5D MKIII specs

jcs said:
The ALL-I clip was 84.8Mbps, and IPB was 30.4Mbps! It's possible IPB is a lower max bitrate, but it sounds like they added B frames, so that increases efficiency (5D2 had a simpler H.264 profile and no B frames running at around 38Mbps). A higher bitrate would be welcome, of course. At least one region looks a little better with ALL-I- the back of the SUV (top red light area: red light is better defined). Without the sharpening and pixel peeping, IPB looks better than ALL-I due to lower noise crawl (and looks more detailed). Perhaps if Canon reduced input noise to ALL-I that might help (however I'd prefer to see higher quality from a long GOP codec to save disk space- more efficient and very high quality (e.g. Canon's 50Mbps 422 MXF for the C300 and XF camcorders)).
Click to expand...

They did add b-frames. I read that in the specs/feature list on a Canon web page. So it's a lot more efficient than the H.264 on the MKII. THe FS100 AVCHD has b-frames running at 24Mbit/s, some people have compared that to prores via the HDMI out and they haven't noticed a bump in quality.
 
Emanuel said:
: D Keep going, Anthony, keep going :-)
Click to expand...

hehe...I'm just warming up!

It will be interesting to see which grades better IPB to prores or ALL-I. I mentioned earlier that someone was saying they had to go to prores444 to get the best transcode out of IPB. I think that's a testament to the IQ of the IPB codec....I'm still hanging for a magic lantern bit-rate boost for both codecs.

I'm a GWMP (guy with a macbook pro).
I'm with you dusty, FCPX is cool...and I used to cut SP.
 
StMad said:
You're calling someone out here, right?



You don't find it ironic that this is your next post?
Click to expand...
Not at all.
One post I tell you what I'm seeing.
Next posts saying it's still really early and new to us, so somebody might come along and say this or that.

Two sides to every coin. Every positive reaction has a negative reaction. That is science and observation.
 
jcs said:
Pro audio test recorded straight into camera (not separate audio):
http://www.youtube.com/watch?v=2bWGogt2Kxg&hd=1
It's clean, though dynamics could be improved (need to experiment with gain stages).
Click to expand...

Sounds like the wife is at work. You're at home playing with expensive stuff. Wife is tired and had a "busy" day. She thinks you play too much with your toys. Therefore, you're in charge of dinner. ****, it's 5pm. I gotta go get dinner ready as my wife will be home any minute and cranky as hell! Ahhhhh, the parallels

The audio sounds good.
 
dustylense said:
Sounds like the wife is at work. You're at home playing with expensive stuff. Wife is tired and had a "busy" day. She thinks you play too much with your toys. Therefore, you're in charge of dinner. s*%t, it's 5pm. I gotta go get dinner ready as my wife will be home any minute and cranky as hell! Ahhhhh, the parallels

The audio sounds good.
Click to expand...

I'm allowed to play with my toys as long as they eventually bring home the bacon....still have to cook dinner though.

jcs said:
I understand what the problem is: the ALL-I codec isn't doing a very good job at VBR: quantizing (throwing away) too much information in the high-frequency domain. I don't think it's necessary to perform additional analysis: isn't the problem already clear? See the DCT-II link I posted earlier in the thread (and watch the GIF animation, all the way through). ALL-I does exactly the same thing except with variable macroblock sizes (effectively a fancier and more efficient MJPEG). This could also be a side effect of too much input noise (as we see in ALL-I footage). It appears Canon is doing less noise processing for ALL-I, and that could break VBR quantization. As others have mentioned, 91Mbps should be decent for an ALL-I implementation.
Click to expand...


My first DSLR was a D90, it has an m-jpeg 24Mbit/s VBR codec. THe bit-rate dropped to 7Mbit/s in some shots and was really nasty. Even at 24Mbit/s it was pretty lousy. The ALL-I codec I'm told drops to the low 30Mbit/s range. A high quality 1080p jpeg still is around 500k. 24x 500k is roughly 90Mbit/s so if ALL-I was CBR there's no reason for it to not look really good. I reckon it could go higher than 90Mb/s too....I say we start another firmware update petition!
 
@ JCS.

