EPIC Noise/Grain

[Lee Dashiell]

Too big to upload, even when zipped. I'll be glad to email it if someone wants to take a look. I have started a Support ticket but haven't heard anything yet.

[Lee Dashiell]

My EPIC boots up and records like this every once in a while as well.

[John Marchant]

Lee would you mind if I post the R3D snapshot you sent me?

I don't want to come off as being mean, but looking at the raw file I can see how it would end up noisy. I'll elaborate if you like, although I'm sure Red support will have an opinion too :)

[Deanan]

My EPIC boots up and records like this every once in a while as well.

Can you send me an r3d still and a log file from the camera?

deanan at red dot com

Thanks.

[Lee Dashiell]

Lee would you mind if I post the R3D snapshot you sent me?

I don't want to come off as being mean, but looking at the raw file I can see how it would end up noisy. I'll elaborate if you like, although I'm sure Red support will have an opinion too :)

HAHA!! I've got pretty thick skin so any suggestion is welcome. What I'm saying is that I've shot in lighting conditions much worse than these with the same glass on RED ONE MX and I've always had an image that I could correct without the noise that I'm getting from EPIC. It just comes as a bit of a surprise.

[Lee Dashiell]

Can you send me an r3d still and a log file from the camera?

deanan at red dot com

Thanks.

Will do Deanan. Thanks!

[John Marchant]

OK Lee, don't want to cause offense - its easily done online :)

Lee's R3D snapshot is at : http://www.kippertie.com/red/leedashiell.zip

My take on it is this. Three main factors - lower than expected resolution, lower than usual colour temperature, low exposure.

First, its a 3k clip. That's about 5mp not 14mp of resolution, so any noise is much chunkier in a 1080p transcode.

Second, the colour temp going by the guy's shirt is around 2200, which makes life hard for the sensor without colour filtration in front of the lens to help you out (assuming sufficient light to do so)

Third, if you check out an approximation of what the sensor is seeing in RCX by rating at 320 ISO rather than 1000, removing the little exposure boost in the metadata too, and placing colour temp at 5k, you'll see that even the brightest points of light aren't illuminating the blue channel more than 33%, the blue level in the white shirt is way down around 15%. The blacks - well lets say as a result they're way down close to the noise floor of the camera, hence all your dancing noise.

I'd say if you had loads of light, then there's been a misunderstanding of the histogram due to camera meta settings. The colour temp evil, and wasn't helping you though, and balancing that in front of the lens would mean your exposure would have been pushed nearer to the sweet spot on the sensor, even with the slightly beyond-safe exposure.

Up side- I'd say with a willingness to beat down the blacks with the shadow slider, some compromise of colour temo to the warm side and some judicious s-curving you will probably hide the issue.

Set RAW check on a friendly button and remember the histogram behaves according to your meta settings at all other times :)

[Tom Gough]

Hi Lee,

I have just looked at the R3D and agree with John's analysis.

The noise on his jacket dances around like crazy!! Can't deliver this to my client.

In my view it can be saved in RCX... I have reset all the original metadata and (roughly) tweaked it - once downsampled to HD it should be usable.

I have zipped a snapshot and an RMD with my tweaks.

http://www.flow-mps.com/uploads.html

Hope it helps!

All the best,

[David Battistella]

I got this, out of that, in about 20 seconds...

No problem with this r3d at all...

Just by using RCX...

[John Marchant]

Nice one David. Long and short of it - thank RED you underexposed that shot on an EPIC. Just kidding Lee, you did say you've got a thick skin ;)