Wednesday, December 15, 2010

TowerJazz Presentation at ISE 2010

I was given TowerJazz presentation at Image Sensor Europe in March 2010 titled:

Foundry business for CMOS Image Sensors
Dr. Avi Strum, VP & GM Specialty BU

Below are few slides from the presentation talking about the company's approach and technology achievements:

A slide showing stitching quality for large sensors:

BSI becomes an option for TowerJazz customers:

Lightpipe is another available option:

All in all, it seems that access to a lot of advanced technologies is now available to the smaller players on the image sensor market.


  1. Anyone can give an idea about the QE FSI/BSI with microlens? The comparison without ML doesn't make a lot of sense.

  2. BSI QE approaches 90%. One can hardly justify microlens in that case. Well, may be only for crosstalk reduction.

  3. what QE can be expected with a good microlens?

  4. Agree. Would like to see comparison of BSI vs FSI+ulens, and to state pixel size.

    One also wonders what QE and angle response looks like for waveguide vs no waveguide.

  5. Oops, must have missed something. Doesn't BSI sensors use microlenses?

  6. Of course BSI sensors use microlenses.
    BSI/FSI gain depends of a lot of factors, particularly wavelengths.

    On Tower marketing charts, a very uncompetitive FSI example has been chosen.
    20% peak response in monochrome corresponds to approximately 40% with an optimized microlens. With color filters, it remains 30% peak QE in green. Best 2.7µm pixel achieve more than 60% QE peaks. For me BSI is not required for 2.7µm.

    Regarding angular response of waveguide, organic material must have a refractive index as high as possible, if it is too close to oxide one (1.6 for example) no guide effect is achieved.


  7. @ "Of course BSI sensors use microlenses."

    Sure, microlenses areessential for color BSI sensors. Talking about large B/W pixels with close to 90% QE, like in Tower presentation, the microlenses necessity is dependent on the imager spec. I can think of many scenarios when microlenses cost is hardly justifiable.

  8. Dear Vladimir,

    Then I would like to know why microlenses are designed on Tower chart!
    I'm not sure that microlens process is so costly compared to BSI process. But there exist scenarii when microlenses are impossible to implement, more for material specs reasons.


  9. Talking about Tower-specific case of large BSI pixels, the microlenses are important for a color sensor version to direct light away from the borders between color filters. Normally these borders are rough in structure and scatter the light all around increasing the color crosstalk. I'm not speaking for Tower, but I believe this is the reason that their slide shows microlenses combined with color filter.

    However, the QE graph shows B&W sensor response, not related to the bottom picture. If QE touches 90%, and the pixel size is 2.7um, what benefit you see in using the microlens? QE increase from 90% to 95%? Minor MTF improvement for large apertures and CRAs?

  10. Well I think that the BSI spectral response of this specific case would be different using color filters, crosstalk would be seen particularly at sensor edge.
    Microlenses would help controlling the light injection and the location of photogenerated carriers decreasing the crosstalk risks. To my opinion it would help sensor color balance. Of course there is a trade off with final sensor specs requirement.



All comments are moderated to avoid spam and personal attacks.