Sharp-eyed AOI
09 February 2009
Because of the continual miniaturisation of electronic components, there are challenging requirements in the production process as well as quality assurance. Since electric test technologies are harder to utilise, Automated Optical Inspection becomes more important as a qualitative test method.
However, due to the structures getting smaller and smaller, the highest requirements are required for the image capturing and processing within such a system.
Basics of detail recognition
The application of CCD-matrix cameras within an AOI system has become established as quasi-standard. Compared to scanner solutions, AOI systems with area scan cameras are more flexible and offer a lower optical distortion at a higher resolution.
AOI systems’ recognition output is often evaluated in pixel resolution with the unit ‘µm/Pixel’ – regardless of how the image was captured. But this size is rather a theoretical value, for which the impact of critical lens parameters such as diffraction or aberration is not considered. A considerable contribution to detail recognition lays with the design of the optical system as well as the way it combines to the applied CCD-matrix.
One of the most important parameters when defining the optical resolution of lenses is their numeric aperture. It can be understood as a scale for the light collection ability of an optical system. Typically, it is determined by the lens aperture. Because of the physical optics theory, they are responsible for occurring diffractions. One of the possible results, for example, is the projection of an object spot as a more or less highly ‘blurred disc’. If there’s a CCD-matrix behind the lens, an additional light spot enlargement based on the pixel size is the result.
When looking at the entire image capturing chain, one finds that in the captured camera image, an area of 3x3 pixels has emerged from the original light spot with minimal diameter. More recently it has become obvious that in such a case an increase in pixel resolution (µm/pixel), from a higher pixel CCD-camera, is of no benefit in regard to improving the detail recognition. In contrast, it is clear that the resolution capability of the applied image sensor has not fully been used.
An analogue situation occurs in projecting a light/dark transition (optical edge). After leaving the lens, such an edge is also ‘blurred’ by a higher number of pixels. The resolution is mainly determined by the quality of the lens used. An increase in pixel numbers would not result in improvement. On the contrary, it would lead to an increase in data. Knowing these physical basics and considering what we know from commercial photography, it emerges that the quality of the applied lens is responsible for high-quality images rather than the number of pixels.
The optimisation of the applied optics is the only useful way to significantly increase the image capturing resolution. Considering the CCD-matrix to be utilised, the lens must be designed in a way that the optics’ blur is smaller than the pixel size of the applied CCD-matrix. An effective resolution increase in terms of detail recognition is possible by such a pixel-adapted lens with a constant pixel number in the camera. The utilisation of a lens built on this basis not only results in improvements in spot-wise projection, but also for image capturing of optical edges.
Utilisation of pixel-adapted lenses in the OptiCon systems’ camera modules
01005 ICs are, at the moment, the smallest applied components and have a dimension of 0.4mm x 0.2mm. Typically, the solder joints are approximately 0.15mm x 0.08mm. At a pixel resolution of 21µm/pixel, this means that the solder meniscus is displayed at ca. 28 pixel (approximately 4x4 pixel) – the entire component only covers 180 pixel, which is sufficient for a safe component detection and solder joint inspection.
In reality, these pixels are not available in the required quality for sufficient feature recognition – limited by the applied optics’ resolution ability. The usage of a lens, which wasn’t ideally designed to the camera’s pixel geometry, leads to a ‘blur’ of the relevant characteristic areas. Just increasing pixel resolution (to 10µm/pixel) doesn’t give an improvement of the recognition results. In contrast, a lens design to the camera’s pixel geometry allows an inspection without increasing the pixel number or pixel resolution.
Based on this physical regularity, the image capturing concept of the OptiCon AOI system family has been consistently developed further. By utilising a pixel-adapted lens, detail recognition is possible that guarantees a safe inspection of 01005 components solder joints on 0.3 pitch. As a matter of course, the well-proven telecentric view for image capturing without any parallax errors is kept.
For an improvement in statistical safety during processing by the respective recognition algorithms, the resolution was increased to 10.5µm/pixel, based on a feature optimised image transformation. Additionally, this higher pixel number enables a better visualisation; and it is beneficial for operations such as the manual adjustment of test areas.
For the inspection of the smallest components and solder joints, the exclusive increase in pixel resolution (by CCD-cameras with higher pixel number) does not give the suggested increase in respect of the required detail recognition. The utilised lens is the typically limiting component in such an optical system. In order to achieve the maximum detail capturing opportunity, it has to be designed to the pixel size of the applied camera in terms of its optical resolution ability. Subsequent transformation processes increase the statistic safety of the used algorithms, and contribute to an ideal visual representation and operability.
Jens Kokott and André Hacke work for GOEPEL electronic GmbH
Contact Details and Archive...
Related Articles...
Most Viewed Articles...