Why Foveon X3 sensor is not popular used

[bingo]

If we have 3 channel of RGB data in a same location pixel, the quality of the images should be more accurate in my understanding. Then why Foveon X3 sensor and SIGMA cameras using Foveon X3 sensors are still not popular?
Is there any problem with the design of the Foveon X3 sensor?

[pnodrog]

There is one devotee of Foveon sensors on this site (no doubt you will hear from him)...the rest of us are far too conservative...

[xpatUSA]

If we have 3 channel of RGB data in a same location pixel, the quality of the images should be more accurate in my understanding. Then why Foveon X3 sensor and SIGMA cameras using Foveon X3 sensors are still not popular?
Is there any problem with the design of the Foveon X3 sensor?

Hello and welcome @bingo, do you have a name?

Here I am, as predicted in Paul's previous post!

I've been using Sigma cameras for six or seven years and have thoroughly researched the technology.

The data is not really RGB, not in the sense that it is generally understood here. The three layers of pixels are often called "RGB" but their response is very broad, so much so that the latest X3 cameras use the "blue" (top) layer as a faux luminance value. The broadness in all Foveon-based cameras requires quite severe raw-to-RGB conversion algorithms and the end result of that is noise.

The quality is more "accurate" simply because the Foveon conversion to RGB does not involve interpolation, well - actually fancy guesswork called "de-mosiacing".

Foveon/Sigma products are unpopular because "quality" comes at a cost:

The X3F files are very large compared to the MP size of the output; and, since many wrongly relate MP to resolution, this is held against the technology.

The proprietary converter gets black marks for being slow (like RawTherapee, every adjustment is done from scratch including all previous) and somehow many people expect more than just conversion. Unlike ACR, it can't even crop or fix tilt and has only recently sprouted Levels!

The sensor requires good exposure ... under- or over-exposure (at the sensor) gets you poor results. The first DSLR only allowed 100, 200, 400 ISO and 400 was a waste of time. Sigma has upped that with all sorts of tricks to 6400 extended but it remains a waste of time except for monochrome shots in dark pubs.

People such as myself always shoot at 100 ISO, kind of like shooting a film camera. There is no "exposure triangle" with the Foveon ... more like a line. .

Sigma cameras do not have an AA filter over the sensor because there is no color aliasing. However there is still luminance aliasing and they have a pretty good response, maybe 30% MTF, at the spatial frequency where aliasing starts (64 lp/mm on mine). Meaning that any regularly-occurring scene detail over that frequency will get quite visible aliasing, if well-focused with a sharp lens.

Summing up, the Foveon takes some work to shoot properly and the reward is worth it for people like me with good Texas light and plenty of time:

.

Taken with a 4.7 MP camera.

[scottelly]

If we have 3 channel of RGB data in a same location pixel, the quality of the images should be more accurate in my understanding. Then why Foveon X3 sensor and SIGMA cameras using Foveon X3 sensors are still not popular?
Is there any problem with the design of the Foveon X3 sensor?

Most people don't know that Sigma makes cameras. They are a small company (but growing), so they don't have the marketing budget of a company like Fuji or Sony. Sigma has been making good value cameras since the film days, but they march to the beat of a different drummer, and it takes a special kind of photographer to even try a Sigma camera. I suggest you try one. There are plenty of used Sigmas available. Just be warned though, they operate slowly.

As far as "problems" with the design of the Foveon sensors, it depends on who you talk to. Is it a problem that they have an advantage in per-pixel acuity? Is it a problem that they are noisy, so produce very noisy high-ISO photos? If these things are not problems for you, then maybe you're a candidate for a camera with a Foveon sensor. Sigma will be releasing a new, full-frame Foveoen sensor on the World next year. Hopefully that camera will help build their camera business and the Foveon sensor reputation to its highest level yet.

I shoot with a Sigma SD Quattro H and a Nikon D810. Both have their advantages.

[Manfred M]

Paul is correct. One of the members here, Ted, is a Sigma camera user and knows the Foveon sensor quite well.

The high level answer to your question is that while one aspect of the Foveon design has advantages, but also significant downsides. The upside is what you have mentioned; the sensor is designed in such a way that it works much in the same way that colour film did. Each area of the sensor can be assigned deliver a specific colour R, G and B for each location without having to go through the hoops of determining a specific pixel value through a process known as demosaicing required of a Bayer sensor. That is unfortunately, where the key advantages of the Foveon design stop.

