2K Soft Focus on all Red Ones

[Jeffrey Lamont Brown]

2k=soft

2k=soft

David,

Please, if you have the time, come over to my studio and we'll fire up some cameras and I'll show you what we're talking about. I'm at New Deal downstairs from Ken at Filmworks, we've seen each other there a few times while you were timing Manure and Stay Cool.

This is a very frustrating issue. When shooting 2k off speeds with subjects at around 15' or greater using a wide lens the picture is unusable for compositing plate work. This is very apparent when shooting VFX elements at 4K and say a 25mm or a 32mm S4 and having to match that same shot at an off speed in 2K. The difference is more than just blowing up the center punch of a 4k sensor to get the 2k image which I've done to compare the differences. The difference is enough to bring out the film cameras and pull the Reds.

I'm actually shooting with a 3D SI2k rig for the next two weeks on our stage and that might be a fun test but the lenses on the SI2k rig are these tiny Schnider C mount necessary to fit the beam splitter rig and wouldn't be a very good comparative test against my S4s but it would be a good test especially if the SI2Ks were actually sharper. Why don't you come by and join us for a look see?

Cheers!
Keith

I'm seeing the same sorts of softness. Our 2k offspeed footy was so soft that I sent the camera to RED for repair. They checked it out. Still, we see a substantial loss of sharpness going from 4k (viewed at 2k) to 2k when all other variables are held the same. So much so that we don't shoot any 2k.
Cheers,
JB

 

[Brian D. Goff]

To me this looks like a debayering issue. I would imagine, that the debayer is optimized for 4k and when shooting below 4k the debayer process is different....

 

[David Mullen ASC]

This is a very frustrating issue. When shooting 2k off speeds with subjects at around 15' or greater using a wide lens the picture is unusable for compositing plate work. This is very apparent when shooting VFX elements at 4K and say a 25mm or a 32mm S4 and having to match that same shot at an off speed in 2K. The difference is more than just blowing up the center punch of a 4k sensor to get the 2k image which I've done to compare the differences. The difference is enough to bring out the film cameras and pull the Reds.

I'm actually shooting with a 3D SI2k rig for the next two weeks on our stage and that might be a fun test but the lenses on the SI2k rig are these tiny Schnider C mount necessary to fit the beam splitter rig and wouldn't be a very good comparative test against my S4s but it would be a good test especially if the SI2Ks were actually sharper. Why don't you come by and join us for a look see?

Cheers!
Keith

Sure, I'd love to see a comparison in 2K between the RED and SI2K, but again, the question is why should it make a difference if you use the same lens, same image... but crop it in post to 2K versus in-camera to 2K? It's the same MTF, same degree of enlargement, same everything. Could the REDCODE compression of the 2K version be an issue then? I just don't understand why there would be a difference.

 

[Brian D. Goff]

Sure, I'd love to see a comparison in 2K between the RED and SI2K, but again, the question is why should it make a difference if you use the same lens, same image... but crop it in post to 2K versus in-camera to 2K? It's the same MTF, same degree of enlargement, same everything. Could the REDCODE compression of the 2K version be an issue then? I just don't understand why there would be a difference.

I'll bet it's a software bug When the project settings in the camera are set to other than 4k, I assume there's a different piece of software doing the debayering - Gramme?

 

[Roberto Lequeux]

I am sure this was brought up but, is there some confusion going on with scaling 4k to 2k vs. cropping the image recorded or using only the 2k area of the sensor? Obviously everyone knows that 4k re sized to by -50% gives you a whole lot more detail than straight up 2k. Also perhaps people are used to getting 2k from a 4k scan that the lab down converts for you to give you better 2k and avoid someone saying they know a lab that gives you better 2k.

Point is I don't understand how someone would have figured out all of a sudden that everyone's Reds give you soft 2k... especially if they believe that at 4k it isn't soft since nothing changes. If anything it should get sharper by using the better part of the glass.

I wonder if what David asked about might be the reason?

Because if I understand correctly, the larger the image the better it compresses.

 

[Evangelos Achillopoulos]

Sure, I'd love to see a comparison in 2K between the RED and SI2K, but again, the question is why should it make a difference if you use the same lens, same image... but crop it in post to 2K versus in-camera to 2K? It's the same MTF, same degree of enlargement, same everything. Could the REDCODE compression of the 2K version be an issue then? I just don't understand why there would be a difference.

