Luminous Landscape Forum
Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: jhmaw on May 20, 2011, 09:15:08 am
-
So there it is. Sigmas SD1. And that price.
But I don't see the price as the main stumbling block. Really it's more the lenses. Now I'm not saying that Sigma doesn't make some decent lenses, but here's the thing. As they move towards higher resolutions the lens quality becomes more of an issue. Sigmas big problem is their quality control (or lack thereof). I am not the only one to have had vastly differing image quality from what should have been near identical products. Sigma cameras are also in the poor position of having to rely only on the body manufacturer for lenses. There are no third party lenses with a Sigma mount (as far as I am aware). A bit ironic when you think that Sigma is best known as a maker of third partly lenses for other people's cameras.
So with unreliable quality control for their lenses and an increase in image resolution can the SD1 be considered as the foundation of a good working cameras system? I wouldn't put money on it and they want an awful lot of money.
-
If this camera is an epic fail - the SA mount is not to blame - it will be the Price 100%. Even I, a Canon/Leica shooter, showed some interest in this camera after a brief experience with the DP1 and seeing how the Foveon sensor renders fine details etc.. but for $9700! I would sooner get an M9 with some cash leftover for the 5DMKIII!
-
As they move towards higher resolutions the lens quality becomes more of an issue.
they need to deliver the light to just 15mp photosites (remember that 3 sensels are "stacked" vertically) on APS-C sized sensors... the same lenses that are being used right now on FF bodies from C&N&S w/ up to 24mp photosites and on APS-C bodies that are up to 18mp photosites (7D, etc)... so where is the problem ?
-
No chance at all... it will fail. Perhaps its only purpose is to be a "showpiece" camera to try and lift perception of the rest of their product line.
But really... 1.5 crop, non-gripped body, and you're stuck using Sigma lenses.. for ten large? I laughed.
-
and you're stuck using Sigma lenses.
that is if you want AF... if not, then SA is pretty much a Pentax K-mount with Canon registration distance (and the lens electronic protocol is the same as Canon, differs only in IS part - so people were modding Canon lenses to work on it and autofocus too)...
-
The price is a bad joke.
The claim of 46 MP is a bad joke, if not a blatant lie.
(The X3 RAW image does definitely NOT transmit the same amount of information than a real 40 MP DB).
Though the working principle behind the X3 sensor is a cool idea, the price and disinformation they practise along with the feeling they give us ("Am I an idiot?") will make it fail along with the other quirks mentioned above.
Its a pity they 're shooting themselves in the knee for such stupid reasons.
-
Maybe they figured it's Armageddon day tomorrow anyway (May 21), so what the hell, let's go out with a splash.
-
I have to admit I don't follow this segment of the market very much (my expertise being medium format capture and software). But I love new cameras and new ideas and am a tech-nerd in general so I went ahead and read through several pages of various sourced info/opinions I am left in utter confusion.
Given that we sell camera kits anywhere from $5k-$60k and to a broad range of kinds of customers I'm the first to tell you that price is not always a stumbling block. But I'm finding it very hard to find the unique selling points that would compel any of our customers to spring for this compared to the other options out there.
I make this post honestly in hopes that someone will chime in with some compelling reasons.
Doug Peterson (e-mail Me) (doug@captureintegration.com)
__________________
Head of Technical Services, Capture Integration
Phase One Partner of the Year
Leaf, Leica, Cambo, Arca Swiss, Canon, Apple, Profoto, Broncolor, Eizo & More
National: 877.217.9870 | Cell: 740.707.2183
Newsletter (http://"http://www.captureintegration.com/our-company/newsletters/") | RSS Feed (http://"http://www.captureintegration.com/2008/08/11/rss-feeds/")
Buy Capture One 6 at 10% off (http://"http://www.captureintegration.com/phase-one/buy-capture-one/")
-
Doug,
There are several compelling reasons for this camera. The resolution and acutance should completely surpass any APS-C camera, and probably give some FF dSLRs a run for their money. The body seems to be competitive with dSLRs in the mid to upper prosumer range. FPS is decent, but the 7 image buffer might be a problem for some.
It's an extraordinarily compelling camera. For $1500. A good deal for $2000. A reasonable buy at a first-released price of $2500. But 4X??? I can't fathom what they're thinking. And I've owned SD10's, SD14's and SD15's---and I'm a huge fan. The fact that they indicated that the release price would be around the price of a 7D (not a pallet of 7D's) has really alienated their customer base.
It'll be interested to see how the camera works and fares in real life, but I'm not buying at that price. I was in at $2K, and would have considered it at $2500.
-
For those who claim Sigma's MP counts are way off, it can also be said of Bayer arrays - the technologies are very different, it's difficult to compare them fairly. I've read knowledgeable people here claim Foveon MP count equaling about [raw Foveon pixel count] x [their advertised multiplier of 3] x [real-world adjustment of 2/3], ie. SD1 would equal roughly 15 x 3 x 2/3 = 30MP of Bayer-produced pixels.
You read it on the internets from some random guy based on hearsay, so this is the gospel and the whole truth :P
I would really like Sigma/Foveon to succeed. We need more competition in the sensor-segment, and new tech is always welcome to push everyone to better achievements. The concept is sound, and am dying to see what DXO and other technical tests say.
Price sounds way too high, but that's what people said about S2 as well and it's killing in its market segment. Unfortunately for Sigma, they don't have the same brand equity as Leica.
I make this post honestly in hopes that someone will chime in with some compelling reasons.
Foveon zealots talk about the tonality, micro-contrast, color discrimination and 3D-quality of pictures produced by the sensors - not unlike the unfalsifiable, ill-defined and/or untested claims many MFDB zealots make.
fredjeang has a Sigma camera, and has given some reserved support for it in the past, with strong caveats on color accuracy in certain situations and colors - maybe he can chime in.
-
that would compel any of our customers
Doug, did any of your customers buy Pentax 645D ? but it sells and it seems good enough to pay off the amount of sensors that Hoya ordered from Kodak
as for SD1 - well, may be somebody who will think that cheap and IS optics on a body that is not that big/heavy to hike in some obscure Japanese wilderness (really not bigger than any FF from C&N&S) and still get around 30mp+ bayer quality... plus as usual w/ Sigma MSRP price might get down by a few $Ks... may be they will sell less than Pentax does - but the mere fact that they still can sell their cameras w/ Foveon sensors for years shows that they probably are OK financially w/ that side of their business even if it is a fraction of the percent of the market - because they have a good cash cow like their lenses.
-
With the D4/D400 and the 1DsIV not yet released, and just possibly subject to delay, and prices elevated across the board, this seems like it was the only opportune time for Sigma to release this. With full-frame sensors expected to top 30MP in the coming months, and APS-c sensors expected to reach 24MP, it's hard to see what place this camera has at its price point, and I have to wonder what their thinking is. A vanity purchase for wealthy hobbyists and a "prestige" item for Sigma?
-
Simply making a bigger sensor won't cure Sigma's greatest weakness, its color gamut. I have used an SD-9 and an SD-14 for years copying paintings because I bought into the Foveon hype. It really has troubles with ochres and straw colors and with greens--oh--and often with certain blues. In other words, it has troubles with red, green and blue. A Canon T2i blows it out of the water in terms of color rendition.
Also, is the BIGGER? chip 14-bit color? It truly needs to be, but I'll be surprised if it is.
Then, as anyone who uses Sigma lenses knows, sharp glass is a hit or miss situation in every lens you buy and the focus mount in lenses like the sharp 50mm f2.8 macro is so sloppy that it will slip out of focus if you hold your mouth wrong.
Now that I shoot with the T2i and a 50 f2 Makro-Planar, my images are far better overall and require far less trouble to get there.
-
Forking out 10K for a cropped camera and then tacking on Sigma glass doesn't appeal to me.
Im finding the lens issue more important as cameras grow in resolution.
-
I suspect the Sigma will fail in the long term (and possibly even in the short term) and the simple reason is its price v. image quality.
Its fine to charge 10k for a camera (or even a lot more), but the image quality and resolution has to match the price tag. MFDB companies like Phase get away with charging completely obscene amounts of money for their backs - BUT, the image quality allows them to get away with it. Simply put those backs produce a file that is superior in quality to anything else on the market and there is always a market for those who want the best irrespective of price. That is not to say these backs are worth the money you have to pay for them; simply that people will pay the asking price because the quality is there.
The Sigma on the other hand (from the sample files on their website) does not appear to be significantly different from anything else in that 10k price bracket (and in that bracket there is quite a lot to choose from). I suspect that many photographers (myself included) shooting with 10k bodies (or thereabouts) already have substantial investments in glass to suit their current cameras. None of them are going to want to switch camps lightly and have to start from scratch with lenses. Taking myself as an example: I would rather buy another 1DS MKIII and put 3 grand in my pocket and be able to use all my existing glass. I suspect Nikon shooters will feel the same way. Pentax 645D users wont consider it either for the same reasons.
So who is this new Sigma camera for?
My best guess is someone entering the 10k body market for the first time is the only real candidate for this camera. And, anyone entering for the first time has quite a lot of options from which to choose. I would not imagine the Sigma would be at the top of the list.
The Sigma comes across like a 'Concept Car' - "Here is what we can do and what we are capable of."
Only problem is concept cars rarely sell; despite the fact they are usually bristling with the latest in technology.
Edit - BTW: Has anyone else noticed that the camera doesnt do Live View?
-
The only rationale explanation would be that they have made some computations about potential market size, production capacity and margin and reached the conclusion that this price point would maximize their profit.
How is that different from Phaseone deciding to price the IQ180 at 43.000 US$? They would sell many more at 30.000 US$... or would they not?
If the SD1 cannot be sold below 3000 US$ for profitability reasons then they might as well go all the way. My prerogative is that the foveon sensor has a very devoted following in Japan. I have had interesting conversations with people when I was exhibiting 400 megapixel panos last year. One guy came and told me he tought his Sigma sd14 had at least medium format resolution and was probably close to my panos. :)
So those guys are typically elderly photo club gurus who need a differentiator gearwise. They will create their own myth and spread it around them. They have the cash and for them the SD1 costing 10.000 US$ is actually preferable as it adds some high end aura to something that was until now just a technical talk on their end. See... it has medium format quality and is priced at a level that corresponds to its abilities.
Considering the buzz this price point has generated, I think they have probably taken the right decision in the Japanese context. I don't believe they really care much about the foreign parts.
Cheers,
Bernard
-
The only rationale explanation would be that they have made some computations about potential market size, production capacity and margin and reached the conclusion that this price point would maximize their profit.
Yup - it could go down in the Biggest ever Corporate Blunders Hall of Fame. ;D
How is that different from Phaseone deciding to price the IQ180 at 43.000 US$? They would sell many more at 30.000 US$... or would they not?
It is different I think. Phase already have a market of followers in that high end price bracket. They are devoted and willing to shell out big dollars for their backs. Sigma have no such captured market. They are effectively starting from scratch in the 10k bracket.
I dont think Phase would sell a whole bunch more IQ180's if they were 30k instead of 43k. More likely those ordering it would just be happier with the price.
So those guys are typically elderly photo club gurus who need a differentiator gearwise. They will create their own myth and spread it around them. They have the cash and for them the SD1 costing 10.000 US$ is actually preferable as it adds some high end aura to something that was until now just a technical talk on their end. See... it has medium format quality and is priced at a level that corresponds to its abilities.
High end aura normally starts with the brand sticker - does it not?
-
If a prestige brand is what the SD1 needs, maybe an alliance with Zeiss is what they are lacking. Or maybe a Sigma-Contax body for this sensor.
Eduardo
High end aura normally starts with the brand sticker - does it not?
-
Sigma is a company that has been able to market product (particularly lenses) of attractive specification and reasonable price combined with poor reliability, generally mediocre performance and extremely inconsistent manufacturing (1 out of 10 meets specification). there have always been plenty of buyers attracted by the specification and price who don't care/ aren't able to judge image quality.
the SD1 will put their management philosophy to the test, but i'm sure they have not invested enough money (just as they don't invest the money in other products to make them consistent or reliable) to suffer seriously if they've mis-judged the gullibility of the market this time
-
WRT the price, I thought Michael's take was pretty good: it's likely proven very difficult to make those sensors. As for whether they can get any significant traction at that price, I'd be very surprised. But who knows, maybe the image quality is really outstanding, and they're just showing poorly converted jpgs on their site.
-
If it's noisy in low light, they're gone.
-
Now three choices of 35mm style camera Leica, Sigma and all the others.
Kevin.
-
I have to admit I don't follow this segment of the market very much (my expertise being medium format capture and software). But I love new cameras and new ideas and am a tech-nerd in general so I went ahead and read through several pages of various sourced info/opinions I am left in utter confusion.
Given that we sell camera kits anywhere from $5k-$60k and to a broad range of kinds of customers I'm the first to tell you that price is not always a stumbling block. But I'm finding it very hard to find the unique selling points that would compel any of our customers to spring for this compared to the other options out there.
I make this post honestly in hopes that someone will chime in with some compelling reasons.
Doug Peterson (e-mail Me) (doug@captureintegration.com)
The Foveon sensor have several advantages, but most of them are unfortunately theoretical. The following assumes that the new sensor is similar to the previous Foveon sensors, which is not a given.
Advantage 1) increased sensitivity.
The color-filters used in a typical Bayer-design will allow approximately 40% of the light to pass. A Foveon sensor should be able to get a about 1,3 stop advantage.
But...
1a) you only get this advantage if you can match the noise-level in the rest of the system.
1b) the color response can not be chosen as flexibly as a Bayer design.
Tech speak: to convert a RAW-image into a typical color-space (e.g. sRGB or Adobe1998), you usually start by applying a "camera matrix". This will increase the noise by a factor called a "condition number" (gory details here: http://en.wikipedia.org/wiki/Condition_number). If the color filters have a large overlap, the condition number is large. Typical Bayer-designs have a condition number in the range 1-2. Previous Foveon designs had condition numbers exceeding 10 (rough estimate). Thus, the color-noise is increased by at least 3,5 stops in practice.
Advantage 2: improved lens correction.
It is significantly easier to fix problems such as lateral chromatic aberration perfectly in software.
But...
2a) the wide color-filters means that the color channels will loose sharpness due to CA. This is similar to a panchromatic BW-image using a lens with a lot of CA.
2b) you need software that will do it.
Sigma could fix 2b... May I suggest cutting a deal with someone who makes an advanced RAW converter ;-)
* * *
The SD1 should be great as a pure BW-camera. All the disadvantages of the Foveon sensor mostly vanish in this case. This does, of course, assume that the software does the right thing and does not add a lot of noise before converting to BW.
-
Now three choices of 35mm style camera Leica, Sigma and all the others.
Kevin.
Sigma is not a 35mm , it can compete certainly with a Sony Nex 5
-
I have no experience with foveon sensors but aren't we as photographers trying to capture what the eye sees?
I understand this sensor mimics color film's layers but the eye has individual cones (RGB) and rods (B&W). It seems to me a sensor with individually colored photosites would mimic the eye better?
I always thought if the RGB sites were arranged in a pseudo random pattern (latin square) vs Bayer pattern we might have something. Other than mimicking film I don't see the advantage of a foveon sensor
Marc
-
I have no experience with foveon sensors but aren't we as photographers trying to capture what the eye sees?
I understand this sensor mimics color film's layers but the eye has individual cones (RGB) and rods (B&W). It seems to me a sensor with individually colored photosites would mimic the eye better?
