Go Back   Steve's Digicams Forums > Digital SLR and Interchangeable Lens Cameras > Sigma dSLR

Closed Thread
 
Thread Tools Search this Thread
Old Sep 1, 2004, 3:58 PM   #21
Administrator
 
Join Date: Jun 2003
Location: Savannah, GA (USA)
Posts: 22,378
Default

cigalaix wrote:
Quote:
Thank for your responses, Lin et JimC.

My english is not good : the Google's translationis better, but sometimes approximative.
Quote:
TRANSLATED BY GOOGLE:
Quote:
My matter was provocative, but it is well the object of a debate. In practice, I currently prefer my EOS 500 associated with a coolscan with Nikon for the good images. On the ground, I work in double with small numerical (currently Ixus 4 méga in spite of its strong distortion evaluated to 5 %) for a rapid result. I am interested by the acquisition of a numerical reflex camera, but I was not convinced by the performances observed. I specify that an image is judged with its personal sensitivity, which is more significant than theoretical technical measurements. The same kind of debate also exists with films argentques and optics. I was disappointed by the photograph obtained with EOS 10D (too smoothed image), and I thus refused to make the purchase of it although I would be equipped in objectives EOS. It appears that EOS 1D mark II is surpenant, but its price (4500 euros the case) and its obstruction (1,2 kg) is too heavy. I could test Nikon D2H; the results are rather satisfactory, without more, with a definition of image neighbors of the SD 10 but the price (3600 euros the case) and the obstruction (1 kg) are also too heavy. On the other hand, I was impressed by some photograph obtained with the SD9 - SD10 for their piqué (I do not know the English term corresponding) and their realism. It is those which gave me the best feelings. Moreover, the Sigma cases have a reputation of robustness even if their completion appears summary. To follow...
It is not unusual for an image to appear "smoothed" (or soft) directly from a camera. Canon uses conservative algorithms for things like contrast and sharpness to maintain as much real detail as possible. Many users prefer to shoot in RAW, then process the images using software later for results matching their individual preferences (for sharpening, contrast, saturation, etc.).

TheSigma SD9/SD10 models must shoot in RAW all the time (using software to process the images later). Steve processed the sample images from the Sigma SD10 with Photo Pro version 2.0, sharpened +1.0, then saved as JPEG quality 12.

The way images are processed (either by the camera, or using software later), can have a big impact on perceived quality (which will vary depending on individual tastes).
JimC is offline  
Old Sep 3, 2004, 3:58 PM   #22
Member
 
Join Date: Aug 2004
Posts: 34
Default

Thanks for taking the time to translate, JimC. I studied French for one year at school, but it seems to be
some time ago. :shock:

cigalaix wrote:

Quote:
I specify that an image is judged with its personal sensitivity, which is more significant than theoretical technical measurements.

I agree. The amount of processing involved with ordinary sensors, with it's associated compromises explains
most of the differences between digicams these days. No measurements yet is able to explain those differences objectively.

I made a reference to audio reproduction earlier, and the same applies in that field, despite the fact that
the audio industry is almost as old as my house. (100 Y's+). Think of it. Have you ever come across an
industry expert, or a set of measurements that fully define the percieved sound of a set of loudspeakers?

Of course not.






code is offline  
Old Sep 19, 2004, 8:06 AM   #23
Member
 
Join Date: Sep 2004
Posts: 66
Default

Hi lin,

It's not an exaggeration.

Foveon is exactly right when they say, "when in the fact the SD9 is delivering image resolution and sharpnessthat is outstanding in the DSLR category of 6 to 14 MP."

Do they know something thatothers don't? Yes. Foveon istalking about full color perofrmance, above, while all the main stream reviews only measure B&W performance.

When imaging a B&W target, a Bayer sensor's performance is greatly exaggerated since all the sensors (R,G,B) can sense black (no reading), so every one is equally useful. But for a color image, each sensor is limited to sensing its color only.

The end result is that for a two color B&W target, a 10.3MP Foveon chip will only sense about the same amount of detail as a 6MP Bayer, because a Foveon sensor always 'dumbly' wastes three sensors to sense black, or that no light is present. If red and blue is involved, the Foveon will sense the amount of same detail as a 14MP Bayer since it has the same number of red and blue sensors as a 14MP Bayer. This is easily illustrated using color tests, and full color results are easy to find on the web where the SD9 very solidly out-performs the 11MP Canon 1Ds in red and blue resolution.

