Cinematography Mailing List - CML
    advanced

Chromatic Aberration/Misaligned CCD's

 

21st July 2005

Hi all,

Could someone please explain to me how to tell the difference between chromatic aberration and misaligned CCD's? Is there a relationship between the two or am I chasing my own tail. I've done all the obvious things such as to put the lens in question onto different cameras and different lenses onto the camera in question. I've done various tests on different charts over & underexposed, various f-stops to wide open, extender on/of etc. etc.

I've tried the internet for examples of misaligned CCD's(unsuccessfully).

After doing so many tests with the different formats and lenses, I've made peace with the fact that I will have some form of chromatic aberration.

The lens in question is ENG style HD, and the camera SD

Jacques Nortier

Wildlife cameraman, South Africa


The problem I've seen the most is that out-of-focus objects smear magenta upwards and green downwards. This happens mostly around very bright objects and areas of contrast. I've been told that's chip misalignment or just an artifact of the chip block itself. I've seen such things happen with nearly every camera, from a D35 with an industrial lens to a Varicam with HD glass, so I'm assuming in that case it is indeed a chip block issue.

Usually the effect is pretty subtle, but I've had one experience where the problem was very dramatic.

What are you seeing?

Art Adams, DP [film|hdtv|sdtv]
San Francisco Bay Area - "Silicon Valley"
http://www.artadams.net/
Local resources : http://www.artadams.net/localcrew


>The problem I've seen the most is that out-of-focus objects smear >magenta upwards and green downwards. This happens mostly around >very bright objects and areas of contrast.

Art,

My observations over the years are:

- the green/magenta smearing is indeed only ever up/down, not left/right;

- the colour of the smearing is one way for dark to light transitions, the other for light to dark;

- the smearing only occurs on out-of-focus transitions;

- the smearing also occurred with three-tube cameras;

- the smearing doesn't happen with single-chip or single-tube colour cameras.

All this suggests its not a chip misalignment issue. Tube cameras could be registered perfectly with an in-focus test chart. If it were a chip-misalignment issue, why wouldn't we get left/right smearing with some cameras? And why wouldn't it show up on in-focus transitions?

My view is that its an artifact caused by the beam-splitting prism. All modern cameras fan their tubes/chips in a vertical fashion - which is sensible given the lesser prismatic diffraction needed for the smaller dimension of the aspect ratio [the "3" in 4:3 or "9" in 16:9].

I'm still keen to catch some archive footage from a Philips LDK-5 studio camera. LDK-5s were the only cameras I'm aware of that used a horizontal fanning of their tubes. My bet is that the green/magenta smearing on these cameras would have been left/right!

I'm also wondering if the cheaper "industrial" three-tube cameras that incorporated beam-splitting dichroic mirrors [no prisms] suffered from this problem?

Food for thought

Clive Woodward
Perfectly registered, today
Perth, Western Australia.


Usually the effect is pretty subtle, but I've had one experience where the problem was very dramatic.

What are you seeing?

Hi Art,

What I see or get is this very defined "hard edge" red line on anything with a bit of overexposure (less than half a stop of overexposure) whether the subject is in or out of focus . A good example is the following : If I expose correctly for a bird perched on a piece of string at midday, everything will be fine except for the top of the string which is a bit more overexposed, the string will have this "electronic" red line all the way through the frame. I can have this string at any angle the red line will stay.

Maybe I'm a bit to critical but on the other hand, all of the equipment in question are current top of the range HD & SD stuff. The problem is once you've noticed it you can't look past it. It didn't concern me as much in the past because I knew my long lens wasn't the best and made a plan to overcome the issue, frame it differently etc. But now that I've invested in HD equipment and still have the problem its a very big issue.

Jacques Nortier


I recently finished post on a feature shot with 3 Sony HDW-F900 cameras. Apparently the prism was misaligned in one of the cameras because we could significantly improve the picture quality by offsetting the Blue channel by a pixel (at 1920x1080).

