EMCCD Tutorial

How EMCCDs work

An EMCCD (electron-multiplying CCD) camera operates by first transforming incident photons into photoelectrons via the photoelectric effect, which takes place in the EMCCD’s silicon body within the camera cryostat. The semiconducting surface is meticulously organized in an array of pixels, a matrix of potential wells in which electrons can accumulate during an exposure. Following the collection of these negative charges, applying voltages to the potential wells forces the electrons to quickly travel row by row, a process known as charge coupling, from the imaging to the storage region. Doing so insures that the acquired image can be processed while a new image is formed during the subsequent exposure time.

In the storage region, the electrons at the bottommost row of the detector travel pixel by pixel into the multiplication register. At this stage, the accumulated photoelectrons are amplified via impact ionization, a type of avalanche effect that generates several thousands of electrons per photoelectron. The charges finally reach an amplifier where they are converted into an electric impulse that is subsequently digitized to form a high quality image.

What parameters influence EMCCD sensitivity?

Compared to other low flux imaging technologies, an EMCCD has the benefits of high quantum efficiency (QE) and low dark current in the same manner as traditional CCDs. More, it presents negligible readout noise, like intensified CCDs, thus providing the best sensitivity. However, an EMCCD is subject to a different noise source, commonly referred to as clock-induced charge (CIC), which can contaminate images up to 200 times more than dark current when only a handful of photons reach the sensor.

Listed below are the primary parameters that influence the sensitivity of low light imaging systems, in addition to the means by which Nüvü Camēras technology is capable of efficiently and elegantly overcoming such obstacles.

Thermal noise

A first parameter that interferes with the proper signal’s acquisition is Thermal noise, or dark current. Ideally, an EMCCD chip would only detect photoelectrons. However, the mere silicon thermal agitation is sufficient to occasionally eject an electron, which will be collected in a potential well. Even in darkness, the wells slowly gather charges as if the sensor had been exposed to a faint light source.

To counter Thermal noise, a widespread technique is to cool down the chip to temperatures well below freezing, near -90 °C. Doing so substantially decreases the amount of dark electrons within the silicon chip, thus improving the photoelectrons’ signal. Still, this is rarely the case in a variety of fields including astronomy and biomedical research, where light sources are extremely dim. Consequently, other noise sources must be taken into consideration and be eliminated.

Nüvü Camēras EMCCD technology. Other EMCCDs.

Acquisition conditions: -85 °C cooling, 10MHz readout speed, EM gain of 5000 for Nüvü Camēras, and 1000 for other EMCCDs.

Readout Noise

Neglecting Thermal noise effects, the amount of accumulated electrons is proportional to the incoming light. Consequently, in low lighting conditions, the sensor must be exposed for longer periods of time for an appreciable quantity of electrons to be captured and detected. Still, the observed phenomenon may sometimes occur over a timescale shorter than the necessary exposure time to reveals its dynamics.

Another noise source — the readout noise — occurs while converting a charge to electric potential. Typically negligible, as it is in the order of 2 to 10 electrons, the readout noise however becomes a limiting factor in achieving high quality image when there are very few photoelectrons in the potential wells for a given exposure.

To increase the sensor’ sensitivity, EMCCDs integrates a charge amplification stage which, with the help of several high voltage elements, boosts the photoelectrons’ signal prior to the charge-potential conversion. The amplification process generates a gain in the order of 1000 to 2000 electrons per photoelectron that, in contrast, renders the readout noise imperceptible. Thus, an EMCCD are ideal ultra low light detectors, since the impact of every incident photon is increased over a thousand-fold. Yet, the stochastic nature of electron avalanche adds the excess noise factor (ENF).

Excess Noise Factor

Being stochastic in nature, the electron avalanche process only allows ascertaining the multiplicative gain mean value. The figure below demonstrates this uncertainty.

Horizontal dashed lines represent ±1 % of mean value.

On the graph, the mean value represented by the solid line is predicted with good accuracy, but individual pixel values are distributed on either side of the mean value, depending on their position and detection time. Such uncertainty induces the excess noise factor (ENF) √2 value, which affects the signal-to-noise ratio (SNR) the same way as halving the system’s quantum efficiency.

Photon counting gets rid of the ENF by applying a detection threshold to the pixels’ values. In this manner, the individual pixel gain is not necessary to determine the number of incident photons; rather, the output becomes binary. If the number of photoelectrons accumulated within a given potential well is greater than the specified threshold, the system translate it as one detected photon. Hence, resorting to Photon counting (PC) mode eliminates the ENF and allows high sensitivity acquisition without losing image quality. Nevertheless, another noise source appears during high-speed acquisition that greatly limits the EMCCD performance — and it is precisely this noise source that Nüvü Camēras’ technology efficiently overcomes.

Clock-Induced Charges

The high voltages and great clock frequencies that operate the EMCCD chip produce clock-induced charges (CIC); these electrons are created during the fast clock variations required for shifting the photoelectrons to the readout register.

Fortunately, Nüvü Camēras’ innovative and patented Controller for Counting Photons (CCCP) generates finer and more versatile signals than those produced by other controllers, allowing it to drastically decrease an EMCCD’s CIC.

Charge Transfer Efficiency

Finally, some photoelectrons may be left behind during charge transfer, especially at high readout speeds: the electron shift from one potential well to the next may be incomplete, creating deferred charges. Consequently, the leftover electrons artificially increase the brightness of certain pixels, thereby diminishing the overall image quality. Nonetheless, on top of decreasing CIC, the CCCP maintains excellent charge transfer efficiency (CTE) at any operating temperature, hence creating images of highly superior quality.

Nüvü Camēras EMCCD technology. Other EMCCDs.

