Canon 5D Mark III

Started by a1ex, June 14, 2012, 04:45:09 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

hjfilmspeed

Its amazing whats been done so far with ml and the 5diii! excellent work everyone! I believe the potential for the 5d3 is great. my biggest strife with its video is h.264. I was messing with the s2 jpegs, shooting at 6fps the opening the files up as a jpeg sequence in a 24p timeline. if you shoot s2 jpegs in video mode the files are cropped at 1920x1080 and are only  .5 to 1 MB in size and were way sharper. could it be possible to somehow  shoot s2 jpeg sequences at 24p say for maybe shorter period of time? if so the sharpness and quality would be insane. i was using marvels cine and htp (which normaly makes uglys) and the quality was awsome. i dought this is possible but the jpeg downsampling was real sharp and so easy to grade. it would probably call for a data rate around 200mbs. so what if playback isnt there it would be worth the quality. thanks again for all your hard work!

hjfilmspeed

Im sure this has been discussed and im sory if it sounds like a broken record but im just curious. in any case i cant wait for the ful ml on the 5d3!!! such exciting times! Thank you so much! The HDR LOOKS SO RAD! so awsome!

csound

The HDR is VERY impressive!!

agsingh

Does this mean HDR will be included in the next alpha build?  :D

a1ex


joxxie

awesome .. any idea when alpha II will be available ?

a1ex


NtiT

Quote from: hjfilmspeed on September 28, 2012, 04:00:19 PM
Its amazing whats been done so far with ml and the 5diii! excellent work everyone! I believe the potential for the 5d3 is great. my biggest strife with its video is h.264. I was messing with the s2 jpegs, shooting at 6fps the opening the files up as a jpeg sequence in a 24p timeline. if you shoot s2 jpegs in video mode the files are cropped at 1920x1080 and are only  .5 to 1 MB in size and were way sharper. could it be possible to somehow  shoot s2 jpeg sequences at 24p say for maybe shorter period of time? if so the sharpness and quality would be insane. i was using marvels cine and htp (which normaly makes uglys) and the quality was awsome. i dought this is possible but the jpeg downsampling was real sharp and so easy to grade. it would probably call for a data rate around 200mbs. so what if playback isnt there it would be worth the quality. thanks again for all your hard work!
Yeah, the more I read about Canon limiting the potential of their products in order to make more money, the more disappointed I get. I really hope that the amazing ML team is able to somehow increase sharpness/resolution of the 5D3, as unlikely as it seems.

Andrzej

Quote from: a1ex on September 28, 2012, 11:52:03 AM
5D3 does proper downsampling at 720p
Is it after ML implementation ?
In my case, 720p producing about 500 lines
I am running Firmware wer. 1.1.2
Is there a hope for true 720p ?
Canon 5dm3, 550d, Canon 24-105mm 4.0 IS, Canon 18-55mm 3.5-5.6 IS, Sigma 30mm 1.4, Sigma 20mm 1.8, Sigma 70-300mm 4.0-5.6, Samyang 14mm 2.8, Falcon 85mm 1.4, Zeiss 180mm 2.8 and more...

a1ex

I couldn't see any noticeable moire on my test chart ( http://graphics.cs.williams.edu/data/images/star-chart.png ), maybe just a tiny bit more than at 24p. On 5D2, in FullHD, moire is clearly visible. So I guess it's proper downsampling.

Andrzej

Yes. 5dm3 is ok for moire, aliasing. But resolution is not real. 5dm3 is producing less lines then 550d. It is clearly visible in 720p mode.  http://www.graphics.cornell.edu/~westin/misc/res-chart.html
Increasing resolution in 720p would be a hudge step.
Check any thrailer in 720p: http://trailers.apple.com/trailers/
There is true resolution and 24fps. (Canon framerate in 720p is: 50 and 60 !)

True video resolution is what we all need in 5dm3
Canon 5dm3, 550d, Canon 24-105mm 4.0 IS, Canon 18-55mm 3.5-5.6 IS, Sigma 30mm 1.4, Sigma 20mm 1.8, Sigma 70-300mm 4.0-5.6, Samyang 14mm 2.8, Falcon 85mm 1.4, Zeiss 180mm 2.8 and more...

a1ex

Then, sell the 5D3 and buy a 550D.

AndreasK


Andrzej

Quote from: a1ex on October 08, 2012, 03:03:23 PM
Then, sell the 5D3 and buy a 550D.
Giving up alredy ? Do i write about 4k like the others?
You did so much amazing things with ML
Canon 5dm3, 550d, Canon 24-105mm 4.0 IS, Canon 18-55mm 3.5-5.6 IS, Sigma 30mm 1.4, Sigma 20mm 1.8, Sigma 70-300mm 4.0-5.6, Samyang 14mm 2.8, Falcon 85mm 1.4, Zeiss 180mm 2.8 and more...

a1ex

Giving up on what? ML never changed sensor scanning modes and similar low-level stuff.

