Talk:Foveon X3 sensor

From Wikipedia, the free encyclopedia

Contents 1 Advantages & Disadvantages? Diagrams? 2 Don't understand the complaint 3 The Nature of the Dispute 4 Sources 5 Concluding Sentence Issues 6 Article Reorganization 7 Claims of controversy 8 Modified lead 9 Light gathering and low-light performance 10 The Foveon X3 shows monochrome moiré patterns 11 Polaroid X530

[edit] Advantages & Disadvantages? Diagrams?

Diagrams of the sensor would really help this article, and listing the major advantages & disadvantages as bullet points might also clean it up.

Dicklyon, do you think Foveon would have an objection to the use of its diagrams in this article? Who should be contacted on this issue? Anoneditor 00:32, 11 January 2007 (UTC)

Take a look at their press downloads page. I doubt that they could give a wikipedia-compatible license for their trademarks, but they could probably be used as fair use for identification. You could write their marketing guy, richard.turner at foveon, and see what he says. Dicklyon 01:12, 11 January 2007 (UTC)

Dicklyon, I thought about your advice of a contact person but have also thought a different kind of diagram might be as useful and not run afoul of intellectual property rights. Do you think this diagram is an accurate representation of the essence of the sensor? If not, what should be changed?

—The preceding unsigned comment was added by Anoneditor (talk • contribs) 23:52, 19 January 2007 (UTC).

I think you've got a good idea there. The scale's not accurate, and the discreteness of it is a bit misleading. It could be fixed up by making a continuous version, where each wavelength is attentuated at its own rate as it penetrates the silicon. The trick is showing what's absorbed, and hence sensed, in each layer. I might try to find time to work on it... Dicklyon 00:04, 20 January 2007 (UTC)

Is this any better, or should I give up? I added a stylized version of the color sensing wells. I noticed that the lateral dimension of the wells get wider as the frequency of the light decreases. Does this mean that the light spreads out in some inverse proportion to it's frequency as it diffuses through the silicon? If so, I guess that could be shown by spatially broadening the spectrum as it goes through the silicon. One of the artifacts of this kind of graphical representation is that it makes it look as if the the light color gathered by each sensor layer is reduced in its lateral dimensions. But I don't know how that can be fixed and keep the same representational scheme. Anoneditor 18:57, 22 January 2007 (UTC)

I've something half-baked in matlab that might be a good starting place. That different-sized-wells thing is long obsolete; see Foveon's more recent patents. Dicklyon 21:44, 22 January 2007 (UTC)

"matlab"? Are you planning to do something on this or shall I slog through the swamp? If the latter, can you post your bakery goods so I cans see what I can do with it? Anoneditor 23:06, 22 January 2007 (UTC)

Try this as a starter. Let me know if you have a use for matlab code. Dicklyon 05:37, 23 January 2007 (UTC)

OK, here's my last stab at it. If this still has technical problems that aren't easy to fix, someone else is going to have to do it; I've hit the wall with this one. If it's OK, my plan is to use the diagram below with the following textual explanation:

The diagram below shows how this works in graphic form. Depicted on the left is the absorbtion of colors of the spectrum according to their wavelength as they pass through the silicon wafer. On the right, a Foveon X3 layered sensor stack in the silicon wafer for each output pixel is shown depicting the colors it detects at each absorbtion level. The color purity and intensity of blue, green and red depicted for the sensors are for ease of illustration. In fact, these attributes of each output pixel reported by a camera using this sensor result from the camera's image processing algorithms.

Thanks for the offer of the matlab codes, but I don't think I'll be needing them. Anoneditor 22:51, 24 January 2007 (UTC)

That looks good. Just one issue: it's drawn taller than wide. In fact, the depth to the bottom of the red detection region is less than 5 microns, and the pixel size in the SD9 and SD10 is 9 microns (sensitive area somewhat less, like 6 or 7 microns). So it's wider than deep. By the way, in any ordinary CCD or CMOS image sensor, the light absorption is at these same depths, but the just don't segregate it vertically. Dicklyon 06:11, 25 January 2007 (UTC)

Thanks. I have greatly appreciate your assistance. Anoneditor 18:26, 25 January 2007 (UTC)

Thanks, Van der Hoorn, for catching the spelling error. I've fixed it in the graphic.Anoneditor 18:35, 30 January 2007 (UTC)

[edit] Don't understand the complaint

I don't understand the complaint that the article reads like an advertisement. It seems to me that it gives the relevant technical details on the differences between this sensor and the Bayer-pattern sensors and its advantages and disadvantages. It also makes a reference to an EDN article that is not wholly complementary. The fact that it exposits the theoretically superior nature of the chip seems justified to me because it is theoretically superior. Also, its practical superiority has been shown in the review of the Sigma SD10 in dpreview.com (http://www.dpreview.com/reviews/sigmasd10/), which seems to be a fairly neutral reviewer.

How else could such an article be written to avoid the complaint?

For the record, I don't own a camera with one of these sensors nor do I have any economic interest in the manuracturer. Anoneditor 18:04, 19 December 2006 (UTC)

[edit] The Nature of the Dispute

It is precisely the "technical details" which are a matter of the greatest dispute.

First of all, dpreview.com in its new announcment of the forthcoming 14-megapixel DSLR (for an astonishing $1599 at Amazon.com) makes a clear distinction between number of effective pixels and number of photo detectors. To my knowledge, Phil Askey of dpreview.com has always been careful to make that distinction.