I always thought if the RGB sites were arranged in a pseudo random pattern (latin square) vs Bayer pattern we might have something. Other than mimicking film I don't see the advantage of a foveon sensor
Marc
If you'd want a sensor mimicking the eye you'd also get halos - they eye does a lateral inhibition which causes these - sometimes one can see them.
-
The price is a bad joke.
Apparently the camera body is cheaper (http://www.sigmaphoto.com/shop/sd1-kits) when bought together with a Sigma lens, or is the price of a body already dropping as well?
Cheers,
Bart
-
Actual tests of the SD15 have found the resolution equivalent to an 8 or 9 megapixel Bayer pattern sensor so the SD1 will have the equivalent of a 24 to 27 megapixel Bayer sensor. 8 to 9 megapixels also happens to be the equivalent of 35 mm film so the SD1 can claim to be a medium format camera based on megapixels. Theoretical calculations have found the Foveon sensor to be equivalent to 2 times the resolution of a Bayer pattern in the red and blue and square root of 2 times a Bayer in the green so the SD1 will have the equivalent of a 22 to 31 megapixel Bayer sensor. The Foveon sensor gathers 3 times the light of a Bayer pattern so the SD1 has the area equivalent of 2.3 times a full frame Bayer sensor so SD1 can claim to be a medium format based on light gathering ability.
-
A crop sensor 35mm body is NOT medium format no matter how you work the slide rule.
-
Actual tests of the SD15 have found the resolution equivalent to an 8 or 9 megapixel Bayer pattern sensor so the SD1 will have the equivalent of a 24 to 27 megapixel Bayer sensor.
Charles,
I'm sorry that I have to step in. Damage control.
I don't know who performed these tests, but they are comparing the performance of a sensor without an AA-filter with a sensor with an AA-filter. Apples and oranges, especially since most MF Digital Backs don't use AA-filters.
8 to 9 megapixels also happens to be the equivalent of 35 mm film ...
Says who? I stopped scanning film when the 35mm sensors surpassed 16MP. Maybe my scans were superior?
Theoretical calculations have found the Foveon sensor to be equivalent to 2 times the resolution of a Bayer pattern in the red and blue ...
In a theoretical worst case scenario of Blue against Red and only when they have similar Luminance, a situation (2 opposite sides of the visual spectrum next to eachother) very uncommon in real life.
... and square root of 2 times a Bayer in the green ...
Sorry, nonsense. A Bayer CFA sensor collects luminosity at each sampling/sensel position. It's the AA-filter that deliberately reduces artifacts that causes a loss, which BTW is to a large extent recoverable with deconvolution sharpening. The Bayer demosaicing is only losing some 6.4% of Luminance resolution, as demonstrated on this page (http://www.xs4all.nl/~bvdwolf/main/foto/bayer/bayer_cfa.htm).
The Foveon sensor gathers 3 times the light of a Bayer pattern ...
Irrelevant, because a Bayer CFA sensor captures Luminance at each sensel position (the spectral bands that are filtered out are calculated and added back in at demosaicing time), only attenuated by the peak transmission of the filter. The Foveon sensor also only uses 1/3rd of the spectrum per color, although it only loses light to the increased amounts of circuitry that produces a smaller fill factor, and the other unutilized photon energy.
Cheers,
Bart
-
Apparently the camera body is cheaper (http://www.sigmaphoto.com/shop/sd1-kits) when bought together with a Sigma lens, or is the price of a body already dropping as well?
Convincing people to buy Sigma lenses is much harder than getting them to buy the body, hence the discount ;D
-
I have a SD9, Sd14 and DP1. My own testing of the SD9 against my 350D showed convincingly that was essentially no visible resolution difference between them (there were visible differences in the type of aliasing artifacts however, which was fun to evaluate).
Likewise, in my Sd14 vs Kodak 14nx article that is quite well known, I showed that in prints up to 24x16 inches there is for all practical purposes, no resolution difference.
Overall, I'm happy with a rule of thumb that says the Sigma cameras have a resolution that roughly, approximately on a par with Bayer sensors of around twice the spatial pixel count. It's what Sigma themselves think when they are not busy marketing it at 3x the count.
Of course, there'll always be something to quibble over and someone to produce a piece of academic research that says differently, but for practical everyday shooting purposes, testing shows the 2x ratio is a reasonable rule of thumb for rough equivalence.
-
Likewise, in my Sd14 vs Kodak 14nx article that is quite well known, I showed that in prints up to 24x16 inches there is for all practical purposes, no resolution difference.
Hi Dave,
I noticed that in that article the Kodak 14n (I don't recall an nx, but suppose its resolution is the same), was shot a f/11. That will produce a diffraction blur pattern with a diameter of 14.9 microns for Green (555nm) light. That's 1.9 sensels diffraction blur in diameter, not really optimal for resolution, shooting it at f/8 would have been significantly better. Raw converters also make a difference.
That the differences were further mitigated by using a good upsampling application is good for that application, but it doesn't allow to compare the camera's themselves well enough to make a quantitative judgement, other than that both can produce output that comes close, if the right software is used to blow the images up.
Cheers,
Bart
-
Sigma is not a 35mm , it can compete certainly with a Sony Nex 5
I never said it was, I said 35mm style.
Now if the thing shoots MF detail without moire and great colour, the size of the sensor hardly matters.
Kevin.
-
I will wait for real World production images before I decide wether the price is a bad joke or the bargain of the century. Mathematics never convince me one way or the other about a cameras capabilities or usefulness on a days work.
I hope it's bloody brilliant and I hope you can get an adapter to use Canon lenses. Why would anyone not wish it to be the best thing ever to capture light, I long for something new and exciting to hit the market, instead of the same 'ol same 'ol slightly better than the one it replaced offerings of everyone else.
Kevin.
-
14nx was a 14n that had a factory fitted sensor upgrade to SLR/n spec.
-
This. The most laughable thing is Sigma saying they're going to be grabbing medium format users....which means they're going after photographers that don't really know better. With that in mind, maybe this will be a huge hit with Ashton Kutcher and all the wealthy hobbyists?
A crop sensor 35mm body is NOT medium format no matter how you work the slide rule.
-
Excellent explanation, thanks. But in that case all future Nikon, Sony and Canon FF could be called digital medium format too.
Eduardo
Actual tests of the SD15 have found the resolution equivalent to an 8 or 9 megapixel Bayer pattern sensor so the SD1 will have the equivalent of a 24 to 27 megapixel Bayer sensor. 8 to 9 megapixels also happens to be the equivalent of 35 mm film so the SD1 can claim to be a medium format camera based on megapixels. Theoretical calculations have found the Foveon sensor to be equivalent to 2 times the resolution of a Bayer pattern in the red and blue and square root of 2 times a Bayer in the green so the SD1 will have the equivalent of a 22 to 31 megapixel Bayer sensor. The Foveon sensor gathers 3 times the light of a Bayer pattern so the SD1 has the area equivalent of 2.3 times a full frame Bayer sensor so SD1 can claim to be a medium format based on light gathering ability.
-
There's more going on with the SD1's launch price than meets the eye.
I'll have an article about it online later today.
Michael
-
There's more going on with the SD1's launch price than meets the eye.
I'll have an article about it online later today.
Michael
Very clear analysis, I think. I hope they will listen - it would be a pity to see this interesting technology go down the gully because of someones stubborness ...
-
Looking at how the Fuji X100 sold at a premium in relation to its specs, maybe some Sigma execs indeed believed they could pull this off too.
-
Excellent explanation, thanks.
Eduardo,
Unless you are being cynical, you were misinformed. There is almost nothing in those statements that's correct (as to why, see my earier response), which is indeed an accomplishment.
Cheers,
Bart
-
Irrelevant, because a Bayer CFA sensor captures Luminance at each sensel position (the spectral bands that are filtered out are calculated and added back in at demosaicing time), only attenuated by the peak transmission of the filter. The Foveon sensor also only uses 1/3rd of the spectrum per color, although it only loses light to the increased amounts of circuitry that produces a smaller fill factor, and the other unutilized photon energy.
Cheers,
Bart
Not following you here.
A theoretic, ideal Bayer sensor must still "throw away" a fraction of the light hitting the sensor area. In practice it will be affected by the selectivity of the color filters, but let us say that 2/3 is thrown away for simplicity.
A theoretic, ideal Foveon sensor need not "throw away" any light, since all light hitting a sensor area could be counted and classified as either red, green or blue.
As a camera is basically a photon-counter (counting where the photons hit, and approximate wavelength), counting all photons seen for a given lense sounds better than not counting them all.
We could compare the two sensors using identical sample grids, and conclude that the idealized Foveon sensor could do more spatial lowpass filtering to effectively remove noise while still having the spatial detail of the Bayer sensor + AA-filter. For simlicity we could assume a 2x2 pixel rectangular filter in the red and green channels, and a effective sqrt(2)xsqrt(2) pixel filter in the green channel.
In practice, limitations of current Foveon technology seems to make this theoretical exercise irrelevant. The problem is not the theoretical gains of having cosited subpixels, but making those subpixels small enough and still of sufficient quality and low enough price...
-
Looking at how the Fuji X100 sold at a premium in relation to its specs, maybe some Sigma execs indeed believed they could pull this off too.
I'm sure Fuji would claim the X100 hasn't a 'premium' price, but it's cost can be justified by it's unique spec and niche market.
For me it was too much and at least £300 in excess of what I'd consider a reasonable price. But if I'd really liked the feel and spec of it*, £300 would have been an affordable premium to pay for it. Michael's assessment of the SD1 being at least $8k too much seems about right and is simply an outrageous and unaffordable premium to pay for the camera.
Maybe Bernard's insight into the specialised nature of the market in Japan might be correct and Sigma's pricing strategy is squarely aimed for the yen paying gullible.
*If they'd got the size right I'd have been in the queue to buy one at an even higher price. I still want a digital version of a truly compact 35mm camera like the Fuji super DL mini/Minox 35/Rollei 35, but it still seems no nearer than it did five years ago :-(
-
Not following you here.
A theoretic, ideal Bayer sensor must still "throw away" a fraction of the light hitting the sensor area. In practice it will be affected by the selectivity of the color filters, but let us say that 2/3 is thrown away for simplicity.
A theoretic, ideal Foveon sensor need not "throw away" any light, since all light hitting a sensor area could be counted and classified as either red, green or blue.
It's simple, really.
Both a Bayer Color Filter Array (CFA) sensor and a Foveon sensor utilize 1/3rd of the spectrum (broadly speaking) for 1/3rd of the RGB output pixel data. Per output pixel,the CFA supplements the 2/3rd of the filtered spectrum by demosaicing, the Foveon has an actual recording of the Luminance in that 2/3rd of the spectrum.
So, e.g. a Green filtered CFA sensel conceptually receives the same amount of the spectrum (= Luminance) as the Green filtered Foveon layer. The only effective difference is in the transmission of the respective CFA filter in the spectal band versus the reduced Fill-factor of the corresponding spectrally sensitive Foveon layer, IOW the quantum efficiency. There is no 3:1 difference in sensitivity due to the technology itself.
Cheers,
Bart
-
Unfortunately I wasn't. I may be misinformed. Actually, I thought his info was accurate and that the form he wrote it "explained" it, and that's what I meant. My response was mainly to question him about his statement saying that the aps-c sensor in the SD1 is some sort of de-facto medium format. It is very possible that the tech data is more that I can understand, but for sure I know apples from oranges. :D
Eduardo
Eduardo,
Unless you are being cynical, you were misinformed. There is almost nothing in those statements that's correct (as to why, see my earier response), which is indeed an accomplishment.
Cheers,
Bart
-
Unfortunately I wasn't. I may be misinformed. Actually, I thought his info was accurate and that the form he wrote it "explained" it, and that's what I meant. My response was mainly to question him about his statement saying that the aps-c sensor in the SD1 is some sort of de-facto medium format. It is very possible that the tech data is more that I can understand, but for sure I know apples from oranges. :D/
Hi Eduardo,
No problem. The benefit of a forum like this is that opinions can/will be challenged, and it will benefit all.
There is a significant amount of disinformation about the Foveon sensor going around, I just try to demystify it with arguments and proof.
Cheers,
Bart
-
After reading Michael's analysis of the SD1 pricing blunder, I find little reason to disagree with him. However, I think the SD1 would have failed anyway -- it's simply too little, too late, even at $2,000. Sigma keeps trying to suggest that it's producing a pro-level camera with revolutionary technology, and reviewers keep saying, "No, you're not." The lenses aren't as good, the system isn't as complete, the software isn't anything special, in comparison to other pro-level competitors, or even the better prosumer cameras like Pentax. Although some are enamored with the sensor tech, it's more the *idea* of the thing, rather than actual performance, that seems to interest people. In terms of performance, Foveons never really kept up with more conventional sensors...not that that perhaps they couldn't have, it's just that in Sigma's hands, they haven't.
It made me wonder if some discouraged Sigma exec didn't do an analysis of the completed camera, and concluded, "You know what? We have just labored mightily and produced a POS, and everybody is going to know it, and it's gonna die an ugly death. Is there anything we can do to attract attention to this thing, to stir up some controversy?"
-
After reading Michael's analysis of the SD1 pricing blunder, I find little reason to disagree with him. However, I think the SD1 would have failed anyway -- it's simply too little, too late, even at $2,000. Sigma keeps trying to suggest that it's producing a pro-level camera with revolutionary technology, and reviewers keep saying, "No, you're not." The lenses aren't as good, the system isn't as complete, the software isn't anything special, in comparison to other pro-level competitors, or even the better prosumer cameras like Pentax. Although some are enamored with the sensor tech, it's more the *idea* of the thing, rather than actual performance, that seems to interest people. In terms of performance, Foveons never really kept up with more conventional sensors...not that that perhaps they couldn't have, it's just that in Sigma's hands, they haven't.
It made me wonder if some discouraged Sigma exec didn't do an analysis of the completed camera, and concluded, "You know what? We have just labored mightily and produced a POS, and everybody is going to know it, and it's gonna die an ugly death. Is there anything we can do to attract attention to this thing, to stir up some controversy?"
Lets give it a chance.
What if it is MF quality but without the moire, it could look a bargain? If it only matches top Canikon quality then they will have to give them out free with a Burger or something. Lets give it a few weeks before writing the obituary.
Not so long ago the The Leica "S" was going to drag the company under when it was announced according to many. The Sigma will hit the streets at much less than the price given out, I reckon they are just enjoying the attention, everyone is talking about and scratching their heads, so far it's great publicity.
Kevin.
-
so far it's great publicity.
I wouldn't call it 'great' publicity. More like a 'Firestorm'
Edit - DP Review has a hands on preview of the Sd1 now on their site
http://www.dpreview.com/previews/sigmasd1/
They offer little in the way of conclusions with this initial 'preview' - I guess more to come...
Initial Conclusion
The SD1 is a very straightforward camera to shoot, as you'd expect from its very still-image-focused feature set. Some of its operation is a little eccentric, such as the inability to directly access exposure compensation using one of the control dials in the A and S modes, but not in a hugely problematic fashion. We're also a bit surprised to see metering mode given its own button in preference to White Balance, and that its position is considerably easier to reach than the more regularly used ISO button.
It seems rather odd to have two distinct quick access systems, especially as neither of them makes great use having twin dials available (the dials remain dedicated to their shutter or aperture roles when in the Func. menu, and simply flick back and forth between the two screens of the QS menu). In fact it's only really in the menus (where the front dial scrolls up and down the menus and the rear dial scrolls left and right across menu tabs), that the camera's twin dials are really fully utilized.
Overall there's little to gripe about though - it's all pretty logical and well-behaved. However the ability to make a few customization decisions about dial operation, or to rededicate either the QS or Func button, would help the camera to suit even more tastes.
