The next gen sensor

[Bérenger Brillante]

We may be in a situation that RED had a more impressive chip design, and Sony has a fabrication process with the ability to form finer chip layouts.

But can they work together ? :mad5:

 

[Patrick Grossien]

A 64mm sensor would be a special camera that would cost very very much ...

Aah, don't worry. Only $45K for the body...

 

[AndreeMarkefors]

In terms of dynamic range…

Don't forget that on a strict sensor level, in linear light, EACH stop demands a doubling of the signal to noise ratio. So from 13 to 14 stops you need TWICE the sensor you had at 13. And then TWICE that again to go to 15... Exponential development gets demanding really fast…

These aren't just incremental improvements you're demanding... each stop is an epic achievement. Especially nowadays, where sensors are already performing at a high level. I don't believe in magic. Signal processing is getting better and there are technical solutions that can help (like maybe double dipping each photo diode with something like dual gain). I don't know, I just don't see any leaps in DR happening soon, that are based on real, actual light gathering.

 

[Jeff Kilgroe]

We don't want to see larger pixels. 6um pixels would be too large, IMO. The 5um current size is more appealing and it makes sense to move toward smaller pixels as the tech allows. Larger pixels can capture/ bin more light, but as we see continuous improvements in sensitivity and efficiency, we can maintain or improve the performance per pixel, even as size decreases.

Global shutter is a possibility and actually completely do-able with current and previous tech. The rolling read-out was originally a necessity, but kept as it created a more organic response from the sensor. However, it's not always desirable, just as global shutter is not always desirable for its overly sterile and rigid appearance. What I would like to see is the ability for the camera operator to create their own shutter blanking effects. Give us the option to set it to a global read-out if we want or a rolling read out if we prefer. For rolling shutter read-out, let us specify the speed, direction, even let us create masks or other options there. We could fully emulate a mechanical mirror shutter with this ability or create new effects. I've proposed this before on several occasions. Even way back before the MX sensor.

I would like to see a move away from Bayer color filtration and use direct wavelength and photon charge detection. That is not hypothetical tech, but something real. It has yet to make it into commercial products, so far only in the realm of lab creations and a few special sensors that have been used by NASA and ESA... We'll get there...

Sensor size... I would like to see an increase in available sensor size to match 645 MF or IMAX as an option. However, the larger formats are not going to be sustainable or even desirable for the vast majority of shooters.

 

[Patrick Grossien]

We don't want to see larger pixels. 6um pixels would be too large, IMO. The 5um current size is more appealing and it makes sense to move toward smaller pixels as the tech allows. Larger pixels can capture/ bin more light, but as we see continuous improvements in sensitivity and efficiency, we can maintain or improve the performance per pixel, even as size decreases.

Global shutter is a possibility and actually completely do-able with current and previous tech. The rolling read-out was originally a necessity, but kept as it created a more organic response from the sensor. However, it's not always desirable, just as global shutter is not always desirable for its overly sterile and rigid appearance. What I would like to see is the ability for the camera operator to create their own shutter blanking effects. Give us the option to set it to a global read-out if we want or a rolling read out if we prefer. For rolling shutter read-out, let us specify the speed, direction, even let us create masks or other options there. We could fully emulate a mechanical mirror shutter with this ability or create new effects. I've proposed this before on several occasions. Even way back before the MX sensor.

I would like to see a move away from Bayer color filtration and use direct wavelength and photon charge detection. That is not hypothetical tech, but something real. It has yet to make it into commercial products, so far only in the realm of lab creations and a few special sensors that have been used by NASA and ESA... We'll get there...

Sensor size... I would like to see an increase in available sensor size to match 645 MF or IMAX as an option. However, the larger formats are not going to be sustainable or even desirable for the vast majority of shooters.

