The 35MM full frame 25MP global shutter CMOS sensor

Postmaster

Postmaster

Dark Side of the Camera
The 35MM full frame 25MP global shutter CMOS sensor

Glimpse of the future - Business Wire and EE Times report, via the excellent Image Sensors World blog that industrial image sensor manufacturer Cypress have just introduced a 35mm global shutter CMOS sensor for video.

With a global shutter, the VITA 25k delivers 53 frames per second at the full 25 megapixel resolution. No scaling. That's video in 5120x5120 resolution without compression, scaling, aliasing or moire.

http://bit.ly/bkJS4O

Frank
 
Last edited:
:cool:
I can only hope they aren't the only ones who've been implementing global shutter on CMOS.
It's important to remember that according to how they've implemented it, if one were to choose to deactivate global shutter and implement rolling shutter, then there's actually an improvement in image quality for noise reduction and dynamic range.
So, there are still some sacrifices to be made going in either direction.
I would still absolutely love to have at least 4k res raw with no rolling shutter. No matter what, I'd take a subtle hit in image quality that I can probably fix in post to ensure I don't ever get jello-vision from the rolling shutter.
 
Last edited:
Nice development.

The article says the sensor is 35mm full frame size CMOS, but the form factor is square (looks like the bottom 25% of the chip is circuitry, not photosensors). Someone should tell them that full frame 35mm is 36 x 24mm. This sensor was probably not targeted to TV or cinema, but that can't be far behind.
 
The Drama is open to anything that connects to a computer, but the storage size and cost goes up, we are talking uncompressed. 1080p is a practical limit for uncompressed today IMHO. My focus is on low budget filmmaking camera one can own, not something that rents for $5,000 a day due to storage restrictions. Resolution is a limitation you have to take in uncompressed imaging.

Most sensors available in the open market are implemented within 1 year or so if the performance is comparable to the older sensors. This sensor should be available in a few months. It could take until January 2012 to see it implemented. It could be as good as rolling shutter in noise, but I suspect it will be inferior, especially in sensitivity. Low light is critical because large aperture DOF increases lighting cost, especially with a sensor that is so large. If it will need 10x more light that 35mm film, it's not an option for low budget filmmaking. It could easily need 10x or 20x more light than ASA500 film. We will see when detailed specs are out.

Oh wait. They have some specs on that blog link.

The pixel is a little small for SLR lenses, but no real problem there for 1080p resampling.

The AD is 10bit. 10bit AD is not enough. The norm is 12bits or 14bits today. It will have lots of alias at 0dB gain, after typical white balance and that's with the most conservative gamma settings which is ITU709 with more than 4x (2 bits of gain) at the shadows. Think 6bits of resolution with a good exposure. You can't have highlight protection with a 10bit AD. The shadows will be very aliased. Under digital gain it will be terrible. The AD is usually engineered to reflect the quality of the system, so that's not a good sign.

A dark noise of 2.19DN is very bad on a 10bit chip (1024 DN). Ok for industrial use, not for quality video. The dynamic range will be less than 8stops in linear signal. About 3 stops inferior to a good sensor. Notice that switching from rolling shutter to global shutter produces 3x higher noise (2.19 vs 0.625DN). There is the global shutter limitation I have discussed before.

The dark signal is 8.33DN. That is 1024/8.33=certainly not enough.

It has a multislope optical dynamic range of 90dB. Yeah, right:) We have heard that one before. That's 15 stops. Even if this works well, with the 10bit of output it would be totally useless, the slope points would produce very unnatural colors in every single slope change. It could be useful for security or monitoring in industrial processes.

Give it two years.
 
Last edited:
It is 23mmx23mm. About twice as large as the Drama sensor (2x10.56mm=21.12mm). Super35 or something like that.
 
well its EOS-HD. its the only site where a baby is born and eos-hd calls it the coming of christ... and claims factual truth
 
Seems to me that the article draws a lot of conclusions that are loosely based in fact... Cool that there could be some cool sensors down the road, though. Personally, I am wondering if black silicon was ever found to be useful for video or photo.
 
Applications

The substance has since been found to be incredibly sensitive to light, leading to a range of exciting plans for commercialization, including night-vision and infra-red imaging systems. According to James Carey, co-founder of Harvard spin-off company SiOnyx, “We have seen a 100 to 500 times increase in sensitivity to light compared to conventional silicon detectors.”

The unusual optical characteristics, combined with the semconducting properties of black silicon make this material interesting for sensor applications. The potential applications include:http://en.wikipedia.org/wiki/Black_silicon#cite_note-3

 
There are lots of global shutter sensors around, they are just not good enough yet.

Cameras are already sensitive enough, some are so sensitive that they are problematic in daylight. More does not mean better. We need good daylight AND lowlight performance on the same sensor. An ISO10,000 sensor would be competely useless, you wouldn't be able to shoot in daylight and you wouldn't even be able to light in lowlight.

For 2/3" an ISO25 daylight and an ISO500 lowlight is ideal IMHO.
On super35 sensors the ideal would be ISO100 daylight and ISO2000 lowlight.

But since we can't have both, an ISO200 for 2/3" and an ISO800 super35 will be a good compromise. A little gain covers low light and some ND covers daylight.
 
Jonathan Briceño said:
Applications

The substance has since been found to be incredibly sensitive to light, leading to a range of exciting plans for commercialization, including night-vision and infra-red imaging systems. According to James Carey, co-founder of Harvard spin-off company SiOnyx, “We have seen a 100 to 500 times increase in sensitivity to light compared to conventional silicon detectors.”

The unusual optical characteristics, combined with the semconducting properties of black silicon make this material interesting for sensor applications. The potential applications include:

  • image sensors with increased sensitivity
  • Thermal imagin cameras
  • Photodetector with high efficiency through increased absorption.
  • Mechanical contacts and interfaces
  • Terahertz applications.
Click to expand...

Don't mind me, just lighting my scene with a godD!@# cell phone:laugh:
 
Otis Grapsas said:
It is 23mmx23mm. About twice as large as the Drama sensor (2x10.56mm=21.12mm). Super35 or something like that.
Click to expand...

I hope it stays below S35mm (or 30mm diagonal) .... because FF35 is lame in my book. I don't need the subsequent shallower DoF and also, all the amazing lenses I love Master Primes, Cooke S4's, Angineuix zooms, and etc would be useless because they don't cover the area. Of course you have Zeiss Compacts, but they are certainly not to spec with the aforementioned lenses. They compare to the quality of 1980's cinema lenses, as far as performance.

Most lenses only cover 30mm diagonal or 32mm diagonal. I believe FF35 is around 36mm.

I really don't believe we need sensors greater than S35... unless you are shooting large resolution, large format, such as specialty 65mm or 70mm film equiv.

But then again, if we are in a pixels arm race, you have to fit them in some sort of frame, so bigger frames would allow for the pixels to gather more light. Perhaps then we can just stop down and get that DoF back.
 
You must log in or register to reply here.
Back
Top