Sony 'boosts low-light picture quality' with white pixels

[Isaac]

Sony is promising improved low-light photography from its forthcoming range of smartphone camera sensors, thanks to the addition of white-light pixels.

[Fine_Art]

That's excellent, as it's perfected they will move it up into the higher quality cameras.

Ive always wondered why manufacturers don't split the light in 2 with part going to a sensor with no color filters and part going to a filtered sensor for color information. I'd assume it would get them a lot more light for the cost of an additional sensor. Its cheaper than a dichroic prism with 3 sensors.

[BJL]

Sony is promising improved low-light photography from its forthcoming range of smartphone camera sensors, thanks to the addition of white-light pixels.

Kodak announced the same idea a few years ago, so I hope that Kodak gets some patent royalties; it needs the money. But maybe Kodak sold the patents as it sold of its sensor operations, first CMOS, then CCD.

[MarkL]

Good to see at least one camera manufacturer is doing something innovative rather than sitting on their heels.

[famalam]

interesting idea

[hjulenissen]

Good to see at least one camera manufacturer is doing something innovative rather than sitting on their heels.

Guess we will have to see if this is innovation or just a marketing-spin.

r+g+b+w seems like a simple thing to do technologically. If it has all benefits and no drawbacks, why did not dr Bayer think of it in the 70s?

-h

[BJL]

r+g+b+w seems like a simple thing to do technologically. If it has all benefits and no drawbacks, why did not dr Bayer think of it in the 70s?

That is a question that is always worth asking, and indeed I have seen this idea proposed in forum discussion some years ago, not to mention Kodak announcing a CMOS sensor with this feature a few years ago, before abandoning its CMOS sensor efforts.

One issue to be dealt with is that with "normal" exposure for the RGB photosites, the white ones will blow-out about a stop or so earlier. Maybe some fancy processing will detect this, like replacing blown-out white signals with interpolated data from nearby RGB photo sites. If instead exposure levels are reduced by a stop or more to hold the white highlights, shadow areas will have more color noise, so I doubt that is what Sony is doing.

[Fine_Art]

That is a question that is always worth asking, and indeed I have seen this idea proposed in forum discussion some years ago, not to mention Kodak announcing a CMOS sensor with this feature a few years ago, before abandoning its CMOS sensor efforts.

One issue to be dealt with is that with "normal" exposure for the RGB photosites, the white ones will blow-out about a stop or so earlier. Maybe some fancy processing will detect this, like replacing blown-out white signals with interpolated data from nearby RGB photo sites. If instead exposure levels are reduced by a stop or more to hold the white highlights, shadow areas will have more color noise, so I doubt that is what Sony is doing.

Which is why a full sensor of white is a better solution. It gives you the extra dynamic range of the color sensor for color + highlights with the full white sensor for shadows. You don't need new algorithms to decide how to blend the white pixel.

[hjulenissen]

Which is why a full sensor of white is a better solution. It gives you the extra dynamic range of the color sensor for color + highlights with the full white sensor for shadows. You don't need new algorithms to decide how to blend the white pixel.

I have speculated on something similar.

Those two sensors would add to the size and cost, and aligning them mechanically, getting similar focus, etc might be an issue ("3CCD" video cameras have far looser spatial resolution requirements).

It seems to me that photographers are generally more interested in a "35mm" Bayer-sensor, than 2x cropped luminance/color sensors, if those have similar size/weight/cost/...

-h

[Fine_Art]

Full frame vs cropped has nothing to do with 1 sensor vs 2. You can do it with either. People are far more likely to understand why a dslr is expensive if it has 2 sensors rather than other features.

Lets say they split 70/30 for the white. For a shot I get 70% +1/3 of 30% = 80% of the light in a scene.
A regular Bayer is 33% of the light by filtering 2/3 at each point.

Suddenly I have far more ISO range with higher DR.

[BJL]

The idea of two (or three) sensors plus beam splitters falls into the category of "obvious and suggested many times over the years, so there is probably a good reason why no camera maker is doing it". Some possible reasons that I have read in previous discussions:
- For less cost and bulk, you can have a single sensor of twice or more the area (though lenses get longer), and this is better overall
- the beam splitter approach increases the optical path from lens to sensor, putting greater constraints on lens design, especially for wide angles.
- there are problems with accuracy of alignment, which has to be accurate to significantly less than the photosite size. This is far harder for stills than with the far lower resolution of three sensor video cameras. Note also that the new generation of higher quality video cameras with large sensors has abandoned the three sensor approach.

[Fine_Art]

Panasonic's 3CCD video cameras are very good. If you have seen their output, it is much better than other cameras at the same price point.

[BJL]

Panasonic's 3CCD video cameras are very good. If you have seen their output, it is much better than other cameras at the same price point.

Those video cameras have three very small sensors in a body far, far larger than any still camera using that sensor size; in fact, bigger, heavier and far more expensive than most DSLRs, even with just 1/3" sensors. Multiple sensors in a still camera would make it as big as a camera with a single far larger sensor. So there is no point comparing to a camera with one sensor of the same size; that is not the alternative that it would complete against on price or size.

I end with the same question as before, and ask "h" raised earlier: the idea is well known, so of it would work well for stil camera, why is no camera maker doing it, including ones like Panasonic and Sony who do it in some small video cameras? Once we eliminate the possibility of horrible industry-wide stupidity or conspiracy, we are left with "it cannot actually compete against the currently used approaches, even if it is not immediately obvious why."

