Thanks Steve. Lately I find the difference of focus shift on differently colored cards(with plastic cover), and then on colored paper (no plastic). So they had been better targets than the shiny things, which I cannot convince myself and the Nikon guys that are valid targets.Now to look at it, I think it's more likely not the refection, but the color/contrast near the eye, that caused the shift. So I still maintain the defect of the device is the right answer. But of course I'm still testing for the right answer and will find Nikon if it turns out to be a defect.

liwei zhen wrote:Thanks Mako for all the help. Glad I posted here. Please see some comments and a little progress below:Mako2011 wrote:If you superimpose the arrays FOV (purple) over the red box in your picture...you'll see the contrasts the arrays are working with can vary greatly. Just something to consider when learning how best to use what we have An example...note how each specific array (of 39) has way more than one contrast to work with in this sceneLook at just the center array. It sees...top of box, bottom of box, lettering left, lettering right. And it's seeing in a different way than we are (RGB). Complicated stuff.I see the point. While my thought is that when we make sure the frame of focus point (shown in view nx/capture nx) are on one surface (parallel to focal plane) only, then we are sure the focus should not be out of that surface. Of course the camera could still slightly shock on my tripod, but not much. In the case of the last pic I pasted, I think the frame is far enough from the battery.The "no 45 degrees" thing is actually the same, IMHO. 45 degree surface is not good, but the focus plane should still be within the scope of the nearest point in the focus frame to the farthest point, while in my previous cases, the plane was shifted way out of the same card. This I think is practical an approach to test the shift from a longer distance: when we need the large reference "ruler" (all the cards hung on the wall) to see the shift for the much thicker DOF.Those said, I did make a similar setup as the blog you posted---but much shabbier. The shift difference when focusing on the differently colored paper still stands. But with more detailed reference than batteries, I can see that the difference when shooting from nearer place too. So the shifts (and difference) do exist at closer distance, just not as much as standing farther. Here is an example of the small difference at close range (forgot how close):closer shot with smaller difference on differently colored paper. 35mm @f1.8. not so good sunlight.Please note the scale on it is not accurate (bigger than real ruler), not sure how long they are at horizontal direction, maybe closer to the real thing.And there are errors since the tool set is light, not well made and I might position it and the camera slightly differently. But still the error doesn't account for all the shift difference on the paper.Tried 18-55 too, but the bigger DOF forced me to try at a close distance, while at the close distance, the focus difference could be very small---at least smaller than 35 at f1.8, and I can not safely tell. So I still need to try some more with much bigger reference object. Will do that with even better light and better tool kit.Much better set up to test focus! Definitely shows the focus shift with the red target. That said, I don't think you need worry about it to much. In both cases the target is in focus and the actual plane of focus is in the DoF. Focus shift like this has been documented before with other bodies and lens combinations. Not a unknown or totally unheard of phenomenon. Again, the 35mm f1.8g is often in the mix when things like this pop up. It is simply not unheard of to see a focus shift with some lenses when the target is in certain color spectrums. In this case, the target is still well within the dof at close distance so I would mark it down to "good/cool to know" and go take more pictures and see how it goes.  Good luck!

liwei zhen wrote:Thanks Steve. Lately I find the difference of focus shift on differently colored cards(with plastic cover), and then on colored paper (no plastic). So they had been better targets than the shiny things, which I cannot convince myself and the Nikon guys that are valid targets.Now to look at it, I think it's more likely not the refection, but the color/contrast near the eye, that caused the shift. So I still maintain the defect of the device is the right answer. But of course I'm still testing for the right answer and will find Nikon if it turns out to be a defect.That bright orange card is an extreme color. I would be more concerned about the front focus your test shot shows when using a more normal black pattern on white background.Am I correct in following your posts that the camera/lens combination back focused originally, but then front focused after an adjustment was made by Nikon service? If so, they just over compensated.I like the way my 35mm f/1.8 focuses on my D300 with an AFFT adjustment of -5. Without it, it back focuses a bit.I may try some different color targets myself someday to see if I can reproduce the AF focus shift. But right now I'm so busy with optics testing at work that I just don't feel much like doing more tests at home.Steve

