3
\$\begingroup\$

Does the number of phase detect auto-focus points affect focus accuracy of a mirrorless camera? And if so, how and why?

I am considering buying either a Nikon Z6 II or a Z7 II. They have 273 and 493 AF points respectively. For comparison, the Sony A7iii has 693 AF points, a lot more. I have read/heard that the Z6 (not the Z6 II) struggles a bit with focusing the iris of the eye. Does this have anything to do with how many AF points? Perhaps less focus points means a lower chance the point will be in the right position on the subject? Or maybe the contrast detect will work together with phase detect and the issue is not the number of points?

I will appreciate any informed answers.

\$\endgroup\$

2 Answers 2

4
\$\begingroup\$

On-sensor PDAF works much more like a split prism focusing screen as compared to the DSLR's dedicated PDAF sensor. A pair of photosites (pixels), or a pair of photosite groupings, have their lens(es) biased to intercept opposing virtual images (left/right, upper/lower, or both), and when those virtual images differ in phase (angular offset/position) the image is determined to be out of focus.

The number of focus points is a tradeoff in terms of size. Due to the offset photosite lenses, on-sensor PDAF points inherently receive less light (virtual image rather than the total image/light), this can result in exposure lines/banding in an underexposed/recovered image. It is NOT necessarily less light than a DSLR's dedicated PDAF sensor receives, it may be more (max aperture AF) or less (stopped down "live view" AF). But the more photosites that are dedicated to PDAF reduces overall fill efficiency (exposure sensitivity). When there is plenty of light this is of minimal concern.

Smaller focus points can detect finer details, but that is equally more dependent on the quality (sharpness/clarity/contrast) of the image being focused. If the lens projected image doesn't have the resolution to project that fine of a detail clearly, then having that small of a focus point is essentially useless (image will still be focused as well as possible).

What is most relevant in terms of the number of focus points is the area of coverage (IMO). I would rather have a larger number of focus points because they cover a larger portion of the image area, as opposed to having a larger number due to them being smaller and covering the same portion of the image area.

The eye AF issues of the Nikon Z series is primarily due to the algorithms in use (IMO). Nikon is relatively new to the advanced mirrorless systems, and eye AF was a firmware add-on for the Z6... Sony has been doing advanced mirrorless for a lot longer.

\$\endgroup\$
3
  • \$\begingroup\$ Split prisms work exactly like PDAF arrays: They redirect light from the edge of opposite sides of a lens, just as the microlenses redirect light from opposite sides of a lens to the lines on a PDAF array. \$\endgroup\$
    – Michael C
    Nov 21, 2020 at 8:48
  • \$\begingroup\$ @MichaelC; They work similarly. Split prisms and on-sensor PDAF use virtual images (a portion of all of the light) on it's way to form a singular image at a focal plane. This system is dependent on the lens' aperture for the amount of light and DoF and usually occurs stopped down. When the image is in focus there is only one images and the virtual images can no longer be compared for phase (position); which is why on-sensor PDAF systems typically switch to contrast detection (CDAF) to finalize/refine the focus. \$\endgroup\$ Nov 30, 2020 at 14:51
  • \$\begingroup\$ The dedicated PDAF system uses multiple real images taken from separate areas of the objective element. These images are dependent on their own aperture limit (virtual apertures) for the amount of light and DoF at the focus sensor (the DoF is much greater even at max Ap. The location of the compared images on the sensor lines is used to determine the direction of focus drive. And single (matching) focus points determine when the same detail is in the same location (in focus). Even when focus is achieved there remain multiple real images which can continually be compared for position (phase). \$\endgroup\$ Nov 30, 2020 at 15:06
3
\$\begingroup\$

As with many things, the marketers have tried to convince us that bigger, more, or higher priced is always "better". Most of the time in photography quality trumps quantity.

The term AF "point" has always been a bit of a misnomer. Rather than a single discrete point with no dimension, all AF systems have areas or zones of sensitivity that they call "points".

  • This is the case with dedicated PDAF sensor arrays, such as those found in DSLRs, which use microlenses to direct light from certain parts of a lens' image circle to pairs of lines on the AF sensor. Each line is either one relatively large "pixel" or several smaller "pixels" wide and many more "pixels" long.
  • This is also the case with imaging sensor based CDAF and hybrid PDAF/CDAF systems used in mirrorless cameras as well as by DSLRs when in Live View. Each AF "point" uses multiple photosites (a/k/a sensels or pixel wells) to measure contrast within the area of sensitivity.

In both cases the camera is usually programmed to bring into focus the area of strongest contrast anywhere within the entire selected area of sensitivity. For many cameras and systems from all the major camera manufacturers, the areas of sensitivity for each "AF Point" are far larger than the little squares, rectangles, or dots that you see in the viewfinder. Often these "points" share different, overlapping sections of longer lines on a dedicated PDAF sensor array or share overlapping areas on the main imaging sensor. It's not at all unusual for various AF areas of sensitivity (i.e. "points") within a specific cameras AF system to be shaped differently than the others around them.

