Page 149
New iStar ICCD
Appendix
NoIse
Noise
is a complex topic, a full exploration of which is beyond the scope of this glossary. Noise may, however, be
broken down into two broad categories:
1. Pixel Noise
2. Fixed Pattern Noise
PIXEL NOISE
Let us first attempt to define pixel noise. Assume that a light signal is falling on a pixel of the CCD. If the charge on the
pixel is read, and the read process is repeated many times, the noise may be taken as the variation in the values read.
The
Root Mean Square (r.m.s.)
of these variations is often used to express a value for noise. As a rule of thumb, the
r.m.s. is four to six times smaller than the peak to peak variations in the count values read from the pixel. Pixel noise has
three main constituents:
•
Readout noise
•
Shot noise from the dark signal
•
Shot noise from the light signal itself
Shot noise cannot be removed because it is due a known noise factor derived from these signals. Most simply defined,
shot noise is the square root of the signal (or dark signal) measured in electrons.
READOUT NOISE
Readout noise is due to the amplifier and electronics: it is independent of dark signal and signal levels; it is only very
slightly dependent on temperature; and it is present on every read, as a result of which it sets a limit on the best
achievable noise performance.
•
Shot noise is dependent on the dark signal. The dark signal is itself dependent on the exposure time and is very
dependent on the temperature.
•
Shot noise from the signal is additionally dependent on the signal level itself. If either the signal or the dark
signal falls to zero, their respective shot noise also falls to zero.
The total pixel noise is not, however, simply the sum of the three main noise components (readout noise, shot noise from
the dark signal, and shot noise from the signal). Rather, the Root Sum Square (r.s.s.) gives a reasonable approximation -
thus:
total = sqrt (readnoise² + darkshot² + sigshot²)
where:
F
ixed
P
attern
N
oise
(FPN
) consists of the differences in count values read out from individual pixels, even if no light is
falling on the detector. These differences remain constant from read to read. The differences are due in part to a variation
in the dark signal produced by each pixel, and in part to small irregularities that arise during the fabrication of the CCD.
Since fixed pattern noise is partly due to dark signal, it will change if the temperature changes, but because it is fixed, it
can be completely removed from a measurement by background subtraction.
total
is the pixel noise
readnoise
is the readout noise
darkshot
is the shot noise of the dark signal
sigshot
is the shot noise of the signal
