Bayer filter

Wouldn't it be more advantageous to change the bayer filter?
the filter is composed of a red green green blue repetitive matrix. green is repeated twice because human sight is more sensitive to green, and it makes sense.
But if one of the two green filters were completely removed, wouldn't there be more advantages than disadvantages? That photosensor would record 66% more light which, although monochromatic, would help significantly in removing noise at low lights.

You mean like what the 12k already does?

And it’s not really monochromatic, more like all the colours.

https://patentimages.storage.googleapis ... 6472A1.pdf

JB

You mean like what the 12k already does?

And it’s not really monochromatic, more like all the colours.

https://patentimages.storage.googleapis ... 6472A1.pdf

JB

Exactly. I didn't know the 12k used this system. I fell behind.

You have a naive understanding of how both bayer sensors and light work. White light on a computer screen is indeed red, blue, and green without overlap, but the amount of light quanta is not 1/3 1/3 1/3, and a sensor is not just an inverse of a display. Quantum Efficiency or QE is just the percent of the light that hits the sensor that the sensor can 'see'.

The 'green' channel on the camera sensor is already broad spectrum, seeing both blue and red light as well, and sensors usually have quantum efficiency over 50%, meaning each green pixel is already counting over half the light that reaches it. Without the filter, the sensor might have 70% or more efficiency, but not 100%.



Here's a sensor similar to what's in the BMPCC4k. The grey curve is what the camera would be sensitive to with no color filter array. It's not much better, a few percent more. The green curve overlaps red and blue and this is great because when we reconstruct the red channel, we can use the green photo sites to help us determine how much red they saw at the green location, rather than only interpolating between red photo sites.

If you're wondering why this sensor sees so much past 700nm, this is why having an IR cut filter that's well tuned is so important for good colors, the sensor behind the filter is easily contaminated because it is designed to also be used in industries that need to see IR, like security cameras.

It’s not really by design, but an inherent property of silicon as a photoreceptor.

I think there are sensors where the color filter array on the chip itself can cut the infrared, right? so it doesn't need another piece of glass in front of the sensor in order to do the IR cut. Its true that the silicon itself is sensitive which is the grey line on the chart for that specific sensor.

For example, the BMPCC4k uses a sony STARVIS sensor. From sony's own marketing - "STARVIS enables image sensors to deliver clear images in a near-infrared light environment. If the image sensor’s NIR sensitivity is sufficiently high, NIR lighting can be lowered, helping to reduce heat generation and power consumption." And it goes on from there.

So the IR contamination of the sensor color filters themselves in this case is by design.