-
Not so long ago the The Leica "S" was going to drag the company under when it was announced according to many.
The difference is that Leica are a premium brand that could justify the price. There was also no doubt that any lenses made for the camera would be of the highest quality and allow the body to deliver it's full potential.
The Sigma will hit the streets at much less than the price given out, I reckon they are just enjoying the attention, everyone is talking about and scratching their heads, so far it's great publicity.
Great publicity ? I don't think so, they've made themselves look stupid and out of touch with this pricing, plus there's been a lot of negativity towards the brand overall. Do Sigma really want all their performance and quality control issues aired so publicly ? It certainly won't help their existing reputation and sales, let alone assist potential sales of the new body.
-
Perhaps this is an attempt at a 'clever' marketing stunt and we have not had the 'punch line' yet.....
Perhaps the idea is that the camera IS suposed to be massively overpriced,...and then carefully selected users will be granted the pivaledge to buy for 'only' 2000USD... That way they will be able to sell at what the camera is actually worth and have no shortage of buyers who will clamber for something they have been persuaded to believe is 'exclusive'.....
.....I can just hear the conversations at golf and photo club meetings;......."I had the chance to get the ten thousand dollar Sigma at Gold club price, so I just snapped one up to see what it's like".......
I know two people who use previous Sigma DSLR's and they are both retired business men who don't know one end of a camera from the other. let alone recognise image qualities,...but they both tell people how exclusive and wonderful their Sigmas are, and how much better than a Canon or Nikon could ever be. Perhaps the Gold Club price will be offered to existing Sigma users first,...I'm sure my two aquaintances would jump at the chance to buy.
-
The difference is that Leica are a premium brand that could justify the price. There was also no doubt that any lenses made for the camera would be of the highest quality and allow the body to deliver it's full potential.Great publicity ? I don't think so, they've made themselves look stupid and out of touch with this pricing, plus there's been a lot of negativity towards the brand overall. Do Sigma really want all their performance and quality control issues aired so publicly ? It certainly won't help their existing reputation and sales, let alone assist potential sales of the new body.
Leica are a premium brand alright, but it was still going to be there toll on the death bell.
I doubt I could have named a Sigma camera before this, now the World and his dog know about the SD1, what ripples did any of their other cameras make on release. Is there a forum not talking about the new Sigma in great length. Hopefully this is a new generation Foveon which might take things on a notch or two. The traditional camera makers don't exactly set the scene alight with innovation these days. I am certainly going to keep an eye on the new Sigma I doubt it will ever be priced on the shop floor anywhere near the announced ticket price. Give it a couple of days and it will be announced as some kind of mix up.
I quite like the idea of a high res camera with no AA filter and little chance of moire, where else can you get that? My biggest concern is the two Sigma lenses I own, don't exactly entice me to buy more of them.
Kevin.
Kevin.
-
It's simple, really.
Both a Bayer Color Filter Array (CFA) sensor and a Foveon sensor utilize 1/3rd of the spectrum (broadly speaking) for 1/3rd of the RGB output pixel data. Per output pixel,the CFA supplements the 2/3rd of the filtered spectrum by demosaicing, the Foveon has an actual recording of the Luminance in that 2/3rd of the spectrum.
So, e.g. a Green filtered CFA sensel conceptually receives the same amount of the spectrum (= Luminance) as the Green filtered Foveon layer. The only effective difference is in the transmission of the respective CFA filter in the spectal band versus the reduced Fill-factor of the corresponding spectrally sensitive Foveon layer, IOW the quantum efficiency. There is no 3:1 difference in sensitivity due to the technology itself.
Cheers,
Bart
Do you agree that (using your numbers) 1/3 of the photons are counted in a Bayer sensor, while all photons are counted in a Foveon sensor?
What you are saying is that the "per sensel" signal/noise level should (theoretically, ideally) be the same, and i agree. What I am saying is that the Foveon sensor would have 3x as many sensels for a sensor of same size and same pixel grid, therefore being able to measure a larger fraction of the signal (energy).
If you have the choice between doing election polls on a group of 100 individuals, or 3 groups of 100 individuals, which would be a less "noisy" prediction of a population of millions?
-h
-
Do you agree that (using your numbers) 1/3 of the photons are counted in a Bayer sensor, while all photons are counted in a Foveon sensor?
Yes, in principle that's the case due to the CFA. The remaining 2/3rds are not contributing to the same sensel position, although due to the AA-filter some is used in neighboring sensels. I left out that latter aspect earlier to not complicate the matter too much, but the Bayer CFA is really elegant in design and allows a significant level of reconstruction of the missing information (although not perfect, also due to different sampling densities of Green versus Red and Blue). Fortunately, chromatic information is generally lower spatial frequency data compared to Luminance, which mimics human vision a bit.
What you are saying is that the "per sensel" signal/noise level should (theoretically, ideally) be the same, and i agree. What I am saying is that the Foveon sensor would have 3x as many sensels for a sensor of same size and same pixel grid, therefore being able to measure a larger fraction of the signal (energy).
That's correct. An unfortunate side effect is of course that the uncompressed Raw data stream is 3x as large, burst shots require to move more data around (limiting the no. shots/sec), and memory cards fill up faster. On the plus side, the file does slightly better when upsampling to a larger size, because the chroma resolution is a bit higher.
If you have the choice between doing election polls on a group of 100 individuals, or 3 groups of 100 individuals, which would be a less "noisy" prediction of a population of millions?
If they were all monochrome samples, the more the merrier. However, we're dealing with filtered spectral bands here. Depending on scene content, some colors may have a low contribution with accompanying photon shot noise levels. And in the case of the Foveon sensor, the channel separation is very poor according to the Foveon documentation. This means that a lot of matrix math with huge negative coefficients is needed, which increases noise a lot. In addition, the fill factor of a Foveon sensor is rather small due to the additional connections and transfer gates (also needs to store 3 charges in the same small space), so the sensels saturate quickly. The comparison of noise performance is too complex to be explained with a simplified model. In practice though, the Foveon sensors have had limited high ISO quality compared to alternative technologies.
Cheers,
Bart
-
Yes, in principle that's the case due to the CFA. The remaining 2/3rds are not contributing to the same sensel position, although due to the AA-filter some is used in neighboring sensels.
If the AA filter is linear, space-invariant and wavelength invariant, the spreading should be equal into and out of any given sensel (area), meaning that no gain or loss occurs in terms of total number of photons counted?
If they were all monochrome samples, the more the merrier. However, we're dealing with filtered spectral bands here.
Let me rephrase. Would you rather sample the wavelength of 400 to 550 nm within a 2n*2m micron area by:
1. Measuring all photons within a n*m area (blocking out the remaining 3 n*m),
2. Measuring 1/4 of the photons within a 2n*2m micron area (hypothetical square AA-filter),
3. Measuring all photons within a 2n*2m area distributed on 4 sensels? (hypothetical foveon)
I would prefer the latter, as it would give me more photons.
The comparison of noise performance is too complex to be explained with a simplified model. In practice though, the Foveon sensors have had limited high ISO quality compared to alternative technologies.
Cheers,
Bart
I agree about all of your practical concerns about Foveon. My objections were about the crude idealized model and how it would (could) have worked.
-
More than often, the price defines the level of attractiveness of something.
My 2 bets are:
1. The SD1 will be a success in Japan,
2. Sigma will make more money with it selling for 9000 US$ than they would have with it priced at 2000 US$.
Indeed, I believe that most people interested in this camera in Japan don't really care about the price. They are looking for differentiation. The price actually helps them
Cheers,
Bernard
-
If this turns to be true that would be another reason to think that photography indeed is starting to suck!
Eduardo
More than often, the price defines the level of attractiveness of something.
My 2 bets are:
1. The SD1 will be a success in Japan,
2. Sigma will make more money with it selling for 9000 US$ than they would have with it priced at 2000 US$.
Indeed, I believe that most people interested in this camera in Japan don't really care about the price. They are looking for differentiation. The price actually helps them
Cheers,
Bernard
-
street price will be less just wait till they start shipping :)
-
street price will be less just wait till they start shipping :)
Yep, already at 630.000 Yen in Japan.
That's even smarter in fact... you have an exclusive camera with the best sensor in the world (claim I predict the owners will lay), one that is worth 10.000 US$, but you did in fact get it for a lot less... still you know people think it is worth 10.000 US$. :)
Cheers,
Bernard
-
Yep, already at 630.000 Yen in Japan.
That's even smarter in fact... you have an exclusive camera with the best sensor in the world (claim I predict the owners will lay), one that is worth 10.000 US$, but you did in fact get it for a lot less... still you know people think it is worth 10.000 US$. :)
One of the oldest tricks in negotiation: ask for 10 times what your outrageous goal is. All of a sudden your original outrageous goal doesn't sound too bad.
-
actually i think on b&h site its already lower and it hasnt shipped yet .. but lets be honest a $1000 camera body type even if its as good as a canon 5d mk2 the sensor as as good as a nikon d3s sensor at 7k$? i think thats the price on b&h site http://www.bhphotovideo.com/c/product/736546-REG/Sigma_SD1_Digital_SLR_Camera.html.. it has to beat 3 cameras i think in order to justify the price
the obvious leica m9 nikon d3s and maybe canon 5ds mk2 or 1ds3
honestly i think the price should be around 2k more or less .. just my 2 cents
-
Perhaps Sigma's theory is that there's no such thing as bad publicity... whatever they were thinking, there will be a LOT of photographers examining test photos and comparisons in microscopic detail.
-
Perhaps Sigma's theory is that there's no such thing as bad publicity... whatever they were thinking, there will be a LOT of photographers examining test photos and comparisons in microscopic detail.
Exactly, give it a month or two and it will be less the $2k. Great stunt Sigma I hope it's a world beater of a camera.
Kevin.
-
My theory is that a Sigma executive noticed that he had to wait months for his Leica M9 because demand exceeded supply at the $7000 price point. So he figured that Sigma could achieve something similar (and charge even more because this camera has autofocus!), but unfortunately his PowerPoint slide got flipped 180 degress and the whole sales structure is now based on supply exceeding demand. I'm assuming he has a good pension plan because people simply do not come up with ideas like a $9000 DSLR if they have to worry about their future.
Steve
-
Exactly, give it a month or two and it will be less the $2k. Great stunt Sigma I hope it's a world beater of a camera.
Kevin.
No way it'll go from 6.9K to 2K in two months. Maybe two years.
-
That's why my recommendation is that Sigma have to drop the price before the first unit ships.
If they don't, no one will buy at $10K, $9K, $8K or whatever because they will fear that the price will eventually come down and they'll have been a sucker.
Even the nut bar photographers in Japan who will buy anything for the sake of exclusivity will come back screaming at their dealers if the price drops by thousands, even a year later.
It's nor or never.
Michael
-
actually i think on b&h site its already lower and it hasnt shipped yet .. but lets be honest a $1000 camera body type even if its as good as a canon 5d mk2 the sensor as as good as a nikon d3s sensor at 7k$? i think thats the price on b&h site http://www.bhphotovideo.com/c/product/736546-REG/Sigma_SD1_Digital_SLR_Camera.html.. it has to beat 3 cameras i think in order to justify the price
the obvious leica m9 nikon d3s and maybe canon 5ds mk2 or 1ds3
honestly i think the price should be around 2k more or less .. just my 2 cents
A crop sensor resolving around the current full frame Canon cameras on the market like Canon 5D mkII, 1Ds mkIII is not bad at all. No other crop sensor can do that. A Canon 7D resolves about the same as the old 5D mkI according to DxO on the best Canon glass available. The Nikon D7000 resolves around the same as the D3s with the best glass. The Bayer sensors seem to resolve about 0.58 of full resolution if you take the resolution numbers from DPreview over the years. I think this number is pretty conservative given the latest RAW converters like ACR 6 and Lightroom 3 which can pull even more detail out from Bayer sensors. This should added together give a resolution of the SD1 around 19MP (full frame equivalent) in the best case on the best possible glas and less with standard zooms. This year we likely will see full frame cameras from Canon and Nikon in the 30MP+ range so there will be little to convince people to buy a Sigma SD1. All the reviews will say this I would expect.
After I looked at the posted photos on the Sigma site I'm not impressed and saying that they compete with a 5D mkII is even a stretch. RAW files may give a little change to the better as long as there are not color shift problems and color moire as on previous models.
So what does this say, if you agree with this analysis? That I agree with the poster here that the price should be no more the 2K and likely less than that. Given that you would need to buy into another lens mount and one that is used very little I think 2K is even very optimistic.
-
Sigma has already damaged themselves with the announced price and the marketing hype they try to generate about the X3 sensor.
A serious and reliable company wouldn't do that.
-
Nice article BTW, Michael. I've had a job in the past, having helped to engineer a great, usable product, and watch it get shot down due to poor marketing, (or actually a lack of it).
For a man whom makes a living behind the lens, and works in a company that would afford him any camera if asked, I still find that a crop sensor Canon does the job. 90% of what I do in the real world hits a double page spread in a brochure, or the web. The 7D does a fine job, and I have some impressive product shots that hit 36" as well - which suits our pull up banners. We no longer print larger than that, having dropped large prints for digital billboards and HD LCD monitors at trade shows, (we affectionately call them "digital posters").
For the most part, these semi-pro DSLRs have been hitting a level of quality that suits majority of our work (including video), and I'm hard pressed to see any real advantage without having to move to a phase one or mamiya, and then, at that level, I don't think that price is really an issue. So, while it may seem odd that someone of my position is found using a prosumer camera (I do have a dusty 1Ds mark I on the shelf), nothing beats a real world working knowledge of a good tool. I guess in a weird way that while price really doesn't matter, it doesn't mean that I have to spend the big bucks either.
-
What possible excuse is feasible to change the asking price in the age of instant communication? Other than admiting they plainly f...... up, Sigma will sustain the price. I think that under current technology, Sigma can't output enough sensors to consider this camera a typical production camera. It is more like an exotic chassis and body made in a garage-factory somewhere in Europe fitted with a production big block american V8. Well, in this case, just switch the exotic side.
In the meantime, they will obtain some revenues and experience they need for a new technology they're already developing. Another possible scenario.
Eduardo
-
I've had a fair amount of experience with Foveon sensor cameras. The first thing I can say is that sharpness is not equal to double a Bayer sensor camera. That is, unlike what Michael expects, this camera will not have the equivalent of 30 MP. It will be closer to 24 MP. Now, that's pretty good. But Sony and Nikon already have cameras out with that resolution. Later this year, Canon will up that, and it's possible that Sony and Nikon will as well.
But I've found a big problem with Foveon sensors. That is, they have poor color purity. Unlike Beyer sensors that rely on filters, the Foveon relies on the selective absorbson of light that silicon exhibits as the light passes through the chip. That is, as the light passes further through, different colors are absorbed, allowing each color sensor to reside behind the one in front. This is what makes the Foveon possible. But, that absorbson isn't perfect, hence, colors poison the other layer sensors, resulting in color mixing that simply can't be fixed.
In addition, the sharpness is higher in some colors, but lower in others, when comparing to Bayer sensors. While Foveon sensors have a higher resolution in the red and blue, it's actually lower in the green, again, when compared to Bayer sensors. That's entirely because Bayer sensors have two green sensing sites to one of each red and green, while Foveon sensors have equal sites for all colors. We are most sensitive to green, specifically a green-yellow. So, in practice, the sharpness advantage of Foveon sensors is highly dependent on the subject matter. Sometimes it appears a lot sharper, and sometimes less so.