The March, 2004 review in Digital Photography Review (dpreview.com) alluded to by the previous contributor does not offer an endorsement of the technical claims made. Rather, it reports the claims made. In any case, Phil Askey of dpreview.com tries to evaluate cameras based on performance. He does not, to my knowledge, offer sufficiently technical explanations that could either confirm or refute technical claims such as those made for the Foveon sensor, nor is that the purpose of his site. The review cited is almost three years old, in any case, and Phil Askey gave the camera in question a "Recommended" rating, not a "Highly Recommeneded" one. He also cited some problems that, in retrospect, could possibly contradict technical claims made about color and digital artifacts.

I have tried to indicate the disputed claims in my edits of the article. Someone with much greater technical expertise than myself would be needed in order to assess the validity of the claims made.

I have never used a camera with such a sensor, but I have used a number of cameras that use either CCD or CMOS sensors. I have no financial stake in this discussion. I certainly have nothing against the Sigma corporation (the primary user of the Foveon sensor), and I use their lenses from time to time. I like them. I certainly do not want to pan their products or components of such products if they have made revolutionary breakthroughs. Rather, I would buy one.

One thing is troubling to me, however, and that is what brought me to Wikipedia to check on the Foveon sensor. Just hours agao, I saw the new forthcoming Sigma 14-megapixel camera advertised on Amazon.com, which is taking orders in advance. I also consulted dpreview.com, which also contains an announcement of the camera just alluded to. What I read from Sigma did not even mention the sensor size, only that the new camera uses the Foveon sensor. Failure to mention sensor size in a new product description is most unusual, and it suggests the possibility that Sigma does not want to be drawn into a discussion of the merits of the sensor. That is a red flag for me in and of itself, but all the more so since the only 14-megapixel or higher sensors so far used have been 24mm x36mm designs (used by both Canon and Kodak). The avoidance of the sensor size issue suggests that the sensor used in the new camera is not a "full-frame" (24x36) sensor, implying that Sigma is claiming to get 14-megapixels from a smaller sensor. That raises questions immediately about pixel count, which is at the core of the dispute about Foveon sensors. Since lots of pixels on small sensors has been linked to claims (unresolved) about digital noise, the possible suggestion is that some of the techinical claims made about the Foveon sensor fly in the face of what has been considered technologically feasible to this point.

I would like to believe the claims made for the Foveon sensor, but any claim that virtually doubles the effective sensor count has to be viewed as being extremely suspect. The Foveon sensor certainly has its zealous adherents, and perhaps they are on to something, but so far the claims made are suspect and probably exaggerated, if not specious.

Readers are referred to Amazon.com (search "Sigma camera") and to dpreview.com under "Sigma." A camera using a revolutionary new sensor that offers 14 megapixels (on a small sensor?) and costs only $1599 is perhaps too good to be true. Even if it should be true, one would want to know about digital noise, low light capability, and high ISO effects on noise. Since the Foveon sensor is used in the camera in question, claims about its actual pixel count are of the utmost importance, as are claims about low noise and other digital artifacts.

I don't think that the entry on the Foveon sensor should be pulled just yet, since the element of doubt has been raised. Even so, questions remain as to whether readers might still read the entry as being an endorsement of products based on Foveon technology.

Other contributors/editors, especially those who understand all technical issues at stake, could be most useful on this one. I am not quite up to the task of doing the kind of rewrite that is really needed.

Landrumkelly 10:44, 23 December 2006 (UTC)

Landrumkelly, I looked over your changes. Thanks for making clear the items you are skeptical about. However, your changes need to be turned into "citation needed" tags, instead of WP:weasel words like "it is claimed." Then, editors can go through and try to fix those items, either adding WP:V refs, or change to "Foveon claims" with a link to where the claim is made, and maybe refs to counter claims when they can be found, or just remove it if it's not verifiable. Where you've said "extremely controversial claims are made" presumably is an example of a place where links to claims and counter claims can be found; it would be much better to have "Foveon claims X" and "Phil Askey disputes it", than your opinion that it's extremely controversial; see Wikipedia:Guidelines for controversial articles. Please fix your edits promptly, or I will revert to a state before the widespread changes to "it is claimed." Dicklyon 17:23, 23 December 2006 (UTC)

As to the sensor size and other technical details of the Sigma SD14, have you considered consulting a neutral source such as wikipedia? Dicklyon 17:27, 23 December 2006 (UTC)

Due to the demands of the holiday season, I will not be able to do this. I will be leaving town later today. It might be better to pull the entry, in my opinion, or let it revert with its caveat that it reads like an advertisement. I have used the phrases only to indicate that the claims are not neutral claims, and to specify which claims are in doubt. If I had the sources at my fingertips, I could make the changes that you request. Unfortunately, though I have followed this controversy for some years now (countng the earlier Foveon sensors), I do not have the sources readily available. If I say "Foveon claims" (which is indeed what I mean), then I will have to back that up, and I cannot do that quickly. Nor can I cite the numerous sources of competing claims (regarding "controversy") in the time available. Some of the controversy can be found in the sources cited for the article. I am sorry that the author of this entry did not incorporate those competing claims into his original article. Revert or change as you will. Thanks.

Landrumkelly 18:28, 23 December 2006 (UTC)

I reverted to the pre-weasel-word state. At this point, we're not asking you to provide sources, just to tag the places that you think need them. That will help others to address your concerns. When you have time, put the {{citationneeded}} (or {{cn}} or {{fact}}) tags after any statements that need to be verified or removed, and others will then be guided to help. Or someone else might help if you're off for a while. Dicklyon 18:39, 23 December 2006 (UTC)

I'm not sure how to add the "citation needed" tag. Perhaps I can find this out upon my return.

