Automating LCD displays testing with machine vision
The human eye soon tires of this and dead pixels are then missed.
Obviously, an automated system is the solution but we could not find an off the shelf solution, so we had to build a Machine Vision-based one using a camera and Vision Option software from National Instruments to detect dead pixels.
The first question was whether or not to test the incoming displays from the supplier or to test them once assembled into products or at both stages. Doing both would have doubled the costs as two sets of cameras and optics would have been required along with two software licenses.
We therefore decided that the most cost effective solution was to test the assembled product at the functional test stage, as this would highlight both any faults in the display’s pixels and any faults that had occurred in the assembly of the control electronics.
This would result in a simple pass or fail for each unit. A fail could be broken into two types. Random pixel failures meant that the display had dead pixels and would be sent back to the supplier. Groups of pixel failures in certain patterns meant that the control electronics had a fault for us to fix.
However, when we set this all up, the camera showed strange coloured patterns, which meant that it was impossible to work out if any pixels were malfunctioning. These are Moiré patterns caused by the lines of sensors in the camera’s CCD optically interacting with the lines of pixels in the display.
Even the slightest difference between the line-up of the two sets of parallel lines resulted in Moiré patterns. A common example of the Moiré effect is a bridge over a motorway where the vertical railings on the near and far sides create a changing pattern as you drive towards the bridge.
Fig. 1: Moiré patterns.
Getting rid of the Moiré effect was a major challenge. There are two variables. The first is the distance between the camera and the display as this affects the relative spacing of the lines and the second is the amount of angle offset between the two, i.e. rotating the camera relative to the display means that the interference between the lines changes.
In theory, one could calculate the optimum solution but this would require detailed knowledge of the physics and optics of the components so the practical approach of adjusting both variables to find a working solution was used and only took a week.
We initially thought that aligning the camera and the display in parallel would mean that there would be no Moiré patterns but actually the best angle is 45 degrees. Combined with the optimal distance between the camera and the display and the Moiré pattern disappears.
As a result, the automatic testing now only takes a couple of seconds per unit and is highly reliable as the human element is removed. Removing the cost of the human element is also important, as several people would have been required on a rotation basis to avoid eye fatigue.
Fig. 2: Rotating by 45 degees removes Moiré patterns.
Automating as many processes as possible is the key to being competitive in contract manufacturing. China is losing its edge of ultra cheap labour and customers are realising that highly automated production facilities can be price competitive with the advantages of being located nearby to ensure quality and a short distance and time for goods to be delivered to their markets.
About the author:
Thomas Eschenmoser is Section Head Test Engineering at ESCATEC – www.escatec.com