Foveon sensor & IR sensitivity

[anselwannab]

I was just reading about the Foveon Sensor . Since the red sensing part is on the bottom, would these cameras have inherently less IR sensitivity?

Less cover glass, less issues with oblique light rays.

Just a question that I couldn't find the answer to.

It seems there are rumors that the 1DsIII might have a sensor like this.

Mark

 

[cornellison]

That is good question. The Foveon sensor seems to be the most logical design to me, a very non-engineer, non-physicist. I wish you could get that sensor in something besides a Sigma.

 

[rxmd]

Less cover glass, less issues with oblique light rays.

The latter is wrong.

The Foveon sensor detects colour by detecting how deep a photon travels. Take a photon that hits the sensor vertically; let's assume it travels to depth d. A photon of the same colour that travels at an oblique angle alpha against the vertical axis will travel a distance of d within the sensor, but because of the oblique angle the depth it actually reaches is only d*cos(alpha), which can be considerably less. So in fact the Foveon sensor has a problem with wideangle lenses. Assuming that red is at the bottom, it also would mean that infrared gets read out as red with increasing distance from the center of the frame, depending on the lens.

You can work around this in software if you know the optical parameters of the lens (focal length, placement of back element, current aperture etc.), so that you know which areas of the sensor are likely to be hit at what angle. Note that since the back element moves, the precise parameters also depend on the focusing distance. This is why the only digital compact with a Foveon sensor on the market, the Sigma DP1, has a fixed focal length lens.

Or you can try to work around it by arranging the sensor layers differently, not in a flat shape, but instead in little pits (like this image suggests). The difficulty with pits is that you don't know where the photon hits the surface, and if it doesn't hit the center of the pit its colour will be misdetected. So you have to mask out much of the surface, making the sensor less sensitive overall. At least that's how I understand it.

So the only way a Foveon sensor could possibly make sense in a rangefinder would be with a lens database in the camera and with coded lenses, and even then it would be very difficult. For instance, the camera would have to read out the distance setting electronically (probably the aperture as well). Maybe this was one of the reasons why Leica introduced lens coding, because they want to retain the option of using Foveon sensors in the future (which I don't think they will do, given the still-existing problems and the comparatively low sensitivity of present-day Foveon sensors).

Philipp

 

[sf]

That is good question. The Foveon sensor seems to be the most logical design to me, a very non-engineer, non-physicist. I wish you could get that sensor in something besides a Sigma.

I know it. SIgma is the last company I expect to produce something of quality. Well...maybe not the last.

I have to say, though, that the Foveon just produces gorgeous images. Just darn good. And the newest incarnation of it is going to be something to have.

I want to DP1 to live up to the hype we have given it!

 

[sf]

.
The latter is wrong.

The Foveon sensor detects colour by detecting how deep a photon travels. Take a photon that hits the sensor vertically; let's assume it travels to depth d. A photon of the same colour that travels at an oblique angle alpha against the vertical axis will travel a distance of 1 within the sensor, but because of the oblique angle the depth it actually reaches is only d*cos(alpha), which can be considerably less. So in fact the Foveon sensor has a problem with wideangle lenses. Assuming that red is at the bottom, it also would mean that infrared gets read out as red with increasing distance from the center of the frame, depending on the lens.

Philipp

that is probably why they chose the fixed lens 28mm for the DP1. they also say that they chose a design that would ahve light "hitting the sensor at a perpendicular angle" which means to me that they intend for the light to be coming in at the steepest possible angle - straight.

 

[MichaelB]

What about those microlenses for each pixel? Maybe they should ask Kodak how to do it. 😉

 

[anselwannab]

The latter is wrong.

The Foveon sensor detects colour by detecting how deep a photon travels. Take a photon that hits the sensor vertically; let's assume it travels to depth d. A photon of the same colour that travels at an oblique angle alpha against the vertical axis will travel a distance of d within the sensor, but because of the oblique angle the depth it actually reaches is only d*cos(alpha), which can be considerably less. So in fact the Foveon sensor has a problem with wideangle lenses. Assuming that red is at the bottom, it also would mean that infrared gets read out as red with increasing distance from the center of the frame, depending on the lens.

You can work around this in software if you know the optical parameters of the lens (focal length, placement of back element, current aperture etc.), so that you know which areas of the sensor are likely to be hit at what angle. Note that since the back element moves, the precise parameters also depend on the focusing distance. This is why the only digital compact with a Foveon sensor on the market, the Sigma DP1, has a fixed focal length lens.

