WP's Photo

Fantastic Colors

High-speed matrix sensor integrated circuit

Enjoy the live photo shot during the camera filter assembly, when putting a glass based stripe filter onto an operating(!) CCD sensor. The action was necessary because the manufacturer did not offer a color version.
And it has worked reproducibly. Thanks to Fraunhofer Institute (FhG) IOF in Jena we could recognize 150 nanometer shifts of the filter.

On the right a photo of the black-and-white CCD EG&G Reticon HS0512JAQ area sensor, resolution 512 pixel x 512 pixel at 1 000 frames/sec. The active are, the dark square in the middle, measures 8.19 mm (almost 1/3 inch) edge length.

Δ

Object

Each photos was made with an opened black-and-white CCD area sensor, which means one with removed glass cover and thus bare silicon chip. View direction: from the active sensor through a drop of glue on the sensor surface onto (and through) a red-green-blue (RGB) stripe filter on glass substrate into the light source. The algorithm for color recovery is partly inactivated yet.

The the photo on the left depicts the glue drop between sensor surface and the stripe filter. The color algorithm is switched off.

Camera

Easy spoken the black-and-white area sensor works as its own camera. The exposure is 1/1 000 sec. The photos are a little bit post processed (contrast and etch enhancing) because too much light would have cured (i.e. harden the glue) before adjusting the filter.
The alignment is done under optical control using the sensor itself, with an accuracy of ±1/4 micrometer, because the fill factor is close to 100%.

What you see

Explanation: the upper left image was made in the clean room during filtering the black-and-white sensor. The bright lines are the green ones of a red - green - blue - green line pattern on the glass substrate. Each line is 16 microns wide, with an overlap of approx. 1 micron to each neighbor line.
Viewing at the stack from the side you would see the sensor at the bottom. Then the glue in an eggcup shape, approx. 0.5 mm high. And on the top the glass substrate (thickness: 2.0 mm) with the filter pattern.
What you see in the image (just a suggestion): the egg-shaped contour shows the original size of the glue drop (on the sensor) before it has gotten in contact with the glass substrate. The dark ring is caused by the flank of the glue eggcup, because the viewing direction is against the light source. Its circumference is the contact area between glue and glass substrate. The inner circle is the optical path through the glue given by the smallest diameter of the eggcup. The bright spot in the middle is caused by the lens effect (focal point) of the glue drop.

It is necessary for the color recovery that the corresponding algorithm is familiar which color belongs to each pixel, thus the color filter is fixed and clearly aligned to the pixel. If this is not given misinterpretation and thus fail colors happen. In the image on the lower left the strip filter is not close enough lowered to the sensor surface. Of course, the not already completely spread glue drop has some influence, too.

The the photo on the left depicts the moment when the stripe filter reaches the position on top of the glue drop, but has not been lowered. The color recovery algorithm is activated.

Δ

Is what you get

Hand optimized color sensor

End work: after finishing the glue fills the whole gap between sensor die and glass substrate. The distance between them is then approx. 20 to 30 microns. The glue is hardened by UV exposure and the cavity of the sensor (open bond wires!) is sealed by ceramic/epoxy filling and cured in an oven.

The photo on the right shows the finished color coated wide ceramic DIP CCD sensor, molded, cured and ready for the integration into the socket of the camera head.

This so modified chip was exclusively used between 1998 and 2001 in the color version of the high-speed camera Weinberger Speedcam Pro. For non-scientific cameras the about three hours effort would have been just to expensive.