When I first read about this camera, I thought that Sigma was making a pro model, in the way we think of pro models, and so even though the price was crazily high, I though it MIGHT be worth it for some. But upon finding that it's just a mid range advanced amateur body, I'm rather shocked. While their high end lenses are good, the rest are just what they seem to be. Unless they up their QC by a good bit, this entire senario is doomed, even if they lower the price.
-
The SD1 should be great as a pure BW-camera. All the disadvantages of the Foveon sensor mostly vanish in this case. This does, of course, assume that the software does the right thing and does not add a lot of noise before converting to BW.
There are a number of issues with your post, but I'll just address that last. The Foveon gains no advantage as a B/W sensor. Used that way, it's just a medium resolution 15 MP sensor.
-
Actual tests of the SD15 have found the resolution equivalent to an 8 or 9 megapixel Bayer pattern sensor so the SD1 will have the equivalent of a 24 to 27 megapixel Bayer sensor. 8 to 9 megapixels also happens to be the equivalent of 35 mm film so the SD1 can claim to be a medium format camera based on megapixels. Theoretical calculations have found the Foveon sensor to be equivalent to 2 times the resolution of a Bayer pattern in the red and blue and square root of 2 times a Bayer in the green so the SD1 will have the equivalent of a 22 to 31 megapixel Bayer sensor. The Foveon sensor gathers 3 times the light of a Bayer pattern so the SD1 has the area equivalent of 2.3 times a full frame Bayer sensor so SD1 can claim to be a medium format based on light gathering ability.
That's not even close to being accurate in practice other than for sharpness. I've found, in using these cameras, that a 50% increase in sharpness over the single layer MP number is about right. The cameras have had poor sensitivity overall, no matter what some theoretical numeration says.
This will be in no way equal to a medium format model, and likely not equal to the current top Nikon.
-
I have a SD9, Sd14 and DP1. My own testing of the SD9 against my 350D showed convincingly that was essentially no visible resolution difference between them (there were visible differences in the type of aliasing artifacts however, which was fun to evaluate).
Likewise, in my Sd14 vs Kodak 14nx article that is quite well known, I showed that in prints up to 24x16 inches there is for all practical purposes, no resolution difference.
Overall, I'm happy with a rule of thumb that says the Sigma cameras have a resolution that roughly, approximately on a par with Bayer sensors of around twice the spatial pixel count. It's what Sigma themselves think when they are not busy marketing it at 3x the count.
Of course, there'll always be something to quibble over and someone to produce a piece of academic research that says differently, but for practical everyday shooting purposes, testing shows the 2x ratio is a reasonable rule of thumb for rough equivalence.
While I haven't read your "well known" article (or for that matter, heard it it), my experience is that it never exhibits twice the resolution of a bayer sensor. I don't know your testing methods, but I ran a fair sized commercial photo lab for a long time, and with my equipment, I ran more than a few tests, and prints. My evaluation of Foveon sensor cameras was "meh!".
I'd like this one to be much better, even though the price is a joke.
-
I don't know your testing methods
and what are yours ? because so far it is opinions only
-
sigma is not about selling xxx amount of units like canon /nikon/sony if they sold 500 units a year they are happy sort of its about being "different" then the canikon/sonys the question is how many or who will spend 6-8k to be different 6k-8k on a less than 1k type of body
we shall find out soon i guess
-
After thinking about the X3 sensors principle I came to the conclusion there is a flaw inherent in the design which will limit its use on the long run:
Its simple: The sensor produces huge files without an according gain in image quality.
The Bayer array sensors have a large ability to gather (selected) luminance data and a limited collection of chrominance data.
Basically this leads to a sort of data compression which is similar to the principle of jpegs: compress the color, keep the luminance.
What appears to be faulty at the first look comes out as the greatest strength of this system:
Getting the most relevant data in appropriate amounts to generate maximum quality images with minimum loss in quality by a clever hardware based compression.
Said differently: The average amount of information per byte of data in a file produced by a bayer pattern sensor is probably much better than in a foveon based system, because of the hardware inherent reduction of information during capture due to the bayer pattern.
This means less disc space, faster storage on the memory card, higher possible framerate.
So - my impression now is, that they try to sell as a magic bullet what really is an incumbering mass of irrelevant data no one really needs.
q.e.d.?
-
That's why my recommendation is that Sigma have to drop the price before the first unit ships.
That's a logical idea however, IMHO there is zero chance it'll happen.
-
Said differently: The average amount of information per byte of data in a file produced by a bayer pattern sensor is probably much better than in a foveon based system, because of the hardware inherent reduction of information during capture due to the bayer pattern.
True, but then any raw file format performs lousy compared to in-camera jpeg if image quality per byte is important.
The reason why we use raw is that it gets us a little more quality and flexibility, at a large filesize cost. Since storage and bandwidth is cheap compared to the number of images and frames per second that most of us need, this tradeoff makes sense.
By this logic, a Foveon sensor that actually gave significant improvements in image quality could still be worthwhile even if it resulted in 3x the filesize.
-h
-
There are a number of issues with your post, but I'll just address that last. The Foveon gains no advantage as a B/W sensor. Used that way, it's just a medium resolution 15 MP sensor.
Well, it does not have an AA filter like pretty much all Bayer sensors. Most B&W users would consider that an improvement. Further, demosaicing algorithms may sacrifice some luminance acuity in order to reduce chances of color artifacts. Foveon avoids this.
-h
-
True, but then any raw file format performs lousy compared to in-camera jpeg if image quality per byte is important.
The reason why we use raw is that it gets us a little more quality and flexibility, at a large filesize cost. Since storage and bandwidth is cheap compared to the number of images and frames per second that most of us need, this tradeoff makes sense.
By this logic, a Foveon sensor that actually gave significant improvements in image quality could still be worthwhile even if it resulted in 3x the filesize.
-h
Exactly.
And that is, what I believe is not delivered for standard situations.
If you want maximum quality on a minimum size sensor that might be a different story.
But for normal photographic application I don't believe the three times bigger files are justified for the gain in sharpness, together with the reported color problems.
I'd prefer to have more "untrue" Bayer-pixels, than 3 times less "true" Foveon pixels.
-
and what are yours ? because so far it is opinions only
Do you want me to list the entire procedure? I will give you a quick idea of how I tested cameras that manufacturers used to give me to check out.
We had a room for the purpose of these tests. In the room was a heavy rolling camera stand on tracks, similar to our 8 x 10 Durst mural printer. This ensured that the camera was always parallel in the horizontal AND the vertical directions. Lighting supplied by four 1200 WS strobes with 36" soft boxes. Various targets were used, depending on the nature of the test. Both color and resolution tests were performed.
As a lot of this is lens dependent, the choice of lenses is important. Aproblem is that not all manufacturers have an exact equivalent of another manufacturer, or, if they do, the quality may not be the same. So that involves testing the lenses as well for sharpness, distortion, color, off centered elements, lack of parallelism to and from the camera body, etc.
We didn't do MTF or other testing of that type, because the photographic quality is more important to me than the numbers themselves.
When printing, a 44" Epson was used for various print sizes. Though we didn't always print a full 44" print. Usually once we went to 20 x 30, we would print out portions of the image at greater magnification. It's rarely needed to print to this size for D-SLR's because of the resolution of those cameras. But for medium format, it's different. I've been beta testing Leaf backs since they first came out, and the latter models surely can print up to that size. I'm astounded as to what a difference 80MP can make in a print, even a smaller one.
That's the essentials. As usual, life is more complex, and so were the tests very often, especially if we found what looked to be an anomaly.
But testing did include taking actual photo's with the cameras, both in our studio, and out in the "real"world".
-
Do you want me to list the entire procedure? I will give you a quick idea of how I tested cameras that manufacturers used to give me to check out.
We had a room for the purpose of these tests. In the room was a heavy rolling camera stand on tracks, similar to our 8 x 10 Durst mural printer. This ensured that the camera was always parallel in the horizontal AND the vertical directions. Lighting supplied by four 1200 WS strobes with 36" soft boxes. Various targets were used, depending on the nature of the test. Both color and resolution tests were performed.
As a lot of this is lens dependent, the choice of lenses is important. Aproblem is that not all manufacturers have an exact equivalent of another manufacturer, or, if they do, the quality may not be the same. So that involves testing the lenses as well for sharpness, distortion, color, off centered elements, lack of parallelism to and from the camera body, etc.
We didn't do MTF or other testing of that type, because the photographic quality is more important to me than the numbers themselves.
When printing, a 44" Epson was used for various print sizes. Though we didn't always print a full 44" print. Usually once we went to 20 x 30, we would print out portions of the image at greater magnification. It's rarely needed to print to this size for D-SLR's because of the resolution of those cameras. But for medium format, it's different. I've been beta testing Leaf backs since they first came out, and the latter models surely can print up to that size. I'm astounded as to what a difference 80MP can make in a print, even a smaller one.
That's the essentials. As usual, life is more complex, and so were the tests very often, especially if we found what looked to be an anomaly.
But testing did include taking actual photo's with the cameras, both in our studio, and out in the "real"world".
I see... "We didn't do MTF or other testing of that type"... so that is just your visual impressions... somebody here was able to see the infamous "6 stops" more DR... did you see it too ?
-
Well, it does not have an AA filter like pretty much all Bayer sensors. Most B&W users would consider that an improvement. Further, demosaicing algorithms may sacrifice some luminance acuity in order to reduce chances of color artifacts. Foveon avoids this.
-h
It has to be remembered that 15MP is the province of under $1,000 cameras today, and that cameras costing just a fraction of this one are already at 24MP, and will likely go higher late this year, or early next. The antialising filters used are getting weaker with every model of higher resolution camera that comes out. The $10,000 Pentax 40MP model doesn't use one, and neither does the 18MP Leica.
The truth is that Foveon camera have moire as well. They don't have the easy to get rid of color moire, but they are subject to luminance moire, as every other digital camera is. The higher the resolution the camera, the less the moire problem is. I found it to be almost impossible to get a moire, even with a bridal gown and veil, usually a major cause, along with men's suites and jackets, with the new Leaf 80MP back.
I've a feeling that as D-SLR resolution goes higher, we will see even weaker filters, and eventually, none.
-
Exactly.
And that is, what I believe is not delivered for standard situations.
If you want maximum quality on a minimum size sensor that might be a different story.
But for normal photographic application I don't believe the three times bigger files are justified for the gain in sharpness, together with the reported color problems.
I'd prefer to have more "untrue" Bayer-pixels, than 3 times less "true" Foveon pixels.
Yes, I am sceptical about the quality.
If one wants the best possible quality for a given sensor size (for cost, size, weight, compability reasons), then perhaps Foveon or something Foveon-like will be the answer sometime in the future, as all photons can in principle be counted (unlike Bayer that must discard 1/2 or 2/3 or something)
For most of us it is more about the best quality per dollar, or the best quality within a system that we are already heavily invested in, or a trade-off between IQ, ergonomy, features, etc.
-h
-
I see... "We didn't do MTF or other testing of that type"... so that is just your visual impressions... somebody here was able to see the infamous "6 stops" more DR... did you see it too ?
MTF tests are much more used for the testing of lenses, not camera bodies. One of the biggest problems in testing bodies are the faults of the lenses themselves. But when testing bodies, I didn't see a purpose in testing for lens defects. Michael and a number of other professionals who have sites such as this one are not enamored of doing technical tests such as MTF tests. They would rather take the camera and use it. I agree to a great extent. There's nothing like taking actual pictures.
You can demean the idea of looking at pictures, and so demean everything Micheal does here when he evaluates cameras and lenses. That's not helpful. Nor is it meaningful. We shot tests in our lab that were repeatable, and would tell us all we needed to know about the items under test. This was carefully done, and repeatable, and used targets that were built up for that purpose.
And exactly what did you do?
-
It has to be remembered that 15MP is the province of under $1,000 cameras today, and that cameras costing just a fraction of this one are already at 24MP, and will likely go higher late this year, or early next. The antialising filters used are getting weaker with every model of higher resolution camera that comes out. The $10,000 Pentax 40MP model doesn't use one, and neither does the 18MP Leica.
I might have misinterpreted you, but your post gave me the impression that you thought that a 15 megapixel Foveon sensor would perform like a 15 megapixel Bayer camera when shooting b&w. I believe that an AA-filter less 15 mp Foveon performs like an a*15 mp sensor Bayer when shooting B&W. "a" is some manufacturer/model-dependent scaling factor >1.
The truth is that Foveon camera have moire as well.
Of course. Nyquist tells us that all digital cameras will need AA-filters that are shaped like sin(x)/x of infinite extent to remove all aliasing in the general case while having maximal passbandwidth.
Luckily we can usually settle for "looks good to me".
-h
-
Interesting way you have of inverting the colour resolution issue :) Foveon sensors have stable resolution across all colours, Bayer CFA sensors have variable resolution across colours. You can see this quite clearly in Mike Chaney's (ddiSoft, makers of Qimage) comparison with his 5D using colour resolution charts. It is the 5D that struggles to match the SD14 resolution. Admittedly this is a worse case, using the kind of highly saturated near primary colours of the flowers he uses but it makes the point, even if it is rare to hit that exact situation.
Personally, I wouldn't go down the route of criticising the Foveon colour resolution, I'd focus more on its tendency to show yellow/green casts that are awkward to dial out, occasional problems with bright reds and some units seem to suffer "Italian flag" green edges.
I've had a fair amount of experience with Foveon sensor cameras. The first thing I can say is that sharpness is not equal to double a Bayer sensor camera. That is, unlike what Michael expects, this camera will not have the equivalent of 30 MP. It will be closer to 24 MP. Now, that's pretty good. But Sony and Nikon already have cameras out with that resolution. Later this year, Canon will up that, and it's possible that Sony and Nikon will as well.
But I've found a big problem with Foveon sensors. That is, they have poor color purity. Unlike Beyer sensors that rely on filters, the Foveon relies on the selective absorbson of light that silicon exhibits as the light passes through the chip. That is, as the light passes further through, different colors are absorbed, allowing each color sensor to reside behind the one in front. This is what makes the Foveon possible. But, that absorbson isn't perfect, hence, colors poison the other layer sensors, resulting in color mixing that simply can't be fixed.
In addition, the sharpness is higher in some colors, but lower in others, when comparing to Bayer sensors. While Foveon sensors have a higher resolution in the red and blue, it's actually lower in the green, again, when compared to Bayer sensors. That's entirely because Bayer sensors have two green sensing sites to one of each red and green, while Foveon sensors have equal sites for all colors. We are most sensitive to green, specifically a green-yellow. So, in practice, the sharpness advantage of Foveon sensors is highly dependent on the subject matter. Sometimes it appears a lot sharper, and sometimes less so.
When I first read about this camera, I thought that Sigma was making a pro model, in the way we think of pro models, and so even though the price was crazily high, I though it MIGHT be worth it for some. But upon finding that it's just a mid range advanced amateur body, I'm rather shocked. While their high end lenses are good, the rest are just what they seem to be. Unless they up their QC by a good bit, this entire senario is doomed, even if they lower the price.
-
"Well-known" is a relative term in this case. What I mean is "well-known" amongst people interested in reading and writing about Foveon technology in the sense that it gets referenced/quoted quite often in such discussions. Here's a link: http://www.whisperingcat.co.uk/scans/sd14vs14nx.htm
If you have the energy to read it, you'll notice that it is a little different than a lot of reviews/comparisons because it was done using actual prints rather than pixel peeping onscreen. No test methodology is perfect, especially a field test like this one but we did our best to make it meaningful by verifying with our own eyes what the scans show. And I stick by our conclusions even today: in the context of prints up to 24 x 16 inches, there is no practical resolution difference between the Sd14 and the Kodak 14nx. Dpreviews tests of the DP series compacts that use the same sensor also agree that there is no detail advantage to 12MP Bayers over the 4.6MP Foveon. So, I'm happy with the rough rule of thumb that Foveon resolves approximately the same as a Bayer sensor of twice the pixel count.