One of the other channels was also slightly misaligned, but I opted not to correct this as moving it in a sub-pixel increment would soften it due to re-sampling.

I wasn't involved with the shoot or the capture, so the error could have been introduced during transfer and compression onto a computer, but I doubt it.

Tim Baier - Compositor
Sydney, Australia


Tim Baier writes :

>Apparently the prism was misaligned in one of the cameras because >we could significantly improve the picture quality by offsetting the Blue >channel by a pixel (at 1920x1080).

Pardon my ignorance, but how could this be introduced in capture? I'm guessing that someone could have manually changed the alignment, just as you did, but could it happen otherwise?

Brent Reynolds
DP / Producer
August Moon Productions
Tampa FL


You can see this with film lenses also; I have a 9.5-57HEC here, a good but not stellar lens which does the exact same thing.

I don't think it's so much the point to say the prism optics in a 3 chip camera "cause" this as it is the prism optics make it a challenge to correct for this aberration.

This would seem to be lateral color, not longitudinal aberration. (I'd love it if someone would confirm if I'm right or not about this).

(as an aside, I do wonder how much the vf optics in a film camera will exaggerate the perceived effect - as opposed to what the film sees)

Why green / magenta I wonder ? I'd guess they magnify furthest from the image plane in achromatic lenses.

Sam Wells


> You can see this with film lenses also; I have a 9.5-57HEC here, a >good but not stellar lens which does the exact same thing.

Normally Chromatic Aberration I see in video and film lenses appears as a blue halo around specular sources. I don't think I have never seen lateral color error.

A friend has a astronomical refractor (Takahashi) that has a fully corrected objective. It cost him a small fortune but images of the planets and star "airy discs" are nothing short of stunning...

>Why green / magenta I wonder ? I'd guess they magnify furthest from >the image plane in achromatic lenses.

Isn't this caused by the pixel offset video camera engineers use to increase resolution? The red and blue channels are offset from the green channel. I have only noticed this on a component monitor direct from a DigiBeta or BetaSP field tape. Watching at home on a good monitor I can't see it.

I could imagine HD being blown-up on a big screen could present a major problem.

Tom McDonnell
Dir/DP
New Orleans, La


>Isn't this caused by the pixel offset video camera engineers use to >increase resolution?

Hi Tom

What I do find interesting in regard to the lens and camera in question is the fact that I can manipulate the amount of "chromatic aberration" in the red channel by merely moving the light source closer or further away from the subject at any f-stop. If I expose correctly for the highlights I don't see any chromatic error regardless whether I am wide open @ f2.8 - 203mm or f5.6 - 406mm where one should notice chromatic aberration at its worst.

Jacques Nortier

Wildlife cameraman, South Africa


>What I see or get is this very defined "hard edge" red line on anything >with a bit of overexposure (less than half a stop of overexposure) >whether the subject is in or out of focus.

Wow! I don't know that I've seen that before. Could it be some sort of detail error? That's very funky.

Art Adams, DP [film|hdtv|sdtv]
San Francisco Bay Area - "Silicon Valley"


Sounds like your flares might be off. CCDs are glued to a prism and don't go out of alignment. If they were "off", you wouldn't even get a picture that was anything near what you described. I say 'off' because the odds of something like what is being thrown around here are pretty slim.

Walter Graff
BlueSky Media, Inc.
www.bluesky-web.com
Offices in NYC and Amherst Mass.


I see this in the raw 12-bit RGB feed from the A/D's on my DVX100.

There are different contributing factors...one of them is the fact that the CCD's might not be perfectly aligned.

The other factor is the beam splitter, which yields slightly different focus and image size for usually one or two channels. Since these cameras use on-board hardware to correct for these problems, it is sometimes too expensive to implement a resizing algorithm to account for the prism, so the images are just aligned as best as possible.

Whether this is normal for your camera or not, it depends. On lower-end 3CCD cameras, some shift is usually present. On the DVX it is visible in the DV footage under specific conditions. The correct alignment involves slightly resizing one channel and then aligning all three. The resizing is too computationally demanding, so it is not done by the camera itself, and this is why the magenta/green edges are sometimes visible in the DV footage.

