Large chip 3-chip designs?

[Graeme Nattress]

That's a good paper from Dalsa.

Graeme

 

[Dean Gough]

Amazing I sign up to Reduser.net and first thread I read is about an item I've been experimenting with for a while now.

The semi transparent mirror you are talking about is a pellicle (see attached)

View attachment 410

View attachment 411

View attachment 412

Pellicles are very thin nitrocellulose membranes bonded to lapped aluminum frames. Ghost images are eliminated by the thinness of the membrane, as the second surface reflection superimposes on the first surface reflection. The uncoated pellicle reflects 8% and transmits 92% throughout the visible and near infrared regions. They are approx 2 mcrons thick.

The transmittance/reflection ratio of the pellicle determines how much light goes to the second sensor.

Experiments are being conducted with a pair of 2/3" Bayer CCD's, output of cameras is 8 bit RGB. The exact set up is still going through changes.

I have also explored briefly the other item of discussion, a rotating mirror, not to increase dynamic range but to increase frame rate. Imagine a 45 deg rotating mirror shutter as in a Arri 435 (or similar). The mirror not only provides the shutter function but in its closed cycle reflects the image to the optical system for the eyepiece, replace the eyepiece assembly with a sensor and you now can have two sensors capturing images 180deg apart or twice the frame rate.

All we have to do now is deal with the flood of data

Talking about 3-chip design.
What about 2-chip design, to increase dynamic range.
Can we have semi transparent mirror in front of the first sensor and second sensor next to it, one responding to lower light levels and the second responding to the higher light levels?

Alternatively, is there a way to read only one CMOS sensor, first fast, for the intense light levels and then slow for the low light levels?

I was thinking to use 120fps in 4 intervals.
We use first ¾ to read low light levels and ¼ to read high levels.
Then to combine it as a one 30fps frame with extra two or 4 stops dynamic range?

 

[Andrew M.]

Dean, this 8% reflection or bit more will be more then enough to capture extra 4 bits of highlights. As to the second sensor to get higher frame rate, I think the CMOS is fast enough to go even 800fps, it is the data processing that is the bottleneck here.
So two chips or one, you have to get this data out somehow and process it.
The best is to advance CMOS technology so it will get 16 bit dynamic range.
For now if we want to beat the film we have to use two sensors.

 

[baro]

Hi Dean. That would be a great way to increase dynamic range. For a seamless integration of the two images from the brighter and the dimmer image, theoretically you would need the two transmission graph to be the same shape ( that is true when your goal is to increase the resolution too). As you experimented with that thing, you may have some experience if the small differences in the transmission shapes has a bad effect in practice on the resulting picture. Or can you apply an optical filter in front of some of the imaging devices that corrects the transmission function so that the two shapes are identical?

 

[Dean Gough]

The project was originally started because of my interest in HDR imaging. Most Digital Cinema cameras (notice I didn't say video cameras) seem to have around 9 - 11 stops of dynamic range and bit depths ranging from 10 to 16.

I don't see any reason this could not be extended to 14 - 16 stops (film is considered 11 - 12). From a purely visual stand-point you're not going to see the extra dynamic range on the screen because of the eyes inability to distinquish large variation in brightness but from a Post perspective the extra latitude may be useful. Same as an original camera negative may capture 11 stops of information but only a small part is ever seen on the screen.

CMOS is definitely fast enough, Vision Researchs Phantom HD camera does 1000fps at HD resolution. Useful but not for your typical Digital Cinema requirements. Thats actually what I like about RED, it has a great practical feature list without going overboard (read as adding unnecessary expense)

Anyway probably getting way of the subject of 3 chip cameras here. My 2 cents worth is that single chip cameras are here to stay and have more than enough quality capability for Digital Cinema use. There are so many other factors which effect the initial captured and displayed image. The digital still market which have used large single CCD and CMOS sensors for years, there are many instances 6 -8 Mpixel cameras have out performed 10 - 12 Mpixel units.

I've seen some pretty awful films shot on 35mm as well. Content is King.

Dean, this 8% reflection or bit more will be more then enough to capture extra 4 bits of highlights. As to the second sensor to get higher frame rate, I think the CMOS is fast enough to go even 800fps, it is the data processing that is the bottleneck here.
So two chips or one, you have to get this data out somehow and process it.
The best is to advance CMOS technology so it will get 16 bit dynamic range.
For now if we want to beat the film we have to use two sensors.

 

[Dean Gough]

Increasing the dynamic range was the primary goal.

One of the reasons I chose the 92/8 pellicle was because the graphs are very flat in there response. I am only experimenting at low resolution (640x480x8bit RGB) as proof of concept.

Now if I had access to a couple Mysterium sensors I could take it to the next level

Hi Dean. That would be a great way to increase dynamic range. For a seamless integration of the two images from the brighter and the dimmer image, theoretically you would need the two transmission graph to be the same shape ( that is true when your goal is to increase the resolution too). As you experimented with that thing, you may have some experience if the small differences in the transmission shapes has a bad effect in practice on the resulting picture. Or can you apply an optical filter in front of some of the imaging devices that corrects the transmission function so that the two shapes are identical?