"Termed a quanta image sensor, the device is implemented in a commercial stacked (3D) backside-illuminated CMOS image sensor process. Without the use of avalanche multiplication, the 1.1 μm pixel-pitch device achieves 0.21e− rms average read noise with average dark count rate per pixel less than 0.2e−/s, and 1040 fps readout rate. This novel platform technology fits the needs of high-speed, high-resolution, and accurate photon-counting imaging for scientific, space, security, and low-light imaging as well as a broader range of other applications."
A remarkable work! Do you have some demo images to share please?
ReplyDeleteThanks for the comment. One 1Mjot binary single-photon image can be found in the paper. BTW, the paper is open-access.
DeleteImpressive! one question here: was the 0.21 e- rms avg read noise measured at 1040fps in full resolution?
ReplyDeleteThe measurement is made with the analog output array that operates at a much slower speed than the digital one. Since the noise seems dominated by SF 1/f noise, slow is not an advantage and also the CDS time is faster (shorter) for the digital array. The jot and SF are the same for both devices. The log D log H curve tells us about the noise for the higher speed single-bit output but determining input-referred e- noise from the on-chip single bit quantizer output alone is complicated.
DeleteWhat's the unit of "probability density"? Why it is not added up to 1?
ReplyDeleteIt should integrate to unity and it looks like that to me. Which figure are you concerned about?
Deletethe y axis of the Fig 5(b). Also in the (a), if simply counting the data dots, it seems less than 1.
DeleteStill looks at least roughly right to me. The green curve should have an area underneath it equal to unity, which is the same area as a rectangle 20 high (full scale of the graph) by 0.05 wide. But, perhaps JJ wants to comment further.
DeleteI guess the confusion here is that one needs to do the integration with the corresponding x-axis scale. As Eric commented, the area of the curve equals to 1. The varied spreads of the data over x-axis make the y-axis scale look quite different in several plots. For example, the x-axis span for CG plots is ~60 and is ~0.2 for read noise plots (Fig. 5). As you normalize the area to 1, you get quite different y-axis scales.
Delete