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 Noise/Grain

Tom Gough said:
Hi Lee,

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

In my view it can be saved... 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.
Click to expand...

I don't think the OP's question was "Can it be saved?" (He's already running it through Neat Video noise reduction.) Though I agree with John's analysis too.

But I thought he was asking why it happened in the first place. I really do suspect something like the scenario I described above, where the xmas bulbs at the left of the frame gave him clipping in the histogram and fooled him into underexposing. Not a camera issue, but an operator error. It's easy to see how that could happen. They're not even in the room, and there's the white shirt and flags to protect from clipping. Who'd notice the xmas bulbs outside the window?
 
I have seen some tests Jim Hare shot and it was clear that compression was adding noise. Same shot was recorded at varying compression rates and there was no doubt the cleanest blacks were at 3:1. The differences on a normally exposed shot were hard to see unti you zoomed in. But if you window the sensor, underexpose or push the sensor hard high compression will only aggravate the noise.
 
3k, very warm colour temp, underexposure, all of these are bad for the picture inherently, but Tom's right also that the same shot at a better compression rate than 10:1 would look better.

I feel that the EPIC's redcode can sometimes have a hard time compressing the inherently random noise down in the shadows. If you go pixel peeping you'd be surprised how often it occurs, but critically it usually occurs NOT to the detriment of the visual. Sometimes a combination of other bad factors like this will put the spotlight on the compression in a way that isn't flattering.

Always more to learn to get the best from even the most forgiving of systems :)
 
Thanks for the help John. I agree with everything you've pointed out. I wasn't in a situation to bring in additional lighting here and the lens was wide open so I should have lowered the compression. I shot at 3K which is very unusual for me but I thought it would be better to go ahead and push in to what would be my widest shot on the sensor. I was on a 200mm lens at 2.8. I was able to correct the image well enough to make it work but I still don't recall my RED One MX pushing this much noise into the image. I think that the 2200 kelvin may have been what hurt me the most, it's hard to say.

Thanks again for the help.
 
Jason Comparetto said:
It means he's shooting it as if his ISO 200. I thought about doing that, too. But then I realized I just paid $40 grand for a camera that claimed to shoot at ISO 800. I stood in denial. :-)
Click to expand...

well, everything is calibrated/zeroed at 320ISO. That is not to say te cam is 320, but that you still can trust your meters down to 320.

You lose all highlight headroom that way, though
 
@Lee

Lee,

If you would like me to send you a copy of the snapshot with the RDM I created, I would be happy to help you out by e-mailing it to you.

Pushing the ISO to 1000 is always going to increase noise in the black areas, but with a few easy post strategies you can get a really good result from this footage, probably without using NEat video.

David
 
David Battistella said:
Highlight headroom can always be preserved with the iris ring on the lens...:)


David
Click to expand...

THAT depends highly on shooting conditions. That's basically why it is good to keep some headroom.
The other reason (on R1 M/MX and just about any other camera I have tested BUT the epic) is that sensors seems to start to behave quite flimsy and un-nice in the last stop or two before clip.
Even though their mathematical/theoretical precission should be at the best at those last stops.

G
 
Gunleik Groven said:
THAT depends highly on shooting conditions. That's basically why it is good to keep some headroom.
The other reason (on R1 M/MX and just about any other camera I have tested BUT the epic) is that sensors seems to start to behave quite flimsy and un-nice in the last stop or two before clip.
Even though their mathematical/theoretical precission should be at the best at those last stops.

G
Click to expand...

Agree. But it's parlour tricks. ISO, HEadroom, whatever. Knowing how and where the sensor reacts is important, but at the end of the day, it's how much (or little) light you let hit the sensor and that is always controlled by aperture, nothing else.

David
 
Ouch David... :)

I regularilly controll it with filters, because I want to keep the aperture at a given stop to controll DOF, be it deep or shallow.
If I shoot for stills, I use the shutter quite a bit to keep the aperture.

Only when desperate, I use the aperture to controll exposure (by either shutting down more or opening up more than I would like to for compositional reasons).

Aperture is part of image carracteristics. Filters and shutter are parts of light/exposure controll.

Just my take on that. I think both our approaches are relevant, but aparture controlls a lot more than exposure...

SOrry to derail the OT... :)
 
Jason Comparetto said:
@Gunleik -- Are you saying that I should set the camera at 320ISO when I shoot instead of 800ISO?
Click to expand...

No.

What I am saying is that 800 @ daylight you have 1,25 stops highlight headroom, which means that if you overexpose slightly, you can recover that in post through either lowering the ISO in post OR grade from redlogFilm, which recovers the highlights in itself.