Despite the above comments, I'm still unconvinced of the merits of the Sigma cameras. They seem to be still too prototype-y for my taste. I had hope that the Sd1 might be the breakthrough product that would take Foveon mainstream but alas it is not to be so. Maybe next time.
While I haven't read your "well known" article (or for that matter, heard it it), my experience is that it never exhibits twice the resolution of a bayer sensor. I don't know your testing methods, but I ran a fair sized commercial photo lab for a long time, and with my equipment, I ran more than a few tests, and prints. My evaluation of Foveon sensor cameras was "meh!".
I'd like this one to be much better, even though the price is a joke.
-
Hmm. In the study you linked to, the Kodak 14nx looks much better to me. It doesn't seem that close.
-
Interesting. When I lay the 24 inch prints out on the floor, I have no idea which is which. They look absolutely identical to me and as they were supplied unmarked I don't know which is which.
In all the time that article has been up, I have invited people to identify the unmarked samples - to date no one has done it.
-
Interesting. When I lay the 24 inch prints out on the floor, I have no idea which is which. They look absolutely identical to me and as they were supplied unmarked I don't know which is which.
In all the time that article has been up, I have invited people to identify the unmarked samples - to date no one has done it.
Hi Dave,
A quick glance suggests an inverted order compared to the smaller prints, unless each pair is in different orders, I'd then have to look at each pair individually and score them per pair. The difference is not huge though. I also see the differences in the smaller prints.
Cheers,
Bart
-
Interesting way you have of inverting the colour resolution issue Smiley Foveon sensors have stable resolution across all colours, Bayer CFA sensors have variable resolution across colours. You can see this quite clearly in Mike Chaney's (ddiSoft, makers of Qimage) comparison with his 5D using colour resolution charts. It is the 5D that struggles to match the SD14 resolution. Admittedly this is a worse case, using the kind of highly saturated near primary colours of the flowers he uses but it makes the point, even if it is rare to hit that exact situation.
Ah. You beat me to it. :)
Sigma's marketing idiocy aside, the sensor has a true 14 million pixels. This is like on your monitor - a pixel is a full-color range location. Same as film if you are scanning it. The problem is compounded by the fact that Bayer pattern sensors, which are sub-pixels are being marketed as the real thing. So we end up with marketing claims that are as clear as mud for the most part.
The only real advantage of the Sigma is that the color channels have identical resolution. This produces very good and realistic colors and a very film-like shoulder as well. There also are no real issues with moire from the color, since there is no real "pattern" for the camera or software to interpolate. So it takes very very good scenery shots.
But it has a fatal flaw in it, like film, because it works like film does. It shoots at one ISO only. Anything other than the base ISO is a bunch of computer trickery. So it is pathetic if you push it more than a stop or two, just like film. Given that the "base" ISO seems to be about 50-100 on it, it's exactly like throwing some Velvia in your camera and trying to use it for absolutely everything. Slow shots will look stunning. Action and low light... yeah... not going to happen.
So it's really the world's best scenery DSLR camera. And for that, if it were priced at a reasonable $2000 or so, it would be a tremendous tool for people who do that sort of thing. (such as myself) But at that price? Not going to touch it.
Me? I'm a bigger fan lately of the Fuji sensors. They are inexpensive and their pixel binning and diagonal pattern combined with their HDR mode that brackets and blends inside the camera essentially gives you the cost advantages of a cheap Bayer sensor but pretty much none of the problems. (especially of note is the way it doesn't completely wash out over-exposed areas) The pictures it takes are *this* close to the Sigma. But the price, well, Fuji's putting that top-end EXR sensor in under $500 cameras.
Given Sigma's so-so glass, the sad thing is that the Fuji HS20EXR has a similar cropped sensor but a very nice lens - and will probably work 80-90% as well as the Sigma SD1 in terms of output and 200% better in terms of action and low light shots. Even without interchangeable lenses. For $500.
Verdict: Love the Sigma. Will buy the Fuji instead. Doubly so if Fuji ever releases a DSLR version of that thing.
EDIT - before everybody jumps all over me, let's do a little "napkin" math:
- The Sigma sensor is being calculated to come in at about 30MP equivalent. Or about 25% better than a 24MP DSLR.
- If you calculate out the geometry and interpolation losses in a Bayer sensor, and then factor in the AA filter, you end up with roughly a 0.65-0.7x multiplier for most DSLRs in each dimension. This gives you a rough idea of how many actual full-color locations you can expect. This allows for easier comparison between technology types.
- The Sony 24MP DSLR comes in at about 11 million actual pixels. The Sigma has 14 million. Or about the same 25% difference.
- The Fuji at 16MP has a closer to 0.8X multiplier (give or take) and comes in at about 11MP as well. (scanned 35mm film is closer to 3-4MP, btw)
- The Sigma has a bit over 14MP and a 1.0x multiplier.
What this means is that the cameras are all about the same other than that 25% bigger picture/resolution for the Sigma. But I think Sigma's glass is 25% *worse* than the competition, making it a wash versus the three technologies. Ouch. (actually the math hurts Sigma more, as worse glass has a myriad of other issues with it in terms of image quality aside from absolute resolution)
$500 bridge camera does almost exactly as well as Sigma's new toy. Man, that's got to hurt... ;D
-
The only real advantage of the Sigma is that the color channels have identical resolution. This produces very good and realistic colors ...
If by good and realistic colors you mean that the technology is well suited to recreate perceptually the color "as seen", then I think that you are very wrong. I have seen suggested that the reason why earlier Sigma cameras did not have in-camera JPEG was because the lousy color properties of the Foveon sensor meant that they could not do the heavy processing inside a camera with limited processing capabilities.
http://forums.dpreview.com/forums/read.asp?forum=1000&message=38502539&q=fossum&qf=m
The Foveon sensor, contrary to kool-aid purveyors, does not measure RGB information. It measures the photons absorbed in three different layers of silicon. Only in cartoon-land do the three layers correspond to R,G and B. In fact, R, G and B is invented MORE for a Foveon sensor than R,G, B interpolation for a Bayer sensor, at least in my book. Besides the immense amount of processing required to generate a semblance of R,G and B for the Foveon pixel, the noise floor on the pixel is quite high due to not being able to do correlated double sampling with complete charge transfer. Not as important under bright light conditions, but ugly under low light.
http://forums.dpreview.com/forums/readflat.asp?forum=1027&message=36397198&changemode=1
This paper from Richard Lyon at Foveon explains some of that.
http://www.foveon.com/files/CIC10_Lyon_Hubel_FINAL.pdf
...
But the big problem with that chart is that it's plotted all pretty in red, green, and blue, and the Foveon outputs aren't really red, green, and blue. Look at the spectral sensitivity curves, figure 7. There's actually wavelengths where a photon can show up equally in the top, middle, or bottom layers of the sensor. So. light of that wavelenght isn't interpreted as a color, it's interpreted as "white". One stimulus that can produce outputs in all the channels is "bad" from a signal processing standpoint. That's called "mutual information", and you have to uncouple the mutual parts and extract the "separable" information before anything makes sense.
If you actually look at the outputs of a Foveon sensor, the three outputs are more like white, yellow, and red. The top layer is all mutual information, you have to make the green and red that "contaminate" it go away before you can see the green.
...
I you mean that the color of processed earlier Sigma/Foveon sensors appealed to you, just like the colors of certain film manufacturers appeals to many, then that is a subjective opinion that one cannot argue.
EDIT - before everybody jumps all over me, let's do a little "napkin" math:
- The Sigma sensor is being calculated to come in at about 30MP equivalent. Or about 25% better than a 24MP DSLR.
- If you calculate out the geometry and interpolation losses in a Bayer sensor, and then factor in the AA filter, you end up with roughly a 0.65-0.7x multiplier for most DSLRs in each dimension. This gives you a rough idea of how many actual full-color locations you can expect. This allows for easier comparison between technology types.
- The Sony 24MP DSLR comes in at about 11 million actual pixels. The Sigma has 14 million. Or about the same 25% difference.
- The Fuji at 16MP has a closer to 0.8X multiplier (give or take) and comes in at about 11MP as well. (scanned 35mm film is closer to 3-4MP, btw)
- The Sigma has a bit over 14MP and a 1.0x multiplier.
Then you are ignoring the physics of how natural scenes tend to reflect light, and the way the human visual system works. If you can ignore those two factors, what is the application for a camera for you? Some industrial assembly-line quality control system?
-h
-
The interpolation losses for each type of sensor are something that you can't overcome, though. Bayer "pixels" are really not full-color locations comparable, to say, pixels on a monitor, printer, or a scanner. So there are some unavoidable losses, which means that 24MP Bayer equals about 16MP Fuji. The sigma is roughly equal to 30MP Bayer.
We're talking about minor differences between the two other than lens quality and features(which the DSLRs do win vs Fuji, no question about), but a ~12% in each dimension difference between the first two and the Sigma. Like comparing 1440*960 to 1680*1050 on your monitor. A slight difference, really. What it comes down to is the optics, and to be honest, I think that the Fuji lens is not any worse than the stuff Sigma puts out.
80%+ the quality of the Sigma for under $500. Usability isn't a factor as the Sigma is *not* a pro camera in features and ease of use. Well, the past models weren't. Maybe this one will be different... *laugh* (wow the past software was horrendous to use...)
It really is becoming who has the better assembly line at this point. The Sony is a tremendous bargain at just under $2000 and with its better lenses will easily equal or beat the Sigma. Sigma needed to price it at no more than 50% more than the Sony to stand a chance.
In any case, arguing aside, the thing is dead in the water. Maybe the talk of Sony coming out with their own take on the technology will pan out. At least they will do a good job of it.
-
There are a number of issues with your post, but I'll just address that last.
Oh please do address the others.
The Foveon gains no advantage as a B/W sensor. Used that way, it's just a medium resolution 15 MP sensor.
I mean, it has the potential to be awesome 15MP B/W camera. Or you could say, an awesome APS B/W camera.
This obviously changes if you compared it to a modern pro-grade full-frame 35mm DSLR.
However, having played with a Phase Achromatic back, I can definitely say that a B/W camera where each pixel records a B/W value is great to work with.
I also stand by my comment that correcting a lens for chromatic aberration should be a lot easier with a Foveon than with a Bayer sensor. This will, eventually, be a limiting factor for the competition regardless of pixel size.
-
:o
Sigma won't sell many of these at the ridiculous MSRP (or for that matter the still-astronomical "street" price related thereto).
The Foveon sensor tech has always been interesting to me, but they still haven't got it to a big enough format (i.e., 24x36 mm) to interest me. It's a joke that they think a less-than-half-frame format camera is going to compete with digital medium format. Sensor tech does not circumvent the laws of physics, after all, and there's no way that an APS-C format camera will be able to resolve 46MP or 30MP of detail, because the limits imposed by diffraction and practical lens resolution will cap it at something more like its actual photosite count (or less) in most cases.
How many (more like how few) photographers will tolerate a mid-range quality body with a relatively modest frame rate and buffer and a crappy interface (read: an APS_C viewfinder, and a less-than-100% one at that, without even the consolation prize of "Live View") at medium format/professional full frame format prices?! I suspect not many.
Another Achilles Heel is the proprietary Sigma lens mount. If the thing has ANY chance of selling in meaningful numbers, they should make one in each of Nikon and Canon mounts and ditch the proprietary Sigma mount, or make all three ONLY if they sell enough Sigma-mount cameras and lenses to justify it. Nobody with meaningful investments in Nikon or Canon gear (and the kind of money to afford cameras in this price range) is going to switch brands for a mid-range APS-C body camera with supposedly better than par image quality (the jury's still out on that, but we'll see) in what will probably be a more limited range of ISO values, given the Foveon track record.
-
Theoretically, Foveon sensors have the potential for 3 times the resolution of a Bayer pattern sensor when both sensors are of equal size. Of course that's theory and you can never actually get to 3X. But 2X is very believable. I think these sensors have fantastic potential, but frankly I was unhappy when Sigma bought the intellectual property rights to this technology. In the hands of Nikon or Canon I think we'd now have much better cameras than we have now. But I don't trust Sigma quality based on reports I've read, and I've got a ton of money invested in non-Sigma lenses that I'm not willing to give up. Even putting all that aside, I'd never pay what they're asking for these cameras. In all, their pricing and proprietary lens mount make it a no-go.
-
Theoretically, Foveon sensors have the potential for 3 times the resolution of a Bayer pattern sensor when both sensors are of equal size. Of course that's theory and you can never actually get to 3X. But 2X is very believable. I think these sensors have fantastic potential, but frankly I was unhappy when Sigma bought the intellectual property rights to this technology. In the hands of Nikon or Canon I think we'd now have much better cameras than we have now. But I don't trust Sigma quality based on reports I've read, and I've got a ton of money invested in non-Sigma lenses that I'm not willing to give up. Even putting all that aside, I'd never pay what they're asking for these cameras. In all, their pricing and proprietary lens mount make it a no-go.
http://www.luminous-landscape.com/tutorials/resolution.shtml
The resolution limit for APS-C sensors (as respects middle spectrum color) for diffraction-limited lenses at f8-f11 (the range that is a reasonable expectation for mass-produced lenses) is thirteen megapixels for f8, and seven megapixels for f11, based on physics. These limits won't be exceeded due to sensor technology, and irrespective of Foveon vs. Bayer "relativity," those maximum resolution limits are here to stay. When Foveon sensors were sporting less than 5 megapixels (of three layer pixels), the headroom, if you will, existed for the effective difference in resolving power to be seen (in terms of Foveon resolving like a higher megapixel count Bayer sensor). When you increase the megapixel count of the Foveon sensor to a point where you already are constrained (in terms of maximum potential resolution) by diffraction limits, you simply won't be getting more "resolution" based on Foveon vs. Bayer technology. We're already at the overkill level for megapixel count on APS-C sensors (at the 16MP level) with Bayer sensors. The real potential of Foveon will not be seen without a bump in sensor size (a 35mm Full Frame Foveon would have resolution limits of 29 megapixels for lenses "diffraction limited" at f8, and 16 megapixels for lenses "diffraction limited" at f11, which would allow more room for additional real-world resolution gains).
I was actually also kind of bummed when Sigma bought the Foveon technology, since it locks up the product in a minority lens mount from a small-time manufacturer that simply isn't going to make the best use of it. I agree that we would probably see much more enticing offerings with this sensor technology (and that some of the issues with it would have a much better chance of being resolved - pardon the pun) in the hands of the "Big Two." Sigma's price selection for this camera has turned the whole thing into something of a freak show, and their reasoning for the justification of the price is a joke.
-
I like your name 24X36, honoring the only true full frame sensor up to this date. All others are some sort of FF just because they were shrunk to accommodate the then new incipient technology.
If Nikon had bought the Foveon technology and developed according to his capabilities, I just imagine the tete a tete with Canon and now with Sony!
Sigma hasn't been of much help for this technology. Too many drawbacks. Sigma should seek big-time investors and sell sensors like Sony do.