To make sure, I would take your camera to whoever is supposed to service it, and have him/her check it. The CCD's themselves are usually not movable by hand, but he might determine that your specific model has too much shift...i.e. some impact caused the CCD's to move or maybe it came like that from the factory.

Juan P. Pertierra
Electrical Engineer, Mad Scientist.


>Pardon my ignorance, but how could this be introduced in capture?

Ideally I'd like to have tested the whole pipeline before pointing the finger at the camera. An error could theoretically be introduced during capture/compression by some sort of programming error in the codec, but I think that would be very unlikely.

Tim Baier - Compositor
Sydney, Australia


> If I expose correctly for the highlights I don't see any chromatic error >regardless whether I am wide open @ f2.8 - 203mm or f5.6 - 406mm

This could be something different.

Tom : I too would think lateral color should not be an issue either. But what am I seeing ?

Sam Wells


Juan Pablo Pertierra wrote :

> There are different contributing factors...one of them is the fact that the >CCD's might not be perfectly aligned.

Hey, Canon calls that a feature -- pixel shift -- to make lower-res CCDs in the XL1 appear to have more resolution.

I think it's an optical problem. You're dealing with complex lenses and a prism... artefacts will often appear (though never ever with fine Zeiss DigiPrimes from BandPro).

Jeff "just credit my account, Michael" Kreines


>I think it's an optical problem. You're dealing with complex lenses and a >prism... artefacts will often appear (though never ever with fine Zeiss >DigiPrimes from BandPro).

Hi Jeff,

Due to the kind of work I do (Natural History etc) its impossible to walk around with a bag full of Digiprimes. So I have to make peace with the fact that - as the Fujinon REP said to me "all lenses have chromatic aberration".

My concern is, at 38 000.00 US$ for a single lens, how much chromatic aberration is acceptable if the problem is indeed with the lens? Besides the fact that the lens manufacturer clearly states in its brochures/internet site that its lenses have NO chromatic aberration.

Jacques Nortier

Wildlife cameraman - South Africa.


Jacques Nortier wrote:

>Due to the kind of work I do (Natural History etc) its impossible to walk >around with a bag full of Digiprimes.

I know, that was just a little joke so Michael wouldn't have to tell us how great Digiprimes are...

Jeff Kreines


>Sounds like your flares might be off. CCDs are glued to a prism and >don't go out of alignment. If they were "off", you wouldn't even get a >picture that was anything near what you described.

Hi Walter, I did give the lens flare issue some thought and experimented a bit but changing the flare settings didn't do much. I must admit I do not have the proper equipment to set it up, and I do realise that if I were to set the flare properly I should get a better exposure latitude which begs me to ask the question. Why are the apparent chromatic lens aberration worse in high contrast area's? A simple answer might be, because its more difficult to contain disturbed detail.

My initial question was, what does the picture look like when the CCD alignment are "off", I don't have any reference. I'll appreciate it if you could refer me to an example

Jacques Nortier

Wildlife cameraman, South Africa.


>I know, that was just a little joke so Michael wouldn't have to tell us how >great Digiprimes are...

Thanks Jeff, I managed to giggle quite a bit after this reply...

Jacques "much too serious" Nortier


> My initial question was, what does the picture look like when the CCD >alignment are "off", I don't have any reference. I'll appreciate it if you >could refer me to an example

The problem is I never heard of CCD alignment being off. If it were, you'd get a lot more of a funky picture than what you are describing (more like watching a 3d movie without glasses. Tell me again what the camera lens combo is?

Walter Graff
BlueSky Media, Inc.


Jacques Nortier writes:

>So I have to make peace with the fact that - as the Fujinon REP said to >me "all lenses have chromatic aberration".

This is of course true. All lenses have every aberration. No lens is perfect. But it's a question of degree; and for $38,000, the amount of chromatic aberration in this lens should be just about invisible.

