C300: Can you tell a difference? 4K vs. 2K and 8-bit vs. 10-bit C300 MK II

D

drummondb

Well-known member
I just recorded a project at work, and it gave me the chance to do a little test. I used a C300 MKII and recorded three versions of the same takes: 2K 8-bit on the SD card, 4K 10-bit on the CFast card, and 2K 10-bit via the SDI to a 7Q+.

Color is pretty straightforward: bt709/Neutral/Wide-DR

Here's all three rendered from a 4K timeline to a 4K h.264 MP4: https://youtu.be/S4OLTp7o834

And here's a version rendered from a 2K timeline to a 2K MP4. This one also includes a 200% crop/blow-up where I assumed the 4K would really stand out: https://youtu.be/LdfSBm0_1_k

Thoughts?

UPDATE: 4K Cineform render at max quality: https://vimeo.com/246504701/5350a0fa0c
 
Last edited:
I downloaded the video, took a dpx image from a frame, then I sampled a copy of this frame down to 1080p and then back up and overlayed the two to spot any differences between the three images. Nothing significant resolution wise showed up.

Could you dropbox the original out of camera files (or extract a few seconds) so we can verify it is not due to processing error, edge "enhancements" or bit starvation?
 
Last edited:
From just a short look the take to the left (1st window) looks like it has a little less color fidelity. Maybe a little less detail. Nothing crazy and on its own could be addressed. I think the whole megapixel race to 4k and beyond is kind of overblown but it is what it is. Give me more DR and stronger codecs over shear resolution. So my guess is the first window is the 2k 8bit.
 
I can try to find somewhere to upload the original camera files, but it's big--about 50GB worth.

I didn't do any processing other than dropped the files on a timeline and cropped them to show all three at once. The log for the two encodes out of Adobe media Encoder are below. Both used max render quality and maximum depth:

12/08/2017 12:17:50 PM : Queue Started


- Source File:xxx\2017-12-07 qz2102 new consultant welcome email series 01.prproj
- Output File: M:\2017-12-07 qz2102 new consultant welcome email series\render\2k8bit-v-2k10bit-4k10bit at 4k.mp4
- Preset Used: Custom
- Video: 3840x2160 (1.0), 23.976 fps, Progressive, 00:01:00:00
- Audio: AAC, 320 kbps, 48 kHz, Stereo
- Bitrate: CBR, Target 100.00 Mbps
- Encoding Time: 00:04:40
12/08/2017 12:22:31 PM : File Successfully Encoded


12/08/2017 12:22:32 PM : Queue Stopped


12/08/2017 01:19:01 PM : Queue Started


- Source File: xxx\2017-12-07 qz2102 new consultant welcome email series 01_1.prproj
- Output File: M:\2017-12-07 qz2102 new consultant welcome email series\render\2k8bit-v-2k10bit-4k10bit at 2k from 2k timeline.mp4
- Preset Used: Custom
- Video: 2048x1080 (1.0), 23.976 fps, Progressive, 00:01:00:00
- Audio: AAC, 320 kbps, 48 kHz, Stereo
- Bitrate: CBR, Target 30.00 Mbps
- Encoding Time: 00:01:53
12/08/2017 01:20:54 PM : File Successfully Encoded


12/08/2017 01:20:55 PM : Queue Stopped
 
drummondb said:
I can try to find somewhere to upload the original camera files, but it's big--about 50GB worth.
Click to expand...
Could you extract just a few seconds of frames?

For instance with ffmpeg you could use:

ffmpeg -i input.mp4 -ss 00:00:00 -to 00:00:10 -async 1 -c copy output.mp4
 
drummondb said:
I'll upload a shorter take to google drive.

Here's a 10-bit 4K Cineform render at maximum quality (you can download it too): https://vimeo.com/246504701/5350a0fa0c
Click to expand...
Wait, that is a 4k video as opposed to an UHD video you uploaded to YouTube.

Did you crop the YouTube video or did you resize since it?
Resizing 4K to UHD is a bad idea due to the ratio being too small (something to do with Nyquist's theorem).
 
Last edited:
Hmm, yes that's true. I'm not sure how that happened. At this point, even I'm losing the plot a little. Still, what do you think about the 4K Cineform render on Vimeo?
 
For your 3 up example, the center looks slightly sharper. C300 II 4K (and any 4K Bayer camera shooting 4K) provides about .66 actual resolution, or about 2.5K, so not much more than true, full HD at 1.9K. http://www.dvxuser.com/V6/showthread.php?356356-Actual-Camera-Resolution-Before-Aliasing

If you want to see really detailed 4K, that's where the other brands compete well. Canon excels with colors and skin tones. Best of both worlds is ARRI 65, F65, and latest RED hardware and post software. On the low end the GH5 is decent.