Well said, Jeff!
As you said, I don't believe raising the pixel pitch would be beneficial to RED. If the pixel design and quality continues to rise as it did since the original Mysterium, I'm a happy camper

When / if a completely new sensor technology will be adapted by RED, even better. But already a CMOS sensor which uses the full pixel count for each color would be sweet. The monochrome sensor is amazingly detailed... Something like a prism split / 3chip would probably make no sense but damn, it would be nice

 

[Nils Ruinet]

I'm super happy withe the 6K Dragon sensor as it is. DR is amazing, resolution is big enough (I still mainly use 5K for most projects), the S35+ sensor size is fine by me, color science has gotten really good over the years (still some room for improvement probably, as always, but it's one of the best out there).
The only thing I miss sometimes compared to other recent cameras, is a higher clean ISO. I usually use the Dragon in the 400-800 ISO range, which is its sweet spot in my opinion. But sometimes I'd like to be able to go higher without bringing up too much noise. 3200 clean ISO would probably be the sweet spot. Looking at what the competition does, it does seem possible now.

So, for the next gen sensor, I would like :
- Higher sensitivity : noiseless 3200 ISO for example
- Improved DR would always be a plus of course, with HDR being around the corner, but it is already pretty amazing as it is.
- Higher framerates is always a plus too, but not the biggest priority in my opinion.
- 6K resolution is fine in a S35+ sensor. I'd rather not go higher and improve the sensitivity. Making photosites smaller would go against that.

Really curious to see what will be next !

 

[Mendes Nabil]

We went far enough in "sensor size, resolution and high frame rates", so the next gen sensor needs (an will i'm sure) focus on the basics :

-Color science

-Color science

-color science

-Texture and texture again, i feel like texture on Dragon sensor is "soul less" like a cold body freshly gone (it wasn't the case with Mx)

-Dynamic range (highlights rolloff need to be seriously improved)

- Higher sensitivity (as a little bonus)


I was asking myself recently : Why is the Red one Mx the "most Red filmic camera" ? The truth is that when they made the old queen, the only reference Red had was "Film camera", now with all these new digital cams their references have certainly changed..

So my word to the Red team is, please, while making the next gen sensor have in mind and as reference "The real and unique king of Cinema camera", aka Film..

 

[Antony Newman]

Heres something that piqued my interest.

A solid state parametric optical amplifier that is only 400nm in diameter

http://phys.org/news/2016-05-unveil-submicroscopic-tunable-optical-amplifier.html

Why is this interesting?

+) A Bayer filter uses a colored filter -> which reduces light levels and reduces signal-to-noise

+) This could replace the Bayer. It works like a LASER ... Optically amplifying the Frequency of light of interest and suppressing out of band colour frequencies. This would Increase light levels, and greatly increase the signal-to-noise

AJ

 

[Antony Newman]

Here something that may be of interest to RED:

http://phys.org/news/2016-05-optics-breakthrough-revamp-night-vision.html

It is a discovery that could be used to increase light absorption (in their example this increased the photon absorption from 7.7% -> 99%).

AJ

 

[Brodie Gwilliam]

I think it would be interesting to have a poll on what sensor people want most... lets call current dragon 16.5 DR for comparison

new sensor poll could look like -

6K FF sensor (current 8k size)
19 stops of dr
120 fps at full frame
5000 iso equivilant to current 2000 on dragon
$9000 upgrade price

or

8k FF (current 8k size)
18 stop DR
85 fps at fullframe
3200 iso equivilant to current 2000 on dragon
$15000 upgrade price


These are just guesses on the top of my head where things might lie and i could be way off on what the comparable specs would be, but i think a poll like this could be helpful to see what users want going forward

 

[Karim D. Ghantous]

Well, I'm thinking about the thread which shows how to increase DR by using a 360 degree shutter with a higher framerate. What about making that a supported feature, right in the camera, with audio recording enabled? It's simple and has nothing to do with new tech, but makes intelligent use of what's already there.

Also, how about a 4:3 sensor which fits into the image circle of VistaVision? This would be useful for IMAX productions. Again, it's not new technology, but just a new way to look at what exists.

 

[Christoffer Glans]

Well, I'm thinking about the thread which shows how to increase DR by using a 360 degree shutter with a higher framerate. What about making that a supported feature, right in the camera, with audio recording enabled? It's simple and has nothing to do with new tech, but makes intelligent use of what's already there.

Also, how about a 4:3 sensor which fits into the image circle of VistaVision? This would be useful for IMAX productions. Again, it's not new technology, but just a new way to look at what exists.

