The optical stack in front of an image sensor of an interchangeable-lens camera generally consists of:
- The first birefringent crystal of the optical low-pass filter, separating the image in the horizontal direction.
- A phase plate combining the vertically and horizontally polarized lights exiting from the first birefringet filter into a circularly polarized light.
- The second birefringent crystal of the OLPF, separating the image in the vertical direction.
- An infrared absorption filter
- A glass plate with a dielectric thin film coating acting as a dichroic infrared filter (hot mirror)
- The glass cover of the chip carrier package containing the silicon image sensor chip proper.
To reduce the thickness of the overall optical stack, some elements might be combined: cameras with a pizeoelectric dust removal system ("dust shaker"), for example, might use as the transparent cover of the chip package the second birefringent crystal of the OLPF, eliminating the need for (6)
Here's a schematic of the optical stack of the Canon 5D Mark II :
In a digital camera, the "focal plane" (FP) shutter curtains actually sit in front of the first birefringent crystal / dust shaker — i.e. not exactly at the "focal plane".
Considering the thickness of the various optical elements and the air gaps that must exist between them, the shutter curtains must be at a non-negligible distance — e.g. five millimeters — from the actual focal plane — i.e. the sensor chip surface.
What effect would that curtain-to-sensor distance have when EFCS is used ?
Let's consider the cone of light emitted by a f/1.4 lens. By definition, all f/1.4 lenses, regardless of their focal length, emit a cone of light of the same "shape", in the sense that these cones of light will all have the same angle at the cones' apex (the apical angle).
If F# is a lens' F-number, the value of that apical angle is 2*arctan(0.5 / F#)
For a f/1.4 lens, that apical angle is about 39 degrees.
When a point source of light is focused by a lens on the image plane (electronic sensor or photosensitive film surface), the result is — if we ignore aberration and diffraction effects — a sharp point of light; that point of light on the image plane is the apex of the cone of light emitted by the lens.
When the point source of light is defocused, a light disk is obtained on the image plane.
The pixels of an image sensor continuously generate, and internally accumulate electricity when they are exposed to light. Prior to taking a picture, the pixels must thus be "reset" — i.e. the electrical charges they've internally accumulated must be discarded, so that a new light accumulation cycle is started from a "blank slate", or a black pixel, if you prefer.
In a film camera, the first shutter curtain, as it moves, uncovers the film, starting the exposure.
After the required delay has elapsed, the 2d shutter curtain starts to move, and covers the film, thus concluding the film's exposure to light.
The normal state of a piece of film in a camera is thus to be protected from the light.
OTOH, in a digital camera operating in EFCS mode, the normal state of the sensor is to be exposed to light; it typically generates the liveview picture being fed to the camera's EVF or rear LCD.
When exposure starts, the EFCS is therefore implemented as an electronic "reset" signal wavefront that travels across the sensor at the same speed as a mechanical shutter curtain would. The EFC — i.e. the travelling reset wavefront — clears the accumulated charges of a pixel, thereby starting a new exposure — i.e. the accumulation of new photoelectrons. The expsosure is completed when the pixel is covered by the mechanical 2d curtain, blocking any additional incoming light energy.
The effective exposure time of a pixel is thus the time between:
- the pixel being reset by the EFC a.k.a. reset wavefront, and therefore starting a new cycle of photoelectron accumulation
and
- the pixel being covered by the 2d curtain — i.e. entering in the shadow area projected by the 2d curtain on the sensor surface.
In the Sony A7 series, the mechanical shutter curtains travel upwards, from the bottom of the camera to the top. The direction of travel of the pixel-resetting wavefront emulating the first shutter curtain must thus also be upwards.
With mechanical FP shutters, the exposure window created by the 1st and 2d shutter curtains is, conceptually, a slit travelling across the film or sensor. The shorter the exposure time requested, the narrower the slit.
The same principle applies to the EFCS: when the exposure time must be short, the 2d (mechanical) shutter curtain follows the electronic first curtain closely.
The EFC, being a reset wavefront travelling across the sensor, can be considered to travel right at the sensor surface — which we might describe as zero altitude.
The 2d curtain, OTOH, "flies" above the OLPF optical stack — i.e. about 5mm above the sensor's surface.
This 5mm altitude difference creates interesting effects when the lightrays are heavily tilted — e.g. in the case of the marginal rays emitted by large-aperture lenses.
Consider a blurred point light source, which should therefore be normally imaged as a light disk.
When the 2d curtain starts to intersect the light cone emitted by the lens, it blocks part of that cone's constitutive lightrays, and therefore projects a shadow on the light disk.
In the figure below, the EFC travels upwards, while the shadow cast by the 2d curtain travels downwards. The pixels in the bottom part of the light disk will enjoy the longest exposure time — i.e. the time between they being reset by the EFC, and then being covered by the downward-travelling shadow zone cast by the 2d curtain.
As we move up in the light disk, the effective exposure time becomes shorter and shorter, as pixels tend to be covered by the travelling 2d curtain shadow almost immediately after they're reset, thus giving them very little time to accumulate any photoelectrons.
In fact, there exists a boundary in the light disk above which the pixels cannot accumulate any photoelectrons, as they are already in the 2d curtain shadow when the EFC reset signal reaches them.
Pixels in the light disk above that boundary will then remain absolutely black, resulting in a light disk with a luminance distribution that goes from bright to dark, and totally black beyond a certain point.
Thus, for large aperture lenses which neessarily have large apical angles, and therefore have heavily tilted marginal rays, if the altitude difference between the 2d curtain and the sensor surface is not that different from the shutter curtain slit width, then the EFC will lead to a significant, uneven blocking of the light rays in the lens' light cone.
Blocked lighrays mean less light reaching the pixels, and a darker picture overall.
For out-of-focus areas, blocked lighrays result in truncated light disks; whether the truncated part will be towards the top or the bottom of the disk will depend on:
- the shutter curtain's travel direction
and
- the defocus position: OOF disks for light sources behind the focus plane and in front of the focus plane will have truncations occuring at diametrically opposite positions — e.g. if back-focused OOF disks are truncated at the top, then front-focused OOF disks will have their bottoms truncated.
Here's picture I've taken with a Sony A7II in EFC mode, with a narrow shutter slit (1/4000s) and a large aperture (f/1.4).
The same scene, in mechanical shutter mode, 1/4000s, f/1.4
It's apparent that the picture taken in EFC mode is generally darker, and that the OOF disks are truncated.
Another example of OOF disk truncation caused by EFC. Sony A7s, 1/8000s, f/1.4
Given the physical principles involed, the effects caused by the EFCS become visible only:
- if the shutter slit is very narrow — i.e. shutter speeds shorter than 1/2000s ?
and
- large apertures — f/1.4-ish ?
At apertures darker than, say, f/4.0, I wouldn't worry too much even at quite high shutter speeds, as the light rays wouldn't be that tilted anyway (the apical angle of the lens' light cone will be pretty narrow).
