Will Quattro be the last of the Foveons?

Roland Karlsson said:

Scottelly said:

By the time you see a 200 MP CFA sensor, the Foveon sensors will be twice the resolution they are today.

Click to expand...

I.e. half the size they need to be competetive resolutionwise.

Roland Karlsson said:

If you don't think that's possible, then just look at the latest from Samsung here:
http://connect.dpreview.com/post/22...-thinner-16mp-sensor-with-1-micrometer-pixels

That sensor has 1 micron pixels. They wouldn't make such a sensor if it didn't work.

Click to expand...

Of course it will work.

Roland Karlsson said:

Sigma could easily make 3 micron pixels,

Click to expand...

Easily?

Click to expand...

Roland Karlsson said:

Roland Karlsson said:

allowing the top layer of their sensor to capture nine times as much light.

Click to expand...

Hardly. Lots of light goes just through that sensor.

Click to expand...

victorgv said:

Hi Tom,

In my of course personal opinion part of the reason that foveon images look different is that there are much more colour overlap between layers comparing to Bayer filters

Click to expand...

jande9 said:

An interesting idea and probably much studied by the camera makers but is it that different than an anti-aliasing filter plus interpolation used now?

Binning 4X4 blocks of pixels would be a problem for their megapixels marketing jargon, making a 24 megapixel interpolated sensor into an 6 mp sensor.

I think the fact that there was so little improvement between the last 2 generations of foveon sensors indicates that the development potential of the sensors has come to an end, while cfa sensors still seem to have much room for improvement.

The foveon look can be approximated with sensor shifting, and I anticipate someone coming up with prism colour splitter blocks for a stills camera, which would are not too big on a small sensor.

Jan

Click to expand...

Roland Karlsson said:

xpatUSA said:

Roland Karlsson said:

Here is a question you might know the answer to: is there a CFA raw converter out there that can do it with color filter quads, without interpolation between the quads?

Click to expand...

Although the question was not asked of me personally, RawDigger can export such image as a TIFF.

Click to expand...

Interesting.

Maybe should test it (even if I do not believe it is a good idea).

Click to expand...

DMillier said:

This all sounds remarkably complicated in my opinion.

I've been playing this Foveon vs Bayer game for enough years now to not really need over-complicated explanations.

There isn't a huge difference between the results from the technologies IMO. How big the difference is depends on how you display your images.

- if you pixel peep on screen, Foveon looks sharper

- if you downsize on screen, bayer looks to all intents and purposes identical to me

- if you print huge you might start to see Bayer demosaic artefacts on a regular basis and maybe Foveon does a better job in skilled hands. Personally, although I thought Rick's print tour prints were very big and very fine, when I've seen somewhat smaller print comparisons I could never really tell them apart

- In smaller prints I think there is zero difference in most cases. Certainly in my testing, I can see no meaningful difference between my 6MP Fuji X10 compact and my DP2M in A4 prints.

- there are always going to be edge cases where you unexpectedly find a big difference: for example I find the DP2M is very good at rendering the rough surface of weathered plaster and stone work. It's not so much Bayer can't do it, more that it comes easy to the DP2M. The danger is that people cherry pick extreme examples and claim they are universal.

IMO, if you downsize a D800 to 18MP, you will get a result that most of the time, most people can't tell from a Foveon.

You might find Keith Cooper's recent 5Ds review to show how in many, many cases people can't see quality differences, nor care about it.

https://luminous-landscape.com/canon-5ds-review-through-print-performance/

Quotes from the article:

"[50Mp vs 16MP]: At ~13×19" there is virtually no difference visible without getting a powerful magnifying glass out, "

"It’s a bit of a tough pill to swallow for some photographers, but most people couldn’t spot the differences in print quality between good and superb if it fell on them."

Modern cameras all produce superb results if used properly. No need for geekish obsession over miniscule unimportant differences.

Click to expand...

DavidMillier said:

Tom Schum said:

atom14 said:

Succinct: No!

Hyperbolic: Never in a million years!

Dramatic: Non!

Emotional: Mais non!

Laconic regional: Nah, mate! Nevvah!

Toffish: Ehw heardly, ewld chap!

Parliamentarily: The question is resolved in the negative.

...

... and so on, and so 4th...

atom15. No! atom14.

Click to expand...

This idea was not equal to bayer.

Bayer has interpolation between pixels. The effects of this can extend several pixels in every direction. By downsizing 6:1, we basically reduce this proximity effect but it cannot be eliminated. So my posted photo is only an approximate example.

The idea is to use the fine structures now being developed for high pixel count sensors, and do a smaller set of complex "pixels" that contain two things: (1) Enough photosites (with color filters above each) to produce complete colorimetry from just the data from the small group (3, 4, 6, or 9 photosites) and (2) Individual diffusers over the group that makes up the "pixel" to prevent color aliasing within the small group.

These two things together give us a full-color "pixel" exactly like the Foveon vertical stack, but planar. There is no interpolation needed between these "pixels" anymore because they each decode the color as well as the intensity of the light striking them.

This is functionally equal to Foveon, perhaps only slightly more complicated, but again each pixel offers full color, and no pixel information is interpolated between these pixels, same as Foveon.

Note this is just a slight evolutionary development of existing CFA technology.

You can't really take a CFA image and compare it to a Foveon image at this point, because sensors as I've described have not yet been made.

I have a suggestion regarding lackluster colors in your bayer images and that is to cut off the dynamic range at top and bottom of the histogram, as I did in my sample image in the first post. Give it a try. It might be quite interesting, at least it was to me.

Back to the sensor idea: Since this is an easy way for CFA technology to divest itself of edge problems and artifacts and compete directly with Foveon at the pixel level, when it appears will Foveon survive? I'm guessing not.

And it would immediately obsolete existing bayer CFA technologies as well.

--
Tom Schum
Celebrate mediocrity (in moderation)

Click to expand...

--
"...while I am tempted to bludgeon you, I would rather have you come away with an improved understanding of how these sensors work" ---- Eric Fossum
Galleries and website: http://www.whisperingcat.co.uk/
Flickr: http://www.flickr.com/photos/davidmillier/

Click to expand...

Scottelly said:

jande9 said:

An interesting idea and probably much studied by the camera makers but is it that different than an anti-aliasing filter plus interpolation used now?

Binning 4X4 blocks of pixels would be a problem for their megapixels marketing jargon, making a 24 megapixel interpolated sensor into an 6 mp sensor.

I think the fact that there was so little improvement between the last 2 generations of foveon sensors indicates that the development potential of the sensors has come to an end, while cfa sensors still seem to have much room for improvement.

The foveon look can be approximated with sensor shifting, and I anticipate someone coming up with prism colour splitter blocks for a stills camera, which would are not too big on a small sensor.

Click to expand...

Click to expand...

Scottelly said:

Yeah, and there was so little difference between the photos produced by the Nikon D200 and the D300 that it proved that the development potential for the APS-C sensor had come to an end too . . . yeah, that's the ticket.

Click to expand...

Kendall Helmstetter Gelner said:

jande9 said:

<...>

I think the fact that there was so little improvement between the last 2 generations of foveon sensors indicates that the development potential of the sensors has come to an end

Click to expand...

How is that in any way a "fact"? The newer sensors deliver improved resolution, better high ISO support, with a totally different sensor structure, allowing for lots more flexibility in future paths for sensor design...

By your metric Bayer sensors have hardly advanced in the last five years.

Click to expand...

Kendall Helmstetter Gelner said:

--
---> Kendall
http://www.flickr.com/photos/kigiphoto/
http://www.pbase.com/kgelner
http://www.pbase.com/sigmadslr/user_home

Click to expand...

xpatUSA said:

This sub-discussion would be easier to follow if y'all could stick to pixel area only.

And "nine times as much light" is meaningless without any units and lacking, as is usual in this forum, a definition for the amount of light.

Click to expand...

xpatUSA said:

Anyway, small pixels have smaller well capacities so they gather less "light" pro-rated by area. And, two sensors of the same area gather the same amount of "light" almost irrespective of effective pixel area. You both knew that, of course.

Click to expand...

Roland Karlsson said:

xpatUSA said:

This sub-discussion would be easier to follow if y'all could stick to pixel area only.

And "nine times as much light" is meaningless without any units and lacking, as is usual in this forum, a definition for the amount of light.

Click to expand...

I do not see the problem. If it really did collect nine times as much light, then it did so independent of the amount of light.

Click to expand...

Scottelly said:

Actually David, I think there IS a need for this "geekish obsession" over miniscule, "unimportant" differences. What you call unimportant is to some people, like me, VERY important. So important in fact, that I plan to move away from digital for my landscape photography, because digital is not capable of capturing the image quality that I want. Visit one of Peter Lik's galleries, and tell me you can make a print like that from your digital shots. You can't. NO digital camera can shoot photos like that . . . unless you have some gigantic 261 megapixel Red Mysterium Monstro 617 sensor in a custom camera, or some other ridiculously expensive piece of equipment that almost no landscape photographer can afford.

If you're limiting your prints to 13x19 then NO, you don't need more than a Nikon D800 or even a Sony A99. If you're planning to print at 20x30 and you want to be able to print at 24x36 (inches), then you'll want the best you can get, so it will make sense to get a Sigma Quattro, a Nikon D810, or a Canon 5Dsr. But if you want to have the ability to make prints that look as good as Peter Lik's prints, you need something better . . . like a 4x5 large format camera. That's why I'm planning to get an 8x10 large format camera. Digital cameras just don't compare. No, I won't be giving up my digital cameras. I like to shoot, and it takes too long and too much money to shoot a lot with an 8x10. But for ultimate quality, nothing beats large format film. Yes, if I could afford a BetterLight back, I could shoot a lot more, but those things are ridiculously expensive and limited in their use even more than film, in some ways.

With all this said, I understand that most people rarely print larger than 13x19, if ever, so for most people what you say is correct and practical. In fact, for most people, a Sony RX100 of some flavor should do it. Those things are great performers and produce high quality images. If someone wants interchangeable lenses, then they could use any of the m4/3 cameras, any APS-C sensor camera that's available today, or any of Sony's mirrorless cameras, and they'd be able to capture high quality images good enough to print up to and including 20x30. So in such cases, which are the vast majority, it is unnecessary to have all this questioning about whether Foveon is better or whether Merrill is as good as Quattro, etc. etc.

Click to expand...

1. Those that genuinely needs it, for some reason.
2. Those that think they need it or believe they might need it some time.
3. Those that think it is fun or have some other geekish motivation.

xpatUSA said:

Roland Karlsson said:

xpatUSA said:

This sub-discussion would be easier to follow if y'all could stick to pixel area only.

And "nine times as much light" is meaningless without any units and lacking, as is usual in this forum, a definition for the amount of light.

Click to expand...

I do not see the problem. If it really did collect nine times as much light, then it did so independent of the amount of light.

Click to expand...

I meant amount of light collected.

If you don't see the problem with the phrase "nine times as much light" then there's not much I can say about it that would mean anything here.

Click to expand...

Roland Karlsson said:

Scottelly said:

By the time you see a 200 MP CFA sensor, the Foveon sensors will be twice the resolution they are today.

Click to expand...

I.e. half the size they need to be competetive resolutionwise.

Click to expand...

Kendall Helmstetter Gelner said:

Which is OK since lenses will only be able to deliver resolution to the level of detail that Foveon sensor captures, not the 200 MP "Homer Sensor".

I find it really amusing to see all the talk about 200 MP cameras when the smaller pixels will simply go into making cell phone cameras a bit higher res.

Click to expand...

Roland Karlsson said:

Kendall Helmstetter Gelner said:

Which is OK since lenses will only be able to deliver resolution to the level of detail that Foveon sensor captures, not the 200 MP "Homer Sensor".

I find it really amusing to see all the talk about 200 MP cameras when the smaller pixels will simply go into making cell phone cameras a bit higher res.

Click to expand...

You are ignoring the power of over sampling and digital filtering.

Click to expand...

Scottelly said:

Laurence Matson said:

xpatUSA said:

From the response, you appear to have wasted your time, Tom.

If it's any consolation, long ago I compared the 3.4MP SD9 with a 12MP Panasonic, both down-sized considerably and could see little difference - which bears considerably in favor of your assertion.

Click to expand...

And if I downsize a Tesla S, the Nisan Leaf can keep right up and go the same distance.

Any more - pardon my ruthlessness, which I know is something you believe should be banned - ridiculous conclusions?

My point is that the comparison mentioned above is as useless as mine.

Take offense if you like. Perhaps you could understand my frustration at the constant useless diversions which ultimately lead to complete misunderstanding.

And if you think I am nuts, go back and read Merrill's one-day postings from about ten years ago. His logic still prevails.

Click to expand...

Click to expand...

Laurence Matson said:

xpatUSA said:

Laurence Matson said:

xpatUSA said:

From the response, you appear to have wasted your time, Tom.

If it's any consolation, long ago I compared the 3.4MP SD9 with a 12MP Panasonic, both down-sized considerably and could see little difference - which bears considerably in favor of your assertion.

Click to expand...

And if I downsize a Tesla S, the Nisan Leaf can keep right up and go the same distance.

Any more - pardon my ruthlessness, which I know is something you believe should be banned - ridiculous conclusions?

My point is that the comparison mentioned above is as useless as mine.

Take offense if you like. Perhaps you could understand my frustration at the constant useless diversions which ultimately lead to complete misunderstanding.

And if you think I am nuts, go back and read Merrill's one-day postings from about ten years ago. His logic still prevails.

Click to expand...

I will not take offense. I will, however, be unable to read your posts for a while. Sorry.

Click to expand...

Nevertheless, I hope you (the other reader) might at least have the goodness to really read this tread:

http://www.dpreview.com/forums/post/29731759

A careful reading might provide better edification and also might help you understand better my frustration at the constant sideshows around here. At one time, this was a forum for deep discussion with many more notable participants than any other on this site.

So, do not read-on MacDuff, but perhaps someone else could.

Click to expand...

jande9 said:

I think the point was more along the lines of, if you view images on a device that displays 2mp, it makes little difference if you start with 4.7mp or 12mp.

Click to expand...

Roland Karlsson said:

Kendall Helmstetter Gelner said:

Which is OK since lenses will only be able to deliver resolution to the level of detail that Foveon sensor captures, not the 200 MP "Homer Sensor".

I find it really amusing to see all the talk about 200 MP cameras when the smaller pixels will simply go into making cell phone cameras a bit higher res.

Click to expand...

You are ignoring the power of over sampling and digital filtering.

Click to expand...