Welcome to our community

Be a part of something great, join today!

Register 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

A

Adam Rosenbloom

Active member
Joined
Oct 12, 2011
Messages
29
Reaction score
0
Points
0
Location
Washington, D.C.
I did a compression test on our Epic today, and I wanted to share my results, and see if this seemed right to all of you.

What I found, in short, was that on difficult scenes, 3:1 was great, 5:1 was good enough, and after that blocky compression started to appear.

Here is a link to download and below is a description of the shoot:
Tiffs: http://www.filesanywhere.com/fs/v.aspx?v=8a70628b5866737c6b9a
R3D: https://www.filesanywhere.com/fs/v.a...8b5d6374b6a5ad

Scene: I shot a streetscape filled with trees, including some foreground bark, and a more isolated shot of the trees. I was hoping for a sunny day, because I read that some others had compression trouble in high contrast scenes. However, from seeing the results, a slightly overcast day worked fine.

Lens: I used a Zeiss cp.2 35mm lens. The aperture remained small (about f/15 to f/16) because I wanted a sharp image that would really test the compression. I did not use an ND filter.

Camera Settings: The ISO was 640. The frame size was 4K HD (this is just because it's how we intend to shoot an upcoming project). The framerate was 29.97 and I used a 180 degree shutter angle. All other settings that I can think of were set to their defaults (except white balance).

Notes: All of the jpegs in the link above were made from the raw tiffs out of RedCine X Pro, and I've verified that there is no visible difference between the jpegs and tiffs. I applied no post-processing as you will see, except I've blurred some signs and license plates in the 1st shot (sorry, I just wanted to be safe). I want you to know also that even though the jpegs are all the first frame of the video, I played all of them and the frames in the rest of each shot look the same.

The compression really showed the in bark. I know this is a rough test to put a camera through, but we intend to shoot some scenes like this in the future.

I would love to hear any of your analysis. Are these results expected? Did I do something wrong in my test?

Thank you in advance!

Adam
 
Last edited:
Shooting with such a small aperture isn't a good idea (IMO). You're reducing the light transmission to the sensor - you'll also loose performance from the lenses themselves at such an extreme.

You've starved the sensor of some light, so you'll add to any compression artifacts. You'd have been better shooting at 800 (or 320 if you prefer), T5-T8 with ND, so the sensor get's a nice even spread of light.

I'm sure others will chime in with their thoughts too.

BTW - what did they look like on motion? Was the camera locked off? - Just interested :)
 
Last edited:
From this example I think I would be fine with up to 8:1 but probably not higher. Especially with all the detail in these shots.

5:1 seems like the most logical and practical choice based on these shots.
 
This is one of the better compression tests I have seen. The deep focus does of course not starve your sensor for light, and exposing at 640 only makes you expose higher, if anything...

The leaves is where I have most interesting results.

I'd love to see the RAW files from this, and would suggest you sent single RAW frames to RED for analysis.

Thanks a lot for the effort!
 
Thank you all very much for your helpful responses.

I guess I did go a little too far with the aperture. I probably won't be able to do another test today, but maybe over the weekend.

In motion, the video at 3:1 really looks nice, clean with a subtle film-like grain. At 6:1 and above the grain turns to blockiness and is just ugly noise.

I would love to post the RAW R3D files, but they're obviously huge. The 3:1 shot of the street is 2.43 GB, and it's only 25 seconds long! Do you think trimming them in REDCINE X to 3 seconds each would be work? Any ideas on which particular clips you would like to see? I'll look for a way to transmit them.

When you say send the raw R3D snapshots to RED, do you mean just email a link to their tech support?

Thank you all very much again,

Adam
 
Just need an r3d snapshot from rcx.
And frankly I don't need them, but would urge you to send them to RED directly.

The aperture is not what is happening here, and this is exactly what needs to be adressed.

Goodone!

G
 
EDIT - DOWNLOAD LINK IS FIXED.

So here's a link to the R3D snapshots from shot 2: https://www.filesanywhere.com/fs/v.aspx?v=8a70628b5d6374b6a5ad (I have added these to the original post)

If it's okay I'd rather not post the RAW of the first shot because of license plates and signs, but if it will really help, I can upload them as well.
I am sending all of my R3D snapshots to Red.

Also, I forgot to mention that yes, the shots were locked down.

Thanks!

Adam
 
Last edited:
Andy White said:
Shooting with such a small aperture isn't a good idea (IMO). You're reducing the light transmission to the sensor - you'll also loose performance from the lenses themselves at such an extreme.

You've starved the sensor of some light, so you'll add to any compression artifacts. You'd have been better shooting at 800 (or 320 if you prefer), T5-T8 with ND, so the sensor get's a nice even spread of light.
Click to expand...

Andy, Your respose here, stated in a very authoritative tone, is total bull-shit. Light is light. A small iris, used to create deep focus, does not in any way starve the sensor of light if the exposure is right, its right!. Get a light meter, learn about proper exposures, before you spout off non-sense again!
You were somewhat correct in that extremely small iris openings can intruduce lens artifacting as a result of diffraction. But the degree of that is lens-to-lens dependent.
 
Paul Nordin said:
Andy, Your respose here, stated in a very authoritative tone, is total bull-shit. Light is light. A small iris, used to create deep focus, does not in any way starve the sensor of light if the exposure is right, its right!. Get a light meter, learn about proper exposures, before you spout off non-sense again!
You were somewhat correct in that extremely small iris openings can intruduce lens artifacting as a result of diffraction. But the degree of that is lens-to-lens dependent.
Click to expand...

