Canon Releases White Paper for Cinema EOS C300 Mark II

[Canon Rumors Guy]

The Canon EOS C300 Mark II is nearly release after being <a href="http://www.canonrumors.com/2015/04/announcement-canon-eos-c300-mark-ii/" target="_blank">announced last April</a> at NAB 2015. The biggest development in the C300 Mark II that DSLR users are hoping for, is the claimed 15 stops of dynamic range in the new camera. Currently Canon DSLRs are running at around 12 stops, which is a few stops below the competition from Sony/Nikon.</p>
<blockquote><p>The original EOS C300 digital cine camcorder employed an innovative new 4K UHD Super 35mm CMOS image sensor developed by Canon. The readout system dissected that 4K sampling into four parallel HD components. The summation of the two green components reduced aliasing that, in turn, supported a higher Luma MTF. The new C300 Mark II preserves these basic strategies but within a totally new 4K CMOS image sensor design. The new camera is intended to significantly extend the overall image performance of HD beyond that of the C300 while further supporting the alternative 2K cinema format. A 15-stop dynamic range is provided by a new photodiode design that simultaneously lowers the noise floor while elevating the saturation level – offering excellent HDR functionality.</p></blockquote>
<p>The Cinema EOS C300 Mark II white paper will give you all the technical information you could possibly want about the new camera.</p>
<p><strong><a href="http://www.usa.canon.com/CUSA/assets/app/pdf/camera/brochures/WhitePaper-imageperformanceenhancements-eosc300markii.pdf" target="_blank">Download the White Paper</a> | Preorder: </strong><strong><a href="http://www.bhphotovideo.com/c/product/1134579-REG/canon_0635c002_eos_c300_mark_ii.html/BI/2466/KBID/3296" target="_blank">Canon EOS C300 Mark II EF</a>| <a href="http://www.bhphotovideo.com/c/product/1134580-REG/canon_0635c009_c300_markii_cinema_eos.html/BI/2466/KBID/3296" target="_blank">Canon EOS C300 Mark II PL</a></strong></p>

 

[sanj]

Brilliant news!

 

[BeenThere]

For the sensor gurus among you, how much of this sensor technology could translate to better still images? The better micro lenses sounds good as well as high s/n ratio analog amps. Anything else?

 

[privatebydesign]

Very interesting. The time and money is being spent on the log calculations and hardware to map 15 stops of DR into 14-8 bits in much more realistic ways, this can only be a good thing for us photographers.

I'd expect shadow detail in jpeg stills (and RAW files) to be much better in future stills cameras once the stills camera sensors inherit those improved diodes and on sensor noise reduction for a lower noise floor and higher saturation levels.

 

[mb66energy]

For the sensor gurus among you, how much of this sensor technology could translate to better still images? The better micro lenses sounds good as well as high s/n ratio analog amps. Anything else?

Yes! (beware: no sensor guru)

(1) "noise floor is lower": electronic noise added during readout "smears" the real values of a pixels electron content - better readout electronics reduce this "noise floor"

(2) "elevated the saturation level of the charge well": a photosite can be charged up to a maximum charge which represents the highest level of light it can detect. If more light enters the photosite, you cannot see changes anymore (highlights are clipped in that case). Canon has more charge capacity in the photosites well or has perhaps found a non-linear approach of saturation ...

 

[privatebydesign]

Very interesting. The time and money is being spent on the log calculations and hardware to map 15 stops of DR into 14-8 bits in much more realistic ways, this can only be a good thing for us photographers.

We don't know yet what their file format is but if it is a 14bit format then either the 15 stops of DR is a lie or Canon are using lossy compression.

That said, if this is true then I'd love to see Canon use this "new sensor" in a DSLR - or rather to see a DSLR with this new photodiode design in its sensor. Maybe the 5DIV?

As you well know you can't fit more than 14 stops of linear capture information into a 14 bit file without some kind of remapping of the information, nobody can. But if they are remapping things we can't see with the human eye to allow space for things we can see with the human eye then I'd take lossy 'RAW'. All lossy 'RAW' is not equal.

But the white paper shows you the curve they are applying to the linear capture information, it seems like a decent sort of human vision type of curve to me. Time for us all to read about gamma curves and how the eye actually sees different tonalities at different luminance levels.

 

[traveller]

I hate to be a killjoy and I'm by no means a sensor expert, but the diagram in Canon's C300 mkii White paper still seems to show the A/D conversation happening off chip. I thought that the fact that Exmor does A/D conversion on the chip prior to signal amplification was one of the main reasons that their sensors can maintain such a low noise floor? Is the 15 stop DR of the C300 mkii more to do with oversampling and processing of a video signal (dual ISO style) than pure sensor performance?

 

[privatebydesign]

I hate to be a killjoy and I'm by no means a sensor expert, but the diagram in Canon's C300 mkii White paper still seems to show the A/D conversation happening off chip. I thought that the fact that Exmor does A/D conversion on the chip prior to signal amplification was one of the main reasons that their sensors can maintain such a low noise floor? Is the 15 stop DR of the C300 mkii more to do with oversampling and processing of a video signal (dual ISO style) than pure sensor performance?

But we don't know what is meant by "on chip noise cancellation technology", or how effective it is.

 

[privatebydesign]

...
Is the 15 stop DR of the C300 mkii more to do with oversampling and processing of a video signal (dual ISO style) than pure sensor performance?

No, they are saying they have a higher full well capacity and a lower noise floor, so the sensor is capable of recording 15 stops of information natively. To fit that 15 bits of information into a 14 bit file, or the 10 bit and 8 bit options also available on the C300 MkII, they are applying a gamma curve to the readout signal information.

Now for us photographers the question becomes if this is applied to a stills camera is the curve smooth enough in the right places to increase the IQ over true linear RAW or will they go with a true 16bit file which won't be as much of an issue at still capture rates. To be sure this is quite a different approach to the one Sony take with their lossy RAW files.

 

[BeenThere]

I hate to be a killjoy and I'm by no means a sensor expert, but the diagram in Canon's C300 mkii White paper still seems to show the A/D conversation happening off chip. I thought that the fact that Exmor does A/D conversion on the chip prior to signal amplification was one of the main reasons that their sensors can maintain such a low noise floor? Is the 15 stop DR of the C300 mkii more to do with oversampling and processing of a video signal (dual ISO style) than pure sensor performance?

The way I read it looking at the block diagram is that there is an analog amp/buffer on the sensor chip and A/D is off chip; then some digital gain off chip.

 

[rfdesigner]

I hate to be a killjoy and I'm by no means a sensor expert, but the diagram in Canon's C300 mkii White paper still seems to show the A/D conversation happening off chip. I thought that the fact that Exmor does A/D conversion on the chip prior to signal amplification was one of the main reasons that their sensors can maintain such a low noise floor? Is the 15 stop DR of the C300 mkii more to do with oversampling and processing of a video signal (dual ISO style) than pure sensor performance?

I design radios.. with high speed ADCs sampling baseband analogue radio waveforms. (I also built a home designed CCD camera many years ago)

It's a straight choice.. ADC in the radio. spewing EMI muck throughout the radio, or ADC in the digital domain meaning you have to feed your delicate analogue baseband signals into the filthy digital domain. ideally the ADC bridges the gap.. but muck still crawls all over the place.

There is no single "right" solution.

Don't let anyone tell you an off sensor ADC is bad and can't acheive a certain performance, it's all about the very fine detail of how it's done.

 

[RGF]

looking forward to seeing this in dSLR

 

[9VIII]

I really wish they would just let us eat the file size and use a 16bit file.

Theoretically 10 bit displays are going to become mainstream in the coming years (to match the UHD Blu-ray spec), which could make the extra detail in gradients more important.
Or maybe I'm just a picky technophile who dislikes compression.

 

[privatebydesign]

I really wish they would just let us eat the file size and use a 16bit file.

Theoretically 10 bit displays are going to become mainstream in the coming years (to match the UHD Blu-ray spec), which could make the extra detail in gradients more important.
Or maybe I'm just a picky technophile who dislikes compression.

I'd happily take the file size, especially if the MP are kept at a reasonable mid 20's. Sure the 5DS/R look like very nice cameras, but I'd definitely take the 16bit 15 stop over the 14bit 50MP any day.

 

[tiger82]

Wouldn't the sensor for this 4K camera be 8 megapixels? Would anyone want that in a DSLR?

 

[traveller]

Wouldn't the sensor for this 4K camera be 8 megapixels? Would anyone want that in a DSLR?

The 22mp sensor in my 5d Mk3 isn't far off 8mp if you crop to Super 35 format.

 

[privatebydesign]

Wouldn't the sensor for this 4K camera be 8 megapixels? Would anyone want that in a DSLR?

Yes it is an 8.9 MP sensor, but that isn't the point, nobody wants that exact sensor for their stills camera, we are interested in if the tech they are showcasing can be transferred to a stills camera sensor.

Like I said, I'd happily take a true 16 bit RAW file of a 25MP sensor that has 15 stops of DR, and I'd probably be happy to take the current 14bit file in a remapped 'RAW' file if it gave me a big improvement in shadow detail, but I'd rather the former.

 

[privatebydesign]

Wouldn't the sensor for this 4K camera be 8 megapixels? Would anyone want that in a DSLR?

The 22mp sensor in my 5d Mk3 isn't far off 8mp if you crop to Super 35 format.

Exactly, the 5D MkIII has a pixel pitch of 6.25µ, the C300 MkII has a pixel pitch of 6.4µm

 

[mb66energy]

I really wish they would just let us eat the file size and use a 16bit file.

Theoretically 10 bit displays are going to become mainstream in the coming years (to match the UHD Blu-ray spec), which could make the extra detail in gradients more important.
Or maybe I'm just a picky technophile who dislikes compression.

I really dislike lossy compression too: I am not a picky technophile but love nature.
If I reproduce an image of nature (etc.) I like it to be as close as possible.
Perhaps it has nothing to do with morbus "picky technophilie"

That's what I observed comparing 5D classic-files with EOS M files (newest cam/sensor I own):
The 5D reproduces fine detail much closer to what I see than EOS M - I don't know exactly why,
but I think it is the precision of sensor-amplifier-ADC + the large photosite advantage.

And that leads me to my conclusion about upgrading to a current 5D: 25-30 MPix in a 5D iv
with the C300 ii-per-pixel quality we would expect from what we read in the white paper would
lead to a very versatile high end DSLR. I would prefer it compared to the 5Ds(r) variants ...
maybe the delay of the 5D iv release is caused by sensor optimization?

 

[Ebrahim Saadawi]

I'd be surprised if I knew the EOS cinema sensor development/manufacturing/designing is separate completely from the stills sensors drsign/development teams. I would bet big money that they're probably the same team/company using their expertise and technology in developing sensors for both the large chip cinema line and DSLRs, which makes me pretty confident that any breakthroughs in sensor performance in the cinema line could very likely be utilized in DSLR sensors.

But just another point, remember that this is a 20 thousand dollars camera. So the sensor technology and performance and accuracy will probably be ahead from the sensor technology put on a 3K DSLR. Just a though. Even the C100 MKII, a 6000$ cinema camera doesn't get these sensor technology upgrades, so more likely to see 15 stops of DR and noise levels of a C300II in a 1DxII rather than a 5D MK IV?

I have no idea, just thinking out and loud.

But again, the main reason these c300II sensors are so expensive is because they shoot 30 frames per second for years, so perhaps a sensor with the same technology and image quality on frame-to-frame basis would be a viable 5D solution, just without the heavy load a cinema/video camera puts on a sensor/processor. A scaled up C300 II sensor to FF would yield about 21mp? (it's 9.3 megapixels APS-C size), without the 30p, 60p, 120p capability, the cinema ridiculous colour gamut covering all film stocks every produced, the intelligent and innovative 2K/HD downsample, could be a cheap one to imploy in a 5D/1D.

Just with normal decent video downsample to 4K like all cameras do now instead of the C300II magical complicated processes (Takes the 14bit data, applies gain, then WB to the 14bit original data, separates the RGB layers into 2 green, 1 Blue, 1 Red layers, upscales each layer to full 4K, applies complex false colour and anti-aliasing algorithms, lens pixel abberation correction in 4K, remaps 14bits into a Canon LOG gamma, then combines the 4 up-sampled 4K layers into a super-sampled 1080p image, then after all of that store it in a 12bit 4:4:4 RGB codec with 405 megabits second, I mean that nuts to do in real time, nobody takes such effort to just produce 1080p but now I know why the C100/300 HD images look better than any other 1080p image out there that are just reading a 1920x1080p pixels and debayering. I always saw it clearly being MUCH superior to all HD cameras but didn't understand until now.

I eagerly await the day that 1080p quality comes to sub 3000$ canons. The closest I've got now is the generation one C100 at 2900$ but the lack of any slowmotion and useless EVF put me off entirely, the C100 MKII is perfect 1080p and ergonomics wise, but out of my range.

Well, maybe the 5D MKIV 4K image will answer my prayers of a consumer Canon LOG cinema line looking image with canon colour science and joy to use and grade,