R3D sample: Folks with more RED experince than I please weigh in on noise performance

[Matt Ryan]

Jeremy,

I was Shooting a doc today on Scarlet (which is waaaay easier than i thought it would be), and i've been exposing far to the right. Overexposing without clipping, the images are "questionable" on set but a bringing it down in RCX makes it sing. Grainless and rich. Its a different way of working. But the rewards are great. Follow the advice of others. It seems unnatural and against your instincts, but it works.

Also, got to try out hdrx today. Its the shit. Love it. Want it in 4k more than anything.....

nice, glad you got exposing for the Scarlet down. Crazy how good the images can be when exposed correctly huh?

 

[Jeremy Hildebrant]

Glad it's working out! Ditto on HDRx in 4k, mmmm.

Jeremy, I'm curious as to how you are viewing your footage. I find that makes a huge difference in perceived noise levels. Are you watching it rendered out of your NLE/finishing system or are you judging off of something like 1/4 playback in Premiere Pro? If you are in a proper grading suite with a calibrated monitor you can ignore this question of course

William, I have viewed it a few different ways, but anytime I have done any critical judging whether it be Premiere Pro or RCX I have made sure my playback and paused resolution are set to full. Also viewed in Photoshop using the beta Adobe Red plug in.

I am all about exposing to the right, been doing this with raw digital before there was RED. I guess where I get bogged down is, I can push that histogram way to the right by increasing the ISO, but it doesn't matter if the sensor is not receiving enough light in the first place, right? In an extreme situation that histogram can be to the right but you can still have the raw noise meter filled, and vice versa for that matter.... right? One is true/raw data and the other is just signal path to the camera display.

Sometimes stuff looks great, other times not so much.
I guess I just need to still get a better feel for it. I am going to do some more testing, try and be real concious of the raw signal, check exposure mode, etc.

 

[William Robinette]

William, I have viewed it a few different ways, but anytime I have done any critical judging whether it be Premiere Pro or RCX I have made sure my playback and paused resolution are set to full. Also viewed in Photoshop using the beta Adobe Red plug in.

I am all about exposing to the right, been doing this with raw digital before there was RED. I guess where I get bogged down is, I can push that histogram way to the right by increasing the ISO, but it doesn't matter if the sensor is not receiving enough light in the first place, right? In an extreme situation that histogram can be to the right but you can still have the raw noise meter filled, and vice versa for that matter.... right? One is true/raw data and the other is just signal path to the camera display.

Sometimes stuff looks great, other times not so much.
I guess I just need to still get a better feel for it. I am going to do some more testing, try and be real concious of the raw signal, check exposure mode, etc.

Right. You know I'm finding it a bit hit or miss as I go. The more I learn and more tests and do and see the less I miss and more I hit but still, I'd like to be 100% all the time (as I'm sure we all would be). So when you say the Scarlet is producing a bit more noise then would would think, what is the comparison too? Raw still frames pulled from motion? The 4k files in motion? Down sampled output to 1080p/2k? Are you shooting 4kHD? 4k?

I put my 7d next to my Scarlet both shooting raw (the Scarlet at 5k for still pulling) at the same scene and compared in PS. The Scarlet files were output as 16 bit TIFFs from REDcineX-Pro as to my knowledge the PS plugin is 8 bit only right now (could be wrong). Anyway, the Scarlet blows the 7d raw out of the water in pretty much every way. I'm getting better and better results out of the video abilities by really just ETTR as you said. Just squeezing every last drop before clipping really seems to be the ticket.

Although I'll say every time I do a full max quality render that is fully graded to 1080p and watch it on an actual television the results are jaw dropping. No noise anywhere for anything close to properly exposed footage.

 

[Jeremy Hildebrant]

Right. You know I'm finding it a bit hit or miss as I go. The more I learn and more tests and do and see the less I miss and more I hit but still, I'd like to be 100% all the time (as I'm sure we all would be). So when you say the Scarlet is producing a bit more noise then would would think, what is the comparison too? Raw still frames pulled from motion? The 4k files in motion? Down sampled output to 1080p/2k? Are you shooting 4kHD? 4k?

