Sensitivity of the new SD1 sensor ?

Started Sep 21, 2010 | Discussions
dala Regular Member • Posts: 411
Sensitivity of the new SD1 sensor ?

Anyone seen any statement of the ISO range of this beast?

 dala's gear list:dala's gear list
Nikon D750 Nikon AF-S Nikkor 70-200mm F4G ED VR Nikon AF-S Nikkor 20mm f/1.8G ED Nikon 24mm F1.8G ED Nikon AF DX Fisheye-Nikkor 10.5mm f/2.8G ED +6 more
MOD Kendall Helmstetter Gelner Forum Pro • Posts: 20,587
Not sure yet...

dala wrote:

Anyone seen any statement of the ISO range of this beast?

This seems like a pretty new sensor so it's hard to be sure of what the range might be. In the past Sigma has waited until near release to decide what they thought the acceptable range would be, although we could see some initial more conservative announcement before long I suppose since everyone will be wanting to know...

 Kendall Helmstetter Gelner's gear list:Kendall Helmstetter Gelner's gear list
Sigma 70-200mm F2.8 EX DG OS HSM Sigma 8-16mm F4.5-5.6 DC HSM Sigma 24-70mm F2.8 EX DG HSM Sigma 50-500mm F4.5-6.3 DG OS HSM Sigma 85mm F1.4 EX DG HSM +4 more
OP dala Regular Member • Posts: 411
Re: Not sure yet...

Yeah, I guess it depend on how fast the processor is and how complicated noise removal algorithms it can handle.

 dala's gear list:dala's gear list
Nikon D750 Nikon AF-S Nikkor 70-200mm F4G ED VR Nikon AF-S Nikkor 20mm f/1.8G ED Nikon 24mm F1.8G ED Nikon AF DX Fisheye-Nikkor 10.5mm f/2.8G ED +6 more
Michael McMurrough Senior Member • Posts: 1,107
Re: Not sure yet...

Just a guess, but I'm betting they go 100-1600 with extended modes for 50 and 3200.

I don't expect high ISO performance will be that much better than the sd15, hopefully no worse, and I really hope they include the iso 50 if only as a custom setting.

maple Veteran Member • Posts: 3,352
On par with SD14?

That’s my top concern, too. Here’s my wishful thinking:

The pixel pitch is 5 micron vs. 7.8 of the SD15, i.e., pixel size is 2.44 times smaller. Allowing 1 stop improvement in sensor base ISO thanks to better technologies (possibly, gapless micro-lens and cleverer circuitry layout?), it is still half a stop worse than SD15. So perhaps on par with SD14?

My other concern is speed. I guess it is reasonable to expect that it has at least two data transmission pipelines working in parallel to spit out 3.3 times the amount of data, each would have 65% more workload than SD15. So 65% slower? That’s no good for a flagship model. Perhaps it needs at least 4 pipelines plus 8 times buffer size, as 2 TRUR II processors can’t keep up with 3.3 times the data volume. Even then, it would only be 17.5% faster than its older sibling. No idea how much more complicated that would make the sensor making.

I do not expect miracles in these two respects, or did I read too much into the omission of these important points in the announcement? I would be very happy with the same ISO and speed as SD14 for 3.3 times the resolution. It is nothing short of a miracle already.

Otherwise, no worries at all. Not even about price.
--
Maple

photoaddict Contributing Member • Posts: 702
Re: Sensitivity of the new SD1 sensor ?

I think it's much harder for X3 to have the same sensitivity as a Bayer sensor because it has to go through three layers and every time you go through a layer, you lose some photons? Wouldn't it be like wearing three sunglasses?

Mostly Lurking Senior Member • Posts: 2,103
I don't think you can assume the pixel size will be smaller. . .

Remember the sensor patent?

Patent Title: "MULTI-COLOR CMOS PIXEL SENSOR WITH SHARED ROW WIRING AND DUAL OUTPUT LINES"

Publication date: 06/24/2010
Patent application number: 20100155576

This is surely the SD-1's sensor!

http://www.google.com/gwt/x?wsc=pr&wsi=6e885fb5d065508b&u=http%3A%2F%2Fwww.faqs.org/patents/app/20100155576&ei=g3wlTMfoG6iwqwO7qJG2Bw&ct=pg1&whp=30
--
William Wilgus

maple Veteran Member • Posts: 3,352
Re: Sensitivity of the new SD1 sensor ?

