Full frame sensors and light falloff: real problem or design flaw?

[lanierb]

This is an optics question (Bob?). Everywhere these days people

cite "light falloff" on wide angles as being a weakness of full frame

cameras. My question is whether there is any theoretical reason that

light falloff would be worse for the same angle of view for a larger

sensor than for a smaller one. I always thought light falloff had to

do with the angle of the light rays leaving the lens. Wouldn't that

be constant across formats for the same angle of view?

 

Second: if there is no theoretical justification for this, then has

this conventional wisdom come to exist simply because there a design

flaw (slight vignetting) in the Canon wide angle lens lineup that

wasn't noticed before?

 

Lanier

[oceanphysics]

No, it's not the same, there is a real issue and it's not Canon's fault. The angle that the light hits the sensor is not the same as the field of view because wide-angle SLR lenses are retrofocus designs.

[gary_jean1]

I ahve always seen light falloff at the corners with wide angle 35mm lenses at maximum aperture. It is no worse on my 5D than it was on my 3. And, if it is objectionable (often it is not), it is corrected in about 5 seconds in ACR.

[daniel_taylor]

It's the same problem as on film for the same reasons.

 

If you look at a full frame or film image with vignetting, you see that it gets worse the further you get from the center of the image. Since APS sensors crop out the edges of the image circle cast by full frame lenses, they also crop out the worst of the falloff.

 

It's always been around, but now we live in an age where anyone and everyone can snap 1,000 test photos with their digital for free, study the results in Photoshop at 300%, and discuss them forever on message boards. So everyone has "noticed" this more, even though it's still only an issue in real pictures when you shoot wide open and the background is uniform and bright. Just like the film days.

[gary_ferguson1]

Daniel's absolutely right. But even if the problem is more apparent in the digital world it's also more easily fixed. With film you just had the option of expensive centre spot filters to cure vignetting. With digital you can still use the filters, or you can simply click a button in DxO and the problem's gone.

[PuppyDigs]

I too noticed it with wide angle lenses when shooting chromes and negs. It wasn't a

big deal as the slide mount covered much of it. Even if it was severe, by the time you

cropped the print to 8 x 10 or 11 x 14 most of it was gone.

[bobatkins]

All lenses vignette wide open and the wider the angle of view, the worse the vignetting is. It's a result both of basic optics and (to some extent) lens construction.

 

Is it worse with digital? Maybe. There are/have been some sensor designs in which the photosites are in shallow "wells". If this is the case then they can be partly shadowed by light rays coming in at a very oblique angle. I don't know if current sensors are all like this or if they've all solved that problem.

 

The major cause of most recetly discovered lens and camera defects is the ability to easily look at images at any desired magnification. The closer you look, the more defects you see.

[paul_repacholi]

It has always been there, but the fall off in the enlargers canceled

out much of it. Now you have the fall off in the camera, but no complementry fall off in the enlarger. A digital camera goes strainght to print, and a scanner generally only projects the neg image in one direction.

 

Have a close look at some of your old negatives, you will be suprised at how dark the corners are.

[henrik.ploug]

Here you can see how bad the vignetting is for different lenses on full frame and on 1,6 crop cameras - the rectangle in the middle is the cropped view:

 

http://www.the-digital-picture.com/Reviews/Lens-Vignetting.aspx?Lens=114&View=3&desc=Canon-EF-16-35mm-f/2.8-L-USM-Lens-Vignetting-Test-Results

 

If it doesn't bother you, get a full frame camera. If it does, don't.

[jamesdb_db]

Would one be able to say that it is Canon's fault in some sense because they could have used a larger image circle than the "minimum" required to cover the 35mm frame? That way even at 35mm the poorer edges would be cropped off. Of course in my mind this would mean building a lens with quality in mind, not just profitability.

 

James

[ben_kriete]

A little off-topic, but I recently saw a thread in another internet forum where someone returned a new 70-200 to B&H because at 400% magnification he thought he *might* be seeing chromatic aberration on a shot of a b&w test chart.

[shambrick007]

"Of course in my mind this would mean building a lens with quality in mind, not just profitability." <p>It would also mean building bigger/more expensive lenses to cure something that's not that big a deal (IMHO) and can be corrected w/software.

[digitmstr]

No, it's not Canon's (or any other company's) fault.

 

Producing an entire line of NEW lenses with a bigger image circle would be super-costly and you can use MF lenses on your EOS as it is, if that is what you want.

 

Light falloff is a fact of life with ALL current optical designs, from any manufacturer. Some specific lenses may exhibit more, some less but, it's a limitation of optics which is visible on both film and digital.

[daniel_taylor]

"A little off-topic, but I recently saw a thread in another internet forum where someone returned a new 70-200 to B&H because at 400% magnification he thought he *might* be seeing chromatic aberration on a shot of a b&w test chart."

 

Sounds about right. I feel sorry for customer service departments that have to deal with these types. People like this need to spend a year with a Holga.

[bobatkins]

A bigger image circle doesn't really help if the vignetting is caused by cos^4 theta falloff. Theta isn't a function of the image circle size, only of the format size and the position of the exit pupil.

 

Highly retrofocus lens designs can lower vignetting, but they have their own problems!

