Welcome to our community

Be a part of something great, join today!

  • 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 :)

Whats the use of 4k in a budget camera?

Although I can see exactly where they're coming from ...

What I don't get in their rationale is, why didn't they decide to deBayer directly the 3840x2160 data to create a higher rez image than 1080p. They must have known this is possible as obviously they must be aware of this 80% stuff with Red cameras. Perhaps they wanted to avoid the OLPF, but that does not sound like the wisest thing to do. If they are keen with this idea of counting pixels, then I can somewhat follow their line of thoughts (altough would not do it the same way myself).
 
@ Eric

And a very important point.

The footage I captured raw in 2008 actually looks better today than the day it was shot. This kind of "evolving quality" sort of existed with film (as scanners got better) but Never with tape or 4:4:4:4 digital capture methods.

We can actually say that R3D gets better over time. What tape stock/DSLR/film stock/ can make that claim.

So is 4k overkill. No. No. No.

R3d is what we needed as a replacement for film. A digital raw image container. Period.

David
 
What I don't get in their rationale is, why didn't they decide to deBayer directly the 3840x2160 data to create a higher rez image than 1080p. They must have known this is possible as obviously they must be aware of this 80% stuff with Red cameras. Perhaps they wanted to avoid the OLPF, but that does not sound like the wisest thing to do. If they are keen with this idea of counting pixels, then I can somewhat follow their line of thoughts (altough would not do it the same way myself).

Well, demosaic+downsample is what they do in their DSLRs for generating JPEGs, so I don't know why they'd not do that with video. That said, I'm not sure on the downsample quality of their JPEGs as I've never really shot JPEG, only RAW with the cameras.

With the direct extraction approach they're using you actually need an OLPF set for 1080p rather than an OLPF set for QuadHD. But You can bet they've set theirs for QuadHD (testing will confirm) to allow for a much sharper (but aliasy) 1080p image.

Graeme
 
Mathew Joki said:
Yes, talking about Scarlet. North america TV (NTSC) broadcasts at 29.97 frames per second.

Content is broadcast at 29.97, but that does not mean that it is its native framerate. A lot of the primetime content is 23.976 with a telecine pattern. A good video processor can remove the telecine pattern and play the content back at 23.976.
 
With the direct extraction approach they're using

This is worse than what I thought and is so amazing that better to stop thinking what they are trying to do. Your test results will anyhow be interesting.

Quite the opposite feeling to what it was when first read about the more specific technical plans for Red One. Still remember so well thinking how clever the plan was, and for the first time started to think way beyond 1080p. Needless to say, I was immediately headed towards getting R1. My only concern was whether somebody will fund the project until you really have a working camera, but then later on while reading Steve Gibby's interview of Jim discovered there was even less reason to worry about the financial side.
 
I posted this in another thread but it is worth reposting here.

Longtime REDUSER Sanjin took this shot with a Leica lens a couple of years ago.

WIth the existing processing tools you can see how RED footage can be "developed" however you wish. When you enter 4K RAW you are entering a world of professional digital imaging. It's not WYSIWYG. It's what you want to see is what you WANT.

Keep in mind this is on an ORIGINAL R1 M sensor.

AS SHOT

A046_C008_1017EW.0003607F.AS_SHOT.jpg


GRADED RCX

A046_C008_1017EW.0003607F.GRADE.jpg
 
This is the old experiment I did to see what sRAW was doing - it's a similar technique, as far as I can tell, to the C300 extraction method.
sRAW_v_RAW.jpg


Then I coded up a simulation of the C300 extraction method and compared it to a full demosaic and proper downsample:

bayer_example.jpg


Once I can get a camera on the test bench I can measure rather than simulate, but given the evidence I have at hand from sRAW and my knowledge of demosaicing, sampling I can make a pretty educated guess.

What my tests don't show you is how fast the quick extraction is compared to full demosaic and downsample. Such expediency is excusable in a $2000 camera, but perhaps not in a $20,000.

Graeme
 
5k Isn't for everyone, it's for everyone who knows what to do with it. 8k, 16k,
whatever, keep it coming, some will put it to very good use. Others will squander it.
That will never change.
 
I wish I'd seen this thread earlier....

Why 4k (why 5k, why more-k)??

If all you're worried about is 1080p, 4k will make a better looking 1080p than any 1080p native system will. If you measure the resolution on a camera designed for 1080p, from the mighty Sony F35 down to the 5D2 you'll find they show less measured resolution than 1080p and show aliasing to some degree. By starting with more (like 4k or 5k) and properly optically filtering and properly downsampling you'll get a better quality higher measured resolution 1080p

4k (or higher) leads to better MTF especially at low frequencies, which leads to a nicer looking image

4k (or higher) gives you more resolution so you can punch in

4k (or higher) gives you more resolution so that you can more effectively post stabilize

4k (or higher) gives you a better key for your greenscreen

All of those benefits are valid for low budget as well as high budget productions, and all are valid if your final output is 1080p not 4k.