Or you can try to work around it by arranging the sensor layers differently, not in a flat shape, but instead in little pits (like this image suggests). The difficulty with pits is that you don't know where the photon hits the surface, and if it doesn't hit the center of the pit its colour will be misdetected. So you have to mask out much of the surface, making the sensor less sensitive overall. At least that's how I understand it.

So the only way a Foveon sensor could possibly make sense in a rangefinder would be with a lens database in the camera and with coded lenses, and even then it would be very difficult. For instance, the camera would have to read out the distance setting electronically (probably the aperture as well). Maybe this was one of the reasons why Leica introduced lens coding, because they want to retain the option of using Foveon sensors in the future (which I don't think they will do, given the still-existing problems and the comparatively low sensitivity of present-day Foveon sensors).

Philipp

Sounds like all the technologies have plusses and minuses. I would think that with the focusing cam position and lens coding you could get a better estimate of how the light rays are coming in.

 

[HAnkg]

If a bayer sensor is problematic for RF because of the proximity of lens and sensor the Foveon would be a much bigger technical challenge. Foveon really requires a lens/camera system designed for digital, ruling it out for an M system.

 

[Sailor Ted]

I would think that with the focusing cam position and lens coding you could get a better estimate of how the light rays are coming in.

Yes but if you think the uproar around Leica's use of lens coding on the M8 is anything can you imagine the howls if the camera hardly worked at all without lens coding on non-Leica, non-coded lenses?

 

[HAnkg]

Well they might get around the RF lens/sensor distance problem with a 2x lens factor. That should really get'im howling.

 

[IGMeanwell]

The Foveon sensor is definitely not free from all IR issues

some say the screen installed in the new SD14 is actually an IR filter (sigma says its a dust screen), because of its tint

Actually they mentioned this on Luminous Landscapes photokina feature....

 

[Gabriel M.A.]

Yes but if you think the uproar around Leica's use of lens coding on the M8 is anything can you imagine the howls if the camera hardly worked at all without lens coding on non-Leica, non-coded lenses?

Yes. We'd see a flood of vignetted images claiming the sensor is a POS. Of course, they would be the same who would not understand the tool that they're shooting with, and expect the same results from a P&S camera, which does not use the same technology, so you'd think they'd know.

"Flaws" will be found and exacerbated to gargantuan proportions until a legend is created.

 

[Sailor Ted]

Gabriel you eluded to the idea in a prior post long past that people are basically looking for a camera that takes "good" pictures for them- this is correct. Many who are attracted to the Leica marquee are not true photographers in as much as they are toy lovers. With the Internet Leica toys have been available at rock bottom used prices (even the R-D1 is quoted as being a $1200 camera by some on RFf) so the M8 and it's tight control of retail pricing comes as somewhat of a shock for many. Since they must pay dearly for the camera most choose not to buy and instead harp on the cameras issues- something all cameras have one way or another; while others purchase the camera, take the same crap photos they always do, and proclaim the camera a POS due to the fact it and their pictures are not perfect.

Me I'm taking the same kind of pictures I always have, but I do notice the IQ of these images are much improved and I in turn am learning every day so I too am getting better as a direct result of this camera. I'll get my second M8, however when the IR issue is sorted. For the moment I am glad of having the M8 to work with right now, and unlike money, I can’t make any more time and this, for me, is the rarest of commodities.

Wow that was OT : )
T

 

[anselwannab]

Yes. We'd see a flood of vignetted images claiming the sensor is a POS. Of course, they would be the same who would not understand the tool that they're shooting with, and expect the same results from a P&S camera, which does not use the same technology, so you'd think they'd know.

"Flaws" will be found and exacerbated to gargantuan proportions until a legend is created.

I guess they call those enduring quirks.

I know everyone is sick of car comparisions, but it seems everytime BMW remakes the 3-series and Jeep redoes the wrangler, there are people screaming about how it has sold its soul yada, yada.

The more and more I read, I starting to think that the essence of Rf photography needs to be taken and a clean piece of paper solution engineered. Everyone bemoans the capital in their lenses. I thought we were all about the final image costs be damned? Why be held hostage by these collectors with Stockholm lens Syndrome. Yes, the camera companies will make more money, but that might be what it takes for them to take this market seriously.