Also for that kind of money, Fujinon should be allow you to choose between several different lenses to see which is the best, not just take the one they hand you.

Is it possible for you to try another camera of the same type with the lens in question? Or other lenses of the same type with your camera?

Brian Heller
IA 600 DP


>This is of course true. All lenses have every aberration. No lens is >perfect. But it's a question of degree; and for $38,000, the amount of >chromatic aberration in this lens should be just about invisible.

This has not been my findings in my 20 years in the business. For years lens manufacturers have had to walk a fine line creating a lens that worked on a video camera while trying to maintain quality. For years that was a tough proposition as video cameras cost $40k and no one was going to may more than around a fourth of that price for a lens. So lens manufactures were trying to make quality lenses while dealing with keeping costs down for mass production. What you had was what I consider the weakest part of the chain in the lens. It seemed to me that for many years, regardless of the manufacturer I used, lenses were pretty crappy. Consider that these folks were trying to make a general purpose zoom (one size fits most). These zoom lenses have upwards of 18 elements in them. They try all kinds of coatings and formulations to help with linear and chromatic aberrations. It was easier in the days of tube cameras as one could adjust the target area of each tube effectively eliminating a lot of chromatic problems.

Then they came out with CCDs and lens manufacturers had a big problem; now it was more up to them to fix inherent aberrations because now your targets were glued to a Schott glass block. But you had to do it while maintaining the retail price of the lens because no one would pay for a lens that cost $20000 for a camera that cost $40000 in the video world. As I said for years I personally thought general purpose mass manufacturer video lenses were pretty bad; they had lots of port holing, etc. In fact in the days when the 79e was the defacto standard camera for video production in the US I learned just how bad many mass produced lenses were. We used to shoot commercials with this camera but used prime lenses and an adapter to make them work with the extended back focus of a video camera and the difference was AMAZING! That's when you could really understand just how weak video lenses were in the equation.

Move along to another aside. Many consider the Ikegami DVW-7 to be the best professional portable DV camcorder made. Problem when it came out was that it cost more to buy the lens than the camera body and that confused folks looking to make a purchase because their minds used to years of paying little for a lens now had to pay more for it. Some folks even said the camera isn't worth it because the lens cost more. Strange thinking but I saw many folks say that. But when you saw what you got for the combo, it was more than worth it, but many video folks are about a range of perceived cost rather than about perceived difference for cost.

Move along to HD. Now manufacturers had an even more serious problem, a target that produced a sharper image let alone they had to produce a vertically wider image. Now lens manufacturers had a problem, keeping a mass produced lens in the formula that video folks find acceptable for a lens/camera price ratio. I remember the first video lenses for HD were no better than what I saw in SD as in the same lousy portholing, aberrations etc. But as HD has been picked up more by film folks who would pay as much if not more for a good quality lens than a body, lens manufacturers have been given some more freedom to make better quality lenses while stretching the acceptable retail price. Now I see mass produced varieties that are much better, but still lack some quality but it's getting better.

Walter Graff
BlueSky Media, Inc.


Brian Heller : for $38,000, the amount of chromatic aberration in this lens should be just about invisible.

Walter Graff writes:

>This has not been my findings in my 20 years in the business. For >years lens manufacturers have had to walk a fine line creating a lens >that worked on a video camera while trying to maintain quality.

That video zooms were poor in quality compared to film lenses is a well established fact. However, as you point out putting a high quality lens on an ordinary camera can make for an astonishing improvement in quality. It works for film as well as video.

Put a Cooke S4, an UltraPrime or a Primo on a decent Arri IIc and no one would be able to tell that it wasn't shot with the newest camera.

As you say, film folks are used to often paying more for a lens than a camera.

However, $38,000 is film lens territory. And Paying $38,000 for a video lens should give you a lens approaching film quality in that there should be no visible aberrations.

That is Fujinon’ s claim and no one should pay that kind of money and accept an inferior lens -- if indeed it is the lens that is at fault. Either the lens is defective, or there is a problem with the camera.