So, both mathematically and empirically, the SD9 is roughly a:
  1. 6.86MP Bayer for two color B&W[/*]
  2. 13.72MP Bayer for Blue and Red[/*]
  3. 6.86MP-13.72MP Bayer for everything in between
[/*]
Not suprisingly, it is a 10.29MP camera--the number of sensors it has. Bayer cameras change with color since they have varying response to different primaries. That is why three bullets are needed above instead of just one: 10.29 MP.


SigmaSD9 is offline  
Old Sep 19, 2004, 8:12 AM   #24
Member
 
Join Date: Sep 2004
Posts: 66
Default

Hi Lin,

You might want to read my reply here:

http://www.stevesforums.com/forums/view_topic.php?id=33393&forum_id=79&jump_t o=178405

The SD9 is a true 10.29MP camera, not a 6MP Bayer, unless you image atwo color B&W target.
SigmaSD9 is offline  
Old Sep 19, 2004, 11:08 AM   #25
Moderator
 
Join Date: Jun 2002
Posts: 1,139
Default

Quote:
This is easily illustrated using color tests, and full color results are easy to find on the web where the SD9 very solidly out-performs the 11MP Canon 1Ds in red and blue resolution.
This argument makes absolutely zero sense. Why don't you post two identical images made at the same focal length of the same subject where the SD9 "out-performs" the 11MP Canon 1DS" :-)

You're talking to someone who uses both these cameras and I've not found even one case where the SD9 "solidly" out performs even my 10D - it doesn't even come close in resolution to my 1DS. I've read your arguments about this on Phil's forum for a long time, and they don't hold up any better here than they did there.

Just look at Steve's own images of the brick building (red) with lots of (blue) sky, etc., and show me exactly where the SD9/SD10 "solidly" out-performs the 1DS or even remotely comes close to the 1DS.

Let's just enjoy the great images from the SD9/SD10 without off-the-wall exaggerations which make us look foolish to the rest of the world.

Lin
Lin Evans is offline  
Old Sep 20, 2004, 1:11 AM   #26
Member
 
Join Date: Sep 2004
Posts: 66
Default

It makes perfect sense, which is why Foveon says it. As for the samples, see

http://www.outbackphoto.com/artofraw/raw_05/essay.html

The 1Ds gets clobbered by the SD9 in red/blue whenusing the same lens. Yes, you can completley give up the$8000 full frame advantage and use a much stronger zoom than the SD9, to equalize the FOV, but it still won't be quite enough to get from 550 lines red and blue to the SD9's1500. The 1Ds only has 2.75MP in Red and Blue, 25% less than the SD9. It takes a 13.72MP Bayer to match the SD9 in red/blue (the Kodak only has 13.5MP)

Like I said, the SD9 is only 10.29MP, so the 11MP 1Ds should outperform it ever so slightly accross the board all else being equal. And btw its not really equal, the 1Ds always uses blurrier full frame glass toward the corners even when the imaged FOV is the same, it employs a blur filter to quell rainbowing, and it's noise is pretty high by modern standards even at ISO 100.

Anyway, to say the SD9 is a 6MP Bayer and that is that, is incorrect. That is for two color B&W only. The world is in color. For color images, the SD9 is precisely a 10.29MP camera.

Foveon is exactly right when they say that it is an outstanding6 to 14MP Bayer equivalent, those are the worst case/best case boundaries since Bayer resolution changes with color. The "avg"full colorimage will fall right at 10.3MP. Shouldn't be a surprise to anyone, since that is the sensor count and digital needs sensors to see.
SigmaSD9 is offline  
Old Sep 20, 2004, 1:20 AM   #27
Member
 
Join Date: Sep 2004
Posts: 66
Default

Here is another 10D vs SD9 test with FOV equalized, the SD9 would perform better using the same lens:

http://photocenter.smugmug.com/gallery/147356

You can measure the results yourself using the test images provided. As you can see,the SD9 has:

Black on White - 106% of the 10D
Black on Red - 127% of the 10D
Black on Blue - 115% of the 10D
Black on Green - 106% of the 10D



Almost exactly the same as dpreview.com's B&W results and outbackphoto.com's color results. The note there is worth noting, it takes about twice the MPs to increase resolution by 25%. Here is the SD9's MP count relative to the 10D:

Overall - 164% of the 10D (= 10.3MP Bayer)
Red - 217% of the 10D (= 13.72 MP Bayer)
Blue - 217% of the 10D (=13.72MP Bayer)
Green - 109% of the 10D (= 6.86MP Bayer)


SigmaSD9 is offline  
Old Sep 20, 2004, 1:50 AM   #28
Moderator
 
Join Date: Jun 2002
Posts: 1,139
Default

Your math makes no sense. Even if the information provided was accurate (it's not), the difference would be the average of the improvements not their sums. Elementary statistics 101 ....

