Following on from an earlier question about bayer filters, it got me wondering:

How are they actually manufactured? How do they apply such a small amount of dye to each sub pixel?

My best guess would be some sort of optical based chemical etching followed by a dye bath... (Per colour)


From the abstract of this article Color filter array for CCD and CMOS image sensors using a chemically amplified thermally cured pre-dyed positive-tone photoresist for 365-nm lithography

Diazonaphthoquinone-novolak photoresist is used to produce these filters by successively depositing and patterning each color layer.

My understanding of this quote is that process is as follows

  1. The photoresist (a polymer and sensitization agent that is relatively transparent in the visible but strongly absorbing in the UV starting at 450nm) with an added dye(red, green or blue) is deposited over the sensor surface
  2. A metal patterned mask is placed over the photoresist
  3. The unmasked region of the photoresist is exposed by a UV light source upon which the diazonaphthoquinone is transformed
  4. The mask is removed
  5. The exposed photoresist is dissolved with an aqueous solvent, the unexposed area does not dissolve as readily and therefore remains
  6. 1-5 are repeated using different dyes and different masks

See en.wikipedia.org/wiki/Diazonaphthoquinone for description of how this photoresist works.

  • I suspect there should be a baking step either between 5 and 6 or after 6. There might also be a baking step between 1 and 2. Aug 7 '15 at 14:07
  • Nice link - However the article is now 16 years old, so given the MASSIVE increase in ccd/cmos chip density since that, this may not be the current method... for example is it physically possible to create a metal mask fine enough for today's chips? Aug 10 '15 at 9:17
  • Has it been a massive increase? I think the pixel density has increased by about a factor of 10. The linear feature size would therefore have only had to decrease by a factor of 3. However, you are probably correct in pointing out that placing the mask directly over the photoresist is probably not how it is done. It is much more like that the mask is imaged onto the photoresist and simultaneously reduced as would be necessary for patterning the electronic devices of the sensor. I will edit the answer to reflect this. Aug 11 '15 at 4:30
  • After further consideration I think a contact mask, as I first described, is more likely used for patterning the color filter. The resolution of the metal mask can better than 100nm. The pixel size is on the order of 1um. Aug 11 '15 at 8:26
  • One issue I have not included is the formation of the micro lens array, which if present, can separate or an integrated part of the color filter. Aug 11 '15 at 8:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.