Today, a large choice of high framing cameras in the visible spectrum (Si-based sensors) are available on the market. The applications are well known and include vehicle impact testing, ballistic motion analysis, tracking of various objects etc. All require a sensor equipped with global shutter and short exposure times (typically in the microsecond range) to avoid motion blur. As a result of the short exposure times, high sensitivity (a combination of quantum efficiency, pixel size and readout noise) as well as dynamic range are important factors when comparing available options.

When it comes high-speed imaging beyond the visible, the options become limited and the choices are not always obvious. With the emergence of infrared-laser-based technologies in the recent years, the need for fast framing cameras with high QE in the 900 – 1700 nm has become increasingly stronger.

Free-space-optical (FSO) telecommunications benefit strongly from short-wave infrared wavelengths compared to visible or near infrared solutions as disturbances inherent to air/free-space (such as weather, water/dust absorption, scattering or scintillation) strongly impact the quality of the optical transmission. The use of SWIR-band lasers is all the more pertinent when it comes to cost/power ratio and eye-safety. As a result, InGaAs focal-plane-arrays are often used on the receiver end as part of an adaptive-optic setup aimed at providing a live correction of the wavefront. Correction in the kHz range requires an InGaAs FPA capable of outputting a large amount of data with low latency and fast electronic shutter.

Still involving lasers in the 900 – 1700 nm range, the need for speed is required for pointing analysis and tracking for instance. The problem with most solutions available on the market today is that one must often compromise on the pixel bit-depth/quantification in order to reach higher frame rates. Most manufacturers often share what I call the ‘best of their best’ results in their data sheets, often leaving out context including key camera settings such as pixel formal (bit-depth), thus letting customers to their own devices when shopping for the best solution for their application.

When choosing a high-speed SWIR camera, one should consider the following key camera attributes:

• shutter mode: in most cases a global shutter where each pixel begins and ends the exposure simultaneously is of the essence

• minimum exposure time: it is often a limitation of the sensor itself. Most InGaAs cameras have a minimum exposure time in the µs range with sometimes the option to go even lower (100 nanoseconds). Most cameras show a reduced shutter efficiency (readout noise goes up, dynamic range goes down) for exposure times under a few µs.

• windowing mode: most cameras will have an increased frame rate as the number of rows is reduced (horizontal direction). Some cameras allow windowing in both horizontal and vertical directions, thus enabling very high frame rates where only a portion of the sensor is read out. In some cases, several areas of interest (AOI) can be read out at once (search for ‘multi-AOI’ or ‘multi-ROI mode’).

• ADC vs. pixel format/bit-depth: the analog-to-digital converter (ADC) quantifies/digitizes the electric signal (electrons) onto a gray scale value.. The range of this gray value is given by the ADC resolution. For example, a 12-bit ADC will quantify a signal onto 4096 levels of gray with 0 being black and 4095 being white (saturation). It is important to understand that though the ADC is often chosen according to the sensor dynamic range, it is not always the case. In other words, ADC converter range does not equal dynamic range. Once a pixel is digitized with a certain bit-depth, the camera can either output raw data by maintaining the bit-depth of the ADC or rescale the pixel value before outputting. In most cases, the pixel value is downscaled in order to reduce the bandwidth and potentially increase the speed.

For example, a camera with a 14-bit ADC will quantify light between 0 and (2^14)-1 or 16,383 levels. The camera can output raw data in Mono14 (14-bit data requiring 2 bytes or 16-bit) or rescale the data to Mono12 (12-bit data also requiring 2 bytes or 16-bit). Some cameras will go as low as Mono8 (8-bit requiring only 1 Byte). The pixel format is specific to each camera. A pixel format smaller than the ADC resolution will always result in a loss of information. It is eventually up to the customer to pick the right pixel format for their application factoring in signal resolution and speed. This often results in a compromise as most manufacturers communicate their frame rates in a lower pixel format than that of the ADC.

• video interface: today the most common interfaces for InGaAs cameras are GigE Ethernet, USB 3, CameraLink and MIPI CSI-2. The video interface defines the max data rate / bandwidth. In some instances, the interface itself limits the data rate as the sensor bandwidth is greater than that of the interface! The interface and latency time are highly correlated.

• dynamic range: the analogy of a water bucket can be used to understand the dynamic range of a pixel. The deeper the bucket, the more water it can accept before overflowing. The same applies to a pixel accumulating charges until the saturation is reached. The maximum number of electrons a pixel can accumulate before saturation is called the well capacity (in e-). The linear dynamic range of a pixel is given by the ratio ‘well capacity / readout noise’. The value in dB is 20 log (well capacity / readout noise).

C-RED 2 Lite and C-RED 3 : Best-in-class high-speed SWIR cameras with few compromises

C-RED 2 and C-RED 3 are high-performance SWIR InGaAs cameras designed for applications requiring fast frame rates without compromising on pixel bit depth, sensitivity or dynamic range. In the table below, we compare the frame rates of 2 cameras using very similar sensors in terms of specs but for which the frame rate of C-RED 2 is greater by a factor of 3.6! C-RED 2 Lite and Goldeye G-034 are both cooled with a TEC1 while C-RED 3 is uncooled.