Why not use dichroic filters in modern solid state cameras?

[sophiecentaur]

In the days of analogue TV with vacuum tube cameras, the colour filtering was done with dichroic filters . Dichroic filters are more efficient than light absorbing filters and characteristics can be tailored better (?) than with buckets of pigment. Looking at a typical three tube camera layout I can see that the combination of dichroic filtering and splitting the incident light uses a high percentage of the light energy than when just using three filters.
I have probably answered my own basic question when a Bayer filter is involved; it would all be too complicated at pixel level. But high quality colour photography could be achieved with three sensors. Size would hardly be a problem when you realise how small a phone camera is. Maybe the optical path would need to be longer to get the dichroics to work properly.

I know that dichroic filters are used for regular photography and astro but not at the level of a Bayer.

I did some searching around but couldn't see an answer to my specific question.

 

[Andy Resnick]

Three chip color cameras do exist, I've used one in the past. Besides requiring more components, the mechanical alignment tolerances are also stringent, so the total cost of the device is significantly higher.

An alternative approach, Foveon (now Sigma) developed the X3 color chip which conceptually is a lot like color film:

https://en.wikipedia.org/wiki/Foveon_X3_sensor

Why this chip has not achieved commercial success is unclear (to me), my guess is the manufacturing complexity is still too high.

 

[sophiecentaur]

conceptually is a lot like color film:

I see what you mean. I imagine it's at the back of their minds and could emerge when the time's ripe and something pops out of research to give it an edge. But it's waited for at least twenty years already.
Bayer is a real fudge imo.

 

[Baluncore]

I did some searching around but couldn't see an answer to my specific question.

The last opportunity to reproduce a realistic colour, comes with the camera filters.

The sensitivity of the eye to different wavelengths is through chemical bleaching, a bit like the response of chemical dyes or filters. It is easier to come up with a chemical dye mix, than it is to deposit many accurate layers for a dichroic filter.

The response of an expensive dichroic filter, is probably too sharp to match the colour transitions of our normal vision.

Maybe in the early days of colour TV, light was more valuable, so colour cameras needed more sensitivity, hence less lossy dichroic filters. Having three separate cameras gave an opportunity to use three dichroic filters, but now, when one image sensor can do it all with dye filters. Printing tens-of-millions of accurately positioned, multilayer dichroic filters, would make the image sensor very expensive.

 

[sophiecentaur]

Printing tens-of-millions of accurately positioned, multilayer dichroic filters, would make the image sensor very expensive.

In the early days that was true but "tens of millions of accurately positioned" is frequently achieved in many electronic devices. However, the depth of the filter layer would be greater than the diameter of each filter element. That would make it difficult.

The response of an expensive dichroic filter, is probably too sharp

You can get any response you want with a time domain filter but it may not be practical for an array of small filters.

Maybe in the early days of colour TV, light was more valuable,

That was certainly true; TV studios needed loads of light and massive lens apertures. SNR is probably significantly less of a problem these days.

But never say never; the Bayer is a bit unsatisfying as an idea (to me) and it would be nice for a new technology to have a turn.