"We expect that overall imaging revenues (CMOS + CCD array sensors) are set to moderately grow at a CAGR of 5% within the next 5 coming years. However, if CCD image sensor sales are set to decrease with time, this is not the case of CMOS image sensors as we forecast CIS sales to grow at a CAGR of 11% from 2010 to 2015. Actually, CMOS image sensors already surpassed CCD sensor revenues in 2007 and this trend will be even accelerated with the emergence of a wider 300mm CIS infrastructure in Asia and the introduction of future CMOS BSI(Backside illumination) technology."
Here is a revenue-based CIS market shares (click on picture to enlarge):
I think that BSI structure makes CIS process definitively different from CMOS process. So the CIS industry is now completely similar to DRAM industry.
ReplyDeleteBSI process module is an addition to a "conventional" CIS process. In that sense it's not that different from other process modules added to the standard CMOS process to make it CIS process.
ReplyDeleteOn the other hand, the standard CMOS logic is going toward 3D integration, including TSV, wafer thinning, bonding, etc., not unlike BSI processing. DRAM and flash processes are also moving toward multichip stakes with TSVs, etc., etc. So all these process modules become "mix and match" and still retain some degree of similarity between all the process flavors.
Well still a lot of logic and conventional CMOS designs can be done with very conventional CMOS process. The need of BSI development will make CIS process very expensive, so if there is no a consequent market share to support this expense the Fab can not survive at the end.
ReplyDeleteThis will make CIS fab very specialized ... Despite of the similarity between DRAM CMOS and standard CMOS, all these fab are specialized.
That is what I would like to say.
Even FSI CIS process deviates from the standard CMOS one. It has a lot of additional process modules. BSI, in theory, can make image process process to be closer to the standard CMOS by using standard metals instead of thinned, eliminate light pipe steps, neglect nitride layers in dielectrics. Instead it adds BSI processing module. It's not immediately clear to me which process is closer to the logic CMOS, FSI or BSI.
ReplyDeleteWell the optional process in FSI uses the same equipements. But the BSI needs special equipements that standard CMOS process doesn't need.
ReplyDeleteI think that you are both right:
ReplyDeleteBSI is closer that FSI to standard CMOS from an architecture stand-point. On the other hand, BSI is more disruptive than FSI from a manufacturing stand-point as specific additional steps such as laser annealing, wafer bonding and thin wafer grinding are necessary for BSI realization.
OK, and with CMOS logic moving toward 3D integration with memory and other logic, I'd guess the difference in equipment would be not that large in 5 years from now.
ReplyDeleteThose market numbers are not right.
ReplyDeleteCanon as a factor in CIS?? C'mon on.
The cy2009 numbers for Omnivison are overstated by $100 million, but the entire market is overstated by nearly a factor of two.
2009 is a throw away year in any case.
Non-captive served market for Omnivision for 2010 should be about $2.7 bil. They will get about 34% of that number. Even that number is probably bullshit because they could get 30% of their sales in modules (whether CCs or conventional modules) so that distorts the numbers.
Omnivision will be at least back to where they were in 2008 or about $800 mil....maybe a lot more.
SEMIWIZ
What is the percentage of every application for CIS? for exam. security market
ReplyDelete