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 :)

Epic Compression Test Results

Zeb B

Zeb B

Well-known member
Joined
Sep 30, 2011
Messages
4,933
Reaction score
1
Points
0
Location
In The Moment
Is there any science out there or just personal prefs and anecdotal advice? Seems to be a bunch of armchair opinions on this subject without much science.

For example: which produces better 1080 delivered results at 120fps?
5K @ RC13:1
3K @ RC6:1

Would love to see extensive compression testing in a lab environment running from 3:1 to 15:1 / 3K to 5K to make better technical decisions. Anyone know if this exists somewhere?

How about some actual scientific test footage running up the compression ration on something like these sets:

1. Flesh tones in studio lighting
2. Flesh tones in low light (candle and mixed light night scene)
3. Harsh sunshine, high dynamic range landscapes and architecture
4. Fine detail like wires, bricks, cloth patterns etc (zoomed in to see detail)
5. Test colors and patterns in controlled lighting

It would be very useful to see where the point of diminishing returns is on compression and pixels. High Frame Rate and RC is the core of the RED Advantage. I would love to have the facts to know how to best apply the technology.
 
Also wondering but this kind of tests is very demanding to make and to monitor proprely. Compression rate choices are made in many cases based on subjective impressions...

Scientific tests would be helpfull indeed.

Patrick
 
Let's just leave the "subjective" and "application/subject matter specific" discussion out of the thread (it will derail it quickly) and just focus on the what is objectively known about the OP
 
Not speaking technically, don't have an EPIC in front of me to do the tests but...

Wouldn't 3K at 6:1 to better than 13:1 5K? I mean resolution being there but 13:1 is less than the 7:1 which is supposedly the equivalent of REDCode 42 on the R1. Anything extremely less would probably see noticeable differences...

If your shrinking to 1080p, 3K at 6:1 seems like a good choice...

Like I said, non scientific since I don't have an EPIC in front of me but hopefully someone would do these tests, would be nice to see the various jumps in compression rates and which is potentially great for 1080p/ 'good enough' in terms of compression...

Edit:

It's been said that 5:1 versus 3:1, the difference isn't noticeable on first sight...I'm not sure if its one of those things that is 'if your REALLY want' but considering most here shoot at 5:1 since its best for compression versus data storage...the difference between 5:1 to 3:1 then isn't that major, but 7:1 to 13:1 I'm sure is a lot more noticeable...
 
Imran Farouk said:
Not speaking technically, don't have an EPIC in front of me to do the tests but...

Wouldn't 3K at 6:1 to better than 13:1 5K? I mean resolution being there but 13:1 is less than the 7:1 which is supposedly the equivalent of REDCode 42 on the R1. Anything extremely less would probably see noticeable differences...

If your shrinking to 1080p, 3K at 6:1 seems like a good choice...
Click to expand...
Don't forget that when you crop to 3K, you're completely changing the field of view. It's not just a matter of compression. Also, noise is more noticeable and larger in the frame.

I'd be happy to set up a non-scientific scene tomorrow and post the results for those who are interested... I agree that what is "acceptable" compression is largely subjective, but you could at least see the differences and make those choices yourself.
 
Häakon said:
Don't forget that when you crop to 3K, you're completely changing the field of view. It's not just a matter of compression. Also, noise is more noticeable and larger in the frame.

I'd be happy to set up a non-scientific scene tomorrow and post the results for those who are interested... I agree that what is "acceptable" compression is largely subjective, but you could at least see the differences and make those choices yourself.
Click to expand...

I would sign-on Pay for View for this if done right! Damn, my "girlfriend" costs me $3.95/min so this would be worth . . . like maybe $4.95/min!!

Thanks for the effort - you have the latest Alpha build on your Emachine?
 
Haha, well even "done right" is going to differ from person to person. Do you want downloadable 1080p prores files, or straight up R3Ds? Personally I feel like it would be rather useful to see what the extreme levels of compression do in a "final output" sort of scenario, but if you want to mess with RAW files I'm happy to post those instead. Just might need some help with a server if that's the case.

My plan would be to set up both an indoor lit scene and an outdoor high-detail natural light scene, with a model and subtle movement; record one take at each of the various levels of compression and share the results. Obviously compression works in all kinds of different ways; this isn't going to scratch the surface of all of the different ways one could shoot. Shooting extreme motion from a vehicle mount, for example, would probably really stress it compared to a locked-off shot. This "test" would be a more "normal" (if that exists) type of shot that might be applicable to narrative work.

Thoughts appreciated as I'll set this up for tomorrow.
 
ProRes should be best - like you say. Amazing that this is the first time this has been done!!
 
This is going to be really really challenging based on use case. I just got Epic 3:1 (uncompressed) greenscreen footage for the first time on this last project I worked on and for keying the uncompressed keys terribly. Like really really bad. The blue is so noisy in the bluescreen that the motion blur and edges are sparkly. Worse than film. But I apply a little wavelet noise reduction to it and it produces nice clean edges and semi-transparency with minimal ringing or other artifacts. Now when I got to the hair the wavelets introduced little artifacts the size of hair strands and keyed pretty poorly so I went back to the uncompressed source and pulled a key for the hair from that using the red channel which had almost no noise.

