Welcome to our community

Be a part of something great, join today!

Register 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 :)

Steve Yedlin, ASC on Resolution

Misha Engel said:
I know and I would like to have a choice to go 1:1 RAW and/or 1:2..1:?? .R3D when DSMC3 comes out.
Click to expand...

This reminds me of the Blackmagic forums when everyone bitched about the firmware switch from 1:1 Raw to 3:1 Raw, as no one understood the concept of lossless versus lossy compression. There is no advantage of using 1:1 over 2:1 (if I'm not mistaken, up to 2.5:1 is lossless). All you're doing is wasting two or three times the storage for what is otherwise a mathematically identical image.

To take what David already said a step further, not only is 1:1 inherently not possible with R3D, but it would be entirely pointless.
 
Pure image noise cannot be compressed losslessly but real world images have some predictability that allows ~2:1 to usually be lossless. Perhaps an out of focus shot of a green screen lit evenly could be 5:1 or more losslessly...

Is the reason wavelets cannot be used for a full raw image because it would take infinite terms to define the waves? Seems like pixels aren't arbitrarily small nor bit depth arbitrarily deep so there should be some limit where it's capable. Or maybe I don't understand the methods involved.
 
Nathan DuMoulin said:
This reminds me of the Blackmagic forums when everyone bitched about the firmware switch from 1:1 Raw to 3:1 Raw, as no one understood the concept of lossless versus lossy compression. There is no advantage of using 1:1 over 2:1 (if I'm not mistaken, up to 2.5:1 is lossless). All you're doing is wasting two or three times the storage for what is otherwise a mathematically identical image.

To take what David already said a step further, not only is 1:1 inherently not possible with R3D, but it would be entirely pointless.
Click to expand...

1:1 RAW

and/or

1:2..1:?? .R3D

REDCODE SETTINGS
RANGE FROM 2:1 UP TO 22:11
5:1 REDCODE for 8K Full Format (8192 x 4320) at 24 fps
12:1 REDCODE for 8K Full Format (8192 x 4320) at 60 fps
2:1 REDCODE for 4K Full Format (4096 x 2160) at 24 fps
3:1 REDCODE for 4K Full Format (4096 x 2160) at 60 fps

Only REDCODE for 4K Full Format (4096 x 2160) at 24 fps can do 2:1 at the moment.

I would like 2:1 with REDCODE for 8K Full Format (8192 x 4320) at 24 fps AND/OR full 1:1 RAW when RED updates to DSMC3.

Or if someone can prove me that 5:1 gives exactly the same output as 2:1 and/or 1:1 RAW(lossless in any way).
 
To derail more...

I believe redcode is Red's strongest feature.. and we don't have ARRI cameras primarily due to their lack of raw compressed... so I am a big fan of compressed Raw and appreciate Reds vision.

BUT please can anyone show me anywhere any red employee states/claims clearly that 2.5 to 1 is mathematically lossless? Not 'basically visually lossless' but actually mathematically lossless?..
 
Compression is done in three stages - transform, quantization and entropy coding. If you're doing mathematically lossless you omit the quantization step. You can also take any lossy compression and turn the quantization off and see what compression ratio you get. Mathematically lossless compression is always variable rate as it is entirely content dependent. It's only by going to lossy compression can you target a fixed data rate.

So what happens in RedCode - you should see increasing file size per frame as you lower the compression. The point at which the file doesn't get any bigger is the point where you've reached mathematical losslessness. That will occur somewhere between 2:1 and 3:1 and is image content dependent.

Graeme
 
Thanks Graeme!

Graeme Nattress said:
Compression is done in three stages - transform, quantization and entropy coding. If you're doing mathematically lossless you omit the quantization step. You can also take any lossy compression and turn the quantization off and see what compression ratio you get. Mathematically lossless compression is always variable rate as it is entirely content dependent. It's only by going to lossy compression can you target a fixed data rate.

So what happens in RedCode - you should see increasing file size per frame as you lower the compression. The point at which the file doesn't get any bigger is the point where you've reached mathematical losslessness. That will occur somewhere between 2:1 and 3:1 and is image content dependent.

Graeme
Click to expand...
 
Björn Benckert said:
Sure does extra pixels does not make any difference what so ever. :)

A002_C003_11289I by Björn Benckert, on Flickr
Click to expand...

This misses the point ENTIRELY.

Man, I get the need to defend the choices you've made in buying a camera, but this is such a strawman.

Steve's point is not that there is no benefit to a higher pixel count, simply that over a certain resolution there is "enough" sharpness and the law of diminishing returns quickly sets in.

Your image is nothing at all to do with that.
 
Also, for all the Monday morning Quarterbacks in this thread.... Steve Yedlin shot Star Wars episode 8.

Remind me which episode you're shooting?
 
Tom Dowler said:
Also, for all the Monday morning Quarterbacks in this thread.... Steve Yedlin shot Star Wars episode 8.

Remind me which episode you're shooting?
Click to expand...

I agree with Yedlin, but this is the definition of an appeal to authority fallacy. Yedlin's position is well defended not because of the work he's done making movies, that is actually completely irrelevant. His position is well defended because he presents evidence of his claims.
 
jake blackstone said:
I always find fascinating the different approach to post when working with commercial productions in US and in Europe. Here in US, I rarely, if ever see or talk to a DP. Post is pretty much an ad agency domain. On the other hand, just about every time I do a job in Europe, it is not only run by a director, but it is also almost always attended by a DP. At the very least, if the DP or director must run to the next gig, we usually get to do at least a couple of hours "look" session, which becomes the basis for the final grade. The whole process working with director and DP makes so much more sense and the result is always much more streamlined.
Click to expand...
In features, typically the director has final say and does supervise, but often the DP is involved and attends some or all of the sessions. But in TV, the DP is generally still working on other episodes, so they can't supervise, and the directors often move on to other series or other episodes. That means in TV, the producer is king and generally has control of the sessions. This counts for both scripted and reality/doco shows.

Commercials are a whole separate animal, and that's usually "color by committee" in my experience: a creative director, an art director, the producer, the director, sometimes the DP, and even the client.

How this relates to resolution, I dunno, but in Yedlin's case it's interesting that he had the choice of any digital camera in the world for Star Wars VIII and chose Panavision 35mm film cameras (and to some extent IMAX film cameras).
 
Well I had to jump in at some point, just posted this on facebook due to many messages:

Resolution for Motion Picture Production
A lot of people have messaged me about Yedlin's recent resolution test and comments.

First, I respect much of what he's all about and in terms of audiences perception he's onto something interesting and he's certainly dead on with exactly how much film is resolving as well as noise levels.

Second, many are missing much of the subtext and even some things he actually points out about having literally more color and tonal detail with higher resolution sensor technology, which you can see when he toggles the images even on the web ready 1080p video he is showcasing (though the video was mastered 4K?). You can also see artifacts and over-sharpening of edges of the upscaled content when going from 2K to 4K.

The apparent question for me is what looks better when mastering 4K content or even "at capture resolution".

Actual source resolution, scaling down, or upscaling are the things we are talking about. The hard truth is scaling down yields a higher quality image especially from a Bayer pattern sensor.

This isn't a RED thing either. ARRI wrote the white paper on this back in the early days of film recording and scanning, it's part of the reason they even made the Alexa65, and more than half the reason we have different film format sizes is to provide a higher quality motion picture image.

For me much of this is going to be an utter non-conversation once a few new cameras drop late this year through half way through next year.

I've created a rather large, not web compressed 15MB .JPG so you can make some observations for yourself on a 4K image from an 8K source as well as going the other way by upscaling from 2K. And I've shown 4 common methods for both directions. Feel free to pass it around.

http://phfx.com/temp/reduser/phfx_RED8K_scalingExamples.jpg



A couple of other points I didn't make online as they are relevant here. Dragon and Helium are literally different sensor technologies and Helium's lower noise floor even with it's smaller pixel pitch alludes to what is actually going on there. You can't simply say smaller pixels are worse these days. Pixel design/sensor design is a major factor.

Also, I want to point it how common re-framing, cropping, digital zooming, and stabilization are when it comes to motion picture production these days. Having more source resolution will benefit you on that front.

As for compression, I agree that we can get there. Though I find nearly no interest in fully uncompressed data, I would love to be about to shoot 2:1 or 3:1 at 8K for instance. At the moment 5K is indeed possible with Weapon at 2:1 at speed, which is great. Though logistically, cost effectiveness, and workflow the good old fashioned rule of 5:1-8:1 REDCODE RAW Compression Ratio is pretty much the sweet spot for everything. I personally don't find myself higher than 12:1 if you are curious.


Anywho, I do think this test is great as it points out much. But Steve is saying there is a subtle difference between resolving power, points it out, and even showcases pretty clearly with his test shots.

The big question you will need to ask yourself at the end of the day is if that subtle difference is how you want to capture your motion picture content. For me, I like higher source resolution for many of the reasons I've stated above.


As for the marketing "purely all the Ks". That's not have I have experienced it, but it's damn well how many people talk about it. I don't mind saying a few things. Dragon and Helium's color and dynamic range is very advanced and a nice evolution from Mysterium-X. For me RED came into their own with Dragon's release. Color, Dynamic Range, and Resolution are all a part of that. I think it would be company suicide if only the resolution has advanced since the M and MX sensors, but truthfully it has been a combination of so many things that lead to higher quality images all around.


One thing I do truly agree with Steve on is testing. Test thoroughly to make your own observations. And I do find it odd that neither of RED's 8K options were included here, but that's for another day.


Side note, I have recently been able to shoot Kodak Vision 3 next to Dragon and Helium. I have some B-Roll I'll be writing something up about.
 
phil ftw

on a side note, what codec are youze guys using to upload 4k web content?

youtube is free and seems to stream pretty well, whereas vimeo is basically not free and seems to lag for me but may be better quality 4k once it finally loads... opinions?
 
Misha,

You are missing some fundamental information.

It is called REDCODE RAW. They are not two separate things REDCODE IS RAW when you are talking about RED. The whole reason you can record RAW to a MINIMAG is because of REDCODE RAW.

REDCODE RAW is a variable compression RAW format. If you record it 10:1 or 2:1 it is still RAW.



David



Misha Engel said:
1:1 RAW

and/or

1:2..1:?? .R3D

REDCODE SETTINGS
RANGE FROM 2:1 UP TO 22:11
5:1 REDCODE for 8K Full Format (8192 x 4320) at 24 fps
12:1 REDCODE for 8K Full Format (8192 x 4320) at 60 fps
2:1 REDCODE for 4K Full Format (4096 x 2160) at 24 fps
3:1 REDCODE for 4K Full Format (4096 x 2160) at 60 fps

Only REDCODE for 4K Full Format (4096 x 2160) at 24 fps can do 2:1 at the moment.

I would like 2:1 with REDCODE for 8K Full Format (8192 x 4320) at 24 fps AND/OR full 1:1 RAW when RED updates to DSMC3.

Or if someone can prove me that 5:1 gives exactly the same output as 2:1 and/or 1:1 RAW(lossless in any way).
Click to expand...
 
What this test shows also is that oversampling is where it is at. The reason we talk about pixels is that having more of them, in the end matters. Also, how much longer are we going to be in HD land over UHD. The user experience Netflix standard has migratied to 4K.

Oversampling for aquisition is where it is at.

The image chain in digital needs to carry the same mentality as the photochemical process.

1.Capture
2.Process
3.Print

If you do not run your images through these three phases prior to production you can not have any real control over your images.

This three step process is where it is at.

Ansel Adams Zone system for exposure still applies.

Ansel Adams is where it is at.

I highly doubt you could have convinced Ansel Adams that creating his images on 35mm neg would have been the same or meant the same thing after taking it through the process.

Large format aquisition is where it is at.

We can all pretty much agree that in this test the Alexa 65 is the DIGITAL camera you would pick, for resolution and flexibility. It is the only true large format digital sensor of the bunch maybe even against 11K film scans!

Sure, it is good to know you can mix and match stuff and get closer, but when you pixel peep the pixels are nicer. Question is, does nicer pixels make nicer images when you come back out to the full, intended image. And that is what I like about this test. He pixel peeps and compares the images as they are intended to be viewed. An audience may not notice the perceived quality difference, but is very important for us to know what happens under the hood.

David
 
Last edited:
You must log in or register to reply here.
Back
Top