Are you thinking about my thread? haha :smiley:
Problem with that is that higher frame rates with HDR exposure makes 360 shutter impossible since the shutter is based on essentially a 100 fps setting; 50 fps double exposure. However, I'm not sure if it would be possible to do a kind of special exposure capture where the X-frame of HDRx would be captured not in between two frames but rather at the same time. If that kind of exposure, a kind of double processing exposure were possible with the next gen sensor that would erase the problems with a native higher DR.

 

[Elsie N]

Are you thinking about my thread? haha :smiley:
Problem with that is that higher frame rates with HDR exposure makes 360 shutter impossible since the shutter is based on essentially a 100 fps setting; 50 fps double exposure. However, I'm not sure if it would be possible to do a kind of special exposure capture where the X-frame of HDRx would be captured not in between two frames but rather at the same time. If that kind of exposure, a kind of double processing exposure were possible with the next gen sensor that would erase the problems with a native higher DR.

Sort of like the X frame exposing for STHL and the A frame exposed for LLO? Choose one or combine the two. Great idea.

 

[Björn Benckert]

Are you thinking about my thread? haha :smiley:
Problem with that is that higher frame rates with HDR exposure makes 360 shutter impossible since the shutter is based on essentially a 100 fps setting; 50 fps double exposure. However, I'm not sure if it would be possible to do a kind of special exposure capture where the X-frame of HDRx would be captured not in between two frames but rather at the same time. If that kind of exposure, a kind of double processing exposure were possible with the next gen sensor that would erase the problems with a native higher DR.

X frame is not in between two frames. its a part readout of the frame so if you got 1 stop HDRx the photo sites are dumping one set of data after half the exposure time and then A frame is the result of the x frame + the rest of the capture for that frame. i.e X is captured within A.

 

[Karim D. Ghantous]

Are you thinking about my thread? haha :smiley:
Problem with that is that higher frame rates with HDR exposure makes 360 shutter impossible since the shutter is based on essentially a 100 fps setting; 50 fps double exposure. However, I'm not sure if it would be possible to do a kind of special exposure capture where the X-frame of HDRx would be captured not in between two frames but rather at the same time. If that kind of exposure, a kind of double processing exposure were possible with the next gen sensor that would erase the problems with a native higher DR.

Yes. But, I wasn't referring to HDRx. Shooting at higher frame rates effectively allows you to increase DR because shadows are cleaner and exposure can be reduced to protect highlights.

Anyway, as Björn says, HDRx is not between frames after all, although I don't think that it is supported at high fps settings.

 

[Björn Benckert]

Yes. But, I wasn't referring to HDRx. Shooting at higher frame rates effectively allows you to increase DR because shadows are cleaner and exposure can be reduced to protect highlights.

Anyway, as Björn says, HDRx is not between frames after all, although I don't think that it is supported at high fps settings.

but you mean shot higher frame rates and compound frames to get less noise. right? as framerate alone does not increase DR. For example shooting phantom 4k at 500fps 360 shutter and then compound / mix or add 10 continues frames together and then skip the next 10 frames and then make frame 2 ot of the next ten frames after that etc. Would give a super clean 25p 180 deg shutter from the phantom but ofcourse you also loose light by a factor of 1/10 but the result image is cleaner than a single frame at 1/500.

 

[Christoffer Glans]

X frame is not in between two frames. its a part readout of the frame so if you got 1 stop HDRx the photo sites are dumping one set of data after half the exposure time and then A frame is the result of the x frame + the rest of the capture for that frame. i.e X is captured within A.

Doesn't seem to be like that when looking at the frames. Maybe that's the way the sensor works with HDRx, but the clips have ghosting in between each A-track, essentially it looks like a faster shutter capture (X) image between normal shutter capture (A) image. This is why shooting at 50 fps removes that ghosting, since A and X seems closer to each other in time when converted down to 25 fps.

But I haven't tried shooting HDRx with a base shutter that's faster, so I'm not sure that the two tracks would allign perfectly if both tracks were sharp by fast shutter. If not, then the X-capture happens at different times (in between A frames), if alligned they happen at the same time.

Yes. But, I wasn't referring to HDRx. Shooting at higher frame rates effectively allows you to increase DR because shadows are cleaner and exposure can be reduced to protect highlights.

Anyway, as Björn says, HDRx is not between frames after all, although I don't think that it is supported at high fps settings.

