Ugly Pink Outline - Help!

[Billy McCannon]

Okay, I haven't been able to carry out extensive tests. Don't know how people can be so sure it is CA because I've never seen anything like this on film.

All I had time to do since I last posted was to recreate the conditions with Canon 35mm f1.4L wide open.

- when I closed the lens down in 1/4 stops the magenta affect had completely gone before I'd closed down a stop.

- Placing a .3 ND filter in front of the lens also eliminated the magenta affect.

There are a bunch more tests I'd like to do to resolve exactly what part of the affect is lens and what part is sensor - but clearly it is not just CA.

 

[Jeff Kilgroe]

Here's your pink outline...

Here's your pink outline...

It's electron bleed or electron overflow. That window and the hot spots on those house lights are so over-exposed that they are literally overloading photosites on the sensor, causing bleed or overflow into adjacent sensor space.

Easy to replicate. Next time, bring up the internal lighting and put some ND material on those windows. You can somewhat control the amount of this effect in post through your ISO and exposure settings, however there is only so much you can do once its already recorded. Even though ISO / exposure are meta, the sensor still has a native ISO of about 320 and that is how it sees the world as it records it.

 

[Jason Ing]

Not CA? That's interesting. But does all blown out areas cause pink overloading? I've seen Red footage of blow out areas with no pink overloading. I've also seen a high dnr scene like this but the blown out windows didn't overload the sensors (well, at least there wasn't any pink outlines). How is this discrepency explained?

 

[Jeff Kilgroe]

The pink overloading occurs on areas adjacent to those that are more than blown-out -- they are literally overloading the sensor to the point of failure. Look at the inside of the flare toward the tip of the sword. The dark center is the "black sun" or "purple sun" anomaly that occasionally rears its head with this camera. And yet, the pink overloading is there.

 

[Jason Ing]

oh, so this is the black sun effect.

so an ND would get rid of the black sun effect but CA would remain.

 

[Jeff Kilgroe]

Yes.

ND would clear it up as it would bring these hyper-exposed elements to a lower level.

 

[Billy McCannon]

Thanks Jeff,

I'm hoping you are right but I fear that ND( I presume you mean on the window) is not the solution. The contrast ratio ( didn't check it ) was not that great. As soon as I have some time I'll run more comprehensive tests and post them.

 

[Albert Armes]

So I took these shots with a Canon 1D MK III and 50mm F1.4 lens. Wide open, we see the purple fringing. ND down 3 stops (compensating the exposure with the shutter) we see the purple fringing. Stopping down 3 stops, the fringing is mostly gone. Stopping down further there was no fringing, but I don't have enough ND to do that test.

Albert

View attachment 9184

 

[Jason Ing]

So I took these shots with a Canon 1D MK III and 50mm F1.4 lens. Wide open, we see the purple fringing. ND down 3 stops (compensating the exposure with the shutter) we see the purple fringing. Stopping down 3 stops, the fringing is mostly gone. Stopping down further there was no fringing, but I don't have enough ND to do that test.

Albert

View attachment 9184

So in this case, it's CA? Seems like it. If it was the black sun phenom, then the middle of that white circle would show it to?

 

[Albert Armes]

Yes, I've always said it's a form of CA. This kind of test fits in line with that. Now, the live view on the Canon showed a black sun while shooting this, but it didn't come through to the picture. The black sun was only in the very middle of the light. So if this were black sun, it would be in the middle not the edge. To me, this looks to relate to the green / purple CA you get on back / front focused objects, just exaggerated by the sharp edge and bright illumination. Perhaps. Either way, it doesn't look to be a camera thing, but a lens thing.

Albert

 

[Mike Prevette]

Do you guys not see how that example doesn't look at all like the original one posted?

Look at how the color fringe changes around the Axial dimension of the lens. Vs the original example with constant bleed through all 360 degrees. These are similar looking effects caused in similar situations but completely different.

 

[Jeff Kilgroe]

Do you guys not see how that example doesn't look at all like the original one posted?

They don't want to believe, Mike...

Albert's test does indeed show purple fringing, but now we're talking about a different camera, different sensor. Do the same test with a RED One and see what happens. You will have to over-expose the light more than what Albert did in his test to achieve the overloading.

The "black sun" artifact is related to sensor overflow, but a dark spot in the middle of a highlight is not always present with the purple outline. And likewise, the purple outline is not always present with the black spot. And this highlight is in fact a full on purple / magenta and not the more subdued blue-purple seen with typical "purple fringe". Every sensor handles overloading differently and there are also factors at play related to how the sensor data is processed. Most all sensors will overload at some point -- a black spot may be visible in a "live view" on a DSLR, but most are programmed to detect overload and simply set those overloaded sites to a 100% white exposure. The RED One has difficulties in this area. It has been vastly improved since early builds. I've seen footage from Builds 8 and 9 where flares would leave black streaks across the sensor.

If you guys still want to believe this is a CA anomaly, do a CA test like I showed above to build a "CA profile" for the lens you are using. Then go point it at something really bright to cause overload or bleed. You will see we are dealing with two completely different things here.

CA is a lens thing... Electron bleed or sensor overflow or "black sun" or whatever you want to call it, is a sensor thing. The CA test I posted above was shot on a Cooke 18-100, the overloaded reflection off a shiny katana from a 2K HMI was also shot with the same Cooke 18-100. I'll see if I can go back and bring up a 1:1 crop of that frame. Should have posted it that way anyway. It's obvious at 1:1 that it's not CA. In the above picture with the purple outline, we're looking at the image at a crop from a 1920 pixel wide frame scaled from the 4K original. So it's about 1/4 actual size.

 

[Albert Armes]

What I did should be fairly easy for someone with a red one to do. Just point the camera into a light and have the lens wide open. I used a .9, but you could really do with more ND as it can take more than 3 stops for the lens to start behaving right.

What I did was not with a red, but it does demonstrate that purple fringing is caused by lenses when they're wide open. It backs up what I said about seeing it in filmproductions also.

Sounds like purple in an image could have many causes. You'd have to go through all causes and eliminate them.

 

[Harmonica]

the purple outline is not always present with the black spot.

This is probably the case when there is a more gradual change in light values surrounding the overloaded part of the sensor, as opposed to the first frame grab showing an abrupt change from the blown out light to the window that is framing it.

 

[JD Holloway]

I'm agreed with the bleed theory here. It definitely does not appear to be CA (though CA may be present as well). Hopefully as sensors get thinner this will become less of a problem as longer wavelengths can't seep up from the depths and tweek short wavelength registers in the shallow zones.

 

[Albert Armes]

Just doesn't look bright enough to be a bleed. I don't have the 35mm F1.5 Canon lens to perform my same experiment. Unless the original poster can do what I did and compare stopping the lens down to F4, or ND.9, we have no way to figure it out. I wonder if they could do this test?