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 :)

Exposure / Noise Epic-W problems - help is welcome

Raphael Maier

Raphael Maier

Member
Joined
May 7, 2016
Messages
23
Reaction score
0
Points
0
Location
Germany
Hey guys!

I've read many tips and tricks, watched many videos all around the RED universe and lastly did some proper blackshading sessions.
But I am still getting that nitty gritty noise in my footage. Maybe I just don't get the concept behind all that - maybe I am an idiot. Help would be very welcome.

This footage (.R3D and .mp4 included) -> DROPBOX was shot with EPIC-W (v7.0.1) ISO 800, 25 fps , 180°/50 shutter, manual blackshade for 1/50, 40°C sensor, IPP2 in use, Sigma 18-35 1.8, 6K WS 10:1

Some of the low light tests feature more challenging light situations with better results for example:

https://vimeo.com/204989672
https://vimeo.com/227493065
https://vimeo.com/189478583
https://vimeo.com/188865905


I am very looking forward to some tips from you! Maybe I am just too dumb. Thanks in advance. :)

(yes I know that shooting in 8K gives better results - but for some projects I can't shoot 8K)
 
Last edited:
Hey buddy,

I'm the one who filmed the first video (stater bros) test. For some reason, I recently upgraded to v7.0.1 and I have realized my blacks are a little more noisy than they used to be and I don't know why.


:/
 
How did you go about your blackshade calibration? Did you bring the camera/sensor up to your target temperature before the blackshade? Or did you do a blackshade immediately after booting up your camera (you shouldn't do)?

Were those settings more or less the same that you blaskshaded to? (sensor/core temps and exposure?) Within 15°C or 1/2 of a second above/below your shutter's exposure time?

I've also found that shooting lower compressions give me a bit better noise performance (try 6:1-8:1)


RED has some documentation which would better explain Black Shading and shooting in low light:
http://www.red.com/learn/red-101/black-shading-calibration
https://www.youtube.com/watch?v=WiM14Hni3nI&t=107s

Hope that helps! Think everyone at first gets confused/unsure about blackshading, what it is, how to optimally do it for your shooting style/conditions, etc.
 
Rich Miro said:
Hey buddy,

I'm the one who filmed the first video (stater bros) test. For some reason, I recently upgraded to v7.0.1 and I have realized my blacks are a little more noisy than they used to be and I don't know why.

:/
Click to expand...

that doesn't sound very good :( btw the test looks very nice!

Chris Kennedy said:
How did you go about your blackshade calibration? Did you bring the camera/sensor up to your target temperature before the blackshade? Or did you do a blackshade immediately after booting up your camera (you shouldn't do)?

Were those settings more or less the same that you blaskshaded to? (sensor/core temps and exposure?) Within 15°C or 1/2 of a second above/below your shutter's exposure time?

I've also found that shooting lower compressions give me a bit better noise performance (try 6:1-8:1)


RED has some documentation which would better explain Black Shading and shooting in low light:
http://www.red.com/learn/red-101/black-shading-calibration
https://www.youtube.com/watch?v=WiM14Hni3nI&t=107s

Hope that helps! Think everyone at first gets confused/unsure about blackshading, what it is, how to optimally do it for your shooting style/conditions, etc.
Click to expand...

Yes camera and sensor were at target temp. The settings were absolutely the same as which I was blackshading to. I've also got some different calibration maps for different shutter times. Same result.
I`ve checked the video and the RED 101 a several times. I just can't get it right... :(
 
Yeah you shouldn't be filming at 10:1 on 6k because it's cropped. I would shoot the lowest possible compression as you start to crop down your sensor. I haven't gotten the extreme noise you're looking at, but I think I might have noticed a little bit more sense the new upgrade.. Would like to see a test done or if someone else is having the same issues.

:)
 
Don't forget about the new ISO changes in recent firmware (to match light meters). Don't know if you have it enabled or not, but if enabled it will give you noise as a one stop higher ISO. Looking at your mp4 it seems consistent to what I get on ISO 1600 (old calibration, I'm still on V6 firmware).
Try shooting ISO 400 and see if it matches closely the exemples you posted.
 
I'm no colorist but looking at the r3d it looks a bit underexposed-there still was quite a bit of room in the highlights. Most digital sensors strive for light and especially when your cropping the sensor you'd risk noisier images-especially at a high compression like you were shooting. Like in the youtube video I linked to said, consider exposing to a brighter exposure and shooting at a lower ISO to create the look your going for.

Try again with more lights, less compression, and in 8k (just for a test since you apparently won't shoot 8k for some projects).

If you absolutely can't shoot in 8k for certain projects have you considered shooting in 8k and a 4k proxy and editing with the proxy instead of the r3d?
 
Rich Miro said:
Yeah you shouldn't be filming at 10:1 on 6k because it's cropped. I would shoot the lowest possible compression as you start to crop down your sensor. I haven't gotten the extreme noise you're looking at, but I think I might have noticed a little bit more sense the new upgrade.. Would like to see a test done or if someone else is having the same issues.

:)
Click to expand...

What does compression have to do with noise?
 
Jacek Zakowicz said:
What does compression have to do with noise?
Click to expand...

At higher ISO ranges in particular and higher compression ratios it can "enhance the energy" of it on a visual level as it encodes chunkier.

That said, you shouldn't have any real issues in the 5:1-12:1 range most of the time if you're oversampled for sure. Going 4K for 4K or even 5K for 4K, I'd recommend as low as you can go.
 
It's very different to compare your untouched raw 6K 10:1 clip to a graded (except for one) 8K clip of unknown compression downscaled to 4K and then compressed to hell on Vimeo.

