3 years Foveon FF Sensor Development

Started 9 months ago | Discussions
rf-design
rf-design Contributing Member • Posts: 552
3 years Foveon FF Sensor Development
6

The fullframe sensor development startet 3 years ago and is still under development. That seems to me a quite long time for a Quattro scaling to a different dimension. I expect that for a 3 years time line a much different sensor architecture was the target. Not from the pixel architecture level but my guess is that the Foveon FF sensor contain for the first time ADCs. That is a very critical component because only digital readout with integrated ADCs allow much lower system power under functionality typical required for modern mirrorless.

If the pixel architecture is based on Quattro density there are 66M diodes to read from a 44M array. For a 4 shots/s camera external ADCs need to convert more than 264MS/s. For operational battery lifetime the energy needed to drive the conversion for a shot taken is critical for DSLRs. For mirrorless the liveview requires a continious readout of a subset of all pixels. If all pixels need to be read for display the number of conversions are 50/4=12.5 times higher than for reading a full picture. So there is a trade-off of display resolution versus power drain and battery lifetime. Similar for PDAF on sensor there must be a readout of some percentage of the pixel array with a higher update rate, typical better than 10ms.

All this is only possible with low power with integrated ADCs. That seems to me the main reason for the 3 years development. The second reason is the "Yamaki Smile" regarding video functionality. So the FF Foveon will have video.

Now lets speculate:

APS-C Quattro: 5424x3616 (same pixel architecture)

FF-Quattro: 8136x5424 (66.2M-diodes)

If the integrated ADCs readout rate is 66.2Msamples each 1/7.5 seconds you can provide the following camera functions:

1. Shooting burst rate 7.5/s (100% ADC-Power)

2. 4k30p video at 1/4 subsampling (100% ADC Power)

3. Liveview 2k resolution at 1/30s update rate and 1/16 subsampling (25% ADC Power)

4. PDAF at 1/120s update with 6.25% of the pixel area (100% ADC Power)

To be specific the patent and actual Quattro use only the top layer for PDAF. So the lower layers do not to be read. But this reduce the rate only by 1/3.

It will be an fantastic camera based on this numbers. It is not in the same class as an A7R3 which could read all diodes and interpolate for 4k30p or the  Lumix S1R which again double the readout over the A7R3.

 rf-design's gear list:rf-design's gear list
Panasonic LX100 Canon EOS 40D Sigma SD10 Sigma SD14 Sigma SD15 +60 more
41mm
41mm Regular Member • Posts: 255
Re: 3 years Foveon FF Sensor Development
3

Let Sony&Panasonic be specialists in Video, and SIGMA in Image Quality from 50-400 ISO.

NOT MORE - not less.

Video and touch screen for the rest of the other companies.

LESS IS MORE (sometimes)

;–)

 41mm's gear list:41mm's gear list
Sigma fp
digi2ap Contributing Member • Posts: 738
Re: 3 years Foveon FF Sensor Development

I've always thought that video will be a considered part of the next sensor architecture but (more than photography) video users requires gain or in other words higher ISO.  Maybe it is an improved video output for the EVF instead.

 digi2ap's gear list:digi2ap's gear list
Sigma dp2 Quattro Sigma dp1 Quattro Panasonic Lumix DMC-GH3 Sigma 18-35mm F1.8 DC HSM Art
D Cox Forum Pro • Posts: 23,442
Re: 3 years Foveon FF Sensor Development
1

I get the impression from this that the new camera will not be available until late next year.

Plenty of time to save up for one.   

xpatUSA
xpatUSA Forum Pro • Posts: 14,895
Re: 3 years Foveon FF Sensor Development
1

A thoughtful piece, Reiner, thank you for posting.

Since the current ADCs are buried in the AFE chip, (I assume), perhaps the FF system would revert to ISO-less?

-- hide signature --

Ted

 xpatUSA's gear list:xpatUSA's gear list
Sigma DP2 Sigma DP2 Merrill Panasonic Lumix DMC-G1 Sigma SD15 Panasonic Lumix DMC-GH1 +16 more
Paul Petersen
Paul Petersen Senior Member • Posts: 1,296
Re: 3 years Foveon FF Sensor Development
2

Nice of you to weigh in and speculate.

My hope and dream for this is they will be able to go to a BSI X3 chip design getting all the wiring out of the way to get more light to the deeper photo diodes. IMHO it would be advantageous return to a 1:1 RGB aka Merrill stack since it just seems to have lost its acuity in the Quattro generation. Maybe Sigma sourcing a Processor chip with power and efficency like the one in the new Fuji X-T3 will bring it all up to speed especially if they want to do video.

My other question is if they will offer a smaller version of chip in APS-C/H to better serve the middle of the market.

Paul

-- hide signature --

A bad day of train chasing is better than a good day at work.
http://peterzpicts.smugmug.com/

 Paul Petersen's gear list:Paul Petersen's gear list
Sigma SD14 Nikon D7500 Nikon AF-S Nikkor 70-300mm f/4.5-5.6G VR Sigma 18-200mm F3.5-6.3 DC OS HSM Sigma 18-35mm F1.8 DC HSM Art +2 more
Dryce Regular Member • Posts: 369
Re: 3 years Foveon FF Sensor Development

Paul Petersen wrote:

My other question is if they will offer a smaller version of chip in APS-C/H to better serve the middle of the market.

I don't think thre is any point in them going for a APS-C or APS-H body with interchangeable lenses once they go FF.

However it may well make sense to them to use an APS-C version of any newer sensor in a new generation DP family.

 Dryce's gear list:Dryce's gear list
Sigma DP1 Merrill Sigma DP2 Merrill Sigma DP3 Merrill Sigma dp0 Quattro Sony Alpha DSLR-A700 +3 more
EEvan
EEvan Regular Member • Posts: 100
Re: 3 years Foveon FF Sensor Development
1

While I wouldn't plan to shoot any video with it, a sensor that can do video with lower power does mean the LV should be improved, which might be nice.

Scottelly
Scottelly Forum Pro • Posts: 12,643
Re: 3 years Foveon FF Sensor Development

Dryce wrote:

Paul Petersen wrote:

My other question is if they will offer a smaller version of chip in APS-C/H to better serve the middle of the market.

I don't think thre is any point in them going for a APS-C or APS-H body with interchangeable lenses once they go FF.

However it may well make sense to them to use an APS-C version of any newer sensor in a new generation DP family.

Agreed, and to keep costs down Sigma is likely to offer something less expensive than the full-frame camera at some point. They already lead the APS-C market for low-ISO image quality, so why not retain that lead with a new camera that can do video too? I mean if they're going to get into the video world with the competition, doesn't it make sense for them to eventually offer a lower-priced camera that can do video in addition to their full-frame camera?

-- hide signature --

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

 Scottelly's gear list:Scottelly's gear list
Sony SLT-A65 Sigma SD1 Merrill Nikon D810 Sigma sd Quattro H Sony DT 18-55mm F3.5-5.6 SAM +21 more
Scottelly
Scottelly Forum Pro • Posts: 12,643
Re: 3 years Foveon FF Sensor Development

Can we assume that taking three years is not because of some technical difficulty? Could it be that they have had trouble with the process of going full-frame? The three year time frame does seem like a long time to do something that they seemed to do in one year with the H. Again, going full-frame may have very well presented quite a lot of difficulty. The development of a camera that could handle the large quantity of data may have been a consideration during that time frame too, so the three year development time may have been so long, because they were in no rush to develop it, because they had to create a camera to handle such a sensor anyway. This makes me think new sensor could be a Quattro that offers very high resolution. I certainly hope that is the case. I think that would fit Mr. Yamaki's desire to create a camera with high resolution, like a medium format or large format camera . . . a camera that can take advantage of the incredible image quality that new 40mm f1.4 Art lens has to offer.

I so wish that interviewer had asked more questions, like "Will there be Wi-Fi in this new camera?" and "Will the new camera have stabilization in the body?" and "Do you plan to introduce some new short flange distance lenses next year with the new camera, or are you concentrating on just making short flange distance versions of the lenses you're making for the DSLR market?"

-- hide signature --

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

 Scottelly's gear list:Scottelly's gear list
Sony SLT-A65 Sigma SD1 Merrill Nikon D810 Sigma sd Quattro H Sony DT 18-55mm F3.5-5.6 SAM +21 more
rf-design
OP rf-design Contributing Member • Posts: 552
Re: 3 years Foveon FF Sensor Development

Scottelly wrote:

Can we assume that taking three years is not because of some technical difficulty? Could it be that they have had trouble with the process of going full-frame? The three year time frame does seem like a long time to do something that they seemed to do in one year with the H.

The APS-C and H do not need stitching. The stitching for FF is done in the fab but there could be nonuniformity effect which are visible as stitching line in the image. So I guess all FF sensors are stitched once on a vertical line aligned in between the pixels. Only if something basic in the pixel have to be changed to overcome negative effects of the stitching could imagine a 3 year timeline.

 rf-design's gear list:rf-design's gear list
Panasonic LX100 Canon EOS 40D Sigma SD10 Sigma SD14 Sigma SD15 +60 more
rf-design
OP rf-design Contributing Member • Posts: 552
Re: 3 years Foveon FF Sensor Development

EEvan wrote:

While I wouldn't plan to shoot any video with it, a sensor that can do video with lower power does mean the LV should be improved, which might be nice.

Yes, that is a point for improvement. Actual the SDQ have less LV resolution than the EVF display. The noise could also be lowered if you read more pixel and integrate them over time. The same is done in Canon LV for a decade now.

 rf-design's gear list:rf-design's gear list
Panasonic LX100 Canon EOS 40D Sigma SD10 Sigma SD14 Sigma SD15 +60 more
rf-design
OP rf-design Contributing Member • Posts: 552
Re: 3 years Foveon FF Sensor Development

xpatUSA wrote:

A thoughtful piece, Reiner, thank you for posting.

Since the current ADCs are buried in the AFE chip, (I assume), perhaps the FF system would revert to ISO-less?

Socionext (Fujitsu before) having only digital sensor inputs. So it must be a separate chip for ADCs up to now. The Milbeaut processor series have a record history of design wins:

Pentax, Nikon, Leica, Sigma, GoPro

https://en.wikipedia.org/wiki/Milbeaut

Implementing a preamplifier to reduce the SNR contribution at lower sensor light is not the most power efficient strategy. You gain only SNR if the noise of the amp is lower than the comparator. That is a hole different story than a Dual-Gain pixel architecture. In this architecture the intrinsic voltage gain of the high-gain pixel configuration is increased typical at the cost of less well capacity. But I am not shure if the latest EXMOR sensors are using dual-gain pixels or dual-gain amps.

If modern asynchronous sucessive approximation is used for ADCs dual-gain could be implemented with reference scaling but there is no SNR advantage.

Coming back to your question; ISO-less or Gain-Step-Less ADCs are more a question of power efficiency. To me it seems that the hole mirroless architecture is dominated by power efficiency of the sensor data conversion.

 rf-design's gear list:rf-design's gear list
Panasonic LX100 Canon EOS 40D Sigma SD10 Sigma SD14 Sigma SD15 +60 more
xpatUSA
xpatUSA Forum Pro • Posts: 14,895
Re: 3 years Foveon FF Sensor Development

rf-design wrote:

xpatUSA wrote:

A thoughtful piece, Reiner, thank you for posting.

Since the current ADCs are buried in the AFE chip, (I assume), perhaps the FF system would revert to ISO-less?

Socionext (Fujitsu before) having only digital sensor inputs. So it must be a separate chip for ADCs up to now. The Milbeaut processor series have a record history of design wins:

Pentax, Nikon, Leica, Sigma, GoPro

https://en.wikipedia.org/wiki/Milbeaut

Implementing a preamplifier to reduce the SNR contribution at lower sensor light is not the most power efficient strategy. You gain only SNR if the noise of the amp is lower than the comparator. That is a hole different story than a Dual-Gain pixel architecture. In this architecture the intrinsic voltage gain of the high-gain pixel configuration is increased typical at the cost of less well capacity. But I am not shure if the latest EXMOR sensors are using dual-gain pixels or dual-gain amps.

If modern asynchronous sucessive approximation is used for ADCs dual-gain could be implemented with reference scaling but there is no SNR advantage.

Coming back to your question; ISO-less or Gain-Step-Less ADCs are more a question of power efficiency. To me it seems that the hole mirroless architecture is dominated by power efficiency of the sensor data conversion.

I do not understand your response, Reiner.

According to figures 4 and 5 analog-to-digital conversion is external to the Milbeaut device:

http://www.fujitsu.com/downloads/EDG/binary/pdf/find/24-1e/3.pdf

There it is called an "AFE" which may or may not be a chip.

Your last paragraph went right over my head and did not help me at all, sorry.

I withdraw my original assumption and related question.

-- hide signature --

Ted

 xpatUSA's gear list:xpatUSA's gear list
Sigma DP2 Sigma DP2 Merrill Panasonic Lumix DMC-G1 Sigma SD15 Panasonic Lumix DMC-GH1 +16 more
rf-design
OP rf-design Contributing Member • Posts: 552
Re: 3 years Foveon FF Sensor Development

xpatUSA wrote:

rf-design wrote:

xpatUSA wrote:

A thoughtful piece, Reiner, thank you for posting.

