The ASIC is first grown on the bottom, then the MEMS mechanical elements are fabricated above and connected with tiny 3 micron vias after which the MEMS element is hermetically sealed with a silicon cap. (Source: mCube)
docdivakar: I put the question of whether mCube's process is "really" monolythic, and they insisted it was and justified their claim with the following statement:
"mCube uses a process that creates a single piece of silicon (hence the word ‘monolithic’) connecting MEMS to CMOS at semiconductor tolerances (3um). The result is a smaller die, that has least connection parasitics and requires only one chip probe test.
Alternative approaches use two fully processed pieces of silicon & are wire bonded or stacked together either by bond wires or bonding completely processed separate MEMS and ASIC wafers at assembly tolerances.
While at package level, the later approaches can make single-chip claim, they suffer order of magnitude larger area for interconnect & higher parasitic capacitances and require two separate chip probes one for CMOS and one for MEMS."
All this adds up, I think, to it being "more monolythic" than the other approaches and in a manner that drastically shriinks the die size.
Thanks so much for the question.
Sorry, I missed a few points in the post below... from a technical sematics point of view, mCube is not quite correct in using 'monolithic' for their gyro product. It is truly a 3D heterogeneous stack with via's, granted they may not be TSV's but via's nonetheless! How ever, I don't want this to overshadow the progress mCube has made by product releases of this type of vertical integration. I am quite happy to see this, long time in coming with all the buzz that started five years ago on 3D stacking. I hope there will be many more products like this by other companies.
If you see page 4, the word 'monolithic' has to be taken with a grain of salt! There is a monocrystalline MEMS Si wafer bonded to the CMOS wafer using MEMS via. The MEMS part is then capped with another wafer. So in effect there are two wafer-to-wafer interfaces that are quite different both in behaviour and function from metal-to-ILD interfaces of truly monolithic Si CMOS process. The reliability of this MEMS interfaces is the one I was questioning.
I did not find anything on their website on rel data, perhaps you can enlighten me more tomorrow?
@Colin_Johnson: regarding "mCube can use three-micron vias to interconnect the ASIC and the MEMS elements of a monolithic chip..." -though the 3um via many not be that impressive in monolithic technology, in this context I think it is a breakthrough! How ever, I would like to see reliability data prove the long term stability and durability of CMOS-integrated MEMS.
The iGyro uses sensor fusion on an accelerometer and magnetometer to synthesize the rotational output of a gyro while using 80% less power, costing 50% less, and taking up almost 60% less space than a hardware gyroscope.
Wow less power, less cost and less space. I am sure this technology will be used many other smartphones and in all kind of wearable devices.
Yes you are right they might only be selling it to China and Taiwan, as in the clients list they are not showing any global player except Acer. And in local clients only one giant ZTE they are mentioning. This may be due to their agreement with purchasers of not disclosing the list using their devices.