Also, ISO 800 on the newest firmware (without intervening in the settings) is one stop brighter on screen than before, making noise one stop more apparent. So what you're shooting today is the same brightness as yesterday's 1600.

At 1:1 pixels under similar conditions your camera seems pretty much in line with my camera.

For dead clean low key stuff, I shoot ISO 250-320 on the current ISO calibration. Otherwise, I'll shoot 500-800 for most other scenarios.

I typically run a temporal denoise and then re-grain anyway. I like texture so that my eyes have something to latch onto, but prefer it to be not so digital-ish. Your clip cleans up just fine in this regard.

Edit:
Here's a 100% punch-in on a 4K project what noise reduction and adding grain back in can do.
28345504979_eae07468da_o.png
 
Last edited:
Fabricio Morato said:
Don't forget about the new ISO changes in recent firmware (to match light meters). Don't know if you have it enabled or not, but if enabled it will give you noise as a one stop higher ISO. Looking at your mp4 it seems consistent to what I get on ISO 1600 (old calibration, I'm still on V6 firmware).
Try shooting ISO 400 and see if it matches closely the exemples you posted.
Click to expand...

omg. I'm retarded. That's totally the issue. I shot that test footage before the iso re-calibration. thanks man.
 
Totally idiot question. Is there any way I can revert ISO readings to old firmware levels? Having been so many years used to Red's interpretation that am struggling a bit to get a hang on uncontrollable lighting situations.
 
Sergio Perez said:
Totally idiot question. Is there any way I can revert ISO readings to old firmware levels? Having been so many years used to Red's interpretation that am struggling a bit to get a hang on uncontrollable lighting situations.
Click to expand...

Yes you can do that: MENU > IMAGE > ISO > use updated ISO calibration
 
Sergio Perez said:
Totally idiot question. Is there any way I can revert ISO readings to old firmware levels? Having been so many years used to Red's interpretation that am struggling a bit to get a hang on uncontrollable lighting situations.
Click to expand...

The change was done to match light meters better (as well as other cameras). Monstro-based cameras released with this setting enabled by default. It's clear that's the direction RED is heading. So, personally, I'd try to get used to the change if you can.

Easiest way to remember it is whatever you're used to shooting in old firmware, cut that number in half, i.e., if you liked shooting 800, shoot at 400 now.

If not, Raphael posted how to change the setting in the menus.
 
Hi Raphael,

My experience with the Helium sensor is not deep but the Dragon sensor is an open book for me. In all the years one thing was obvious: Red's ISO missed lightmeters by about good 1 stop. E.g., ISO set to 800 in camera and likewise the lightmeter you get a 2.8 with the lightmeter but the camera's tools / picture brightness suggest a 4. I never understood why Red didn't close that gap (you don't have that with Alexa).

For juicy pictures and no visible noise Dragon needs to be set tp ISO 320. The same applies to Helium but with just about 1 stop more sensitivity. I find ISO 800 workable (though 640 better) but not the suggested 1,600. As mentioned above Red seemed to understand that this was an issue and re-calibarted their ISO apparently. So the new default is ISO 800 and not a underexposure prone 1,600. Please validate that for yourself.

This said, there are other things than mere ISO that may influence the noise level considerably. First off, in different environments the same ISO does not lead to equal results regarding noise. Exposure varies from environment to environment. In old film times it was a good habit to rate stock ½ - 1 stop slower for night exteriors. The same applies to digital sensors. But compared to film-stock which allows you to choose between daylight or tungsten sensitivity SI based sensors like Red's sensors need colour temperatures around daylight to feed the the channels evenly. Not doing so will will starve a channel and lead to noise. In the end a good exposure is all about feeding the sensor with a sufficient amount of light. Please bare in mind that open the lens by 1 stop is doubling the amount of light that hits the sensor. Correct exposure matters.

This is not the case in your example. Assuming the walls are painted in bright white and lit quite evenly the picture should be pretty bright, which it is not. Please have a look at the histogram and compare the position of the assumed whites in the picture with the histograms reality. I think you missed the right exposure by 2-3 stops. Also the red channel is starving considerably due to the tungsten lighting which is costing you about 1 stop.


I would go outside and shoot in a typical every day environment such as a street in bright daylight. Get somebody into the picture for skin tones and compose for some darker parts in the picture. Dial in ISO 800. Now shoot the same scene with different stops. 3 under and 3 over. Check out the results. You should get noise when you are underexposing deliberately and are pushing the picture by ISO for a correct picture. But the picture should look clean in the default exposure that looks right on the screen. If this is not the case contact RED.

I once had a broken Dragon sensor which exhibited much to much noise and showed a subtle grid. But the latter isn't case in your example. Black shading is an issue with Red cameras but only if you don't do it right, forgetting to put on a lens cap etc...


Hans
 
Aaron Lochert said:
Here's a 1:1 crop comparison of what noise reduction and adding grain back in can do.
Click to expand...

Curious whether you tried turned DEB on (which really helps coloured noise) and left it at that. Eyeballing that it looks like it would do the same as your denoise/regrain going on (of course you can grain to any kind of size or stock)

cheers
Paul
 
paulcurtis said:
Curious whether you tried turned DEB on (which really helps coloured noise) and left it at that. Eyeballing that it looks like it would do the same as your denoise/regrain going on (of course you can grain to any kind of size or stock)

cheers
Paul
Click to expand...

Sure. I personally have never been a fan of D.E.B. Notice the green splotches that happen in the corner of the room, especially as compared to the denoised+regrained still. I also feel like those two corners on that wall are stronger, where they are somewhat undefined with D.E.B.

25270539097_5ac7f2816c_o.png
 
You must log in or register to reply here.
Back
Top