David, Keith is doing off speeds, I guess, that means slow motion... 100fps... so thats why he is forced to use the windowed 2K mode...

Hello Keith... its being some time... 3d... wow... I wish I was there, to play...

 

[keicol]

Evangelos,

Thanks for clarifying that for me, yes, I only use 2k is only for Slow Motion effects.

Oh and, I may have another film out for you soon.

Let's Skype second week in July.

Keith

 

[D Namir]

when shooting 2K and especially wide shot always use a 16mm lens format and not a 35mm format.

 

[David Mullen ASC]

David, Keith is doing off speeds, I guess, that means slow motion... 100fps... so thats why he is forced to use the windowed 2K mode...

Hello Keith... its being some time... 3d... wow... I wish I was there, to play...

Yes, I understand that, but that doesn't answer the question -- why would there be a sharpness difference if you did a shot on the same lens and cropped it from 4K to 2K in post versus in-camera? It's the same lens, the same stop, the same distance focused, it's just the difference between cropping the sensor to 2K before compression as RAW, versus converting 4K RAW to RGB and then cropping to 2K. Either way, the debayering is happening in post, so why would there be a difference? Is it the REDCODE compression of the 2K version? Is it the debayering of the 2K version?

The issue isn't the use of a 16mm lens versus a 35mm lens because from what people are saying, the same 35mm lens looks softer when the crop is done in-camera instead of in post. Now why would that be?

 

[Harry Lipnick]

I'm with David here; this must be an electronic/digital issue. Quite simply, there is no optical/physical rule that would explain why this is happening. I wonder if Graeme has any ideas here, of course he's the real tamer of the magic pixies anyway...
Peace,

-Harry

 

[keicol]

I'm hoping it could be something as simple as a software firmware update to the camera that could solve this problem. Or at least have Red made aware of this issue so the next cameras they make will have taken this into consideration and hopefully put this issue to rest.

By the way, I've tried the latest version of Zeiss S16 Ultra primes from Keslow and found the same problem.

 

[David Birdy]

I'd like to see a focus chart test to settle this question...I haven't noticed this issue and we shot a lot of 2k High frame rate material.

The sensor scales, so you get more resolution as the frame scales up from 2k to 3k to 4K...IE the resolution of the CMOS sensor should be linear across the entire sensor, or you would see softness in the center of the 4k image..

Put the camera on a focus chart and keep all variables the same, lens, frame rate, distance, etc then scale from 2k to 3k to 4k... If each resolution is sharp try it again and increase the frame rate and lighting level to compensate for the higher frame rate.

If there is an issue I would think it would have to do with the high frame rate scan, not the sensor resolution...

Dave

 

[conrad gaunt]

I don't see why it would make a difference whether you cropped 4K to 2K in-camera or in post other than you have less info to compress if you do it in-camera.

But remember that the low-pass filter was optimized for the full 4K sensor.

To compare something remotely similar, you'd have to compare the Red in 2K to the SI-2K to see if a sensor designed for 2K, with an OPLF designed for that resolution, produces better results or not. Otherwise, I can't think of any other single-sensor 2K camera anyway to tell you if there is any way to get a better 2K RGB image when only starting out with 2K RAW.

I'm no expert, but using what logic I have.. surely the OLPF shouldn't make a difference when comparing a 2k "crop" to a 4k "full frame" from the same sensor, it should only matter when comparing a 2k "full frame" image to a 4k "full frame image" from the same sensor (impossible!), since one (the 4k one) would be trying to resolve higher frequencies than the other.

The OLPF is simply there to help resolve high frequency details that fall between the cracks, and can't be resolved properly by the sensor, so..

..since the cracks (photosites etc) are the same size in the 2k crop as the 4k full frame, surely the OLPF shouldn't make a difference?

I have to say, those TIFFs all look awfull to me, are you using a lossy form of TIFF?

I noticed when I applied auto contrast to all three images in Photoshop, two of the images then had bluish table clothes and one had a decidedly greener colour?

 

[Mitch Gross]

I have done the same test David describes, shoot in 4K and pull a center crop in post and compare that to a 2K file shooting at the same subject with the same lens and other settings. The images resulting are essentially identical. There are probably some compression efficiency differences but they are effectively a non-issue.

2K on the RED is a bit soft & flarey. It is really a 4K camera, with an OLPF designed to prevent alaising at that resolution. When you take a lesser resolution and blow it up the effect of the OLPF is greatly magnified. Think of it as a diffusion filter -- zoom into the center and the diffusing effect is stronger. Simple optical physics.

