Twenty years ago, you would have needed a VHS video recorder, a TV, a record player, a camera, a landline phone and a home computer to perform the same functions as Apple’s new iPhone, not to mention a whole rack of CDs and tapes to store the music. Now all of these functions, combined with a mobile internet connection, are contained within a handheld package of the same weight as a small bar of chocolate.
This trend for concentrating higher powers in smaller devices is global and it is evident in everything from our mobile phones and laptops to the safety mechanisms in our cars. These advances are almost entirely down to new production techniques that have allowed manufacturers to downsize their electronic components while massively increasing the processing powers and memory storage.
It was inevitable that the technology used to inspect the production line would have to develop accordingly. Whereas each component was once manually inspected, to ensure that all the different components were placed on the circuit board and soldered correctly, much of this inspection is now performed automatically by a machine vision system.
The new systems can inspect smaller components at a faster rate than had previously been possible to allow for the high throughput of these devices that has ultimately led to their affordable prices. Many of the new vision technologies can even see through components
to find faults within the circuit board itself or behind a component’s casing. For manual inspection, which is still necessary in certain circumstances, vision suppliers are now offering ergonomic microscopes that reduce the eyestrain associated with inspecting miniscule circuit boards for hours at a time, day after day.
Machine vision is applied in all the steps of assembling a circuit board with components, the first of which is to check that the unpopulated board has been manufactured correctly. Printed circuit boards are typically made from very thin alternate layers of an insulating substrate and copper, which are layered alternately in specific patterns to provide a conducting pathway for the circuit in place of wires. Tiny holes, coated with copper, provide a link between the different layers and act as contacts for the different components.
Manufacturers need to inspect the production of the printed circuit board, often layer by layer, to check that the correct patterns have been created.
‘Automated optical inspection systems are used to inspect the routing of each layer, the inner and outer layer circuit patterns, the drill hole patterns and the solder masks,’ says Christophe Robinet, strategic marketing manager for machine vision at e2v, which provides sensors for these systems.
Further downstream, vision systems are used to inspect the components to ensure that they are placed into the correct part of the circuit, and that the solder has been applied correctly.
As with all machine vision applications, the ability for one of these systems to judge reliably whether a part is faulty is determined by the illumination, the cameras and the software used in the installation. A good source of illumination is crucial to ensure that the images provide the correct information. Many applications make use of structured lighting, made with rows of LEDs that are positioned in ingenious ways to highlight the parts of components that need to be inspected.
Even with excellent illumination, it is still essential to choose the camera correctly to discern the different components. ‘You need enough grey-levels in each pixel to discriminate the components from a dark substrate. If you are inspecting metal packages, this is easy, but with dark packages it is difficult to discriminate the components from the substrates when it is viewed from the top,’ says Robinet.
However, many applications also require the cameras to transfer the data at very high speed. A camera that can discern more levels of grey also needs to transfer more data, and this can bog the system down so that it cannot function at the high speeds of the production line. ‘It is a challenge to serve both needs. As with life, it is difficult to do something quickly very well,’ he says.
An automated optical inspection system from Viscom.
This raises the question of what type of camera, area or line scan, is suitable to accommodate these requirements. Line scan cameras capture a single line of roughly 1,000 pixels. The components would typically run past the camera on a conveyor belt, and the system would build a 2D image of the circuit board from the successive rows of pixels that have been captured. This is ideal for the production line, which can run continuously in front of the camera without stopping.
Area scan cameras however, which may contain an array of more than one million pixels, capture each image instantaneously. This requires the conveyor to stop in front of the camera each time an image is captured, which can slow down the process. However, because area scan cameras capture all the information in one go, rather than in successive lines, they can afford to use a longer exposure time per pixel, to provide an image with a better contrast.
Once the image has been acquired, it must be interrogated by image analysis software to determine whether the component matches a preset standard. Often, this standard takes the form of a ‘golden image’ of a perfect circuit board, against which each new component is compared through a pattern matching process.
The alternative is to enter the parameters into the algorithms oneself, which may be more complicated for the user, but it is also more robust against insignificant alterations to the production line. ‘The algorithmic approach measures the position of each component by analysing its outline,’ says Gerd Rademann, head of international sales at Viscom. ‘Predetermined values are held in a part library, but the user can enter the tolerance himself. It’s more tolerant against a change in the process.’
