Common platforms deliver optimised solutions for machine vision systems
06 December 2018
The quality increases and productivity improvements associated with the adoption of industrial machine vision systems have been an integral driver in the development of Industry 4.0 and the IIoT. And as this piece explains, as the demands placed on machine vision systems have evolved over time, so too have the requirements placed on the image sensors that drive them – leading ultimately to broad diversity in their features and capabilities.
For the digital issue of this piece, please visit this link – or click here to register for EPDT's magazine.
While such diversity can provide end customers with image sensors designed specifically for given applications, it can also impose considerable stress on camera manufacturers, forcing them to develop and maintain camera solutions for a wide range of image sensor families and variants.
If not properly managed, this fragmentation can consume resources and delay the development of new camera lines, meaning end customers may not benefit from the latest technological advancements as soon as they could otherwise. Ultimately, this can impact the introduction of new solutions to the industrial sector, delaying development of machine vision systems and the benefits they can provide.
Diversity in machine vision
The need for a wide variety of different imaging solutions arises directly from the close alignment of the performance needs of a machine vision system to its targeted application, where the relative importance of parameters such as resolution, bandwidth, light sensitivity and more can vary from application to application.
One type of automated inspection may prioritise image resolution, while another may require high dynamic range – each to ensure that accurate measurements and decisions are made. Overall light sensitivity may be critical in situations where lighting cannot be fully controlled, while other applications may require sensitivity in specific spectral bands, such as near-infrared or ultraviolet.
In applications such as an assembly line process, moreover, maintaining a minimum frame rate may be critical to ensure that the machine vision system does not act as a bottleneck to the overall manufacturing flow. Requirements such as these need to be evaluated independently and then optimised for each specific end use case (see Figure 1 – links to ActiveMag).
While it may seem simple to assume that for each of these parameters ‘more’ always means ‘better’, in reality, many of these features are interrelated. Increased resolution slows down frame rate, unless additional changes are made to the sensor to increase bandwidth – which then increases the power required by the device.
In addition, having ‘more’ imaging performance is not always beneficial to the overall system, as unneeded resolution or frame rate can clog network resources. In the end, a balanced approach is needed to developing and optimising an imaging system.
Moreover, the requirements being placed on machine vision systems continue to evolve, leading to a corresponding need to drive the development of additional technologies. Low-noise architectures are required to extend linear dynamic range for the capture of high-contrast scenes. Depth imaging is needed not only for pick-and-place automation, but for proper characterisation and volume measurement in automated logistics systems.
Multispectral and hyperspectral imaging enable new levels of automated sorting. Each of these emerging technologies adds an additional layer of complexity to the task of developing cameras for this market or application.
Adopting the best approach
With the wide range of performance needs associated with the machine vision market, camera manufacturers are faced with a challenge: how to develop a portfolio of cameras while minimising camera development efforts and expenses, and accelerating time-to-market.
A clear option is to leverage camera development by platform, modularising camera features such as output interface (GigE, Camera Link, CoaXpress, USB and so on) to allow them to be easily deployed. But to be truly impactful, this type of platform design needs to propagate to the image sensor level, allowing the camera electronics to be leveraged in support of different image sensors – with options for resolutions, bandwidths, pixel performance and more.
Figure 2 illustrates how this type of integrated image sensor design can work in practice. Instead of requiring a separate camera design for each image sensor, ON Semiconductor’s PYTHON family of image sensors allows a single camera design to be leveraged in support of resolutions from VGA to 25 megapixels, and incorporating additional options for multiple spectral sensitivities and output bandwidth. Camera manufacturers who design to this image sensor family can minimise camera development costs, control inventory and speed time-to-market.
The new X-Class platform then extends this functionality by supporting, not only multiple resolutions, spectral sensitivities and speed grades, but also different pixel functionality as well – such as global or rolling shutter pixels, extended dynamic range, low noise and so on.
Because all members of this platform share both a common high bandwidth and a low-power image sensor frame, camera manufacturers who work within this platform can easily leverage their camera designs to support a full portfolio of cameras: one that spans a variety of different resolutions, spectral sensitivities and pixel functionalities.
As an example, the first two devices in the X-Class platform are based on the 3.2 µm Global Shutter XGS pixel. The XGS 8000 provides 4K/UHD resolution (4096 x 2160 pixels) at up to 130 fps, while the 12 megapixel (4096 x 3072 pixel) XGS 12000 operates at up to 90 fps. Both devices are housed in a compact package that, when combined with the low-power footprint of the X-Class platform, enables the development of cameras with a 29 x 29 mm2 footprint.
Camera manufacturers who design to this platform can support not only these initial devices, but will soon be able to add additional resolutions based on the 3.2 µm XGS pixel, as well as new pixels being added to the platform. This will simplify the work needed by camera manufacturers to develop a full portfolio of advanced imaging cameras – appropriate not only for machine vision, but applications such as intelligent transportation systems (ITS) and broadcast markets as well.
Selecting the most appropriate technology for any given application is not a simple task, and, in any case, it may actually be the wrong starting point. What is perhaps more important today is choosing a platform that can grow and evolve, supporting a path of continuous improvement that has the lowest impact on product development.
Image sensor designs such as the X-Class platform provide the flexibility to grow, by includingother resolutions and pixel features – allowing camera manufacturers to continue delivering the benefits of a common platform for many camera generations.
Contact Details and Archive...