comparison with a DLSR please?

[Álex Montoya]

The RED is truly an amazing camera.

 

[Phil D]

Haakon, again, great test. Thanks for sharing.

 

[Mike Gifford]

Excellent test, Haakon!

Phil B - check your PM

 

[jbeale]

getting the best focus

getting the best focus

My method was to use the long end of the lens combined with the "2x zoom" function to see what areas were sharp - and they look fine oversampled to the 720-1080 range (really, let's be realistic with what we're getting here!), but at full 4K they definitely suffer.

In the test I did on Jan. 8 using Red #103 with the Red 18-50 zoom, I noticed the focus would shift slightly when zooming tight vs. going wide. Perhaps it was just that particular lens, though. I was shooting high-contrast test charts from a distance of about 8 feet. I got the best results when focusing manually based on the LCD viewscreen at maximum aperture, in the 2x view, and then stopping down. I wish there was a 4x or 8x view.

 

[Rudi Herbert]

Haakon,

Great test man, thanks a lot really. Now, I have reread the whole thread and couldn't find an answer to what I'm wondering, so I'll ask you and apologize if it has been already explained. Do the Canon cameras already add a bit of sharpening? I don't know about Canon, but my Nikons are preset to add a 25% sharpening, which is removable, automatically. Could this account for the sharper look of the Canon shots? Also, as good a lens as the RED 18-50 continually proves to be, one has to wonder if the Canon glass is not a bit sharper anyway...

 

[Elizabeth Lowrey]

Thanks, Haakon, for your efforts and for such an informative comparison. Also have to say that I LOVE your shots of that temple. Nice eye!

 

[Häakon]

Thanks everyone for the feedback and compliments!

The RED seems to clip highlights a bit earlier than the DSLR's. In the city shot, although the RED shot seems darker, the whites are clipped while in the DSLR's they are not. But then again, these are shot with an original RED #79 so there could be some improvement with the new cameras. Or maybe there is some room in RAW processing.

I'm pretty sure there is room. Keep in mind that all of these images come in RAW and look so radically different from each other that something is bound to get lost in the translation when I matched them to each other (quickly, at that), the best I could. The RED images, specifically, are extremely flat and desaturated right out of the camera - which gives you good room to play - but it also means that pretty much everything you shoot has to be tweaked in some way. Perhaps I tweaked that particular shot too much, I don't know. In particular, I'm surprised the RED shots appear consistently darker here than the others - I matched them pretty well on this side. Maybe it's a color space thing? I hate color spaces... why can't blue just be blue?

What I can do is upload the .R3D files and .CR2 files for those who are interested and you can grade them away to your hearts content. Note that I wasn't trying to grade these images individually for any kind of particular mood (or even how I would present them), just try to get them to look as close as possible to each other for a basic comparison. There are a million ways to do these kinds of tests!

Noise seems to be very well controlled, more akin the 5D than the XTi which is quite amazing since the 5D's sensor is twice the size of the RED one's. And that's without the camera upgrade so it might even get better than that.

Keep in mind that the RED's "base" is 320 (moving to 100 or 200 can actually clip highlights), whereas the Canon camera base is 100. That could have put the two dSLRs at a disadvantage noise-wise; you'll notice that at a 320 rating, they are more sensitive than the RED, so had I shot them at 100 and matched the shutter speeds, they'd definitely be cleaner. My intent was to use the same variables (ISO, f/stop, glass, etc.) as much as possible, but again, with different cameras, different formats, and even different sensor sizes, nothing is going to be an exact match.

Any focus position between 20'-infinity will compose an overall sharp image. I would suspect 1 of 2 things. 1 a badly collimated focal flange distance so your FFD may be an issue? 2 Jim told us no sharpening is added so that may just be the way they are.

I definitely shot the cityscape at an f/11 between 20' and infinity. I remember that it was not set all the way to infinity, but I did double and triple-check the monitor and it was sharper to my eye just a hair pulled back. I know that some of the original batch of cameras did have collimation issues, but I have been shooting regularly and checking distances each time and my results have been spot-on until this point. I even uploaded a 2K sequence (LINK), shot with the same lens and camera just a couple weeks ago and it looks tack sharp to me. However, this 4K TIFF from the same shoot - although in focus - exhibits quite a bit more of that softer look. I think it largely has to do with the things that you, Anders, and Deanan have mentioned - that the camera is tweaked best for moving pictures and not still images, does not apply any sharpening in-camera, and has properties that don't lend themselves best to *this* kind of a test. It is rare that I would shoot any motion picture at 1/2000th of a second!

Haakon, do you live in Coronado?

Nope, but good eye! My parents live in San Diego (it's where I grew up), and I was just visiting for the holidays. I'm back up to L.A. tonight to shoot more "real" stuff.

In the test I did on Jan. 8 using Red #103 with the Red 18-50 zoom, I noticed the focus would shift slightly when zooming tight vs. going wide. I got the best results when focusing manually based on the LCD viewscreen at maximum aperture, in the 2x view, and then stopping down. I wish there was a 4x or 8x view.

I know that if backfocus is an issue, changing the focal lengths might throw things off - however I shot several of these images *at* 50mm, so I was judging focus at the same distance I was shooting (I just happened to be zoomed in already). That being said, I wish there was an 8x view also. I've been begging for it for months.

Do the Canon cameras already add a bit of sharpening?

The camera doesn't (I have the setting disabled, anyway), but I believe the RAW conversion software does. Whatever method it uses, it's a lot more pleasing to my eye than the results I get from various sharpening methods in Photoshop. I can see that it would be helpful for some to see the RAW, untouched files, so I will provide those. The thing to remember is that even the RAW images have been shot a certain way - based on the ISO, shutter, lens, etc., and while I tried to match them the best way possible, it's not an apples-to-oranges comparison. RED RAW != Canon RAW. That's why I tried to match the images myself, because in an untouched state, they look like they were shot completely differently.

Thanks for the questions - it seems that even though it's difficult to make a test like this "perfect," it has been helpful for many and thus was useful to have carried out. I've learned a lot from the process myself - both about the limitations and advantages to RED as well as how I would approach shots like this in the future. Most of all, I'm just really floored by how much dynamic range is present in all of the cameras - we are definitely in a new era of hardware than we were even just a few years ago!

 

[jbeale]

the reason for only 2x "focus assist" view (?)

the reason for only 2x "focus assist" view (?)

...I wish there was an 8x view also. I've been begging for it for months.

It just occurred to me... if the active sensor area (including the "look around" part that isn't recorded) is 4520 pixels, and you are displaying that on (for example) 1130 active image pixels horizontally on a 720p LCD viewscreen, that is a factor of 4. It means you can zoom in by a factor of 2 and still only need 1/2 the resolution of the sensor. Meaning, right now the camera is not doing any debayer process for the viewscreen. There is no need for it. BUT if they zoomed in any more, they would need to either do some kind of debayer, or interpolate, or duplicate pixels. If they did the latter, I can just imagine the forum comments about Digital 8, Fischer-Price, etc. They may not have room in the FPGA for a good debayer algorithm for this purpose.

Now, that's just a guess, but it makes sense to me. Or maybe all this has already been explained, and I missed it.

