next up previous contents
Next: Extinction and Sky Brightness Up: Exposure Times Previous: Array Linearity

Saturation

  An important thing to be aware of is array saturation. With WHIRCAM, saturation is defined by the full well capacity of the pixels, NOT by the dynamic range of the A-D converter (65,535 DN). The full well capacity is controlled by the substrate voltage (called the bias voltage), but as this also affects the number of hot pixels, a larger well capacity is paid for by a larger number of hot pixels. The default voltage setting (0.68 V) provides a reasonable well depth and a bearable number of hot pixels. If fewer hot pixels are required, then you can select a lower bias voltage by selecting Imaging, set_ALICE, recall_setup, then change recall_input to FPA40348_SHALLOW, followed by select_enable = YES. This will change the bias voltage to 0.60 V. The default is FPA40348_DEEP, with a bias voltage of 0.68 V.

To make life interesting, the well depths, and hence saturation, vary across the chip, and vary with wavelength. And because of the non-linearity above 80% of full well capacity, the effective saturation level is even lower. In the following, all saturation figures are defined as 80% of full well capacity.

For ND_STARE mode with a bias of 0.68 V, the minimum saturation in Ks is 9,500 DN (along the lower edge of the chip), and the minimum saturation within the central tex2html_wrap_inline1343 pixels is 12,000 DN. There is also an odd/even column effect, giving rise to two peaks in the saturation levels, at 13,500 and 17,500 DN. At J, these figures tend to increase, with the minimum saturation being 11,000 DN, the minimum within the central tex2html_wrap_inline1343 being 14,000 DN, and the two odd/even column peaks being at 15,500 and 17,500 DN. Figure 5, plots the non-linearity for the even and odd columns, but shows simply the means of these columns. Users should be very careful to estimate exposure times such that the maximum counts will not saturate on the chip in the area of interest. As most observations will be taken as combinations of dithers, then the best practice would be to aim to have final counts always below 10,000 DN in ND_STARE. For STARE mode, the fixed pattern noise, dominated by the reset values, has a range of 20,500 to 28,000 DN, making a comparison with ND_STARE difficult. With a bias of 0.68 V and a Ks filter, STARE mode has a minimum 80% full well saturation above reset level (ie 100% of reset level plus 80% from there to saturation) is tex2html_wrap_inline1117 37,500 DN, with less well-resolved peaks at 38,000 DN and 40,000 DN.

The approximate 80% well depths and number of hot pixels for each Bias voltage are given in Table 5.2.

  table282

A problem with this array is that the odd columns tend to have a lower well depth than the even columns. So that both sets of columns will have 80% saturation at similar DN values, the reset level has been set up such that it is higher for the odd columns (see Figure 5).

In the thermal near-IR (3-5 tex2html_wrap_inline1109 m), the background level is always high and dominates any astronomical signal. Short on-chip integrations are always required.

Although the detector is a direct readout device with 4 readouts which each address every fourth column, the minimum exposure time for full frame readout is 120 ms, limited by the multiplexer performance, and it is this minimum exposure time and finite well depth that defines lower magnitude limits below which stars will saturate the chip. Table 5.2 gives the minimum magnitudes of stars which will saturate at this minimum exposure time (in 0.5'' seeing and a scale of 0.24''/pix):

  table293

Faster exposures can be made (of brighter stars) if subarrays are employed.


next up previous contents
Next: Extinction and Sky Brightness Up: Exposure Times Previous: Array Linearity

Shaun Hughes
Thu Mar 14 14:48:38 GMT 1996