Sunday, May 23, 2021

Sony SenSWIR Reverse Engineered

TechInsights publishes a nice analysis "Groundbreaking SenSWIR Sensor by Sony - IMX990/IMX991."

"By moving away from pixel-level bump bonds and taking advantage of greater miniaturization in Cu-Cu Direct Bond Interconnect (DBI), Sony was able to reduce the pixel size of the InGaAs/ROIC SWIR imagers down to 5.0 ┬Ám. This makes the IMX990/IMX991 the smallest pixel-pitch InGaAs-based SWIR image sensors commercially available on the market.

Sony Semiconductor Solutions Die-to-Wafer Hybridization Advanced Packaging Process
Additionally, Sony has developed a highly scalable Die-to-Wafer Hybridization process, with wafer-level hybrid bonding to help manufacture its new SWIR imagers. In Die-to-Wafer Hybridization, a handle wafer supports an array of pre-manufactured SWIR InGaAs Die so that hybrid bonding can take place much like conventional CMOS Image Sensors CIS-to-ISP wafer-level bonding.

This approach presents two advantages, the Cu-Cu DBI can help reduce the overall height of the Die while Die-to-Wafer hybridization can help reduce the per-Die cost, thereby facilitating greater utilization of Sony’s SWIR technology for a wide range of applications."

6 comments:

  1. Do you know some medium to high volume for these SWIR sensors?
    I mean with such resolution.

    This market seems quite close to be a very very small one.
    Any killer app?

    ReplyDelete
    Replies
    1. silicon wafer crack chipping inspection.

      Delete
    2. There is one aspect apart from this being a SWIR sensor: Cu-Cu hybrid die to wafer bonding of 2 components of completely different process technology, even based on different materials. The traditional enabler of 'moore's law' - smaller structures - is nearing its physical end, single digit nanometer is in the magnitude of a few (10s) SI atoms. It will continue for a while but will become more and more expensive (the economy of scale is a important part of moore's law). I think we will see a continuation of moore's law through advanced packaging technology like the one used here by Sony in this InGaAs-SI hybridization. There is a huge drive visible in papers and announcements in this topic, not only in the image sensor world, not only cu-cu hybrid bonding, also other technologies to high density interconnect of "chiplets". First companies are ramping this into higher volume, tools seem to get available. Take Applied Materials and Besi for example: http://image-sensors-world.blogspot.com/2020/12/besi-and-evg-address-die-to-die-and-die.html, it seems volume production of die to wafer hybrid bonding with single digit micrometer contact pitch - 1Mio contacts per square mm - is around the corner, they talk of placemant accuracy in the 100nm range. Some numbers are given in current Besi investor presentation page 31-33 (https://www.besi.com/fileadmin/data/Investor_Relations/Investor_Presentations/Investor_Presentation_May_2021.pdf). We will see this in many applications quite soon and this is only the starting point. Contact pitch will go down into the 100s nm over the next years, maybe even stacked over multiple layers. When you imagine what you can build whit this, your fantasy is the limit.

      Delete
    3. Did you see the recent Computex TW speech of AMDs Lisa Su? https://www.youtube.com/watch?v=gqAYMx34euU - starting around 33:00. And related report at https://www.anandtech.com/show/16725/amd-demonstrates-stacked-vcache-technology-2-tbsec-for-15-gaming showing a 900nm TSV pitch for hybrid bond. If TSMC can make Mrs. Sus announcement reality, hybrid bonding with contact pitch in the um matnitude is about to enter volume production with Ryzen 5900x by end of 2021.

      Delete
  2. Where us your limit for "medium size"? I think the potential market for this sensors and the next generations to come is quite attractive. As mentioned, it is more scaleable than the microbump based hybridization. The sensor feels like a killer for existing sensors e.g. in the target automation apps, food inspection, waste sorting, semiconductors. There are nice effects you can utilize, silicon gets transparent, water opaque. There is no sunlight at 1400nm imagine what apps you can realize with that? A dtof at 1400nm could use much higher illumination energy, water absorbs light before it hits the retina. Etc... i think as prices for swir sensors drop due to such progress like demonstrated by sony we will see plenty of apps for it

    ReplyDelete
  3. Just fogp/haze penetration capabilities for automotive & drone applications could justify the development of these sensors.

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.