Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Will it be FOVEON?

I think bayer.

Foveon is a super interesting technology but actually is implemented only in SIGMA products.
 
So... Greame says there's still some problems with Foveon. But the concept it's a lot better than bayer.

I did a job 2 weeks a go and I notice that even with zeiss ultra prime lenses the image is not that sharpen. Course, we don't want sharp edges, but it will better if the detais is a little more sharpen.

But any way... the movie was shot in 4K the final material will be viewed in 1920x1080... So... at the end I have a very fine detail image in 1920x1080.

Thanks to RED!!!
 
A working Foveon-like sensor would be great for digital cinema.

I don't like everything about Foveon images, but the color there looks very filmic, IMHO.
 
I found this: http://www.foveon.com/files/F13_image_sensor_Product_Flier_RevD.pdf 640x480 at 30fps sounds an awful way off being useful for digital cinema to me. 265x1768 at 5fps doesn't sound too useful either. Thanks Alexander for that thread link. It sounds like there are some serious and still un-addressed issues with Foveon technology. I mean, it sounds cool enough, but I bet the devil is in the details.
 
Albert Armes said:
I found this: http://www.foveon.com/files/F13_image_sensor_Product_Flier_RevD.pdf 640x480 at 30fps sounds an awful way off being useful for digital cinema to me. 265x1768 at 5fps doesn't sound too useful either. Thanks Alexander for that thread link. It sounds like there are some serious and still un-addressed issues with Foveon technology. I mean, it sounds cool enough, but I bet the devil is in the details.
Click to expand...

Of course that specification is not good enough for a RED. But maybe has a way to improve the quality of the sensor.

Would be wonderful if the RED folks project a sensor with no subsampling. That's all. Do not matter if foveon or whaterever!!!

Hope so!!!
 
Albert Armes said:
I found this: http://www.foveon.com/files/F13_image_sensor_Product_Flier_RevD.pdf 640x480 at 30fps sounds an awful way off being useful for digital cinema to me. 265x1768 at 5fps doesn't sound too useful either. Thanks Alexander for that thread link. It sounds like there are some serious and still un-addressed issues with Foveon technology. I mean, it sounds cool enough, but I bet the devil is in the details.
Click to expand...

Interesting stuff. Obviously 5fps isn't very useful, but once they can bring that up to at least 24/25fps things should get interesting. Also Albert, as I understand it 2652x1768 on a Foveon sensor is more resolution than you'll get on a Bayer-Pattern Mysterium sensor (because each of those pixel has full RGB colour).
 
"once they get it up to 24fps" - happy waiting :-)

Although in a Foveon, both luma and chroma is sampled at each and every pixel location, the chroma is too noisy to use as-is, and hence goes through a spatial denoise algorithm, so the end RGB image, although having per-pixel luma, does not have per-pixel chroma. The RAW image off a Foveon is incredibly under-saturated because silicon is not a good colour filter. That means a heavy matrix is needed to saturate and colour correct the colours, which is what creates the heavy noise burden on the chroma.

The resolution advantage of not demosaicing is not as great as it seems as the only company to really use the Foveon sensor, Sigma (who now owns Foveon) omits the necessary optical low pass filter from the camera, and hence all images from their camera are (when in focus) very aliased. In that respect a three chip system is better than a Foveon because the dichroic filters used are much superior to the silicon used by Foveon and hence the spatial chroma denoise is not necessary. Prisms have their own problems, but...

Measured luma resolution on the RED One is 3.2k, which is greater than the horizontal pixel count of the Foveon chip. Due to the maths of sampling theory, and the physics of light, you cannot achieve the rated pixel resolution of a chip without significant aliasing.

Graeme
 
So, for a working 3xLayer kind of cinema sensor, it has to be made from a material with better color filtering characteristics?
 
Yes, better colour filtering is needed, also, you need to be able to read off the sensor an awful lot faster than you can now to cope with the 3x amount of RAW data.
 
Graeme Nattress said:
"once they get it up to 24fps" - happy waiting :-)

Although in a Foveon, both luma and chroma is sampled at each and every pixel location, the chroma is too noisy to use as-is, and hence goes through a spatial denoise algorithm, so the end RGB image, although having per-pixel luma, does not have per-pixel chroma. The RAW image off a Foveon is incredibly under-saturated because silicon is not a good colour filter. That means a heavy matrix is needed to saturate and colour correct the colours, which is what creates the heavy noise burden on the chroma.

The resolution advantage of not demosaicing is not as great as it seems as the only company to really use the Foveon sensor, Sigma (who now owns Foveon) omits the necessary optical low pass filter from the camera, and hence all images from their camera are (when in focus) very aliased. In that respect a three chip system is better than a Foveon because the dichroic filters used are much superior to the silicon used by Foveon and hence the spatial chroma denoise is not necessary. Prisms have their own problems, but...

Measured luma resolution on the RED One is 3.2k, which is greater than the horizontal pixel count of the Foveon chip. Due to the maths of sampling theory, and the physics of light, you cannot achieve the rated pixel resolution of a chip without significant aliasing.

Graeme
Click to expand...

*sigh*

Sometimes I fear we're never going to have that cake and eat it too... :waaa: Thanks for the explanation Graeme.
 
Nope, you never get to have cake and eat it. However, I do believe the route that RED is the right one.

Graeme
 
Still it's nice to see a 4.6MP camera outresolve a 12MP camera (Sigma DP1 versus Canon G9). DP1 photos look amazing with that larger sensor.
 
Does it, though? 2640 Foveon pixels across, compared to 4000 bayer pixels across. Measured resolution on the DP1 cannot exceed 2640, however, lack of optical low pass filtering means that if you have details in the scene which would have to be recorded at above that resolution, they significantly corrupt and artifact all lower details. There are plenty of examples of such corruption in images from Sigma cameras.

Using a similar factor to what we measure on the RED, the Canon should do about 3.2k measured luma resolution with minimal aliasing beyond that, certainly not the the extremes you get with the DP1. Chroma resolution will max out at 2k.

If the DP1 was properly filtered to avoid aliasing, you'd measure less chroma and luma resolution than a 12mp bayer sensor. Because it's not, it cheats more, and causes artifacts that cannot later be removed from the image. No imaging system is perfect, but this is an obvious problem that is easily avoided.

Problem is many photographers like the look of aliases....

But the look of aliases is one reason why many video cameras look the way they do, in a negative way. Aliases also interfere negatively with motion estimated algorithms and compression.

Yes, larger sensors, on the whole, are a good thing...

Graeme
 
The G9 fares very well despite the tiny sensor and occasionally seems to have more detail, but then you look at things like purple fringing, dynamic range etc. and the DP1 is clearly better. Too bad the slow handling is killing the DP1. The G9 is a dream to use in comparison.
 
Well, given the vast sensor size difference, you can't really compare. DR is usually very closely related to sensor size, all other things being equal.
 
You must log in or register to reply here.
Back
Top