Andrzej

Quote from: a1ex on October 08, 2012, 03:54:16 PM
Giving up on what?
Trying to increase resolution. I was wondering if thats possible witch ML
Canon 5dm3, 550d, Canon 24-105mm 4.0 IS, Canon 18-55mm 3.5-5.6 IS, Sigma 30mm 1.4, Sigma 20mm 1.8, Sigma 70-300mm 4.0-5.6, Samyang 14mm 2.8, Falcon 85mm 1.4, Zeiss 180mm 2.8 and more...

nanomad

EOS 1100D | EOS 650 (No, I didn't forget the D) | Ye Olde Canon EF Lenses ('87): 50 f/1.8 - 28 f/2.8 - 70-210 f/4 | EF-S 18-55 f/3.5-5.6 | Metz 36 AF-5

hjfilmspeed

ugg it just bothers me to see the camera can produce tack sharp 1080 jpegs (processed in camera). i guess it cant do it at 24p for any length of time. I was hoping this could be possible with silent stills but judging by these last few posts i guess not.

to be fair the 5d3 wasnt designed to be a full frame cine camera. it was designed for photos. you want cine from canon you better be ready to pay for it. It would be awsome if canon could make a 16mm version of the c300. then guys like me might be able to afford it (after i sell sum camera stuff ha ha) But ML is for making the most out of these dslrs for cine work thats for sure. love it so much! Thanks again!

Andrzej

Here is an intresting post from a differend forum about 5dm3 olpf:
Quotehow can this OLPF filter be strong enough that the video looks so soft out of the camera, yet weak enough that it doesn't affect stills? with such a strong filter, RAW stills should be a mush of non-detailed identical pixels, but they're not - which leads me to think... (you know I'm always on the optimistic side) that filter actually changes from stills to video mode, it is electronically controlled somehow... and Magic Lantern will give as a way to control it without tearing the camera or getting rid of the dust-removal mechanism

http://philipbloom.net/2012/04/01/a-drastic-solution-to-increasing-sharpness-with-the-5dmkiii/
Canon 5dm3, 550d, Canon 24-105mm 4.0 IS, Canon 18-55mm 3.5-5.6 IS, Sigma 30mm 1.4, Sigma 20mm 1.8, Sigma 70-300mm 4.0-5.6, Samyang 14mm 2.8, Falcon 85mm 1.4, Zeiss 180mm 2.8 and more...

hjfilmspeed

hmmmm i dont think its that. Even with the filter 100% removed the 1080 vid is only a slight bit better but not nearly as sharp as the s2 jpeg. its deffinetly canons method of downsampling in vid mode for sure. You never know ML always surprises but this one sounds tricky.

Andrzej

Test is based on Neutral preset zero sharpness, | -2 contrast, | -2 Sats
Second similar filter was untouched
Canon 5dm3, 550d, Canon 24-105mm 4.0 IS, Canon 18-55mm 3.5-5.6 IS, Sigma 30mm 1.4, Sigma 20mm 1.8, Sigma 70-300mm 4.0-5.6, Samyang 14mm 2.8, Falcon 85mm 1.4, Zeiss 180mm 2.8 and more...

hjfilmspeed

from the tests ive seen removing the olpf is not
the answer. if theres proof otherwise i think
the olpf thing was busted.

Andrzej

I dont know what kind of tests you have seen.
Canon 5dm3, 550d, Canon 24-105mm 4.0 IS, Canon 18-55mm 3.5-5.6 IS, Sigma 30mm 1.4, Sigma 20mm 1.8, Sigma 70-300mm 4.0-5.6, Samyang 14mm 2.8, Falcon 85mm 1.4, Zeiss 180mm 2.8 and more...

hjfilmspeed

These tests. If they are true anyways.

http://www.eoshd.com/content/8201/canon-5d-mark-iii-without-optical-low-pass-filter-the-verdict

The results are very minor. There is a way bigger jump comparing the s2 jpeg shot in video mode to the all-i.

I realy think its like a1ex said. The 5d3 rez issue boils down to the downsampling/sensor scanning method and maybe codec. As alex has posted ML cant change the type of codec. He himself was very optimistic in saying that there might be a very very slim chance of mjpeg or realy short 422 clips but again that in itself would be a crazy tuff code to crack.

Andrzej

The company who did this modification replaced olpf with anoder piece of glass... Also there is a calibration issue.
I think Andrew's test is realible, but this was done diffriend way.
Canon 5dm3, 550d, Canon 24-105mm 4.0 IS, Canon 18-55mm 3.5-5.6 IS, Sigma 30mm 1.4, Sigma 20mm 1.8, Sigma 70-300mm 4.0-5.6, Samyang 14mm 2.8, Falcon 85mm 1.4, Zeiss 180mm 2.8 and more...