Out of sight

Machine vision has long been an integral part of electronics manufacturing, but improved systems and lower costs have made it more accessible. Mark Venables reports.


Two of the most common uses of machine vision are traceability and visual inspection of solder joints — both tasks that have benefited from recently-introduced technology.

German electronic circuit board (ECB) producer EbV Elektronik uses a network of 42 In-Sight 1010 vision sensors from Cognex, which were integrated into its Burbach production line to make the process smoother and more efficient.

At this fully-automated plant the production of ECBs is structured so that small quantities can be handled with extreme flexibility, and large-scale series can be processed at high throughput speeds. Zero tolerance for defects is top priority. Quality and efficient automation are the defining factors, with traceability and documentation from cradle-to-grave — by means of various codes on the components and circuit boards — vital in giving the company a competitive edge in a tough global market.

‘On one hand, we must be competitive in the consumer-markets of low-cost products, the so called China-markets,’ said EbV’s Burbach managing director Andreas Fries. ‘On the other we must meet and fulfil the highest quality demands. And all this is topped with the customer’s requirement for individual and distinctive flexibility in very short reaction times. Even a small fault in production could trigger a stop on the line and this we must avoid at all costs’.

The company’s primary goal since opening the site has been to enhance circuit board production and raise automation and quality to a new level. It has also placed a large emphasis on development and production of control systems for original equipment manufacturers (OEMs) used in such areas as heating, control engineering, weighing technology and white goods.

The objective was to integrate a fault-free continuous structure of total traceability throughout the process chain, from beginning to end. The permanent documentation of all process steps must include an exact quality control of the written direct part marks (DPM). Immediately after being marked with a laser, the data matrix codes are checked for a number of testing-characteristics by means of vision sensors. Quick and reliable reading of the codes in all stages of production is of fundamental importance for the tracking process.

The results are then stored as archives and include all fabrication data collected up to the time the product is delivered to the customer. With the integrated Fast Ethernet-bus and with minimal spatial requirements, the complete product family of In-Sight vision sensors unites the advantages of intelligent cameras and complex image processing systems.

The sensors were integrated throughout the complete project, Elektronik claimed, because of the reliability, accuracy and sensitive performance of the algorithms and the Cognex vision tools, which achieved 100 per cent reading rate quality. Even if 60 per cent of the matrix codes are obstructed or damaged, they still can be identified. The code-reading tools are already integrated in the camera as a library, and are based on Cognex’s PatMax vision software package.

The project, from planning through to installation, was carried out within a year and is said to be unique on this scale. The simple installation and the ergonomic user interface of the sensors mean that if updates of the respective batches in Printed Circuit Board (PCB) production should become necessary, simple programming by means of a spreadsheet will permit the machine operators to react promptly.

The permanent and reliable documentation of manufactured products provides complete transparency and understanding of the production process and any eventual problems. Easily adapted parameters enable quick process modifications. Focus is on prevention and damage control and any problems can be identified and localised at an early stage, allowing the right decisions to be made at the right time

One of the most complicated machine vision tasks to complete is solder inspection of PCBs. This is primarily due to the issues surrounding reflections from the metal, the organic nature of the solder flow and the additional problem of flux on the part. All these give vision systems a tough time when it comes to automatic inspection of solder. Historically PCB manufacturers have had to use very expensive custom-made inspection machines that give the end-user little scope in setting up and making changes to the inspection criteria and screening levels of acceptance.

Industrial Vision Systems of Oxfordshire has created a generic machine concept to allow robust inspection for a fraction of the price of some of the larger automated optical inspection machines on the market. The latest generation of NeuroCheck Firewire cameras and systems software gives users the latest generation of machine vision that can cope more reliably with changes in light and surface quality.

Solder inspection requires high magnification of the joint to accurately confirm whether it is good or bad. This system provides the ability to concentrate light around the joint area with a combination of darkfield, on-axis and direct LED lighting. The strobing light arrangement captures the solder in differing lighting conditions through the image tray facility of NeuroCheck. Each condition can then be processed according to a different check routine.

Typically, four solder joints are checked at the same time, giving an approximate field of view 9-10mm square. Each check within the global routine offers a unique set of image pre-processing functions to allow for differing faults to be analysed.

Optimised LED lighting was used to give overall clarity of the image and offered the best overall solution to the changing surface conditions. Upon acquisition of the image, appropriate pre-processing algorithms are assigned, based on the quality of the captured image. The next stage utilises neural network classification for reliable detection of the solder position and other features. Neural networks enhance the solution with the capability of cognitive intelligence.

The network can automatically train itself against the presented images, but does not utilise specific differences for individual samples — thus offering an advantage where surface quality is poor and changing. This information is enough to develop a system with the necessary intelligence to guarantee reliable recognition.