Skip to main content

The pick of the pixels

Displays are getting larger with higher resolution. Most produced now are at least HD resolution, with ultra-high-definition and 4K screens coming through. All of this puts demands on the inspection technology, especially regarding the resolutions of the screens, where manufacturers are now able to produce pixels in the single digit micron scale. This poses an interesting question: how do you inspect a pixel that is smaller than the camera sensor pixels used to carry out the inspection?

‘Display pixels are now becoming so small that it is getting to the point where the object size is actually smaller than the pixel size on the image sensor. That’s making us change how we might inspect displays in the future. Things change when your object pixel size is smaller than the image sensor pixel size,’ commented Mark Butler, product management team leader at camera manufacturer Teledyne Dalsa.

Teledyne Dalsa’s line scan cameras are used in various flat panel display inspection applications, where, along with the higher pixel counts, display makers also want to inspect these panels extremely fast. The inspection speeds mean the responsivity of the camera becomes a major issue, noted Butler, which is why time delay and integration (TDI) cameras are used in this area. Multi-line CMOS is another option, such as the Eliixa+ cameras from e2v. Whether TDI CCD or multi-line CMOS, multiple exposures are necessary to get enough light for a good image at the inspection speeds found in flat panel display production.

‘You’re imaging so fast now – we’re pushing 100kHz and more – and you can’t get a lot of light onto a chip when you’re running that fast. So if you use multiple exposures, you capture the light multiple times,’ explained Butler.

Teledyne Dalsa’s Piranha TDI cameras can reach 256 stages, which provide effectively 256 exposures that all get summed together noiselessly. These cameras therefore give, theoretically, 256 times the responsivity boost compared to a single line.

E2v’s multi-line CMOS Eliixa+ line scan cameras offer 16k pixel resolutions, with the latest version summing the charge from four CMOS lines. The idea is the same as TDI CCDs, in that, at higher inspection speeds, higher sensitivity is needed. ‘The customer wants to inspect with a certain signal-to-noise ratio (SNR) – the SNR guarantees the robustness of the algorithms,’ explains Paul Danini, marketing manager at e2v. ‘Fewer photons will reach the sensor when inspecting at higher speeds and so, in order to get good SNR, the signal from multiple lines is summed.

‘The goal for running at high speeds is to have the maximum charge saturation, which can be achieved by adding more lines onto a multi-line CMOS sensor, by increasing the quantum efficiency of the sensor, or by increasing the full well capacity of the sensor,’ Danini continued.

Teledyne Dalsa also provides multi-line CMOS cameras with its Piranha XL 16k CMOS TDI model. Butler explained the difficulties with increasing the number of lines in multi-line CMOS sensors: ‘The CMOS TDI is summed in the digital domain, which isn’t as good because while you get higher responsivity, your noise goes up by the square root of the number of lines. While there’s some signal-to-noise benefit, overall you will get to a point where the total noise is too large. With CMOS there are generally problems getting beyond four lines of exposure. You have to take each of those lines and digitise them all in the time that it takes for the camera to read out one line. So getting beyond four lines is very challenging, as the internal speed requirements of the sensor become extremely high.’

He added that a TDI CCD is used where lots of responsivity is required, whereas a multi-line CMOS solution might be used in slightly less demanding inspection tasks, but where a better responsivity than a single line is an advantage.

Relying on line scan

There are multiple inspection points in flat panel display manufacture: Butler listed the glass, the LCD thin-film transistor (TFT), the colour filter array, the polarising film, and the actual panels themselves are all examined, among other checks. Most of these inspection points use a 16k-pixel line scan camera. Area scan cameras can also be used off-line.

