By P. Merken, D. De Gaspari, K. Jacobs and B. Grietens
Industrial image processing in the near infrared (NIR) spectrum is gaining in importance as a viable, non-destructive analysis tool in production lines geared for quality assurance, higher productivity and environmental responsibility. In this broad context, a new NIR camera platform based on high-resolution linear sensors and reprogrammable during operation enables flexible deployment. The camera marks a breakthrough in simplification and cost reduction for NIR technology, which opens numerous applications in the realm of industrial manufacturing.
NIR Inspection Yields Detailed Information
Spectroscopy and image processing in the near infrared are proving themselves as reliable manufacturing tools for inline inspection and classification of a multitude of products. Since cameras operating in the visible spectrum deliver just information on RGB color differences, they can not distinguish between objects of like colors but different materials. This limits their usability for industrial image processing. Whereas the visible spectrum permits the analysis of an object's surface layers, radiation in the near infrared can penetrate deeper into the object and provides valid information about its molecular structure.
In this sense, every molecule identifies itself by its own spectral "fingerprint". Thus, NIR analysis tools can safely distinguish between various materials exhibiting just minor differences in their molecular structures by observing how they absorb radiation at specific wavelengths.
As a consequence, high-performance InGaAs area-array and linear sensors can provide important information for the effective control of manufacturing processes, realizing higher yields and efficiencies. One of the latest InGaAs line cameras, the Lynx 2048, is raising the technology to new peaks. The Lynx camera, featuring a linear 2048 pixel array arranged in a 12.5mm layout, delivers detail-rich signals, which up to now could be achieved only through complex multi-camera solutions. Lynx reduces not just system complexity; it also significantly lowers system cost.
High Resolution NIR Line Sensor
Industrial image processing in the NIR spectrum puts high demands on line cameras. They require high line frequencies and high resolution as well as low noise combined with a broad dynamic range. This primarily pertains to the sensor – if the sensor misses something in the original signal it cannot be reconstructed by the signal processing chain.
Therefore, the new Lynx camera is based on a linear detector array built by Xenics in InGaAs technology. The camera's signal output is pre-processed in a specifically designed CMOS read-out integrated circuit (ROIC). This combination of linear sensor and ROIC realizes the advantages of two technologies. First, the sensitivity of an InGaAs sensor extends well into the near infrared covering wavelengths between 0.9 and 1.7μm – far beyond the reach of CMOS and CCD (Figure 1). And second, an ROIC in CMOS technology optimizes the analog sensor signal pre-processing by complementing it with highly integrated digital functionality.
A first version of the Lynx line sensor (Xlin-1.7-1024) sporting 1,024 pixels was introduced at the Vision 2009 show. Meanwhile there are two versions out offering resolutions of 1,024 and 2,048 pixels, as well as four pixel formats from 12.5 or 25μm wide to 12.5 or 250μm high. The smaller pixel formats are destined for line scan applications at high resolutions. Due to their high sensitivity and quantum efficiency they operate at very low illumination levels. Linear sensors with largeformat pixels are primarily used in spectroscopy applications.
The high frame rate of up to 40 kHz (1,024-pixel version) and 10 kHz (2,048-pixel version) enables a high resolution in the time domain. In many cases, this leaves sufficient time resources for multiple sampling, thereby increasing the dynamic range through superframing.
In a standard configuration the sensor is fitted with one-stage thermoelectric cooling. If an extremely high signal-to-noise ratio is required, such as in spectroscopy applications under critical lighting conditions, the camera can be upgraded to threestage thermoelectric cooling.
Flexible Read-out IC
The ROIC used in the Lynx camera is based on an optimized detector interface. It includes digital functionality to adjust numerous operational parameters. Therefore, the camera designer can accommodate various pixel sizes. But, far more important to the user, it can also be tailored for a wide range of different applications.
The simplified ROIC block diagram in Figure 2 shows the five pre-processing steps of the sensor signal:
- The InGaAs photodiode with pixel sizes of 12.5 x 12.5μm up to 25 x 250μm (area ratio of 1:40) works with n-well capacities of 7 x 105 up to 2 x 107 electrons (capacity ratio of 1:28).
- The current/voltage converter CVC, which holds the photodiode to a constant charge during the integration phase, is equipped with five different capacities as charge integrators. These can be selected at runtime, singly or in any combination, yielding values of 5 fF, 25 fF, 100 fF, 500 fF and 2 pF. This results in a wide variety of conversion characteristics to adapt pixel size to application requirements.