Landrumkelly 18:47, 23 December 2006 (UTC)

Just put the text I indicated above, where a reference would go; the double squiggly brackets mean that a template is envoked. See Template:Citationneeded, which also lists some other templates you can consider. Dicklyon 21:55, 23 December 2006 (UTC)

Landrumkelly, I agree that Mr. Askey has always been careful to make the distinction between the number of effective pixels and number of photo detectors. However, in my opinion, the way he does it doesn't support your argument. In the opening paragraph of the review of the SD10, he states, "...in our previous experience the X3 sensor delivers approximately twice the resolution of a standard mosaic sensor with the same number of horizontal and vertical pixel locations." Strictly construed, that statement would mean that the X3 sensor has a Bayer equivalent resolution of 13.7 megapixels (the product of twice each dimension. However, I think his meaning is that it equals a Bayer sensor with twice as many pixels. This would amount to a Beyer-equivalent resolution of approximately 4.9 megapixels (sqrt(2)* each dimension) or 3207 x 2138. This comports nicely with the results of his comparison with the 6 megapixel Canon EOS 10D in the review. As he states on p. 15 of the SD10 review, "What's clear to see from these comparisons is that the X3 sensor technology achieves amazing levels of detail and resolution pixel per pixel compared to the six megapixel sensor of the EOS 10D. Indeed it's fair to say that the EOS 10D doesn't truly exhibit any more visible detail than the SD10 (just a larger image, which we'll examine more on the next pages)."

Moreover, contrary to the statement made in your second paragraph, with respect to the X3's resolution, Mr. Askey does offer sufficiently technical explanation to confirm the technical claims made for this sensor. For instance, in the second paragraph of the commentary on p. 18 of the SD10 review, he states that the higher resolution of the Foveon sensor results from its lack of an anti-aliasing filter.

To me, your speculation about the size of the sensor in the newly announced SD14 is interesting but irrelevant to the question of whether or not the sensor technology is better. (I have no doubt that it is smaller than a full frame sensor.)

Finally, I don't think the fact that the SD10 was not rated "Highly Recommended" helps your argument much. That rating probably had more to do with perceived deficiencies in the camera other than deficiencies with its sensor: seven of the nine "Cons" in the review's conclusion relate to other aspects of the camera.

It still doesn't read like an advertisement to me. I don't think it should be pulled at all. Anoneditor 22:14, 23 December 2006 (UTC)

The previous editorial complaint about this article was that it read like an advertisement. I responded to that in two prior postings. The first noted that the article gave both pros and cons on the sensor involved and that the sensor was fairly well reviewed by one independent reviewer. That brought on an editorial response that I thought was not well reasoned and misread the material it used as its purported basis and I mentioned this in my second posting.

Now, the article is accused of using "peacock words" and advancing a "limited or personal interpretation" of the subject matter.

Peacock words. After following the link to the definition of "peacock words," I find that the only words in the article that could possibly be classified thus are: The word "advantage," in the first sentence of the fifth paragraph and the word "elegant" in the last paragraph. Couldn't these words simply have been changed rather than make all the fuss with the multi-colored headline banners?

Limited or personal interpretation. How does this assessment square with the text of the article?

• Paragraphs 1 & 2: Provide factual information on the subject sensor.
• Paragraph 3: Notes the difficulties in comparing the resolution of the sensor with the conventional Bayer sensor, setting out the way different pixel counting methods may underestimate or overestimate the sensor's resolution.
• Paragraph 4: Notes the factual reality that demosaicing, anti-aliasing and sharpening are not required due to the design of the filter.
• Paragraph 5: Notes the technical differences between the two designs with respect to color absorption, and the advantages and disadvantages of the subject sensor in this respect.
• Paragraph 6: Takes up the pixel count resolution issue again, comparing the two types of sensors from a different perspective.
• Paragraph 7: Notes how the physical characteristics of the subject sensor affect its performance respecting focusing, chromatic aberration and sharpness in longer wavelengths.
• Paragraph 8: Notes a book written on the subject.
• Paragraph 9: Notes one reason for the lack of great commercial success of the subject sensor.

If this is a limited interpretation, then that appellation could be applied to any technical article that doesn't completely cover every facet of its subject. Also, I cannot understand why it would be categorized as personal as the vast majority of articles necessarily contain the opinions of their authors. Perhaps paragraph 8 should be a reference and paragraph 9 could mention other possible reasons.

As a result of my analysis above, and with all due respect, I think that the editorial response to this article has been quite disproportional to whatever defects it may have. Anoneditor 05:46, 24 December 2006 (UTC)

Sorry. I probably should have explained myself earlier, but the holidays don't allow me much free time. While I'm not going so far as to directly question the validity of specific claims like Landrumkelly, I do agree that the article doesn't meet the standards of an encyclopedia entry, lacking the formal tone that is to be expected. NPOV policy is not "find what appears in your opinion to be a fair and balanced assessment of the subject, then endorse it." It is not the place in an encyclopedia to assert anything but objective fact. All opinionated or otherwise remotely disputable information should be attributed to external parties by name. Words like "difficult", "advantage", and "elegant" are subjective terms, and open to debate, hence the {{peacock}} template I added.