No. It is Bayer sensor that wares sunglasses though.

I don't know whether I understand it right: Foveon uses all the light fall upon a given area, less gaps between pixels or micro lenses and areas used up by wiring (which maybe quite a bit more than Bayer) whereas Bayer uses only 1/3 of the light even if they are able to use 100% of that given area.
--
Maple

stanislaw stitchanow Senior Member • Posts: 1,956
Re: Sensitivity of the new SD1 sensor ?

Might be 100-1600 maybe till 3200, afaik all ISO above 1600 is anyway just interpolated on those Nikons and Canons, those sensors have no real sensivity for ISO 12800 or higher.
So you can do it yourself and apply have NR, even with a Apple Quick Take

Joseph S Wisniewski Forum Pro • Posts: 34,385
Foveon uses about 8% of the light...
3

maple wrote:

No. It is Bayer sensor that wares sunglasses though.

I don't know whether I understand it right: Foveon uses all the light fall upon a given area, less gaps between pixels or micro lenses and areas used up by wiring (which maybe quite a bit more than Bayer) whereas Bayer uses only 1/3 of the light even if they are able to use 100% of that given area.

A Foveon sensor uses about 8% of the light that it received.

There's two things going on. First is the photons that simply aren't absorbed in any of the sensitive layers.

This paper from Richard Lyon at Foveon explains some of that.

http://www.foveon.com/files/CIC10_Lyon_Hubel_FINAL.pdf

Have a look at page 3 of 7, figure 6 wavelenght vs. quantum efficiency. That actually looks a lot like the QE plot of a decent Bayer sensor, except we'd expect closer to 45% in the green.

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.

The diagram on slide 26 of the AeroSense 2003 presentation by David Gilblom (Alternate Vision Corp), Sang Keun Yoo (HanVision), and Peter Ventura (Foveon) explained it pretty well. The link was:
http://www.alt-vision.com/r/documents/AeroSense_2003_Oral.pdf

But that appears dead. I'm not sure about the legalities of hosting the copy that I have.

The way it works is that a red photon can be absorbed by any of the three layers, at random. The red layer actually has the least probability of absorbing the red photon. But let's make the probabilities all equal, for the example.

So, imagine 30 red photons hit a pixel. 10 are absorbed in the first (white) layer. 10 are absorbed in the second (yellow) layer. 10 are absorbed in the bottom (red) layer.

Basically, 2/3 of the photons have turned into "noise" that must be subtracted from the white and yellow layers before they can become blue or green.

Same thing with green photons. They cause as much "noise" in the white layer as "signal" in the yellow layer.

The real number is actually 4.1 of red has to be subtracted away from the white channel. That's 1/4.1 or 24% of stuff that you get to keep.

Multiply that by the 32% QE, and you've only got 7.8% of your light actually used.

A Bayer sensor uses closer to 35% of photons. Virtually no destructive mutual information. About 4x the efficiency of a Foveon.

There's also the issue of differential sensing and uncorrelated noise, making reading low values from the sensor twice as noisy as a non-differential Bayer, so it's really more like 8x (or that the Foveon actually only uses about 4% of the light, in a low light situation).

-- hide signature --

Rahon Klavanian 1912-2008.

Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed.

Ciao! Joseph

http://www.swissarmyfork.com

 Joseph S Wisniewski's gear list:Joseph S Wisniewski's gear list
Nikon D90 Nikon D2X Nikon D3 Nikon D100 Nikon Z7 +45 more
omr Senior Member • Posts: 1,118
Corrected Link

Joseph S Wisniewski wrote:

... 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.

The diagram on slide 26 of the AeroSense 2003 presentation by David Gilblom (Alternate Vision Corp), Sang Keun Yoo (HanVision), and Peter Ventura (Foveon) explained it pretty well. The link was:
http://www.alt-vision.com/r/documents/AeroSense_2003_Oral.pdf
But that appears dead. ...

That presentation can be found here:

http://www.alt-vision.com/documentation/AeroSense-2003-Oral.pdf

and the paper is here:

http://www.alt-vision.com/documentation/5074-35.pdf

-- hide signature --

omr

AMomcil Senior Member • Posts: 1,018
Joseph, a few questions for you

Hi Joseph,

I have been following your posts about Foveon these days. Being an EE myself, I appreciate the way you are describing things, nicely explained.