Hi Paul. No problem :) I did say IMO in the post, but thanks for the authoritative and absolute response in your reply.
 
Sorry for the tone of my post, Andy. I could have been more politically correct, a la Gunliek's reply. Exposure as relates to tstop, and theoretically to fstop, is not an opinion. Its a mathematical fact. Nuf said
 
ah...

I should have added that the deep focus and richness of detail in the whole image is exactly what makes this a good test for the compression.

The high stop, is what gives us the info, intended or not.
 
Glad to know the aperture seems to not be an issue! That was my intention to keep it small so as to have a wide depth of field.

I did leave the iris as open as it needed to be for proper exposure, though, using the histogram. I think I irised until one of the clipping traffic lights on the LCD came on, then I backed off until they were gone. Also, there was no noise in the camera's raw noise level bar and no clippping in the raw clip meter.

I sent the R3D snapshots to RED tech support, so we'll see if I get a response (other than the automated reply). If you have any advice on something else I should be doing, please let me know.

Thanks!

Adam
 
Small update: a tech support rep got back to me and said he will review the snapshots. He then asked if I had a chance to upgrade to firmware 3.0.6 and run my test again. So this is probably a good time to mention that I am on 3.0.0, the latest release version, because I figured it's still a recent build, and why risk installing beta software which is not recommended for actual use?

I'll see about running a test on the beta software when I have a chance.
 
I had the same thing with a ticket, tech supp asked me to move to 2.0.8 (the beta at the time) to see if the errors could be repeated (they could :) I only moved to 3.0.0 myself recently (2.0.8 was working OK for me, even though it was beta). First thing I noticed with 3 was the gui reporting error (it'll mis-report in the system status/system configuration on your Epic - RG3 and it's reported as Redlogfilm colorspace).

If you're doing the test again, are you looking to replicate or expand your original parameters?
 
If I do this again to test the firmware update as Red suggests, I believe I should aim to replicate the exact conditions and setup as before, right down to the overcast weather. That way, we'll know if the firmware update did anything. I mean, right?

If it's okay, I was thinking of just doing the shot of the trees (shot 2), rather than worry about blurring signs and license plates.

But if I were to expand my parameters, do you have any suggestions for one or two setups you would like to see? Ideally, the only thing I would be changing for each recording in each setup is the compression ratio.
 
Hi Adam :)

You're right in replicating as much as you can, IMO, as it'll give an excellent compare to the original.

Are overcast conditions common where you are? (it feels like we're getting ready for Noah's Ark here in my part of the UK... it's been raining off and on everyday, and we've had the wettest April for over 100 years), so it would be a PITA to replicate something :)

Shot 02 (trees) will save the hassle blurring plates etc - but I've seen lots of tests were people haven't bothered doing this, so it's nice that you're aware of it.

In terms of other stuff - maybe a pan (repeated at the various compression ratios) - ie motion of the camera itself vs motion within the frame? Perhaps an ND'd version (different exposure)... really on that, it depends if the shots you're testing for will be this way - if not, deep focus & lock off it is :)

Just a thought... the new firmware update is due imminently (saw mention from Jarred somewhere) - so it might be wise holding off until then? Otherwise you might get caught in a groundhog day of testing loops :smile:
 
Good call on the new firmware update. I think I'll hold off in that case (gives me an excuse to stall!! hehe).

I'm in the Washington D.C. area, so the weather can be pretty erratic (yesterday beautiful, today blech), but an overcast day is not hard to find. It wasn't a dark day when I did my test, it's just the sun wasn't shining.

Thanks for your suggestions, and I'll see what I can do!

Adam
 
Adam Rosenbloom said:
What I found, in short, was that on difficult scenes, 3:1 was great, 5:1 was good enough, and after that blocky compression started to appear.
Click to expand...
This is pretty much what we've told our post clients. This is not a Red problem -- it's the same with any JPEG2000-based imagery. I think 5:1 is a realistic limit for feature projects.

I've had a gut feeling for a few years that this is kind of a "magic ratio" for compression. This is also the compression rate of standard-def Digital Betacam, which (for SD) looks very good and holds up for many, many generations. HDCam-SR also uses about 5:1 compression, and likewise works fine for broadcast. In audio, 320kbps sound files sound very good, and are about 1/4th the size of an uncompressed WAV file.

Any more than that, you can see, hear, and/or measure the compression artifacts. They aren't always deal-breakers, but they are there.
 
Marc Wielage said:
This is pretty much what we've told our post clients. This is not a Red problem -- it's the same with any JPEG2000-based imagery. I think 5:1 is a realistic limit for feature projects.

I've had a gut feeling for a few years that this is kind of a "magic ratio" for compression. This is also the compression rate of standard-def Digital Betacam, which (for SD) looks very good and holds up for many, many generations. HDCam-SR also uses about 5:1 compression, and likewise works fine for broadcast. In audio, 320kbps sound files sound very good, and are about 1/4th the size of an uncompressed WAV file.

Any more than that, you can see, hear, and/or measure the compression artifacts. They aren't always deal-breakers, but they are there.
Click to expand...

Interesting points of thought.
 
You must log in or register to reply here.
Back
Top