Friday, February 06, 2009

WLC Challenges Cataloged

Semiconductor International published a very nice article "Inside Wafer-Level Cameras". The authors from SUSS summarize the successes and problems in creating WLCs, with an emphasis on problems. Their list of challenges is frighteningly long:

Optical design
  • Wafer-level cameras require novel strategies for optical design and stray light management. Very few optical designers have WLC experience.
  • Integration of actuators for autofocus, mechanical zoom or shutters is difficult or impossible.
Lens material
  • The lens material must be suitable for high-throughput imprint lithography, UV-curable, reflowable at 260°C, and must have long-term stability in harsh environments.
  • Hybrid wafers from polymer lens material and glass substrates show issues during wafer-level integration (warping, bowing) due to thermal expansion mismatch. Glass-only solutions may have a competitive advantage.
  • For chromatic correction, different lens materials (crown and flint glass types) must be available.
  • Additonal material problems are introduced with antireflective coatings on polymers, apertures or baffle layers and IR cut filter.
Metrology
  • Optical testing of aspherical and free-form microlenses is extremely difficult.
  • Wafer-level, high-throughput lens testing is mandatory to maintain high production yields.
Wafer-level packaging
  • Opto-wafers are usually not perfectly planar. Bow, warp and double-sided optics make it especially difficult to mount a stack of 3–6 different wafers.
  • Gluing and sealing for the wafer stack is not trivial in terms of reflow temperature restrictions.
  • Dicing of the complete WLC stack consisting of different materials is complicated and significantly impacts yield.
Legal aspects
  • WLC technology and through-silicon via (TSV) technology is protected by numerous patents.
  • The profit margins for mobile phone camera suppliers are already small. WLC manufacturers might simply not earn enough to finance the new technology ramp-up and to pay the license fees.


Thanks to R.C. for pointing me to this.

2 comments:

  1. That paper has been out for a while. The part that I thought was interesting was that you needed to have TSV or BSI to make it work, but most people are just concentrating on TSV.

    Here's another thing to consider. As far as I know, only Omnivision (in a patent application) thinks that they can make modules at the wafer level and then dice them. Everybody else seems to want to build them one at a time. What's the point of doing them one at a time? Isn't labor a big part of the module cost right now? Maybe you could bring cost down through automation, but I'd still think you'd need to do builds at the wafer level to have much of a cost advantage.

    ReplyDelete
  2. innovations are infinite.It is important to be updated..Thanks for the post:)

    ReplyDelete

All comments are moderated to avoid spam.