Debugging D800 AF issues w/Speedlight AF assist

Started 7 months ago | Discussions thread
ForumParentFirstPreviousNextNext unread
Flat view
Horshack
Veteran MemberPosts: 4,096
Like?
Debugging D800 AF issues w/Speedlight AF assist
7 months ago

There have been several threads about D800 Autofocus problems when using the red AF assist beam on various Nikon Speedlights. I'm performing experiments to help characterize the issue. I started this thread to report the ongoing results of those experiments and also to have a place for everyone to troubleshoot and brainstorm the problem together. My goal is for us to fully characterize the issue to the point where we know the specific conditions that cause the problem, along with coming up with possible workarounds for it.

The first interesting data point is that the AF problems occur even if the AF assist beam is blocked (ie, covered so it can't reach the subject). I believe 'olyflyer' was the first to report this. This implies that the PDAF algorithms are designed to act differently when the camera thinks the target will be painted with the red assist beam. There could be several reasons for this, including adjusting for the red spectrum of the assist light, adjusting for the expected reflectivity of the assist light, adjusting for the timing of firing the assist light, etc...

Focusing first on the assist's light's color/reflectivity, I wondered how the AF would be affected when the assist light is the primary illuminant for the target ("low LV" condition) vs when the assist light has to contend with ambient tungsten light ("normal LV" condition). The working theory is perhaps the color/reflectivity adjustment the PDAF system is making for the assist beam is not accounting for ambient light levels and thus is expecting a presumed spectrum/reflectivity on the target and not responding well when that spectrum/reflectivity is not primarily the red assist beam.

The setup is a D800 + 50G f/1.4 + SB-700 mounted on a tripod with the center point covering a variable-width cross-hatched black and white target positioned 7'3" away. I AF tuned the body+lens at -16. At this setting the camera has 100% shot-to-shot precision on the target under all lighting conditions when the AF assist beam is off...so that is the control. The experiment is to then enable the AF assist beam and perform the following 4 tests:

  • Low LV lighting, AF assist beam un-obscured
  • Low LV lighting, AF assist beam obscured/blocked
  • Normal LV lighting, AF assist beam un-obscured
  • Normal LV lighting, AF assist beam obscured/blocked

Again, in the "Low LV" lighting the AF beam is the primary illuminant for the target (very dark room). In the "Normal LV" condition the ambient light represents the majority of illuminant visible on the target (only a faint red from the assist beam is visible). I took 3 shots for each of the 4 cases listed above (12 total). Here are the representative samples of each case (100% crops):

These results indicate AF is good when the AF assist beam is the primary illuminant. The results become worse when the AF assist beam is balanced with ambient tungsten lighting (I believe this is representative of the reported problem). The results become much worse when none the AF assist beam is allowed to reach the target, irrespective of LV.

More experiments to follow.

Nikon D800
If you believe there are incorrect tags, please send us this post using our feedback form.
ForumParentFirstPreviousNextNext unread
Flat view
Post (hide subjects)Posted by
ForumParentFirstPreviousNextNext unread
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow