Rob Coppinger looks at the inspection systems found on electronics production lines, both visible and X-ray variants, and finds that, when it comes to manufacturing electronics, throughput is everything
Electronics make the world go round, from dishwashers to cash machines and the stock exchange’s high frequency trading computers. In electronics manufacturing, productivity is key – and vision can ensure that quality control and traceability are maintained over production runs of millions of units per year.
Semiconductors, printed circuit boards, flat-panel displays and electronic components are packed with the aid of dispensing machines. Nordson Asymtek, based in Carlsbad, California, provides such automated dispensing machines. These require vision systems for dispenser alignment so components can be easily located.
To determine the precise location of a workpiece and automatically correct its misalignment, Nordson Asymtek uses vision system components including cameras, image processors, and software. This inspection system works using a camera that acquires an image of the part and, using a greyscale correlation vision algorithm, locates the workpiece position. The vision system then guides a dispensing head to a specific location on the workpiece for processing.
One way that the vision system identifies the correct workpiece orientation is by detecting small circular pads, called fiducials, which are used as reference points. The software can also aid with orientation detection. Within the software, the workpiece’s edges can be identified as fiducials. In production, the vision system captures an image of each workpiece after it is loaded in the machine, and determines both the corner location and the angle of those two sides. The vision system compares the measured workpiece locations with the locations specified when the dispensing software program was created. It calculates the distance the dispenser has to move and the angle it has to rotate.
The company behind that technology is Waterloo, Ontario-based digital imaging and semiconductor company Teledyne Dalsa. Brad Finney, vice president for sales in North America, tells Imaging and Machine Vision Europe about the developments that are making vision systems for electronics inspection even faster.
Lighting has changed dramatically in the last five years, moving from halogen lights to the LED lighting that can be purchased: ‘LEDs are inexpensive, low-heat, low-cost and it’s so amazing at how fast they can turn on and off,’ says Finney. Users can charge a capacitor with all the power the LED is able to handle and then deliver it over a short period of time for strobed systems. ‘They [lighting systems] were much more difficult to achieve in the past and would have been expensive and hot and tricky, but now with LEDs you get it almost for free,’ states Finney.
Strobed systems help speed up inspection because before, users would have to stop the object to image it – but now the object can continue on the conveyor and it’s the very short duration of light pulse that freezes the motion instead of a physically stopping the part. ‘This opens up new imaging modalities that can be a lot faster,’ says Finney. ‘If you were going to have many images to cover the whole region to look for fiducials or special parts, instead of step and repeat [stoppages], you can just fly and use the strobe to freeze motion and you can finish that panel much faster.’
Having such an ability with a strobed light has driven the requirement for higher frame rates that Finney’s clients are expecting – and, subsequently, total throughout and processing capacity. ‘If frame rates are going from 10 frames to 60 frames a second and if you want better resolution, you’ve got a double whammy: higher speed and higher resolution, so an increase in total bandwidth demand,’ says Finney.
Computers are keeping pace with that processing capacity requirement as they move from 32-bit operating systems to 64-bit and, one day, 128-bit. But between each computational generation, machine vision companies have to tackle the transition. ‘Some of the trends that are occurring in the industry are a move from 32-bit to 64-bit and that is interesting at a hardware level, for frame grabbers that are going for higher speed and whatever is beyond,’ says Finney. ‘As soon as you add a frame grabber you are not at the user level, you are at a kernel level, and so we have to deal with the 64-bit application and the 32-bit operating system. The 16-bit to 32-bit transition was a pain; we’re 15 years on from that. Now we’re in the 32- to 64-bit transition and in a couple of years everything will be 64-bit. Maybe 10 years from now we’ll do a 128-bit transition.’
Such improvements, in Finney’s view, are bringing an end to the problem that inspection is the bottleneck in the production system. Once the products can flow past the camera without stopping, combined with greater computing power, vision systems make inspection faster than any other process. ‘Automated inspection has benefited tremendously; the throughput has increased tremendously. Throughput is everything when you talk about the factory floor, because if a machine has a higher throughput, you can have fewer machines,’ says Finney.
Another use of vision inspection in electronics production is imaging barcodes and data codes. Asys, an international company that designs and manufactures handling and marking systems, uses imaging systems from Microscan for its products. Asys’ machines carry out additional traceability checks using barcode scanners and imagers. On the production line, the codes on the electronics boards that go into a machine can have additional information, such as pertaining to its components. Those components can be counted to ensure that the correct number has been received, with inspection machines flagging up when a board has not enough components so that corrective action can be taken.
For all these imaging systems, footprint size is important. ‘We are required to build smaller and smaller machines,’ says Swen Nothjunge, Asys’ product manager. ‘When Microscan launched the Quardus Mini, we quickly realised its reading potential and saw that it could easily be integrated into our machines.’ The Quardus Mini imager reads all standard 2D and linear barcodes, and it can also decode multiple codes per capture. In addition to its reading performance, it measures 25.4 x 45.7 x 53.3mm and weighs 57 grams.
Asys also uses Microscan’s Mini Hawk imagers in their machines. With multiple resolutions and high speed configurations available, Nothjunge says: ‘We plan to recommend the Mini Hawk to our customers from now on thanks to its reading performance.’
Inspecting deeper with X-rays
While strobing lights may ease image capture in a fast-flowing production line and barcodes enable better traceability, sometimes a deeper inspection is needed. For this, X-ray machines are employed to see through the parts of the electronics.
Inspection of electronic components needs to be quick and accurate. Credit: Teledyne Dalsa
General Electric (GE) offers an X-ray machine called Phoenix. Not inline, like Microscan and Teledyne Dalsa’s products, Phoenix will sit to one side of the production line and be loaded and unloaded manually. Phoenix allows for the non-destructive inspection of electronics solder joints. ‘We have a fully shielded system, lead shielded, with an X-ray tube inside. We have a manipulation system to move the sample and a detector to collect the shadow. That is what we are looking at, the shadow,’ says GE’s Dr Tobias Neubrand. That shadow can be magnified by moving the sample towards the X-ray tube, but keeping the detector at the same distance. This can be useful if a certain area of the subject under study needs to be magnified for closer inspection. However, the Phoenix has a 1μm resolution, which is far greater than the normal 2-5μm resolution used for electronics inspection. ‘We have an X-ray source with a very small focus spot down below one micrometre,’ says Neubrand. ‘You can have a complete view of the solder joints and the solder quality. When you have a new solder paste, you can check for voiding in the solder joints and if the solder is placed right.’
Making the X-ray shadow as easy to interpret as possible is key for the inspection personnel. To get a very clear X-ray image requires a very good detector. As Neubrand explains: ‘When you have a sensitive low-noise detector it will give you very good images, which will be very easy to interpret and make a decision to recognise what is in the image.’
X-rays can also be useful when a solder joint is not easily visible due to other components being in the way. This is one situation where the X-ray is more useful than optical alternatives, according to Neubrand.
Although not part of the production line, the X-ray process can also be automated. ‘You can automate the system in certain steps; the machine will move the part for predefined settings and the operator can check the quality,’ says Neubrand. ‘The next step is to apply image processing to classify the solder joint, and the computer makes a decision to quantify the quality of solder.’
Automation is helping to make the inspection process as quick as possible. Electronics manufacturing is a high-productivity process. Being able to inspect a small sample of products rapidly, and confirm that their properties such as solder joint quality are within tolerances, is important. ‘We see the most critical part is to keep the time for inspection as short as possible. You can do this with automation; the computer doesn’t have to think, it just calculates,’ says Neubrand. But it’s not always possible. ‘On the other hand there are certain restrictions; as the electronic boards become more complex, very often you can’t do automatic inspection.’
Whether automatic or manual inspection is used, imaging will continue to play a role in inspecting electronics for quality and traceability purposes. The high productivity of the electronics production line lends itself to automation, but the human element that can cope with the unexpected cannot be ruled out. While vision systems have advanced greatly, the human brain is still the best interpreter.