Unravelling Sensitivity, Signal to Noise and Dynamic Range – EMCCD vs CCD
4. Effects Of Background Photons On Detection Limit
4.1 Background Photon Considerations
-
Deep-cooled, high-end EMCCDs represent the most sensitive detectors available on
the market and perform exceptionally well in ultra low-light experiments. It is
crucially important to understand however that under such conditions, it is not
enough simply to have the most sensitive detector, but it is essential also to restrict
as far as possible the amount of undesirable background photons that are reaching
the detector.
-
Some sources of background photons may be
-
Excitation light
-
Scatter from microscope slides
-
Unblocked ambient light
-
Light from outside of the z focal plane of the imaged specimen
-
Unspecific binding of fluorescently labelled biomolecules.
-
Background photons, if not minimized, have the potential to contribute significantly
to the noise floor and yield a detection limit that is defined by the background
shot noise, rather than the read noise and dark current sources of the detector.
Depending on the flux of background photons and also the technology used (standard
CCD or EMCCD), this contribution can affect the detection limit with varying severity.
-
It is important to learn to recognise the degree of contribution of background photons
to the detection limit. This is accomplished first by defining the true camera noise
floor in total light tight conditions, using the same exposure times/gain levels
that are to be used in the actual measurement. True light tight conditions can be
achieved by screwing a blanking plate onto the c-mount adaptor of the camera. The
‘2D Line Intensity Profile’ within the Andor MCD software is very useful for reading
the instrument noise characteristics.
4.2 CCD Cameras and Background Photons
Following the procedure given above for defining the true instrument noise floor,
the intensity profile at any point across the image should represent largely read
noise with a small contribution from darkcurrent noise (higher contribution at very
long exposures). The read noise amplitude in counts can in fact be predicted from
the pre-amplifier gain, available on the specification sheets that accompany Andor
cameras.
Upon exposure to the signal, it only requires background photoelectron levels of
say > 5 to 20 electrons/pixel for the noise floor to begin to become dominated by
photon shot noise characteristics. If this is the case, both the detection sensitivity
and the dynamic range become compromised.
An increase in QE is unlikely to help significantly in this case since it will make
the camera more sensitive to background photons also (unless the background photons
are of a different wavelength to the signal photons).
4.3 EMCCD Cameras and Background Photons
The EMCCD cameras, representing the world’s most sensitive CCDs, are particularly
susceptible to background photon noise when EMCCD gain is turned on. Never forget
that the EMCCD is capable of detecting even a single photon!
There are a couple of different levels of background photon noise to watch for under
high EMCCD gain:
1. Background photons eliminated
this is rarely the case in an experimental set-up but it can be simulated with a
good blanking plate or shutter over the sensor and keeping exposure times to an
absolute minimum. The noise profile in this case should consist of a read noise
base level, approx 10 counts peak to peak with the iXonEM+ range, with occasional
spurious CIC noise spikes amplified well above this level (darkcurrent spikes are
negligible). If the EM gain were to be turned down, the CIC spikes should disappear
leaving only the read noise floor, of the same peak to peak magnitude as with gain
applied.
2. Some background photons
many Total Internal Reflection Microscopy set-ups or confocal microscopy set-ups
will create this limited background situation, i.e. the illumination plane in each
configuration is highly restricted, drastically reducing the acceptance of background
photons from outside the focal plane. Nevertheless, background photons will still
be present to varying degrees and can be recognised as additional ‘spurious’ spikes
alongside the expected level of CIC spikes. It is still essential to eliminate CIC
and darkcurrent contribution as far as possible.
3. Background shot noise dominated
If insufficient measures have been taken to reduce background levels, it is likely
that the noise profile, under conditions pf applied EM gain, will be dominated entirely
by background shot noise. Under highest EM gain setting, this will have a peak to
peak value of several hundreds or even thousands of counts and this magnitude will
decrease as EM gain is reduced, possibly to a point where it is buried in the read
noise when gain is low or off.
This is a somewhat less desirable situation for optimal operation of EMCCD technology
– the EM gain may have been used to amplify a weak signal clear of the read noise,
but the background photon shot noise will be amplified also and application of further
EM gain will do nothing to improve this limiting Signal/Background Noise ratio.
That is not to say that EMCCD are not still highly effective in optimizing most
low light signals, whether limited ultimately by background photons or not. Even
set-ups such as widefield fluorescence microscopy (where the limiting background
is from out of plane fluorescence) can still benefit enormously from application
of EM gain, as the read noise floor can still contribute strongly (often overwhelmingly)
in its absence.