There is usually never a reason to lower the ISO on set from 800 and NEVER a good reason to set the ISO lower than 640 at tungsten/320 at daylight.
 
Gunleik Groven said:
No.

What I am saying is that 800 @ daylight you have 1,25 stops highlight headroom, which means that if you overexpose slightly, you can recover that in post through either lowering the ISO in post OR grade from redlogFilm, which recovers the highlights in itself.

There is usually never a reason to lower the ISO on set from 800 and NEVER a good reason to set the ISO lower than 640 at tungsten/320 at daylight.
Click to expand...


We are of two minds on this but I do not want to derail the thread either.

Obviously filters come into it and shutter speeds (electronic at that)...

My point was more about how much light you let hit the sensor.

As this relates to the original post.

Let's say you rated at 800 ISO, and have your camera set there and you open to f4 and it looks good. Well you are underexposing (or getting 1.25 stops of headroom) as you put it. In this case it would have been better to open up to 2.8 (gain the stop) and have more light hit the sensor (rather than starve it one stop) in a tungsten situation.

I know there are many variables, but on shoots like the one in this thread, you need to do two things.

1. get more light on there.
2. Use a faster lens.

That's it.

David
 
We do agree that up'ing the ISO does no-good :)

There is some issues with full exposure at tungsten if you WB to 3200, as you risk putting a cap on the RED channel and get some unwanted artifacts from that.
But we are into details very far off the OP, I agree... :)

The OPoster just needed more light hitting the sensor...
Agreed on all points as to that.
And it gets even worse with the crop-factor as to noise.

But this seems to take some time to get aquianted to, so I wouldn't really call it a mistake, more like part of the process of gaining experience... :)
 
I agree Gunleik.

There are no mistakes, just best practices.

If I was in Lee's shoes I might have.

1. Captured 4khd over 3k ( for less noise and punch ins later if needed)
2. Captured at 8:1 over 10:1
3. Tried to get a bit more light in there.
4. A 200mm F2.8 is pretty fast for a long lens.

Boosting ISO in post is a bit like using db gain on a traditional ENG camera. Any amount you push it will increase noise and noise is often most visible in the blacks.

All of this depends on the situation lee was facing and the decisions he has to make to do the shoot. Let's say 3k might have been chosen to save disk space. I don't know. But it's best to know how the camera will react by testing.

David.
 
David Battistella said:
@Lee

Lee,

If you would like me to send you a copy of the snapshot with the RDM I created, I would be happy to help you out by e-mailing it to you.

Pushing the ISO to 1000 is always going to increase noise in the black areas, but with a few easy post strategies you can get a really good result from this footage, probably without using NEat video.

David
Click to expand...

I ended up using Redcine-X to make the correction after seeing that the Neat Video render time was going to be 4 days. There's a slight bit of noise that still exists in the image but I'm pretty pleased with what I have considering what it looked like when I first opened the files. I will say that the 1080 image being viewed on a 17" Panasonic monitor via SDI was very clean. If you look back at the shot with Leonardo Decaprio shot on Epic lighting a match http://www.youtube.com/watch?v=rLxQiI8c1Bs compared to the shot I have that's dropping a hard shadow, bright enough for a large room full of people to see a speaker at a podium, you would have thought that I would have been safe.
 
I'm glad it worked out lee.

I'm constantly surprised at how sharp the rendered footage looks when you drop the resolution.

A lot can be done in RCX just with ISO and curves and white balance.

David
 
I do appreciate all of the feedback, suggestions, and experience that is shared here. I've been in this business for over 20 years and learn something new every day. The day I think I know it all is the day I should depart from the industry. That being said.....I think I'll be here for a long time.
 
Hi Lee,

I had a quick crack at the image to get the following result (1920x1080 JPG, right click to save full size version):

A003_C001_1208IO.0012153F.1920x1080.jpg


Here's my RMD settings so you can apply them to your clip if you like what you see. Nothing done to the clip outside of RedCine-X PRO build 8 adjustments.

HTH!

Cheers from Japan,

Paul :)
 
David Battistella said:
Agree. But it's parlour tricks. ISO, HEadroom, whatever. Knowing how and where the sensor reacts is important, but at the end of the day, it's how much (or little) light you let hit the sensor and that is always controlled by aperture, nothing else.

David
Click to expand...

This not true unless you are shooting rough ENG work. Otherwise use NDs.
 
You must log in or register to reply here.
Back
Top