- The stage performing a correlated double scan (CDS) compensates for offset variations in the current/voltage converter and eliminates reset noise. Since kTC noise is predominant especially with low integration capacities, say 5 fF, CDS is a necessity. With larger integration capacities, read-out noise is more of a concern.
- The sample/hold stage decouples integration and read-out, so that the charge of the actual frame is integrated while the previously integrated frame is read out.
- An analog multiplexer and pad driver feed all pixel values sequentially to the output and further to an external analog/digital converter.
Figure 3 offers a glimpse at the interior of the otherwise hermetically sealed package, which is fitted with a clear window. On the right-hand side there is a detailed view of the connection of the sensor line with the two ROICs. The two read-out circuits are placed above and below the sensor line, contacting even and odd photo diodes, resp. Thus, the ROIC connection matrix might be coarser than that of the sensor.
Versatile NIR Line Camera
With various sensor layouts and flexible ROIC, the new Lynx camera is well suited for a very wide range of scientific, industrial and medical applications, offering direct imaging and spectroscopy. Today's complex industrial manufacturing processes demand non-destructive inspection methods. Spectroscopy and imaging tasks enabled by line cameras permit continuous and consistent real-time supervision of complex manufacturing processes. The objective is process management with maximized production yields.
For convenient integration in the application environment the camera platform offers a very flexible user interface. Besides analog video signals, it also delivers 16-bit wide digital image data via high-speed CameraLink as well as a Gigabit Ethernet connection, which is GigE Vision compatible. Parameter adjustment and camera control are done via serial interface. Synchronizing images with external events is ensured by the trigger input, which offers several trigger functions. A trigger output delivers information about the start and end points of a frame and/or integrator status.
Among the camera's pre-processing capabilities is a superframing mode, which increases the effective dynamic range. If one or several pixels within a frame tend to go into saturation it is obvious that the illumination level could be still larger and that it would make sense to capture and resolve this range as well. In this case a renewed sampling at lower sensitivity (by means of a shortened integration period) will yield more accurate data. Up to four of these subframes are feasible, and in many applications acceptable, due to the high line sampling frequency.
The camera housing (Figure 4) is suited for a variety of practical uses. It can be fitted to a spectrometer via threaded bushings. It also can be fitted with various C-mount compatible lenses.
The Lynx camera belongs to a class of products that can be used in many flexible ways and excel by their systemic openness. To support system designers and end users alike, Xenics has devised several software products for the camera. The graphical user interface Xeneth performs easy camera control and image capturing. It stores still images and videos in various data formats. It also calculates histograms, line and time profiles and enables false-color rendering with various different color profiles.
The Xeneth-SDK (System Development Kit) is available as an application development tool for Windows platforms (interoperates with C, C++, C#, Visual Studio, Visual Basic and Delphi) as well as Linux (C and C++). Sample codes and Labview prescriptions are rounding off this userfriendly functionality. As a result, the camera functions can be very accurately adapted to the application requirements.
Sample Application: NIR Examination of Thinfilm Solar Cells
Electricity generation using CGIS (copper-gallium-diselenide) thinfilm solar cells shows significant benefits such as lightweight and flexible modules. They save energy and materials in the manufacturing process and yield and good efficiencies over a broad spectral range, even under unfavourable weather conditions.
Solar modules can be examined by analysing their near-infrared electroluminescent radiation through the use of highly sensitive InGaAs cameras. This shortens the module development cycle. Module production will benefit by ensuring component uniformity, aiming for highly reliable, efficient and long-life photovoltaic systems.
The significance of examining the electroluminescent radiation in the near infrared is clearly recognizable in Figure 5. At the start of a corrosion test (left) the solar module radiates across its entire surface. After a prolonged hot steam treatment over 1,000 hours, the test item - obviously not assembled in optimal fashion – shows severe corrosion (right-hand side), especially along its edges. This practically cuts the energy output in half.
With their inherent high resolution and usage flexibility, the Lynx 1024 and Lynx 2048 NIR line cameras can substitute for costly multiple-camera solutions. More important, they enable many new sensor projects, which up to now had to be relegated due to high cost. System designers can now benefit from the high validity of infrared examination in various industrial manufacturing processes.