[Fine_Art]

Those video cameras have three very small sensors in a body far, far larger than any still camera using that sensor size; in fact, bigger, heavier and far more expensive than most DSLRs, even with just 1/3" sensors. Multiple sensors in a still camera would make it as big as a camera with a single far larger sensor. So there is no point comparing to a camera with one sensor of the same size; that is not the alternative that it would complete against on price or size.

I end with the same question as before, and ask "h" raised earlier: the idea is well known, so of it would work well for stil camera, why is no camera maker doing it, including ones like Panasonic and Sony who do it in some small video cameras? Once we eliminate the possibility of horrible industry-wide stupidity or conspiracy, we are left with "it cannot actually compete against the currently used approaches, even if it is not immediately obvious why."

That provides no more information. It is actually well known in business circles that companies usually find it hard to switch to a new path once they have started down another. They have massive investments in the bayer method.

Let someone who works in the industry fill in the blanks. If "it's not being done, therefore it cant work" had been the conclusion through history we would still be grunting around the fire in a cave.

[hjulenissen]

Full frame vs cropped has nothing to do with 1 sensor vs 2. You can do it with either. People are far more likely to understand why a dslr is expensive if it has 2 sensors rather than other features.

You did not seem to understand what I was talking about.

2 sensors will (everything else equal) increase the price, size and weight.

One larger sensor will (everything else equal) increase the price, size and weight.

For the manufacturer, the benefits of having 2 or more sensors could be weighed up against having one larger sensor, compared to an existing produce.

-h

[Fine_Art]

You did not seem to understand what I was talking about.

2 sensors will (everything else equal) increase the price, size and weight.

One larger sensor will (everything else equal) increase the price, size and weight.

For the manufacturer, the benefits of having 2 or more sensors could be weighed up against having one larger sensor, compared to an existing produce.

-h

Not really because by increasing the sensor size you are forcing everyone to buy a full set of new lenses. Most people won't do it. Therefore the savings in manufacturing cost yields very little new revenue.

For example how many people are buying an S2 vs buying a new mirrorless with adapter for their Leica lenses?

[BJL]

Not really because by increasing the sensor size you are forcing everyone to buy a full set of new lenses.

What scenario are you talking about now? We started with a small sensor fixed lens compact camera, where the issue of a new set of lenses is irrelevant. There, the point is that for the same bulk and cost, a camera with one larger sensor and all the light going to the color sensor would probably do a better job than delivering only about 70% of the light to that sensor, getting worse color noise in low light situations.

But of you are talking about interchangeable lens systems
1) to do this with an SLR optical viewfinder requires at least two mirrors or beam splitters, making for a bulkier camera and as fad as I can tell, increasing the optical path length from lens to sensor, so requiring new lenses.
2) if done without an OVF, the now standard mirrorless single-sensor designs can be far more compact than if a mirror or beam splitter is needed. The Sony SLT cameras are in effect an example of the bulk added by a mirror, compare to the Sony NEX cameras.

P.S. on the argument that the multiple sensor and beam splitting approach has not been done out of sheer conservatism, staying with the current technology rather investing in a new better approach, I am skeptical for several reasons:
A. Panasonic, Sony, and Canon already have this technology, for video cameras.
B. Sony has used a beam splitter for different purposes in its SLT cameras, but using the space between lens and focal plane already there for SLRs.
C. Companies like Panasonic and Sony in particular have been quite vigorously investing in and developing new approaches in their attempt to make inroads against the established SLR makers, and Olympus, Pentax, Samsung, and Nikon have also invested substantially in the new technologies of their new mirrorless systems, including new lens systems, so the cynical view that "no camera maker is willing to spend the money on a new, superior approach" does not fit the facts very well.

[Fine_Art]

Time will tell where they take their white pixels.

[hjulenissen]

Not really because by increasing the sensor size you are forcing everyone to buy a full set of new lenses.

Fine, then turn it around. I have a Canon 7D featuring a 1.6x cropped sensor. Say that for the next-generation, Canon have the choice between:
1. Throwing in one (hopefully improved) 1.6x crop sensor
2. Throwing in two e.g. 2.2x crop sensors, using your suggested light splitting

Lets assume that both alternatives result in somewhat similar bulk, weight, price, etc. Lets also assume that I can still use my EF/EF-S lenses.

Like I said, the cost vs benefit of doing two sensors, would have to be weighed up against the cost vs benefit of doing one larger sensor (or a better dsp, or a million other trade-offs). I am sure that Canon people are thinking quite hard about such problems.

For example how many people are buying an S2 vs buying a new mirrorless with adapter for their Leica lenses?

How many are considering a Pentax MFDB instead of Canon/Nikon 35mm? Quite a few, I think. Not enough to make a dent in Canon/nikon sales, but then they have products that are allready very streamlined for the majority of customers.


-h

[Fine_Art]

Fine, then turn it around. I have a Canon 7D featuring a 1.6x cropped sensor. Say that for the next-generation, Canon have the choice between:
1. Throwing in one (hopefully improved) 1.6x crop sensor
2. Throwing in two e.g. 2.2x crop sensors, using your suggested light splitting

Lets assume that both alternatives result in somewhat similar bulk, weight, price, etc. Lets also assume that I can still use my EF/EF-S lenses.

Like I said, the cost vs benefit of doing two sensors, would have to be weighed up against the cost vs benefit of doing one larger sensor (or a better dsp, or a million other trade-offs). I am sure that Canon people are thinking quite hard about such problems.
How many are considering a Pentax MFDB instead of Canon/Nikon 35mm? Quite a few, I think. Not enough to make a dent in Canon/nikon sales, but then they have products that are allready very streamlined for the majority of customers.


-h

Sure, sounds reasonable.