There are three layers of sensitivity that relate to being red, green and blue that are stacked on top of each other and the light has to travel through each layer to reach the one below it through a colour filter and a substrate that contains the detectors which include the elements on the semiconductor that capture and transmit the data to the camera. These are not transparent and lead to a level of light loss.

To compensate for this drawback, the newest Foveon sensor pixel is larger than is found the high end sensors that Sony, Nikon, etc. use in their most recent cameras. The pixel pitch for the Foveon X3 Fx17-78-F13 sensor is 7.8 μm whereas the Nikon D850 sensor has a pixel pitch of 4.3 µm. So while the Foveon sensor has a advantage in terms of resolution, it is not as large as it might seem.

The standard Bayer type sensor has all this hardware on the same plane, so this design of sensor is inherently more sensitive to light for all colours. This means that these cameras will be more sensitive to light and inherently less noisy. Sensors have a single sensitivity and higher ISO levels are created by amplifying the signal. A noisy signal can't be amplified as much as a signal that is cleaner, so this limits the effective maximum ISO that can be used. I believe Sigma tops out at ISO 8000 while my Nikon D810 goes to a usable ISO 128,000; other cameras have even more ludicrous ISO ratings.

Finally, Sigma owns Foveon (they bought it in 2008), so this makes it somewhat less likely that other camera manufacturers that have committed to Bayer style sensors to jump on the Foveon bandwagon. The sensor needs to offer some clear advantages over the current technology that these companies use to jump on that bandwagon. It seems that in the other manufacturer's view, those advantages are not there.

[xpatUSA]

Paul is correct. One of the members here, Ted, is a Sigma camera user and knows the Foveon sensor quite well.

<>
There are three layers of sensitivity that relate to being red, green and blue that are stacked on top of each other and the light has to travel through each layer to reach the one below it through a colour filter and a substrate that contains the detectors which include the elements on the semiconductor that capture and transmit the data to the camera. <>

Just to be clear, there is no "color filter" per se - unlike the color filter array that sits on top of non-Foveon sensors. Each of the three Foveon layers responds to the whole visible light spectrum (say 400-700nm), albeit non-linearly. The "RGB" designations refer to the point of maximum QE for each layer. Non-intuitively, the lower layer has the highest QE of the three layers, in spite of what many think.

Please pardon the clarification.

I think that @bingo would be better served if we all refrain from giving our own versions of "how the Foveon works" and instead address his/her questions as simply as possible. Should shorten this thread quite a bit because, personally, I could go on about loose terminology and minor technical points for ever ad nauseam.

[Manfred M]

Just to be clear, there is no "color filter" per se - unlike the color filter array that sits on top of non-Foveon sensors. Each of the three Foveon layers responds to the whole visible light spectrum (say 400-700nm), albeit non-linearly. The "RGB" designations refer to the point of maximum QE for each layer. Non-intuitively, the lower layer has the highest QE of the three layers, in spite of what many think.

Please pardon the clarification.

I think that @bingo would be better served if we all refrain from giving our own versions of "how the Foveon works" and instead address his/her questions as simply as possible. Should shorten this thread quite a bit because, personally, I could go on about loose terminology and minor technical points for ever ad nauseam.

Thanks for the explanation Ted - you have a much more detailed knowledge about this technology than most.

The effect that you where specific wavelengths are filtered out based on the depth that light penetrates a material is a phenomena that is well known by anyone that scuba dives. The reds and yellows are filtered out first in relatively shallow water and the blues and violet wavelengths penetrate the deepest. Coral reefs look boring and colourless unless one brings along a dive light to provide the wavelengths that are missing at depth. Silicon does the same thing but the optical properties allow it do so so much more quickly.

[xpatUSA]

Thanks for the explanation Ted - you have a much more detailed knowledge about this technology than most.

The effect that you where specific wavelengths are filtered out based on the depth that light penetrates a material is a phenomena that is well known by anyone that scuba dives. The reds and yellows are filtered out first in relatively shallow water and the blues and violet wavelengths penetrate the deepest. Coral reefs look boring and colourless unless one brings along a dive light to provide the wavelengths that are missing at depth. Silicon does the same thing but the optical properties allow it do so so much more quickly.

Thanks Manfred. So seawater works backasswards to silicon, it seems, silicon being opaque to UV at about 250nm and transparent to IR>1150nm. I have no idea why that should be ...

[Manfred M]

Thanks Manfred. So seawater works backasswards to silicon, it seems, silicon being opaque to UV at about 250nm and transparent to IR>1150nm. I have no idea why that should be ...

