The HIRES CCD can be run at one of two gains, low or high. The high gain setting will have fewer electrons per DN than the low gain setting. What makes one astronomer choose high gain and another choose low gain?

If you are looking at faint objects, which will not be close to saturation, then much of your data may be "read-noise" limited, which means that the read-out noise will be a significant factor in the quality of the data. There is a similar number sometimes called "digitization noise" which may also be a limit; when the electrons are converted into a Digital Number (or Data Number, often called "DN"), they are scaled and rounded to an integer. If you only have a few DN in an interesting part of your image, this roundoff may be significant. Using the high gain setting will produce more DN for the same number of photons, hence will decrease the effect of the roundoff.

Conversely, if you have plenty of photons, you can use the low gain setting to increase the "headroom" of your data. By decreasing the number of DN produced by a given number of photons, you can push more electrons through the analog-to-digital converter (ADC), hence observe brighter objects than you can with the high gain setting. This is particularly important since non-linearity starts to become significant above approximately half of full well.

Limitations are the read-out noise for high gain setting, and CCD well depth for the low gain setting. If the read-out noise (constant in number of electrons) dominates the digitization noise, then pushing to higher gain settings does not help. And once you have a gain low enough so that the CCD's well depth will fit into the ADC, there is little point in pushing to lower gain settings. Hence many CCD systems provide only the two settings.

The default gain setting for HIRES is low. The blue, green, and red CCD's have low gains of 1.95, 2.09, and 2.09 e-/DN respectively. The corresponding high gain settings are 0.78, 0.84, and 0.89 e-/DN.

A description of how to change from one setting to another can be found here.

.