I put my 7d next to my Scarlet both shooting raw (the Scarlet at 5k for still pulling) at the same scene and compared in PS. The Scarlet files were output as 16 bit TIFFs from REDcineX-Pro as to my knowledge the PS plugin is 8 bit only right now (could be wrong). Anyway, the Scarlet blows the 7d raw out of the water in pretty much every way. I'm getting better and better results out of the video abilities by really just ETTR as you said. Just squeezing every last drop before clipping really seems to be the ticket.

Although I'll say every time I do a full max quality render that is fully graded to 1080p and watch it on an actual television the results are jaw dropping. No noise anywhere for anything close to properly exposed footage.

William,
Well, when Timur started this feed and provided the R3D snapshot, I thought to myself "hey that looks like I would expect to look in the environment that he shot it in". I didn't see anything displeasing about it. I saw some noise but it was pleasant like film grain. FYI, I like grain and I am not looking for super smooth clean images like I would expect from a still image captured by a Phase One digital back. I guess I am saying, don't think that I have unreasonable expectations

The funny thing is when I do encounter noise from the Scarlet, i actually like how it looks better at 1/2 res in RCX, because it is more granular then. I also think that noise is more apparent based on the color and tone of the subject matter you are shooting. Noise is not going to be as apparent on something that has some texture a variation in tones to it, than it would be on a flat grey wall. You know what I mean? And blue channel noise is going to be more apparent on a warm flat surface than it would in scene with a lot of depth that already has a cold tones to it. Like I said I just need to have more time to get a better feel for it. I know the limitations and advantages of a 5DmkII, I am still discovering the Scarlet.

As far as how I have compared is, usually I have had the camera 4kHD. I have looked at them full res in RCX, Photoshop, Premiere Pro, and exported 1080p QT prores files. Only on a computer monitor though. I have not done an accurate side by side comparison, but yesterday I pointed the Scarlet and a 5D at a underlit set that was being built. I knew ahead of time that part of the Scarlet's sensor was going to be in "noise" But I also instinctively thought that the 5D would handle it fine, at least on a perceptual level. Both cameras opened up to 2.8, both at 1600, 1/48 and 1/50, 3200k. Same lens. Marvels picture style on the 5D. Looked at them both next to each other in a 1080p timeline, and the 5D appeared to be more pleasing even though I know it smearing the hell out of noise, but edges detail is still there, and is a pretty clean image. Field of view was different on both cameras because I was doing other things at the same time, so I want to redo something like it to compare apples to apples.

Now I know that if there was more light that it would have been a completely different situation, but I am trying to feel out the limitations. I have seen some beautiful footage from the Scarlet in lighting environments where it is pretty much all practicals and/or some window light. And I think it goes back to what I was saying before about subject matter itself, real life shoot situations are much different than say shooting tests of a color card or something....

Sure is a lot of talk about proper exposure and noise on reduser.... In the near future I expect there to be less.

 

[Björn Benckert]

Timur, I guess your noise meter / goal post was quite high when you shot this since your left side of the image is all in in low exposure range for the sensor. If you would lift that light just a little bit, putting more fill in there then you would be without problems. Using the floor of the sensor is something at least something I try to stay away from.
IF you shoot in the given light you can only open up your lens and cheat it with shutter speed but exempt for that there is not much you can do exept for noise reduction or crunching the noise away but that leaves you contrasty image that does not look that good.. So try to stay within the red bars.. or do make sure they do not show while you shoot...

 

[William Robinette]

Thanks for the replies Jeremy! Hope you don't think I'm coming off as anything but curious, just want to gather as much information as to people's experiences as possible! It sounds like we are going though about the same things.

 

[Jeremy Hildebrant]

Thanks for the replies Jeremy! Hope you don't think I'm coming off as anything but curious, just want to gather as much information as to people's experiences as possible! It sounds like we are going though about the same things.

I'm right there with ya William! Curious, of others experiences.

 

[Timur Civan]

Thank you for all the help folks.

I'm used to shooting for look and not so used to grading. So I'm forced to retrain how I do things. Y'all been a great help!