds South and 795 arcseconds East of your target co-ordinates.
|Camera Number||Position in Focal Plane||Array ID Number||Gain [electrons/ADU]||Full Well [electrons]||Full Well [ADU]|
|1||SW||60||4.6 +/- 0.2||185,000||40,000|
|2||SE||63||4.5 +/- 0.2||181,000||40,000|
|3||NE||76||4.7 +/- 0.2||186,000||40,000|
|4||NW||42||5.6 +/- 0.2||156,000||27,900|
The WFCAM IR arrays are each divided into 4 electronically isolated quadrants, each of which is subdivided into 8 segments. All 32 segments are read out simultaneously through the 32 parallel channels of one SDSU array controller. There are 4 such controllers, one for each array. The 4 contollers are synchronised to minimise pick up of readout clock signals between the arrays. Figure 2.1 shows the layout of the channels on the array and the orientation of each of the science arrays in the focal plane.
WFCAM offers a limited number of readout modes. STARE mode is for use in engineering only, science readout modes are limited to two types of NDSTARE modes: CDS and MNDR.
CDS – Correlated Double Sampling is the default readmode and should be used for all broad-band observations. Basically, this is a reset-read-read sequence; the array is reset, then read out as soon as possible. As soon as the first read starts, the exposure timer starts. After the specified exposure time, a second readout takes place. The second readout is subtracted from the first readout to give the data values for that exposure. By doing two readouts like this, we optimally remove the array bias signal and reduce readnoise. Each readout step takes approximately 0.61 seconds to cover the entire array. Although all the pixels are not read out simultaneously, the interval between the two reads of any given pixel is exactly equal to the exposure time. Figure 2.2 illustrates this. The overhead for an exposure is approximately 1.2 seconds (reset-time plus read-time).
Figure 2.2. CDS readout timing
MNDR – Multiple Non-Destructive Reads will probably be used for the longer exposure time modes, i.e. the narrow band filters. It’s possible that we’ll use it with the broad band filters too. This will be determined during commissioning. MNDR is similar to CDS, except that readouts occur more frequently right through the exposure. Once the exposure has finished, a straight line is fit to all the readouts of each pixel in turn. The gradient of the line represtents the flux arriving at that pixel. This helps to further reduce the read noise in the exposure, though the increased readout activity could cause pickup problems. More recent measurements on this subject (20090630):
- Overhead for a CDS exposure: 1.39 +/- 0.05 seconds
- Overhead for an NDR exposure: 1.61 + 0.075 * exptime (seconds). For example, 5 seconds: 1.97 seconds, 10 seconds: 2.36 seconds, 20 seconds: 3.14 seconds, 30 seconds: 3.85 seconds.
- Delay before a new image is started: 0.8 +/- 0.15 seconds
Total overhead between two consecutive images if a new guide star needs to be acquired:
- 10 seconds if within the same 4-points tile
- 11 seconds if starting a new tile
Total overhead between two consecutive images if a filter change is involved:
- 30 seconds
The array read noise is approximately 25 electrons for CDS exposures. NDR exposures have somewhat lower read noise, but not as low as would be expected due to some systematic effects. Typical NDR read noise is 15 electrons.
The readout gain is approximately 4.6 electrons per data number for cameras 1, 2 and 3 and 5.6 for camera 4. Because of inter-pixel capacitance, adjacent pixels are coupled by approximately 20%, meaning that a simple Poisson noise analysis will over-estimate the gain by about a factor of 20%.
The fundamental system parameters for sensitivity calculations are the telescope collecting area, the filter bandwidth and the total system throughput. For the purposes of these calculations, the telescope is considered to have a collecting area of 9.6m2. The filter bandwidth is given in Table 1.1. Table 3.2 gives the mean total system throughput under photometric conditions. For the narrow band filters, assume similar throughput to K band.