Talk:Sigma SD14

Sourcing

This article needs some serious sourcing. It also needs removal of the Sigma-fanboy perspective that 4.3 megapixels of true 3-color pixels equate to 14 megapixels. It does not. There is no logical way that that makes sense.

Yes, there is a discrepancy between a 3-color sensor pixel on a Foveon chip and a 1-color sensor pixel on a Bayer-matrix chip. This discrepancy means, yes, that comparing the two is hard, but embracing advertising inaccuracy doesn't help make a good Wikipedia article. Matthew Brown (Morven) (T:C) 22:34, 28 September 2006 (UTC)

The 4.7 megapixel figure was edited out by Dicklyon, and my not putting that back in was an oversight. However, I don't agree that the language in the article represents a "Sigma-fanboy perspective". I made a deliberate effort to use language that was similar to the language in existing wikipedia articles on the SD10 and the Foveon X3. It is an indisputable fact that Sigma and Foveon both claim that the sensor used in the SD14 is a 14 megapixel sensor, and the logical basis for that claim is that they are counting each RGB element, just like for every other digital camera sensor in the marketplace. If the figure of 14 MP is misleading (without qualification) than so is the figure of 4.7 (again, without qualification). So I say, put both figures in there and qualify them with an explanation on where they both come from. Furthermore, I would question if your perspective is neutral, given that you took the total number of pixels in the RAW file, and divided by 1024^2 to arrive at an even more misleading figure for the number of megapixels than 4.7. RanmaSaotome 03:05, 29 September 2006 (UTC)

That's just my being used to megabytes and gigabytes, and without thinking using the binary thousands. No bias involved. I've several times had to edit the SD10 and Foveon articles for bias, so simply using those articles as a beginning isn't necessarily perfect, but ...

Anyway, it would be nice to get sources for all of this - without that, it just reads like the opinions of Wikipedians. And don't think I'm against the SD14 or Foveon; I'd love one, though I'm not rich enough to afford one right now and I'm not sure I want to invest in a lens system only usable on Sigma cameras. Matthew Brown (Morven) (T:C) 04:37, 29 September 2006 (UTC)

I gave references for both Sigma and Fujifilm claims of megapixel counts. Matthew Brown (Morven) (T:C) 05:19, 29 September 2006 (UTC)

I recommend adding {{fact}} tags where you think things need a citation. Then, if no citation is forthcoming, you will be justified in removing or rephrasing it. Dicklyon 06:26, 29 September 2006 (UTC)

I think what bothers me about Sigma/Foveon claiming each color sensor at the same location is a separate pixel is that with the other systems (Bayer, Fuji) there is at least a different geographical location for each claimed "pixel", even if it does not have full information at that location. It seems to me that Sigma is actually getting 4.7 full megapixels, and that the other technologies are getting partial pixels - it would make more sense to discount other manufacturer's pixels by a factor than to multiply the Foveon's.

Be that as it may, the most important thing is that this article (a) be clear about what is actually provided in terms of sensor elements and output pixels in different formats, and (b) accurately cite every claim. It's certainly become much better, and I'm pretty sure we can get it most if not all of the way there. Matthew Brown (Morven) (T:C) 18:53, 29 September 2006 (UTC)

Heh, I wish I could afford one also. If this process leads to the best article, so much the better! RanmaSaotome 20:34, 29 September 2006 (UTC)

RAW File Size

I think the wording in the paragraph with the RAW file size is a little bit awkward, specifically the "2640 × 1760 × 3 samples" phrasing, which is inconsistent with the phrasing used for the JPEG file size and the SD10 RAW file size, both of which are quoted in terms of megapixels. If we changed those phrasings to match, then the whole paragraph would be horribly awkward, so I favor changing it back to "4.6 megapixels, or 2640 x 1760 pixels". I think that once sample photos are available, and reviews appear, we will be able to firmly establish how the per pixel IQ compares between the SD14 and other DSLRs. RanmaSaotome 21:13, 29 September 2006 (UTC)

