Latest news about FF sigma sensor: dual pixel design

Started Dec 6, 2018 | Discussions
LaMeule Regular Member • Posts: 384
Latest news about FF sigma sensor: dual pixel design
4
 LaMeule's gear list:LaMeule's gear list
Ricoh GR Sigma SD1 Merrill Canon EOS 6D Panasonic Lumix DMC-G6 Olympus OM-D E-M10 II
Tiger1 Contributing Member • Posts: 560
Re: Latest news about FF sigma sensor: dual pixel design
14

RIP 1:1:1 and 3 layer design....... 

 Tiger1's gear list:Tiger1's gear list
Sony RX1 Canon EOS 5D Mark II Fujifilm FinePix S5 Pro Sigma SD14 Sigma SD1 Merrill +14 more
bizi clop
bizi clop Contributing Member • Posts: 558
Re: Latest news about FF sigma sensor: dual pixel design
1

This is not funny

Each pixel has two layers of "B and G" or "G and R" as shown in a figure of embodiment

All pixels have G layer (a) and missing colors of R or B at each pixels are interpolated as virtual Bayer array (b).

 bizi clop's gear list:bizi clop's gear list
Canon PowerShot G2 Sigma DP2 Merrill Canon PowerShot SX50 HS
Johan Borg Veteran Member • Posts: 3,280
Re: Latest news about FF sigma sensor: dual pixel design

bizi clop wrote:

This is not funny

Each pixel has two layers of "B and G" or "G and R" as shown in a figure of embodiment

All pixels have G layer (a) and missing colors of R or B at each pixels are interpolated as virtual Bayer array (b).

Interesting! So this is a bit like a Quattro sensor where every pixel has luminance information, but instead of 2 values for the lower layers, you get 4 per 2x2 pixel block.

Judging from earlier patents, I sort of expected a 3 layer Foveon sensor with a CFA on top for the full frame sensor, but this is more elegant.

 Johan Borg's gear list:Johan Borg's gear list
Sigma dp2 Quattro Sigma fp Sigma 45mm F2.8 DG DN
Johan Borg Veteran Member • Posts: 3,280
Re: Latest news about FF sigma sensor: dual pixel design
1

Johan Borg wrote:

bizi clop wrote:

This is not funny

Each pixel has two layers of "B and G" or "G and R" as shown in a figure of embodiment

All pixels have G layer (a) and missing colors of R or B at each pixels are interpolated as virtual Bayer array (b).

Interesting! So this is a bit like a Quattro sensor where every pixel has luminance information, but instead of 2 values for the lower layers, you get 4 per 2x2 pixel block.

Judging from earlier patents, I sort of expected a 3 layer Foveon sensor with a CFA on top for the full frame sensor, but this is more elegant.

After having a brief look at the patent it seems like they expect two uses:

- Foveon processing pipeline that prevents false colours and

- Bayer processing pipeline for high speed at lower cost.

Maybe this is how you get 4K video from a Sigma camera? Or how you feed a high definition EVF?

 Johan Borg's gear list:Johan Borg's gear list
Sigma dp2 Quattro Sigma fp Sigma 45mm F2.8 DG DN
Josh Leavitt
Josh Leavitt Regular Member • Posts: 276
Re: Latest news about FF sigma sensor: dual pixel design

It certainly is a different spin on Foveon, that's for sure. I would lament the loss of the 3-layer sensor design if this new architecture made it's way into future Sigma cameras. But... It's difficult to overlook the potential gains this system offers. Luminance data is still gathered from every photodiode, while blue and red pixels are exactly half of the overall green/luma channel. It partly resembles a Bayer CFA since demosaic algorithms will certainly be required, but has an edge due to twice the luminance resolution.

Let me put this way - I'd most definitely prefer Foveon Quattro for the upcoming L-mount camera (being stills-oriented I'm assuming), but I could get behind a sensor like this for a video/cinema solution. Especially if it meant higher FPS, and lower noise at higher signal amplification. I'm a bit curious how Sigma would go about demosaicing with the blue and red channels in a striped array rather than a conventional checkered array. It seems like bicubic interpolation would be necessary to avoid artifacts.

 Josh Leavitt's gear list:Josh Leavitt's gear list
Canon EOS 5D Canon EOS 30D Fujifilm X-T1 Canon 6D Mark II Canon EF 35mm F2.0 +7 more
veebegin Forum Member • Posts: 63
Re: Latest news about FF sigma sensor: dual pixel design

You can see Full text on this link here

This is not 2 layer type,this is 2:1:1 Bayer simulation mode on three layer with Bayer filter on top layer type.