Eduardo
http://www.luminous-landscape.com/tutorials/resolution.shtml
I was actually also kind of bummed when Sigma bought the Foveon technology, since it locks up the product in a minority lens mount from a small-time manufacturer that simply isn't going to make the best use of it. I agree that we would probably see much more enticing offerings with this sensor technology (and that some of the issues with it would have a much better chance of being resolved - pardon the pun) in the hands of the "Big Two." Sigma's price selection for this camera has turned the whole thing into something of a freak show, and their reasoning for the justification of the price is a joke.
-
Theoretically, Foveon sensors have the potential for 3 times the resolution of a Bayer pattern sensor when both sensors are of equal size. ...
According to what theory, and what kind of resolution? Foveon for certain does not generally have the potential for 3x the number of lp/ph for monochrome input. For most applications (such as human viewers), the luminance resolution is by far the most important one.
-h
-
I like your name 24X36, honoring the only true full frame sensor up to this date. All others are some sort of FF just because they were shrunk to accommodate the then new incipient technology.
Eduardo
Thanks. :)
I've always disliked the "digital shrinkage" of formats to deal with issues of cost, and the attempt to pass the shrinkage off as an "advantage," though I have to admit the marketing people did do a heck of a job with that (the legions that parrot the marketing campaigns as gospel speaks volumes of their success in that regard). I'd like to see more effort focused (pardon the pun) on making larger imaging chips with lower cost and higher yield (which go hand-in-hand, of course), so that all of the formats can be restored to their actual film equivalents.
-
I'd like to see more effort focused (pardon the pun) on making larger imaging chips with lower cost and higher yield (which go hand-in-hand, of course), so that all of the formats can be restored to their actual film equivalents.
Given laws of economics and physics, those efforts will also benefit smaller sensors. And given the mathematics involved, they benefit smaller sensors more than larger ones - ie. improving yields of larger sensors by 10% will improve yields of smaller sensors by >10%.
-
Thanks. :)
I've always disliked the "digital shrinkage" of formats to deal with issues of cost, and the attempt to pass the shrinkage off as an "advantage," though I have to admit the marketing people did do a heck of a job with that (the legions that parrot the marketing campaigns as gospel speaks volumes of their success in that regard). I'd like to see more effort focused (pardon the pun) on making larger imaging chips with lower cost and higher yield (which go hand-in-hand, of course), so that all of the formats can be restored to their actual film equivalents.
Do you believe that the 24x36 format is some sort of technology-independent ideal size, or that it simply is what made for economic, portable, high-quality end-results in the film era? If bigger is better, why stop at 24x36? It seems to me that at least part of the reason why 24x36 was so popular was due to limitations in the film medium that cannot be directly transferred to digital.
Or are you argueing more pragmatically based on availablility of lenses (and their "sweetspot"), the ability of people to learn how many mm are needed for "wide angle" etc?
-h
-
Given laws of economics and physics, those efforts will also benefit smaller sensors. And given the mathematics involved, they benefit smaller sensors more than larger ones - ie. improving yields of larger sensors by 10% will improve yields of smaller sensors by >10%.
This is not true.
Let us compute the yield assuming different levels of defects. Let p be the probability of a defect per mm2 of the sensor. The yield is then given by (1-p)A/1mm2, where A is the area of the sensor. (This is simplified but not too far from reality).
Here is a table for different sensor sizes:
| Format | Area | Yield (p=0.05%) | Yield (p=0.01%) | Yield (p=0.001%) |
| IQ180 | 2169 mm2 | 33.8% | 80.5% | 97.9% |
| 35mm | 864 mm2 | 64.9% | 91.7% | 99.1% |
| Nikon DX | 372 mm2 | 83.0% | 96.3% | 99.6% |
| 1/1.8 | 38 mm2 | 98.1% | 99.6% | 99.9% |
Clearly, larger formats gets more competitive with smaller formats if the yield can be improved. Going from a yield of 98% to 100% is not nearly as big as an improvement as going from 34% to 98% is...
-
Clearly, larger formats gets more competitive with smaller formats if the yield can be improved. Going from a yield of 98% to 100% is not nearly as big as an improvement as going from 34% to 98% is...
Thanks for bothering to check my head math, which clearly is inferior to your actual number crunching :)
-
The beauty of digital sensors is that they can be any size but enormously restrained to legacy glass as well as current lens lines in manufacturing (same thing).
Eduardo
Do you believe that the 24x36 format is some sort of technology-independent ideal size, or that it simply is what made for economic, portable, high-quality end-results in the film era? If bigger is better, why stop at 24x36? It seems to me that at least part of the reason why 24x36 was so popular was due to limitations in the film medium that cannot be directly transferred to digital.
Or are you argueing more pragmatically based on availablility of lenses (and their "sweetspot"), the ability of people to learn how many mm are needed for "wide angle" etc?
-h
-
The resolution limit for APS-C sensors (as respects middle spectrum color) for diffraction-limited lenses at f8-f11 (the range that is a reasonable expectation for mass-produced lenses) is thirteen megapixels for f8, and seven megapixels for f11, based on physics.
And though, one can see a rather small loss of sharpness in photos taken with 18MP APSC sensor @f/11?
Could it be than the presence of an AA filter, and the associated capture sharpening, are able to somewhat offset (or dilute) the loss of sharpness?
-
And though, one can see a rather small loss of sharpness in photos taken with 18MP APSC sensor @f/11?
Could it be than the presence of an AA filter, and the associated capture sharpening, are able to somewhat offset (or dilute) the loss of sharpness?
I dont think that the AA filter will ever increase sharpness. Rather, it will decrease sharpness (together with lense/diffraction, pixel count etc).
Capture sharpening can seemingly give arbitrary levels of sharpness if the signal to noise level is sufficient. I dont think that it can ever give arbitrary levels of real details, though.
-h
-
http://www.luminous-landscape.com/tutorials/resolution.shtml
The resolution limit for APS-C sensors (as respects middle spectrum color) for diffraction-limited lenses at f8-f11 (the range that is a reasonable expectation for mass-produced lenses) is thirteen megapixels for f8, and seven megapixels for f11, based on physics. These limits won't be exceeded due to sensor technology, and irrespective of Foveon vs. Bayer "relativity," those maximum resolution limits are here to stay.
Interesting how people throw around numbers without actually looking at test data. If lenses were really diffraction limited at such apertures, then resolution would not improve with increasing pixel density. Instead, if on looks at Photozone tests done with 8MP and with 15MP (Canon, Bayer pattern) sensors, one sees a substantial increase in resolution even at f22:
8MP:
(http://www.photozone.de/images/8Reviews/lenses/canon_100_28/mtf.gif)
15MP:
(http://www.photozone.de/images/8Reviews/lenses/canon_100_28_50d/mtf.png)
There is a difference between the point beyond which "diffraction blurs the image more as the aperture shrinks" and the point beyond which "diffraction prevents any further increase in resolution by increasing the pixel density". The Photozone data shows this quite concretely. Yes, diffraction decreases the resolution beyond about f8, and yes, increasing pixel density increases resolution even in the regime well beyond f8.
The physics here is that MTF decreases linearly with aperture, but it is not zero at Nyquist for either sensor, and therefore there is a benefit to finer pixel pitch.
-
This is sooo interesting! Thank you Emil for pointing this out. Gimmee more pixels!!!! ;D
Eduardo
Interesting how people throw around numbers without actually looking at test data. If lenses were really diffraction limited at such apertures, then resolution would not improve with increasing pixel density. Instead, if on looks at Photozone tests done with 8MP and with 15MP (Canon, Bayer pattern) sensors, one sees a substantial increase in resolution even at f22:
There is a difference between the point beyond which "diffraction blurs the image more as the aperture shrinks" and the point beyond which "diffraction prevents any further increase in resolution by increasing the pixel density". The Photozone data shows this quite concretely. Yes, diffraction decreases the resolution beyond about f8, and yes, increasing pixel density increases resolution even in the regime well beyond f8.
The physics here is that MTF decreases linearly with aperture, but it is not zero at Nyquist for either sensor, and therefore there is a benefit to finer pixel pitch.
-
I dont think that the AA filter will ever increase sharpness.
No, but I'd think it can mask a small loss of sharpness.
Anyway, Emil answerded much better than I did. ;D
-
Thank you ej.
I took the liberty of plotting the data that interested me (center resolution) in one common plot. Interesting things happen at the single f/22 measurement point for the 50D. Is this just a measurement error/noise/outlier? The datapoints (sadly) dont overlap, so it could be that my "linearizing" lines are fooling me, and that somewhere between f/16 and f/32, the "diffraction limited" effect kicks in for real. Naiively, one could interpret this as the 50D being more limited by diffraction (on an absolute scale) than the 350D. I do not believe that it is so, nevertheless, some might interpret your data this way.
If you connect two low-order RLC lowpass-filters in series, you should get a total filter response that has lower cutoff frequency than either. If the cutoff frequencies are very different, the lowest one will dominate the output. If the cutoff frequencies are of similar frequency, you will typically see significant changes in the total cutoff by varying the cutoff of either filter. I believe that is what happens here. We are trying to estimate the total system MTF based on simplified cutoff frequencies as if those were "ideal" brickwall lowpass filter. Neither nature nor humans seems to like designing perfect brickwall filters...
Edit:
Added a graph of center sharpness for various macro lenses for FF, 1.6x crop and 2.0x crop. All from photozone. Assuming that one only cares about the maximum center resolution that can be crammed into one single image, and willing/able to adjust distance/focal length freely, bigger sensor AND higher pixel-count seems to be the thing. It seems that increasing the MP count as well as buying a better lense will both typically increase the real center resolution (or the MTF50, at least) up until at least f/16 for 1.6x crop.
In the figure, I have tried to suggest the point of "being diffraction limited" according to http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm . I think it is clear that diffraction starts to limit resolution before this point is reached, and that it continues to pull resolution downwards after this point. So the "diffraction limit" should perhaps serve as a vague guide about where you should venture only with caution, and not an absolute physical brickwall, as it is sometimes described on the interwebs?
-h
-
I suspect the f/32 on the one test is a typo and should be f/22 (or perhaps the other way around). The data is more consistent if so; the graphs should not cross -- the higher density sensor will never have less resolution due to diffraction. Perhaps best to compare some other lenses using Photozone tests to make sure.
Edit: Diffraction is a brick wall at some point -- when MTF drops to zero due to diffraction, there is no detail to be found beyond that. What the graphs show is that current sensors are rather far from this point. I think what many photographers consider to be "the diffraction limit" is a personal choice on how much diffraction blur they are willing to tolerate on a given system (and as the data shows, the resolution at that point is system dependent). It's similar to the debates about DR -- there is a quantitative definition of DR, 11-14 stops in current models; and there is the portion of that range that is useful to a given photographer, based on a personal standard of image quality.
-
Sigma executive apologizes on his twitter page for SD1 price. This is another quake & tsunami for Sigma. This is not funny anymore. I feel sorry for them
Eduardo
-
Interesting how people throw around numbers without actually looking at test data. If lenses were really diffraction limited at such apertures, then resolution would not improve with increasing pixel density. Instead, if on looks at Photozone tests done with 8MP and with 15MP (Canon, Bayer pattern)...
But the new sensor from Sigma is roughly 15MP - and full-color ones at that. It should do quite well, especially as you open up the lens to F 4.0 and larger. It will probably easily outclass the typical DSLRs - that is, if you mounted the same glass on the Canon. But Sigma's glass is *not* a good Zeiss or similar, and that will kill any real advantage.
Better sensor + crappy glass = Good sensor and fantastic glass, in this case.
end result - just get the normal DSLR for a fraction of the price. Oh, and have 10x the glass to use on it that's any good, that is.
At $2000, Sigma would do some damage. At what they are charging, it's a toy for yuppies.
-
But the new sensor from Sigma is roughly 15MP - and full-color ones at that. It should do quite well, especially as you open up the lens to F 4.0 and larger. It will probably easily outclass the typical DSLRs - that is, if you mounted the same glass on the Canon. But Sigma's glass is *not* a good Zeiss or similar, and that will kill any real advantage.
If the DSLRs didn't have an AA-filter, the difference in resolving ability between Foveon and Bayer would be quite miniscule with modern demosaicing algorihms. The AA-filter does give some advantage to the Foveon, though how much remains to be seen, as the natural tendency for AA-filters is to weaken with smaller pixel pitches.
(On just about every other metric the traditional sensors easily outdo the SD1 sensor due to the inherit limitations of the technology used in the Foveon.)
At $2000, Sigma would do some damage. At what they are charging, it's a toy for yuppies.
At 2k it would sell in similar numbers to what it's predecessors, that is maybe 10-20 thousand units per year. Maybe a bit more due to many upgrades. However it would be very uninteresting to the general public due to it not having all the features of the competition and it not having similar shelf presence, advertisement, peer-pressure network, lens collection and so on while also not having any compelling advantages and having clear disadvantages in image quality as well (color accuracy and noise in color imagery - both unfixable in a sensor of Foveon-like design).
The price they launched sounds like Sigma is about to end their adventure in camera sales and concentrate on their real bussiness, lenses.
-
Do you believe that the 24x36 format is some sort of technology-independent ideal size, or that it simply is what made for economic, portable, high-quality end-results in the film era? If bigger is better, why stop at 24x36? It seems to me that at least part of the reason why 24x36 was so popular was due to limitations in the film medium that cannot be directly transferred to digital.
Or are you arguing more pragmatically based on availability of lenses (and their "sweetspot"), the ability of people to learn how many mm are needed for "wide angle" etc?
-h
I think the already existing format sizes shouldn't be decreased simply to deal with a fabrication cost issue which is a temporary aspect of a relatively new technology. If someone were to suggest to you that it were a great idea to carry around medium format equipment to shoot on a 24x36 sensor, would you agree with that notion, or dismiss it? That is the same thing, to me, as the idea of carrying equipment sized for a 24x36 format to shoot in a less-than-half-frame format, i.e., APS-C.
I do think 24x36 happens to be the "sweet spot" in terms of size/weight/cost of equipment, and that replacing film with digital sensors does nothing to change that fact. The array of available optics for the 24x36 format is second to nothing, and if digital makes for enhancement of image quality to boot, the more the merrier! If I was a medium format shooter, I'd be just as annoyed about the unwanted/unrequested cropping of my image format as I am using 35mm.
In short, bigger IS better, but there are practical limitations. Ever see a 300-800 zoom for large format? For medium format? Nope! You never will, because nobody could afford one, and you would need a gun motor carriage to transport and use it. I never moved to a larger format because the available tools were too few, never mind the cost of doing so and the logistics of transporting it in the field. Speed is another factor; how many medium or larger format systems can shoot at high frame rates for moving subjects? They are more specialized tools that pose certain limitations as to what the photographer can do. The 35mm format is the format offering the greatest array of tools for the photographer, and the greatest range of capabilities/versatility, while maximizing image quality within the practical constraints of size/weight/cost of equipment. APS-C format is nothing more than a 35mm camera with an undersized sensor, which may reduce cost (mainly of the camera body, as most of the lenses are 35mm anyway) but makes no appreciable dent in size/weight of equipment, and suffers a loss of image quality as well. Furthermore, it suffers (due to the format size reduction) from an unacceptable user interface, i.e., the viewfinders are too small. I found myself unable to focus on anything that wasn't pretty much "infinity" focus with lenses shorter than about 70mm or so with an APS-C viewfinder, which made it intolerable for me (having all manual focus lenses in particular; and since one can't relay on the ability of autofocus to work in all situations, I would find it intolerable even if I had autofocus lenses).