As true photographers, and artists, we should be advocating for the best system, not a "Horseless carriage" that is a blend of old an new.

Mark

 

[jaapv]

Me I'm taking the same kind of pictures I always have, but I do notice the IQ of these images are much improved and I in turn am learning every day so I too am getting better as a direct result of this camera.

Hear-hear! You made the post I did not dare to make for fear of being shot down in flames, but you are SO right!

 

[sf]

The Foveon sensor is definitely not free from all IR issues

some say the screen installed in the new SD14 is actually an IR filter (sigma says its a dust screen), because of its tint

Actually they mentioned this on Luminous Landscapes photokina feature....

I follow. I think you're probably right. That makes alot of sense.

 

[HAnkg]

the essence of Rf photography needs to be taken and a clean piece of paper solution engineered. Everyone bemoans the capital in their lenses. I thought we were all about the final image costs be damned? Why be held hostage by these collectors with Stockholm lens Syndrome.

The problem is Leica has an economic incentive to produce an RF digital because of it's installed base of customers -M lens users (collectors with Stockholm lens Syndrome). What would be the financial motivation for another company to create a digital RF?

The small distance between lens and sensor that gives RF an edge in the analogue world - higher quality, smaller lenses together with a camera body that is more compact (the best performing and smallest lens in medium format are the Mamiya RF) are a huge disadvantage in digital. Starting with a clean sheet of paper would probably mean SLR type (big) retrofocal lenses better suited to digital. Canon could make an RF 5D (with a 'retro' look body design) that used the Canon prime SLR lenses - the body would be only a little more compact then a 5D (no prism). You would get RF viewing in an SLR sized package and a less expensive lens line that can leverage the economies of scale of a much bigger market. Somehow I don't think that would be very appealing to many RF users.

 

[clintock]

Would the angled incident light be a concern for a sensor if color was not an issue? A black and white only sensor?
I'd give up color in a second if it meant a full frame sensor with no kluge-coding compensation or artifacts..

 

[jaapv]

Yes- vignetting in exposure value. It is correctable in post-processing and on the M8 in firmware as well, but that loses dynamic range in the corners.

 

[Olsen]

Gabriel you eluded to the idea in a prior post long past that people are basically looking for a camera that takes "good" pictures for them- this is correct. Many who are attracted to the Leica marquee are not true photographers in as much as they are toy lovers. With the Internet Leica toys have been available at rock bottom used prices (even the R-D1 is quoted as being a $1200 camera by some on RFf) so the M8 and it's tight control of retail pricing comes as somewhat of a shock for many. Since they must pay dearly for the camera most choose not to buy and instead harp on the cameras issues- something all cameras have one way or another; while others purchase the camera, take the same crap photos they always do, and proclaim the camera a POS due to the fact it and their pictures are not perfect.

Me I'm taking the same kind of pictures I always have, but I do notice the IQ of these images are much improved and I in turn am learning every day so I too am getting better as a direct result of this camera. I'll get my second M8, however when the IR issue is sorted. For the moment I am glad of having the M8 to work with right now, and unlike money, I can’t make any more time and this, for me, is the rarest of commodities.

Wow that was OT : )
T

I think you are a bit arrogant on behalf of your fellow amateur photographers.

Further; I disagree with you. I see a clear trend showing most Leica shooters being far better photographers than those with, say, cheap P&S. And: What is 'good photography' to you is not neccessarily 'good photography' to me.

When our photography hobby has brought us all the way to these RRF pages it is far beond 'just taking pictures'. Collecting cameras, lenses and accessories are an integral part of that hobby. That more toys makes us less photographers is not true. They are tools and inspiration to take even better pictures.

Leica has no perfect fix for the IR casting on their new M8.

If you read their statements you see that this is what they are saying. To avoid soft and dark corners they had to cut out IR filters in front of the sensor. To avoid this they had to accept that. Some react to this with astonishment since most digital cameras has very heavy IR filtering. From 12 to 16 times IR filtering. It is a design flaw that M8 users will have to live with and try to compensate somewhat by putting IR filters on the lenses. As you will see, neither is this a perfect solution.

Since you already have a M8 you are in a better position than most to describe what a problem the lack of these IR filters really mean in day to day amateur photography. For quite a few, particularly the professional photographers, it means a hell of a lot. It is not accepted in any newspaper to come back from a press conferance at the UN building with picture files showing the new secretary general in a purple suit, - when we all can see that it is black.