[0019]
An eighth invention for solving the above-mentioned problems is an image signal processing apparatus, which detects by using a stacked solid-state image pickup device in which three or more photoelectric conversion layers are laminated on a substrate in one pixel An image signal processing apparatus for processing a stacked type image signal, comprising: an image signal reading section for reading two layers of image signals for each pixel of the stacked solid state image pickup element; and 2 And an image signal rearranging unit that rearranges the read image signal of the layer to a Bayer type image signal, wherein the two layer image signals read out by the image signal reading unit are two or more sets A combination of the image signal processing apparatus and the image signal processing apparatus.

[0050]
Control of the read control unit 1011 in step # 1 selects and reads out two layers of image signals out of the three layers of image signals in each pixel of the stacked solid state imaging device 101.

 veebegin's gear list:veebegin's gear list
Sigma DP3 Merrill Sigma dp2 Quattro Sigma sd Quattro H Panasonic S1 Canon EF 24-70mm F2.8L II USM +1 more
nick webster
nick webster Senior Member • Posts: 2,180
It's only a rumour
2

LaMeule wrote:

https://www.mirrorlessrumors.com/new-sigma-foveon-sensor-patent-discloses-dual-pixel-design/#disqus_thread

cheers

If true, at the rate Sigma are moving away from the original 1:1:1 design they might as well just go Bayer now

Nick

Johan Borg Veteran Member • Posts: 3,280
Re: Latest news about FF sigma sensor: dual pixel design
1

veebegin wrote:

You can see Full text on this link here

This is not 2 layer type,this is 2:1:1 Bayer simulation mode on three layer with Bayer filter on top layer type.

I expected a Bayer filter there, but it sounds like they rather simulate a Bayer filter by reading alternating B/R layers. But yes, looks like a 3 layer sensor where not every layer is read for every pixel row.

 Johan Borg's gear list:Johan Borg's gear list
Sigma dp2 Quattro Sigma fp Sigma 45mm F2.8 DG DN
tagscuderia
tagscuderia Senior Member • Posts: 1,952
Intellectual Property exercise
10

With Sony and Canon both rumoured to have Foveon-esque sensors in R&D phase, hopefully this is just SIGMA submitting patents to protect their IP and monopoly on multi-layered sensors.

I want my Foveon luminance data to be captured across all layers so hints of anything other than a true X3 sensor makes me glad that I bought an SD1M when I did!

 tagscuderia's gear list:tagscuderia's gear list
Sigma DP3 Merrill Sigma SD1 Merrill Sigma 15mm F2.8 EX DG Diagonal Fisheye Sigma 18-35mm F1.8 DC HSM Art Sigma 50-100mm F1.8 DC HSM Art
Johan Borg Veteran Member • Posts: 3,280
Re: It's only a rumour
3

nick webster wrote:

LaMeule wrote:

https://www.mirrorlessrumors.com/new-sigma-foveon-sensor-patent-discloses-dual-pixel-design/#disqus_thread

cheers

If true, at the rate Sigma are moving away from the original 1:1:1 design they might as well just go Bayer now

Well, if Merrill is 4:4:4 and Quattro is 4:1:1, then this is 4:2:2, so sort of a compromise

 Johan Borg's gear list:Johan Borg's gear list
Sigma dp2 Quattro Sigma fp Sigma 45mm F2.8 DG DN
target5
target5 Contributing Member • Posts: 813
Re: Latest news about FF sigma sensor: dual pixel design
11

If these rumors will be confirmed from production, Sigma again took the wrong path, creating a mockery product. The right path is to invest in developing custom-made fast D.S.P. processors to speed up the read-out of the original 1:1:1 classic Foveon design and made from here the FF cameras. We will have to laugh again when Sigma will claim that this new bastardization tech will have results like the original classic 1:1:1 structure layer Foveon (if not better)!

I guess that at the same time the refinement of Quattro based cameras will end: firmwares and SPP too.

 target5's gear list:target5's gear list
Sigma DP2 Sigma DP2s Sigma DP2 Merrill Sigma DP3 Merrill Sigma SD15 +4 more
Johan Borg Veteran Member • Posts: 3,280
Re: Latest news about FF sigma sensor: dual pixel design

target5 wrote:

If these rumors will be confirmed from production, Sigma again took the wrong path, creating a mockery product. The right path is to invest in developing custom-made fast D.S.P. processors to speed up the read-out of the original 1:1:1 classic Foveon design and made from here the FF cameras.

It's presumably a lot cheaper to use the same Fujitsu processor family as Leica and Nikon. Making better use of that processor makes sense.

We will have to laugh again when Sigma will claim that this new bastardization tech will have results like the original classic 1:1:1 structure layer Foveon (if not better)!

Let's see

I guess that at the same time the refinement of Quattro based cameras will end: firmwares and SPP too.

Yes, firmware updates tend to end just before launch of a new generation, but 5 years of updates isn't bad in these smartphone days.

 Johan Borg's gear list:Johan Borg's gear list
Sigma dp2 Quattro Sigma fp Sigma 45mm F2.8 DG DN
FDecker Senior Member • Posts: 2,264
What....?
9

I don‘t get it. This design corrupts the idea behind the Foveon design. The Quattro was the first step away from full color information at full resolution with the two lower layers binned to a lower resolution. But at least, at every location, all three colors were acquired. This new design even skips full color information at any location on the sensor. In addition, there is a severe misconception. The blue/green pixels works ok. But green/red either needs a blue filter or they sort out the blue photons by absorption in silicon but that means these could have been converted to a signal just like in a normal Foveon. So, why throwing away this information? Just to save processing and writing time? In this case, go ahead, why not just use one color per pixel and go with Bayer filter.

Repeat: I don‘t get it. The Sigma concept was about the best achievable image quality. Even if this is only possible at low ISO with a lot of light. And now, frames per second is of higher importance?

Writing times can be reduced by limiting the pixel count. With very high quality pixels and high acuity, this is the way to go. Oversampling and filtering/decimation of data can easier be done with Bayer sensors.

Just my 2 cents. I hope, this will not be the FF sensor from Sigma.

Hornbrille
Hornbrille Regular Member • Posts: 371
Re: Latest news about FF sigma sensor: dual pixel design
2

It looks like a brilliant concept to me. Less noise, much faster processing and raw files are probably Lightroom compatible. I also assume a lack of color artifacts as there is no color interpolation. The stacked pixels are just put beneath each other. Maybe the FF Sigma camera will be a 75MP Bayer equivalent with usable 1200 ISO

 Hornbrille's gear list:Hornbrille's gear list
Sony a7 II Sigma sd Quattro H Sony a7 III Panasonic Lumix DC-S1R Sony FE 35mm F2.8 +6 more
Scottelly
Scottelly Forum Pro • Posts: 16,684
Re: Latest news about FF sigma sensor: dual pixel design

This looks like Sigma is blending Foveon into a CFA concept without the filters, which could actually be an advantage in time. The color filters of CFA cameras may fade over time. If so, that's a bad thing, while the Foveon sensor colors should stay the same over time. Still, this is interesting. As long as the image quality is as good or better than the image quality of Quattro, I guess it is acceptable. I wonder what the reasoning is for switching. I'm guessing it's a way to reduce the quantity of circuitry. It does seem like there are going to be a lot more green pixels (four times as many). It also looks like there will be twice as many red pixels. This is definitely a major departure from Quattro.

-- hide signature --

Scott Barton Kennelly
http://www.bigprintphotos.com

 Scottelly's gear list:Scottelly's gear list
Sony SLT-A65 Nikon D810 Sigma sd Quattro H Sony DT 18-55mm F3.5-5.6 SAM Sigma 18-200mm F3.5-6.3 DC OS HSM +26 more
Scottelly
Scottelly Forum Pro • Posts: 16,684
Re: Latest news about FF sigma sensor: dual pixel design

Hornbrille wrote:

It looks like a brilliant concept to me. Less noise, much faster processing and raw files are probably Lightroom compatible. I also assume a lack of color artifacts as there is no color interpolation.

Where do you get the idea that there is no color interpolation? With red missing from half the pixels and blue missing from the other half of pixels, this design will require demosaicing of the image . . . kind of the way a Bayer pattern CFA does. This means there will most likely be color moiré in some photos . . . though maybe far less often than there is with a Bayer pattern CFA image. I guess those who said the last TRUE Foveon was Merrill may have indeed been right after-all . . . though Quattro in low-res mode is indeed a true Foveon concept imaging machine . . . though unfortunately lower resolution than the sensors in the Merrill cameras. I certainly hope the new full-frame camera has a Quattro or a 1:1:1 sensor and not this weird thing . . . though maybe I'm being hasty and just don't understand the meaning of this new thing. Maybe the full-frame camera will offer 4Kp60 video and much faster operation, because of this sensor design. Maybe this sensor design will be what launches Foveon into the world of mainstream imaging.