Compare this to an SI-2K, with an OLPF that is designed for that sensor to be used at 2K. Also the pixel size is different (SI-2K is larger). The SI-2K is noticeably, dramatically sharper, but then again it cannot shoot in 4K and if it did with the same strength OLPF it would likely suffer from aliasing issues. There is another 2K camera around, the Phantom HD. It has a vastly larger pixel size, with each photosite 12.5 microns instead of the RED One's 5.4 microns. That's a huge difference. And it has an appropriate strength OLPF for it's size, which makes it a very sharp camera. In many ways it is as sharp as the R1 in 4K, even though it is a 2K camera.

 

[keicol]

I haven't noticed this issue and we shot a lot of 2k High frame rate material

It's most apparent when shooting wider shots with the subject at least 13' to 15' away.

 

[keicol]

Think of it as a diffusion filter -- zoom into the center and the diffusing effect is stronger. Simple optical physics.

That's great when your out shooting your own work and you don't have to answer to anyone who might expect to maybe pull a matte from an off speed 2k plate shot. it's impossible, like you said, simple optical physics.

 

[Roberto Lequeux]

It has to be compression being less productive/effective in smaller images.

I seriously doubt that Red would make a camera reach a higher framerate by making compromises because they are fully aware that an overcrank is displayed at the same 24fps speed as the rest of the movie. It would make no sense whatsoever and it would open up the door for criticisms of the line-skipping kind if you know what I am talking about. However a small difference in compression effectiveness of their Red Code when used in the 2k flavor makes a whole lot more sense when you consider how much Red Code is doing compression wise to keep the hard drives from filling up right away.

With Red Ray and the myriad of Red Code flavors announced one would be smart to think that that things could change but most likely RC36 will remain unchanged as their effort will likely go into the new heavier RCs. Also, with these heavier amounts of data the 2k issue might be gone. Unless the ratio or detail difference from 2k to the higher resolutions remains untouched... I am glad to see numbers like 125fps with Epic-X at 4k or 100fps at 5k. Or 125fps 5k and 150 4k with Epic FF35.

I have done the same test David describes, shoot in 4K and pull a center crop in post and compare that to a 2K file shooting at the same subject with the same lens and other settings. The images resulting are essentially identical. There are probably some compression efficiency differences but they are effectively a non-issue.

2K on the RED is a bit soft & flarey. It is really a 4K camera, with an OLPF designed to prevent alaising at that resolution. When you take a lesser resolution and blow it up the effect of the OLPF is greatly magnified. Think of it as a diffusion filter -- zoom into the center and the diffusing effect is stronger. Simple optical physics.

Thanks Mitch. It makes a lot more sense now.

 

[conrad gaunt]

I have done the same test David describes, shoot in 4K and pull a center crop in post and compare that to a 2K file shooting at the same subject with the same lens and other settings. The images resulting are essentially identical. There are probably some compression efficiency differences but they are effectively a non-issue.

2K on the RED is a bit soft & flarey. It is really a 4K camera, with an OLPF designed to prevent alaising at that resolution. When you take a lesser resolution and blow it up the effect of the OLPF is greatly magnified. Think of it as a diffusion filter -- zoom into the center and the diffusing effect is stronger. Simple optical physics.

Compare this to an SI-2K, with an OLPF that is designed for that sensor to be used at 2K. Also the pixel size is different (SI-2K is larger). The SI-2K is noticeably, dramatically sharper, but then again it cannot shoot in 4K and if it did with the same strength OLPF it would likely suffer from aliasing issues. There is another 2K camera around, the Phantom HD. It has a vastly larger pixel size, with each photosite 12.5 microns instead of the RED One's 5.4 microns. That's a huge difference. And it has an appropriate strength OLPF for it's size, which makes it a very sharp camera. In many ways it is as sharp as the R1 in 4K, even though it is a 2K camera.

Yes, that does make sense. the olpf is never going to be 100% efficient, and a 2k olpf design would make for soft 4k footy.. but when blown up to 2k, might not be strong enough..

 

[keicol]

So, we all agree that 2K is soft focus on all Red Ones. Exactly my point.

It specifically falls out of usefulness when trying to shoot wide shots with subjects at about 15' out.

 

[Roberto Lequeux]

Well not soft focus, no. Just less detail, and the why is still unexplained... however it is not in the optics so soft focus is not correct at all.