sensors

[david distefano]

i know there are many tech knowledgeable people on this site and i have a question on sensor design. there is rumors floating around of sony developing a 54mp camera with the foveon styled sensor. some say it will be a ff sensor and some rumors have it as a mf sensor. in my readings it would seem that a foveon sensor records all the info which should produce a more accurate image, color wise, than the bayer pattern sensor. is the foveon sensor more expensive to produces? is theory of the sensor and actual output of the sensor not similar? are mf manufacturers looking to produce a foveon sensor back in their lineup?

[Bart_van_der_Wolf]

i know there are many tech knowledgeable people on this site and i have a question on sensor design. there is rumors floating around of sony developing a 54mp camera with the foveon styled sensor. some say it will be a ff sensor and some rumors have it as a mf sensor. in my readings it would seem that a foveon sensor records all the info which should produce a more accurate image, color wise, than the bayer pattern sensor. is the foveon sensor more expensive to produces? is theory of the sensor and actual output of the sensor not similar? are mf manufacturers looking to produce a foveon sensor back in their lineup?

Hi David,

The Foveon design, which translates travel distance through silicon as a kind of color separator, depends on all rays more or less entering at a similar angle of incidence. Otherwise the more oblique corner rays would need to be calibrated for the longer travel distance, compared to the center rays. That would also mean a higher light fall-off (underexposure) towards the corners and more color shift due to the significant amount of per photosite infrastructure required to read the signals.

IOW, it seems very unlikely that they will take that direction for a larger sensor array size.

Cheers,
Bart

[ErikKaffehr]

Hi,

My impression is that the Foveon design is built in three layers, making it much more complex than a Bayer design. It also depends on photons of different energy penetrating to different depths, which means that colour is depending very heavily on math.

There are some patents on the colour pixels depending on different principles, not all similar to Foveon.

Sony is said to have samples of a 54 MP non Bayer sensor, but it is said to be very expensive to make. That sensor would be full frame 135, AFAIK.

Best regards
Erik

i know there are many tech knowledgeable people on this site and i have a question on sensor design. there is rumors floating around of sony developing a 54mp camera with the foveon styled sensor. some say it will be a ff sensor and some rumors have it as a mf sensor. in my readings it would seem that a foveon sensor records all the info which should produce a more accurate image, color wise, than the bayer pattern sensor. is the foveon sensor more expensive to produces? is theory of the sensor and actual output of the sensor not similar? are mf manufacturers looking to produce a foveon sensor back in their lineup?

[uaiomex]

Considering that the Merrils can compete in IQ with FF sensors with double the pixel count, Sony could produce an A7f (Foveon type) with 36+ mp to challenge a P1 IQ260 and perhaps a 280 too, it won't matter if it cost 8 or 10 grand, it would still be the bargain of the century.                                          Eduardo.                                                                      

Hi,

My impression is that the Foveon design is built in three layers, making it much more complex than a Bayer design. It also depends on photons of different energy penetrating to different depths, which means that colour is depending very heavily on math.

There are some patents on the colour pixels depending on different principles, not all similar to Foveon.

Sony is said to have samples of a 54 MP non Bayer sensor, but it is said to be very expensive to make. That sensor would be full frame 135, AFAIK.

Best regards
Erik

[BernardLanguillier]

For now, the new Sigma DP2/DP3 Quattro (45/75mm equivalent lens) at base ISO will probably be close to the P45+ and best technical lenses in a single shot.

Add a good light pano head to the set up and you're good for those extremely impressive 60 inch landscape prints for ~2,000+ US$ and less than 1 kg worth of equipment.

Cheers,
Bernard

[marcmccalmont]

The Foveon sensor is credited as being more "film" like, do to the 3 color layers, I'd propose a Bayer or pseudo random conventional sensor is more "eye" like with individual cones? Are we trying to record what film sees or what the eye sees? Just a thought.
Marc

[torger]

The Foveon sensor is credited as being more "film" like, do to the 3 color layers, I'd propose a Bayer or pseudo random conventional sensor is more "eye" like with individual cones? Are we trying to record what film sees or what the eye sees? Just a thought.

