Major frustration in D800 AF Tuning even with DotTune Method

[ZionImage]

I'm trying to AF Fine Tune my D800 with various Nikon lens using DotTune method developed by Horshack described in his youtube video DotTune: Autofocus fine tuning in under 5 minutes

Initially I followed the steps strictly but I found very unpredictable results. Then I read through every threads on the http://www.dpreview.com/forums/post/50883023

Based on the info I learned, I modified the steps slightly to try addressing the issues I'm seeing. But it is still giving me lots of problems. So instead of 5 mins AF tuning, it turns into days. It is so frustrating and I hope that you guys could give me some help.

How I set it up:

I put my target (Bob Atkin's focus chart suggested by Horshack) on a white wall in my studio. I use the 2 modeling lights with my speedotron with 2 softboxes attached to ensure the target had the proper contrast. And the target has consistent lighting to avoid any variation with the lights. I put my D800 on a Bogen 3221 professional tripod without using the center pole to minimize any camera shake. I use the levels on my Bogen tripod head and the virtual horizon on my d800 Liveview to ensure the camera is level correctly and pointing squarely to the target. Set the camera to focus only using AF-ON button.

Here are my steps:

1. Set the camera on the tripod at a distance using a tape measure
2. Make sure the AF Tune is set to 0 and using the AF-S and with the center focus point
3. Establish critical focus in Live View by zooming in 100% with the AF-On Button
4. Evaluating Viewfinder Focus Confirmation by counting to 10 for having the solid green dot (I made sure I only touch the shutter release button very lightly to avoid shaking or moving the camera)
5. Cycle AF Tune values to find the edges of the range
6. Repeat testing the point surrounding the edge values to ensure the same result could be repeated 3 times in a row
7. Mark down the edge values on paper
8. Repeat the same steps for 3+ times to until similar edge values are recorded with previous tests at that distance
9. Move my camera to another distance and repeat 1-8 again.
10. Put all the data in Excel and had it calculate the miidpoint on each test run.

Here are my results for one of the lens - Nikon 135mm f/2.0D AF DC-Nikkor Lens. Based on previous testing, I found that I need to set it to +2 using the Defocus ring or it will be completely out of +20 range.



As you could see the calc AF mid point range from 6.5 t0 -3.5. So it is a very wide range. I'm not done with the testing yet but want to make sure I'm doing it right before continue spending hours in this. While I only showed the results of this one lens, other lens (except the Nikon AF 24-70mm f2.8 which is my most recent purchase) have similar results.

What AF Tuning point should I use for this lens? What did I do wrong? Is my D800 just having focus problem and need repair? Or my lens are just too old and not suitable to use DotTune?

Some footnotes:

1. Why I chose these distances? Because they are 50x of the focal length of the lens that I own. Given my previous bad results, I decided to test all lens with all these distance as I don't really shoot just one distance anyway.

2. My understanding is that shortest focus distance will have the least DOF. It seems to me that if we need to achieve best AF point, we should use the shortest focus distance. And I don't think I typically shoot so far away for a particular len but I went along with 50x anyway. If someone could help explain why use 50x, I would appreciate that too.

Thanks a lot for reading this long post and thanks in advance for your help and suggestions. I very much appreciate that.

Joshua

--
Just trying to improve my draft one image at a time
You could check our more of my photos at http://www.joshuaphotographystudio.com

 

[Robin Casady]

Set the Fine Tune value at the midpoint in your variations and take some real-world photos. Evaluate if they are sharp at the point you focused. If so, stop worrying and take pictures.

This is a mass-produced camera. It isn't a hand-built NASA Mars Rover. There are going to be variations due to mfg. tolerances, and PDAF isn't perfect.

The D800 is the highest resolution FX camera Nikon has made with PDAF. It is going to show PDAF variation better than any other Nikon. After the D800 was announced, but before it hit the streets, I recall reading some punditry saying that PDAF might not be accurate enough for 36 MP FX. Well, guess what? It is good, but it can't do zero tolerances.

My own experience, The Dot Tune method is tedius and frustrating if you expect perfect results. Few things in this world come in with zero variation in the data. Most scientific results have variation in the data.

If you absolutely need to nail the focus point at an exact location with a large aperture, verify focus with Live View at 100% (a few clicks down from max).

BTW, I find Live View's CDAF to not be very accurate. Often it is farther out than PDAF. Visually with a CamRanger is how I nail precise focus in macro shots.

Also, a DK-17M magnifying eyepiece helps with visual focus through the viewfinder.

--
Robin Casady
http://www.robincasady.com/Photo/index.html
"The whole problem with the world is that fools and fanatics are always so certain of themselves, and wiser people so full of doubts."
— Bertrand Russell

 

[Robin Casady]

I should add that in looking for perfection you will learn that different focal lengths in zoom lenses, and different focus distance may require different Fine Tune values. You have to make compromises.

Probably best to set a lens for the distance it will be most often used at. While close subjects are more critical, don't set it at closest focus if you rarely use that distance with that lens.

 

[Horshack]

Sounds like you're doing everything right. I'm not sure why you're midpoints from test-to-test at the same distance have outliers but if I average your results for each distance I get 1.58 for distance "137" (so use either +1 or +2), -.25 for distance "167" (use either -1 or 0), and -1.66 for distance "236.5" (so use -2 or -1). Even with the outliers you seem to be getting a pretty consistent average midpoint across all distances, and the average midpoint looks good.

Regarding your midpoint outliers, try this target in place of the Atkins target and see if you get more consistent results from test-to-test. I've been doing experiments with different targets and found the cross-hatch one to produce the most consistent results. Please report back on whether it helped you.

As for focus distance, it is true that the shorter distances produces the lease DOF and so intuition might say that shorter is better since it would reveal focus differences more. In practice I've found the opposite is true - even though longer distances have more DOF, long distances also many times reveal the need for AF tune much more than shorter distances do. At first blush this makes the short vs med/long distance results look like distance-specific tuning variation (and thus the impulse is to average the midpoint distances), but in actuality I've found that the tune value derived at the mid-to-longer distance works well for short distances as well whereas the opposite is not true (using the short-distance tune value for other distances doesn't work well). My conclusion then is to try tuning at mid to long distances for cases when the short distance seems to produce a significant different midpoint than other distances. As for what specific distance to use for "mid" to "long" distance...for some lenses (like the 50G f/1.8 and f/1.4), I've actually found infinity sometimes works best...and so I have to use a outdoor targets in these instances (like a far tree, house, etc...).

 

[ZionImage]

I should add that in looking for perfection you will learn that different focal lengths in zoom lenses, and different focus distance may require different Fine Tune values. You have to make compromises.

Probably best to set a lens for the distance it will be most often used at. While close subjects are more critical, don't set it at closest focus if you rarely use that distance with that lens.

 

[ZionImage]

Sounds like you're doing everything right. I'm not sure why you're midpoints from test-to-test at the same distance have outliers but if I average your results for each distance I get 1.58 for distance "137" (so use either +1 or +2), -.25 for distance "167" (use either -1 or 0), and -1.66 for distance "236.5" (so use -2 or -1). Even with the outliers you seem to be getting a pretty consistent average midpoint across all distances, and the average midpoint looks good.

Regarding your midpoint outliers, try this target in place of the Atkins target and see if you get more consistent results from test-to-test. I've been doing experiments with different targets and found the cross-hatch one to produce the most consistent results. Please report back on whether it helped you.

As for focus distance, it is true that the shorter distances produces the lease DOF and so intuition might say that shorter is better since it would reveal focus differences more. In practice I've found the opposite is true - even though longer distances have more DOF, long distances also many times reveal the need for AF tune much more than shorter distances do. At first blush this makes the short vs med/long distance results look like distance-specific tuning variation (and thus the impulse is to average the midpoint distances), but in actuality I've found that the tune value derived at the mid-to-longer distance works well for short distances as well whereas the opposite is not true (using the short-distance tune value for other distances doesn't work well). My conclusion then is to try tuning at mid to long distances for cases when the short distance seems to produce a significant different midpoint than other distances. As for what specific distance to use for "mid" to "long" distance...for some lenses (like the 50G f/1.8 and f/1.4), I've actually found infinity sometimes works best...and so I have to use a outdoor targets in these instances (like a far tree, house, etc...).

Horshack,

First of all, thanks for developing this method. It is a time saver for sure compare to other methods I tried. It is still tedious but systematic testing is always tedious. Very much appreciate your confirmation on my method as I felt quite lost.

I like to shoot shallow DOF with kids and family portraits. (Yeah, I know I'm asking for work. But well that is my style that is different than others.) Most of the time I do wide open if shooting a single kid. So landing the critical focus is important and I can't precisely control my shooting distance. But I do know that I don't shoot 50x away but relatively close to the subject to interact with them.

It is very interesting on your comment about tuning for longer distance would work better than shorter one. Is your conclusion partial due to the AF focus inconsistency? With longer distance, the focus inconsistency will have less of an effect? At least my result of this lens support that theory.

How do you define short, mid and long? In my type of shooting, would you still suggest longer range might be 50x, 70x or 100x? I'm concerned about doing it outdoor as there are so many more variations. I'm not looking for perfection and certain expect some variations. Hence I do multiple test run. If you look at my test result for this lens, my AF Range is between 4-7 and the Calc AF Midpoint has a 10 point range. So there is a high chance that my value will hit outside of the AF Range. I feel it is more just a gamble.

But in any case, I will retest with your new target and report back. Ah .. Joy of testing.

Thanks a lot.

Joshua

 

[Schroedingers_cat]

[No message]

 

[GodSpeaks]

I use FoCal (Pro). It just works.

 

[Truman Prevatt]

I'm trying to AF Fine Tune my D800 with various Nikon lens using DotTune method developed by Horshackdescribed in his youtube video DotTune: Autofocus fine tuning in under 5 minutes

What you will find is there is a board range where the "dot is still on." Which sitting is right. Then if you are using a zoom you change the focal length and things will change. The you change the distance and things will change.

That is physics. The goal should not be to fine AF offsets for lenses but to take photographs.

First thing is to forget all this forces target nonsense - it would be great if the only thing you took pictures of was focus targets. Go out and find a target - one you would normally take a picture of. Make sure it has a lot of detail. Set up a tripod at a nominal distance you will most often use with a lens. Forget the rules. Do the live view focus and take a shot. Better yet manual focus on live view and take an image. Then if you have some idea of what the fine tune should be and start there. Take images at settings +/- 8 on either side. If you have no idea take raw images at -20 to +20. Down load the images into you raw converter and covert them with no sharping.

Find when are the sharpest. That is you fine tuning.

Paper targets and rulers are wonderful if all you ever plan to photograph is paper targets and rulers.

--

Truman
www.pbase.com/tprevatt

 

[ZionImage]

--
http://www.nightstreets.com
-
One day long ago, a sorcerer and a warrior did battle, as such things were commonplace in that age. Sometimes the sorcerer won, and the sum value of human abilities was improved some trifling amount. Sometimes the warrior won, and again the human race improved by some insignificant amount, for a sorcerer who can't defeat one miserable warrior is a poor sorcerer indeed.

Corona, Do you mean the noted AF problem with geometric pattern?

Horshack, Isn't your new target is more geometric pattern than the one from Bob Akin's? What do you think about this?







--
Just trying to improve my draft one image at a time
You could check our more of my photos at http://www.joshuaphotographystudio.com

 

[Frank C.]

The AF fine tuning should not be considered as a panacea for the camera's AF woes, I think it was put there to act as a security blanket and appease owners more than anything else

 

[ZionImage]

I'm trying to AF Fine Tune my D800 with various Nikon lens using DotTune method developed by Horshackdescribed in his youtube video DotTune: Autofocus fine tuning in under 5 minutes

What you will find is there is a board range where the "dot is still on." Which sitting is right. Then if you are using a zoom you change the focal length and things will change. The you change the distance and things will change.

That is physics. The goal should not be to fine AF offsets for lenses but to take photographs.

First thing is to forget all this forces target nonsense - it would be great if the only thing you took pictures of was focus targets. Go out and find a target - one you would normally take a picture of. Make sure it has a lot of detail. Set up a tripod at a nominal distance you will most often use with a lens. Forget the rules. Do the live view focus and take a shot. Better yet manual focus on live view and take an image. Then if you have some idea of what the fine tune should be and start there. Take images at settings +/- 8 on either side. If you have no idea take raw images at -20 to +20. Down load the images into you raw converter and covert them with no sharping.

Find when are the sharpest. That is you fine tuning.

Paper targets and rulers are wonderful if all you ever plan to photograph is paper targets and rulers.

--

Truman
www.pbase.com/tprevatt

Truman,

I totally appreciate your perspective. In fact, I agreed with you that I shouldn't have to care about this AF Fine Tuning, Monitor Calibration, Color profile all these stuff. Back 15-20 years ago, when I was shooting with my D4S, I didn't have to do any of these and still get great sharp images. Unfortunately time changed with digital camera that comes with lot of goods but also these stuff that took me away my time from taking photos.

When I first my D800 over a year ago, I didn't care about test the AF system and fine tune my camera. I did a job with a good customer who is very influential in the community. Some of the most important images were out of focus. He was so upset that I lost his business and all the referral I could have from him. I did some test and found that the focusing system has some issues and ended up sending it back to Nikon to repair.

Over the last year, I tried different AF tuning method and it has been frustration. Out of focus images keep popping up and forced me to now only use 2 lens out of 8 I have. I have a couple shots coming up that I really would like to use my 135/f2 and 80-200/f2.8. Well, that forced me to do these fine tuning again.

Unfortunately, I mostly shoot people and kids. I don't have the luxury to do what you suggested in a control environment to do the adjustment. I'm starting to think that may be the only way out is to shed some money to buy Focal Pro as Godspeak suggested.

 

[Horshack]

Horshack,

First of all, thanks for developing this method. It is a time saver for sure compare to other methods I tried. It is still tedious but systematic testing is always tedious. Very much appreciate your confirmation on my method as I felt quite lost.

If you're finding that DotTune is tedious then you're doing it wrong It should be a painless process after just a little practice. Just relax the muscles a bit and zen your way through it. I have it down to 90 seconds or less

I like to shoot shallow DOF with kids and family portraits. (Yeah, I know I'm asking for work. But well that is my style that is different than others.) Most of the time I do wide open if shooting a single kid. So landing the critical focus is important and I can't precisely control my shooting distance. But I do know that I don't shoot 50x away but relatively close to the subject to interact with them.

If most of your shooting with the 135mm is close then you can just stick to tuning at 50x and be fine.

It is very interesting on your comment about tuning for longer distance would work better than shorter one. Is your conclusion partial due to the AF focus inconsistency? With longer distance, the focus inconsistency will have less of an effect? At least my result of this lens support that theory.

Longer distances means more DOF which imply that focus misses would be less obvious but again in my testing the longer distances are much more susceptible to tuning requirements so it actually works out that longer distances are more demanding of a proper tune value for good results.

How do you define short, mid and long? In my type of shooting, would you still suggest longer range might be 50x, 70x or 100x? I'm concerned about doing it outdoor as there are so many more variations. I'm not looking for perfection and certain expect some variations. Hence I do multiple test run. If you look at my test result for this lens, my AF Range is between 4-7 and the Calc AF Midpoint has a 10 point range. So there is a high chance that my value will hit outside of the AF Range. I feel it is more just a gamble.

Short is 50x, medium is somewhere between 50x and infinity, and long is infinity or close to it. i wouldn't stress about all the distances since you're using it mostly for close shooting...and even if you do use it for longer distances it shouldn't take much for you to arrive at the correct tune value.

But in any case, I will retest with your new target and report back. Ah .. Joy of testing.

Definitely would like to hear back on whether your test-to-test midpoints are more consistent with the new target.

 

[Horshack]

Over the last year, I tried different AF tuning method and it has been frustration. Out of focus images keep popping up and forced me to now only use 2 lens out of 8 I have. I have a couple shots coming up that I really would like to use my 135/f2 and 80-200/f2.8. Well, that forced me to do these fine tuning again.

Unfortunately, I mostly shoot people and kids. I don't have the luxury to do what you suggested in a control environment to do the adjustment. I'm starting to think that may be the only way out is to shed some money to buy Focal Pro as Godspeak suggested.

I've never used FoCal but others have reported great results with it so I would never steer someone away from it, particularly since it has some automation that some prefer. Most who have used Focal and DotTune side by side have told me they produce the same tune result for a given lens/distance.

 

[Horshack]

--
http://www.nightstreets.com
-
One day long ago, a sorcerer and a warrior did battle, as such things were commonplace in that age. Sometimes the sorcerer won, and the sum value of human abilities was improved some trifling amount. Sometimes the warrior won, and again the human race improved by some insignificant amount, for a sorcerer who can't defeat one miserable warrior is a poor sorcerer indeed.

Corona, Do you mean the noted AF problem with geometric pattern?

Horshack, Isn't your new target is more geometric pattern than the one from Bob Akin's? What do you think about this?



--
Just trying to improve my draft one image at a time
You could check our more of my photos at http://www.joshuaphotographystudio.com

I'm guessing they reference geometric patterns because they're concerned there will not be the proper orientation of detail relative to the AF sensor depending on the size/distance of the pattern. The irregular cross-hatch I linked to earlier in this thread has been working very well for me. I plan to start a thread in a few weeks with results from various targets.

 

[Leonard Shepherd]

Why use the AF target you are using?

On page 100 of your D800 instructions Nikon say this with this type of subject "Autofocus does not perform well".

With no sample images it is not possible to know if you change focus distance with each change in focal length, but you need to do this even with a better test target.

Maybe with a good test target and good technique your D800 will instantly cease to indicate a possible fine tune issue.

 

[ZionImage]

Horshack,

First of all, thanks for developing this method. It is a time saver for sure compare to other methods I tried. It is still tedious but systematic testing is always tedious. Very much appreciate your confirmation on my method as I felt quite lost.

If you're finding that DotTune is tedious then you're doing it wrong It should be a painless process after just a little practice. Just relax the muscles a bit and zen your way through it. I have it down to 90 seconds or less

I like to shoot shallow DOF with kids and family portraits. (Yeah, I know I'm asking for work. But well that is my style that is different than others.) Most of the time I do wide open if shooting a single kid. So landing the critical focus is important and I can't precisely control my shooting distance. But I do know that I don't shoot 50x away but relatively close to the subject to interact with them.

If most of your shooting with the 135mm is close then you can just stick to tuning at 50x and be fine.

It is very interesting on your comment about tuning for longer distance would work better than shorter one. Is your conclusion partial due to the AF focus inconsistency? With longer distance, the focus inconsistency will have less of an effect? At least my result of this lens support that theory.

Longer distances means more DOF which imply that focus misses would be less obvious but again in my testing the longer distances are much more susceptible to tuning requirements so it actually works out that longer distances are more demanding of a proper tune value for good results.

How do you define short, mid and long? In my type of shooting, would you still suggest longer range might be 50x, 70x or 100x? I'm concerned about doing it outdoor as there are so many more variations. I'm not looking for perfection and certain expect some variations. Hence I do multiple test run. If you look at my test result for this lens, my AF Range is between 4-7 and the Calc AF Midpoint has a 10 point range. So there is a high chance that my value will hit outside of the AF Range. I feel it is more just a gamble.

Short is 50x, medium is somewhere between 50x and infinity, and long is infinity or close to it. i wouldn't stress about all the distances since you're using it mostly for close shooting...and even if you do use it for longer distances it shouldn't take much for you to arrive at the correct tune value.

But in any case, I will retest with your new target and report back. Ah .. Joy of testing.

Definitely would like to hear back on whether your test-to-test midpoints are more consistent with the new target.

Horshack,

I focused my testing so far just on the 135-DC as there has been the results I showed. I only have time to test with 2 distances so far as I did have to go shoot and need to process my images too (for those who concerns that I only do testing all day long. I want to report back some partial result and interesting observations. And I have concerns.

1. my D800 does seem to have an easier time focusing on this new target. The AF Range (with solid green dot) is larger.

2. The new target does reduce the variation of between the midpoint calculation on each run by 50%.

So it does support your thinking that this is a better target for testing.

However, I notice something that I haven't seen being brought up.

1. After I focused with LV, I hit the button to switch back to view finder. In this process, it seems like there are enough camera shake that the focus is already off. (I test it by immediately switch back to LV and found that the green rectangle turned red.) I tried to reduce the camera shake by putting a sand bag on top of the D800, but it doesn't seem to help enough to avoid that issue.

2. When I test the edge for each test, it seems to have some unpredictable variation. For example, if +5 is the edge, within the same test run, sometimes it shows solid but sometimes it shows blinks. And I changed the AF Tuning setting again, it could be solid again. So when we change the AF Tuning setting, does it has any physical moving parts? That might support this type of variation.

Partial result with the new target:









--
Just trying to improve my draft one image at a time
You could check our more of my photos at http://www.joshuaphotographystudio.com

 

[caspianm]

Did you set exposure delay on? Works well.

 

[ZionImage]

Did you set exposure delay on? Works well.

With this Dottune method, I'm not taking any photo at all. Would that still help?

 

[caspianm]

Never used the dot tune before and wasn't aware of what was involved in procedure. Good luck