Hi. The last comparison I posted was to show that at a closer distance, there are shift difference too. Plus it doesn't really show which target leads to better result: it looks that the red one cause some back focus, but at a longer distance, the black + white one cause (more stable) large front focus.At a longer distance, the focus shift/difference are largely amplified. This is understandable I think since the physical movement of the focusing group in the lens should be smaller when the focus place is farther. Let me show some more comparison with a much stable test setup.First the setup:The number on the reference scale are CMs. And the wooden box is much heavier so when changing the target, I won't move the thing by accident.And I got some time take some shoot this morning with modest sun light. And in that light the shifts are rather stable with each targets. The distance was like 1.7M-2M:The white card has the focus at around 15. The red card at 5-10.And more obvious comparison:The blue one is around -5. So it's a 20 CM, or 1X CM difference in real.The blue card was the original one causing deep back focus before the camera tweak. It's covered with plastic, but not much noticeable with eye from the angle and distance.With fluorescent, the shift are rather random even with the white + black card. So I won't post anything here.But with the 18-55 under fluorescent last night, the difference are not obvious. So there is a chance that all the shift randomness/differences are caused by the lens.

SteveL54 wrote:the right answer and will find Nikon if it turns out to be a defect.That bright orange card is an extreme color. I would be more concerned about the front focus your test shot shows when using a more normal black pattern on white background.Am I correct in following your posts that the camera/lens combination back focused originally, but then front focused after an adjustment was made by Nikon service? If so, they just over compensated.I like the way my 35mm f/1.8 focuses on my D300 with an AFFT adjustment of -5. Without it, it back focuses a bit.I may try some different color targets myself someday to see if I can reproduce the AF focus shift. But right now I'm so busy with optics testing at work that I just don't feel much like doing more tests at home.SteveHi, you are a real pro then, great.First the back/front focus tweak at the service, yes, they compensated, but I don't think overly, but rather wrongly. Because the focus was not consistent (random, different shift on different thing) before, and most of the high contrast targets are okay. After the tweak, those good targets (like the black/white card) becomes bad.I'm not concerned about the front focus, but rather the inconsistency.And the bright orange card issue: good to hear the point. I didn't use DSLR much. The only camera with dedicated PDAF focus module I experienced was a sony A58 + SAL35f18 (yes, I'm into cheap 50'ish lenses for my baby). Used for several month and sold, didn't notice anything like the shift difference.And some mirrorlesses: anything with that clear drawing---even with a bit shinny plastic cover, or computer-monitor-reflectiveness---are cozy target for them even in bad light (80W fluorescent for example).I put some more tests result with better setup in Mako's thread (or right above this post with flat view), please could you take some look?