 

[balaclava9]

hi all, i'm responding to the DSLR RED comparison idea. i was the still photographer for Asylum Seekers. Off Hollywood's first RED production.

http://www.reduser.net/forum/showthread.php?t=5327&highlight=asylum+seekers

i shot a nikon D80 and on old leica M3 with fuji ProH 400 and a little large format too. the dp shot alot of scenes at asa 320 or 500 at f/2.5 or so. i would quickly ask the ac what the rating and f-stop were, and i would dial in my iso and aperture to match. then i would set my shutter speed to 1/50 and when i did that, pointing into the scene or the sweet spot of the lighting setup, i would get just beatifully exposed images.

in that sense the D80 and the RED were pretty comparable. the image i would get and the images i've seen from the footage are very similar. of course the RED is capturing at 24fps not my still rate! the DP's lighting was excellent, so i wasn't comparing failure limits between the RED and the D80. he never had the red anywhere close to failure limits, except for maybe at the end of the day in this one forest where we were shooting. and at that point he just called it, anyway.

i noticed in the dailies that certain scenes were pretty high on magenta. just out of curiosity, i was matching my still to the daily, and i had the tint in adobe camera raw at like +50 magenta. many other scenes were +0 magenta and +10 magenta to match my RAW file (with WB at either day or tungsten) with the daily. still noise is not at all an issue. if anything, i think my d80 was slightly noisier than the RED. i only looked at a few comparison frames that the on-set red tech had on his machine while unloading the cf cards.

another interesting comparison was to my fuji 400h. the response curves are different in film. the highlights, among other things, are different. still i had my old leica at 1/50s or 1/60s (my asa was always 400 rather than the red's 320 or 500) and f2 or f2.8 to match the whatever the RED was doing. these shots look fantastic. it occurred to me that it may be that you can be more liberal with putting highlights into a composition when you have fuji proH in the camera rather than a sensor.

of course the scenes were frequently surrounded with lots of darkness, and the still photographer's problems multiply (ISO1600, pushing RAW files and film in post, handholding 1/15s, etc.) when pointing into the darkness behind the scenes!

 

[mezmo]

Hi Haakon,
Yes no sharpening in the Red camera and the OLPF will give
you that slight off-focus look prior to post sharpening.
Primes may be a better deal for wide shots as a zoom can go out
if dumped in it's case a liitle heavy or not secure or supported in it's mount
when shooting in rushed conditions. Primes tend to sit in a mount better
and be less focus critical due to less glass and in some cases, better quality glass. Primes are easier to scale and all lenses should be on-scale before going out to shoot. This is a big deal with Red as eye focus on a 720p LCD
may not be totally reliable. If scale is correct on your lenses just set to
a calculated distance and shoot. This could be the only go, given the OLPF effect , with eyefocus, how do you know where 'tack sharp' really is.

Mezmo

 

[FELIPE.AGUILAR]

Amazig Test Häakon! Thanks!

 

[Michael Mann]

Thanks a lot for your tests, Häakon!
Frankly, I find the RED images much too soft. Even after adding much too much sharpening (500% unsharp mask) in Photoshop I don't get close to the sharpness of the 5D-image. For instance "Union Bank" (Diego skyline) and the other building signs are simply not readable on the RED image, even after maximum sharpening.
Is it just me?

 

[Phil D]

This is not an attack, nor am I a Red fan boy but...

I do feel sorry for RED, it seems to have has been compared, scrutinised and tested against every known photographic capture device ever invented, and if it doesn't measure up to them all, is seen to be lacking some how!

It would have been interesting to see as part of this test what sort of sharpness was present in an equivalent (asa rated) 35mm film frame, as that is what RED is really looking to emulate, or dare I say, supersede.

As interesting and as informative as Haakon's excellent test is, lets not forget the 100% pixel for pixel blow up he did are the equivalent of watching a film with your nose press hard against the screen. Which your never going to do. (unless your a fly)

Considering the state of most cinema prints, which lets face it, are pretty dreadful quality wise by the time they arrive in your local multiplex, digital cinema is by comparison, going to make a huge difference to the viewing pleasure of the majority of the cinema goers.

I guess what I'm really trying to say is, does it matter if its not a pin sharpe image?

 

[sander kamp]

This test was intended more for people like me, who know DSLR's very well and wanted an indication how they compare to RED. It was not intended nor set up as a 'who is better' test. After all, DSLR's can't make movies.

The softness is interesting, though. I didn't find it in the shot of the flowers but the cityscape was noticeably softer than the DSLR. It could be a matter of focus, lens, different OLPF filter or wavelet compression.

Knowing what the reason is is interesting as a way of getting to know the camera and getting the most out of it, like knowing how much sharpening should be applied in post. It might also be critical for people who are planning to crop in post or who are doing special effects.

But anyway, since this was not a scientific test it can hardly be used as critique on RED. This is more about the 'feel' of the images, about wether they are comparable in color and dynamic range to a DSLR. And I think they are, which is good.

 

[Phil D]

Very true. I have a very bad habit of hijacking threads, which I must stop.

As a comparison test I think what Haakon did is incredible useful, I didn't mean to imply it wasn't. I was just musing about the tests red has been subjected to in general. Please don't take it personally.

Long may these tests continue!

 

[Häakon]

It would have been interesting to see as part of this test what sort of sharpness was present in an equivalent (asa rated) 35mm film frame, as that is what RED is really looking to emulate, or dare I say, supersede.

Hang on tight, it's coming.

 

[sander kamp]

Cool, Häakon! These tests are a lot of fun

 

[mezmo]

This is not an attack, nor am I a Red fan boy but...

I do feel sorry for RED, it seems to have has been compared, scrutinised and tested against every known photographic capture device ever invented, and if it doesn't measure up to them all, is seen to be lacking some how!

It would have been interesting to see as part of this test what sort of sharpness was present in an equivalent (asa rated) 35mm film frame, as that is what RED is really looking to emulate, or dare I say, supersede.

As interesting and as informative as Haakon's excellent test is, lets not forget the 100% pixel for pixel blow up he did are the equivalent of watching a film with your nose press hard against the screen. Which your never going to do. (unless your a fly)

Considering the state of most cinema prints, which lets face it, are pretty dreadful quality wise by the time they arrive in your local multiplex, digital cinema is by comparison, going to make a huge difference to the viewing pleasure of the majority of the cinema goers.

I guess what I'm really trying to say is, does it matter if its not a pin sharpe image?

Hi Phil,
Yes it does matter,
Sharp is one of the most important things the Red system has to do,
camera and post processing. This is a 4K camera and sold as such.
The advantage of shooting 4K is increased resolution.
If it can't achieve this with certain lenses and shooting conditions we need
to know now and move towards a solution.
Cameras like the Arri D20 have an optical viewfinder and output high quality 1080p for reference. that's how serious Arri are about sharp.
Also The film prints you see in a theatre have gone three or more
generations before you view them. the original Neg would have been pin sharp.
If Red material starts off just a little bit soft it will be mush at the end of a
multi-generational film-out and print run. We don't want that do we.
Let the tests continue and I hope this helps. Mezmo

 

[Phil D]

Yeah I agree, obviously Red should be able to capture images that are sharp and in focus, that goes without saying.

Hi Phil,
The film prints you see in a theatre have gone three or more
generations before you view them. the original Neg would have been pin sharp.

That's sort of what I was saying too. The print a director and DP sees and the the print the average Joe sees are poles apart, because, like you say, the multiple generations the print has gone through.
So it should be possible for digital cinema to deliver digital "prints" that are much closer in final delivery to the quality that comes out of the camera.

But that said, I worked on Corpse Bride an animated feature film that was shot on Canon DSLR's and used Nikon lens'. The work flow was not that different to that of Red One, but with much less dynamic range. I saw both a digital projection of the film, and a version that was printed to film stock for final cinema release, and I personally preferred the film print version because of the softness it introduced to the image.

So I,m thinking that absolute pin sharpness from the sensor is really not that desirable. Maybe the fact the RedOne looks so good is because of this perceived "softness" which gives the images it produces a more organic feel and look. Just a theory.

 

[sander kamp]

Oh wow, Corpse Bride looked amazing. First movie made with a DLSR, right?

Interesting observation about preferring some softness for moving images. Moving images are different from stills and perhaps they should be judged differently.