EMCCD Vs ICCD
ICCDs edge out electron-multiplying CCDs in low light? (We don’t think so!)
A recent article has been published in Laser Focus World in September 2004 by David
Dussault and Paul Hoess - David is an application engineer at Paul Hoess (Munich,
Germany) and Paul is president of Stanford Computer Optics (Berkeley, CA) and general
partner of Paul Hoess KG.
Read laser focus world article
The article claims that EMCCDs perform better than ICCDs only at light levels
above those that require cooling or slow scan. Andor Technology, as the pioneers
of EMCCD technology would like to set the record straight.
David and Paul claim “ICCDs edge out EMCCDs in low light” because “charge-transfer
noise greatly reduces their noise performance” (Laser Focus World, September 2004).
They present theory to show this should be so and also have a series of pictures
to confirm the theory; all seems very compelling. The trouble is that the numbers
used in the theory are erroneous and the pictures for the EMCCD do not look like
low light level images should from our cameras, strongly indicating that something
is seriously wrong with the EMCCD images. Without access to the original measured
data we can only guess that either there was no gain, or that the signal was lost
in darkcurrent because the camera was not cooled, or the camera was faulty. We know
from our own measurements, and those of independent users, that BI EMCCDs outperform
ICCDs.
The theory mistake is easy to spot, it is because they use a “charge–transfer noise
of 5.4e”, we were not involved with any of this testing and cannot even guess were
they get this number from, the true value for the camera model quoted is 0.02 (0.05
at worst). When this number is used in their theory the camera model quoted will
outperform the ICCD for all signal ranges shown, and this is without considering
their other assumptions. (For reference: the charge-transfer noise arises from the
Clock Induced Charge or Spurious Charge.) They choose the peak QEs so that the EMCCD
is only 1.8 times better, but this does not do justice to the high and broad nature
of BI EMCCD QE, something not matched by any photocathode.
They appear to have other misunderstandings of EM technology; they state “very slow
pixel rates” are required “to minimize the impact of charge-transfer noise”. Again
we do not know where this comes from because the opposite is in fact true. (They
may be confusing some old assumptions from non-EMCCDs.)
They make a curious statement “Hence, we consider chip cooling of an ICCD camera
as being unnecessary and a useless cost increase of the system”. They make a valid
point but this statement goes over the edge; they mention the dark-field subtraction
method required by the ICCD, but they have forgotten that for this to work best
the CCD needs to be temperature stabilized; and this is best done by cooling. Modest
cooling also reduces the ICCD fixed pattern noise and blemishes making the process
more accurate.