Acceptable level of color fringing, Cooke/Zeiss

[Andrew M.]

Today I tested Cooke S4/i primes 14, 27 and 65mm
The tests were identical to the test we did with Zeiss R8 last week, see post:
http://www.reduser.net/forum/showthread.php?t=1381

Without going in to the details, I am disappointed with chromatic aberration of both lenses makes.

Anybody knows what is the acceptable chromatic aberration expressed in Lp/mm please?

Cooke has bit less chromatic aberration as compared with Zeiss but noticeably lower resolving power at the periphery of the frame.
65 mm had much better performance then anything I tested so far a specially on the chromatic aberration side, but anything on the wide angle side 28 and down is less then 100Lp/mm

I am scared to do the ultra light zoom tests of both, Zeiss and Cooke, I am already disappointed.
Maybe I am doing something wrong so I went to the source for some answers.
So far, the only what I got is, that they will get the new lenses, hand picked for the test.

All the lenses tested so far in 8 to 27 mm are great for 2K format, but for 4K we will have trouble to find something that will not degrade the resolving power down to 2K. I will keep searching.

[steevo435]

Andrew-
How are you testing these lenses?

[Andrew M.]

Andrew-
How are you testing these lenses?

Like any rental shop will do the test if you ask them to do it.
Putting the lenses on the projector, projector has test chart, laser etched transparency in it and projecting it on the white screen 5 to 14’ away depending from the focal length of the lenses tested. The background of the test chart is black (not transparent).
Andrew

[steevo435]

Without going in to the details, I am disappointed with chromatic aberration of both lenses makes.

I am scared to do the ultra light zoom tests of both, Zeiss and Cooke, I am already disappointed.

All the lenses tested so far in 8 to 27 mm are great for 2K format, but for 4K we will have trouble to find something that will not degrade the resolving power down to 2K. I will keep searching.

Andrew-
You will always see some things on a test projector that you will never see in an image....the projector is a great tool, but is somewat subjective to opinion. You're never going to get pure white light through a projector.
You are looking at the best of the best lenses- They don't get better than the Cookes or Zeiss. The Angenieux Optimo's are excellent zoom lenses, too.
Chromatic Abberation has become such a buzz word in this industry...it's much worse when it comes to 2/3" lenses because of the 3 ccd design. The Red, Green, and Blue ccd's are at different depths (only by a few microns, but different) in the ccd block, and the lenses designed for the 3ccd system are designed to "focus" each color on each of the 3 ccd's. It's never been much of an issue in motion picture because of the single plane design.
Unless you are dealing with a MP lens that has been dropped or is seriously out of alignment, CA shouldn't be much of an issue.

[Bruce Allen]

Hopefully CA won't be much of an issue, although I do see it on 1-CCD digital SLR still camera shots... even with high-end lenses. Also, the 2K Nikon Red example that was posted on Dvxuser a while back showed some CA. Even the beautiful 4K Milk Girls still shot with a Cooke shows a little bit of fringing. Not much but if you're obsessed about doing a greenscreen key at 4K for some reason, you might want to compensate. So I think we still need to watch for CA in our lens choices.

I agree the average 3-CCD camera's CA sucks way more, though! When doing greenscreen work FX1 footage, I most often end up running separate lens distortion corrections on each channel just to get everything to line up for a better key. It looks nicer, too.

Bruce Allen
www.boacinema.com

[Andrew M.]

but if you're obsessed about doing a greenscreen key at 4K for some reason, you might want to compensate. So I think we still need to watch for CA in our lens choices.

most often end up running separate lens distortion corrections on each channel just to get everything to line up for a better key. It looks nicer, too.

Bruce Allen
www.boacinema.com



Bruce, I couldn’t agree more with you here.

What I see happening is that we will get in to the 4K and in terms of green screen keying and sharp details on wide screen shoots, we will be still at 1080p, even with very good lenses. More I search on this subject and more I learn, more I see that shooting on film was optimized for maximum 1000 lines (vertical) on final copy but masters were more like 1500 lines equivalent. I know for film guys it sounds strange but this is where the whole industry is going and we have to learn to count pixels and lines now.
Now we have 4K coming with 2000 lines (vertical) equivalent and there are no lenses out there to do it right. So we have to use a different, sometimes time consuming technique, till lenses will catch up with sensor technology.

