DotTune: AF tune technique without taking photos

Started Jan 31, 2013 | Discussions thread
Horshack
Veteran MemberPosts: 3,988
Like?
Why AF mistune looks like AF inconsistency
In reply to Horshack, Feb 2, 2013

There have been miscellaneous reports online about the D800 having inconsistent AF. For example Lloyd Chambers reported that the D800 has poor precision with large aperture primes at mid-to-far distances (report here). However when I tested my D800 in a similar fashion I found the AF to be very consistent (thread here).

There have also been recent threads in this forum about D800 AF inconsistency (thread here), along with a thread about how that inconsistency was solved when the lenses were AF tuned (thread here).

Many of us might believe that AF tune and AF inconsistency are unrelated, which isn't unreasonable. If a body+lens are out of AF tune, how could that cause some photos to be in focus whereas others are OOF? Shouldn't all photos be OOF in that scenario?

In the section below I will demonstrate how a non-AF tuned body+lens combo can lead to inconsistent shot-to-shot AF. At least how I theorize it to be, based on a tenuous understanding of how this stuff works Warning: all this may already be obvious to many of you...however it wasn't immediately obvious to me until I thought it through, so for other slow-witted people like me I post

Boring Exposition

Phase Detect Autofocus (PDAF) is part optical (mirrors, beam splitters, sensors) and part mechanical (lens elements, lens motors, servos, mounts). You can read about it (here), and play with a neat java-app depiction of it (here). It's reasonable to assume that both the optical and mechanical elements are subject to manufacturing and environmental tolerances, some of which may be static in nature (mirror/sensor/AF point/mount misalignment, lens aberrations) and some dynamic (imprecise AF motor movements/movement sensors/servos, non-linear positioning errors that vary based on element movement distances, thermal variation). I'm only theorizing about the true source of all the tolerances - I really don't know for sure. Regardless of their source, it's reasonable to assume these tolerances can lead to shot-to-shot AF variability, particularly dynamic tolerances. The magnitude of that variability is also something I can only theorize on, but for the purpose of this discussion I'm going to state that variability in terms of AF tune units, and arbitrarily say the shot-to-shot variability is +/- 4 AF tune units, which is a complete swag but the absolute numbers aren't important, just the concept. A +/- 4 AF tune variability means that focus from one shot to the next has a standard deviation equal in magnitude to the amount of focus adjustment provided by a range of 4 fewer or 4 more AF units from its current AF tune setting (8 total units).

Exciting Diagrams

If we state AF variability in terms of an AF tune range, we can then depict how this variability would manifest in actual photographs for various combinations of actual vs necessary AF tune values. Here's are three scenarios:

In scenario #1, a body+lens need no AF tuning, and AF tune is set to disable (or zero). If the PDAF variability is +/- 4 AF tune units, every photograph should have an "acceptable" level of focus. When the PDAF variability falls in the mid-point of the range (0), AF will be theoretically perfect. When it falls at the edge of the range (either -4 or +4), focus will still be considered "acceptable", which can mean either you can't see the difference vs "0" or that the difference is small enough to not be noticeable or unacceptable for a given print/viewing size.

In scenario #2, a body+lens need an AF tune value of 4, but the AF tune is still set to disabled (or zero). This is the scenario that depicts why an untuned body+lens combo can produce inconsistent AF, with some photos in focus and others OOF. Because the AF tune is off by 4 units, representing half the magnitude of PDAF variability, the odds of achieving an in-focus shot drop by half, limited to the photographs where the PDAF variance falls within the positive half of its range (0 to +4, green to yellow area). Photographs where the PDAF variance falls in the negative half of its range (0 to -4, yellow to red area) will be unacceptably OOF.

In scenario #3, a body+lens need an AF tune value of 4, and AF tune is set to a value of 4. Just like in scenario #1, the AF tune's "needed vs actual" is centered over the mid-point of the PDAF variability range of +/-4, meaning that all photographs should have acceptable focus.

Now all the provisos...the actual PDAF variability is unknown, so we can only theorize its true magnitude based on empirical in-focus rates. Plus what is deemed "acceptable" regarding focus is relative to print size, DOF, viewing distance, etc...Also, outside the general shot-to-shot variability there are huge PDAF outliers that sometimes occur for reasons unknown. Like that one shot out of 10 (20?) that is completely OOF for no particular reason - I've excluded those from this discussion.

Here is a real-world example of the above concept, on a D800+50G whose AF tune needed+actual value is set to 10.

AF series

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
MuchNew
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow