The miniaturization of pixel size impacts performance of light sensitivity and SNR in today's 1.12um pixel image sensors. BSI technology helps improve sensitivity, but is said to fall short on elevating image quality. Toshiba developed its newest image sensor with BSI and CNR integrated on the sensor to address both low-light sensitivity and SNR. As a result, the Toshiba CMOS image sensor provides approximately 1.5 times higher SNR than a 1.12um pixel sensor with no CNR.
Samples of the new Toshiba sensor, the T4K37, will be available in December 2012. Sample pricing begins at $20.00.
Update: It appears that Japanese PR has much more info. Among the pictures, it shows the CNR effect:
Here is the Google translation of CNR principle:
"CNR Our circuit has adopted the type to compare the image between frames by a digital filter circuit to suppress the color noise, without compromising a sense of resolution, of 1.12 micrometer pixel circuit not installed CNR SN ratio of about 1.5 times to accomplish, to achieve the SN ratio of the pixel corresponding to 1.4 micrometers."
The sensor is fast, it delivers 30fps at full 13MP resolution. The Japanese PR also has a table with two other sensors, apparently based on the same 1.12um pixel:
Update: Toshiba published a product page of the 1/3.07-inch T4K37 sensor. The block diagram shows a location of Color Noise Reduction part (CNR). The output format is either 8b or 10b.
Why does CNR seem to increase sensitivity (brightness) of the image too? The image comparison is not easy to see to begin with but the only difference seems to be the signal level
ReplyDeleteI believe the sensitivity difference is from RGB VS RGBW color filter array
ReplyDeletewhat has RGBW to do with CNR?
ReplyDeleteThe square root of the brightness ratios is about 1.5 or so, corresponding to the reported SNR improvement. Maybe CNR adds several frames together with motion cancelled out, or something like that. Maybe I am thinking too much about digital integration sensing these days...
ReplyDelete