[lanierb]

Bob's response is closest to what I was thinking. "Vignetting" means that some physical part of the lens is causing light falloff. This can always be cured by making the lens bigger. The problem I was thinking of was the cos^4 light falloff issue, which is an optical phenomenon that cannot be cured. My hypothesis was that this problem would be identical on smaller format sensors for equivalent angle of view, but I'm still not sure about that. I know it can be fixed easily in Photoshop. I was just curious if there's something "real" to the issue.

[oceanphysics]

As I indicated, cos^4 falloff is not a simple function of viewing angle with retrofocus designs. That's the reason non-retrofocus rangefinder wide-angles require a center filter, whereas the same focal length in an SLR lens may not.

[jo7hs2]

Ben Kriete,

I second the Holga suggestion, but raise you a "but also make them use it in a zip-loc bag."

[michael_beckmann]

This alleged light falloff problem with full frame cameras is complete bs because the lenses work as designed with these cameras.

 

Of course the light falloff is less using the same lens on a 1.6x crop cameras, but that was not what the lens was designed for, so what is the point of this comparison. You might as well just crop your full frame image and enjoy less light falloff.

 

A lens designed for 1.6x crop sensors will have the same amount of light falloff on a 1.6x camera as the regular lens shows on a full frame camera.

[KenPapai]

Shot with film, 16-35L lens at 16mm, shows zero light fall-off "vignetting" at the corners (but I likely used 1/500 at f.8...).<div></div>

[mark u]

"Wouldn't that be constant across formats for the same angle of view?"

 

No. The angles of incidence of light rays depend primarily on two factors: the distance of the exit pupil from the film plane relative to the distance of the image point from the image centre defines the angle for the principal ray. The exit pupil is the apparent point at which, viewed from the back of the lens, all image rays cross over (often approximated by the apparent location of the aperture as viewed from the back of the lens). The second main factor is the aperture in use, since each point of the image consists of a cone of rays limited by the aperture focussed on to it. Very wide apertures produces light cones with a broad apex angle, so light strikes the image point over a wide variety of angles. Narrow apertures have a small apex angle, so all rays strike at angles that are close to the angle of the principal ray (approximately the centre axis of the cone). The narrow cone produces a searchlight effect that highlights dust on the sensor filter, while the broad cone produces a diffuse lighting of the dust that therefore doesn't cast sharp shadows on the sensor itself. An excellent discussion of the various causes of lens vignetting can be found here:

 

http://www.vanwalree.com/optics/vignetting.html

 

Those who claim that there is no difference between film and digital sensors as regards the consequences are paying no attention to the different contributions of the sensor optics compared with the film emulsion. All light sensitive layers of film emulsion, together with the layers that overlay them are typically no thicker than 3 ten thousandths of an inch according to Kodak (about 7.5 microns - similar to the size of a sensor pixel):

 

http://www.kodak.com/US/en/motion/students/handbook/identification2.jhtml?id=0.1.4.9.6&lc=en

 

Provided the film is held flat in the film gate (or at least well within the depth of focus), the effects of spreading due to refraction, diffraction and reflection within the film are very modest. There is almost no additional vignetting caused by light striking the film at a more oblique angle - vignetting can be regarded as entirely due to the lens.

 

By contrast, the sensor optics start with the filter layers (IR/UV and anti-alias) that are typically 1-2mm in front of the actual silicon. These layers exhibit a degree of birefringence (changes in refractive index with light wavelength). This can contribute to upsetting the lens' corrections for chromatic aberration with oblique rays. The anti-alias layers are designed to spread an image point over a 4 pixel area, so as to ensure that the colour of each image point is sampled in all three RGB channels. Immediately atop the silicon is the Bayer colour filter array, and above that are the microlenses, which are roughly hemispherical. If you actually ray trace a hemispherical lens, you will find that oblique rays are poorly focussed off to the side (and therefore not on to the photodiodes), whereas normal rays are focussed correctly - this obviously contributes additional vignetting.

 

Indeed, Leica designed the microlens layer for the DMR back with microlenses that are progressively offset like shift lenses, precisely to counteract this effect. Meanwhile, Olympus have built in software corrections to reduce the effects of vignetting on their 4/3rds system, based on parameters for each lens stored in its ROM, as well as adopting a design for their lenses that pushes the exit pupil further from the sensor plane. Canon have made no public comment (e.g. in the white papers they produce for their cameras) about any measures they may have taken to address the problems, but you can be sure they are aware of them.

[lanierb]

Thanks for the link Mark! That was super informative. So the cos^4 law depends a lot on lens design, which is a trade-off I wasn't aware of. I'm guessing that in the future, changes in sensor design as well as lens design will improve the light fall-off issue somewhat then.

 

I still have one question which I was not able to figure out though, which is: Holding lens design and sensor design constant, if you were to scale down the system by a factor of 1.6, is there any theoretical difference in light fall-off? It seems to be conventional wisdom that there is, but I'm wondering if it that perception has simply resulted from people using full-frame lenses on 1.6 crop cameras.

[michael_beckmann]

Lanier,

 

I have not seen a single case where the alleged excessive light fall-off attributed to full frame cameras was explained by anything other than finding less light fall-off when the same lens was used on a 1.6x crop camera.

 

Still the argument is repeated over and over, especially in amateur photo magazines. They makes lists about pros and cons of full frame vs. 1.6x crop cameras because their readers want to know the difference. Vignetting is always listed as a con for full frame cameras.

 

I guess this will be repeated till the end of time in camera reviews (online and in magazines), because the logic is so simple.