liwei zhen wrote:SteveL54 wrote:the right answer and will find Nikon if it turns out to be a defect.That bright orange card is an extreme color. I would be more concerned about the front focus your test shot shows when using a more normal black pattern on white background.Am I correct in following your posts that the camera/lens combination back focused originally, but then front focused after an adjustment was made by Nikon service? If so, they just over compensated.I like the way my 35mm f/1.8 focuses on my D300 with an AFFT adjustment of -5. Without it, it back focuses a bit.I may try some different color targets myself someday to see if I can reproduce the AF focus shift. But right now I'm so busy with optics testing at work that I just don't feel much like doing more tests at home.SteveHi, you are a real pro then, great.Well, optical related applications have been a large part of my engineering work for the last couple of years. However, I can't claim to be an expert with Nikon autofocus issues. In fact, when using Nikon cameras at work I do not rely on AF. I typically use manual focus lenses, and once focus is achieved I often tape the focus ring to ensure it doesn't change during tests. But much of my tests involve machine vision type cameras where the lenses have a set screw to keep focus constant.First the back/front focus tweak at the service, yes, they compensated, but I don't think overly, but rather wrongly. Because the focus was not consistent (random, different shift on different thing) before, and most of the high contrast targets are okay. After the tweak, those good targets (like the black/white card) becomes bad.I'm not concerned about the front focus, but rather the inconsistency.The inconsistency is an important distinction. If you get obvious inconsistent results when using a well defined target where both camera and target are stable then something is really wrong with the camera. There will always be some slight shot-to-shot variation with AF, but it shouldn't be enough such that you would immediately recognize the difference except perhaps when viewing at 100%.Then there's the case where AF errors are related to differences in distance, color, and focal length setting. I'm surprised your camera responds so differently to the extreme colors of your target, but I haven't tested it myself. I have noted some lenses respond differently to a black and white target illuminated with different color temperature lamps. But it was not at the level that made an obvious case of AF failure.It's quite a challenge making a PDAF module these days that can obtain the precision required for a 24 MP APS-C sensor under a wide range of conditions.The problem might be something as simple as debris on the entrance to the AF module or the mirror.It could be that the mirrors are not aligned correctly. When you sent the camera in for service they should have checked alignment of the main and submirror. (I know for the D70 and like cameras they have a specialized device that projects a collimated beam of light into the mirror box. An inspection mirror placed in the bottom of the mirror box returns the collimated light and the position of the returned image verifies mirror alignment and flatness.) Then once the mirrors are aligned they can calibrate the AF of the camera with a computer.Or ir could be something isn't aligned correctly inside the PDAF unit.Anyway, I think it's worthwhile sending the camera and lens back in for a second look. I get the impression that they may do a quick once-over servicing on the first attempt in the hope that easily resolves the issue and they use the customer as the final QC check. The service technicians may be under some pressure to quickly turn around the jobs. They'll probably take a closer look at it if you return it a second time. Unfortunately, I've seen too many posts here where individuals have to send their equipment in multiple times to get their unit fixed.And the bright orange card issue: good to hear the point. I didn't use DSLR much. The only camera with dedicated PDAF focus module I experienced was a sony A58 + SAL35f18 (yes, I'm into cheap 50'ish lenses for my baby). Used for several month and sold, didn't notice anything like the shift difference.And some mirrorlesses: anything with that clear drawing---even with a bit shinny plastic cover, or computer-monitor-reflectiveness---are cozy target for them even in bad light (80W fluorescent for example).The contrast detection AF in other cameras (or live view with Nikon dSLRs) will generally be more accurate than PDAF. But they will have a harder time following your child around as he/she gets older and more active. I gave up on CDAF and went with a dSLR as soon as my sons started to play sports.I put some more tests result with better setup in Mako's thread (or right above this post with flat view), please could you take some look?Yeah, I'll take a look later today.Steve

Can you find another 35mm f/1.8 to test? That would help confirm if it is the lens or camera.It looks to me as though both camera and lens need to go in for adjustmentA suggestion for posting any future test photos. Do not combine photos together for comparison. It strips the EXIF meta data. The EXIF data can hold valuable clues to the problem. Post the photos as jpegs straight from camera. The DPR galleries limit the photo size so I'd try another photo posting service like PBase. (You should be able to get a short term trial account)Steve

Okay, first, during the days I tested focus with different focus points; and used line with good contrast to "lure" the center focus point to use its upper, bottom, left or right edge. The results show no trace of unaligned mirrors---If I understood correctly.Then today we finally got some good sun light here. I shot some more today in the office with much better targets designed according to the recent messy tests.The targets are printed papers, with red or blue blocks on it, and a white bar run almost through it. They cause quite some degree of focus difference, and the behaviors are quite stable.I uploaded the pics right to this web site. First the blue figure: you can see the focus is around 0-5.Then the red one, the focus was from 15-20 I think. So it's a 15 CM difference, and the reference board was at a less than 45 degree of angle with the table, so it's in real a 10CM+ difference.I also run into this one (reproducible): the focus looks very close to the red one, which confused me a little: could it be the horizontal detect line in the cross focus point behave differently from the vertical one on blue color? Then some later tests shows that it is not true, it could simply be that I covered less blue in this shot.Or it could be the positioning of the focus frame which covered blue, white, then blue again, then white again. With red figure, even the one showed above, whose focus frame wasn't totally in the red area too, the degree of shift is the same as other test with the red figure.Now it looks to me some heavy chromatic aberration that caused the problem: red and blue light focus on different position on the axis, and there is no correction in the focus sensor/camera that compensate the behavior. But could it be this heavy?Another not-so-certain observation is that the stronger the light was, the bigger difference was made by the colors. I tested also at 5PM at home (at around 1/200s, iso 400), the red one caused smaller shift.But in relatively strong light (both noon and 5PM), the black and white paper which frontfocus a lot, behave similarly to the blue one.Yes I feel like a free tester for Nikon. I clicked 3000+ pics in all kinds of conditions for them. Anyway I will go to Nikon to ask them to check again, and will bring the colored papers.