Q1: What do you thing, knowing the Foveon physical model, what would be their future? Only low-light, high colour resolution applications (like landscapes)? Except described low-light problems, do you see more gotchas that would prevent Foveon to spread around more widely?

Q2: If we remove the colour array filter completely from the sensor (as Kodak used to long time ago) to produce a B/W (gray) camera, if I understood properly, it should be around 3 times more light sensitive, which would give around 1.5 stops better high ISO performances, comparing to just converting the picture to grayscale from RAW? In that case, there would also be no need for AA filter, I assume?

Q3: Would it be possible to create an electrically controlled CFA, that could be transparent (grayscale sensor) or color filtered, by e.g. direction or intensity of some controlling current. If so, would it be applicable to this small sensor pitches we have today?

Well, it turned out much more than three questions, but it is a very interesting topic for me..

Joseph S Wisniewski wrote:

The real number is actually 4.1 of red has to be subtracted away from the white channel. That's 1/4.1 or 24% of stuff that you get to keep.

Multiply that by the 32% QE, and you've only got 7.8% of your light actually used.

A Bayer sensor uses closer to 35% of photons. Virtually no destructive mutual information. About 4x the efficiency of a Foveon.

There's also the issue of differential sensing and uncorrelated noise, making reading low values from the sensor twice as noisy as a non-differential Bayer, so it's really more like 8x (or that the Foveon actually only uses about 4% of the light, in a low light situation).

-- hide signature --

Greetings, Aleksandar

 AMomcil's gear list:AMomcil's gear list
Sony RX100
maple Veteran Member • Posts: 3,352
Re: Foveon uses about 8% of the light...

Thank you, Joseph. I'm not sure I understand all of it, I read it twice to get most of it. Not trying to defend myself, what I was saying is just that all the light went to the sensor of the Foveon design, instead of being screened out by 2/3 before it reaches the pixels as is the case with Bayer. What you have described is how efficient a 3 layer sensor is in using the 100% of light it receives.

It is known that low light imaging is not Foveon's strength. What it does well is clarify at pixel level. As pixel density of the sensors eventually reaches the limit by diffraction, Foveon's advantage will become more pronounced. At the same pixel density of a given diffraction limit, a Bayer sensor delivers only about half of Foveon's resolution. In that sense, Foveon holds a bright future, at least for some type of photography.

-- hide signature --

Maple

Ipsofoto Forum Member • Posts: 55
Re: Foveon uses about 8% of the light...

I'm amazed how nerdy and technical photography has become. I bet the likes of Cartier-Bresson wouldn't give a flying about photons and pixel pitch.

Erik Magnuson Forum Pro • Posts: 12,247
Re: Joseph, a few questions for you

AMomcil wrote:

Except described low-light problems, do you see more gotchas that would prevent Foveon to spread around more widely?

The gotchas are as much political as technical. For large format sensors Canon, Nikon, Sony, and Panasonic design their own. They are not likely to change or license someone else's tech unless there is a huge performance gap.

If the technology spreads, it will be in different ways. For example, Canon is now using a 2-layer silicon separation sensor for the color metering sensor of their latest cameras.

Q2: If we remove the colour array filter completely from the sensor (as Kodak used to long time ago) to produce a B/W (gray) camera,

To get this with current Foveon sensors, all you need to do is add the 3 channels together. Someone has produced some SW to do this: google Monoveon.

In that case, there would also be no need for AA filter, I assume?

There is still aliasing - it's just not often visually objectionable.

Q3: Would it be possible to create an electrically controlled CFA, that could be transparent (grayscale sensor) or color filtered, by e.g. direction or intensity of some controlling current. If so, would it be applicable to this small sensor pitches we have today?

I suspect the problem would be similar to Foveon's - getting good color separation from that type of filtration.

-- hide signature --

Erik

Erik Magnuson Forum Pro • Posts: 12,247
Re: Foveon uses about 8% of the light...

Ipsofoto wrote:

I'm amazed how nerdy and technical photography has become.

If you haven't noticed, this is a gear forum ( "the place to discuss Sigma digital cameras (and Foveon X3 technology)" and not a photography forum.

-- hide signature --

Erik

TomFla Regular Member • Posts: 489
Re: Foveon uses about 8% of the light...

Joseph S Wisniewski wrote:
SNIP

So, imagine 30 red photons hit a pixel. 10 are absorbed in the first (white) layer. 10 are absorbed in the second (yellow) layer. 10 are absorbed in the bottom (red) layer.

