Blog_banner_beyond_the_datasheet

OEM metrology cameras enabling 2 Gigapixels per second at low power – how is it possible?

Posted by Gretchen Alper on Fri, Oct 17, 2014

The state-of-the-art high-resolution global shutter CMOS image sensors (and subsequent metrology cameras) enable performance leaps for many measurement methods relied on in semiconductor and electronics process control systems.  BUT, the extreme data rates present major challenges for OEMs in their new innovative systems. For camera builders these data rates are just as challenging in terms of camera component selection, small physical outline design, increased power dissipation, heat management within the camera and more.

In order to take full advantage of these image sensors such as the CMOSIS CMV12000, major advances were required in several supporting technologies including camera interfaces, FPGAs, internal camera data buses, etc.  For instance, the data buses, which go between the image sensor and the FPGA need to be able to run at frequencies of 500 MHz and FPGAs need to run at very high clock speeds.  Now the supporting components are available in the market as well.  These high speeds are sustained with advances in the CoaXPress high-speed machine interface standard.

Adimec quartz cmv12000 cxp low power metrology camera

So now that the technology is available, the challenge is to put it all together in a small, low power, reliable camera.  Because of the high clock speeds on the FPGAs and the multiple data buses required there is a heat build-up in the components.  If this is not managed, it can negatively affect the performance of the camera.  Some camera manufacturers implement a fan for cooling, but this increases power and reliability issues as well as size requirements. Similarly active cooling elements require more power and generate more heat that needs to be removed from the image sensor.

Other ways to manage the heat may require a complete redesign of the electronics platform to process 2 Gigapixels per second efficiently.  This is the approach Adimec has taken with our new QUARTZ Q-12A180.  One of the goals with this new platform was to enable all of the camera functions necessary for best system design without any memory requirements in the camera; adding memory requires more power and generates more heat. This reduces the power requirement significantly, but in order to do this the image sensor must be thoroughly characterized in the initial design. With a full understanding of the sensor, calibrations can be done efficiently without bogging down the data coming out of the sensor.  We designed our image processing in such as way that only a few corrections (dark image and bright image corrections in offset and gain) are needed to preserve the accuracy of the information in every single pixel across the image.

After more than 6 years of CoaXPress design experience, we created a brand new design where the entire Adimec CoaXPress core is implemented.  We created a highly integrated 2nd generation CoaXPress V1.1 compliant 20 Gb/s CoaXPress Quad interface. This interface is also fully backward compatible to existing V1.0 frame grabbers through a configuration tool for all major frame grabber brands.

These are some of the design considerations that resulted in more than 30% lower power usage than other cameras at full frame speed (180 frames/second) and in the smallest mechanical outline available in the market resulting in the smallest camera with the best image quality.  This camera will be on display at the Vision Show 2014 – please stop by to check it out!!

Topics: New Adimec Products

Previous blog:

Next blog: