Interesting Panasonic - patent

[GaryLH]

http://www.43rumors.com/panasonic-p...latform-alike-camera-with-three-layer-sensor/

Whether Panasonic really follows Olympus n this open source... What is really interesting is theirs uses three separate sensors (one for each RGB color).. Uses a prism to split the paths to each sesnor. Not foveon, but should be much better for high iso, but definitely going to add bulk (cube makes sense for this design it guess).

Gary

 

[tsiklonaut]

I think all sensor manufacturers secretly know and admit the mathematically interpolating sensors suck big time, but amazing why they'd want to go such lenghts to get a simple non-interpolating 100% full information on each pixel sensor - this would not only be a huge bulky camera (if the sensors are APS-sized or bigger) but also would cost a lot!

Hasselblad has "solved" this fundamental issue by a shifting sensor - basically taking 3 or more shots to compile a non-interpolated shot, with sensor shifting by a pixel row distance so they end up 100% info collected on each pixel. But the introduced time-lag obviously creates other issues and limitations.

I wonder if Foveon technology is too forbidden patent-wise to be used by other manufacturers? Their simple but elegant design inspired by proper old fashioned color film, can't see why not or they just lack the key know-how how on building such sensor(?)

 

[GaryLH]

I agree this will be bulky.. Since Panasonic makes their own m43 sensors.. I suspect the bulk will not be too bad compared to a apsc or ff design.

In terms of other manufacturers, I think there is too much infrastructure based on Bayer that gives them pause. The original foveon patents should be ending soon I would hope.

My wish list
- other manufacturers make foveon like sensors
- a csc style camera using a foveon like design.
-- this open source may be closest I will ever c.
- a monochrom sensor from someone else other than Leica (too expensive for my taste).. While the foveon makes great monochrom pictures (much better than Bayer sesnor), there is no csc camera.

Gary

 

[YYV_146]

I think all sensor manufacturers secretly know and admit the mathematically interpolating sensors suck big time, but amazing why they'd want to go such lenghts to get a simple non-interpolating 100% full information on each pixel sensor - this would not only be a huge bulky camera (if the sensors are APS-sized or bigger) but also would cost a lot!

Hasselblad has "solved" this fundamental issue by a shifting sensor - basically taking 3 or more shots to compile a non-interpolated shot, with sensor shifting by a pixel row distance so they end up 100% info collected on each pixel. But the introduced time-lag obviously creates other issues and limitations.

I wonder if Foveon technology is too forbidden patent-wise to be used by other manufacturers? Their simple but elegant design inspired by proper old fashioned color film, can't see why not or they just lack the key know-how how on building such sensor(?)

Foveon is as flawed as Bayer sensors. A lot of light is lost once photons get to the innermost layer, and the gain has to be boosted to even out the difference. When the front layer is operating at iso 100, the back layer (green, I think) has to operate at close to iso 400. High ISO performance on Foveon cameras will therefore never be as good as a normal Bayer camera.

The layering also creates the same issues as microlenses on mirrorless cameras. Using short-register, fast lenses on them will result in large amounts of falloff. I suspect this is why the DP cameras have long-register, slow lenses.

 

[GaryLH]

As I remember, a Chinese company did a dp to Leica rf csc conversion.. I don't remember seeing the results of this conversion.. I would suspect pretty bad... 🙁.

If someone were going to do a csc foveon camera, the native lenses should account of the type of issues u mention.

I find the foveon images have a look that is hard to beat from any Bayer sensor camera below a 36mp image. Even that being said..not everything I shoot needs to be shot w/ a foveon based camera or anything as high as even 24mp.

Gary

 

[tsiklonaut]

The layering also creates the same issues as microlenses on mirrorless cameras. Using short-register, fast lenses on them will result in large amounts of falloff. I suspect this is why the DP cameras have long-register, slow lenses.

True, but I always though the in-camera "signal-boost" (noise reduction) algortithms have gotten so advanced latery (i.e. ISO 100 000+ cameras are becoming a norm, and cameras going as far as ISO 512 000 are for sale these days) that this isn't really a big issue.

On Foveon the lowest layer should be red and as I understand the red spectrum has much better "penetration" properties (i.e. to go to extremes: infrared that surpasses thick non transparent materials vs UV that gets stopped with just with thin transparent glass), so blue has the worst penetration properties that sits on top, with B->G->R they are placed like this purposefully, not all "colors" (wavelenghts) being equal this optical effect should balance out some of that falloff(?)

I find the foveon images have a look that is hard to beat from any Bayer sensor camera below a 36mp image. Even that being said..not everything I shoot needs to be shot w/ a foveon based camera or anything as high as even 24mp.

I think that's the mark of non-interpolated image. My ancient 4,7MP Sigma runs circles around on most 10-12MP Bayers with similar sensor size by the look and richness of image IMHO, I've never gotten so clear and chrispy looking shots yet with excellent color-depth from any Bayer, I need to resize down to some 3X to get similar "feel" with a Bayer images. Interpolation really is clear when you compare them.