What criteria are being used to judge "better"? What common references are being used for comparison? Number of pixels based on RGB format conversion? Sensor size? Sensor cost? Device cost? Short-exposure performance? (From what I've read, cameras that use X3 sensors tend to produce prohibitively noisy images when using long exposures.) Energy efficiency? These would all produce different results. To me, "theoretically superior" often means "better under a particular set of conditions that exclude relevant real-world factors." Phrases such as "on the other hand" can imply primacy of one thing over another. Stating that "prospective camera buyers need to understand the pixel counting definition" is wrong. They don't need to understand anything.

As Dicklyon pointed, the real problem is the lack of rigorous verification of all the statements in the article. Each assertion should have a reliable source that directly asserts the claim. For example: "Regardless of the elegant design, the Foveon X3 sensor has not yet achieved great commercial success because most camera manufacturers use their own image sensors." Is that really why the sensor has not achieved great commercial success? Says who? How did they arrive at that conclusion? What is "great commercial success" anyway? Was this really regardless of the design? Was the design really that elegant? How elegant? You say that is valid as one reason. The article does not specify that it is just one reason, though. In truth, I didn't find that anything was provided to verify it as one possible factor.

Landrumkelly did try to fix things by changing the wording. Dicklyon was right in saying that it was merely weasel-wording that does nothing to address the real problem of attributing sources. Some of the statements are not supported by the existing references, and without sources, those statements could be construed as original research.

While you can probably find dozens of examples of personally asserted opinions found in Wikipedia articles (there is certainly a tendency towards among editors, in my experience), I dispute that they are there "necessarily". While I haven't checked myself, I'm betting the better articles manage to keep this to a minimum. As for other articles, if you look at my contributions, I doubt you'll find that I'm any less harsh on articles that I edit, and I don't think I should be held responsible for articles that I don't edit. The condition of this article should be argued on its own merits. Dancter 09:24, 24 December 2006 (UTC)

Dancter, Please accept my apology; I should have realized that saying I thought the editorial position taken was unfair might be taken as a challenge to the editor's integrity. That was not my intent and I'm sorry if it came out that way.

Also, thank you for taking the time to fully explain the meaning of your tag, which now makes sense to me, and I greatly appreciate it. I would suggest that the definition of "peacock words" be expanded to include the terms you mention because, as it stands, they are those "that merely show off the subject of the article without imparting real information." Maybe a new category is necessary. Anoneditor 19:52, 24 December 2006 (UTC)

[edit] Sources

There are lots of sources to draw on to fix the article. Books, Foveon tech papers, other papers, patents. It shouldn't be hard to find sources for statements, to say which ones are Foveon claims, or someone else's claims, when they are not obviously factual, etc. Dicklyon 18:58, 24 December 2006 (UTC)

TOO MUCH EMOTION ON A TECHNICAL SUBJECT

I've gotten back from the holidays and read the above commentary, as well as seen the changes from "Reads like an advertisement" to references to "peacock words." Someone with more technical skill than I possess needs to work on this article. Unfortunately, in my opinion, the piece is clearly promotional, i.e., does sound either like an advertisement or a piece by an enthusiast of the cameras that use this sensor. In any case, I can't fix the article, but it definitely needs fixing. This is definitely not an encyclopedia entry. It can only come to be that if written by someone who is both (1) technically proficient and (2) neutral and objective. I shall have to leave this entire entry with commentary to someone else. Right now it is still a mess, and there is too much emotion in the discussion of claims that could be verified or disverified empirically. Citations will not solve the problem. I am not even sure that the topic deserves an entry at this point. Promotional literature of unproven technological innovations is surely not what Wikipedia is about. Barely disguised zealots can be expected to keep coming back to this one, creating an editorial nightmare. Right now this one is a laugher. Landrumkelly 04:20, 31 December 2006 (UTC)

Landrumkelly, it's unclear to me why you sense emotion here. Seems like a reasonably dispassionate discussion to me, at least by comparison with other Foveon X3 discussions that I have seen. Why don't you go ahead and mark with fact tags the items that you see as needing sourcing or correction or removal? There's no need for you to take on the fixing, just point out the problems. I can help with the technical expertise and finding sources, but I've avoided doing a lot of editing here because I am not sufficiently neutral on it. I'll look for sources where requested, correct technical errors, and comment in the discussion, but I'm not going to rewrite the article. Dicklyon 04:53, 31 December 2006 (UTC)

One problem here, however, is that few people seem to be paying attention to this article, and nobody really working on it. So let me point out a few things, and we'll see if we should make some changes. First, as to the statement about the sharpening filter, I don't buy it; a sharpening filter is not REQUIRED with either Bayer or Foveon type images, but is usually used with both. So I'd leave that out. As to an AA fitler not being needed, that's more true; an AA filter is not used in the Sigma cameras, unless you count the anti-aliasing effect of the area over which the microlenses integrate the optical signal; their are plenty of sources to back up the ideas that this is close to the best that can be implemented for an image sensor, unlike the situation with Bayer sensors, though such sources might need considerable interpretation; would you like me to find one? As to the issue about "effective pixels", that's a tar pit; Askey has his own unique solution; it's probably best to just state what Foveon and Sigma say, and to state what Askey says as an alternative; at present, nothing in the article mentions effective pixels, but the question of how to count pixels is discussed (in a somewhat POV way, though, saying what a buyer needs to do). Most of the rest of what you had put "it is claimed" on are simple points of physics or optics, that as far as I can see are about right (for example, the depth of absorption of light in silicon is no different in Foveon's silicon than in anyone else's). Please feel free to ask here, or with tags, about any that you'd like to see sources on. One of the items you questioned already has a reference (on image comparisons) that I added a while back. Dicklyon 06:03, 31 December 2006 (UTC)

