You would get excellent results if you hold a 10" x 15" piece of matt black cardboard to photograph the ghosty images! (note: the shutter release button on the camera was at the top) In live view mode orient the lens so as to get the longest streak from the edge of the field:
The image on the left was shot outdoors with a bright sunny sky. The image on the right was shot indoors next to a French window lit from a cloudy sky.
The goal of the experiment is to verify that there is a visible difference when the camera is oriented vertically with the battery on top and when it is pivoted upside down.
(Too) many complex features related to flare co-exist on the images at the top of this page that hampers the analysis of the phenomenon in a serene and effective way. To get a definitive clue out of the images shown above because the cause of the phenomenon seems quite elusive. I therefore decided to build an experimental setup to better control and analyze the impact of the Flare in an attempt to determine the cause of it.
To expand on the observations that were already made,let us see what happens when1) the light rays from a bright light source(s) are hitting the focal sensor plane just outside but near the border of the CMOS sensor. 2) In addition, a dark-shadowed background for the photos would make the flare effect more visible than if it was more in the high key side..
I came up with the following first experimental setup (I had all the components at hand already):
The 60 LED flashlight is supported on the top of a pair of firedogs on the front step of a large open fireplace where soot makes a rather dark background.
Canon 8-15 mm @ 15 mm
Canon 50 mm f:1.8
Samyang 14 mm f2.8
The benefit of having discrete (i.e. separate) points of light is to be able to see how the veiling glare is constructed. If a large and long uniform light source had been illuminating, this phenomenon could not have neen discerned. The center part of the LED array was directed through the entrance pupil of the lens. As a consequence and because the LED have all a parallel active axis, the other LEDs that are closer to the ends of the flashlight are not directly oriented toward the entrance pupil of the lens. Thus the streaks of light become fainter when going toward the lower corners of the sensor frame.
Two Object VR Object movies are presented below:
Canon EF 8-15 mm f4 L on EOS 5D2
Tokina 10-17 mm (shaved) on EOS 5D2
In the course of the experiment, the camera was pivoted 360 deg. around the axis of the lens: the reader may hit left and right arrow keys on the keyboard (step by step) to move the rays of stray light around the image center and observe the moving effect of flare and veiling glare.
These animations put IMO squarely in evidence a very important fact: whilst the bright light source is willingly kept at a constant angle from the axis of the lens, the impact of the veiling glare on the image obviously varies and depends strongly on the roll angle (when it should not if the lens was the main and only contributing factor). Veiling glare is obviously heftier when the light comes
Note that the streaks of stray light are all perpendicular to the lower side of the image. They are not radially converging toward the center of the image. We must conclude from all these observations that another factor than the lens itself is mainly contributing to the creation of the phenomenon.
The white horizontal lines (parallel and close to the lower side of the thumbnail image of the animation here above) are probably due to diffraction and/or refraction on the edges of the thin filters(of different sizes) that are stacked in front of the CMOS and are all larger than the sensor itself. This wild guess has to be confirmed though.
Not every body have a 60 LED flashlight. However there are other cheap ways to experiment. One method allows to perform tests in better controlled conditions than using sunlight for "parasitic" lighting:
You may buy for instance this Energizer X-Focus personal mini-flashlight (aka Energizer X113L ) which price is only a few bucks.
The main advantage of this small device is its single LED lamp that can be very easily uncapped by unscrewing the tiny reflector module:
>>>> ======= >>>>
For the purpose of this experiment (where no "measurement" shall be actually attempted), the lone LED can be assimilated to a single point of light. Then two alternate processes can be used:
The background of the picture should preferably be as dark as possible (the sheet of cardboard is OK if the general scene lighting is dimmed).