Try another lens and see if the problem goes away.

Brian Heller
IA 600 DP


I wrote :

> Try another lens and see if the problem goes away.

What I meant to say was have Fujinon give you duplicate lens to test to see if it is the lens. They are an excellent company and should stand behind there products.

The salesman is blowing smoke by telling you: "all lenses have chromatic aberrations."

Brian Heller
IA 600 DP


>Why are the apparent chromatic lens aberration worse in high contrast >area's?

Is it possible the problem is not a lens aberration but a frequency response problem in one of the camera or viewing device circuits? Is it present using a different playback device and/or monitor?

It seems that going from high contrast bright to dark may be causing an overshoot or undershoot of the proper voltage. Admittedly this is usually seen as a white or black edge in the picture but I wonder if it is possible that poor frequency response in one of the color channels could present a specific color fringing.

I'm no electronics designer but maybe someone on the list might be able to confirm or deny this possibility.

Allen Emer
Lighting Cameraman
Holmdel, NJ


All lenses do have chromatic aberrations and a color problem that is only on vertical or horizontal lines in one color in a shot is not a lens problem but a camera problem. Folks seem to forget that camera registration is physical not electronic.

Some folks think that you buy a video camera and it never has to be tweaked for the remainder of the time you own it. None of this is true. I don't know your problem is but if you could send a jpeg of it, we could probably learn more and save the hundred posts we've made trying to figure it all out.

Walter Graff
BlueSky Media, Inc.


Walter Graff wrote :

>Some folks think that you buy a video camera and it never has to be >tweaked for the remainder of the time you own it. None of this is true.

Agreed. Digital innards have made video camera owners including me "confidently spoiled lazy" about camera maintenance other than normal tape deck and head service compared to the old tube or BVP-5 days. However, it is nice to have a camera maintain my exact dialled in look day after day after day after day.....

If I think a camera has a problem the first thing I do is save my menu settings and default back to factory presets. If this does not eliminate my problem then it's off to the shop.

Tom McDonnell
Dir/DP
New Orleans, La


Jacques Nortier wrote :

>What I do find interesting in regard to the lens and camera in question >is the fact that I can manipulate the amount of "chromatic aberration" in >the red channel by merely moving the light source closer or further away >from the subject at any f-stop.

Hi Jacques,

What type of camera/Fujilens combo are you using?

Tom McDonnell
Dir/DP
New Orleans, La


>What I meant to say was have Fujinon give you duplicate lens to test to >see if it is the lens. They are an excellent company and should stand >behind there products.

Hi Brian, just be careful, I never said that the lens in question is from Fujinon. I only repeated the comment of the REP from Fujinon in regard to chromatic aberration in general. As difficult as it is to stay away from mentioning product names - I'm trying.

Jacques Nortier


Jacques Nortier writes :

>Hi Brian, just be careful, I never said that the lens in question is from >Fujinon. I only repeated the comment of the REP from Fujinon in regard >to chromatic aberration in general. As difficult as it is to stay away from >mentioning product names - I'm trying.

My apologies for assuming it was Fujinon.. They're still an excellent company that stands behind their products. And no $38,000 lens should produce visible chromatic aberration, whoever made it.

Brian Heller
IA 600 DP


Clive Woodward writes:

>LDK-5s were the only cameras I'm aware of that used a horizontal >fanning of their tubes.

The earliest RCA studio color cameras (TK40 & TK41) and their immediate successors also used horizontal fanning. The TK40/41s were gargantuan, lumbering affairs with three big orthicon tubes (which occasionally exploded). I believe they used dichroic mirrors.

You might say they were TV's equivalent of the huge three-strip Technicolor camera. (And they required a similar light level.)

For some wonderful pix see :

http://www.kingoftheroad.net/colorTV/TVcams-in-action.html

Dan "history for the masses" Drasin
Producer/DP
Marin County, CA


Brian Heller wrote :

>And no $38,000 lens should produce visible chromatic aberration, >whoever made it.