Based on discussion here, I have removed the peacock and essay tags, and put a sources tag instead. Dicklyon 20:03, 31 December 2006 (UTC)

Landrumkelly, if this topic doesn't deserve an entry, then neither do the CCD or CMOS sensors currently in use in digital photography. In my view, the X3 is clearly an important innovation in digital photosensor technology as it avoids the color interpolation of the current sensors. It may not be a perfect product, but it certainly is noteworthy. Moreover, I find it hard to see how this article falls in the class of "Promotional literature of unproven technological innovations." As to "unproven", the sensor is currently in use in a commercially marketed camera and at least one outside reviewer has demonstrated image sharpness equivalent to current 6MP Bayer sensors. As to "promotional," though the article may have some technical issues, it isn't as one-sided, as you would expect a promo piece to be. I think Dicklyon has done some good recent editing on the text and, contrary to your view, I think citations will help. Anoneditor 22:11, 31 December 2006 (UTC)

Nobody has suggested removing the topic. Get over that misconception. Dicklyon 22:50, 31 December 2006 (UTC)

I don't recall even hinting that anyone had suggested it. I was merely taking issue with Landrumkelly's observation. What prompts you to think I have a "misconception" about this? Anoneditor 23:17, 31 December 2006 (UTC)

I was interpreting your earlier statement "I don't think it should be pulled at all" and your more recent "if this topic doesn't deserve an entry" as rebuttals to an imagined suggestion that this topic doesn't deserve an entry or should be removed. Sorry if I read you wrong. Dicklyon 23:51, 31 December 2006 (UTC)

Dicklyon has indeed "done some good recent editing on the text," and it does indeed read much better than when I first saw it about a week ago. I'm sorry that I did not notice that before responding again. I do think that it should stay, and the sensor is indeed a promising development. Just how promising remains to be seen. I apologize for the confusion. Landrumkelly 02:31, 1 January 2007 (UTC)

Dicklyon, I would like to remove the non-disclosure-of-sources tag on this article. In one of your replies to Landrum Kelly, with respect to anti-aliasing filters, you said, "...an AA filter is not used in the Sigma cameras, unless you count the anti-aliasing effect of the area over which the microlenses integrate the optical signal; their are plenty of sources to back up the ideas that this is close to the best that can be implemented for an image sensor, unlike the situation with Bayer sensors, though such sources might need considerable interpretation; would you like me to find one?" Would it be too much trouble for you to find one for me so I can see if this article can be finished? Anoneditor 03:57, 1 March 2007 (UTC)

See ref 2 of Anti-aliasing filter (first paragraph of section 26.5). If you need more, Pratt's book on Digital Image Procesing and Sequin & Thomsett's book on CCDs also make the point, if I recall correctly. Or see this video lecture on silicon image sensors (at 56:25). Dicklyon 05:37, 1 March 2007 (UTC)

Thanks so much. I'll run them down now that I've reorganized the article. Anoneditor 20:29, 4 March 2007 (UTC)

[edit] Concluding Sentence Issues

"The Foveon X3 sensor has not yet achieved great commercial success because most camera manufacturers use their own image sensors or buy them from large semiconductor companies such as Sony."

The sentence above does not make sense. The reason most camera manufacturers prefer to "use their own image sensors or buy them from large semiconductor companies such as Sony" would be why the X3 is not successful. The article quoted for the concludimg statement even states that one of the few brands that DID adopt the Foveon was recalled due to "reported image-quality problems". So the reasons given by the article that the X3 is unsuccessful are that 1) Foveon is still an unproven technology and 2) the recall did not look good for this still unproven technology. 24.83.178.11 06:55, 29 January 2007 (UTC)KnowledgeSeeker

I'm deleting this paragraph and its reference for the following reasons:

• "[G]reat commercial success" is undefined. (See comment by Dancter, above.)

• I think that the comment above, that the reference doesn't support the assertion made in the paragraph, is true.

• The reference itself is tainted because it doesn't cite any source for its assertion of the recall on quality issues of one of the cameras using the sensor, implying the sensor was at fault. (My quick research indicates it may have been based on the April 15, 2005 news item in Imaging Resource.com [1] which ascribed the report to an anonymous reader. It also raised the inference that the problem may have been with the image processing firmware used in early camera samples. If so, this doesn't equate to a problem inherent in the sensor itself.) Anoneditor 23:46, 1 February 2007 (UTC)

[edit] Article Reorganization

I have substantially reorganized the article for clarity by separating the description of the sensor and its operation from the comparison of the sensor to the Bayer sensor. Then, I broke that comparison down into three sub-categories. I've also eliminated the conceptual redundancy and lack of continuity formerly existing in the area of "megapixel" counting and supplied more references and more explanatory text.