It's not just sea water, but fresh water as well. This seems to be a property of water and a bit of salinity doesn't change much. The main issue is that the critters that grow in fresh water tend to be a lot less colourful. However, modern steel shipwrecks tend to look gray at depth and a dive light will show off the rust colour.

[odds]

...Each of the three Foveon layers responds to the whole visible light spectrum (say 400-700nm), albeit non-linearly.

Some 15 years ago Nikon filed a patent application on a sensor where light was mirrored sideways through a chain of tree mirrors that reflected a portion of the light to individual photo sites ("rgb"). Perhaps a variant of the Foveon? I don't know if the Nikon invention was ever used in a product, see https://www.dpreview.com/articles/65...konimagesensor

--
Odd S.

[xpatUSA]

Some 15 years ago Nikon filed a patent application on a sensor where light was mirrored sideways through a chain of tree mirrors that reflected a portion of the light to individual photo sites ("rgb"). Perhaps a variant of the Foveon? I don't know if the Nikon invention was ever used in a product, see https://www.dpreview.com/articles/65...konimagesensor

--
Odd S.

Hello Odd,

Thanks for the Nikon info ... but it is not a variation of the Foveon CMOS sensor because it does not rely on the absorption characteristic of silicon to "separate" R,G and B. Very interesting, though.

Foveon (the Company) had a similar camera prior to the invention of their CMOS layer technology:

"Foveon’s first product was not an image sensor but instead a complete digital camera.In this original system, a beam-splitter prism assembly separated the incoming light into its three primary colors, passing the red, green, and blue beams through separate color filters and directing them to three large image sensors. An extremely high-resolution image was then assembled from the data of the three colors.

This camera was extraordinary, high-end technology, but it was expensive to manufacture and ultimately too costly for the end user. By the time Foveon stopped prism digital camera production, it had already created a patented technology that would ensure its preeminence in the coming era of image processing research."

Go to this link and scroll down a little.
.

[Mike Earussi]

The Foveon sensor has a strong following among a few photographers because of its unique look. Because each pixel is completely separate from its neighbors it does not have to interpolate color information and so can detect minute gradations in color a Bayer sensor simply cannot see because it has to "take a vote" from its surrounding pixels in order to determine its color. This averaging process tends to smear colors wiping put any subtle differences. As such a Bayer image looks more flat and dead and 2 dimensional instead of 3 dimensional and alive. A good comparison would be the look of slide film to color negative. The slide film because it lacks the red mask has much better/purer color saturation and more overall pop. Another advantage is Foveon's ability to detect more colors because its not limited to the fixed color masks that a Bayer chip has to use.


But of course this extra quality comes a price, that of processing speed (my SD1M takes 7 seconds to process one file), DR, about 2-3 stops lower than the best (Sony) Bayer, and iso, where it's only really good at base iso 100 and even iso 400 shows noticeable degradation in image quality. Another drawback is that Foveon raws are so complicated to process that ACR can't open them. As such only those photographers who value image quality over usability bother with Foveon.

[Manfred M]

T A good comparison would be the look of slide film to color negative. The slide film because it lacks the red mask has much better/purer color saturation and more overall pop.

Actually that is incorrect. Slides are transmitted light technology and colour prints using traditional photo papers use reflected light. It's much the same as viewing an image on a high quality computer screen versus looking at an inkjet print. That is why there is a "pop" associated with slide film.

As someone who printed both slides and negatives in the wet colour darkroom, this is definitely the case. In fact the negative prints were much more subtle and cleaner than printed slides using Ektachrome paper. Slide film had a lower dynamic range than print film, so that explains part of the difference. The other part comes from the two filter layers that behave much the way that the absorption of specific wavelengths of light. The film actually has both a yellow and a red filter, so the film has an orange colour. The yellow filter blocked blue light and the red filter blocked the green wavelengths. The blue sensitive layer was closest to the lens and the red sensitive layer closest to the back plane of the camera with the green sensitive layer sandwiched in the middle. End effect was cleaner colours in with print film.

The one exception was the Cibachrome (later Ilfochrome) process where the printed transparencies had an almost magical look, if one could tame the high contrast associated with that process. Ilfochrome paper used a dye-destructive process. All the dyes used in creating the image were coated onto the medium and unneeded dyes were bleached out. I have 40 year-old Cibachrome prints that show no signs of deterioration. Regular colour print and reversal papers used a photochemical reaction to activate the dyes in these papers. Current inkjet technology definitely gives better results than the old colour silver halide papers.