OPTICAL PRODUCT
SCIENTIFIC HIGH-SPEED CAMERAS
in pixel responsivity has allowed the incorporation of ultrashort exposure times, now down to 38 ns.
SCIENTIFIC APPROACH TO SENSOR PERFORMANCE Some camera manufacturers use the EMVA 1288 standard to characterize image sensors. This is a scientific approach and currently is the best means for guiding camera users to an optimal image sensor for a given imaging application. Some of the key terms derived from EMVA testing are in Table 1.
SNR VS SIGNAL PLOT The Signal-to-Noise Ratio( SNR) versus signal plot, as defined by the EMVA 1288 procedure, is a crucial evaluation method for assessing the performance of image sensors. According to the EMVA 1288 standard, this plot is constructed by measuring the SNR at various signal levels, typically obtained from a series of test images captured at different light intensities( irradiances). The irradiation level is plotted on the x-axis, while the SNR is plotted on the y-axis. The SNR is calculated by the ratio of the signal amplitude( usually the mean pixel value) to the respective noise at that signal level. This plot reveals several critical values, namely the absolute sensitivity threshold or AST. This demarks precisely how many photons( at 50 µ s integration) are required to produce an SNR = 1, Figure 4, blue arrow. This is the key specification for low-light applications, like in fluorescence, bioluminescence, screen imaging, or applications demanding ultra-short exposure times. This plot also provides an SNR across the‘ mid-gray’ region of the sensor, and thus users can discern precisely how sensitive the sensors are to change at any light-level, Figure 4, green arrow. Lastly, this plot also provides how many photons are required to bring the pixel to saturation, which is critical information for those who are characterizing scenes with wide
Figure 4: The EMVA 1288 plot illustrates the relationship between the Signal-to-Noise Ratio( SNR) and incident irradiation on a sensor. The curve demonstrates four key performance characteristics: AST( absolute sensitivity threshold), Dynamic Range of the sensor, SNR at all irradiation levels, and the Saturation Capacity.
Table 1: Scientific Sensor Specifications( EMVA 1288)
TERM UNIT DESCRIPTION
Quantum efficiency( QE × FF)
Temporal dark noise( TDN)
Signal-to-noise ratio( SNR max)
Absolute sensitivity threshold( AST)
Saturation capacity
Dynamic range
‘ scene-dynamic-ranges’, see Figure 4, purple arrow. The higher the saturation capacity, and the lower the absolute sensitivity threshold, the larger the sensor dynamic range, Figure 4, red arrow.
TEMPORAL, SPATIAL, AND LIGHT-LEVEL RESOLUTION The sensor specifications should always match or exceed the requirements of the imaging application. To ensure the sensor performance is at least minimally viable( in terms of temporal, spatial, and gray-scale resolution), it always helps to utilize the Nyquist rate( f Nyquist) to define the minimum frame rate needed to prevent aliasing of the event rate( f event), where:
f Nyquist = 2 × f event
The Nyquist rate mandates that the event or entity being measured is sampled at 2 × the rate of the event. For
% Percent of photons incident on a pixel that get converted to electrons at the specified wavelength( λ). EMVA bundles QE and fill factor( FF) value together, and a singular QE x FF-value is generally reported, where λ ~ 532 nm. This term has a direct linear correlation to sensor responsivity.
e- Noise present in an image when there is no incident light on the sensor( i. e., lens cap on). This value is signal-independent, and represents the lowest noise value on a sensor, and is also traditionally known as‘ Read Noise’. This term is paramount for defining sensor effectiveness in low light applications.
ratio dB bits
p
Ke- Kp
ratio dB bits
Maximum signal-to-noise( SNR) ratio a pixel can produce. This value is extracted from the highest pixel response( i. e., right before saturation) since SNR trends with the square root of the signal. A higher value indicates better image quality and higher light-levelresolution in the mid-gray and bright parts of an image. Quantity of photons( p) required for a pixel to generate a signal that is equal to the noise( SNR = 1). This is measured for integration times of 50 µ s and fixed wavelength( i. e., 532 nm). The lower the number, the more sensitive the sensor, and the better a given sensor will perform in low-light applications like fluorescence, bioluminescence, ill-lit scenes & ultra-short exposure times. Amount of charge( Ke-) or photons( Kp) a pixel can take just before saturating. This is also known as Full Well Capacity( FWC). Since most sensors are Shot noise limited, the SNR max generally directly correlates with square root of the saturation capacity. Ratio between the max pixel signal measurable to the lowest signal resolvable. Represented in ratio as saturation capacity( SC): temporal dark noise( DN) OR SC: TDN; in units of dB as-20 log( SC / TDN), or in units of bits or‘ stops’ we have n = log 2( SC / TDN).
74 www. photoniques. com I Photoniques 131