The high 16k resolution is needed because the goal in flat panel display inspection is to have as low a takt time – the number of panels that can be manufactured and inspected in a certain time – as possible, commented Danini. This can be achieved either by increasing the resolution or increasing the speed of the camera. The latest Eliixa+ models operate at 16k pixels at 140kHz, or 200kHz at reduced resolution of 11k pixels. Danini said that the cameras are currently limited by the bandwidth of the CoaXPress 6Gb/s interface, but added that the next generation of the standard with 10Gb/s and 12Gb/s links will overcome this bandwidth limitation.

Flat panel display inspection OEMs use high-power light sources. Martin Hund at Chromasens commented that flat panel display inspection has many challenges with regards to lighting. ‘Most of the time it is a combination of several light configurations that require extremely homogeneous brightfield illumination, (to look for inclusions and scratches), and very powerful darkfield illumination (to check for particle contaminations). Both are very challenging and require a sophisticated design and good cooling [in the lighting system].’ Chromasens provides cameras and lighting for flat panel display inspection. The company’s Corona II series is a powerful illumination solution for line scan applications.

Automated production

Along with inspection of the actual displays, the mobile phone market is opening up a lot of other applications that use machine vision, noted Butler. These include inspecting the skins and other components of mobile phones, the edge of the glass, the casing, and so on. ‘The mass production of mobile devices is now so automated that pretty much every part of a mobile device is inspected, not just the screen,’ he said.

In terms of the screen, however, Hund noted that, in general, to be able to detect the smallest defects, the camera resolutions will have to be increased. However, it is still yet to be determined how to cope with display pixel sizes that are smaller than the image sensor’s pixels.

‘If the display pixel size under inspection is smaller than the pixel size of the camera sensor, either smaller sensor pixels must be used or higher magnification of the optics,’ commented Danini. ‘However, increasing the optical magnification leads to a reduced field of view, which in turn means the inspection has to be faster to cover the entire panel.’ Checks are made over different areas of the panel with a magnified inspection process.

‘There are multiple manufacturing processes, some of which have a field of view of 10µm which is more than the pixel size of the camera,’ Danini continued. ‘But for some of the more advanced manufacturing processes relating to the TFT array, for example, the display pixel size can be as small as 3µm which is smaller than the pixel size of the camera. In this case, display manufacturers need a higher magnification and, here, a higher speed camera is useful.’

Butler commented that a concrete solution for inspecting the smallest pixel components has yet to be found. ‘People haven’t quite figured out what the ideal solution is yet,’ he said. ‘There are many different variables going on at the same time; you’d want to try and shrink your pixel to do that, but once you start shrinking the pixel you get less responsivity, so there are limits there. At the same time, lenses start to be a limiting factor – there are many layers to that whole question.’

Different surface inspection tasks will present their own specific imaging demands, whether that’s inspecting flat panel displays or webs of material. Finnish company Sapotech has developed an inspection system for in-line monitoring of hot steel slabs during a continuous casting process. The system uses Cavitar’s Cavilux laser light source for area illumination to image the metal slabs for surface defects such as cracks and holes.

‘The advantage of inspecting the metal while it’s still hot is that defects can be detected in the early phases of steel production and a decision made quickly about how much more processing is required,’ explained Dr Taito Alahautala, CEO of Finnish laser lighting company Cavitar.

The Sapotech system can inspect all surfaces of the metal slab, providing visual topographical information in the images. Each laser is used to illuminate a small portion of the target and the resulting images are stitched together using Sapotech’s image processing algorithms to provide an image of the entire sheet.

The images are sent to a cloud-based service which gives the operator an overview of the part from which they can zoom in to review the surface of the slab in very high detail. Potential weaknesses in the slab can be identified easily.

The laser used is a pulsed 640nm red laser with 300-400W of output power. The light produces a narrowband wavelength, which means all the other radiation from the hot metal can be filtered out, explained Alahautala. ‘We’re applying the same technology for welding monitoring applications,’ he said.

Cavitar provides relatively high-power lasers for this application, although Alahautala added that, with short pulses and suitable illumination optics, the system itself is eye-safe (3R) and thus easy to use on the factory floor.


Media Partners