Acquisition conditions: -85 °C cooling, 10MHz readout speed, EM gain of 5000 for Nüvü Camēras, and 1000 for other EMCCDs.

EMCCD operations

Conventional mode

Nüvü Camēras EMCCD cameras operate on 2 different types of digitizers: conventional or electron-multiplying mode. The conventional mode is carried out in the same manner as a CCD camera: captured photons are converted into photoelectrons and then scan pixel by pixel. Such mode is ideal for bright conditions where noise is negligible in comparison with the input signal.

Electron-multiplying mode

Conversely, in order to eliminate readout noise, the electron-multiplying mode uses the EMCCD multiplication register to amplify the charges contained in each pixel, and that, before scanning. Such digitizer is suitable for low lighting imaging, where only a few photons or less are detected. In similar conditions, the conventional mode readout noise would be as high as the observed phenomenon input signal!

Processing operations

Each digitizer includes a subset of linear processing operations. In analog mode (AM), each pixel’s intensity reflects the amount of light detected during exposure. In addition, linear mode comprises an image correction option, the bias mode, where a series of closed-shutter frames are combined then subtracted from an image to discard artefacts.

The electron-multiplying mode includes the same processes as described above as well as Photon counting (PC) operations. PC is ideal for extreme low light conditions: it removes the resulting stochastic nature of the EM gain, the excess noise factor. Binary instead of linear, this operating mode assigns a value of 1 for pixels that have detected one or more photons, while black pixels are assigned a 0 value. Accumulating and stacking several binary frames generates high precision images with minimum noise levels and high SNR.

IMO or NIMO: Modes of operation that strongly determine overall noise

A CCD inverted mode of operation (IMO) is based on the voltage application across the CCD detector in order to generate a population of holes within the silicon substrate. Holes then recombine with the dark electrons (the one produced by thermal excitation within the chip) prior to readout. As a result, IMO significantly decreases dark current.

The IMO trade-off is that the moving holes generate more CIC during the electrons vertical transfer towards the readout register. That’s why the IMO is best suited for long exposure times, where dark current dominates the total background noise.

On the other hand, non-inverted mode of operations (NIMO) is ideal for short acquisition periods where dark current has little time to accumulate. Of course, this is all true except in the case where the high voltage signals delivered by the EMCCD controller are finely tune to limit the excess CIC resulting from IMO. In fact, Nüvü Camēras’ patented CCCP — exclusively operating in IMO — is a state of the art EMCCD camera controller capable of generating such precise signals. A camera driven by the CCCP therefore allows a much greater SNR than any other sensor, whether operated in IMO or NIMO, no matter the integration time.

Note that IMO and NIMO are mutually exclusive operating modes; it is impossible to obtain the best of both worlds. As such, it is essential to understand how an EMCCD was operated while being characterized. In the case of Nüvü Camēras products, all data are obtained in IMO. However, other EMCCD manufacturers may provide CIC data in NIMO, while dark current data is obtained in IMO, thus giving the false impression of low total background noise. Given that both modes are mutually exclusive, the resulting background signal will inevitably be greater than expected when such manufacturer’s cameras are put to use.

Photon counting Mode

When expecting a 1 photon per pixel per exposure or less signal, the key is to implement a binary acquisition mode that includes an algorithm to analyze the image on a pixel-by-pixel basis. An arbitrary yet relevant threshold allows to statistically determine whether a given pixel did in fact originally contain a genuine photoelectron or not. The generated images are binary due to the involved thresholding. The stochastic nature of the EM process is no longer relevant to the SNR calculations, and the ENF is completely suppressed. By summing several frames together, it is possible to recover the camera’s dynamic range, which drops due to the nature of this acquisition mode.

Performances

In extremely low light applications, image acquisition using Photon counting (PC) possesses a significant advantage over traditional linear processing: the elimination of the excess noise factor (ENF). Still, in order to provide its theoretical plus-value, an EMCCD camera running in PC mode must also be capable of high frame rates; produce minimal clock-induced charges (CIC), and use an extremely high EM gain for optimal thresholding — all in order to detect the most incoming photons with the least noise. When properly combined, these elements yield the highest SNR in extreme low light imaging, as shown in the comparison below.

PC mode compared to linear mode, or analog mode (AM), using a standard USAF 1951 target. The first two images were acquired using Nüvü Camēras’ EM N2 512, while the third image is a simulation based on the best available specifications of other EMCCD cameras (without specifications as to IMO or NIMO operation). Accumulating and superposing 10 binary frames, each with an exposure time of 50 ms at a 10 MHz pixel rate and EM gain of 5000, hence producing the PC image. The AM image is a single acquisition of 500 ms again with a 10 MHz pixel rate and EM gain of 5000.

When every photon counts

Nüvü Camēras’ imaging sensitivity is based on its CCD Controller for Counting Photons (CCCP) that makes EMCCD devices faster and more accurate for low lighting applications, such as astronomy and biomedical research. CCCP’s great precision significantly decreases clock-induced charge (CIC), the main EMCCD camera noise source while maintaining rapid readout rates.

More

The obstacles in achieving high quality images in low light conditions principally lie in noise sources that are not apparent in standard operating conditions, but rather become predominant in the nearly total absence of light. Furthermore, in such conditions, high frame rates are primordial to obtain accurate and representative images of the observed phenomenon.

These two major obstacles are precisely those that Nüvü Camēras inventively addressed with the design of the innovative CCCP camera controller. As such, Nüvü Camēras’ imaging devices modernize the fields of astronomy and biomedical research: they provide a better understanding of the physical principles behind the stars and galaxies, but also detect even the faintest luminous signals emitted by medical diagnostic tools such as fluorescent markers or laser light scattered by in vivo tissues. In turn, these signals may be able to indicate early signs of diseases such as macular degeneration or early stage cancer.