[button]

Hello everyone,

Can those of you who own both a k10d and k100d (or other Pentax dslr) try this experiment for me please:

Shoot a high contrast scene with a lens prone to PF (DA* 16-50, DA 12-24, etc) with the k10d, and find some bad PF- example: tree branches against a noon sky, white on black in harsh light, etc. Then, put that same lens on your other Pentax, and reshoot the scene (same focal length, apeture, ISO, and so on, of course). Please post 100% crops of the k10d w/ PF and the same crop from the other camera. I'd do it myself, but I have only the k10d.

Reason: I suspect the k10d is prone to PF. Interestingly, as shown on Ken Rockwell's site, PF from a Nikon D200 (same sensor as k10d) disappears on a D300 (new Sony 12 mp sensor). I also read anecdotally that PF may not be a problem on the Pentax 6 mp sensor cameras. If the culprit is the k10d, then we can at least hope for improved (less) PF in the next generation, simply by virtue of its new sensor.

Thanks,
John

[metalfab]

Just from seeing images from both cameras and numerous lenses (on PBase.com alone) its the lens, not the body of the camera.

[MPrince]

Purple fringing - Wikipedia, the free encyclopedia

[carpents]

I tested the FA77/1.8 Limited on the K10 against the *istDS, and couldn't detect any difference.

[mer]

Its from both the lens and camera , more from the lens though .

The digital sensors , in simple terms , multiply the purple fringing because of the reflectivity of the microlenses . Film doesn't show that much purple fringing .

Thats how I understand it .

Have fun

[ryan s]

And we all know how trustworthy Rockwell is

[codiac2600]

Hello everyone,

Can those of you who own both a k10d and k100d (or other Pentax dslr) try this experiment for me please:

Shoot a high contrast scene with a lens prone to PF (DA* 16-50, DA 12-24, etc) with the k10d, and find some bad PF- example: tree branches against a noon sky, white on black in harsh light, etc. Then, put that same lens on your other Pentax, and reshoot the scene (same focal length, apeture, ISO, and so on, of course). Please post 100% crops of the k10d w/ PF and the same crop from the other camera. I'd do it myself, but I have only the k10d.

Reason: I suspect the k10d is prone to PF. Interestingly, as shown on Ken Rockwell's site, PF from a Nikon D200 (same sensor as k10d) disappears on a D300 (new Sony 12 mp sensor). I also read anecdotally that PF may not be a problem on the Pentax 6 mp sensor cameras. If the culprit is the k10d, then we can at least hope for improved (less) PF in the next generation, simply by virtue of its new sensor.

Thanks,

John

The D300 has software CA removal which knowing him he probably has turned on and didn't know it.

[barendvl]

The D300 has software CA removal which knowing him he probably has turned on and didn't know it.

quote from KR:

What is this weird juju? This even works on my old manual focus 8mm fisheye. Is it a magic sensor, magic firmware keyed to the specific lens (sorry 3rd party lens users) or is it clever firmware that looks for odd fringes on any image? No one knows, but it does work, and works perfectly with every lens I've used on the D300, both auto and manual focus.

My reaction: LOL.

[Olrocker]

I just got a DA 18-250; it shows some pretty heavy PF at high contrast areas, similar to what I'm getting on my FA 35/2. I'm starting to think it could be the sensor. Is this something I'll just have to live with, or should I have Pentax check it out when I send it in for "Tilted Image Syndrome"?

[Lowell Goudge]

largely chromatic aborations from the lenses, and if you strongly over expose some also from the sensor.

Good editing programs like PSP X2 and the adobe series of products can remove CA in post processing

[wlachan]

How about some PF on film?

FA*24/2 chromatic aberration photo - Alan Chan photos at pbase.com

[falconeye]

Purple Fringe can come from both, the lens and the sensor.

If it is the lens, the direction always is the same if seen from the center (i.e., if on the left side of the image, the PF is on the left border of a dark forground against a bright background, it is on the right side on the right side of the image).

If it is the sensor, it will always be on the same side irrespective of its location within the image.

*istDS and K20D (I have no experience with the others) have no detectable PF from sensor (as far as I can tell), except that the K20D has strong sensor PF in images shot in burst mode.


Tools which remove PF may do it by masking border areas and reduce the magenta or green color tones. However, this may have a bad influence on image quality overall (I tried it with DxO and found it useless for K20D burst images).

Tools which remove PF may also do it by optical correction algorithms for a particular known lens/body combination. DxO claims to be able to do it for a few Pentax lenses. The results may be better but I have no experience with it.

[falconeye]

Good editing programs like PSP X2 and the adobe series of products can remove CA in post processing

Lowell, how do you do it in PS CS3 (not ACR)? I mean, except adjustments to some color channel saturation values?

I am actually dissatisfied with the state of the art. Even if PF is reduced by fidling at the colors, the resolution will still be negatively impacted. It wouldn't be too hard to derive the mathematical formula of abberation from observed PF within an image(*) and apply the inverse function to remove it. Do you know of any tool able to do it? Should give superior results compared to current methods (i.e., would even deliver improvements when viewed b&w).

EDIT:
(*) Ok, for the curious of us... The red, green, blue channels, when evaluated separetely from one another and when plotting intensity against radius (distance from the center) would give three very similiar curves with SLIGHTLY diverging scale. Adjusting scales to match and recombining into one image would maximally reduce the PF.

[blende8]

If it is the lens, the direction always is the same if seen from the center (i.e., if on the left side of the image, the PF is on the left border of a dark forground against a bright background, it is on the right side on the right side of the image).

No.
See image below.

PF is a longitudinal CA effect.

The cause of PF lies in the lens.
Almost all lenses are not very well CA corrected for wavelengths below 450nm and above 750nm.
Fortunately the sensor is not very sensitive for these wavelengths (< 0.1%).
But if you drastically overexpose the sensor by say a 1000 times, this <0.1% becomes visible in the images as defocussed far blue and far red light, which combines into purple.

[falconeye]

No.

Hmmmh.
We can probably agree that we used different definitions for purple fringe.

(1) In your definition (PF = longitudinal CA) you are right.

(2) In my definition (PF = purple-colored contrast borders) you only mention one possible cause.

Both definitions are used. Personally, I prefer #2, because #1 already got a name (longitudinal CA).

So, to be more precise, there are at least 4 possible causes within the scope of definition #2:
- longitudinal CA (removed by small f stops) (rotationally invariant)
- out-of-focus CA (which scientifically, is different from the former, but removed with blur by small f stops, too) (rotationally invariant)
- lateral CA (not removed by small f stops) (rotationally invariant)
- sensor demosaicing errors (not removed by small f stops) (translationally invariant)

The longitudinal CA doesn't have this opposite side purple/green effect, I agree. Still, it is rotationally invariant (your image isn't in contradiction having same fringe color on both sides).

In many cases, I found lateral CA to be more of concern for PF, e.g., see the hole in the "R" in the attached image. It is a 400% crop near the top-side border from a DA 40mm Ltd. at blende8 (f/8.0)