SCARLET NO NOISE AT 4K, 3K, 2K

[Nick Morrison]

On the lens, try using ND's and keeping your 'sweet spot' Fstop lower... lenses are often softer wide open and also fully closed (again, a recent thread had some examples of this).

It might be a combination of things - try some more tests with ND's to control your exposure (rather than f22) - try and stay F8.

Waveform wise, I'd be looking for a fatter exposure (more mid and to the right) if this was shot on a sunny day outside... of course, I've no idea what the image is - so I could be way off

Exactly right - yeah traditionally, lenses get sharpest two stops from the bottom or top. So a 2.8 lens commonly gets tack sharp at 5.6, and similarly if it maxes at f22, then it comes in sharper at f11.

That being said, zooms are by their nature a compromise (even though some are excellent!)...and at the same fstop a good prime will normally out-resolve them.

So...as Al was at f22 on a zoom...it can only get better from there!

 

[Les C.]

I'm hoping it's just me and being a bit of a noob with this camera. I hate to say it but I've been far from impressed with my first bit of footage on this thing. First soft images and now noise. Not happy

I'm going to upload a clip in a short while so you can see how bad it is. BTW I did shoot RC 12:1 so not sure if that introduces too much noise?

Increasing the compression will actually remove noise, as well as detail you want to keep.

But, when shooting at 4K, you are really shooting at 3.2K due to the Bayer pattern. Missing pixels get interpolated and that introduces errors when viewed at a resolution higher than 3.2K.

http://www.reduser.net/forum/showthread.php?40067-Thoughts-and-Tips&p=539848&viewfull=1#post539848

That's why you have to shoot at 5K to get 4K

Also check out this:

http://magazine.creativecow.net/article/the-truth-about-2k-4k-the-future-of-pixels

 

[Mick van Rossum NSC]

I try to avoid 2K, but if I need to I set the ISO to 320. Don't know why, but the image looks much cleaner if you overexpose a bit in 2K. This tip came from Ted actually.....

 

[Scott Crawley]

Al is a Scarlet owner. 5k is not an option and it is not necessary. If delivering 4k meant that you had to deliver a frame that was of actual 4k dimensions I might agree with you, but as it stands "4k" does not have a precise definition.
Will shooting 5k yield better 4k? No doubt, but that isn't what you are advocating. What you are saying is that Scarlet can not deliver to 4k, and that is FUD.

Increasing the compression will actually remove noise, as well as detail you want to keep.

But, when shooting at 4K, you are really shooting at 3.2K due to the Bayer pattern. Missing pixels get interpolated and that introduces errors when viewed at a resolution higher than 3.2K.

http://www.reduser.net/forum/showthread.php?40067-Thoughts-and-Tips&p=539848&viewfull=1#post539848

That's why you have to shoot at 5K to get 4K

Also check out this:

http://magazine.creativecow.net/article/the-truth-about-2k-4k-the-future-of-pixels

 

[Jake Wilganowski]

Red doesn't apply sharpening, so you have to sharpen to taste

 

[Lliam Worthington]

Jakes correct of course. Though I'll just add, with good lenses used effectively and proper focus, you will most certainly get sharp images on scarlet right out of the box too.

I try to avoid 2K, but if I need to I set the ISO to 320. Don't know why, but the image looks much cleaner if you overexpose a bit in 2K. This tip came from Ted actually.....

That's interesting... Love to hear some of the more tech savvy users on here extrapolate on this.
Overexposing 320 ISO sounds like a dangerous practice when it comes to highlights?

Anyone?

 

[Les C.]

Al is a Scarlet owner. 5k is not an option and it is not necessary. If delivering 4k meant that you had to deliver a frame that was of actual 4k dimensions I might agree with you, but as it stands "4k" does not have a precise definition.
Will shooting 5k yield better 4k? No doubt, but that isn't what you are advocating. What you are saying is that Scarlet can not deliver to 4k, and that is FUD.

Scott - did you look at the link? The 'FUD' that you only get 3.2K from shooting 4K is from Jannard, not me.

http://www.reduser.net/forum/showthread.php?40067-Thoughts-and-Tips&p=539848&viewfull=1#post539848

 

[Scott Crawley]

Scott - did you look at the link? The 'FUD' that you only get 3.2K from shooting 4K is from Jannard, not me.

http://www.reduser.net/forum/showthread.php?40067-Thoughts-and-Tips&p=539848&viewfull=1#post539848

I'll admit that I did not read the link. I've heard the arguments many times before. I was responding to the notion that shooting 4k on a Red Bayer pattern is not enough for "4k" delivery, which is what you appeared to be saying:

That's why you have to shoot at 5K to get 4K

 

[Les C.]

I'll admit that I did not read the link. I've heard the arguments many times before. I was responding to the notion that shooting 4k on a Red Bayer pattern is not enough for "4k" delivery, which is what you appeared to be saying:

When you only get 3.2K actual resolution, if you deliver at 4K, what are you delivering?

I'm not saying it's bad, I'm just saying it's not really 4K. You can't argue with science - and can't really create information where it doesn't exist without introducing something that diverges from reality that you can call noise.

BTW - sometimes you actually want to introduce noise to get a better trans-coding (and less 'noise'), but that's an entirely different topic.

Scarlet's still a great camera, it's just not designed to give you 4K resolution at 24p.

I believe the 'Girl with the Dragon Tattoo' was shot in a mixture of 4K and 5K, so even shooting in 4K certainly gives you usable 4K (as it was projected at 4K in many theaters) - especially as you are sitting back so far in the theater that you can't really discern the details.

Just don't look at 3.2K of resolution @ 100% in 4K size and expect it to be perfect. Try scaling it down 80% and see how the noise looks

Hopefully, Dragon will put all this to rest. Should be an interesting NAB.

 

[Scott Crawley]

When you only get 3.2K actual resolution, if you deliver at 4K, what are you delivering?

I'm not saying it's bad, I'm just saying it's not really 4K. You can't argue with science - and can't really create information where it doesn't exist without introducing something that diverges from reality that you can call noise.

What is 4k really? Is it 4096x2304? 4096x2048? 3840x2160, 2816x2304... 4096x2400? The point being that it depends upon aspect ratio among other things.

You don't have to add anything to Scarlet's 4k to project it at "4k". Red's 4k frame is 4096x2304. Barco apparently calls a 4096x2400 frame 4k (4kHD) at 16:9... So Scarlet's 4k frame fits right in there. Trim a little more off the top and bottom and you have what Red calls 4k 2:1 4096x2048. Have you added anything to get there least of all noise? No you have not. The width is equal in all cases, so you don't need to enlarge the frame at all. I don't know what you are talking about.

I'm no projectionist but WTF?

 

[Scott C]

I believe Les' point is well made. Mathematically if you shoot at 100% of any given target size, and retain 80% of the information from that target size, you will not have 100% of that size, even if it is scaled back up. It is very simple. If you are to make the argument otherwise, I could say that an old JVC handycam can shoot 4K, simply because I can go into an NLE and scale it up. Look, 4K.

Arguing about the precise definition of 4K resolution to a small % tolerance is semantics and aside from the point.

What is 4k really? Is it 4096x2304? 4096x2048? 3840x2160, 2816x2304... 4096x2400? The point being that it depends upon aspect ratio among other things.

You don't have to add anything to Scarlet's 4k to project it at "4k". Red's 4k frame is 4096x2304. Barco apparently calls a 4096x2400 frame 4k (4kHD) at 16:9... Trim a little off the top and bottom and you have what Red calls 4k 2:1 4096x2048. Have you added anything to get there least of all noise? No you have not. I don't know what you are talking about.

 

[Scott Crawley]

Walk into a "4k" projection with your "4k" Red master and nobody is going to bat an eye. Red "4k" is accepted as 4k, period. Semantics and particulars aside, you do not need to shoot 5k to deliver in todays accepted 4k format. That is ultimately the point, and the only point that matters. Whether or not the math equals 4k apparently doesn't mean shit. So unless I am missing something significant here, his point is moot.

I believe Les' point is well made. Mathematically if you shoot at 100% of any given target size, and retain 80% of the information from that target size, you will not have 100% of that size, even if it is scaled back up. It is very simple. If you are to make the argument otherwise, I could say that an old JVC handycam can shoot 4K, simply because I can go into an NLE and scale it up. Look, 4K.

Arguing about the precise definition of 4K resolution to a small % tolerance is semantics and aside from the point.

 

[Scott C]

You are making a different argument at this point. What I am saying is that the math is valid. The validity of his point as moot is your opinion, one I do not share.