First, we can try to shot all at T5.6, forget about this Master Primes nice T1.3
Second if you use wide angle lenses and getting a lot of details in to the frame, avoid high contrast details outside of the center.
Next, don’t freeze the frame still, keep moving, even very very slow pan will do.
Interior shots, better have the good lights for these T5.6
Don’t forget that Bayer likes greens and doesn’t like stripes and fine patterns.
Experience with 4K will bring more of these techniques out, so soon we will have all tricks out there in the open.

I think in few years 22MP sensors will be very popular and once there is a need for better lenses lens manufacture will deliver.

If this will not happen then 4K format will be just a good format to capture it in and downers it to 2K so 2K looks nicer.

The one thing I learned though, that caught me by surprise, that film final copies in best are 2K equivalent!!

[steevo435]

If this will not happen then 4K format will be just a good format to capture it in and downers it to 2K so 2K looks nicer.

The one thing I learned though, that caught me by surprise, that film final copies in best are 2K equivalent!!

Shot with Ultra Primes, 4k DI:
http://www.laserpacific.com/projectGallery.html

http://www.filmlight.ltd.uk/content/c_northlight.html

http://www.digitalcinemasociety.com/downloads/HowManyKsDoINeed.pdf

http://www.thomsongrassvalley.com/products/film/spirit_4k/

[Andrew M.]

Steevo, good find, so we were watching 2K all this time in the cinemas.
4K is just for DPs when they watch master copies.
I guess not anymore. 4K is coming to the place near you.......

[Bruce Allen]

The one thing I learned though, that caught me by surprise, that film final copies in best are 2K equivalent!!

Andrew, the res of film is debatable but I do agree with you that there are a lot of fantastic movies, cinematography Oscar winners, etc that are 2K at best (eg Master and Commander which won but was a 2K finish). I have stated elsewhere that 99% of movie trailers are done at 2K or less. I have done graphics at 4K for only one trailer (Ocean's 13 teaser), but even then I'm not positive that they actually finished at 4K. Did it look much sharper than all the other (presumably 2K) trailers? No. I didn't really notice a difference - and I saw it at a good cinema.

Bruce Allen
www.boacinema.com

[Andrew M.]

Bruce, I agree with you, the 8mm or OMNIMAX can equally impress people as a good/story/acting material, but if you are in the nature/science stories most of the time, you can impress the people much easier, showing tons of details on the screen so your home theater screen looks more like a window to the outside world not like a TV screen and people go wowww! Though you need wide angle lenses most of the time for this wowww!!

Andrew

[Bruce Allen]

Okay, I admit, I like the wowwww feeling too :)

Bruce

[Andrew M.]

Bruce, I think Jim said once that RED will differentiate between babies and the men.
I think what he meant was, that we have to get some practice, practice and practice again.
I will add to it, we have to get experience/knowledge and tons of it.
Top guns in this industry will always push envelope even if equipment will be 10 times exceeding the needs.
There will be a trick or two that experienced guy will use to push whatever is available out there even bit further.
So far I see that industry is producing only as much as required minimum, I guess it is good business to do so.
That is why we have to learn how to go around the obstacles, and it looks like for every obstacle, there is way around it. We just have to learn how to do it.

BTW are you going to NAB next week?

[Michael Hastings]

We've talked about the possibility of post processing to remove CA.

Anybody notice Panasonic's new AG-HPX500 which has built-in real time CAC?

AG-HPX500 2/3" 3-CCD 16:9 P2 HD $14000

Chromatic Aberration Compensation function allows the camera to automatically compensate the registration error that is caused mainly by lens chromatic aberration and minimize the circumjacent blur.
The CAC is lens specific - Panasonic and Fujinon and Canon developed special (lower cost) lenses with the info on CA for that specific lens.

Maybe this can be added to future REDs - obviously it would need to be a really fast processor to handle 12 times the number of pixels and would have to have lens info - but maybe this could be done for the RED lenses. Andrew: somehow we have to convince Jim that his next prime should be a 8-10mm.


Hopefully CA won't be much of an issue, although I do see it on 1-CCD digital SLR still camera shots... even with high-end lenses. Also, the 2K Nikon Red example that was posted on Dvxuser a while back showed some CA. Even the beautiful 4K Milk Girls still shot with a Cooke shows a little bit of fringing. Not much but if you're obsessed about doing a greenscreen key at 4K for some reason, you might want to compensate. So I think we still need to watch for CA in our lens choices.