Yes, it can get awkward. But the important thing is to be clear and correct. With JPEG files there's no difficulty in comparing "pixels" since they all contain RGB pixels. With sensors, the meaning of the term "pixel" is often debated, so we can report what is there, and what Sigma calls it. In the case of the raw file, same things, except that as far as I know Sigma doesn't call it anything. What's important is to report the full truth, which is that the SD14 raw file has 14 million measurements stored in it, in a different organization from Bayer raw files; choose a neutral terminology to describe it so that the debate about the meaning of pixel is not dragged into it. I don't see where the SD10 raw file size quoted in terms of megapixels; if it is, perhaps it should be fixed. Dicklyon 21:35, 29 September 2006 (UTC)

The thing is the only purpose of the RAW file is to produce a TIFF or a JPEG, both of which have a clear definition of pixel. If the RAW conversion software works like the previous version, the user will have the option of half size, 1320 x 880 (1.2 megapixels), full size, 2640 x 1760 (4.6 megapixels) or double size, 5280 x 3520 (18.6 megapixels). It has been demonstrated with the SD10 that the double size output is no better than taking the full size (3.4 megapixel in this case) output and then upscaling in Photoshop. (On the other hand, it is clearly better than taking the 3.4 megapixel output of a Bayer-sensor camera and upscaling to the same double size). On the other hand, we know that the SD14 will be able to produce a 14 megapixel JPEG file in-camera, but so far we know nothing about the quality of that file. I would say that unless and until we can source some evidence that the RAW conversion software for the SD14 can take the RAW file and output a TIFF with greater than 4.6 megapixels, and that TIFF is actually better quality than what you would get if you upscaled the 4.6 megapixel output in Photoshop; we should call the RAW file 4.6 megapixels. RanmaSaotome 13:01, 30 September 2006 (UTC)

Your logic makes no sense; there is no definition of pixel counting that depends on subjective image quality. We just need to describe what's in the raw file, since the JPEG that can result from processing can be any size. Dicklyon 16:49, 30 September 2006 (UTC)

I found where you referred to about the SD10 raw file being called 3.4 MP, and removed the ambiguous pixel count language from that one, too, replacing it with 10.2 M samples. Dicklyon 17:40, 30 September 2006 (UTC)

RanmaSaotome, your new edit applies a presumed definition of "pixel" as it applies to raw files. Since no such definition is in evidence, and since it has no parallel in Bayer sensor raw files, it represents only a opinion of an interpretation, it seems to me. I'm going to revert it for now. Let us know if there's a source that verifies this interpretation. I know there's an international ISO standard that directly contradicts it, and I can dig that up if you like. Dicklyon 20:29, 30 September 2006 (UTC)

Once you have the file in the PC, you are in a realm where "pixel" is solidly defined, and unless you are prepared to say that in TIFF file you need to count each sub-pixel as a pixel, then the X3F file has 4.6 megapixels. Wikipedia is not a place for blindly accepting any companies marketing-speak, so I will revert any attempt you make to muddle the straight forward wording, unless you are prepaired to offer a bona fide improvement. Oh, and RAW is usually capitalized in this context, even though it is not an acronym. Good job on the information table, though, and on the clarification to the Super CCD reference.RanmaSaotome 20:34, 30 September 2006 (UTC)

Once you have an "image" file I might agree. But a "raw" file is different. A Bayer raw file does not contain ANY RGB pixels, so they are still difficult to compare. Take a look at ISO 12231 Photography -- Electronic still picture imaging -- Vocabulary, 2005-02-01, which says the digitized photoelement values are pixels: "2.1 addressable photoelements — number of active photoelements on an image sensor. This is equal to the number of active lines of photoelements times the number of active photoelements per line. NOTE The term resolution should not be used when referring to the number of addressable photoelements on an image sensor. It is possible that the number of addressable photoelements may be different for the different colour records of an image. When the signal values of the photoelements are digitized, the digitized code values may be referred to as picture elements, or pixels." I have no problem with people having dissenting opinions to this, but if we use the term "pixel" where its meaning is ambiguous, the article will become a debate forum, and that what we should seek to avoid. So we need to use a more neutral terminology that is clear and explicit without taking sides on the pixel debate. Right? Dicklyon 20:43, 30 September 2006 (UTC)

