| Budget Amount *help |
¥27,820,000 (Direct Cost: ¥21,400,000、Indirect Cost: ¥6,420,000)
Fiscal Year 2006: ¥7,150,000 (Direct Cost: ¥5,500,000、Indirect Cost: ¥1,650,000)
Fiscal Year 2005: ¥5,850,000 (Direct Cost: ¥4,500,000、Indirect Cost: ¥1,350,000)
Fiscal Year 2004: ¥14,820,000 (Direct Cost: ¥11,400,000、Indirect Cost: ¥3,420,000)
|
|---|
| Research Abstract |
In this study, a super high-sensitivity image sensor using a new noise reduction method called adaptively quantized correlated multiple sampling has been proposed and demonstrated through the theoretical analysis and implementation of image sensor chips. A switched capacitor amplifier whose gain is given by the capacitance ratio of the input capacitance to feedback capacitance is analyzed as a frontend circuit of the column parallel signal readout circuits and it is found that a very low-noise signal readout is possible if the freeze noise component of the frontend amplifier is cancelled in the second stage using a two-stage noise cancelling architecture. The effect of the noise cancelling has been demonstrated with the implementation of a prototype image sensor chip. By extending this idea to a pre-amplified correlated multiple sampling, an extremely low-noise signal readout circuit has been proposed. A noise analysis of the readout circuits with correlated multiple sampling shows that the thermal noise can be greatly reduced by increasing the pre-amplifier gain and the number of sampling. The input-referred noise level can be reduced to 0.2 electrons if the pre-amplifier gain of 32 and the integration times of 16 are used. The most severe problem in CMOS image sensors is an anomalously large noise generated in the in-pixel source follower amplifier, so-called random telegraph signal (RTS) noise. Techniques for reducing the RTS noise is investigated and it is found that a histogram-based noise RTS noise reduction method can effectively reduce the RTS noise with relatively slow relaxation rate. A technique for achieving both extremely low-noise readout and wide linear dynamic range is also proposed and its effectiveness is confirmed with noise analysis.
|
