Your Digital Camera Is Obsolete: Japanese Image Sensor 100x More Sensitive Than Current Chips

Illustration for article titled Your Digital Camera Is Obsolete: Japanese Image Sensor 100x More Sensitive Than Current Chips

Right now, your camera either has a CCD (most point and shoots) or a CMOS image sensor (lots of DSLRs) inside, which converts pretty pictures into an electrical signal. Japan's Research Center for Photovoltaics has developed a CIGS image sensor that's 100 times more light-sensitive than the silicon chip inside your cam. It's able to shoot in environments as dark as 0.001 lux, or about as dark as a "moonless clear night." Obviously, it'll be great for night vision gear, but it also picks up infrared, giving this some serious Sam Fisher applications. Check out the comparison shot between a CMOS and CIGS below, it's insane. Chen won't need that invisible coat, just a good zoom lens.

Advertisement
Illustration for article titled Your Digital Camera Is Obsolete: Japanese Image Sensor 100x More Sensitive Than Current Chips

[Tech On]

DISCUSSION

LittleJon

You can't get 100x sensitivity of CMOS, not for the same pixel size and lens aperture.

A color CMOS sensor with smallish (say 2.5um) pixels will have a quantum efficiency (QE) around 50% (peak green response). That means that on average, every 2 photons generate an electron. To achieve 100x improvement on this you'd need to have a QE of 5000%. Seeing a 100% QE is the theoretically maximum, that's clearly impossible!

A monochrome sensor has a higher QE because there are no colored filters over each pixel that block much of the light. So for a mono sensor the maximum QE improvement that could be achieved is even less.

Sure, sensitivity is not the same as QE, as it also takes in to account other aspects of the pixel design, but sensitivity measures the sensor's response in electrons per lux second. A larger pixel will have a higher sensitivity as it's receiving more photons (lux) per second, and a higher QE will also produce a higher sensitivity, but as I explained above there's not much room for improvement available.

What these researchers are probably doing is extending the spectral response in to the near IR. While that's a valid thing to do, it's also a common trick used to create BS spec numbers even in CMOS or CCDs.

You remove the IR-blocking filter (which really messes up the colors) to capture light beyond the human visible range. However, when you take a light meter reading, you just measure the light in the visible range and get say 0.001lux. But the sensor lux is really much higher than this as the near-IR counts as lux too, just not lux that we humans can see. Voila, you've got a great press release or BS spec to put in your marketing presentation!