Talking all things Sigma cameras and Foveon Podcast
Quote from James Warner on September 25, 2024, 12:32 pmHey all!
I finally got around to editing a conversation I had with Cory about Sigma/Foveon cameras. If you're looking for something to listen to while doing chores then check it out!
https://youtu.be/NvWj-T8eFAQ
Hey all!
I finally got around to editing a conversation I had with Cory about Sigma/Foveon cameras. If you're looking for something to listen to while doing chores then check it out!
Quote from Cory Maben on October 1, 2024, 5:10 amI'm glad you were able to get an episode out of it! It was a ton of fun!
I'm glad you were able to get an episode out of it! It was a ton of fun!
Quote from Christopher Chang on October 5, 2024, 9:52 pmGood stuff!
One day I gotta try a sigma camera again (last one I had was an SD10 some 11 years ago). I do got to say modern conventional Bayer Filter CMOS sensors have gotten insanely good just via brute force caused by their relative cost savings / energy consumption compared to say CCDs. I would say once the industry started moving away from AA filters was about when we hit peak CMOS.
I would disagree that they are out of places to go though, although modern sensors seem like they aren't making as big gains in where we'd normally expect like pixel density or dynamic range. They are doing something more subtle but I'd say just as important. Sensor read out speed. Hilariously enough CCDs didn't have this problem, atleast not in the way CMOS does. CCD's can have all their pixels storing data at once (Global shutter) but the sensor has to dump that data sequentially. That's where the reading out takes it's time, but in affect it's a global shutter with a very bad refresh before the sensor can do more photo-ing.
Not to mention the random dual or tri-gain sensors, those are very neat.
Modern CMOS sensors read the pixel data directly line by line as the photo is taking place, with more and more camera's being billed as hybrid video shooters this is something of an important thing, but more than that, with newer very fast read speeds or even global shutters on CMOS you can also open up new opportunities for photography, infinitely syncing flashes, extremely fast e-shutter speeds that don't take a Dynamic range hit, and have no jellloooooo.
So there is room, I do agree we are probably getting pretty close to an ideal pixel density though.
I also wanna point out Panasonic's new sensor tech does emphasis color accuracy, but it does so by optimizing input for the what is implied to be a Bayer Filter. Rather than what Foveon is doing.
https://news.panasonic.com/global/topics/13982
What they seem to be doing is reducing signal spill over from adjacent pixels with their coating and thus provides better raw data for the camera to do it's magic with its Bayer interpolation. It also appears to have the benefit of being pretty thin. The only thing that is a bit strange is that it's demonstrated in their press release on a normal FSI sensor and not on a modern BSI sensor which they do use as an example as a conventional sensor.
Anyways that is not to slag the Foveon, I think it's one of the most interesting sensor designs out there. But it was always going to be a uphill battle. The 3 layers on their own is a difficult issue to deal with, aside from the fact that there probably aren't that many fabs that can at scale make them their monstrously deep sensor. The sensor's poor ISO if I remember right was partly due to the depth of the lower level pixels, the light transmission wasn't perfect going down the sensor, it seems that even sigma kind of agree, looking at the way they've shifted the Quattro series Foveon sensors where they kind if split the difference between original idea of 3 layers of equal pixel density to 1 layer of spatial/color (mostly blue) pixels and 2 layers of color (red/green), meaning they are also now doing some interesting interpolation.
At any rate this new sensor lay out seems to have worked and it seems to be a very good move to help those 2 lower layers gather photons better.
At any rate I do hope we can see a full frame Foveon one day. However far in the future that day is. Until that day I'll probably see if I can hitch a ride on the Quattro H train and enjoy some APS-H goodness.
Good stuff!
One day I gotta try a sigma camera again (last one I had was an SD10 some 11 years ago). I do got to say modern conventional Bayer Filter CMOS sensors have gotten insanely good just via brute force caused by their relative cost savings / energy consumption compared to say CCDs. I would say once the industry started moving away from AA filters was about when we hit peak CMOS.
I would disagree that they are out of places to go though, although modern sensors seem like they aren't making as big gains in where we'd normally expect like pixel density or dynamic range. They are doing something more subtle but I'd say just as important. Sensor read out speed. Hilariously enough CCDs didn't have this problem, atleast not in the way CMOS does. CCD's can have all their pixels storing data at once (Global shutter) but the sensor has to dump that data sequentially. That's where the reading out takes it's time, but in affect it's a global shutter with a very bad refresh before the sensor can do more photo-ing.
Not to mention the random dual or tri-gain sensors, those are very neat.
Modern CMOS sensors read the pixel data directly line by line as the photo is taking place, with more and more camera's being billed as hybrid video shooters this is something of an important thing, but more than that, with newer very fast read speeds or even global shutters on CMOS you can also open up new opportunities for photography, infinitely syncing flashes, extremely fast e-shutter speeds that don't take a Dynamic range hit, and have no jellloooooo.
So there is room, I do agree we are probably getting pretty close to an ideal pixel density though.
I also wanna point out Panasonic's new sensor tech does emphasis color accuracy, but it does so by optimizing input for the what is implied to be a Bayer Filter. Rather than what Foveon is doing.
https://news.panasonic.com/global/topics/13982
What they seem to be doing is reducing signal spill over from adjacent pixels with their coating and thus provides better raw data for the camera to do it's magic with its Bayer interpolation. It also appears to have the benefit of being pretty thin. The only thing that is a bit strange is that it's demonstrated in their press release on a normal FSI sensor and not on a modern BSI sensor which they do use as an example as a conventional sensor.
Anyways that is not to slag the Foveon, I think it's one of the most interesting sensor designs out there. But it was always going to be a uphill battle. The 3 layers on their own is a difficult issue to deal with, aside from the fact that there probably aren't that many fabs that can at scale make them their monstrously deep sensor. The sensor's poor ISO if I remember right was partly due to the depth of the lower level pixels, the light transmission wasn't perfect going down the sensor, it seems that even sigma kind of agree, looking at the way they've shifted the Quattro series Foveon sensors where they kind if split the difference between original idea of 3 layers of equal pixel density to 1 layer of spatial/color (mostly blue) pixels and 2 layers of color (red/green), meaning they are also now doing some interesting interpolation.
At any rate this new sensor lay out seems to have worked and it seems to be a very good move to help those 2 lower layers gather photons better.
At any rate I do hope we can see a full frame Foveon one day. However far in the future that day is. Until that day I'll probably see if I can hitch a ride on the Quattro H train and enjoy some APS-H goodness.