Maximizing Frame Rate Performance in EMCCDs
The iXon3 is capable of market leading frame rate performance, achieved from ‘over-
clocking’ the vertical shifts during readout. Furthermore, fastest possible continuous sub-
region frame rates can be attained using the new ‘cropped sensor’ mode.
Part 1. Fastest vertical shifts for fastest speeds
Maximum frame rate performance in EMCCDs is a function of two
parameters; (1) Pixel Readout Speed (horizontal); (2) Vertical Clock
Speed. The former dictates how rapidly charge is pushed horizontally
through the EM gain register and the remaining readout electronics,
while the latter dictates the speed at which charge is vertically shifted
down through both the exposed sensor area and masked frame transfer
area of the chip. Significant advantages are gained through optimizing
the camera electronics to enable vertical shifts to be speeded up.
Andor’s iXon3 vertical shift speeds are the industry fastest.
Table 1 shows frame rates available from the iXon3 897 frame rates
versus those of two other competing EMCCD brands with the same
sensor. It is clear that while iXon3 still delivers the fastest full frame
rate (512 x 512; no binning), the beneficial effect of ‘overclocking’
the iXon3 vertical shift speeds are most apparent under conditions of
sub-array and/or binning, e.g. ~60% faster with 128 x 128 sub-array
and 2 x 2 binning. This is particularly significant as many EMCCD
applications require routine use of sub-arrays to examine a smaller
active region of the sample under faster speeds (also used to reduce
the file size of a large kinetic series of images).
iXon3 industry-fastest vertical shift speeds result in
faster frame rates and reduced smearing, markedly so
under commonly employed conditions of sub-array/
binning. ‘Overclocking’ the vertical shifts also reduces
‘clock induced charge’ spurious noise.
Part 2. Pushing Frame Rates with Cropped Sensor Mode
The iXon3 offers Cropped Sensor Mode, which carries the following
advantages:
- Specialized readout mode for achieving very fast frame rates
(sub-millisecond exposures) from ‘standard’ cameras.
- Continuous rapid spooling of images/spectra to hard disk.
- User selectable cropped sensor size – highly intuitive software
definition.
- Ideal for super-resolution microscopy, ion signalling, voltage
sensitive dyes and adaptive optics.
- The iXon3 is now available with the complementary
OptoMask accessory, which can be used to shield the region of
the sensor outside of the cropped area.
If an experiment demands fast temporal resolution, but cannot be
constrained by the maximum storage size of the sensor (as is the
case for ‘Fast Kinetics Mode’ of readout), then it is possible to
readout the iXon3 in ‘Cropped Sensor Mode’. In this mode, the user
defines a ‘sub-array’ size from within the full image sensor area,
such that it encompasses the region of the image where change is
rapidly occurring (e.g. a ‘calcium spark’ within a cell). The sensor
subsequently “imagines” that it is of this smaller defined array size,
achieved through software executing special readout patterns, and
reads out at a proportionally faster frame rate. The smaller the defined
array size, the faster the frame rate achievable. Table 2 shows frame
rates that are achievable with the iXon3 897 when in Cropped Sensor
Mode.
In order to use Cropped Sensor Mode, one has to ensure that no light
is falling on the light sensitive area outside of the defined region. Any
light collected outside the cropped area could corrupt the images that
were acquired in this mode. For microscopy set-ups, this is now aided
with an accessory called OptoMask, available from Andor.
Cropped Sensor Mode has the end result of achieving a much faster
frame rate than that obtainable in a conventional ‘sub-array’ / ROI
readout (during which we would still have to vertically shift the
unwanted rows). The frame rate increase is achieved by not reading
out (i.e. discarding) the unwanted pixels.
Cropped Sensor Mode is ideally suited to a number of challenging
applications across many diverse fields of research. In terms
of technology matches, cropped sensor is well suited to rapid
applications of EMCCD cameras. The fundamental advantage and
a distinct feature of EMCCD technology is its ability to virtually
eliminate the camera readout noise detection limit at any readout
speed. This allows EMCCD detectors to be successfully used for
applications where raw sensitivity and exposure time requirements
ultimately prevent the use of conventional CCD systems.
In biological imaging Cropped Sensor Mode can be successfully
used to enhance performance and throughput in super-resolution
‘nanoscopic’ applications including STORM and PALMIRA.
Imaging frame rates exceeding 1000 / s can be achieved with a
sufficiently small crop area. A series of measurements done on the
Andor iXon3 885 have demonstrated that Cropped Sensor Mode in
conjunction with binning has pushed the speed beyond 4000 frames
per second. Table 3 shows a comparison of frame rates achievable
by the iXon3 885 camera in standard sub-array readout versus the
same size of region selected in Cropped Sensor Mode. It is apparent that frame rates of between x2 and x4 faster are readily achievable in
Cropped Sensor Mode.
Cropped Sensor Mode can also be employed to achieve extremely
fast temporal resolution in ion signalling measurements, such as
observing calcium sparks. Samples labelled with voltage sensitive
dyes also benefit from extremely fast imaging, with thousands of
frames per second not being uncommon.
EMCCD-based adaptive optics, for which smaller format EMCCD
sensors are often used, can benefit from cropped sensor readout.
Small area EMCCDs can already operate at >500 fps and can be
flexibly optimized in cropped mode to exceed 2000 fps. Use of
cropped sensor mode opens new possibilities for very fast adaptive
optics imaging enabling the users to reach into several thousands of
frames per second.
There is also potential to use cropped EMCCDs for multi-spectral
fluorescence confocal scanning, as an alternative to the arrays of
PMTs that have traditionally been used in this approach. The greater
than 90% Quantum Efficiency of the back-illuminated sensor, single
photon sensitivity, array architecture and rapid pixel readout speed
can be exploited to markedly improve this approach. The laser dwell-
time should be set to coincide with the time to expose and read-out a
short row of approximately 32 pixels - sufficient spectral channels to
yield effective un-mixing of several known emitting dyes, resulting
in a data cube of 512 x 512 x 32 (spectral) taking less than 1 second
to generate. There is a clear sensitivity advantage of EMCCD pixels
over the usually employed PMT-technology, which is circa 5-fold in
the blue-green and up to tenfold in the red.