ALl-I Or IPB on MKIII?

Until Canon updates the codecs, this is still true (I work with video codecs and write image processing software in my day job- I analyzed this in detail):

From the result of my tests:

Use ALL-I for 720p60 and 1080p24 with fast motion.
Use IPB with relatively static camera shots or when you can't see any improvement with ALL-I.

While it's possible to see a quality difference with ALL-I for fast camera motion when looking at still frames, it's not that easy to notice a difference when played back in real-time (except with 720p60: use ALL-I all the time). If disk space is an issue (editing on a laptop or SSDs), IPB quality is overall excellent at ~3x less disk space used.
Click to expand...


In my spare time I work on projects like this: http://www.youtube.com/watch?v=BLbZ_nwm8-I (100% IPB)
 
The REAL WORLD is full of people who live in a permanent state of cognitive dissonance due to good perception management. John's assessment is spot on. The VBR implementation of ALL-I is flawed but I'm no fan of IPB either. I'll be recording everything straight to Prores come April.
 
The REAL WORLD is full of people who live in a permanent state of cognitive dissonance due to good perception management
Click to expand...
Thanks squig for refreshing my conceptual toolbox with the notion of "cognitive dissonance" and thanks John for formulating clearly what I had not verbalized in my conceptual mist. Yes, when I purchased the 5DMkIII I wanted to believe that ALL-I would yield much better footage...

There are, yet, as far as I can see, still two issues left:

1. Is the claim that ALL-I yields a better temporal resolution when filming high-motion scenes thoroughly and repeatedly verified by tests applied to a variety of light, detail and motion alternatives? Any links and pointers are welcome!

2. If the temporal resolution gain happens to be negligible and the only remaining benefit of using ALL-I is that the footage is less painful to display on weak computers than IPB, why then has Canon implemented it? Should the conclusion that ALL-I is only a marketing tricks, a kind of bait for credulous buyers who were led to believe (wrongly) that ALL-I would yield much higher quality "almost" fulfilling (but arrrgh the bit rate is just to slow) Discovery and BBC standards for acceptable footage, keeping the uncompressed clean HDMI option as a joker to counter the upcoming competition?

As you might read in between the lines – I dislike very much to be "framed" (pun intended) my a major corporation especially when the price of the 5DMK3 is quite stiff.

PS. It seems that the ALL-I option in D800 and the GH3 is no better. Or is it?
 
Last edited:
danap said:
1. Is the claim that ALL-I yields a better temporal resolution when filming high-motion scenes thoroughly and repeatedly verified by tests applied to a variety of light, detail and motion alternatives? Any links and pointers are welcome!
Click to expand...

I film explosions regularly and ALL-I is significantly better. I personally also see better quality in low motion scenes too. IPB tends to smooth out details and causes a more plasticky look. ALL-I also let's slightly more detail and noise through and shadow noise is quite a bit better than IPB. In motion (leaves, walking etc) the difference is even more pronounced.

There are people who don't even see an external recorder with prores being better than internal AVCHD so I don't concern myself with those.

There is no ALL-I in D800.
 
Here are some examples showing IPB retaining more information vs. ALL-I for very high detail, low motion footage: http://www.dvxuser.com/V6/showthrea...more-detail-and-has-less-artifacts-than-ALL-I

Canon's current implementation of ALL-I doesn't have a high enough bitrate especially given the way they implemented VBR. Consider that the 50mbps MPEG2 (422!) codec used for the C300 does a superior job in every way which is not an ALL-I codec, but rather a highly tuned long GOP MPEG2 design. MPEG2 is less efficient than than H.264. Thus, a tuned H.264 422 codec can do much better than MPEG2 at the same or lower bitrate. Sony is moving forward with such codecs. It's unclear if current NLEs can handle the new codecs, even though they are implemented as part of the existing H.264 spec.
 
danap said:
1. Is the claim that ALL-I yields a better temporal resolution when filming high-motion scenes thoroughly and repeatedly verified by tests applied to a variety of light, detail and motion alternatives? Any links and pointers are welcome!

PS. It seems that the ALL-I option in D800 and the GH3 is no better. Or is it?
Click to expand...

The GH3 early adopters have found the same problems and the general consensus is that IPB looks better. IPB on the GH3 has a higher bit-rate than the MKIII @ 50Mbps.

ALL-I works well with fast motion because the VBR maxes out @ 91Mbps and intra-frame codecs don't produce mud in fast motion like inter-frame codecs do. VBR ALL-I is a bit dumb when it sees something like a big patch of blue sky, it decides not to throw a lot of bits at the sky and that's where you can get nasty macro blocking.

If ALL-I had a constant bit-rate even at 91Mbps it would look much better, but the hardware should easily be capable of 150Mbps 4:2:2 which would yield Prores like quality. Magic Lantern have an alpha hack in progress but they haven't cracked the codec yet.
 
maarek said:
There are people who don't even see an external recorder with prores being better than internal AVCHD so I don't concern myself with those.
Click to expand...

People who don't do much grading perhaps. I've been following the discussions of the FS100 guys and the difference is night and day in post. I've got a Nex-5n and I can tell you from experience that AVCHD falls apart pretty quickly in the grade, it's not pretty. H.264 holds up better. I've noticed some extra artifacts at times when I've converted ALL-I footage to Prores HQ. I had to go to Proress 4444 to get rid of it. I won't touch the internal codecs once the uncompressed HDMI out firmware comes out unless Magic Lantern can crack them. The MKIII is crippled by it's consumer grade codecs just like the MKII was crippled with 30p and moire.
 
Yep, it's not just the compression. People who compare the 4:2:0 onboard footage with 4:2:2 external prores footage will often say "I don't see any difference". And that's because they don't. Did you know that only a small percentage of the cones in your eye can see "blue"? But just because your eye isn't good at seeing chroma, doesn't mean the grading program can't use that information. So, for significant grading, it's wise to use an external recorder whether or not you can perceptually see a difference in coloration (as well as compression).

squig said:
People who don't do much grading perhaps. I've been following the discussions of the FS100 guys and the difference is night and day in post. I've got a Nex-5n and I can tell you from experience that AVCHD falls apart pretty quickly in the grade, it's not pretty. H.264 holds up better. I've noticed some extra artifacts at times when I've converted ALL-I footage to Prores HQ. I had to go to Proress 4444 to get rid of it. I won't touch the internal codecs once the uncompressed HDMI out firmware comes out unless Magic Lantern can crack them. The MKIII is crippled by it's consumer grade codecs just like the MKII was crippled with 30p and moire.
Click to expand...
 
squig said:
People who don't do much grading perhaps. I've been following the discussions of the FS100 guys and the difference is night and day in post.
Click to expand...

I agree. That's why I use All-I. But I don't think the ALL-i is a crippling codec. I believe that ALL-i is considerably better than the AVCHD implementation in FS100 or the C100. Ofcourse, the cam is so soft that the codec doesn't really get challenged anyway. But I've seen horrible motion artifacts with even moderate movement on the FS700 and the 5d mark III was completely clean on the same shoot. So yeah. It's not crippling at all. Atleast for me.
 
You must log in or register to reply here.
Back
Top