5D MKIII specs

The thing is this with the move towards Intra - I frames and the faster processors the hack has a real opportunity to do what was done with the GH2 on the 5D Mark III. Pushing the bit rate can definitely improve things further in a way that could not be done on the Mark II by the Magic Lantern community. I personally view the 90 Mbits/s on the Mark III, as a big part of that which gives the image greater stability, not just a faster read time of the sensor. On that front, we may have a little bit of skew left, but there is nothing resembling the jello that I have seen on the Mark II or even the GH2 for that matter. (I own two of each, so believe me I know). Clearly, I am waiting to see how the C-DSLR helps things along, but the only question remains for me how well the 5DMIII will function as B-Roll to a 4K camera. A bit like trying to cut a lav mic with a shotgun. Some combinations cut together better, and not always from the same manufacturer.
 
yoclay said:
The thing is this with the move towards Intra - I frames and the faster processors the hack has a real opportunity to do what was done with the GH2 on the 5D Mark III. Pushing the bit rate can definitely improve things further in a way that could not be done on the Mark II by the Magic Lantern community.
Click to expand...

Magic lantern has been reliable up to 70Mbit/s in CBR mode. There might be some more buffer/write speed in the MKIII that can be exploited by magic lantern but it's a shame they didn't go with XQD media http://en.wikipedia.org/wiki/XQD_card. Perhaps the C-DSLR will write to XQD. Canon announced future XQD implementation ages ago. The D4 with its XQD media ia ripe for an intra hack.
 
http://www.arihazeghiphotography.com/AH_CW_interview/

Chuck Westfall, technical advisor at the Professional Engineering & Solutions Division, Canon U.S.A.:

"... We also doubled the throughput speed of image data from the CMOS sensor to the rest of the image processing chain by increasing the number of simultaneous readout channels from 4 to 8. This change has a side effect of helping to improve EOS Movie quality as well by cutting the level of rolling shutter artifacts in half. ..."
 
Peter J. DeCrescenzo said:
http://www.arihazeghiphotography.com/AH_CW_interview/

Chuck Westfall, technical advisor at the Professional Engineering & Solutions Division, Canon U.S.A.:

"... We also doubled the throughput speed of image data from the CMOS sensor to the rest of the image processing chain by increasing the number of simultaneous readout channels from 4 to 8. This change has a side effect of helping to improve EOS Movie quality as well by cutting the level of rolling shutter artifacts in half. ..."
Click to expand...

So the 1DX with a 16 channel readout should have negligible rolling shutter?
 
Compact flash will have enough bandwidth for 200+ mbit easily. I have the driftwood hack at 176 on a 30mbs sandisk class 10 SD card on my gh2. So the bandwidth should be enough for higher numbers on the Canon if they hack it.
 
if the sensor readout is twice as fast... that's bad news: on the old canons, sensor readout was around 60ms, cut to 20ms through line skipping (read only 1/3 of the lines); if the new sensor has a read out time of around 30ms, and rolling shutter is about half (there are some tests on this already), it means there's still line skipping (read only 1/3 of the lines to get it to 10ms), and it's a crazy math algorigthm that leads to greatly reduced aliasing/moire - thus the softness :(
 
NormanBates said:
if the sensor readout is twice as fast... that's bad news: on the old canons, sensor readout was around 60ms, cut to 20ms through line skipping (read only 1/3 of the lines); if the new sensor has a read out time of around 30ms, and rolling shutter is about half (there are some tests on this already), it means there's still line skipping (read only 1/3 of the lines to get it to 10ms), and it's a crazy math algorigthm that leads to greatly reduced aliasing/moire - thus the softness :(
Click to expand...

Wrong. The sensor readout was 60ms AFTER the line skipping on the 5d.

And it doesn't need crazy math as the pixels are exactly 3x. You can just drop those pixels and there will be no aliasing. Offcourse it should be sharper that way.
 
ok, I think I'm not, but I could be wron on the 60ms thing; but it still has half the rolling shutter with half the readout speed, so my conclusion remains in any case: 5D3 seems to be using line skipping just like the 5D2 did - and yes, this would lead to aliasing/moire if you didn't apply crazy math at the debayering step, even if it is 3x 1080p - aliasing/moire appears because of issues similar to the following: if you skip 2 out of every 3 lines, your R and B samples are very far apart from each other, one may be falliing on a brick and the other in the mortar in between them, slight movements change this, and you end up seeing horrible rainbows in the debayered image
 
NormanBates said:
ok, I think I'm not, but I could be wron on the 60ms thing; but it still has half the rolling shutter with half the readout speed, so my conclusion remains in any case: 5D3 seems to be using line skipping just like the 5D2 did - and yes, this would lead to aliasing/moire if you didn't apply crazy math at the debayering step, even if it is 3x 1080p - aliasing/moire appears because of issues similar to the following: if you skip 2 out of every 3 lines, your R and B samples are very far apart from each other, one may be falliing on a brick and the other in the mortar in between them, slight movements change this, and you end up seeing horrible rainbows in the debayered image
Click to expand...

I know that they specifically concentrated on removing the chromatic noise related to moire. Honestly, I think you would need to take a test drive to determine if it satisfies you or not. Personally, I was pleasantly surprised at how many non-issues there were. That said, obviously I couldn't download anything I shot...
 
squig said:
Some new test footage for all you peepers out there-

http://www.youtube.com/watch?v=TbzBC8X593A
http://www.youtube.com/watch?v=qr4H7QycmZU
http://www.youtube.com/watch?v=J-uOqMiD21I
Click to expand...

Thanks man. But, this is not impressive at all. Not even a lil' bit unfortunately.
It looks almost exactly like the MKII in every way and even giving a benefit of the doubt to youtube compression for artifacts, it doesn't affect the resolution by much, if at all.

Maybe the C-DSLR will be better. But, now w/ the upcoming Bolex (potentially being cheaper and superior in every way filming wise), even that C-series camera had better be on point :-/
So far, the MKIII seemed like alotta' waiting for a lil' itty bitty pay off. Very typical Japanese business practice (coming from the gaming industry here).

No bueno.
 
I think thats the point of the c-dslr. Thats the point of Canon's new department 'Cinema EOS'

Yes they try and fix a few problems here and there with the video on a still camera - but its is going to remain a still camera.

Cinema EOS = Cinema + canon dslr
 
maarek said:
So the 1DX with a 16 channel readout should have negligible rolling shutter?
Click to expand...

5Dm2 was around 25ms readout, so that might means (assuming no other speed increase to the technology)

12.5ms 5Dm3

6.25ms 1Dx

Given working with 14ms on the FS/AF100 was fine for me, wow, 12.5ms would be AWESOME to have.

6.25ms would be REALLY awesome if real.

That said, you'll still be able to make the image skew panning too fast, because the skew disappears slower than the perception of jello.

If you panned real fast and got a slant that was 100 pixels top to bottom, that would be 50 pixels on 2x and 25ps on 4x speed. Meaning, with a fast enough shutter speed, you'd still see it. Of course, it will be greatly reduced.

What is interesting to me is thinking about motion blur/rolling shutter and where the inflection point is at 1/48 where you don't notice the skew as much because the motion blur is so great.

I really want to get my hands on a C300 so I can test the speed of the rolling shutter, as it looks great. Maybe it's 6.25ms as well if they used the same technology as the 1Dx for how many lanes they are using.
 
You must log in or register to reply here.
Back
Top