I shot a scene with horizontal blinds across the street, stucco wall, bricks and foliage. Graded heavily (now I'm into pushing the codecs) and saw that if you want to stay safe, my thoughts are shoot with IBP as the same scene in I-frame did show some moireing (not disastrous) and the IBP didn't in the horizontal blinds. Also, shoot test scenes with statics at first and them make sure to move the camera with pans up and down to see vert and hort res and artifacts. In that same shot, the I-frame had more uniform resolution (stucco wall with FINE details), but as soon as I started camera movement the slight moire appeared in the blinds. The IBP stayed clean throughout but at the cost of "uniform" resolution. But then again, this is just on scene. That is why I thinks it's too early to say "this is the best settings to shoot the Mk3 with".
 
When I read the MKIII has a 91Mbit/s intra codec I was like :2vrolijk_08:

When I heard it was VBR I was like :nads:

There is still hope with the rebel alliance (magic lantern) :kali: Take that Darth Canon!
 
dustylense said:
@ JCS.

I shot a scene with horizontal blinds across the street, stucco wall, bricks and foliage. Graded heavily (now I'm into pushing the codecs) and saw that if you want to stay safe, my thoughts are shoot with IBP as the same scene in I-frame did show some moireing (not disastrous) and the IBP didn't in the horizontal blinds. Also, shoot test scenes with statics at first and them make sure to move the camera with pans up and down to see vert and hort res and artifacts. In that same shot, the I-frame had more uniform resolution (stucco wall with FINE details), but as soon as I started camera movement the slight moire appeared in the blinds. The IBP stayed clean throughout but at the cost of "uniform" resolution. But then again, this is just on scene. That is why I thinks it's too early to say "this is the best settings to shoot the Mk3 with".
Click to expand...

Cool. I'm shootin' IPB until Canon updates the firmware. Sharpened IPB looks excellent- no surprises; preamped mics work well with internal audio set to a gain of 1; good enough to start shootin'. Time to open the box for the second 5D3 and test dual camera & timecode (had considered returning the 2nd 5D3 and getting an FS100 or better or a traditional pro camcorder. 5D3's full frame has that magic look and works with all my lenses; found no deal breakers after extensive testing and analysis).

Question on lenses: the 24-105 F4L maintains focus while zooming. The new 70-200 F2.8L II does not. Is this proper behavior for this lens?
 
Those of us that know a bit about codecs are calling for Canon to change the ALL-I codec to CBR (constant bit-rate) from VBR (variable bit-rate). Canons published specs on the MKII state that ALL-I encodes at 235MB/min (91Mbit/s). Legally speaking it could be argued that this is a misrepresentation because in real world tests done by users on this forum and cinema5D the VBR ALL-I codec is averaging around 35Mbit/s. A CBR firmware update would keep the bit-rate very close to an average 90Mbit/s and should address the macroblocking/noise issues users are experiencing. A menu option for higher ALL-I and IPB bit-rates would also be a much welcomed update to the firmware. We petitioned for 24p and manual controls and Canon listened and made a tonne of money so I urge everyone to contact Canon by phone, email, bullhorn, sit ins, flash mobs, picket lines, or any other non-violent tactic to get the most out of this camera. Vive la revolution!

Pablov @ cinema5D wrote-

CANON 5D Mark 3 Next Firmware Update:

- I'm concerned about this:

Canon HAS OFFICIALLY admitted (via e-mail) that there is an issue in current 5D3 firmware with two lenses (IS not working properly), and they are working to fix it.

Now, if we don't send our requests to Canon about All-I and in-camera Sharpness bad implementation / poor performance, I'm not sure if they will fix it soon.

So I encourage everyone interested to contact Canon and send them an email. Many of us already did it. But we need to "push" a bit for a fix, otherwise it may be skipped.

- CONTACT INFORMATION (e-mail & phone) is here: viewtopic.php?f=56&t=39967&p=227101#p227101
 
Last edited:
40,000 Philip Bloom disciples can see the tweet I sent him so I expect Canon are going to be inundated real soon :thumbsup:. They fixed the MKII 2.0.3 custom dial sound bug I discovered really quickly so hopefully we won't have to wait long for HQ jpeg looking CBR ALL-I. If they don't respond by next week I propose we start an online petition. Call the corporations to account, 91Mbit/s should be 91Mbit/s!

Now where's my Cuban cigars...
 
I'm playing with the D800 now: >13 stops of DR, anyone? (i.e.: first Flaat_13p_beta1 results in: this thing works!) (edit: well, not as nicely as it seemed at first sight: the picture control works but results are not as good as expected; similar to what the canons do, the camera is doing some nasty transformation to the color space before applying the curve and sending the result to the codec; this kills a lot of the headrom I have to play with in the RAW stills)
 
Last edited:
You must log in or register to reply here.
Back
Top