Just my $ 0.02 ;D
-
Interesting how people throw around numbers without actually looking at test data. If lenses were really diffraction limited at such apertures, then resolution would not improve with increasing pixel density. Instead, if on looks at Photozone tests done with 8MP and with 15MP (Canon, Bayer pattern) sensors, one sees a substantial increase in resolution even at f22:
8MP:
(http://www.photozone.de/images/8Reviews/lenses/canon_100_28/mtf.gif)
15MP:
(http://www.photozone.de/images/8Reviews/lenses/canon_100_28_50d/mtf.png)
There is a difference between the point beyond which "diffraction blurs the image more as the aperture shrinks" and the point beyond which "diffraction prevents any further increase in resolution by increasing the pixel density". The Photozone data shows this quite concretely. Yes, diffraction decreases the resolution beyond about f8, and yes, increasing pixel density increases resolution even in the regime well beyond f8.
The physics here is that MTF decreases linearly with aperture, but it is not zero at Nyquist for either sensor, and therefore there is a benefit to finer pixel pitch.
The trouble is, this "comparison" requires making quite the assumption - that is, that the only difference between the two sensors is pixel count. In fact, we're talking about sensors two generations apart. Differences in sensor technology, demosaicing algorithms, and so forth may have more to do with any increase in resolution than pixel count does, for all we know. Unless we can compare two sensors from the same generation in the same format using the same technology from the same manufacturer with different pixel counts, we're comparing apples and oranges. We also don't have the "lines" related in any way to what the maximum resolution as limited by diffraction might be (both cameras may be below it, for example, which means you could see no limitation in these "tests," even though it would apply if the cameras performed well enough for the limitations to be seen). Quite frankly in the digital age I'm not so confident that "lines" are actually being resolved, as opposed to being interpolated, so I'm not especially convinced by such "data." I'm skeptical of something so easy for square pixels to artificially replicate through algorithm guesswork (i.e., straight lines) as being a meaningful subject of comparison - a more challenging, non-linear subject is more likely to be a realistic test of real-world resolving power.
-
The trouble is, this "comparison" requires making quite the assumption - that is, that the only difference between the two sensors is pixel count. In fact, we're talking about sensors two generations apart. Differences in sensor technology, demosaicing algorithms, and so forth may have more to do with any increase in resolution than pixel count does, for all we know.
I assume that the test was carried out using raw files, ruling out demosaicing as an unknown in you list. Differences in micro lenses, AA filter etc could very well cause some of the differences. It is often claimed that diffraction is a "hard" limit, pointing to some web calculator or simplified formula. It is often claimed upon new camera releases from Canon/Nikon that one will have to use f/4.0 to get any improvement compared to the previous generation. This measurement shows this not to be the case with Canon 8MP vs 15MP sensors, and my bet is that something similar will be seen when going to the next jump in sensor resolution: it is a soft limit where improvements will be getting smaller and smaller compared to the cost (cost in the wide-sense interpretation).
Unless we can compare two sensors from the same generation in the same format using the same technology from the same manufacturer with different pixel counts, we're comparing apples and oranges.
Would have been nice, but probably wont happen. So when something is hard to measure directly, what are we to do? Proclaim that it cannot be known, and that everyones guess is equally good, or try to measure and model what we can?
We also don't have the "lines" related in any way to what the maximum resolution as limited by diffraction might be (both cameras may be below it, for example, which means you could see no limitation in these "tests," even though it would apply if the cameras performed well enough for the limitations to be seen).
It might be that both cameras are limited by something else than diffraction. In that case, claiming that the "diffraction limit" is a real problem for those cameras become dubious.
Quite frankly in the digital age I'm not so confident that "lines" are actually being resolved, as opposed to being interpolated, so I'm not especially convinced by such "data." I'm skeptical of something so easy for square pixels to artificially replicate through algorithm guesswork (i.e., straight lines) as being a meaningful subject of comparison - a more challenging, non-linear subject is more likely to be a realistic test of real-world resolving power.
I dont subscribe to the idea of "interpolating to create real detail". But measuring monochrome targets of regular lines, then using MTF50 to say something about sharpness is for sure a simplified measure that lends itself well to manufacturer "cheating" (or test conductor/interpreter ignoring essential mechanisms).
An alternative would be a 2-d randomized pattern. Measuring the 2-d response to "noise" input, or (even better) comparing the image to the known reference.
-
I think the already existing format sizes shouldn't be decreased simply to deal with a fabrication cost issue which is a temporary aspect of a relatively new technology.
I think that anything that improves the quality-to-cost ratio should and will be considered. Larger sensors have always, and will probably always be disproportionally more expensive than small ones. For film cameras the one-time cost of the sensor was zero (however, it did cost something per shot). For DSLR, the sensor seems to be a large fraction of the cost.
If someone were to suggest to you that it were a great idea to carry around medium format equipment to shoot on a 24x36 sensor, would you agree with that notion, or dismiss it? That is the same thing, to me, as the idea of carrying equipment sized for a 24x36 format to shoot in a less-than-half-frame format, i.e., APS-C.
Well, a Canon 1100D with a 18-55mm EF-S IS and a 55-250 EF-S is not "24x36 equipment". The house will accept 24x36-type lenses, but will only use a cropped portion. The lenses cannot be used on 24x36mm houses, and their cost/size/weight is probably decreased as a result. Someone with more technical insight than me might claim that the EF/EF-S lens mount makes it harder to construct good wide-angle lenses suitable for 1.6x crop than an imaginary mount tailormade for 1.6x crop could have, but I think thats about it. Empiry seems to suggest that there are good, sensibly priced wide-angle crops, although we will never know if they could have been better/cheaper if the mount was different.
Purchasing a 5D with a 24-105mm f/4.0 L IS would have been out of the question for most beginners 5 years ago. Film may have been out of the question. A 350D with a normal zoom would have been a reasonably good entry into DSLR, where they could carry over flashes, some lenses etc into FF if and when they felt the urge.
I do think 24x36 happens to be the "sweet spot" in terms of size/weight/cost of equipment, and that replacing film with digital sensors does nothing to change that fact.
I think that film/sensor is a part of the optimization, and that you are wrong. 35mm may have been the sweet-spot partially because of the cost/performance of film. With digital the cost/performance is radically different, and it is reasonable to expect that if one could design from scratch, other trade-offs would make more sense. I dont think there ever were film cameras using the miniature sensor size of todays compact/cell-phone camera yet with the cost, resolution and speed of todays small cams? Perhaps we will never see 8x10 digital cameras (at least at non-NASA prices) even though the format made sense (for some) on film?e
The array of available optics for the 24x36 format is second to nothing,
Agreed. But I believe that have not been the case since beginning of time? If e.g. mFT grows really popular, I would expect a bunch of lenses (although perhaps mostly lower cost ones.)
In the end, I think 35mm may be a good trade-off for you (and quite a few others). I think that many of the absolute arguments ("bigger is better") could be used for MF against 35mm equally well. So is 35mm "bad" because MF is larger? Of course not, just like 1.5x/1.6x crop is not "bad" just because 35mm is bigger.
-
The trouble is, this "comparison" requires making quite the assumption - that is, that the only difference between the two sensors is pixel count. In fact, we're talking about sensors two generations apart. Differences in sensor technology, demosaicing algorithms, and so forth may have more to do with any increase in resolution than pixel count does, for all we know. Unless we can compare two sensors from the same generation in the same format using the same technology from the same manufacturer with different pixel counts, we're comparing apples and oranges.
The test result as referenced by Emil is normalized as far as possible. It uses a method to determine the resolution, based on a robust MTF determination (essentially equivalent to the ISO prodedure for testing spatial resolution in digital cameras. The inevitable differences that remain are that we use different camera systems, with their particular AA-filter configuration.
We also don't have the "lines" related in any way to what the maximum resolution as limited by diffraction might be (both cameras may be below it, for example, which means you could see no limitation in these "tests," even though it would apply if the cameras performed well enough for the limitations to be seen).
While the presentation used doesn't show the relation to diffraction limitation on limiting resolution, the same test procedure I mentioned above does show that in a different graphical chart that the Imatest program can produce. Unfortunately, it is not easy (if at all possible) to present that in a summary graphical chart for multiple apertures. What the charts do show is the effect of diffraction on the MTF50 metric as an approximation of perceived sharpness.
Quite frankly in the digital age I'm not so confident that "lines" are actually being resolved, as opposed to being interpolated, so I'm not especially convinced by such "data." I'm skeptical of something so easy for square pixels to artificially replicate through algorithm guesswork (i.e., straight lines) as being a meaningful subject of comparison - a more challenging, non-linear subject is more likely to be a realistic test of real-world resolving power.
The resolving power is not determined with a line pattern, it is merely a (numerically converted metric) reference to other ISO test charts that do use hyperbolic line patterns for visual intepretation. The underlying spatial resolution metric is based on an MTF determnation, and more specifically the spatial resolution at 50% MTF response. That is by no means close to the limiting resolutions these systems can resolve, but rather a reasonable indication of perceived resolution. As much as an MTF curve is already a simplified representation, a single point on the MTF curve is even more so.
The important point is that we do not compare based on a line pattern being resolved or not, but on the gradual reduction of contrast as we approach the limiting resolution, and we stop almost half way to pick a reference point that's still very well resolved, and indicative for perceived resolution.
The fact that we compare actual camera's and lenses, will unfortunately make a purely theoretical distinction of a single isolated factor more difficult, because all other factors are not exactly equal. Nevertheless, there is strong enough evidence for us to draw careful conclusions. One such conclusion is that increased sampling density will improve absolute resolution (although it may also influence other characteristics such as dynamic range). Obviously, output magnification will inversely impact resolution, so larger sensor arrays benefit from that. Another conclusion is that diffraction blur has a negative impact on per pixel resolution, but there is no hard limit. However, it is possible to indicate for a given sampling density from which aperture number on the diffraction will visually impact that per pixel resolution. Smaller sensels will already be hurt at wider apertures. How that impacts the combined effect of the whole system, can only be seen at the final output size.
Cheers,
Bart
-
a more challenging, non-linear subject is more likely to be a realistic test of real-world resolving power.
Yes. However, resolution measurement metrics would need to be devised for such "real world" imagery, if that is what you mean. The good thing is that one can devise such measures, though. For example, the image below measures the "sharpness difference" between several pairs of images using JISR.
(http://djjoofa.com/data/images/jisr_pks.jpg)
Variations of such approaches can be used if one wants to have alternate notions of such metrics, which even work on real images, compared to "legacy" stuff such as line charts, MTF50, etc.
Joofa
-
Also, those resolution tests should clearly be in *color*. The biggest advantage of the Sigma sensor is that all of the colors have identical resolution. IF they get the color issues tweaked or worked out (possibly through in-camera processing or similar), it could really be a stunning piece of technology.
But the price... Nobody's going to buy it.
-
Also, those resolution tests should clearly be in *color*. The biggest advantage of the Sigma sensor is that all of the colors have identical resolution. IF they get the color issues tweaked or worked out (possibly through in-camera processing or similar), it could really be a stunning piece of technology.
But the price... Nobody's going to buy it.
The resolution of the Foveon is actually not quite identical for all the colors due to differing S/N of the readings of the different layers. Basicly red-heavy data has slightly lower resolution in principle, though not necessarily under ideal conditions, but in the shadows and near the edges and corners of the sensor.
The relatively low resolution of Bayer equipped sensors for red/blue is largely a myth. First, human eye is far more sensitive to luminance variations than changes in color, second, the color filters do have quite significant overlap (see DxO data for individual sensor data), third, in real life almost all the object data is formed from rather wide spectra information - for exaple the light reflecting from red flowers tends to have quite significant amount of green and/or blue as well. All this makes it much easier for a modern demosaiccing algorithm to reconstruct the image with significant accuracy regardless of the perceived color of the subject.
If there is a significant diffrrence between a Foveon and Bayer of similar pixel counts, the real culprit is not the Bayer CFA, but the anti-alias filter.
As for the color issues of the Foveon - they can not be solved with any amount of processing. It would take several more readout points per pixel making the sensor even more expensive to manufacture and creating even more data to be moved, stored and processed just to gain parity with the competition.
-
I have no experience with foveon sensors but aren't we as photographers trying to capture what the eye sees?
Yeah, but the eye is connected to a brain, which does a lot of processing that a humble camera cannot, so mimicing the physiology of the eye in a camera sensor is not necessarily the best way to go.
-
I have no experience with foveon sensors but aren't we as photographers trying to capture what the eye sees?
Yeah, but the eye is connected to a brain, which does a lot of processing that a humble camera cannot, so mimicing the physiology of the eye in a camera sensor is not necessarily the best way to go.
Actually, the physiology of the eye is much closer to Bayer than to Foveon.
-
so mimicing the physiology of the eye in a camera sensor is not necessarily the best way to go.
This is true if one knows what are the intended applications. If the criterion is to mimic the eye, then the CFA's used on many Bayer-type sensors have a closer response to the human eye than a "bare" Foveon sensor. However, "mimicing" the eye has not always been the goal of engineering applications based upon color vision, in devices ranging from NTSC color TV to more modern stuff such as certain REC specifications on how to do color space transformations. The underlying idea in such situations has been to forgo mimcing eye if an otherwise "incorrect" set of parameters resulted in a more pleasing image or a technically more manageable image.
Joofa
-
so mimicing the physiology of the eye in a camera sensor is not necessarily the best way to go.
This is true if one knows what are the intended applications. If the criterion is to mimic the eye, then the CFA's used on many Bayer-type sensors have a closer response to the human eye than a "bare" Foveon sensor. However, "mimicing" the eye has not always been the goal of engineering applications based upon color vision, in devices ranging from NTSC color TV to more modern stuff such as certain REC specifications on how to do color space transformations. The underlying idea in such situations has been to forgo mimcing eye if an otherwise "incorrect" set of parameters resulted in a more pleasing image or a technically more manageable image.
Joofa
-
so mimicing the physiology of the eye in a camera sensor is not necessarily the best way to go.
Actually mimicking the eye would
- cause halos (lateral inhibition)
- require different pixels for color and for luminosity (cones and rods)
- require concepts of shape to be represented within camera intelligence
(primary sensory fields in the brain are strongly influenced by higher centres)
- require a sharp center (fovea) and a rapid falloff to the corner
- require a blind spot (optic nerve exit)
-
so mimicing the physiology of the eye in a camera sensor is not necessarily the best way to go.
This is true if one knows what are the intended applications. If the criterion is to mimic the eye, then the CFA's used on many Bayer-type sensors have a closer response to the human eye than a "bare" Foveon sensor. However, "mimicing" the eye has not always been the goal of engineering applications based upon color vision, in devices ranging from NTSC color TV to more modern stuff such as certain REC specifications on how to do color space transformations. The underlying idea in such situations has been to forgo mimcing eye if an otherwise "incorrect" set of parameters resulted in a more pleasing image or a technically more manageable image.
Joofa
The last times when I used to get so much striking out on my response was when I was in school. I thought I had passed that terrible phase in my life. ???
Anyway the answer is still what was there before: "This is true if one knows what are the intended applications."
Joofa
-
Actually mimicking the eye would
- cause halos (lateral inhibition)
- require different pixels for color and for luminosity (cones and rods)
- require concepts of shape to be represented within camera intelligence
(primary sensory fields in the brain are strongly influenced by higher centres)
- require a sharp center (fovea) and a rapid falloff to the corner
- require a blind spot (optic nerve exit)
What is meant by "mimicking" the eye here is not the complete visual processing as it happens in the human visual system, a lot of which is still not known fully. What is meant is the "low-level" or the signal acquisition stage - that is why the argument only stays around the color response of the Bayer CFA or the Foveon. Furthermore, "luminosity", as in daylight, is mostly derived from cones, and not from rods. Rods do provide an achromatic signal but is not contributing in a certain sense.