So in that case my answer (within the same frame) would have been "No you don't need uncompressed, and yes it helps to have uncompressed". :D

Which is a long way of saying "I would say that there are as many answers to this as questions" so I would encourage you to find as specific of compression situations as possible to get the best answers. A well formed inquiry will often produce the best studies.
 
Well since this test seems to have never been done publicly at any level - whatever Haakon is offering is a great place to start.
 
I think there are good reasons that this kind of test isn't widely publicised - it can be ugly; akin to crash testing a car. The results aren't necessarily for general consumption and could easily be misinterpreted or taken out of context.

In view of many recent threads reminding us that we're in Red's house when posting here, I'd suggest running any compression-busting test results past one of the RED dudes here present before splurging them online.

I've got plenty of archived tests on this subject, but if I were to share them with anyone at all it would be in a 'facilitated' way, so I can present the information and circumstances fairly. I'm actually not sure thats possible here. I know this isn't quite in the sharing spirit, but particularly with regard to what amounts to destructive testing of technology, the only tests worth anything are your own tests.
 
It's quite a few parameters to think of.
1. FOV change
2. DOF change to compensate for #1.
3. amount of data per frame.
4. what is shot, i.e what is infront of the camera.
5. things i have not thought of.

So there is more than one answer to the question.

Basically as long as you max out the data rate you are somewhat good.
Then if you shot 5k high speed wide lens moving camera with a lot of detail then a smaller sensor/ less pixels saved with lower compression would be a better alternative.

if you shoot 5k with alot of out of areas and low detail then you would benefit to stay on a bigger sensor.

but basically you can say they are all as bad/good since you have a maximum data output and when hitting that bar quality per frame is limited.
 
There are literally a billion ways one could "test" the compression, but the only thing that matters is how you are personally shooting at any given moment. My setup is not going to attempt to address all of the different "what ifs," but to show a very general example of how each of the compression settings might look in a given scene. That's essentially what I gathered the OP was looking for. Don't read into the footage more than that. :)
 
I'd definitely expect 5k 13:1 (~50MBps) to produce more pleasing images than 3k 6:1 (~32MBps), partly because the 5k would be nearly twice the bitrate and partly because my understanding is that compression tends to become more efficient as resolution increases. Also wavelet tends to have the effect of softening as it compresses, but the likelihood of 5k softening to look like even 3k uncompressed seems very low to me at ratios no higher than 18:1 (higher is not even available). I would even guess that 5k 18:1 would look better than even 3k 3:1 in most ways even though its bitrate is similar to that of 3k 5:1.
 
Hardly scientific, but Jim Hare shot some comparative shots that we were able to pixel peep at our Redusers meeting last week. To cut the long story short, if you are going to HD delivery, RED RAW 15:1 at 5k is as clean as it gets. You can't see any noise. You can see the noise at full 5k right until about 6:1 and it becomes academic for lower compression ratios.

The compression noise is most evident in shadows. A difficult scene with lots of shadows and detail (cat's fur) was chosen. ISO 800.

If you shoot outdoors or good lighting...you can get away with much higher compression before you can see compression artefacts.
 
Noise will be more of a problem if you will be doing pushed grades using secondary qualifiers.

That said, moving fields of grass are great compression-scheme breakers, as are long-shots of moving water.

Great thread.
 
John Marchant said:
I know this isn't quite in the sharing spirit, but particularly with regard to what amounts to destructive testing of technology, the only tests worth anything are your own tests.
Click to expand...

Very well said. I'm looking forward to seeing Häakon's tests tomorrow, very grateful for the work put into them. Very, very, very much looking forward to taking a Scarlet on location scouts and lighting tests, into the snow and with strange motion, feeling out the limits of the higher compression ratios.
 
Last edited:
Evening REDs,
To the point of RED being concerned about compression tests being made public: Really? The thrust of the experiment is to find the sweet spot of compression vs bandwidth - not a competitive shoot out (which happens every year over at Zacuto) or anything with a narrative or agenda. I really don't think there is any need to get approval from RED to post results from independent research. There's nothing preventing a moron from renting a Epic, running a "test" and posting it to YouTube.

To the point of "only your own testing is relevant to what you do": Everything technical I know I have learned from others - why throw that successful model out now? Should we have our children home-school themselves?

If this experiment holds no value to you personally please don't comment. I appreciate the hard work and efforts that go into research for basic RED community knowledge and respectfully withhold comment unless I can provide a better solution to a problem. Let's provide positive support for efforts that can bring us knowledge.

For those posting test scenerio suggestions (grass/waves) or known challenges (green screen) etc I thank you for your input. My goal is to have a clearing house knowledge on a real basic setting we all make a choice on every time boot up the Epic.

Cheers!
 
You must log in or register to reply here.
Back
Top