I agree the average 3-CCD camera's CA sucks way more, though! When doing greenscreen work FX1 footage, I most often end up running separate lens distortion corrections on each channel just to get everything to line up for a better key. It looks nicer, too.

Bruce Allen
www.boacinema.com

[Bruce Allen]

BTW are you going to NAB next week?
Andrew, no NAB - booked solid doing more movie titles 'n trailers. Hopefully will meet you guys next year. Have a great party for me. I think the socializing will be better than the announcements - we all know Red will utterly rock and they will surprise us with cool stuff, so there is no news there ;) Unless the other camera companies announce something truly huge, I don't think I'll find their stuff too interesting either...



Anybody notice Panasonic's new AG-HPX500 which has built-in real time CAC?


No, that is very interesting!



The CAC is lens specific - Panasonic and Fujinon and Canon developed special (lower cost) lenses with the info on CA for that specific lens.

Maybe this can be added to future REDs - obviously it would need to be a really fast processor to handle 12 times the number of pixels and would have to have lens info - but maybe this could be done for the RED lenses.

How about this for a compromise - put this functionality in RedCine? With the Cooke /i system built in, the lens data is embedded as metadata into the footage, right? So RedCine could say "ahh" you're using this lens at this f-stop, so we need to process CA accordingly...

One of the problems with doing de-CA in-camera while simultaneously recording compressed RAW is that with RAW, you're storing only red, green or blue data per pixel (because that's what the Bayer sensor sees). If you start doing CA fixing in-camera and the green pixels on the outside need to move inwards by 2 pixels, then you're screwed because now you have to start recording RGB for every pixel. Panasonic doesn't face this because they're not recording RAW anyway.

Bruce Allen
www.boacinema.com

[Andrew M.]

And then you bang the lenses a bit and CA will loose the lenses CA key patterns.
Doesn’t take much, just one pixel off.
The unmapped CA doesn’t work good.
Also on PL mount you have to make sure that lenses are mounted always the same way, pointing the mark up.
I wonder if they have CA auto-map function on some test pattern?

[Bruce Allen]

Very true, Andrew! Okay, here is my CA auto-map program proposal: It would be very simple - it would have a drop-down menu box where you choose the lens, then a "calculate CA for this lens" button. When you hit the button, the program would start with the defaults for the lens, then try different distortion values for r, g, and b and see which one gave the lowest sum amount of pixel difference between channels (starting with a portion of the image for speed, then moving to evaluate the entire image later) - kinda like using the difference transfer mode when lining up two images in After Effects / Photoshop.

Once it finishes, you have an option to tweak the CA fix, then hit OK. The program then stores this info to a little database file - hopefully it would then try those updated values next time you put in the same lens, with the hopes that the CA hasn't changed. Also, it would remember which ones you use most, so after a while you wouldn't have to tell it which lens you were using - it would just take a look at the image and most of the time be able to auto-detect the CA by running through the top 10 lens CA fixes already in the database. It would also be cool if it uploaded this CA database to the Internet so that gradually it would be able to recognize ballpark CA values for most lenses (would probably need to add a "this lens has been bashed and is not typical" button ;)

Agreed, a way to auto-map the CA on a test pattern would be perfect. Eg just shoot the test pattern with all of the lenses you're going to be using on the shoot, put this at the head of the Quicktime you're feeding to the de-CA program, and then it will a) recognize and calculate de-CA values for the lens set and then b) go through the shots (via scene detection or hopefully some metadata from RedCine / your editing package), checking to see which lens each shot was done with (via the sum difference method), de-CA-ing merrily.

What would be best would be if this were built into RedCine or offered as a plug-in, of course. Graeme, you write plugins... are you going to let people write plugins for RedCine?

Cheers

Bruce Allen
www.boacinema.com

[Andrew M.]

And you know what, good CA mapping/correcting program will be worth 10K or more.
Just imagine how much money you will save on the lenses.

[Bruce Allen]

Okay, Andrew, maybe you guys will convince me to actually write this.

By the way, if any other people who know how to code are out there and thinking of doing the same thing, please send me a message! Then I can stop coding it myself ;) I want to make films, not write code, but this is something that needs fixin'...

I also have a nice day job that pays decently (even though I still can't afford a Red ;), plus my attempts at directing take up what's left of my free time... so please don't expect something overnight!

Bruce Allen
www.boacinema.com