It doesn't increase DR for the acquisition but you can use the captured DR much better in post if the shadows are cleaner. But shooting HDRx 50 fps removes the artifacts since X and A tracks are closer to each other in time when converted down to 25 fps (for 24 fps it's 48 fps HDRx).

 

[Karim D. Ghantous]

but you mean shot higher frame rates and compound frames to get less noise. right?

Yes, that's what I meant.

But shooting HDRx 50 fps removes the artifacts since X and A tracks are closer to each other in time when converted down to 25 fps (for 24 fps it's 48 fps HDRx).

Interesting. So higher acquisition frame rates really have so many benefits in so many ways.

 

[Björn Benckert]

Doesn't seem to be like that when looking at the frames. Maybe that's the way the sensor works with HDRx, but the clips have ghosting in between each A-track, essentially it looks like a faster shutter capture (X) image between normal shutter capture (A) image. This is why shooting at 50 fps removes that ghosting, since A and X seems closer to each other in time when converted down to 25 fps.

But I haven't tried shooting HDRx with a base shutter that's faster, so I'm not sure that the two tracks would allign perfectly if both tracks were sharp by fast shutter. If not, then the X-capture happens at different times (in between A frames), if alligned they happen at the same time.




It doesn't increase DR for the acquisition but you can use the captured DR much better in post if the shadows are cleaner. But shooting HDRx 50 fps removes the artifacts since X and A tracks are closer to each other in time when converted down to 25 fps (for 24 fps it's 48 fps HDRx).

100% that 1 stop HDRx is the first 50% of the foto bucket fill of the A frame. So 1 stop HDRx X track is the same as lowering your shutter angle to half. Then the A track is the result of the photo buckets being filled for the remain of the exposure time.

Shooting HDRx and merge in between the two tracks result in a Glitch in the motion blur. Basically one layer has a different motion blur / shorter motion blur than the other. Shooting 360 shutter on the other hand makes it possible to perfectly blend two or more frames together. So shooting 100fps and then make 50/50 blend between frame one and two and then leave out 3 and 4 and then blend 5 and 6... and so on gives you a perfect 25fps 180deg shutter where the frames are built up by half the exposure and dual readout. Strange enough in a lot of situations that would actually trump doing one readout with double the light and half the compression. simply due to the fact that the two readout / frames mixed with each other get a "noise mix" that basically cut the moving noise in half. Also this recording procedure works like an ND 3 for the sensor. So it´s a way to prevent over exposure at a certain stop without having filters in the mix.

Now as the camera does not do 100 fps at full frame size etc we could debate that this is not to much gain and the use would be quite narrow. But as data and frame rates goes up the frame summing and averaging functions of cameras will be huge. Simply put when we can record enough frames per second and the cameras have the built in frame averaging functions, then the ND filters will be far less in demand. Instead you can just have a slider that devided down the exposure time without loosing the motion blur.

 

[Elsie N]

100% that 1 stop HDRx is the first 50% of the foto bucket fill of the A frame. So I stop HDRx X track is the same as lowering your shutter angle to half. Then the A track is the result of the photo buckets being filled for the remain of the exposure time.

Shooting HDRx and merge in between the two tracks result in a Glitch in the motion blur. Basically one layer has a different motion blur / shorter motion blur than the other. Shooting 360 shutter on the other hand makes it possible to perfectly blend two or more frames together. So shooting 100fps and then make 50/50 blend between frame one and two and then leave out 3 and 4 and then blend 5 and 6... and so on gives you a perfect 25fps 180deg shutter where the frames are built up by half the exposure and dual readout. Strange enough in a lot of situations that would actually trump doing one readout with double the light and half the compression. simply due to the fact that the two readout / frames mixed with each other get a "noise mix" that basically cut the moving noise in half. Also this recording procedure works like an ND 3 for the sensor. So it´s a way to prevent over exposure at a certain stop without having filters in the mix.

Now as the camera does not do 100 fps at full frame size etc we could debate that this is not to much gain and the use would be quite narrow. But as data and frame rates goes up the frame summing and averaging functions of cameras will be huge. Simply put when we can record enough frames per second and the cameras have the built in frame averaging functions, then the ND filters will be far less in demand. Instead you can just have a slider that devided down the exposure time without loosing the motion blur.

For me anyway, this clears up a lot about HDRx.

Thanks Björn.