C100: What exactly does "4K sensor" mean?

Omar N

Active member
Hello community,

I've poked around online but haven't found any conclusive answers, so I just need to clarify one thing. I understand the Canon C100 and C300 have a 4K sensor, but only output 1080p. What I am not sure about is what this means exactly. Is this the same as shooting in native 4K like you would with a RED and then downrezing to HD for final output? If so, then am I to assume there should be no difference resolution wise between RED footage and C100 footage when viewed on an HD video monitor?

I understand the difference between the C100/C300's 4K sensor and the Canon 5D Mark 3's 5K sensor is simply the bayer pattern. The 5D line skips to get up to HD, whereas the C100/300 downres correct?

All help greatly appreciated. Thank you.

~Omar
 
Hi Omar,

In the case of the C100 and C300, the sensor is 4K but these cameras can only record at a maximum resolution of 1920 x 1080. They're sampling HD (1920x1080) off a 4K (4096x2160) sensor which also explains why these cameras perform so well in low light. But they're not actually recording at 4K resolution. Whereas RED cameras actually do record at 4K (Red One, Scarlet) and 5K (Epic) and do so in raw format. Canon 1DC records in 4K but in compressed Motion Jpeg.
 
Last edited:
I'm preparing my review of the C100 and I'll be addressing this at some lenght. But in short:

The sensor in the Cinema EOS line is new. It's not a sensor from the DSLR range. It has 4216x2340 photosites and uses 3840x2160 actively for the image (quad HD). You can think of a photosite as a pixel on screen. This is the 4k sensor. It has 4k photosites.

A RED Scarlet shoots 4096x2160 (as an example). This is slightly wider than quad HD 4k. It is a similar resolution.

Both Canon and RED then overlays these photosites with a color filter array - the familiar Bayer pattern with alternating green-blue and green red color filters.

In the Canon sensor, due to it being perfect 4x HD resolution, generates the following: two 1920x1080 green lattices (the two green lattices are separated with half a pixel which help with aliasing.), one 1920x1080 red lattice and one 1920x1080 blue lattice. There are in fact two times as many green "pixels" as red or blue. This is standard Bayer pattern.

Now to the significant part: Canon isn't striving for any type of 4k output resolution here. They simly overlay each of the lattices during image processing so that two green, one red and one blue photosite generates one pixel on screen. There is no de-bayering, or luminance/color averaging taking place here. That is how 4k photosites "only" end up as 1920x1080 pixels. It's beaufiul, if HD is your target. These 1920x1080 pixels are full color. They each come from 2 green, one red and one blue photosite.

In basically every other camera, including RED's, they output all the 4k photosites image data so that you end up with 4k pixels. The thing is though, those pixles are still arranged in the Bayer pattern, so it's alternating lines of green/blue and green/red. Not 4k full color pixels. You can NOT compare these 4k pixels to the 1080p pixles from the Canon output.

Still, the advantage of the Bayer output is that all 4k pixles has unique luminance information. Maybe you can say that it is 4k luminance and 1080p color information? Now all of these pixles need to be de-bayerd so that you end up with 4k full color, color averaged pixels. With good color technology and filtering algorithms you get an excellent result.

Still... the methods differ between our Cinema EOS sensors and other normal sensors. Canon wanted to create a perfect 1080p sensor and they are coming close here. The C100/300 really resolves a lot of detail and there's no real artifacting (color fringing on C300 is fixed I believe).
 
I think RED says that their 4K is actually 3.2K. Our RED users here can explain. I am also not sure the actual resolution in terms of TV lines. Andree is right about Canon's C300 sensor. The signal processing due to the 2 greens, one red and one blue is equivalent to a three chip RGB system. Have a look here for the diagram and explanation: http://www.indiawilds.com/diary/review-canon-c300-for-wildlife-filming/

After looking at the details in the C300 files, don't feel like looking at the DSLR files.
 
Wow thanks for the awesome responses guys! This has all been very helpful. Based on the information above, I understand that if HD is the intended delivery format, the C300 may surpass RED in terms of its color reproduction, although having the benefit of 4K and RAW could potentially make up for it depending on how the footage is treated in post.
 
3.2K is resolving power. 5K debayers to nice 4K material.


I think RED says that their 4K is actually 3.2K. Our RED users here can explain. I am also not sure the actual resolution in terms of TV lines. Andree is right about Canon's C300 sensor. The signal processing due to the 2 greens, one red and one blue is equivalent to a three chip RGB system. Have a look here for the diagram and explanation: http://www.indiawilds.com/diary/review-canon-c300-for-wildlife-filming/

After looking at the details in the C300 files, don't feel like looking at the DSLR files.
 
This has me interested in the Panasonic HPX600. It has a single 2/3" CMOS chip. It has to be Bayer filtered, so if we assume it's a 1920x1080 chip, will it have less than that after de-Bayering?
I have no idea if it is a 1920x1080 chip, there's no mention of it in their literature.
 
The signal processing due to the 2 greens, one red and one blue is equivalent to a three chip RGB system.

Correct, but to be clear, that RGB system would have to have 3 x S35 full raster 1920x1080 chips. For example the Panasonic HPX3700 that is listed on the BHPhoto website today for 60K, has 3 x 2/3" 1920x1080 CCDS (still a beautiful image IMO). Or the Sony F900 for 80K.
 
Back
Top