The Electron-Multiplying CCD (EMCCD) technology renders readout noise negligible thanks to EM Gain, which allows reaching significantly higher frame rates than conventional CCDs by using higher pixel rates. Other than the readout rate, the number of pixels read as well as the size of the sensor will affect the frame rate.
Moreover, the EMCCD sensor’s architecture also gives it a high level of flexibility to create custom readout sequences. This allows to further maximize frame rate when using only portions of the sensors (Regions of Interest – ROIs) or even to reach extremely low exposure times, down to the few microseconds, by minimizing the movement of the charges on the sensor.
The structure of an EMCCD acquisition is highly important in understanding what limits frame rate. EMCCDs use a “frame transfer” architecture, which makes use of 2 regions on the sensor: an imaging area and a storage area. While the imaging area is exposed to light, the storage area is shielded, which enables the readout of the photoelectrons of a previous exposure even while acquiring a new image. Consequently, the shortest exposure possible for the imaging area is limited by the processing time of the image, when the storage area is free.
Once the imaging area has been illuminated for the selected exposure time, the first step after exposure is to rapidly transfer the charges from the exposed area to the storage area. The larger the sensor the more rows on the sensor, the longer this frame transfer takes (~10 us to 1 ms) as there are more rows of charges that must be transferred. The second step is to read all the pixels. The readout time will depend both on the number of pixels to read and the sensor’s readout rate (~0.5 ms to 100 ms for a full sensor). As such, two elements affect frame rate: the size of the sensor (number of pixels to read & move on the sensor) and the readout rate (rate at which the pixels are read).
In most acquisition modes, the readout time will be the main factor limiting the maximum frame rate. Higher readout rates can be used to speed up the process, but EMCCD readout is a complex process and care must be taken to preserve image quality at higher speeds. Thanks to better control over the readout & noise sources, Nüvü Camēras is the only EMCCD manufacturer to offer the highest readout rate of 30 MHz (30 million pixels per second) on all sensor formats.
Another way to reduce readout time is to reduce the number of pixels to read. This is achieved by using a Region of Interest (ROI) which restricts readout to section of the detector relevant to the acquisition. Nüvü™ also supports the use of multiple regions of interest (mROI) for more flexibility. Using pixel binning is another way to increase frame rate, although it is not recommended in low-light acquisitions since the resulting super-pixel includes the noise of all sub-pixels whereas there is not necessarily signal in each sub-pixel.
An important particularity of ROIs and binning is that the horizontal dimension (i.e. the number of columns) of the ROI or super-pixel has no influence on imaging speeds. This means a 32 x 32 ROI or 32 x 512 ROI and 2 x 2 binning or 2 x 1 binning achieve the same frame rate. This is because when a line is selected to be read, the entirety of line must be shifted pixel-by-pixel through the horizontal multiplication register even if they are not of interest.
Other possible improvements are to reduce the frame transfer time. In these cases, it is necessary to use an optical mask, which shields all the imaging area of the sensor from light except for the ROI. It is then possible to use Crop Mode, which optimizes the charge transfer in order to conduct a faster, more efficient readout of the exposed pixels. There is a small delay before the first image is available since it takes several partial frame transfers before it reaches the bottom of the sensor where it can be read.
Another method using an optical mask is Fast Kinetic Mode (FKM), which both optimizes the charge transfer and omits readout to instead store multiple ROI images on the sensor. Once the required number of images has been acquired, all images stored on the sensor are read out at once. Thus, FKM allows to reach minimal exposure times down to a few microseconds in these bursts of images. While these exposure times are unmatched by other acquisition methods, FKM does not allow continuous acquisitions due to the delayed readout.