SNIP

Well that is kinda hard for me to imagine.

According to Dick Lyon a pixel is a term derived from the German word picture bit first used around 1900, and the term pixel comes from picture element and first appeared in literature in the 1950s in some Xerox stuff out of Menlo Park. But that is neither here nor there. Google "Dick Lyon" and "Pixels and Me".

Photons do fall into wells of detectors on a sensor in a digital camera. Some of the best BW astrophotography cameras are claiming a detector can be activated by as few 5-8 photons being collected in the well. Non research grade cameras are no where that efficient. When a detector is activated (by what ever number of photons needed) this information is stored in a RAW file and eventually converted to a more common image file format. I am sure you already know this, and probably more.

But what I would like you to do is comment on how binning, something I seem to remember Foveon sensors can do at the hardware level, could possibly increase the effective sensitivity. I know Canon and Nikon are getting better at binning Bayer sensors, and am also wondering how you assess this race.

It seems to me Sigma/Foveon are missing the boat by not producing a camera that does both binning and HDMI video. The bigger chip in the sd1 would seem to give Sigma a real advantage in these two areas and address the clean high ISO problem past Sigma cameras have had.

Course I could be wrong about all of thiss, I am still on my first cup of coffee.

-- hide signature --

Those who forget history are condemned to go to summer school.

Joseph S Wisniewski Forum Pro • Posts: 34,385
Aleksandar, I was debating not answering...

AMomcil wrote:

Hi Joseph,

I have been following your posts about Foveon these days. Being an EE myself, I appreciate the way you are describing things, nicely explained.

Thank you.

I was debating not answering your questions. They're point blank in areas that I often avoid commenting.

Q1: What do you thing, knowing the Foveon physical model, what would be their future? Only low-light, high colour resolution applications (like landscapes)? Except described low-light problems, do you see more gotchas that would prevent Foveon to spread around more widely?

Knowing the physics of the thing, I don't believe that they have much of a future.

The second "gotcha" is the issue of color accuracy. Sigma made what I consider to be an inaccurate claim in their press release.

http://www.dpreview.com/news/1009/10092129sigmasd1.asp

"it efficiently reproduces colour more accurately"

It is neither efficient (at best, less than 25% of the efficiency of a CFA (Bayer or EXR) sensor, nor more accurate (best case metamerism index 0.199, according to its own creators).

Sigma (and Foveon) got away with these claims because they were, until recently, below the radar of the big sensor makers, Canon, Nikon, Panasonic, and Sony. The competitors honestly didn't care what silly claims Foveon, SMaL, FillFactory, Aptina, etc. made. Sigma may have just done something that's going to attract attention. Once you attract the attention of the large players, they complain to government agencies (and provide the necessary research reports. I'm one supplier of such reports) to get the advertising claims changed.

I don't thing the Foveon sensor has really ever (advertising claims aside, talking to the developers about what we can actually do with it) been about either resolution or color accuracy.

It's been about three things:

  • binning. That's the big thing. Instantly reconfiguring the sensor to video mode, with no loss of silicon area and no complex post processing. Liveview and video, on command.

  • cost. Manufactured in equal quantities, it should have been cheaper than a Bayer sensor of equal size, due to having less organic layers over the silicon layers. No color filters, and, in the first version, no microlenses.

  • simplicity. No Bayer demosaicing, just a 3x3 matrix multiply to get colorimetric color, and you're there. On a smaller image that should deliver resolution to a processed Bayer image with twice the pixel count (the original Foveon claim was only 1.7x. Foveon fans "grew" it to 3-4x. )

The problem was that every one of those things turned out just not be there.

  • binning. The biggest customer ignored it completely. No liveview, no video mode, and they mixed the binned 4:1 shooting mode with a variety of gloopy shoot full res, scale down 3:2 or 2:1 modes. Other customers, once we looked at the low QE, realized that a Bayer sensor, simply tossing 3/4 of the pixels for a video or liveview mode, still had at least a stop of sensitivity advantage over the Foveon "super pixels".

  • cost. The quantities never came, so the price never fell. And political wars with the first two foundries (all tooling at Nat Semi destroyed in a suspicious accident?) kept prices even higher. Some organics (microlenses) turned out to be necessary. The price never even hit the point where a 1.7x Foveon got down as cheap as a 1.5x Bayer with 28% larger area.

