AF tunning what actually happens?????

Started 9 months ago | Discussions thread
Roy Kikuta
Contributing MemberPosts: 869Gear list
Like?
Re: AF tunning what actually happens?????
In reply to Roy Kikuta, 9 months ago

Roy Kikuta wrote:

Horshack wrote:

Roy Kikuta wrote:

Horshack wrote:

Roy Kikuta wrote:

Do you have more details - for example, how does the AF module/CPU communicate with the "silent wave" (ultrasonic?) focusing system in an AF-S lens? Since AF-S lenses contain CPU's, I assume the AF module in the body communicates digitally with the CPU in the lens - has anybody hacked the lens CPU or signal to actually understand what's going on? Is the fine tune "correction" applied during the phase detection/calculation, or is it applied in the signal sent to the lens - stated another way, is the fine tune "correction" applied in the body's AF module - or is it actually executed by the lens CPU? I've heard that only one lens drive command is applied - with no secondary phase check(s)/lens drive command(s) applied - in order to speed up AF operations - is this true? I'm not criticizing your explanation - I'm simply looking for more details because it's a subject that's dear to my heart

Evidence strongly suggests that fine tuning affects only the phase detection part of the process and does not alter the commands sent to the lens. The first piece of evidence is WilbaW's excellent articledescribing how PDAF is a partially closed-looped system that relies on multiple phase sensing:lens movement sequences...in such a system you can't reliably alter the outcome by biasing an input. The second piece of evidence is that a body+lens can be accurately tuned without ever engaging the AF motor (DotTune), with some background information here.

Nikon's own description http://www.nikon.com/about/technology/rd/core/software/caf/index.htm seems to imply that only a single lens drive command is applied - especially in AFS mode. Even in AFC mode, I get the impression that multiple phase detect/lens movement processes are done ONLY when movement is detected. To put it crudely - the phase detect module calculates what direction and how much to move the lens to achieve focus - then issues a single drive command - and once the lens movement is done, a confirmation signal is sent to the body - which we see as the green AF confirmation dot in the VF. Unfortunately, there appears to be no subsequent phase detection to confirm that precise AF has in fact been achieved. My understanding is that the AF-S "motor" inside lenses are a form of linear piezoelectric (ultrasonic) drive motors - where a voltage signal excites the crystal which in turn causes lens (group) elements to move a certain distance and direction. Surely, there must be a difference in how much lens movement is generated by a given drive signal depending on whether I happen to have the lens/body in a vertical position when shooting BIF's going overhead - as opposed to shooting a bird on the ground with lens/body in the horizontal position? What happens with older AF-S lenses as lubricants deteriorate - in time, a given drive signal will result in smaller movement as friction increases? It seems a rather haphazard way to achieve precision AF (i.e., not using a closed loop feedback system) - I assume that Nikon has done a lot of testing over the years, and concluded that the single drive AF works "good enough" - and necessary in order to achieve 10~12 FPS in the pro bodies. A lot of guesses and (half-baked?) assumptions on my part - I just wish I had more definitive info. I'm not a Canon user and don't visit their Forums - I wonder if AF is also a significant discussion item there as well?

I don't see anything in Nikon's article which implies a single movement - just some phrasing limited for the sake of the article's brevity. You seemed convinced that PDAF uses a single motion. It doesn't. I'd suggest performing your own experiments to settle the matter to your satisfaction. WilbaW describes some good experiments you can try in his article.

I've used about a half dozen nikon DSLR models thus far, but have never seen the AF system (using PDAF) do more than a single focus operation to achieve (what it thinks) is in-focus status. Looking thru the VF as I press the AF-ON button, I would expect to see something akin to "hunting for focus" operation if the AF system really performs a close-loop feedback operation to do iterative phase detect comparisons and issue multiple drive commands as necessary to achieve perfect focus - and I've never seen this. All the bodies I've used so far perform one continuous focus operation, lights up the green focus confirmation dot in the VF, then stops - unless you're in AFC mode and the the subject is moving - on static subjects, never. I guess I'll keep on looking for more info. Thanks for the discussion. Aloha!

It looks like I have to eat my words re. how many iterative phase comparison/drives are performed in a PDAF operation according to this article by Roger Cicala of LensRental: http://www.lensrentals.com/blog/2010/07/how-autofocus-often-worksAccording to Roger, the PDAF system is so fast it CAN perform multiple cycles of phase comparison/drive commands in order to achieve perfect focus.  Faster than I could see when looking thru the VF, anyway    I'll have to take time to read the article a couple of times to digest it fully.  Cheers!

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow