Home

Please support mflenses.com if you need any graphic related work order it from us, click on above banner to order!
SearchSearch MemberlistMemberlist RegisterRegister ProfileProfile Log in to check your private messagesLog in to check your private messages Log inLog in

Feelings about the Sigma in Real Life
Post new topic   Reply to topic
Home   Digital SLR Cameras
View previous topic :: View next topic  


PostPosted: Mon Nov 23, 2009 8:39 pm    Post subject: Reply with quote
A very enjoyable thread, and a BIG thank you to Steaphanie for
providing some of the physics data. I am not knowledgeable
in anything "electro***" but I do find interest in reading the
various graphs. I find it fascinating to read material from someone
like Steaphanie, who is at least attempting to show the scientific
basis that ends up in being a simple image.

I also enjoy the "counters" to Steaphanie's treatise regarding
light handling behavior on sensors.


PostPosted: Mon Nov 23, 2009 9:20 pm    Post subject: Reply with quote
rbelyell wrote:

there is no absolute science in this as is perfectly exemplified by steaphany and sichko going back and forth differing over their subjective analysis of the exact same technical data.


The fact that steaphany and I disagree about something does not mean that we are both wrong. One of us might be right ! Readers can make up their own minds.

So for example steapahny suggests that Einsten received the Nobel prize for the photovoltaic property of a pn junction. My own intrepretation is that the prize was for his understanding and intrepretation of the photoelectric effect - a different phenomenon. As far as I can discover, the photovoltaic effect in a pn junction postdates the award of Einstein's prize by about 20 years. But you can check for yourself.

Quote:
we pretend there is science to justify our subjective opinions, perhaps because we are insecure about them.


You must speak for yourself.


PostPosted: Mon Nov 23, 2009 10:00 pm    Post subject: Reply with quote
john no offense intended. my point was pretty simple, that people can interpret the same technical data in very different ways. of course we can all decide what the data mean for ourselves, thats what we all do all the time. i object to people using their interpretation of technical data as a springboard to proclaiming a universal truth about one piece of equipment over another in this particular realm of image reproduction. when interpretations differ as to fact, no one can proclaim universal truth. i stand by that objection.
tony


PostPosted: Mon Nov 23, 2009 10:59 pm    Post subject: Reply with quote
rbelyell wrote:
john no offense intended.

And none taken. However you state...
Quote:
i object to people using their interpretation of technical data as a springboard to proclaiming a universal truth about one piece of equipment over another in this particular realm of image reproduction.

...and I agree with you. Hence the nature of my response(s).

This thread is about the superiority of the Foveon sensor over the Bayer sensor. Look at what the OP wrote (to Nikon, and repeated here)..

steaphany wrote:
There is nothing wrong with Bayer masked imager based cameras, if the photographer is willing to compromise on image quality by allowing interpolation algorithms to guess what colors are actually being focused on the imager. ...

....Sigma with their Foveon X3 based SD14 employs a stacked photodiode junctions to achieve color image sensing


In support of this superiority there has been presented a mixture of "facts" . Some of these are simply untrue - in any sense. Their use questions the quality of the arguments which have been presented.

You say..

Quote:
when interpretations differ as to fact, no one can proclaim universal truth.


In the real world it's sometimes useful to believe one side or the other. If you are infected with HIV which do you prefer ? An anti-retroviral drug or garlic and beetroot ?


PostPosted: Tue Nov 24, 2009 2:09 pm    Post subject: Reply with quote
very hard choice, since i love garlic but intensely dislike beets...Laughing

and yes, the thread was about the superiority of the foveon sensor--the thread was started as a result of my ? to steaphany about how she FEELS about using the foveon sensor, especially as it entails a number of both technical and user compromises. to state my opinion in this regard hopefully more clearly than i have the past several times, i do not believe one can PROVE picture quality with technical data. this is a FEEL game to me. thats MY opinion, im not forcing it on anyone, but i certainly dont want anyone forcing debatable 'facts' on me designed to disprove what i SEE. thats what was happening in my half of the steaphany/sichko, tony/anu debate.

sure, we all analyze conflicting opinions about data and make decisions. i did when i bought my sigma. i first read through the technical discussions that whetted my appetite, and then looked at many pictures taken with the camera that imho confirmed the opinion of one side in that technical debate. my final decision was based on what i saw in those images, which is totally subjective. thats all...


PostPosted: Tue Nov 24, 2009 3:07 pm    Post subject: Reply with quote
rbelyell wrote:
anu it was fun debating with you, but i think absolute statements, like the foveon advantage was a myth, are, to quote you, just silly. there is no absolute science in this as is perfectly exemplified by

Foveon advantage is and was a myth. That is a fact. In science things are measured and measured again, and these measurements are analyzed, compared and so on. So far there are no measurements that would show that Foveon would have given an advantage when it was hot. Plenty of measurements show clearly otherwise.

Foveon sensor idea of getting RGB out of single pixel site is solid, however, the choice of method of doing this is very problematic and also the implementations were always a bit lacking.

Quote:

justify our subjective opinions, perhaps because we are insecure about them. i dont need to point to something external to justify my choices,

Sorry, not insecurity here regarding this. It is just about observations and measurements and if one wants to know, about finding the answer to question why, which in this case is also rather interesting.

I must admit, that if I had a Sigma camera and someone told me that Foveon stinks, I might feel defensive and insecure Smile

Quote:

nor do i feel it necessary to take my subjective analysis of technical data to prove someone else's choice of hardware is inferior. there is a whole community of very good photographers, and very good engineers, who think your analysis of foveon is full of beans. doesnt mean they're right; doesnt mean you're right; means there really isnt a 'right'.


When talking about technologica advantages and disadvantages, we are not interested in the art of photography, but about measurements and the analysis of the results. If the measurements and anylysis are done properly, there is no subjectivity involved, but facts. The fact is, that Foveon sensor was even at it's best an inferior product compared to the best bayer-sensors of the time. It was an innovative and novel product, but that's it.

No engineer who measures Foveon results, and compares them to for example Canon sensor results of the same era, will disagree with me on this. Sorry. I mean, it is not a close comparison, but like comparing Lada and Mercedes Benz. Still, if one just uses the car for grocery shopping, there is nothing wrong with Lada.


PostPosted: Tue Nov 24, 2009 4:04 pm    Post subject: Reply with quote
anu, youre just plain wrong, and the way i read it, we have right here an engineer who does indeed disagree with your interpretation of foveon vs other sensors and thats steaphany. and although i think its needlessly strident, and somewhat foolish, im not defensive about an opinion that says foveon stinks, im defensive about youre determination to have your OPINION regarded as fact. nothing more to be gained from this exchange, have enjoyed it up to now, so thats all you will hear from me on this topic.
tony


PostPosted: Wed Nov 25, 2009 8:44 pm    Post subject: Reply with quote
Anu wrote:
rbelyell wrote:
ok, now this is getting good!!!!

anu, i agree the statement canikon leaves everything in the dust is silly. why isnt the statement 'foveon sensors are quite bad by modern standards' not the same silly??!!

Because it is a fact Smile

Compared to modern APS-C sensors Foveon sensors are inferior in pretty much all categories from noise and resolution to color. Off the top of my head I can not think of a single quality of Foveon that is superior to modern sensors from other manufacturers, when it comes to taking photographs.

rbelyell wrote:

now if you want to say sigma's CAMERAS are quite bad by modern standards, i can more agree with you. i am finding way too many limitations in the camera compared to their competitors. but frankly the only thing that has kept me from slamming my sd14 into the wall IS the foveon sensor.


Why do you like that sensor?

rbelyell wrote:

lets keep heating this up!!


More heat is good here in Finnish winter - it is so cold here that polar bears are appearing in the streets already, searching for food.


Yes. The post asked for opinions.


PostPosted: Tue Dec 01, 2009 8:35 pm    Post subject: Reply with quote
Sorry for being MIA, busy leading into Thanksgiving and my ISP had an outage which started early Saturday when all their tech staff were off for the holiday. Things only came back today and I will be catching up.


PostPosted: Wed Dec 02, 2009 12:43 pm    Post subject: Reply with quote
3 simple facts:

1. bayer mask is 33 years old invention

2. not any single branch of graphics - with the only exception of photography - considers this approach as acceptable. Rendering of 1/3 of the image + following interpolation is not accepted by any other branch as quality method. And even the fact, that the same hardware resources can produce "3-times higher" resolution, isn't considered as satisfactory argument to utilize this approach...

3. majority of manufacturers tell us, that megapixels are most important for image quality, professional work, etc...

They created quite difficult situation - not for Sigma (Sigma customers doesn't care of bayer-style megapixels), but for themselves. It's obvious, that 8MP full color sensor would wipe out any current bayer-based APS-C (and likely FF) product in terms of image quality. But what would the manufacturers tell to their customers?

Majority of customers wouldn't accept upgrade from 16MP camera to 8MP camera. They would feel, that their manufacturer is cheating them (or that they were mystificated previously). It would be very risky for Canon, Nikon and other brands to introduce full-color camera. Who can say, if a Nikon customer, who uses 14MP product, will switch to new Nikon 8MP full-color product and not a new Canon 17MP camera?

Nikon has full-color sensor, Fuji has full-color sensor, some sources told, that Canon too, but they have no reason to launch it, because it would be more risky then to continue in well-tried megapixel war.

I'm sure they will release it in the moment, when Sigma - or any other brand - will introduce full-color sensor more than competitive to current solutions.


PostPosted: Wed Dec 02, 2009 3:36 pm    Post subject: Reply with quote
no-X wrote:

They created quite difficult situation - not for Sigma (Sigma customers doesn't care of bayer-style megapixels)

Sigma customers have allowed themselves to be conditioned into believing that having aliasing and false detail is actually a good thing and that by comparing at pixel level 4 million foveon pixels to 4 million bayer pixels, one is making intellectually honest comparison when bayer sensors have three times the pixel count. Honestly, IQwise there is little or no reason to buy a Sigma camera.
Quote:

, but for themselves. It's obvious, that 8MP full color sensor would wipe out any current bayer-based APS-C (and likely FF) product in terms of image quality. But what would the manufacturers tell to their customers?

This is vast oversimplification and also incorrect.

If you seriously think that 8Mp (foveon style?) full color sensor would likely wipe the floor with 24Mp full frame CFA-sensors, you really should stop using the drugs you're using Smile

Today's full color sensors have less than 5 Mp, and the image quality is at most worth about 10 CFA megapixels. At most, probably more like 8.

Second, it is not easy to make a good full color pixel sensors, not easy at all. Foveon is a good example of one solution which basicly stinks (compared to bayer sensors of the same generation) - very bad color, very noisy, low dynamic range.

I hope that some day we can ditch the CFA, but I'm afraid the day won't come anytime soon.

Quote:

Majority of customers wouldn't accept upgrade from 16MP camera to 8MP camera. They would feel, that their manufacturer is cheating them (or

I thought that Foveon sensor is occasionally advertised as having three times the pixels it actually has.

Anyhow, people will buy anything, no matter what the technical specifications say. There is no reason to believe that one would not same or lower numbers of pixels in a new camera if the new camera looks sexier Smile

Quote:

that they were mystificated previously). It would be very risky for Canon, Nikon and other brands to introduce full-color camera. Who can say, if a Nikon customer, who uses 14MP product, will switch to new Nikon 8MP full-color product and not a new Canon 17MP camera?

Well, they would not offer a 8Mp full color camera to replace a 18Mp bayer based camera as it would lower the resolution (and if using the Foveon tehcnology, it would lower pretty much all the other aspects of IQ too), while increasing the amount of data transfer need.

Regarding that amount of data - single color pixels require three times the data transfer speed Bayer-sensor pixels need: one can think Bayer as a compression method, not too unlike from jpeg - a little detail is lost, but the rewards are well worth it.

Quote:

Nikon has full-color sensor, Fuji has full-color sensor, some sources told, that Canon too, but they have no reason to launch it, because it would be more risky then to continue in well-tried megapixel war.

Nikon doesn't, nor does Fuji, have a full-color sensor. Nor does Canon. Well, they might have one being under developement in the labs, but not in production, nor officially announced. Unless you have some inside information - not likely - you can not say they have full color sensors. You can speculate, yes, but not state it as a fact. Yes, Nikon has the patent everyone likes to mention, but having a patent means nothing.

Would it not be weird, if one of the big companies had these "super sensors" ready, but they would refuse to put them in the top of the line pro-cameras because of ignorance of average-joe-consumer?

Quote:

I'm sure they will release it in the moment, when Sigma - or any other brand - will introduce full-color sensor more than competitive to current solutions.

Sigma will never release such a solution. Yes, never. They don't have the expertise or the resources. Foveon design is simply a dead end.

Also, since there is zero evidence, that for example Canon or Nikon has a full color sensor (even in the works), it is rather courageous to be sure of something like what you're are sure about above.


PostPosted: Wed Dec 02, 2009 3:50 pm    Post subject: Reply with quote
no-X wrote:
3 simple facts:

1. bayer mask is 33 years old invention

And? Combustion engine is even older, yet in use even though it has it's problems. Replacing those problems with the problems of for example hydrogen economy, would be very problematic today.

Quote:

2. not any single branch of graphics - with the only exception of photography - considers this approach as acceptable. Rendering of 1/3 of the image + following interpolation is not accepted by any other branch as quality method. And even the fact, that the same hardware resources can produce "3-times higher" resolution, isn't considered as satisfactory argument to utilize this approach...


I just don't understand what you mean by rendering 1/3 of the image + following interpolation.

What other branch of "graphics" you mean? If lossy data compression bad? Are you saying that the image quality of, let's say 3 million pixel uncompressed TIFF file is superior to image quality of moderately compressed 10 million pixel JPEG (in photogtaphy)? Think of Bayer CFA as a lossy compression method. Yes, Foveon doesn't use this compression method, but the trade off is far fewer pixels, inferior color and noise performance (essentially they lose data in a different way), more difficult to manufacture sensor and higher datarate demands per pixel.


Quote:

3. majority of manufacturers tell us, that megapixels are most important for image quality, professional work, etc...

This may be news to you, but professionals tend to use the best equipement for their needs, regardless of pixel count. It just happens to be, that more pixels tend to equal better image quality. Fewer pixels give more visible false detail.


PostPosted: Wed Dec 02, 2009 8:07 pm    Post subject: Reply with quote
Bayer mask cannot be considered as lossy compression method. There are no data lost, because the data are never created/recorded.

Anu wrote:
no-X wrote:
1. bayer mask is 33 years old invention

And?

Second-rate product, which is developed and improved constantly for >30 years can beat product based on first class idea, which is only a few yeas old and developed by a small company with very limited budget.

But that doesn't mean, that the new product is worse - that does mean, that it needs further investments and development.

Anu wrote:
I just don't understand what you mean by rendering 1/3 of the image + following interpolation.

Bayer sensor sees only 1/3 of the actual image and the 2/3 missing color values are created by interpolation.



Imagine, how easy would it be, if computer graphics would be based on this system. Computer would render only 1/3 of the image and the rest would be interpolated by simple circuit. It would safe 2/3 of hardware resources, or bring 3-times higher performance using the same resources. Why does it nobody use? Because it would look like this...




Well, computer graphics doesn't use it, because results are awful. Photography uses this approach. It is the "best", what the giants like Canon and Nikon can offer. They have no reason to work on anything new, which would be better for customers. They sell what is best for them. Their marketing departments are strong enough to convince customers, that this 30 years old system is the best thing under the Sun, and additional 2 megapixels, 137th auto-focusing point and brand new face detection system (which can recognize up-to 20 faces including your dog) are the key reasons for camera upgrade.

Anu wrote:
Second, it is not easy to make a good full color pixel sensors, not easy at all. Foveon is a good example of one solution which basicly stinks (compared to bayer sensors of the same generation) - very bad color, very noisy, low dynamic range.

One thing are color tables, another thing is how people perceive it. I posted many images on local forums and many members told me, that they feel it very realistic and asked me, how did I do it... Well, now I know, that the reasons of visual reality are bad color, high noise and low dynamic range of the Foveon sensor... Laughing

Anyway, Sigma has its typical colors. During development of DP1 Sigma tried to get closer to color charts, but all Sigma fans, who beta-tested the camera, disliked it a find the resulting colors more unrealistic, than traditional Foveon/Sigma approach offers. Sigma reworked the processing pipeline (that was the reason of DP1 delay).

Anu wrote:
I thought that Foveon sensor is occasionally advertised as having three times the pixels it actually has.

Just like bayer. All other branches of graphics consider single-channel color value as subpixel (at best), not pixel.

Anu wrote:
single color pixels require three times the data transfer speed Bayer-sensor pixels need

Again, that's cost advantage for manufacturer, not for customer. On the other hand, bayer sensor needs low-pass filter, which blurs image before it reaches the sensor itself:

Image part, which should lie within 1 pixel is blurred and covers 9 pixels...



This is necessary to avoid color artifacting on fine details and moire on textures:





We buy the most expensive optics with the highest possible resolution so the camera could blur it, because its sensor (based on 30 years old concept) cannot handle fine details without creating tons of artifacts.

Low-pass filter blurred the image. Now the sensor "photographes" 1/3 of total color value of the blurred image because of bayer-mask limitation. Fine, 2/3 of our image are ignored. After that, camera processor applies debayering interpolation (which is - in fact - based on blurring), some aditional filtering (denoise and additional reduction of artifacts) and finally terrible amount of sharpening to compensate low-pass filtering (blurring), debayering (blurring) and artifact removal (blurring).

Example of old bayer-based camera system (without low-pass filter):



A: real image
B: image taken by bayer sensor
C: image processed (debayering) by the camera - quite sharp, but full of artifacts

Example of current bayer-based camera system:



A: real image
B: what does the low-pass filter do (fine details are removed as prevention of artifacting)
C: image taken by bayer sensor
D: image processed by the camera (debayering)
E: final processing (sharpening)

One of the reasons people still accept it is they never saw real RAW - the green blurry pixelated thing...

The need of high pixel density on bayer cameras is in the multi-level blurring, which destroys all fine details. Pixel count just compensates it. Without these blurring, much lower resolution would be sufficient to capture the same level of detail.

Even the best prime lenses don't have such a huge resolution, as todays cameras. You scoff at <5MP Foveon sensor, but the only lens I have ever tried, which is able to outresolve it from edge to edge is Macro APO Lanthar. Any other lens (including many prime 50mm lenses) isn't able to offer per pixel sharpness from edge to edge. Conclusion is simple - real 4,7MP resolution of the sesor isn't as limiting, as many people think...


PostPosted: Wed Dec 02, 2009 8:28 pm    Post subject: Reply with quote
nice no x, very nice! Very Happy
tony


PostPosted: Thu Dec 03, 2009 3:01 pm    Post subject: Reply with quote
no-X wrote:
Bayer mask cannot be considered as lossy compression method. There are no data lost, because the data are never created/recorded.

It is irrelevant at which point the data loss happens - it doesn't matter if it hits the sensor or not. Bayer is basicly a lossy data compresson.
Quote:

Anu wrote:
no-X wrote:
1. bayer mask is 33 years old invention

And?

Second-rate product, which is developed and improved constantly for >30 years can beat product based on first class idea, which is only a few yeas old and developed by a small company with very limited budget.

Bayer is a pragmatic and well functioning solution to the problem of color separation - it does very nice job at this as the color filters are easy to calibrate for a good color response. Not so with silicon separation of Foveon.

Anyhow, nice to know that you admit that Bayer beats Foveon Smile

(Please note, that I am not against full color pixels, just bad implementations.)

Quote:

But that doesn't mean, that the new product is worse - that does mean, that it needs further investments and development.

There are basic unsolvable problems with using silicon as a mean to separate colors. Foveon technology is a dead end technology.

Quote:

Anu wrote:
I just don't understand what you mean by rendering 1/3 of the image + following interpolation.

Bayer sensor sees only 1/3 of the actual image and the 2/3 missing color values are created by interpolation.

Vast oversimplification. Especially since bayer sensor has about 3 times, or more the pixels compared to Foveon counterparts. Let's say 3 times: now, foveon gets some imperfect information from 3 pixels where Foveon only gets single unit of imperfect infomation. Also the filters down block 2/3rds of the signal, but less - there is plenty overlap, more on some, less on some. Canon's for example tend to allow pixels to capture plenty of "false" wavelenghts. This does improve luminance information, but of course it hampers chroma accuracy.