For example, a manufacturer may decide to change the supplier of one of its components. These components may look slightly different, with a different colour, even though they have exactly the same specifications. A golden image may not be able to account for this, but it is likely that the algorithmic approach would not be susceptible to such small aesthetic changes.
It is also important that the systems are robust against changes to the set-up of the inspection system, otherwise an alteration to the inspection parameters or the vision equipment that improves the detection of one type of error could introduce errors into another part of the inspection task.
To combat this, Viscom installs systems with ‘integrated error verification’. Rademann explains: ‘We have a system where we can make changes to optimise the set-up and then measure whether the errors have been eliminated and check whether it still finds the past errors. We call this a “trusted change”.’
So far, all the applications discussed have concerned vision in the visible spectrum, but many applications require x-ray and infrared vision to see faults that would otherwise be invisible. This is particularly important for applications such as automotive safety with a very low tolerance threshold.
‘The majority of the market uses cameras for visible light, but with some components the solder joins are not visible. X-ray vision is getting very popular, particularly because the automotive market is so demanding,’ says Rademann. ‘X-ray vision would be too slow [to complete all the inspection tasks] so we use both visible and xray vision to share the task and achieve a high throughput.’
Whereas x-ray vision is used to enhance visible inspection, infrared vision is used in an entirely separate area of the electronics market – the semiconductor industry. Most semiconductor wafers are transparent to infrared radiation, but like glass in visible light, cracks and faults show up because they scatter the light.
The production of smaller and smaller devices is providing a greater need for this kind of inspection. This new micro- and nanotechnology includes MEMS (micro electromechanical systems), which are commonly used as motion sensors used to control devices such as the Nintendo Wii and the iPhone. MEMS are constructed in large arrays on a single silicon wafer, where they are vacuum-sealed within a silicon casing and diced into separate entities.
The IR cameras can see through the casing to find problems within the devices. ‘The MEMS market is exploding,’ says Taufiq Habib, a technology and applications wafer inspection engineer at Viscom. ‘One important task is to check the seal. Even using x-rays, it is very hard to see a difference, whereas our method gives a clear insight.’
The solar cell industry is another growing market for these applications, where IR imaging is finding usage alongside UV imaging to find faults below the surface of the wafer. ‘Many manufacturers use deep-UV to look for small defects. However, this can only find small defects on the surface, whereas we can look inside for bigger defects within the wafer,’ says Habib.
In addition to these new technological markets, it seems that the need for vision in electronics inspection is growing geographically too. ‘The market is looking very good in Asia,’ says Rademann.
Whatever the application, it appears the overall demands for these systems are consistent throughout the different markets. ‘The market is requesting higher throughputs. As cycle times get shorter and shorter, the inspection times are getting less, so we need to use larger cameras with a greater field of view and better resolutions. It’s an ongoing process,’ he says.
It is generally accepted that one day there will be a limit to this miniaturisation; the devices can only become so small until a completely new technology will be necessary to take the devices in a completely different direction. While it will be interesting to note how the inspection systems adapt to this technology, machine vision has proved itself so essential to the current developments that we can rest assured it will always play a key part in assuring the quality of our electronic devices.
‘The best computer in the world is the brain. If you couple that with the best quality instruments, you have the best viewing system possible,’ says Colin Wells, international product manager of Vision Engineering.
While automated vision systems do perform the majority of the inspection of circuit boards on the production line, it is still sometimes necessary for a human to view the circuit boards themselves, often to check the results of the automated system. But it is a job few would envy, spending hours at a time hunched over a microscope inspecting intricate circuit designs. The result is often severe eyestrain and a bad posture.
To combat this, Vision Engineering is now providing ergonomically-designed microscopes that reduce these problems. ‘Our machines have a different head that allows the user to work under ambient light in an ergonomic position,’ says Wells.
An ergonomic microscope from Vision Engineering aids PCB inspection.
The company has achieved this by providing a bigger eyepiece, meaning that the user’s head is not restricted to one position to view the sample. ‘The typical microscope would have an output pupil of a few millimetres, so the user would have to stay in the same position (while viewing a sample). Our microscope has an output pupil of 30mm, so the user can get comfortable in their working position,’ explains Wells.
‘The trend is that automatic optical systems will be used, but there will also be the requirement for manual systems, and the highest quality viewers with the best ergonomic designs will be necessary for health and safety measures,’ he predicts.