When people read resolution they rightly think detail, but there's more to the capabilities of resolution than just detail.

Graeme

Graeme, you didn't include lower noise from high ISO when downsampling from 4K to 1080. Is there any reason to expect 1080 downsampled ISO 1600 or 3200 images to have comparable noise to 4K ISO 800 images?

And is there anything RED could do to optimize high ISO downsampled output, either in hardware (short of changing the entire sensor) or software?

If the answer to the above is yes, is there any way to get that cleaner high ISO 1080p output more-or-less real time, now or somewhere down the roadmap?

Before anyone tells me to look for a different camera, I'm just trying to see if I can dual-purpose Scarlet. I also would prefer not to invest in a second set of 16mm glass to shoot on a windowed sensor, especially since my efp work is generally in uncontrolled light-challenged settings.

BTW, any thread where Graeme is offering up insights is not a waste of time and definitely should not be deleted!
 
Yes, a downsampled image will show lower noise, but it also means you can denoise at the higher resolution, then downsample and get much better results than if you'd only had the downsampled image to work on.

I'm not sure what we can do, but it's always something we can look at. I guess one of the problems with camera development is that there's always something else we could be looking at, but RED's a pretty small company and there's only so many of us. I've asked Jim for a clone, but the technology isn't there yet. I have made a couple of semi-clones (people generally call them children) but they're not quite ready to help me out with coding yet.

Graeme
 
Keep in mind this is on an ORIGINAL R1 M sensor.

David, nice example. I did also some experiments with the old R1 M footages and the newest RedCine-X Pro version. It was such a pleasure to discover what expecially Clarity does to old footages. At the same time graded scanned still color positive images. What a pain in the neck compared to the raw files!

What my tests don't show you is how fast the quick extraction is compared to full demosaic and downsample.

Very illustrative examples. So one reason for the choices could be limited processing power of the camera, which also explained why the battery manages to run the camera so long.
 
Very illustrative examples. So one reason for the choices could be limited processing power of the camera, which also explained why the battery manages to run the camera so long.

Power, heat and size are all elements, as is cost and time to market.

Graeme
 
What my tests don't show you is how fast the quick extraction is compared to full demosaic and downsample. Such expediency is excusable in a $2000 camera, but perhaps not in a $20,000.

Graeme

Graeme - this might be seen as a bit off topic but I think it fits here...
Can you give us any insight as to how the 1080p SDI output from Scarlet-X is derived, and how it might compare to a 1080p downsample created in post from 4K masters?
 
In camera we do a full demosaic and downsample. You'll get a little better in software, but it's pretty darn good in camera.

Graeme
 
R1 still did a full demosaic and downsample though.

Graeme
 
Power, heat and size are all elements

So your suggestion is, when ever one finds the fan bit distracting, one should ask oneself whether one is willing to make a trade-off between image quality, fan noise, and size. ;-)

About in-camera things, why the live 2K image in R1 looks so much better/sharper than the same footage through playback?
 
All camera design is trade-off. At RED we don't like to trade image quality. Playback in primarily for confidence on record and review, and the chips are optimized for codec recording rather than codec playback.

Graeme
 
10 Reasons why the Scarlet is worth it over other 1080p cameras

10 Reasons why the Scarlet is worth it over other 1080p cameras

First off, I think you point out a valid question and just want the reasoning behind this camera :) So here's my reasoning, these are not in order of what's most important.

#1. Yes, there really aren't many home screens to really view video better than 1080p, except on the 27" apple screen I'm using right now. In the near future that will be changing.

#2. Five years from now when there is a new bigger (or smaller) and better camera, I would still like to have the work I shoot over the next couple of years to hold up. Just because I shot it 5 years ago, doesn't mean it won't be worth putting in my portfolio.

#3. 13.5 stops of dynamic range at 4k

#4. HDRx boosts it up to 18 stops of dynamic range making it more possible to shoot in difficult lighting situations. It's at 3k for the moment, but that will also change.

#5. RAW video means it's more forgiving where as if I make a mistake with my HDSLR, I'm just screwed.

#6. Over the next few years as screens progress, internet speeds and video will continue to progress as well. Even today you can show off 2k+ videos online if you have the bandwidth or are willing to wait for the load.

#7. Shooting with a RED will increase my workflow time until I get a more powerful machine, but I hate the idea of sacrificing quality for shaving off a day or two on a project if it actually means something to me.

#8. Although the things I shoot now are relatively low budget, it's not my goal to allow it to look low budget and I don't plan on "always" shooting low budget.

#9. When it comes to detail shots and then compressing them for the current youtube or vimeo upload, you're not going to see a big difference between a DSLR and a Red Scarlet/Epic. But shoot a wide shot with lot's of scenery and the story changes. I love my HDSLR, but it isn't true 1080p.

#10. In this industry, it's important to keep up to date and the Scarlet is riding the cutting edge at an expensive, but attainable price for someone who is dedicated.

I can't wait to put in my order for the Scarlet (which should happen no later than December).
 
Back
Top