  • simplicity. By year one (when I got into the game, LOL) Foveon had backed down from the "ultra simple" software claims to having an SDK with a space-to-space interpolator for color, red sharpening (long story), and the beginning of what would turn out to be an insanely complex blown highlight compensation system. Customer one (That's Sigma, goes with "year one", LOL) launched three generations of Foveon camera without JPEG processing, because the processing was to complex to be self contained.

-- hide signature --

Rahon Klavanian 1912-2008.

Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed.

Ciao! Joseph

http://www.swissarmyfork.com

 Joseph S Wisniewski's gear list:Joseph S Wisniewski's gear list
Nikon D90 Nikon D2X Nikon D3 Nikon D100 Nikon Z7 +45 more
Joseph S Wisniewski Forum Pro • Posts: 34,385
But questions 2 and 3 are safer...

AMomcil wrote:

Q2: If we remove the colour array filter completely from the sensor (as Kodak used to long time ago) to produce a B/W (gray) camera, if I understood properly, it should be around 3 times more light sensitive, which would give around 1.5 stops better high ISO performances, comparing to just converting the picture to grayscale from RAW? In that case, there would also be no need for AA filter, I assume?

Closer to 1.5 to 2.4 times more sensitive. Modern Bayer sensors derive most of their luminosity information from the green channel, and not only is that twice as numerous as red or blue, but it has a filter that passes about 40% of light (and closer to 60%, when you equalize for photopic response, but that's an issue for another day). So, doing a new ISO test puts you 1/0.40x, or about 2.5x more sensitive. I actually measured 2.4x on the ICX413AQ.

But, having made and shot B&W cameras, the 2.4x greater sensitivity is a myth. Without filters, a sensor has a truly ugly spectral response. It peaks in the orange-red, and is several stops down in the blue. It's rather like shooting everything through a fairly strong red filter. Yes, clouds are pretty and "dramatic", but people tend to be "chalky", most males look like they have razor stubble even when cleanly shaved (remember what we learned about filters and portrait work). It's like shooting, full time, with a high red sensitivity film like Kodak Tech Pan or Ilford 820 So, you find yourself using a photopic green filter for most B&W photography, the classic X0, which cuts sensitivity down to the point where it's broad green only lets in 50% more light than the narrow (more like an X1 or Wratten 58) Bayer green.

Q3: Would it be possible to create an electrically controlled CFA, that could be transparent (grayscale sensor) or color filtered, by e.g. direction or intensity of some controlling current. If so, would it be applicable to this small sensor pitches we have today?

It's not really practical. Electronically controlled filters are pretty inefficient (and also generally polarization dependent, meaning a 60% light loss for high efficiency polarizers, and the need to have a phase retarder in the optical path as a "polarization scrambler".

Well, it turned out much more than three questions, but it is a very interesting topic for me..

Me too.

-- hide signature --

Rahon Klavanian 1912-2008.

Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed.

Ciao! Joseph

http://www.swissarmyfork.com

 Joseph S Wisniewski's gear list:Joseph S Wisniewski's gear list
Nikon D90 Nikon D2X Nikon D3 Nikon D100 Nikon Z7 +45 more
TomFla Regular Member • Posts: 489
Re: Aleksandar, I was debating not answering...

Joseph S Wisniewski wrote:
SNIP

The problem was that every one of those things turned out just not be there.

  • binning. The biggest customer ignored it completely. No liveview, no video mode, and they mixed the binned 4:1 shooting mode with a variety of gloopy shoot full res, scale down 3:2 or 2:1 modes. Other customers, once we looked at the low QE, realized that a Bayer sensor, simply tossing 3/4 of the pixels for a video or liveview mode, still had at least a stop of sensitivity advantage over the Foveon "super pixels".

SNIP

Ever since I got a 7d and then a 1d4 I have been a huge fan of liveview and video mode. Since I also have a history with Sigma bodies I really hope the sd1 has live view and hdmi video.

But my experience with video on the Canons leads me to think Sigma being a stop behind a Bayer sensor is not such a big deal. It is common to shoot video at a shutter speed close to twice the fps, so for 60fps you shoot 1/120 and for 30fps you shoot 1/60. Since one claimed advantage of using a dslr for video is the ability to open up those nice dslr lens to 2.8 or even wider I often wind up shooting at ISO 100. Not to mention I will probably have to spring for some variable ND filters unless I want to get a matte box which can also double as a babe magnet.
--
Those who forget history are condemned to go to summer school.

Keyboard shortcuts:
FForum MMy threads