Since the current ADCs are buried in the AFE chip, (I assume), perhaps the FF system would revert to ISO-less?

Socionext (Fujitsu before) having only digital sensor inputs. So it must be a separate chip for ADCs up to now. The Milbeaut processor series have a record history of design wins:

Pentax, Nikon, Leica, Sigma, GoPro

https://en.wikipedia.org/wiki/Milbeaut

Implementing a preamplifier to reduce the SNR contribution at lower sensor light is not the most power efficient strategy. You gain only SNR if the noise of the amp is lower than the comparator. That is a hole different story than a Dual-Gain pixel architecture. In this architecture the intrinsic voltage gain of the high-gain pixel configuration is increased typical at the cost of less well capacity. But I am not shure if the latest EXMOR sensors are using dual-gain pixels or dual-gain amps.

If modern asynchronous sucessive approximation is used for ADCs dual-gain could be implemented with reference scaling but there is no SNR advantage.

Coming back to your question; ISO-less or Gain-Step-Less ADCs are more a question of power efficiency. To me it seems that the hole mirroless architecture is dominated by power efficiency of the sensor data conversion.

I do not understand your response, Reiner.

According to figures 4 and 5 analog-to-digital conversion is external to the Milbeaut device:

http://www.fujitsu.com/downloads/EDG/binary/pdf/find/24-1e/3.pdf

There it is called an "AFE" which may or may not be a chip.

Your last paragraph went right over my head and did not help me at all, sorry.

I withdraw my original assumption and related question.

Ted,

ISO-less is a system decision that the SNR advantage of a switchable gain, either in pixel, read-amplifier or in the AFE external to the sensor is worth the effort. So ISO-less is simply underexposure a single-gain processing path.

For the coming FF Foveon there is no clear indication for ISO-less operation.

BR

Reiner

 rf-design's gear list:rf-design's gear list
Panasonic LX100 Canon EOS 40D Sigma SD10 Sigma SD14 Sigma SD15 +60 more
xpatUSA
xpatUSA Forum Pro • Posts: 14,895
Re: 3 years Foveon FF Sensor Development

rf-design wrote:

xpatUSA wrote:

rf-design wrote:

xpatUSA wrote:

A thoughtful piece, Reiner, thank you for posting.

Since the current ADCs are buried in the AFE chip, (I assume), perhaps the FF system would revert to ISO-less?

Socionext (Fujitsu before) having only digital sensor inputs. So it must be a separate chip for ADCs up to now. The Milbeaut processor series have a record history of design wins:

Pentax, Nikon, Leica, Sigma, GoPro

https://en.wikipedia.org/wiki/Milbeaut

Implementing a preamplifier to reduce the SNR contribution at lower sensor light is not the most power efficient strategy. You gain only SNR if the noise of the amp is lower than the comparator. That is a hole different story than a Dual-Gain pixel architecture. In this architecture the intrinsic voltage gain of the high-gain pixel configuration is increased typical at the cost of less well capacity. But I am not shure if the latest EXMOR sensors are using dual-gain pixels or dual-gain amps.

If modern asynchronous sucessive approximation is used for ADCs dual-gain could be implemented with reference scaling but there is no SNR advantage.

Coming back to your question; ISO-less or Gain-Step-Less ADCs are more a question of power efficiency. To me it seems that the hole mirroless architecture is dominated by power efficiency of the sensor data conversion.

I do not understand your response, Reiner.

According to figures 4 and 5 analog-to-digital conversion is external to the Milbeaut device:

http://www.fujitsu.com/downloads/EDG/binary/pdf/find/24-1e/3.pdf

There it is called an "AFE" which may or may not be a chip.

Your last paragraph went right over my head and did not help me at all, sorry.

I withdraw my original assumption and related question.

Ted,

ISO-less is a system decision that the SNR advantage of a switchable gain, either in pixel, read-amplifier or in the AFE external to the sensor is worth the effort. So ISO-less is simply underexposure a single-gain processing path.

Thanks, Reiner. By "ISO-less" I meant the same as SD9/SD10/SD14/SD1M that have no programmable gain amplifier between the sensor and the ADCs. Is that similar to what you said?

For the coming FF Foveon there is no clear indication for ISO-less operation.

Indeed, all we know about FF is that it FF!

-- hide signature --

Ted

 xpatUSA's gear list:xpatUSA's gear list
Sigma DP2 Sigma DP2 Merrill Panasonic Lumix DMC-G1 Sigma SD15 Panasonic Lumix DMC-GH1 +16 more
Keyboard shortcuts:
FForum MMy threads