The eye registers one thing which then is processed by the brain to produce a result. The brain affects the "raw" material quite much, white balance the most obvious. When you look at a print there will be a different viewing condition, considerably less contrast in the print than in reality, different perspective, 2D rather than 3D etc and the brain will interpret things differently producing a different result.

This makes it largely meaningless to record the same way the eye does, even if it did it would not necessarily become a truer representation of what the eye-brain saw at the scene. An additional problem is that the eye's "filters" are heavily overlapping so if sensors used the same overlapping there would be issues with noise. It's better to record in a way that suits the current technology and convert to color and see what works.

Personally I think the Foveon sensor is over-hyped, due to too much focus on pixel-peeping. The old well-established bayer array still produces the best all-around result, especially when equipped with an AA filter (so you don't get much pixel-peep artifacts). I rather use a higher resolution bayer filter sensor with more dynamic range than a lower resolution foveon sensor (with higher color resolution) with less dynamic range.

With a bayer array and normal scenes you lose very little luminance resolution, and the lower color resolution matches well the eye/brain property.

Foveon type of sensors will also be difficult to make with photo diodes close to the surface, making them unsuitable for technical wides, or any other small lightweight wide angle lens design. Already today there exists more than one technology to make bayer array pixels with very high fill factor and photo diode much closer to the surface than today. This has not yet found it's way to larger sensors, but I think it will get there. All Foveon designs I've seen seems to have very deep pixels.

[Ajoy Roy]

Another type of sensor being researched is the vertical accumulation well where the charge accumulation can be accommodated in the whole material rather than on the surface only. What this means is that the sensel surface area need not be large to have a large S/N ratio. Further is the sensel surface are is small, you can group three or four sensels each having R, G, B filter (just like the CRT or the LCD display). In one stroke you eliminate artifacts due to demosaicing and increase the S/N. Much like shooting with multi shot only it is a single shot.

I hope this comes out from the R&D fast, so that we can get clean high ISO high resolution sensors.

[BJL]

Another type of sensor being researched is the vertical accumulation well where the charge accumulation can be accommodated in the whole material rather than on the surface only. What this means is that the sensel surface area need not be large to have a large S/N ratio. ...
I hope this comes out from the R&D fast, so that we can get clean high ISO high resolution sensors.

The ability to count more photons in photosites of a given size by making them "deeper" only increases DR and maxim SNR by increasing highlight headroom -- that is, reducing the base ISO speed. It does nothing for SNR at a given level of exposure (a given exposure index, so-called "ISO"). So this would not help with "clean high ISO high resolution sensors"; the only ways to improve low light performance are increasing quantum efficiency and reducing the "dark" noise introduced by the sensor and in subsequent processing.

[eronald]

One look at the pictures, and one look at pricing ($500 for Merrills now) and you know Sigma are doing something right.

Edmund

[Telecaster]

The Foveon sensor is credited as being more "film" like, due to the 3 color layers, I'd propose a Bayer or pseudo random conventional sensor is more "eye" like with individual cones? Are we trying to record what film sees or what the eye sees? Just a thought.

I don't think it's an either/or scenario. Sometimes I want a photo to look like "what I saw." Other times not. Sometimes I like using cameras that produce highly malleable files. I've been impressed with Sony's A7r in this regard. Sometimes I prefer a camera with a particular look. My new (old) Leica M8, with the RAW files run through PhotoRAW on my iPad, makes photos with a tonality and acuity not unlike Kodachrome 64. As K64 was a film I used & liked a lot I'm a happy camper with this combo. Sigma's Foveon cameras produce a distinctive look, no doubt determined to a large extent by the technology. But that's how it's always been. I like what I've seen from the Merrills but don't own one.

To note: attempting to reproduce visual reality, at least as our eye/brain system perceives & interprets it, is a choice rather than a mandate.

-Dave-