Walk into a "4k" projection with your "4k" Red master and nobody is going to bat an eye. Red "4k" is accepted as 4k, period. Semantics and particulars aside, you do not need to shoot 5k to deliver in todays accepted 4k format. That is ultimately the point, and the only point that matters. Whether or not the math equals 4k apparently doesn't mean shit. So unless I am missing something significant here, his point is moot.

Les said this:

Originally Posted by Les C.
When you only get 3.2K actual resolution, if you deliver at 4K, what are you delivering?

I'm not saying it's bad, I'm just saying it's not really 4K. You can't argue with science - and can't really create information where it doesn't exist without introducing something that diverges from reality that you can call noise.

and your response was to call into question the definition of 4K. His point is correct.

 

[Gunleik Groven]

Like HD for a while practically meant "more than SD" (and at a later stage dipped far bellow SD), 4k currently means "more" than 2k.

Which is good!

 

[Les C.]

What is 4k really? Is it 4096x2304? 4096x2048? 3840x2160, 2816x2304... 4096x2400? The point being that it depends upon aspect ratio among other things.

You don't have to add anything to Scarlet's 4k to project it at "4k". Red's 4k frame is 4096x2304. Barco apparently calls a 4096x2400 frame 4k (4kHD) at 16:9... So Scarlet's 4k frame fits right in there. Trim a little more off the top and bottom and you have what Red calls 4k 2:1 4096x2048. Have you added anything to get there least of all noise? No you have not. The width is equal in all cases, so you don't need to enlarge the frame at all. I don't know what you are talking about.

I'm no projectionist but WTF?

Scott - when you don't have an RGB for each pixel, and have to interpolate the missing ones ... that means there's missing data. That's why you don't get true 4K resolution.

In Jannard's words:

"You can use 80% to figure out exactly what the final measured resolution will be. For example, 4K-RAW will end up as roughly 3.2K (4096 x 80% = 3276) measured resolution."

Read the CreativeCow article. The author has an impressive resume. Maybe the rather long article will make more sense to you.

http://magazine.creativecow.net/article/the-truth-about-2k-4k-the-future-of-pixels

But again, look at the "4K" images at 80% instead of 100% and much of the 'noise' should be gone.

As most of us probably re-purpose for 1080p, after de-noising with NeatVideo, the noise issue is really a non-issue for a properly lighted scene.

 

[Gunleik Groven]

Scott - when you don't have an RGB for each pixel, and have to interpolate the missing ones ... that means there's missing data. That's why you don't get true 4K resolution.
/QUOTE]

Sure kinda... The right conclusion, but the wrong premise...

You can get 4k rez out of a 4k sensor, it just won't look very nice. Or... at least it will look pretty hard. Thus the OLPF which slightly softens the image.

As to the colorsampling, you are spot on...

 

[Scott Crawley]

DCI only calls for 4K (4096×2160) Scarlet's 4k exceeds that. Nothing else need matter to me. If pixel peepers want to insist that it isn't truly 4k they can go nuts with that. As long as I can meet the specs for deliverables you can call it whatever you want. :beer:

 

[Igor Ryabchuk]

Correct me if I'm wrong, but as far as I know when you capture anything with Bayer-pattern based sensor, you get green at half-res and red/blue at quarter-res, but you DO get luminance at each photosite at full-res. And that's quite important - take any photo, convert it to LAB color mode in Photoshop and then blur the L channel and compare it to blurring the A/B channels. When you blur Luminance (L) you can hardly tell what you see on the image, while blurring the Chrominance (A/B) will give you a painty look but with sharp shapes of all the objects. So what you get with RED as 4K camera is 4K of perceptual resolution with 3.2K of color resolution. Again, I might get it wrong.

 

[Scott C]

This I can agree on. =)

If pixel peepers want to insist that it isn't truly 4k they can go nuts with that. As long as I can meet the specs for deliverables you can call it whatever you want. :beer:

 

[Scott Crawley]

...the noise issue is really a non-issue for a properly lighted scene.

I think this speaks more to this thread and the noise questions that new Scarlet users are having, since in the vast majority of cases they are not monitoring in 4k or otherwise scaling up to fit 3.5k in a 4k frame, or whatever. They are most likely making exposure and post processing errors. Folks need to remember to expose to the right and at least do a proper 1 light to bring black values down to where they belong. If they do this most of the noise issues will likely resolve themselves.

That's my guess.