Is this $38,000 US Dollars? That would have to be a studio lens rig at that price. No way a handheld HD lens is going to cost $38,000 or am I out of touch?

I find all recent SD video zooms, broadcast or industrial have chromatic aberration. I never use to notice it on older cams. I guess Walters right, the cameras are getting better but the lens technology is falling behind. I can't understand why however with CAD design and all the improvements with exotic glasses. Guess it all comes down to money, "No bucks no Buck Rogers".

Tom McDonnell
Dir/DP
New Orleans, La


> I can't understand why however with CAD design and all the >improvements with exotic glasses. Guess it all comes down to money, >"No bucks no Buck Rogers".

CADs a tool, nothing more. It doesn't design any better than a draftsman, only easier. It's the designer that does that (the human mind) And exotic glass is expensive just as exotic ladies are (Geoff can probably verify that). Now try to make back R&D and a margin selling 1500 lenses but know that you have to price them in a range that cuts your return or no one will buy them. The few CML’ers might but for 99% of people that shoot video, they just ain't gonna pay and hence manufacturers can only make things so perfect. I'm not saying they don't strive for the best only that there is so much tinkering you can do when you are watching the margin drop.

Don't confuse a brand with quality. Just because you have a name every one knows doesn't mean you always make 'the best' stuff. And just because everyone knows what the name is doesn't mean everyone wants it, just look at cancer, we all know what it is (great recognition) but no one wants it.

Walter Graff
BlueSky Media, Inc.


Tom McDonnell writes:

> Is this $38,000 US Dollars? That would have to be a studio >lens rig at >that price. No way a handheld HD lens is going to cost $38,000 or am I >out of touch?

I suspect the latter. Have you priced any DigiPrimes lately.

Anyway, I believe Jacques is talking about a film style zoom, not an ENG style.

Brian Heller
IA 600 DP


Brian Heller wrote :

>Have you priced any DigiPrimes lately

I can understand that. I'm not that out of touch...!

>Anyway, I believe Jacques is talking about a film style zoom, not an >ENG style.

A film style zoom? Ahh makes sense now. I can't imagine any lens sales person saying a film style lens would have CA. ENG version sure but Film version no way. Come on!, this sales rep has to know how picky DP's are.

BTW Did Jacques ever mention what camera/lens combo he was using?

Tom McDonnell
Dir/DP
New Orleans, La


I'm still researching this issue.

My preliminary explanation of why its always "green and magenta" rather than any other colours is that the first split in any prism block is to separate green out from what's left - red plus blue being magenta. Green is split first because its the predominant component of white light in encoded colour television systems [NTSC, PAL etc]. Remember Y = 59% green + 30% red + 11% blue [approximate figures]. So green carries most of the detail and hence its best to give it the cleanest, most direct path to the tube/chip.

More to come when I get time...

And Dan, Thanks for the link, fantastic site for anyone feeling nostalgic

Those cameras did, indeed, use dichroic mirrors, up until the "C" version which boasted : "TK-41C color TV camera, standard of the industry. New features for 1962 include stabilized circuits for simplified operation, precision yokes for precise registration, and prism optics for sharp clear color pictures." [Quoting from that web-site.]

Given the problems in keeping colour balance on a three-IO tube camera I expect green/magenta contrast transitions were the least of their worries in those days! But I'm betting the earlier dichroic mirror optics didn't have the problem.

>The earliest RCA studio color cameras (TK40 & TK41) and their >immediate successors also used horizontal fanning. The TK40/41s >were gargantuan, lumbering affairs with three big orthicon tubes (which >occasionally exploded). I believe they used dichroic mirrors.

The link is worth repeating for a trip down memory lane - and a reminder to PAL countries why they were better off delaying the introduction of colour to their airwaves

>For some wonderful pix see :
> http://www.kingoftheroad.net/colorTV/TVcams-in-action.html

Cheers,

Clive Woodward,
Perth, Western Australia.



Copyright © CML. All rights reserved.