Regardless of how many focus "points" a camera has, the camera's system primarily uses one measure of contrast at any particular time to actually focus the lens. It may be an area of contrast within a single "point" manually selected by you or it may be an area of contrast inside one of a large number of AF "points" that are active based on the selected settings. If you have multiple focus points selected and more than one light up in the viewfinder, it only means that the camera is telling you that the subjects behind all of those focus points are at roughly the same distance.

What is more important than how many focus points the camera has is how sensitive the focus points are, and how accurately the camera/lens is able to move the elements in the lens to match the instructions from the camera, and how consistently it can do this from shot-to-shot. In some cases how quickly the system can do this is also vital.

How many AF points a camera maker claims a mirrorless model they sell has is based more on the way they decide to count them than about how many discrete areas of contrast sensitivity the sensor contains. It's more a marketing department decision than an engineering and design specification. Design decisions are more along the lines of:

  • How much of the total area of the sensor is used, as well as how many of the individual photosites (a/k/a pixels wells or sensels) in a particular zone are used, to detect contrast to be used by the AF system
  • How the total area of the sensor that is capable of detecting contrast is divided up into various zones
  • How large each of the zones are, how each zone is shaped, and how much they overlap with adjacent zones sharing some of the same photosites
  • How many of those zones are user selectable and how many are available to the camera's AF point autoselection routine

Pretty much all of these things can be modified later by only changing the firmware that controls the camera's hardware.

How many "AF points" a mirrorless camera has doesn't necessarily affect focus accuracy (or consistency - they're two different things), either. If your intended subject is located in an area of the frame that has no AF sensitivity, that's not an accuracy issue at all.

How well a camera can maintain AF on a specific feature of your subjects, such as the iris of an eye, as it moves around in the frame is more about the capabilities of the software controlling the AF system than it is about the sensor itself. There have been more than a few cases when various camera manufacturers have released new mirrorless cameras and then later upgraded the AF system, particularly with regard to Eye AF, using only firmware updates.

If your primary concern is getting the camera that does the best job at focusing on a subject's eye then your best course of action is to read reviews from various sources that compare the models you are considering and/or rent each of the models you are considering and test them yourself. The number of advertised AF points doesn't really affect which will be better or worse at that.

You should be able to find out how much of the entire frame is available to the AF system by looking at each camera's specs. If you like to place your subject's eyes in the outer 10% on either end of the frame and the camera advertises that "80% of the sensor's width is included in the AF area", then you know that the 10% remaining on either end will not be covered at all.

\$\endgroup\$
14
  • \$\begingroup\$ I have tested the AF of many cameras to verify the location of the actual AF point in relation to the displayed location in the viewfinder. I have never found the actual AF area to be larger than the depicted point. A AF sensor line is an array of photosites, very much like those on the imaging sensor. And a focus point can be a single (pair) of photosites, or two (pairs of) photosites. Or it can be a larger grouping, or larger photosites. And while contrast/clarity is still relevant, it is not the primary factor; phase (position) is. But none of this is terribly relevant to mirrorless. \$\endgroup\$ Nov 21, 2020 at 4:18
  • \$\begingroup\$ Almost all digital cameras have larger AF "points" than the little squares in the viewfinder. It's rather well documented from a wide variety of sources. This should be rather obvious to anyone when one can focus on an object that is between two AF point indicators in the viewfinder. \$\endgroup\$
    – Michael C
    Nov 21, 2020 at 8:47
  • \$\begingroup\$ For more about the size of AF "points" versus their depiction in a viewfinder, please see this entry at Andre's Blog. For how PDAF systems with large numbers of AF "points" share different portions of lines in the sensor array, please see this answer. \$\endgroup\$
    – Michael C
    Dec 27, 2020 at 5:08
  • \$\begingroup\$ A PDAF sensor line is not a focus point; i.e. the cross type focus point is only where the two directional lines intersect (4 actually). All of the sensor lines, in their entirety are used to determine the direction of focus drive; but only the focus point in question is used to achieve/lock/confirm. Additionally, the location/spacing of the lines on the sensor is not relevant to their sensitivity. What is relevant is the aperture area of the objective lens where the images are taken from. E.g the diagonal cross central point for Canon is from the f/2.8 region. \$\endgroup\$ Dec 27, 2020 at 20:53
  • \$\begingroup\$ I can only assume there is a testing error in that blog post. E.g. for a Nikon in any dynamic focus mode (which starts as single point) it will use surrounding focus points to achieve focus if the selected point cannot. Canon's AI servo mode may behave similarly (?). The best way to get an accurate representation is to use manual single point/single drive to see when focus is achieved. The system may not report focus lost nearly as accurately/quickly when moving off of the focus target (which is what was used). I have tested Canons before; I do not have one to test now. \$\endgroup\$ Dec 27, 2020 at 21:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.