You seem to think that interpolation is something evil. It's not.

Quote:



Imagine, how easy would it be, if computer graphics would be based on this system. Computer would render only 1/3 of the image and the rest would be interpolated by simple circuit. It would safe 2/3 of hardware resources, or bring 3-times higher performance using the same resources. Why does it nobody use? Because it would look like this...




Quote:

Well, computer graphics doesn't use it, because results are awful.

Usually it is best to use the right tool for the job. At this time of history Bayer-based technology is the best solution for cameras. Maybe in future there will be prismatic color separation, or somehing, but Foveon's way of using silicon separation is not too good.

Anyhow, Canon 7D has 18 millon pixels. Just bin 4 pixels into superpixels and you'll have 4.5 million Foveon pixels. Yet the image quality will be both superior to what Foveon will achieve (less noise, higher DR, more accurate color), and at the same time far inferior to anything even a relatively modest interpolation algorithm will achieve.

Anyhow, nice image to show that Fofeon produces good pixel level sharpness, but you forgot that Bayers have far more pixels. You should have used smaller pixels in the bayer-versions. It is unfortunately a very common mistake of comparing sensors of different pixel counts by zooming in to 100% view. This is not intellectually honest.

Quote:

Photography uses this approach. It is the "best", what the giants like Canon and Nikon can offer. They have no reason to work on anything new, which would be better for customers. They sell what is best for

Really. Silly me, I thought that Nikon, Canon and others are in competition with each other and especially the professionals like to use the best equipment available. And the margins are by far the highest with those products.

Quote:

them. Their marketing departments are strong enough to convince customers, that this 30 years old system is the best thing under the Sun, and additional 2 megapixels, 137th auto-focusing point and brand new face detection system (which can recognize up-to 20 faces including your dog) are the key reasons for camera upgrade.

You're being a bit naive here, sorry: marketing men do have some influence, but they do not run the engineering departments. They do not have the power to stop far superior technologies just because they'd find it easier to advertise the old ones.

Quote:

Anu wrote:
Second, it is not easy to make a good full color pixel sensors, not easy at all. Foveon is a good example of one solution which basicly stinks (compared to bayer sensors of the same generation) - very bad color, very noisy, low dynamic range.

One thing are color tables, another thing is how people perceive it. I
posted many images on local forums and many members told me, that they feel it very realistic and asked me, how did I do it... Well, now I know, that the reasons of visual reality are bad color, high noise and low dynamic range of the Foveon sensor... Laughing

Or maybe they just saw post processing? It is a scientific fact that the color response of Foveon style silicon based color separation leads to massive problems with color accuracy and massive need to process the signal the sensor sees to make it even roughly similar to the signal our eyes think is natural. This processing is not only inefficient, but also increases noise. Signal to noise ratio of Foveon is lousy. Push the shadows a bit and you'll see (though, you'll probably see what's left after strong noise reduction), or try high ISO. On the other hand, Bayer filters give rather accurate color without need for strong color manipulation.

Please humor me and take a picture of a colourful subject (preferrably a color target), then another with it turned upside down. In raw. Make them available for dowload (the raws). (I hope there are free RAW-converters, or that LR understands Sigma raw files.)

Quote:

Anyway, Sigma has its typical colors. During development of DP1 Sigma tried to get closer to color charts, but all Sigma fans, who beta-tested the camera, disliked it a find the resulting colors more unrealistic, than traditional Foveon/Sigma approach offers. Sigma reworked the processing pipeline (that was the reason of DP1 delay).

Foveon sensor produces by far the worst colors of all current SLR-sensors. It is results of post processing that produces the colors Sigma fans see.
Quote:

Anu wrote:
I thought that Foveon sensor is occasionally advertised as having three times the pixels it actually has.

Just like bayer. All other branches of graphics consider single-channel color value as subpixel (at best), not pixel.