"Higher cognitive representations" such as shape / object identification, etc., that you mention, are not a concern in the lower-level processing that is being talked about in "mimicking the eye" in this thread.
Joofa
-
mimicing the physiology of the eye in a camera sensor is not necessarily the best way to go.
The last times when I used to get so much striking out on my response was when I was in school. I thought I had passed that terrible phase in my life. ???
Anyway the answer is still what was there before: "This is true if one knows what are the intended applications."
Joofa
Clearly it is not. Note that I said is not necessarily the best way to go., which is (maybe trivially) correct regardless of any application.
Anyway, enough pedantry.
The larger point is that the human visual system is dominated by a super-sophisticated processing system which can overcome severe shortcomings in the acquisistion stage which would be disastrous in a camera.
-
What is meant by "mimicking" the eye here is not the complete visual processing as it happens in the human visual system, a lot of which is still not known fully. What is meant is the "low-level" or the signal acquisition stage - that is why the argument only stays around the color response of the Bayer CFA or the Foveon. Furthermore, "luminosity", as in daylight, is mostly derived from cones, and not from rods. Rods do provide an achromatic signal but is not contributing in a certain sense.
"Higher cognitive representations" such as shape / object identification, etc., that you mention, are not a concern in the lower-level processing that is being talked about in "mimicking the eye" in this thread.
Joofa
I think that if one gives the concept a little thought, it becomes clear that a camera "mimicing the eye" is nothing to strive for. A much more sensible goal would be a camera that, upon reproduction allowed a human observer an experience that mimic "having been there". My point is that even if my sight has limitations (e.g. limited resolution), I have no desire for my camera to have the same limitation if those two limitations add up.
-h
-
The larger point is that the human visual system is dominated by a super-sophisticated processing system which can overcome severe shortcomings in the acquisistion stage which would be disastrous in a camera.
As I mentioned above, higher-level semantic processing of human vision may not be mixed with the lower-level signal acquisition of a camera system, unless it is clear what is the intent. The usual application in photograpghy, as I understand, is acquiring pleasing images, aided with some simple tasks such as color corrections, profile assignments, etc. And, judging by people's response, the visual appeal or difference between Bayer CFA and Foveon is important for many. However, processing at this level does not require higher level human visual processing as in "image / scene understanding", if that type of processing is what you are talking about when you say "super-sophisticated processing system". And, where higher level scene understanding is involved, e.g., hierarchical semantic relationship of primitive image elements to each other, shape recognition, object identification, motion analysis, computer vision, etc, I believe the difference between Bayer CFA and Foveon color response is perhaps not always important. Example, a challenging task in computer vision is object recognition, say just identifying if a particular picture has a cat in it. How much does it matter under ordinary imaging conditions if the picture was taken with Bayer CFA or Fovean sensor camera?
Joofa
-
What is meant by "mimicking" the eye here is not the complete visual processing as it happens in the human visual system, a lot of which is still not known fully. What is meant is the "low-level" or the signal acquisition stage - that is why the argument only stays around the color response of the Bayer CFA or the Foveon. Furthermore, "luminosity", as in daylight, is mostly derived from cones, and not from rods. Rods do provide an achromatic signal but is not contributing in a certain sense.
"Higher cognitive representations" such as shape / object identification, etc., that you mention, are not a concern in the lower-level processing that is being talked about in "mimicking the eye" in this thread.
Joofa
Everything I was mentioning are things happening in the retina and the first level representation in the brain.
So this indeed is the very basic level of signal acquisition.
E.G. there are optical illusions, like this:
(http://4.bp.blogspot.com/_t7lQZ9jD6C4/SQuDwiAkcXI/AAAAAAAAAAM/xE9DNM_hWt4/s320/300px-Kanizsa_triangle.jpg)
(image linked from 4.bp.blogspot.com/_t7lQZ9jD6C4/SQuDwiAkcXI/AAAAAAAAAAM/xE9DNM_hWt4/s320/300px-Kanizsa_triangle.jpg)
where the shape of the triangle (which doesn't exist in the image) is fully represented in the first level brain area, which is directly supplied by the nervus opticus.
Even at the very basic level optical concepts come into play.
The luminosity information from the rods plays an important role when its dark, since they are more sensitive.
As a consequence we can see our color recognition suffers greatly when its dark.
Of course you might leave out that first level brain thingy, however - what I wanted to stress is, that mimicking the eye might be not so desireable than it might look on first sight.
-
I'm regretting resurrecting this topic, and I apologise to Joofa if my responses seem aggressive. I think we are mostly saying the same thing, in fact, but my original coment was in response to this:
I have no experience with foveon sensors but aren't we as photographers trying to capture what the eye sees?
I understand this sensor mimics color film's layers but the eye has individual cones (RGB) and rods (B&W). It seems to me a sensor with individually colored photosites would mimic the eye better?
I always thought if the RGB sites were arranged in a pseudo random pattern (latin square) vs Bayer pattern we might have something. Other than mimicking film I don't see the advantage of a foveon sensor
which seems to suggest that it is desirable to make a sensor that physically resembles the human eye in some sense, which I think is a mistake.
-
which seems to suggest that it is desirable to make a sensor that physically resembles the human eye in some sense, which I think is a mistake.
I'd say that it would be more useful to have a sensor that allows to output something that resembles what the human eye saw, with minimal processing. That doesn't necessarily mean that the sensor response should resemble our eye's, although it likely would help.
Cheers,
Bart
-
Not sure about that, Bart. Red-green separation in color response of the eye is quite poor, having evolved quite late in our evolutionary history; if I'm not mistaken, this is related to red-green colorblindness issues when one is on the tail of the distribution of color visual response. Having a CFA with even the normal human response would I suspect lead to a lot of chroma noise when one attempts to separate the strongly overlapping color channels.
I thought one of the things people liked about MFDB's is their willingness to sacrifice some QE for good color separation between the channels. I think the engineering challenge in digital photography is really quite different from that of human vision, so wouldn't necessarily expect that the solution embodied in the latter to be optimal for the former.
-
Example, a challenging task in computer vision is object recognition, say just identifying if a particular picture has a cat in it.
Joofa
How could I use a lowly word, such as an "object", in relation to a cat? My sincere appologies to anyone whose feline sensibilities are hurt.
Joofa
-
If there is a significant diffrrence between a Foveon and Bayer of similar pixel counts, the real culprit is not the Bayer CFA, but the anti-alias filter.
easy to test - take any MF(DB) image (no AA) and zoom 100% to a pixel 1:1 level, then do the same w/ Foveon image (not SD1 - take production camera like DP2 for example) and have the subjects of the same size (in pixel terms - subject that occupies the same amount of pixels on both sensors in the image) of course... at that pixel peeping level MF(DB) is destroyed by a simple P&S DP camera, regardless of the absence of AA... there is no replacement for displacement (demosaicking vs no demosaicking).... printing or resizing (down to 4.6mp of the production Foveon) of course will show the different picture, there the sheer amount of sensels that MF(DB) has will give it an advantage, but not @ 100% zoom on screen
-
easy to test - take any MF(DB) image (no AA) and zoom 100% to a pixel 1:1 level, then do the same w/ Foveon image (not SD1 - take production camera like DP2 for example) and have the subjects of the same size (in pixel terms - subject that occupies the same amount of pixels on both sensors in the image) of course... at that pixel peeping level MF(DB) is destroyed by a simple P&S DP camera, regardless of the absence of AA... there is no replacement for displacement (demosaicking vs no demosaicking).... printing or resizing (down to 4.6mp of the production Foveon) of course will show the different picture, there the sheer amount of sensels that MF(DB) has will give it an advantage, but not @ 100% zoom on screen
Apart from the differences in optics, that seems to be a fair test. Have you done it? Is that how you reached your conclusion of "at that pixel peeping level MF(DB) is destroyed by a simple P&S DP camera"?
All I've seen produced by others are assumptions. I've personally done a test (http://bvdwolf.home.xs4all.nl/main/foto/bayer/bayer_cfa.htm) of the effect that demosaicing has on Luminance resolution, and it's not that dramatic as some would like us to believe. Only in the case of equal luminance will a loss of Luminance resolution be seen, because modern demosaicing algorithms tend to optimize more for Luminance resolution than for Chroma resolution.
Cheers,
Bart
-
I've personally done a test (http://bvdwolf.home.xs4all.nl/main/foto/bayer/bayer_cfa.htm) of the effect that demosaicing has on Luminance resolution, and it's not that dramatic as some would like us to believe.
Here is a little different actual test of "resolution" with and without demosaicing on real images on two exact same sensors, with the only difference being the presence of CFA on one, and the other being mono.
(http://djjoofa.com/data/images/bayer_mono_test.jpg)
AA filter: None.
IR filter: None.
Optics: Same paramters between the two; such as f-stop, exp time, etc.
Demosaic method: Bilinear interpolation.
Note that bilinear interpolation is not the best method, hence the two images kind of show two extremes with a CFA and without.
Joofa
-
Note that bilinear interpolation is not the best method, hence the two images kind of show two extremes with a CFA and without.
Hi 'Joofa',
Indeed, clearly visible but not all that earth-shattering especially with a bit of sharpening added to your JPEG.
Cheers,
Bart
-
Here is a little different actual test of "resolution" with and without demosaicing on real images on two exact same sensors, with the only difference being the presence of CFA on one, and the other being mono.
Demosaic method: Bilinear interpolation.
Note that bilinear interpolation is not the best method, hence the two images kind of show two extremes with a CFA and without.
Hardly a test of CFA effect on image quality, but a good example on what a very simple interpolation algorithm achieves from the recorded data.
-
Bilinear will soften edges a lot more than a typical demosaic method used in any good raw converter, as well as introduce a lot more aliasing.
-
For what it is worth, here are a a couple of sample images attached, taken from different distances to allow for comparison of details. Also included a 200% vs. 500% comparison.
Camera used: Pentax K20D - so AA-filter and Bayer CFA were present during imaging ;)
Target was a 20 euro bill - blue-red areas to make resolving more difficult for the demosaicing algorighm. Raw-conversion was done with Lightroom 3 with all settings zeroed, ie. no sharpening what so ever.
Edit: added a sharpened version of the small one just to make it clearer how well details are resolved.
To my eyes it seems like a sensor with a Bayer CFA will resolve quite nicely even when unfavorable detail color.
-
Hardly a test of CFA effect on image quality ...
It is not a test of how good the CFA response can be made using different demosaic algorithms, if that is what you mean here. The intent was to show the kind of two "extremes", i.e., any good demosaicing algorithm will be operating within these two loose end points.
Joofa
-
Hi 'Joofa',
Indeed, clearly visible but not all that earth-shattering especially with a bit of sharpening added to your JPEG.
Cheers,
Bart
Hi "Bart",
True indeed. There is an interesting evaluation of different deMosaic algorithms on the following link:
http://www.libraw.org/articles/bayer-moire.html (http://www.libraw.org/articles/bayer-moire.html)
The author recommends LMMSE for further analysis given its good response on "closer to reality" images (author's words). And, which is not surprising given the blurring effect of AA + other effects on an image before demosaic happens as drawn in a simplified linear fashion below:
(http://djjoofa.com/data/images/bayer_demosaic_filtering.jpg)
Joofa
-
The typical test of demosaic algorithms is to take a three-color image (eg from a scan), 'forget' 2/3 of the colors leaving only those in the Bayer pattern, and then run the resulting mosaic image through the demosaic algorithm of choice. The original image provides the reference 'true' image, the mosaic/demosaic route provides a test of the faithfulness of the demosaic image to that 'truth'. In principle one can also test the effects of OLPF filtering by applying the relevant anti-alias blur operation between mosaic and demosaic (though I have not done that on the images below).
Here is a standard test image often used in the literature on demosaic algorithms:
(http://theory.uchicago.edu/~ejm/pix/20d/posts/ojo/kodim19.png)
and here is what comes from a good demosaic algorithm applied to the mosaiced image one gets by forgetting all but the RGGB pattern from each quartet of pixels:
(http://theory.uchicago.edu/~ejm/pix/20d/posts/ojo/K19_05-08-09.png)
-
I'd say that it would be more useful to have a sensor that allows to output something that resembles what the human eye saw, with minimal processing.
I'd say at least that with our limited chroma resolution, we generally may not remark that an image has a chroma resolution a bit behind its luminance resolution?
-
Here is a standard test image often used in the literature on demosaic algorithms:
[SNIPPED images for brevity]
and here is what comes from a good demosaic algorithm applied to the mosaiced image one gets by forgetting all but the RGGB pattern from each quartet of pixels:
Hi Emil,
Very nice demonstration, thanks. Is that with your full version of AMaZE in MatLab?
My old test was done with Aqua, but it's not around for download anymore.
Cheers,
Bart
-
Hi Emil,
Very nice demonstration, thanks. Is that with your full version of AMaZE in MatLab?
My old test was done with Aqua, but it's not around for download anymore.
Cheers,
Bart
Yes. Actually, in Mathematica rather than Matlab. As one can see, it does decently well with high frequency, greyscale detail such as the picket fence. Errors are largest with the colored edges of the life preserver and the yellow rope since the details are largely in more sparsely sampled reds; and in the grass (more random high frequency structure is harder to infer than regular, more patterned structure).
-
I'd say at least that with our limited chroma resolution, we generally may not remark that an image has a chroma resolution a bit behind its luminance resolution?
Hi Nicolas,
While true, things aren't as black and white (pun intended).
Restricted to the Fovea centralis (http://en.wikipedia.org/wiki/Fovea_centralis), which is occupied mostly with cones (for colorvision), human visual color resolution peaks when enough light is available. Color information by itself usually has a low spatial frequency contribution to our vision and in scenes of natural objects (it's more of a warning signal at a distance, because e.g. Red may be warm/hot/blood). That is easily demonstrated in a Lab (or HSB or HSL) respresentation of an average image, there is little resolution to be seen.
However, the human visual system (HVS) depends heavily on pattern detection (to avoid information overload) for detail discrimination, and therefore luminosity as a means to convey the important (edge detail) message is rated higher in our brain. So counting cones and/or rods alone is not telling the whole story. Contrast, or MTF, is very important for our detail resolving abilities.
The HVS, which involves the brain, therefore has a variable resolution depending on average scene brightness and luminosity contrast. The brilliant weighting of a Bayer CFA is a very good (although not perfect) instrument to allow the most relevant information to be recorded. The fact that the cones of our eyes are more randomly distributed complicates a direct comparison with an aliasing prone ordered sampling method.
Nevertheless, as also Emil has demonstrated above, it is possible to achieve a visually very good/convincing reconstruction of the actual scene with only a faction of the data (which saves a lot of storage and speeds up the storade (frames/second) and transfer of data). In fact, when one analyses the R, G, and B, MTF resolution of an image (e.g. in Imatest) with adequate luminance contrast, the RGB resolutions of a Bayer CFA image are virtually identical to the luminance resolution. Only in the worst case scenario of chroma resolution in the absence of most luminosity contrast info will there be a benefit for full RGB resolution/sampling. When enlarging the resulting image, there is an obvious, although limited, benefit to having as much accurate (also RGB) data available.
Cheers,
Bart
-
Of course, it could be that the real reason why Sigma priced the DSP1 at US$9,700 was to generate some pre-publicity.
This thread is an example of such pre-publicity. The camera hasn't even been released, but we have 150 odd posts on LL discussing its possible properties.
-
Take heart, you can pre-order from from BhPhoto for an amazingly affordable $6,899! Seems like a bargain. When did you say the 5d MKIII is coming out?