Here I compare 4K and HD (with simple post sharpening) on the 1DX II:

https://www.youtube.com/watch?v=ZIw1CpYKk0c

4K/HD on YouTube isn't as important as overall look, and especially color/skin tones (to me). Without post sharpening, there's a big difference between 4K and HD on the 1DX II (bigger difference than on the C300 II 4K/HD).

We can increase perceived detail/sharpness using concepts from acutance. Here I increase contrast in 3 frequencies ("Multi-Spectral Detail Enhancement") to uprez 1DX II HD to 4K (exaggerated for effect. In practice we wouldn't go that far in most cases):
1DXII_Soft.jpg
1DXII_MS_Sharpen.jpg

For our online content, unless it's a wide shot on green screen, etc., we shoot everything in HD on the C300 II (10-bit 422 at 50Mbps!). The same is true for the 1DX II- I prefer shallower DOF and smaller files most of the time with full-frame HD. For wide shots, especially outdoors/nature, the 1DX II crop 4K works much better than the soft HD (1DX II 4K is still 'low detail' compared to the GH5, for example, however it looks much better in terms of being filmic, color/skin tones, and native shallow DOF without adapters (and DPAF is super useful)). All our Panasonic and Sony gear was sold; picked up a Canon 80D (great little camera for the price!) to complement the 1DX II. Canon's color/look is best for us right now; resolution is less important (for higher end we'd go ARRI).
 
Left has the highest resolution, middle is lower, right is similar to the middle but with more noise.

comparison.jpg

I would still be interested in seeing a few seconds of out of camera 4k to compare it with.
 
Last edited:
Thanks for uploading!

The 4K 4:2:2 has more resolution but the difference is minimal, I think that at those resolutions exact focus and the lens (aperture size) start to become a factor.
HD 4:2:2 ProRes has a nice clean image but the HD 4:2:0 is in my opinion below any standard not in the least due to the very low bitrate.
 
They all look the same. Really. They all look the same. Two smart people in this thread picked the wrong sample as being 4K. It takes someone cropping in to an eyeball from a medium shot to tell which is 4K, and even then the difference is negligible.

8-bit 2K scales up nicely to 10-bit 4K to the point it is virtually indistinguishable from 10-bit 4K in a side-by-side comparison. And 4K cropped in 200% on a 2K timeline looks about as good/crappy as the 2k 8-bit version.
 
Interesting. I am viewing on a 1080p monitor so the detail is not that different. The 1st windows still looks a little different in terms of skin tone to me. This is a great comparison where the sensor and overall camera package outweigh any recording hurdles. Kind of proves 8bit is fine for baked in color. 10bit is probably more apparent with LOG shooting or with more pushing around in post.
 
Bassman2003 said:
Interesting. I am viewing on a 1080p monitor so the detail is not that different. The 1st windows still looks a little different in terms of skin tone to me. This is a great comparison where the sensor and overall camera package outweigh any recording hurdles. Kind of proves 8bit is fine for baked in color. 10bit is probably more apparent with LOG shooting or with more pushing around in post.
Click to expand...

Right- 8-bit is fine unless significant post grading. Even 8-bit 420 100Mbps 4K SLog2 on the A7S II was fine most of the time (exposed +2 over). 50Mbps 422 10-bit HD C300 II has worked well for us with green screen (static camera) and YouTube delivery. When shooting full-body/wide on green screen, we use 4K (~410Mbps) and it makes a significant difference (Canon 24-70 F2.8 II (around 24) or 16-35 F2.8 II (around 16) at around F6.3-7.1, ISO 400).

The lesson we learned after going to razor-sharp Panasonic and Sony (not quite as sharp) was that we care most about color and overall look vs. maximum resolution. When one studies film, it's typically not that 'sharp' per se, instead it's sharp enough, and has detail in the grain, which is a pleasing combination, especially for people and close-ups. For wide landscapes, opening shots, all the detail you can get is helpful (as long as there are no sharpening artifacts or aliasing)!
 
Yeah! Delivery is now the great leveller because camera capabilities are far outstripping the ability to deliver the content to viewers.

I know any camera I buy (C300 MKIII !) I'm buying it to make my life easier through new features or ease in post, or feed my ego, not because it will make things significantly better for my viewers.

I think it's better to focus on things that make the content better (autofocus, colour, lenses) than pure numbers (4k, x mb\s) because we already passed the point where many modern cameras can hit those numbers.
 
You must log in or register to reply here.
Back
Top