However, there are still two matters that are unclear to me, both of which appear in the subsection on "Light gathering and low-light performance." First, Is there really a significant difference in difficulty between demosaicing the Bayer cell output and "matrixing" the Foveon X3 output? Second, does the "matrixing" really have anything to do with low light performance of the Foveon sensor, as also stated in that paragraph? The latter seems unlikely but, not being a scientist or engineer in this field, I don't know. I left those assertions in because they were in the earlier version of the article and there were no complaints about them. I've asked for citations for both. Anoneditor 20:32, 4 March 2007 (UTC)

[edit] Claims of controversy

Anon, you've made much of so called controversies. I toned it down a bit, tried to explain better what the controversy might be about, and added citation needed tags. We really should find reliable sources if we're going to say such things. Furthermore, your calling the SD10 3.4 is taking sides against the published and accepted specs found on all retailer sites, so is rather POV. See if the way I fixed this discussion looks OK. The issue of the MP count needs to be discussed, and what the companies call it is certainly backed up by verifiable sources. But these should not be confused with issues of resolution. Dicklyon 05:06, 7 April 2007 (UTC)

Sorry not to have gotten back to you sooner but I've been out of the country and am now seriously jet-lagged. I'll try to respond in a couple of days when I get my head on straight. Anoneditor 12:11, 24 April 2007 (UTC)

Dick, I was fascinated by your commentary accompanying your edit of April 7 in which you say that my statement of the argument over comparison of the spatial resolution of the X3 sensor and the Bayer filter sensors is erroneous. I had stated:

The argument has been over whether it is accurate to compare the number of photosites in each sensor's array or to compare the number of Bayer sensor photosites with the total number of photodiodes in the Foveon X3 sensor array.

Your comment was, if there is an argument, that was not it; remove footnotes of made-up stuff.

It seems beyond doubt that there has been such an argument, and I think my statement of it was accurate. It also seems to me that your revision, though it uses somewhat different terminology, doesn't add much to the understanding of the issue that hasn't already been stated. By the way, what exactly is the made-up stuff in the footnotes you removed?

The "made up stuff in the footnotes" included these:

• "There is a dispute about the counting of megapixels as a universal measure of the spatial resolution of image sensors. Nevertheless, these terms are used here because they are ubiquitous in photographic parlance."

It's true that megapixels is ubiquitous, but I don't know of any dispute about it being a measure of the spatial resolution of image sensors. The industry standards say it is not to be used that way. Has that been disputed? Where?

• "Because it is similar to an area measure, all other things being equal, the difference in resolution between two image sensors varies as the square root of the ratio of the number of their photosites. In this case, sqrt(10.3 million ÷ 3.4 million) rounds off to 1.7."

It's not really clear what you're trying to say here. The term "photosites" has the same definitional problems as pixels or megapixels, potentially, but no matter how you define it, there is no such proportionality relationship when comparing across different photosite organizations.

I have no problem with your "toning down," language. However, I think you're shooting the wrong messenger when you lecture me about my statement that the X3 sensor has been advertised as a 10.3 MP unit being unsupported by a reliable reference. A statement to this effect has been in this article since at least since July 17, 2005, when the following text appeared:

The question of whether to count each stack of three photosensors as a pixel, versus counting each individual single-color photosensor as a pixel as is done in Bayer-mosaic sensors, has been a point of controvery for Foveon X3 sensors and for the specifications of cameras that use them. For example, the Sigma SD10, which produces a native RAW file size of 3.4 Million RGB pixels, is advertised as a 10.2 MP camera, sometimes with the clarification 3.4 MP Red + 3.4 MP Green + 3.4 MP Blue; an 8 MP Bayer-mosaic camera would similarly be clarified to be 2 MP Red + 4 MP Green + 2 MP Blue.

This text continued in the article until my reorganization in March of this year. All I did was restate it in slightly different language.

I'm not trying to blame anyone, just trying to make it more objective. I believe that old version you quote is one I wrote myself, and while it did refer to a controversy without reference, which I now regret, its point was to state the issue in a way that was not confused with resolution. The confusion of pixel counting with resolution is mostly what I didn't like in your version.

Also, my calling the sensor in the SD 10 a 3.4 MP unit wasn't intended to exhibit a POV. It was merely a statement of the number of photosites that the sensor contains, where the term photosites means the elements in the sensor's grid. And I did say in the next sentence that the SD 10's sensor grid consists of a stack of three photodiodes, one for each color. So, I guess I don't understand your complaint.

But here you've implicitly chosen a definition for photosite and then for pixel. If there is a controversy, you shouldn't take a side on it. If there's not, go with the number that the manufacturer uses, which is supported by the industry standards. I understand the POV was not intentional, but it is still one side, where the other side says 10.2 million photosites or pixels. It just depends on what you decide to count. If there's a controversy, that's what it's about. If there's not a verifiable source about such a controversy, however, let's just stick to what the industry practices call it.

By the way, in your subsequent revision, you say that the Sigma SD14 camera is the only currently available consumer camera using the X3 sensor. Did you know that today, Polaroid lists the X530 as a current camera using X3 technology? The URL is: http://www.polaroid.com/global/detail.jsp?PRODUCT%3C%3Eprd_id=845524441763611&FOLDER%3C%3Efolder_id=2534374302029037&bmUID=1177812494726&bmLocale=en_US

I didn't know that. I know they stopped producing and selling them more than a year ago. That link doesn't work for me, perhaps because it denied that site the right to save cookies. But if they say it's current, feel free to add it.

Don't you think that you're stretching McNamara's statement about the SD14's noise performance in your latest edit? You say that, More recently, reviewers are judging the Sigma SD14 camera with the new 14 MP Foveon X3 sensor to be very low noise. However, what McNamara actually says in this respect is: ...based on the SD14's heritage, you can expect it to capture extremely accurate colors, fine details in shadow areas, and incredibly low noise. We'll reserve judgment, though, until we've run a production unit through the Pop Photo Lab. It may very well be that the SD14 has superb noise performance but I don't think this source supports that assertion. Is there another reviewer you've not cited?

I think I found something beyond that forward-looking statement. I'll have to go back and check.

