Quattro marketing screwed up already?

Started Jun 16, 2014 | Discussions thread
Truman Prevatt
Truman Prevatt Veteran Member • Posts: 7,656
Re: comparing foveon to cfa is the problem

JohnLindroth wrote:

I started reading threads in this forum 6 years ago, and it's always in the technical comparison of a foveon sensor to a bayer cfa sensor that causes issues, and what are now long and extended, and rather unhelpful discussions.

If you count pixels as light measuring elements, then a bayer sensor equals the base output resolution. They use interpolation to fill in RGB elements to create a 3 level output image, so most people think that one RGB is a pixel. If you measure the same in a foveon sensor, then in most it is 3x the base output count, or in the Quattro 1.5 times, which is equivalent if you count pixels as light measuring elements, not output RGB elements. Most people base their concept of digital resolution back to the cfa, so having more pixels collected that output is foreign, and therefore people say Sigma is cheating or misrepresenting their cameras.

That's not quite right - to add to the confusion.  For example lets take the Leica M9.  It is an 18 MP CFA.  The spatial sampling matrix is 18 MP so the spatial frequency response is defined by 18 MP.  The CFA consists of a overlay of three color filters.  Like the Foveon, these aren't brick wall filters.  Some red & green photons get through the blue filters to the detectors and some green & blue photons get through the red filters to the detectors and some red and blue photons get through the green filters. That is each color filter has a unique spectral response and the raw processing uses this spectral response to calculate three colors per pixel.  If you would remove the CFA you get a pure 18 MP monochrome - oh yes they have done that it's called the Leica M9 Monochrome.

If you compared a B&W conversion of an M9 image to the same image taken by the M9 Mono you would find very little degradation.  Steve Huff has done some side by sides and found the results are very similar.  Of course the M9 Mono is better at monochrome since you do get a little degradation in going to color through the CFA to RGB processing then to B&W.  His take is the M9 Mono and the converted M240 images produce very similar results so resolution and sharpness wise the 24 MP M240 is about equivalent to the 18 MP (true resolution) M9 Mono.

Now the Foveon does not operate in a true RGB color space.  There is a detector at each pixel location for each of the three color elements of the color space.  The spatial resolution is defined by not the number of detectors but the number of detector triples. That is 15 MP of the SD1M is the spatial sampling lattice and the other 30 MP add to the color response. The two 15 MP lattices in the second and third levels do not alter the resolution - they are for color only.   There is a conversion process that takes place that turns the three layers with slightly different spectral responses into the RGB color space for follow on processing.  The Foveon does not require processing like the CFA to combine the spectral responses of the neighboring pixels to estimate color.  I expect some of the issues with noise and color shifts, etc., especially in not white light reported on this site and others of the Foveon has to do with the fact these spectral responses of the Foveon layers are fairly close.

The Sigma engineers seem to believe that by defining the resolution by the top level and the color by the sub sampled lower they can match the resolution of the SD1M while getting better color performance.  The Q sensor seems to have taken a page out of Dr. Bayer's play book in using subsampling of two of the spectral responses to estimate color at all the locations of the top sensor. This is not exactly the same approach as the CFA but it is similar.  The one thing about the SD1M and DPXM is that independent of color - the resolution was the same.  How it will work in the "real world" remains to be seen.  I am looking forward to seeing the real world samples.

I read about the concept of the Foveon sensor prior to the first camera release, and decided from a totally geek-oriented personality, that when I purchased a DSLR, it would have to use a 3 level sensor, because it was such a novel idea, and I appreciated due to my film photography start. And I did in 2008, a little late to the party.

But there is no need to claim that Sigma is messing up in trying to find a way to describe something that most people never even think about in the digital cameras.

And if I understand the words of Sigma's CEO, Mr. Kazuto Yamaki, correctly, continuing with the line of cameras is an honor to family in continuing his father's dream. I have no worries about them stopping because people don't understand.


 Truman Prevatt's gear list:Truman Prevatt's gear list
Fujifilm X-Pro2 Fujifilm X-H1 Fujifilm XF 18mm F2 R Fujifilm XF 35mm F1.4 R Fujifilm XF 56mm F1.2 R +8 more
Post (hide subjects) Posted by
(unknown member)
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow