d800 focus testing with different targets

Started May 15, 2013 | Discussions thread
Per Baekgaard
Contributing MemberPosts: 889Gear list
Like?
Re: d800 focus testing with different targets
In reply to RajahX, May 15, 2013

RajahX wrote:

Per Baekgaard wrote:

But the results above would suggest a tune value of -3, -4.5, -7 or -10. [...]

... which would indicate values between -5.5, -7, -11.5 and -6.5 at this distance.  [...]

I think the actual tune value depends on the distance to the targert, and the relative posisions between your targets 1, 2 and 3 are roughly the same as above (in the 0.5m test you end up with 67:100:156:222 whereas in the 1m test you have 79:100:157:86 -- so 4 is clearly an outlier).

How are you calculating these values?

Sorry for not being clear on this. I randomly took the 2nd value (-4.5 and -7 respectively for the two series) as 100, and just calculated the ratio to the others in the same series.

So -3, -4.5, -7 and -10 becomes 3*100/4.5, 4.5*100/4.5, 7*100/4.5 and 10*100/4.5 respectively -- or 67:100:156:222.

For the 2nd set of numbers, I did the same calculation and it showed the second set of ratios: 79:100:157:86.

Comparing those, then the 2nd and 3rd numbers are almost equal, the 1st ones are close enough but the last one rather random.

I have used the checkerboard for tuning before, and yes, it looks terrible as a target

Yeah, that looks pretty bad

i agree. i guess what i am trying to figure out is if there a pattern that the PDAF likes --- that way we can remove that source of variability from the picture, and focus on calibrating for the differences in the path that light takes to the PDAF and the actual sensor.

Indeed.

As you might be aware, the PDAF works by comparing two paths of light from the optics vs. one another. Think of it like recording the intensity along a (horizontal/vertical) path and ending up with two curves. You then "shift" one of the curves until you find the position where they are as equal as possible (you could fold them, i.e. "*" mathematically speaking). This will give you a phase offset that indicates how far off you are from "best focus". When in focus, the two internal light paths should result in zero phase shift.

If you have repeating pattern, it can fool the system. The same if you have no clear pattern or a too detailed/too coarse pattern.

 Per Baekgaard's gear list:Per Baekgaard's gear list
Nikon D800 Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF-S Nikkor 14-24mm f/2.8G ED Nikon AF-S Nikkor 24-70mm f/2.8G ED Nikon AF-S Micro-Nikkor 105mm f/2.8G IF-ED VR +15 more
Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow