C300: C 300 MK II - White Paper (PDF)
- Thread starter DLD
- Start date
jm.gardiner
Well-known member
I was studying this today. Everything sounds great with the exception of the fps over 60p. I'm hopping the cropped frame rates don't suffer too much image quality loss. When I demoed the camera a few weeks ago sharpness didn't seem to suffer.
AndreeOnline
Veteran
When reading the white paper on the first pass, it might come across as if the 120fps mode is severely cropped when Larry talks about 1024x540 frames. Don't forget that at this level we are talking about discreet color components/diodes of the Bayer pattern. In 'pixels' these 'frames' would be 2048x1080. Just like in 4k, the Red, Green (x2) and Blue components are 2048x1080 each in a 4096x2160 pixel frame.
In all modes Canon is then up sampling the color components (i.e. filling the holes in the Bayer grids) so that 1024x540 diodes become 2048x1080 pixels. A great cure against aliasing.
Ebrahim Saadawi
Carbonite Member
From early reports/samples there doesn't seem to be loss in image quality, just a crop. A perfectly clear 1080p image with no crazy aliasing/moire (none) or softness as on cameras that do it with a full sensor binning/skipping (FS7/F5/F55).
I prefer a crop vs a lesser/different image anyday and I am glad they went that way instead of the other.
The only thing this camera is missing vs the competition is 4K 60p. I am not sure how important that mode is to people.
I prefer a crop vs a lesser/different image anyday and I am glad they went that way instead of the other.
The only thing this camera is missing vs the competition is 4K 60p. I am not sure how important that mode is to people.
Ebrahim Saadawi
Carbonite Member
Are the other subsequent white papers at the end of the report available?
AndreeOnline
Veteran
Reading this white paper, it quickly becomes clear that the C300 mkII is NOT Canon's 4k product. They are pretty proud of the C300 mkII for sure, but still they refer to its 4k capabilities as "modest".
To me, it seems clear that a C500 mkII is lurking in the shadows.
To me, it seems clear that a C500 mkII is lurking in the shadows.
jm.gardiner
Well-known member
AndreeOnline, in regards to that crop. That makes sense. When I was doing some basic testing with it, I really didn't notice any significant loss in IQ on the crop. I think just seeing those lower resolution numbers gave me pause.
C500mk2 is certainly around the corner, I imagine it'll be something like the Varicam35. I do see the C300mk2 getting along well with the Amira in the mean time, that FaceAF will be a blessing on gimbal shoots. I too would like to see that 4k 60p added at somepoint. Even a 4k widescreen 48fps.
A few other notable take aways:
- 67db Signal to Noise ratio. That out does the Sony F55 (a very clean 57s/n) by a pretty great margin.
- Impact resistant front plate to absorb lens stress.
- Canon LOG2 maxing out at 92IRE. I've been very confused by this as even in the demos it seems the camera clips around 90IRE and with this spelled out, I still don't understand how or why it clips at 92 rather than 109, but it does explain a lot of the demo videos with hight light clipping. If you're used to letting your hightlights reach all the way up to 95IRE on a waveform, but keeping it under 100 (as I've done on other cameras for high contrast scenes) you'll be dissappointed by the performance.
C500mk2 is certainly around the corner, I imagine it'll be something like the Varicam35. I do see the C300mk2 getting along well with the Amira in the mean time, that FaceAF will be a blessing on gimbal shoots. I too would like to see that 4k 60p added at somepoint. Even a 4k widescreen 48fps.
A few other notable take aways:
- 67db Signal to Noise ratio. That out does the Sony F55 (a very clean 57s/n) by a pretty great margin.
- Impact resistant front plate to absorb lens stress.
- Canon LOG2 maxing out at 92IRE. I've been very confused by this as even in the demos it seems the camera clips around 90IRE and with this spelled out, I still don't understand how or why it clips at 92 rather than 109, but it does explain a lot of the demo videos with hight light clipping. If you're used to letting your hightlights reach all the way up to 95IRE on a waveform, but keeping it under 100 (as I've done on other cameras for high contrast scenes) you'll be dissappointed by the performance.
AndreeOnline
Veteran
It's notable that the C300 mkII maintains 58dB even at ISO 25600.
According to the paper, new full well capacity is now at 1600% IRE compared to 800% for previous Cinema EOS cameras. And double amount in linear light equates to 1 stop more worth of highlights. The other two stops (Canon Log2 has 3 stops more than Canon Log) are then found in the shadows. The paper also specifically mentions this new capacity.
I guess Canon felt that with internal recording now being 10bit, or better, they could comfortably deliver a flatter Log file for more options in post. DaVinci has a great engine that keeps image information in 32bit space internally, even if a node should push the highlights way beyond 'clipping'. Lesser tools might benefit from some head room in the file when re-applying contrast or LUTs or whatever.
mraaron895
Well-known member
I think canon is taking one from the Alexa book. The high bitrate up-sampling process yields a much more pleasing to look at image. Yes, RED has HFR 4k and has all the bits a MG artist could want. But at the end of the day you can pick up this camera, slap a small internal battery on it which lasts over an hour, and shoot everything from Narrative, commercial, eng, docu, all without having to add or remove components. And if I want real HFR I would shoot with an actual high speed camera.
FrameFarmMedia
Veteran
Sony Slog3 is similar in that regard and maxes out at 93IRE. Its all part of matching the Cineon curve & future potential to hold 16stops in the same Gamma or so says Sony so I imagine the Clog2 reasoning would be along similar lines.
Matt Turner
Well-known member
Interesting to hear your thoughts on this:
From my understanding (early days) UHD/4K broadcast will be 50p-120p, all the trials except one are > 30p (Sweden @ 30p)
https://en.wikipedia.org/wiki/Ultra...ision#Field_trials_of_UHDTV_over_DTT_networks
Would this mean that the C300MKII will technically not be UHD/4K "broadcast ready"?
Since broadcast ready 1080p was a selling point of the C300, it might put people off from a C300MKII.
But on the other hand by the time UHD/4K broadcast is ready, everyone will have moved on to 8K or a C300MKIII will be available.
jm.gardiner
Well-known member
I wonder if that's refering to the "delivery pipe" and not the acquisition/captured frame rate. Like the way that US broadcast is 1080i/60, but you can capture in any frame rate you want. Delivery specs being different than what's shot on set.
It looks to be the max performance of the next generation of the over-the-air (OTA) broadcasts. Looking at the specs off that Wikipedia page, they're trying to get everything crammed into 50 Mbps using HEVC compression. I believe the 4K Blu-Ray will go twice as fast and, technically speaking, a download service like My Passport can offer anything they feel like. And should the gigabit Internet ever get here ...
MichaelFischer.at
Active member
Hi Everyone,
I read the papers when they came out, but could only reply now.
Canon says that the C300M2 can capture 15 stops of DR. Is it just me or how should this be possible, if they do a 14 bit A/D conversion?
Technically with 14 bits (linear) you can get 12 stops max, 10 if you want to have proper tonality in the shadows...
I read the papers when they came out, but could only reply now.
Canon says that the C300M2 can capture 15 stops of DR. Is it just me or how should this be possible, if they do a 14 bit A/D conversion?
Technically with 14 bits (linear) you can get 12 stops max, 10 if you want to have proper tonality in the shadows...
AndreeOnline
Veteran
14 bit A/D would give you a contrast of 16384:1 (2[SUP]14[/SUP])which corresponds to 14 stops in theory. And it would be easy to understand how 14 stops could be maintained even at lower 10 bit with a logarithmic curve such as Canon Log 2.
It's not quite as easy to see—as you state—how the DR can exceed that of the maximum internal precision. Yet, it has been confirmed several times that the stops are there. Also by generally unbiased Geoff Boyle at CML.
One solution might be along the lines of dual gain. Perhaps a technology that would allow you to read the diodes two times at different gain settings, buffer the linear light values, and then store a combined product logarithmically...
Ebrahim Saadawi
Carbonite Member
Yes perhaps the broadcast UHD standard is going to be 50/60p, but as you know non of today's cameras will be around when UHD broadcast begins. We're still 720p and recovering from the cost of that jump from 360/480p. I give it to 2020 where we will 'see' one or two examples in the globe that are doing a UHD broadcast (probably Tokyo olympics) then at least 5 years later for it to be used by broadcasters around the world. And at that time a C900XZP-FE will probably shooting 8K at 60p with 20stops of Raw DR.
The actual feature for 60p UHD is 2x slowmotion. It is a feature that some can use but most seem to be going to the >60p now for 90% of the time at 1080p resolution.
It's the only features the C300II lacks vs the world. Canon's approach was never to cram everything up in the spec sheet but they seem to have done it here for the first time even for their stills division.
It actually leaves no where for the C500 successor they're planning to go, other than jumping to 6K and 8K resolution and Ultra Slowmotion 240p-480p. What other features could it possibly have vs the C300II?
The actual feature for 60p UHD is 2x slowmotion. It is a feature that some can use but most seem to be going to the >60p now for 90% of the time at 1080p resolution.
It's the only features the C300II lacks vs the world. Canon's approach was never to cram everything up in the spec sheet but they seem to have done it here for the first time even for their stills division.
It actually leaves no where for the C500 successor they're planning to go, other than jumping to 6K and 8K resolution and Ultra Slowmotion 240p-480p. What other features could it possibly have vs the C300II?
cpreston
Veteran
Why is it important that the mathematical limit of the contrast detection match the measured contrast? I'm not an electrical engineer, but it doesn't seem hard in my uneducated brain to just set-up a way of reading the sensor that remaps 15 stops of detected contrast into 14bits. Just don't read the sensor in a linear fashion. I've seen this stops vs bits comparison a couple of times and I can't quite grasp the implication.
Netflix and Amazon are 'broadcasting' 4K now. They're even 'broadcasting' in HDR, or they are starting that as soon as HDR TVs are shipping. That's the reality of tomorrow's networks. Adoption of new standards with internet speed.
The issue I have with the lack of 50/60P UHD and the C300mk2 is that Sports will drive adoption of live UHD and Sports is 50/60P now and will remain 50/60P in UHD. The big question for C300mk2 owners is if those sports channels decide to roll their interviews in 23.98/25P (like now), or if they want 50/60P, or if the lack of 50/60P will just eliminate the C300mk2 from consideration. I don't film much Sports, but it affects market size, residual value, rental opportunities, etc, etc, etc.
IMHO, the C300mk2 isn't just an upgrade, it's also catering to a very different crowd than the C300mk1.
Last edited:
They have a dual-readout sensor, which is sort of like - rough analogy - separating the tweeters and woofers on an audio speaker.
I kind of agree but, after the already existing 4K streaming and download services, the satellite companies (DirecTV and Dish in the US) will be offering 4K-to-home. The OTA 4K is being tested but, in the US, not many homes are set up for only OTA. Most households have cable - with 4K set top boxes coming out - or a small digital receiver.
AndreeOnline
Veteran
I'd say, most of the time stops vs bits won't match, but measured contrast will always be less than the theoretical limit (e.g. due to various forms of noise). Here are your linear stops in 14 bits:
1-2 = 1
2-4 = 2
4-8 = 3
8-16 = 4
16-32 = 5
32-64 = 6
64-128 = 7
128-256 = 8
256-512 = 9
512-1024 = 10
1024-2048 = 11
2048-4096 = 12
4096-8192 = 13
8192-16834 = 14
And once we see them like this, it's also obvious why linear mapping won't work. The first stop only has 1 single value to represent it while the last stop gets 8192 values. This is the nature of how light works with each new stop being twice as much light—a doubling.
Like you say: you don't want to map them like this, but rather logarithmically so that each stop gets a more even value distribution. But you can't magically create more stops. The stops aren't hardcoded like: value so and so is stop 3 and value so and so is stop 4. It's a smooth curve that, when viewed on a vectorscope (shooting a wedge), a certain stop falls in a certain value region.
I know it's tempting to just allocate fewer values per stop and think you get more stops, but you don't. Those would just be fractions of a stop in relation to real light.