The stacked pixels are just put beneath each other. Maybe the FF Sigma camera will be a 75MP Bayer equivalent with usable 1200 ISO

-- hide signature --

Scott Barton Kennelly
http://www.bigprintphotos.com

 Scottelly's gear list:Scottelly's gear list
Sony SLT-A65 Nikon D810 Sigma sd Quattro H Sony DT 18-55mm F3.5-5.6 SAM Sigma 18-200mm F3.5-6.3 DC OS HSM +26 more
nick webster
nick webster Senior Member • Posts: 2,180
Re: What....?
1

FDecker wrote:

I don‘t get it. This design corrupts the idea behind the Foveon design. The Quattro was the first step away from full color information at full resolution with the two lower layers binned to a lower resolution. But at least, at every location, all three colors were acquired. This new design even skips full color information at any location on the sensor. In addition, there is a severe misconception. The blue/green pixels works ok. But green/red either needs a blue filter or they sort out the blue photons by absorption in silicon but that means these could have been converted to a signal just like in a normal Foveon. So, why throwing away this information? Just to save processing and writing time? In this case, go ahead, why not just use one color per pixel and go with Bayer filter.

Repeat: I don‘t get it. The Sigma concept was about the best achievable image quality. Even if this is only possible at low ISO with a lot of light. And now, frames per second is of higher importance?

Writing times can be reduced by limiting the pixel count. With very high quality pixels and high acuity, this is the way to go. Oversampling and filtering/decimation of data can easier be done with Bayer sensors.

Just my 2 cents. I hope, this will not be the FF sensor from Sigma.

I thought much the same - if they are heading down this route just use a Bayer sensor - Sony do some lovely ones that are low noise, fast read outs etc. Or I imagine Panasonic will put a couple of nice ones in their two upcoming L-Mount cameras.

I value my Merrills for the look they produce and put up with the other quirks - if the output is going to start resembling Bayer sensors then I might as well just stick with my Bayer cameras with good AF, good high ISO ... and save my money.

Nick

jimbocourier Regular Member • Posts: 170
Re: Latest news about FF sigma sensor: dual pixel design

Scottelly wrote:

This looks like Sigma is blending Foveon into a CFA concept without the filters, which could actually be an advantage in time. The color filters of CFA cameras may fade over time. If so, that's a bad thing, while the Foveon sensor colors should stay the same over time. Still, this is interesting. As long as the image quality is as good or better than the image quality of Quattro, I guess it is acceptable. I wonder what the reasoning is for switching. I'm guessing it's a way to reduce the quantity of circuitry. It does seem like there are going to be a lot more green pixels (four times as many). It also looks like there will be twice as many red pixels. This is definitely a major departure from Quattro.

Hey Scott do you think this is for the full frame coming next year or for future generations? I assume the sensor being used next year has already been developed (and patented)?

 jimbocourier's gear list:jimbocourier's gear list
Sigma DP2 Merrill
Scottelly
Scottelly Forum Pro • Posts: 16,684
Re: It's only a rumour

nick webster wrote:

LaMeule wrote:

https://www.mirrorlessrumors.com/new-sigma-foveon-sensor-patent-discloses-dual-pixel-design/#disqus_thread

cheers

If true, at the rate Sigma are moving away from the original 1:1:1 design they might as well just go Bayer now

Maybe that's the plan. They may have decided long ago that they were going to have to move their customers away from Foveon, and the Quattro was just the first step in a three-step plan to switch to the same sensor design that pretty much everyone else is using. They may have decided there is not enough difference in the cost of the sensors or the image quality that it is worth continuing to pursue the Foveon sensor technology. The next sensor may indeed be a Bayer patterned CFA sensor, like all the others have. This may allow them to create cameras that are as fast as those from other manufacturers, offer the same dynamic range as those from other manufacturers, can do video, like the cameras from other manufacturers, and are basically low-cost cameras that compete with the rest, rather than kings of one thing, but pawns of most other things.

Nick

If Quattro was indeed the beginning of the end of Foveon, I am sad about it . . . but I'll reserve my sadness until I know what's going on for certain . . . which may take a long time to find out. Certainly I'm pretty happy with my SD Quattro H, and I will use it most of the time, even though I now have a Nikon D810 to use, a past dream of mine. Still, I will buy a 24-85mm f3.5-4.5 VR G for that Nikon, and I'll use it some of the time, until I can get the Panasonic S1R and the Sigma full-frame L-mount camera. I frankly don't know what I'll do with my Sony A7 now. Maybe I'll sell it, but I guess it might be worth keeping, just in case I want to shoot some starscapes with my Sigma SA lenses.

Anyway, if the new full-frame sensor is this weird new design, I look forward to seeing what the images are like. I sure hope they're better than Quattro images. There will be a new version of SPP, no doubt.

-- hide signature --

Scott Barton Kennelly
http://www.bigprintphotos.com

 Scottelly's gear list:Scottelly's gear list
Sony SLT-A65 Nikon D810 Sigma sd Quattro H Sony DT 18-55mm F3.5-5.6 SAM Sigma 18-200mm F3.5-6.3 DC OS HSM +26 more
Keyboard shortcuts:
FForum MMy threads