- March 27, 2008 at 9:52 pm
Part 2 of my first finds with the XDCAM EX 1080 codec. Now onto the 720 codec.
As you would have guessed, 720p24 is the best looking format in terms of compression artifacts. The amount of artifacts is quite minimal actually which is really nice to see.
720p24 generally looks better than 1080p24 scaled down to 720p24 in post:
1080p24 scaled down to 720p24 in post retains some of the 1080’s compression artifacts which the native 720p24 format inherently has less of than 1080. However (!), when scaling down 1080 to 720 in post, some of the rough mosquito noise (common in red edges and such) gets smoothed out when scaled down to 720 (but again, the compression artifacts are still present). Something to possibly consider when working with motion graphics (although there’s obviously no magical increase in chroma sub-sampling, so the color fidelity will still be the same).
So far, I’m suggesting this…
1080 deployment: shoot in 1080
720 deployment: shoot in 720 (1080 scaled down to 720 in post may actually be worse than shooting 720 natively)
480 deployment: shoot in 720
Notes about shooting 720 for 480:
There’s actually enough compression artifacts in 1080 that are retained down to 480 and since 720 has fewer compression artifacts (and requires less file size space), shooting 720 for 480 seems to be a wiser choice.
I should note too that these synthetic tests I perform push the codecs to their limits and are not entirely reflective of what real-world shooting encounters. In most cases, trying to find the compression artifact differences on an HD video monitor between 1080 and 720 out of the EX1 will actually be quite difficult. Use these results as merely a guide but not a rule-of-thumb.
Another interesting note; even at 1080p24, the compression makeup is generally better looking than that of DV25 in many cases, which is pretty significant considering XDCAM EX has only 10 megabits-per-second more of bandwidth than DV25 yet has 6 times the pixel count (2.07 MP versus 0.34 MP). In general, I’d say the XDCAM EX codec is pretty good for the limited amount of bandwidth it has, but I wonder why Sony didn’t simply push the bandwidth limit to 50 as opposed to 35. For 720 HD, 35 Mb/s is adequate, but 1080 HD could use just a “bit” more.
- March 28, 2008 at 2:07 am
[Marco Solorio] “720 deployment: shoot in 720”
Although sometimes shooting 1080 and putting it into 720 timeline allow you room to keyframe some moves.
The same goes for 1080 in 480 timeline. Makes for some interesting pseudo pan and dolly moves.
I thought it was important to mention some exceptions to the rule.
- March 28, 2008 at 4:27 am
[Marco Solorio] “but I wonder why Sony didn’t simply push the bandwidth limit to 50 as opposed to 35. For 720 HD, 35 Mb/s is adequate, but 1080 HD could use just a “bit” more. “
Most likely because of the data rate limitation of recording to the SxS Pro Cards.
Good stuff, Marco.
- March 28, 2008 at 8:23 am
i don’t think that 35mbps has to do with the sxs cards,
for in an expresscard slot copy times are 5x faster than realtime..
so there’s plenty of room for transfer/encoding speed…
to me i wonder why it isn’t possible to get 60fps at 1080 –
at least over sdi.
these are progressive chips, and it’s possible to to 60i,
so the processors can be clocked to 60fps at 1080. Seems
that either the compressing chips can’t handle it, or Sony
simply doesn’t want it in a camera at such a price point..
- March 28, 2008 at 9:44 am
Thanks for sharing the tests Marcos. Very interesting.
[Marco Solorio] ” I wonder why Sony didn’t simply push the bandwidth limit to 50 as opposed to 35″
When shooting 720p24/60, the SxS cards are recording near 100Mbps. This machine could be recording even MPG2 422 without problem.
Mac OX 10.5.2-FC 6.02-QT 7.4.1
G5 2x2Gh 4GbRAM-BlackMagic Extreme
PMBP 17″Core2Duo 4GbRAM-AJA ioHD
SONY EX-1 . SONY PD170
..and always a big mess on top of the table.
- March 28, 2008 at 5:11 pm
Interesting stuff and thanks for the post!
I am noticing almost exactly the opposite. Most of the artifacting is minimized when 1080 is reduced to 720, and produces a cleaner final 720 result.
I had settled on “use 720 only when shooting for slow motion”.
- March 28, 2008 at 6:22 pm
[Randy Strome] “I had settled on “use 720 only when shooting for slow motion”. “
- March 28, 2008 at 7:32 pm
What and how you’re shooting can have impact on overall results. If you’re not pushing the codec 1080 may look better than 720.
If I were shooting sports I might choose 720p60 because that kind of motion may well result in more obvious artifacts at 1080.
Note Marco seems to have been testing “extremes” as far as I can tell.
Like anything involving compression, there’s always a juggle. If you’re dealing with fairly low motion I’d bet 1080 will look better in those cases. Shooting fast motion 720 may be better.
Keep in mind that while 720p24 may have the most bits per frame there are potential motion issues incumbent to the frame rate also to take into account.
It’s also why I don’t like compressing video for web at 15fps. Given a common data rate, 15fps may have more data per frame compared to 29.97 but there’s also greater distance between frame motion and of course the codec may have to work harder since the motion distance between frames is greater.
In other words more bits per frame isn’t necessarily better when you’re dealing with moving subjects or moving camera. The answer is . . . it depends.
- April 6, 2008 at 12:35 am
[Craig Seeman] “Note Marco seems to have been testing “extremes” as far as I can tell.”
Yes, this very important to know. I’m *just* testing the codec itself with synthesized sources, which doesn’t not account for any other conversions like RGB (from the imagers) to the Y’CbCr video format and such. And yes, as I mentioned, in most real-world scenarios, these compression artifacts I found wont be found in actual shot footage (but you’d be surprised at some of the stuff I found).
But one thing that wont change is the fact that under the same data-rate of 35 Mb/s, there is more compression in 1080 than 720. Whether or not you can see it may be a moot point, but it’s there.
Me? It depends on (A) the delivery format, (B) the type of production, (C) the type of action being shot or (D) a specific client request that may go against my recommendation of A B or C. :-p
Log in to reply.