It is irrelevant what other branches of graphics do. Bayer sensors require interpolation - it is a natural feature of them. You however seem to insist that there is something natually evil and unfair about interpolation and thus this required step of processing should be ignored or something like that. Yet you're perfectly happy to accept massive Foveon color processing.
Quote:

Anu wrote:
single color pixels require three times the data transfer speed Bayer-sensor pixels need

Again, that's cost advantage for manufacturer, not for customer. On the other hand, bayer sensor needs low-pass filter, which blurs image before it reaches the sensor itself:

I don't really need all these pictures. I know what AA filter is. And I do also know that diagonal lines in Foveon tend to create massive staircasing - false detail. Once you see this fault in an image, it kind of ruins it.

Quote:

Image part, which should lie within 1 pixel is blurred and covers 9 pixels...

You like staircases?

Quote:

One of the reasons people still accept it is they never saw real RAW - the green blurry pixelated thing...

Just have a look at Foveon raw and you'll be surprised to learn that the color quality is absolutely horrible and the image is filled with noise.

Quote:

The need of high pixel density on bayer cameras is in the multi-level blurring, which destroys all fine details. Pixel count just compensates it. Without these blurring, much lower resolution would be sufficient to capture the same level of detail.

AA-filter does minor blurring, yes. It does not do some awful overblurring that destroys all the fine detail. Absolute nonsense. Pixel count does indeed compensate - and as you know, there are way more pixels in Bayer sensors.

Low pixel count without AA-filter leads to (suprise surprise) aliasing, not detail, but false detail.

Quote:

Even the best prime lenses don't have such a huge resolution, as todays cameras. You scoff at <5MP Foveon sensor, but the only lens I have ever

Today's sensors are not even close the resolving power of todays best lenses. All but two of my lenses are far beyond the ability of my camera to resolve detail.

When the lenses no longer can match the sensors, the manufacturers will throw away AA-filters.

Quote:

tried, which is able to outresolve it from edge to edge is Macro APO Lanthar. Any other lens (including many prime 50mm lenses) isn't able to

Then you have either faulty lenses or bad technique. Or it is a problem of Foveon sensor: because of the photo diode layout there are optical problems, that can not be solved easily, if at all. Let me quote a paper by the folks behind Foveon: "First, there is a potential with stacked
photodiodes to have the sensitivity vary with f-number because lower-angle rays might escape detection by escaping
from the side of the diode.". And "the presence of low-angle rays might shift the color response due to the variation in ray path length
through the silicon" and "the quality of the optics used with the stacked-photodiode
sensors can seriously impact image quality". There are plenty of interesting papers of different sensors and sensor technologies in the interwebs, you know, the series of tubes, one can read Wink

Anyhow, since the photodiodes are stacked and need certain amount of vertical space for color separtation, it is either impossible or very difficult and expensive to solve the low-angle ray problems.

An APS-C sensor needs something in the neighbourhood of 100-200 million pixels to reach the limits of today's glass. And my telelenses (but one mirror lens), do give absolute pixel level sharpness wide open, corner to corner (APS-C).

Quote:

offer per pixel sharpness from edge to edge. Conclusion is simple - real 4,7MP resolution of the sesor isn't as limiting, as many people think...

It depends on what you use your camera for. For web images it's more than enough. For professional imaging it tends not to be. The larger the output size and the more post processing (like rotating) is done, the less worth those 4.7M pixels are.

A couple of things:

Foveon produces lots of false detail - artifacts. A low pixel count makes these far more visible that the artifacts of a high resolution bayer were.

Foveon has lots of noise due to the nature of the sensor. Shadows are horrible, as are high-ISO images. Noise reduces DR. Colors need to ne processed to look proper (and still don't at higher ISOs), this increases noise.

Even though high-ISO might be beyond the scope of this argument, would you plese give your view on the Foveon performance in photon limited situations (shadows and high ISO).

Regarding the images you posted, if you want to compare Foveon output and Bayer outout, you might want to change the pixel sizes - give bayer sensors 3-4 times the pixels as that's what the situation is in real life.

You failed to explain me of your knowledge of the major players having "Foveon" chips of their own ready and just waiting to be used in case they are needed.
Post new topic   Reply to topic
Home   Digital SLR Cameras
Goto page Previous  1, 2
Page 2 of 2