Lin


Here is the actual quote from resolution measurements from the SD9 compared to various six megapixel bayer cameras:




and here are the Canon 1DS figures also from dPReview:



Real world tests on thousands of images bear out the facts, not the "fantasy" world you've created. The SD9 and SD10 come "close" to six megapixel bayer resolution but are not even in the same universe with the 1DS or 14 megapixel Kodaks. You keep talking about "black and white" but resolution charts photographed with any camera are quite representative of what those cameras do in "real life." It's neither necessary nor desirable to "create" some specialized chart to attempt to elevate the SD9/SD10 to resolution levels they can't obtain. Detail is detail - the camera either "can" resolve it or it can't. If the SD9/SD10 can't resolve it on a standardized resolution chart, they can't resolve it in real life which is a fact borne out by practical real world experience with all these cameras.

False detail beyond Nyquist is sometimes useful as in landscapes where it serves much the same function as "marker pixels," but it's not "real detail" and that fact is immediately apparent when enlarging to the point where this false detail becomes visible to the naked eye. The bad stairstep aliasing is readily apparent in many shots on the samples in your link. Look at the end of the cat's whiskers and they look like twisted ropes even at 100% This only gets worse as you enlarge and there is little you can do to "fix" it.

The SD9/SD10 provide wonderfully clean,sharp images and have tremendous resolution for the matrix size of the file, but they don't out-resolve even six megapixel bayer cameras and not in your wildest dreams do they even remotely approach the11 megapixel EOS-1DS or 14 megapixel SLR/n. To continue to try and make this foolish case regardless of whether it'syou or Foveon is only to invite criticism from those who have both cameras and use them professionally.

This "color resolution" versus "black & white" resolution issue has been thoroughly debunked by numerous physicists.A black and white resolution chart is photographed in "color" justas a color resolution chart would be photographed were there such a thing. Resolution is and always has been defined in photography by either line pairs per mm or lines per image height. If you want tocontinue to believe that your SD9 out resolves a Canon 1DS, then nothing is going to change your mind, but you're certainly not going to convince anyone who uses both camera professionally becausethe facts speak for themselves. Sorry.....

Lin
Lin Evans is offline  
Old Sep 20, 2004, 7:51 AM   #29
Member
 
Join Date: Sep 2004
Posts: 66
Default

You quoted a 2 color B&W test and it matches my numbers exactly. So I'm not sure what you are trying to show there, sorry.

Are you saying that a Bayer can sense true color detail without enough sensors to support its interpolated recorded output? For example, that the Canon 10D can ouput 6M optically controlledfull color pixels, with only 6M monochrome sensors? I assure you it cannot. Nor can it come anywhere close.

It can, however, sense a full 6Mblack pixels with them (no light). It cannot sense white at each location, but nevertheless a two color B&W chart is still its best case by a huge margin. As for summing the full color results rather than taking the SD9's best case of 13.7MP, I'm all for it. That's why one should callthe SD9 a 10.3MP camera and not a 14MP camera, or a 6.8 MP camera.

Or alternatively, go ahead and call the SD9 a 3.43MP camera, and the 10D then correctly becomes a 2.10MP camera instead of a 6.30MP camera.

This is all made way more complicated than it needs to be, mostly because of the Bayer defacto standard of calling any "X"MP-monochrome-interpolated camera, an "X"-MP optical full color camera. Obviously, that can never be the case. A digital image requires 3 optical sensors per recorded ouput pixel (RGB), anything less is interpolated to some degree. Computer interpolation, done after the shutter closes,never has any effect or improvement whatsoever on optical resolution. Never.
SigmaSD9 is offline  
Old Sep 21, 2004, 3:44 PM   #30
Member
 
Join Date: Aug 2004
Posts: 34
Default

Thanks, SigmaSd9 for your clarifications! Lin, you accused me of not understanding the processes involved in digital photography, but I'm beginning to suspect that you don't either.

I write programs for the Windows platform for fun, and love algorithms, especially when it gets as tough as solvingproblems built into old time sensors. (Bayer et al).

To be honest, I hope they die. Zoom into a few tens of pixels each side and see for yourself what they use them for. You can literally see the guesswork involved. :lol:

code is offline  
 
Closed Thread


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 9:30 PM.