That is a very sketchy semantic argument, and it falls down because while the RAW file data isn't in terms of R, G, and B, (note that many image formats are in CMYK or LAB or some other color space) anyway, while the RAW file data isn't in terms of R, G, and B, it is in a form that is one simple mathamatical transfomation away from RGB. Calling each location here a pixel is on solid rational ground, and I argue that the way I wrote that paragraph is clear, concise, honest, and in a language that most people can understand.RanmaSaotome 20:57, 30 September 2006 (UTC)

I had read it before reverting, and found it contained not only the presumed pixel definitions but also two expressed opinions about quality. So this time instead of reverting I edited all that. It's a bit awkward, and can use some improvement, but using terminology that takes sides on the semantic issue will not stand. What you wrote may have been rational, clear, concise, and honest, but was not WP:NPOV nor WP:V. I have cited a source; if you find one the supports the other interpretation in camera raw files, and feel compelled to rely on it in the article, then we'll have to also put the other side in. Is that where we want to go? Why not just use a more neutral terminology? Dicklyon 21:04, 30 September 2006 (UTC)

Good call on the POV statement, “contrasts with” reads better than “is superior than” anyway. Nevertheless, the source you have quoted from regarding pixels in RAW files is not available for free, so I cannot evaluate the context of your excerpt. I did, however find this source: [1] which is a specification for the X3F file format, and from that there is this excerpt: “To decompress the pixel data, use the following algorithm for each row of each color channel, where Ci is the unsigned 8-bit value of a color channel in the ith pixel on that row and Di is the unsigned 8-bit decompressed value of that color channel for that pixel”. That seems to imply to me anyway that a pixel may have more than one color channel in an X3F file, and it proves that Foveon considers X3F files to have pixels.RanmaSaotome 22:50, 30 September 2006 (UTC)

That document does not say how to decode the raw data. The passage you found refers to reading the 8-bit RGB preview image, not the raw data. Sorry about the ISO standard; it's been a while since I've seen a copy, too. The bit I quoted was the entirety of the section 2.1, addressable photoelements; no other section attempted to define pixel. Dicklyon 23:12, 30 September 2006 (UTC)

The pixel debates

It would be best to avoid the debates common in the forums about what should be counted as a "pixel" in cameras and sensors. User:RanmaSaotome has added the words "pixel count of 4.7 megapixels" which very clearly is not WP:NPOV, by taking one side on this issue, just as calling it "14.1 megapixels" would take the other side. We all have our opinions, but wikipedia is not the place for them. Instead, we just need to be clear about what is there, and also report what Sigma calls it (they call it 14 MP, in case anyone was wondering). I'd prefer that someone else take a crack at the wording... Dicklyon 21:42, 29 September 2006 (UTC)

When I noticed the change in the article, I wanted some precise information. I went to the megapixel (redirected to pixel) article and haven't found this. So I rewrote the section, with cited references. And lost one hour of work by accidentally closing the window. Ouch. I will redo it. --Marc Lacoste 22:25, 29 September 2006 (UTC)

So far here I have been accused of having a "Sigma-fanboy perspective" and now you are accusing me of the opposite? It is not appropriate for any "side" to be taken here, the proper thing to do is report both the 4.7 and 14 megapixel figures and then explain in the article where they come from and why. This is what I have attempted to do. Sigma, of course, has an interest in maintaining a consistent marketing message that the SD14 is a 14 megapixel camera, but the mission of Wikipedia is orthogonal to that. RanmaSaotome 13:01, 30 September 2006 (UTC)

The only thing I'm accusing you of is expressing one of two opposing points of view on the pixel count. It's not about you, but about how to describe the camera in the article. We agree on what needs to be done. Dicklyon 16:47, 30 September 2006 (UTC)