Finally, for what it's worth, I wish photosensors could be evaluated exclusive of the image processing provided by the particular cameras in which they are used. I think it's pretty commonly known that some of the camera manufacturers do a better job decoding the Bayer filter sensors than others and it might be that another camera manufacturer could do a better job than Sigma or Polaroid in processing the raw data from the X3. All of this, it seems to me, makes the direct comparison of image quality from the different sensor types much more difficult and more contentious.Anoneditor 02:56, 29 April 2007 (UTC)

Indeed, I totally agree. The best we can hope is to quote some reliable and reputable reviewers, and that's not always going to get close to any fundamental truth.

My main issue here was to avoid the kind of rancorous discussions you see on forum sites, by sticking to what's verifiable. That means that if there's a controversy to be discussed, it should be discussed in terms of some published reliable source, not on our reading of the forums. I do appreciate your help here in this article, but these points wer bugging me for a while and needed to be addressed.

Dicklyon 05:31, 29 April 2007 (UTC)

Dick, Thanks for your response. To make it easier to see my responses to your responses, I've extracted your responses and added mine to them in boldface.

The "made up stuff in the footnotes" included these:

• "There is a dispute about the counting of megapixels as a universal measure of the spatial resolution of image sensors. Nevertheless, these terms are used here because they are ubiquitous in photographic parlance."

It's true that megapixels is ubiquitous, but I don't know of any dispute about it being a measure of the spatial resolution of image sensors. The industry standards say it is not to be used that way. Has that been disputed? Where?

Well, maybe argument isn't the right word. What I meant in an abbreviated way was there's a common-sense notion that the finer the two-dimensional grid of photosites the greater the resolution of the image projected upon it by the lens. Thus, the count of megapixels in the public mind is fairly synonymous with the camera's resolution. (I think many camera manufacturers play to this belief, regardless of industry standards, by creating and advertising cameras with ever-increasing megapixel counts. And, in my understanding, this notion is what Foveon has been fighting from the inception of the SD9.) The argument then, is that this ignores the greater noise often created by jamming more photodiodes into the same size image sensor, degrading its resolution. And it also ignores the fact that some sensors, like the Fujifilm Super CCD and the X3 use a different sensor organization to produce effective resolutions that exceed what the common-sense notion would predict.

• "Because it is similar to an area measure, all other things being equal, the difference in resolution between two image sensors varies as the square root of the ratio of the number of their photosites. In this case, sqrt(10.3 million ÷ 3.4 million) rounds off to 1.7."

It's not really clear what you're trying to say here. The term "photosites" has the same definitional problems as pixels or megapixels, potentially, but no matter how you define it, there is no such proportionality relationship when comparing across different photosite organizations.

I think the proportionality relationship exists if you accept my stated premise of all other things being equal. This includes the equality of the photosite organizational scheme, too. Suppose you had two identically sized Bayer filter sensors, one having twice the number of photosites as the other and both having equal noise performance, decoding engines, etc. Would you agree that the sensor with twice the number of pixels would have a resolution of the sqrt(2) times that of the other sensor? I was simply trying to show how the "common sense" understanding of sensor resolution would compare the two on this basis before going into why it would not be correct. Maybe I took too short a shortcut. Anoneditor 02:15, 30 April 2007 (UTC)

Yes, I see what you're trying to do. The trouble is that when you say "all other things being equal" without specifically saying that the sensor organizations are obviously not equal, then you give the impression that a Foveon 3.4 MP x 3 layers sensors should be in some sense comparable to a 3.4 MP Bayer sensor. You even draw that comparison explicitly. Since that impression is the main problem, it's best not to make up explanations that lead that way. You could have gone the other way, and said that, all other things being equal, the resolution is proportional to the square root of the number of photodiodes (which is just as true, and more nearly accurate across the different organizations).

But actually, it's best not to make up anything, and just stick to what is supported by reliable sources. For example, I found where Sigma themselves had referred to a controversy about pixel counting, so I referenced and quoted it; if you can find anything else on that, from some other perspective, that would be good to use, too. But summarizing what you read in discussion forums is not OK; you have to report someone else's reliable analysis of this stuff. So, when I say made up, I don't mean necessarily wrong, but not based on a reliable source. Dicklyon 03:10, 30 April 2007 (UTC)

[edit] Modified lead

RenniePet, your mod to the lead is certainly a better conceptualization of how it works; the idea of a sensor responding to primary colors is a common confusion of concepts of sensing versus reproduction of color. I'd quibble with the word choice "color profiles," however, since that term already has a use in this field. How about "spectral sensitivities" or "sensitivity curves" or something like that? Also, on the "blue, green and red", that order was probably inherited from the top-to-bottom sensor order; I recommend the more conventional "red, green, and blue" (with the serial comma, too). Dicklyon 22:18, 25 June 2007 (UTC)

Thanks, better now. By the way, the three detector layers are customarily referred to as the red, green, and blue layers, even though they have soft overlapping spectral sensitivity curves, not responses to primary colors. So I don't think you need to modify the image to make up new names for them. I went ahead and put the primaries into RGB order in a few other places; even for the X3 sensor, it's a sensible order in terms of the silicon, since the deep diodes get formed first. Only the incident light sees order BGR. Dicklyon 23:52, 25 June 2007 (UTC)

[edit] Light gathering and low-light performance

Rennie Pet, in your latest edit, you say:

However, although the Foveon X3 has greater light gathering ability, the color-indicating information in the sensor raw data requires matrixing to produce color data in a standard color space, which can be problematic in low-light situations.

Problematic in what way? Also, why would this be any more problematic in low light situations than in others. Am I missing somehting? Anoneditor 19:53, 15 August 2007 (UTC)

Because the Foveon sensor has poor color response (the color response of the three stacked sensors overlap each other a lot) you need a statistically significant amount of data (number of photon hits) to deduce the color accurately. In low light situations there are not enough photon hits to provide good color rendition.

At least, that's my understanding of the situation.

Maybe this can be worded differently to make it more clear... --RenniePet 20:31, 15 August 2007 (UTC)

Forgot to mention that the fundamental reason for this problem is that depth of penetration for photons is very random. A photon of a certain color (wavelength) may penetrate to quite different depths before "dying" and releasing its energy. That is what causes the widely overlapping color response curves of the three sensors.

The color filters on the Bayer sensor do a better job of determining color, although photon randomness is also wrecking havoc there too.  :-) --RenniePet 20:43, 15 August 2007 (UTC)

It's simplest and most accurate to just consider the spectral response curves of the channels and then the resulting shot noise and matrixing. That captures the randomness of the depth of absorption just fine, and more simply, and a uniform way for both sensor types. Dicklyon 23:20, 15 August 2007 (UTC)

Thanks for the explanation, RenniePet. Dicklyon, I seem to remember seeing an explanation of "agressive" matrixing somewhere in the exchages on this subject, but now can't find it. Shouldn't there be a footnote for this? Anoneditor 02:10, 16 August 2007 (UTC)

Yes, there is a footnote. The referenced document says it "means large off axis multipliers". There may be other better ways to characterize the matrix factors and their effect on noise multiplication, but I haven't found it. If you find a better source to write this up from, by all means cite it; or work on clarifying what's there if you see a better way. Dicklyon 05:44, 16 August 2007 (UTC)

The footnote to this is a lot more enlightening than what is in the wikipedia entry. '"aggressive" matrixing' really doesnt explain anything to the layman. Can't the information from the article linked in the footnote 16 somehow incorporated in this section of the wikipedia article? Its the most lucid explanation i have read so far as to what happens in a foveon sensor in reality and why it has more moise than a bayer sensor (because in theory it should have less).

[edit] The Foveon X3 shows monochrome moiré patterns

http://www.imatest.com/docs/print_sharpness.html "Moire in Foveon sensors is far less bothersome because it's monochrome and because the effective Nyquist frequency of the Red and Blue channels is lower than for Bayer sensors."

http://www.dpreview.com/reviews/SigmaSD9/page23.asp "However, because the X3 sensor doesn't use a low pass (anti-alias) filter it is able to resolve detail all the way up to Nyquist. Beyond Nyquist the system will alias without any objectionable color moiré. Where a Bayer sensor camera would turn detail beyond Nyquist (such as distant grass texture) into a single plane of blurred color the SD9 will continue to reproduce some individual pixel detail (without color moiré)."

-- Shotgunlee (talk) 16:32, 18 December 2007 (UTC)

Those are good sources. But I reverted due to the statement "...the Foveon X3 sensor with no anti-aliasing filter produces monochrome moiré patterns," since it is slightly wrong on two points: (1) there is actually a pretty effective anti-aliasing filter, as described at Anti-aliasing_filter#Optical_anti-aliasing_filter; (2) whether it produces any aliasing is a function of lots of things; it would be more accurate to say it can produce slight moiré. And you could check with Norman Koren for what he meant by "the effective Nyquist frequency of the Red and Blue channels is lower than for Bayer sensors" which sounds like he got it backwards. Dicklyon (talk) 16:44, 18 December 2007 (UTC)

(1) We can not say that the foveon x3 sensor has an AA filter. The high fill factor may provide AA effect, but it is certainly not an AA filter. The dpreview link shows strong aliasing. The fill factor of foveon sensor may close to 80%, but not to 100%.[2] (2) What are the things that affect production of aliasing? Why do we need can and slight when the dpreview link shows strong aliasing? I agree on the Nyquist frequency issue. -- Shotgunlee (talk) 18:56, 18 December 2007 (UTC)

I just checked which dpreview page you cited, and it's from the SD9, the first camera with X3 technology, which had no microlenses and so had more aliasing. The microlenses can properly be viewed as an antialiasing filter, since that is part of their purpose in the SD10 and SD14 cameras; the fill factor is about 90%, which is quite effective. Look at the corresponding SD10 review page; no mention of visible aliasing in the SD10 res chart, even though this is exactly the kind of pattern that would most provoke it. If you want to add something, this difference between the generations would be worth a mention. Book refs on Foveon and moire generally say "less prone to moiré" or "reduces" or something to that effect, but in discussing Bayer sensors say "sometimes..." because you don't always get moiré. It would be most appropriate to not reinterpret that Foveon advantage as a negative. Dicklyon (talk) 05:25, 19 December 2007 (UTC)

[edit] Polaroid X530

The article mentions the Polaroid X530 being in production as of March 2008. I know it's hard to prove a negative, but Google suggests that the camera was recalled shortly after release, back in 2005. This news story deals with the problem from a UK perspective, and although Polaroid's UK and US websites have a product information page about the camera, they list no dealers, and I can't find any on Google (I can't even find any on eBay.com). I reckon that the camera is no longer in production or on sale. -Ashley Pomeroy (talk) 18:24, 16 May 2008 (UTC)

The camera was re-released after the recall. I have one, and it works great. But I don't think they're currently available, unless in some backwater shop some place. Dicklyon (talk) 20:58, 16 May 2008 (UTC)

This store claims to have it available for $199. Get it before it's too late :) Dicklyon (talk) 20:59, 16 May 2008 (UTC)