Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Quick RED Compression Ratio Comparison

M

Mike P.

Well-known member
Joined
Dec 17, 2010
Messages
4,344
Reaction score
32
Points
48
Location
Toronto & Vancouver

Quick little test comparing 6:1, 12:1, and 18:1... I wanted to see how noticeable the differences would be at the delivery stage (generally compressed and 1080p, not 4k/RAW). Figured there would be some soon-to-be-Scarlet owners (or 32gig SSD owners) that would want to check it out in order to maximize their REDmags. I cover most of the details in the video's description.

Must say I was pleasantly surprised seeing as how we're dealing with approximately ~3gigs/min vs. ~2gigs/min vs. ~1gig/min... Holds up extremely well in the grade, too. Granted there was tons of ambient ~5600k light, which definitely helps, but it's good to see that you aren't sacrificing too much quality whilst literally getting 3x the amount of recording time out of the same REDmag. It certainly doesn't seem to be 1/3rd the quality at the deliverable end (they all look pretty much the same, which I guess is to be expected when they end up being crammed into the same 10mbit h264 codec regardless of the source ratio, but still).
 
If anything, in post, I've noticed that a higher compression ratio = more visible noise. Perhaps the compression ratio needed is situationally dependent? Recording in a darker setting could mean you'll want to bump your compression ratio down?
 
Hi Mike,

Most of the frame is static. I bet you will see more of a difference if there is more motion as a greater percentage of the frame. I had done a quick test of a talking head set up and found 18-1 held up really well but also not much change per frame. Motion like a pan would probably show issues. Thanks for the test.
 
Water and smoke are also good torture tests for compression issues. Another good one is a stadium shot, with lots of little teeny-tiny people in the grandstands. I've seen certain MPEG images just completely fall apart with motion artifacts and dropped frames with complex shots like that.
 
What used to be compression stressing for the R1, would be something like a shot in a forest with high contrast, tons of detail (leaves and branches) and some motion.

The field and the stadium and smoke sounds good, too...
 
You've gotta love the distillery district as a location though. Such an interesting place to shoot!
 
I remember the day they had a casting call for tall skinny dudes when they were shooting the concentration camp scene for the first X-men here. My friend got a part as an extra. Not sure why I didn't??? Oh yeah, I'm not white, tall, or skinny! Ha ha!
 
Gunleik Groven said:
What used to be compression stressing for the R1, would be something like a shot in a forest with high contrast, tons of detail (leaves and branches) and some motion.
Click to expand...

Hi Gunleik,

why do you think it needs motion? RED's compression is strictly Intra.
 
Uli Plank said:
Hi Gunleik,

why do you think it needs motion? RED's compression is strictly Intra.
Click to expand...

Sure! But with tons of small motion in leaves or grass or waves in the sea, the difference in compression between high-detail frames, used to be a stress test. Early on, you could even make the camera go down under such conditions, because it didn't handle them... :) (Looooooong time ago).

I am in no way sure of this (!!!), but I have been wondering that as it is a VBR codec, sometimes the "next" frame could suffer if the "1st" was "hard" to get right...

Haven't run these tests on Epic, and my guess is (again due to all the computing overhead apparantly available inside that box) that it won't be an issue at all! But yoy could trigger t at higher compressionratesm still... as the wavelet "evens out" stuff. Add motion, and that could differ from frame to frame...

Maybe I should try! :)

G

(Just contemplating ways for breaking beauty...)
 
Uli Plank said:
why do you think it needs motion? RED's compression is strictly Intra.
Click to expand...
I could be wrong, but I'd think that if the content of the scene is a lot of little detail that is constantly changing, it'll be easier to spot the compression artifacts as they change from frame to frame. In a still scene with lots of little detail the problems with high compression might not be as apparent, as the artifacts would be similar for each frame shot.
 
After reading everyone's comments, I'm actually kind of pissed at myself for completely gapping with this comparison (the shot has a decent amount of fine detail in the brickwork and the few trees, but because of the lack of movement I can see why the compression really isn't being stressed all that much.)

I will say that in the few shots I have at high compression ratios with movement they still hold up when encoded to a deliverable codec and in the grade (which is really what I wanted to get out of the comparison.) Unfortunately they're standalone shots, so I don't have the same shot at 6:1, 12:1, and/or 18:1. They do exhibit more noise (in the shadows), but again, only at 4k/FullRes RAW; as soon as you go to even 35mbit BluRay the difference is diminished substantially (which, in my opinion, is really good).

Wheat field or forest would be a very interesting real world test... Or, how about cherry blossoms at High Park (*cough*Nick*cough*)? Unfortunately the Epic wasn't mine (one can dream, right!) so I can't follow up with a better comp. A lot of what I shot that day was 12:1, and judging by the footage (at full res) I'd be very comfortable shooting anything at that... Unless low-light/tungsten/high-compression really goes the hell.
 
You must log in or register to reply here.
Back
Top