That being said Sigma's going down-hill with the new generation Foveon Quattro that is also fundamentally interpolating sensor now as well unlike the Foveon Merrill that I though they could improve with their next generation that never came - instead of improving they try to make it a cost-effective sensor with huge compromises just like Bayer manufacturers have done since the beginning.

I remember Canon promised a multi-layer sensor around a year ago, where's that project, forgotten or proved to be too costly?

 

[YYV_146]

As I remember, a Chinese company did a dp to Leica rf csc conversion.. I don't remember seeing the results of this conversion.. I would suspect pretty bad... 🙁.

If someone were going to do a csc foveon camera, the native lenses should account of the type of issues u mention.

I find the foveon images have a look that is hard to beat from any Bayer sensor camera below a 36mp image. Even that being said..not everything I shoot needs to be shot w/ a foveon based camera or anything as high as even 24mp.

Gary

The results are not good - extreme toned falloff with basically any type of short register lens. Some RF wides have 3-4 stops of extra vignetting...makes the A7r's corner performance look stellar in comparison.

Seems that without some sort of optical optimization, it is only possible to use SLR lenses with these cameras - at the point which you'd be better served with a SD body.

 

[GaryLH]

That being said Sigma's going down-hill with the new generation Foveon Quattro that is also fundamentally interpolating sensor now as well unlike the Foveon Merrill that I though they could improve with their next generation that never came - instead of improving they try to make it a cost-effective sensor with huge compromises just like Bayer manufacturers have done since the beginning.

I remember Canon promised a multi-layer sensor around a year ago, where's that project, forgotten or proved to be too costly?

I guess I am the minority of the Quattro sensor then... I don't like the camera design as much as the Merrill, but I am fine w/ the sensor. Now that I have had the dp2q for a while.. I have come to the conclusion that the Merrill cameras are staying.. They are going to be my mainstay landscape cameras.

On the other hand, the Quattro is not being used very often in color or I large raw (where the 4 to 1 color algo kicks in), but in the small raw format which give 1 to 1 to 1 (RGB).. Mainly monochrome mode as well. This is essentially your dp2x but w/ far better high iso capability and better af speed.

I actually like what I c out of the 5 mp (15mp foveon speak) then what I cu out of any 16mp Bayer or xtran sensor. Even the 24mp sensor of the a7 is hard put to beat it.. Most of the time it is at best its equal.

The large raw still looks better than a lot of 36mp camera out there IMHO. Is it not better than the Merrill design.. Over not really except for higher iso capability, faster af and if u use small raw, much faster sd card write time.

If u don't need the higher iso..get the Merrill, u will be happy. For my other types of shooting, I need faster af and higher iso where the dp2q has a better chance of being used more often.

Gary

 

[GaryLH]

Seems that without some sort of optical optimization, it is only possible to use SLR lenses with these cameras - at the point which you'd be better served with a SD body.

Thanks for the update on the results.. I assumed that this was the case.

I have an sd1 Merrill w/ a range of lenses.. It comes w/ all the dslr burden w/o the super fast dslr af speed.. At best it is around d200 af speed. And no live view.

As I said in the prior post, if someone ever makes a csc foveon based sensor, the native (not legacy rf) lenses should work fine. Sigma already makes a nice set of apsc e and m4/3 mount lenses (19, 30, and 60 f2.8s) for example. Hopefully, something like that would be the basis of a csc from them. I would love for them to introduce a 16f2.8 or a 16-70 f2.8-4 for the csc.. Probably never going to happen, sadly...

Gary

 

[YYV_146]

Thanks for the update on the results.. I assumed that this was the case.

I have an sd1 Merrill w/ a range of lenses.. It comes w/ all the dslr burden w/o the super fast dslr af speed.. At best it is around d200 af speed. And no live view.

As I said in the prior post, if someone ever makes a csc foveon based sensor, the native (not legacy rf) lenses should work fine. Sigma already makes a nice set of apsc e and m4/3 mount lenses (19, 30, and 60 f2.8s) for example. Hopefully, something like that would be the basis of a csc from them. I would love for them to introduce a 16f2.8 or a 16-70 f2.8-4 for the csc.. Probably never going to happen, sadly...

Gary

I agree, although (for me) being limited to F2.8 really reduces the appeal of the DP bodies. I'd assume that it's a design compromise to have small lenses and decent wide-open performance, but for someone who's used to shoot at F2 or even F1.4 on an FF camera, the DPs have a double penalty on any work that isn't done in perfect light.

I'm not even sure the different merril cameras have the exact same sensor - it might be the case that Sigma had to optimize the sensor (microlenses and/or coverglass) for each DP body. If that's the case, we'll never see a Foveon CSC.

 

[Particular]

The split prism arrangement is the standard in 3ccd video cameras. Panasonic and others have been using this system for decades. It would be a novel approach in a still camera. I am a little skeptical though if there is enough room behind the m43 mount to fit the prisms.

 

[GaryLH]

It maybe a 4/3 design instead of a m43 since the sensor is essentially the same size, except what is needed for the difference (what they used on their early 43 dslrs).

Gary