Chapter 10
Binning
69
EAR99
Technology
Subject
to
Restrictions
Contained
on
the
Cover
Page.
10.2 Software Binning
One limitation of hardware binning is that the shift register pixels and the output node
are typically only 2-3 times the size of imaging pixels. Consequently, if the total charge
binned together exceeds the capacity of the shift register or output node, the data will
be corrupted.
This restriction strongly limits the number of pixels that may be binned in cases where
there is a small signal superimposed on a large background, such as signals with a large
fluorescence. Ideally, one would like to bin many pixels to increase the S/N ratio of the
weak peaks but this cannot be done because the fluorescence would quickly saturate
the CCD.
The solution is to perform the binning in software. Limited hardware binning may be
used when reading out the CCD. Additional binning is accomplished in software,
producing a result that represents many more photons than was possible using
hardware binning.
Software averaging can improve the S/N ratio by as much as the square-root of the
number of pixels binned. Unfortunately, with a high number of pixels binned, i.e., above
100, camera 1/f noise may reduce the actual S/N ratio to slightly below this theoretical
value. Also, if the light source used is photon-flicker limited rather than photon
shot-noise limited, this theoretical signal improvement cannot be fully realized. Again,
background subtraction from the raw data is necessary.
This technique is also useful in high light level experiments (e.g., absorbance
spectroscopy,) where the camera is again photon shot-noise limited. Summing multiple
pixels in software corresponds to collecting more photons, and results in a better S/N
ratio in the measurement.