Actual sensor counts

See sensor data. Total Pixels 14.5M, 2688 x 1792 x 3 -- Effective Pixels 14.1M, 2652 x 1768 x 3 -- for definition of effective pixels see DPReview. So it's 4.7 M x 3, not 4.6 M x 3, at least at the "effective" RAW data level. The total raw data probably is more than that, possibly the full "total" 14.5 M, but that's not the preferred number to mention. Dicklyon 21:38, 30 September 2006 (UTC)

The specifications on the DPReview site say that effective pixels are 2640 x 1760, which is clearly 4.6 million, but obviously the Foveon reference contradicts that. We shall see what the X3F file has when sample images are released.RanmaSaotome 22:14, 30 September 2006 (UTC)

The sensor has 2652 × 1768 active elements, and a few more in the borders for dark level calculation. This is visible using dcraw in "dark frame dump mode", for example. Sigma Photo Pro only outputs 2640 x 1760, but that's a software choice, not a hardware fact - Rnbc 11:27, 29 October 2007 (UTC)

Moot Disccussion--well, almost. . .

---

Well, the points in dispute are not moot yet, but they may be soon. When this camera does hit the market in massive numbers and is subjected to truly strong critical scrutiny, including but not limited to resolution tests, then this technical debate about how to count pixels will probably begin to recede into the past. We are already getting to the point that "pixel count" becomes a bit more irrelevant every year in terms of determining image quality. The jury is still out on this one, but it will be in soon. I now shoot primarily full-frame CMOS (after starting out with CCD-based digital cameras), but I can see the day coming when the use of Foveon sensors will simply be a fact of life across many camera brands. The dispute over pixel counts has been a complicated and difficult technical issue for lay persons such as myself to understand, but even we who are not technophiles do understand picture quality and overall resolution--and judgments about those qualities will soon be made for this camera and its sensor. I look forward to Phil Askey's reviews at dpreview.com, among many others, very soon now. Above all, I look forward to trying a camera with the Foveon sensor in it for the very first time, so that I can see for myself.Landrumkelly 02:22, 2 January 2007 (UTC)

Sigma Fanboys

I agree the presentation should be as neutral as possible. However, there is an agreement among almost all Sigma users why they prefer a RGB pixel stacked sensor based cameras image over that of a Bayer sensor based camera. I feel there should be a section where those opinions should be listed (particulary color-richtum, pixel sharpness, 3D look, lack of soft noise effects causing 3D look). Unfortunately the SD14 is still not very useable for action photography, and to my experience limited action can be rather slow on a random base. While the low light capabilities of the SD14 are improved about one to two stop over the SD10, the latest DSLRs should be also much better than the SD14 for extreme poor (low) light situations.

More or less there is no perfect camera (cheap, fast, DR, best IQ...) so the Sigma fanboys would agree the SD14 is best used for still photography and when image quality matters.

Regarding numbers of pixels - the common indication as a resolution indicator is unfortunately wrong IMO. The SD14 has true 4.7 MPixel full color pixel resolution, and a 12 MPixel Bayer sensor camera has a resolution of 4.7-5.2(well, thats a guess) full RGB pixel sensor. On the other hand one can say the SD14 has a resolution of a 10-12 Mpixel Bayer camera. For real prints there are good experiences with surprisingly big formats. —Preceding unsigned comment added by Spalbird (talk • contribs) 23:27, 3 July 2008 (UTC)

The article is a Mess!!!

Someone can clean up this mess? Format the article at least... otherwise i'll just undo the last changes to the article nex time I come up here. Rnbc (talk) 23:51, 21 July 2008 (UTC)

I cleaned it up a little bit, but this article needs more than formatting help. There's odd phraseology here and there that makes it seem... we'll, I can't put my finger on it. 63.230.167.170 (talk) 03:20, 22 July 2008 (UTC)

The second half of the opening paragraph reads like advertising. Without references it doesn't read like an encyclopedia entry... EyeExplore (talk) 04:09, 9 April 2010 (UTC)

I've cleaned the article a bit... It still looks like a mess, but less like advertising. Rnbc (talk) 00:40, 18 February 2011 (UTC)