The advantages over the lens-based OIS are:
- Not only camera yaw and pitch, but also roll is compensated
- Very fast sensor movements, under 15ms time required for image stabilization, said to be 3 times faster than lens-based OIS
- Power consumption as low as 10mW, said to be 50 times lower than lens-based OIS
- High accuracy of of compensation of 0.3um, while lens-based systems are said to be 10 times less accurate
"So why is the SmartSensor just now arriving? Let's have a look at just how difficult it was to bring about this technology.
To achieve three-axis sensor-based stabilization, the image sensor needs to be suspended in a very small space to allow for rotation around a central axis.
Another challenge brought by suspension is signal transmission. The SmartSensor presents the challenge of having to connect over 200 signal wires from the floating sensor to the main board.
This was the key puzzle that OPPO and MEMS Drive solved through years of research."
GSM Arena reports from the MWC 2016 Oppo booth: "Oppo SmartSensor image stabilization can correct up to 1.5° of motion and works on sensors ranging from 1/1.8" (bigger than any phone yet) and 1/3" (typical of mid-range phones). It's also smaller than a typical OIS module, so we may see it in thinner phones." While being a very impressive technology, 1.5deg is somewhat limited range for a high quality OIS.
Apparently, the company that supplied the technology to Oppo is MEMS Drive, based in Pasadena, CA.
So the bonding wires need to be very solid !!
ReplyDeleteI don't know how they solved that, but I would suggest a flexi cable over the gap.
DeleteIt makes sense to use flexible interconnection symmetrically. Otherwise coupling between X and Y axis will kill this OIS technology.
DeleteGSM arena must mean that 1/1,8" is bigger then in any of _Oppos_ phones yet. There are several examples of other brands that have larger imaging sensors. Panasonic CM1 with 1", Nokia pureview 808 with 1/1,2", Nokia pureview 1020 with 1/1,5" and Nokia N8 with 1/1,8".
ReplyDeleteLooking at system cameras like Pentax K-1 its obvious that sensor based stabilisation can offer much more then lens based stabilisation.
ReplyDelete- Rotational stabilisation, including horizon correction
- Astrotracer, moving the sensor according to the earths rotation to avoid star trails when photographing the night sky
- Shifting the sensor one pixel at the time, taking several exposures and combining them will provide true RGB in every pixel, avoiding color moire and increasing the true detail level.
- Shifting the sensor like a shift lens to correct perspective.
- Shifting the sensor for subject tracking (not background)
Not implemented in the K-1:
- Shifting the sensor to take one exposure from each corner and combine these images to simulate a larger sensor.
In Olympus OM-D E-M5 markII they do the pixel shifting with another twist:
- Shifting 1/2 a pixels at a time, taking 8 exposures to combine them in a larger resolution file.
These techniques can also be combined with ordinary camera shake compensation or combined with each other.
I guess its much faster to do full readouts with small sensors compared to larger sensors. That will make the multi exposure features more useful on small sensors.
But sensor based is not so good for wide angle lenses - at least when the lens is a 'normal' lens and not a fisheye.
DeleteI think most of the applications you are describing would require linear translation range well beyond what MEMS actuators (and interconnects) can provide.
DeleteLooks like MEMS Drive is a small bunch of former DOC people. It will be interesting to see how the IP side works out.
ReplyDeleteI like to think of them as Siimpel folk.
Delete