One more thing I forgot to mention: someone on the Chinese focum xitek also posted similar complaint about focus shifted to the farther eye when photoing people.The lens is 35f18g too, but the camera a D300s. I think the focus module is rather different from D5300's.

liwei zhen wrote:One more thing I forgot to mention: someone on the Chinese focum xitek also posted similar complaint about focus shifted to the farther eye when photoing people.The lens is 35f18g too, but the camera a D300s. I think the focus module is rather different from D5300's.1st... PDAF sensors are sensitive to red green and blue channels differently. That is normal and your tests show it. It's normal2nd... your targets are poor target for PDAF testing as there really is poor areas of contrast (white to blue/border). This is reflected in the way results differ as you rotate the targets. Again, normal and to be expected when using poor contrast target for PDAF testing.3rd... There is a AF-sub mirror behind the main mirror that you can not see. Slight variations in it's calibration can affect focus in different light and color channels. That's normal4th... dirt/debris on the AF unit at the bottom of the mirror box can affect focus and cause inconsistencies in different light and color channels.5th... Lens design can affect focus. The 35mm f1.8g is often one that brings focus "issues" up.Based on your testing to date...I don't think you have an issue that requires service.See link below for a better explanation.affect of color on PDAF

Mako2011 wrote:liwei zhen wrote:One more thing I forgot to mention: someone on the Chinese focum xitek also posted similar complaint about focus shifted to the farther eye when photoing people.The lens is 35f18g too, but the camera a D300s. I think the focus module is rather different from D5300's.1st... PDAF sensors are sensitive to red green and blue channels differently. That is normal and your tests show it. It's normal2nd... your targets are poor target for PDAF testing as there really is poor areas of contrast (white to blue/border). This is reflected in the way results differ as you rotate the targets. Again, normal and to be expected when using poor contrast target for PDAF testing.3rd... There is a AF-sub mirror behind the main mirror that you can not see. Slight variations in it's calibration can affect focus in different light and color channels. That's normal4th... dirt/debris on the AF unit at the bottom of the mirror box can affect focus and cause inconsistencies in different light and color channels.5th... Lens design can affect focus. The 35mm f1.8g is often one that brings focus "issues" up.Based on your testing to date...I don't think you have an issue that requires service.See link below for a better explanation.affect of color on PDAFThanks, the color/light condition affecting PDAF thing is nice to know. Still I wonder if it's common in DSLR world, that they may affect the focus that much, like the one shown with the eye of my boy, at f4.5.Or if it's normal for 35 18g, then maybe I should just stop using the lens. But in the situation, it may be silly to buy another lens.And if your point 3-4 are right, I do need some service for the camera. Besides, for the bad tweak they gave me alone, I almost cannot take any in focus picture now.As for the rotation thing: I tested later with vertical direction, and even 45 degree rotation with the blue paper, so long as I put the focus frame in the blue area, the behaviors are identical to the horizontal positioning.

liwei zhen wrote:Mako2011 wrote:liwei zhen wrote:One more thing I forgot to mention: someone on the Chinese focum xitek also posted similar complaint about focus shifted to the farther eye when photoing people.The lens is 35f18g too, but the camera a D300s. I think the focus module is rather different from D5300's.1st... PDAF sensors are sensitive to red green and blue channels differently. That is normal and your tests show it. It's normal2nd... your targets are poor target for PDAF testing as there really is poor areas of contrast (white to blue/border). This is reflected in the way results differ as you rotate the targets. Again, normal and to be expected when using poor contrast target for PDAF testing.3rd... There is a AF-sub mirror behind the main mirror that you can not see. Slight variations in it's calibration can affect focus in different light and color channels. That's normal4th... dirt/debris on the AF unit at the bottom of the mirror box can affect focus and cause inconsistencies in different light and color channels.5th... Lens design can affect focus. The 35mm f1.8g is often one that brings focus "issues" up.Based on your testing to date...I don't think you have an issue that requires service.See link below for a better explanation.affect of color on PDAFThanks, the color/light condition affecting PDAF thing is nice to know. Still I wonder if it's common in DSLR world, that they may affect the focus that much, like the one shown with the eye of my boy, at f4.5.Yes, a known and common phenomenon.Or if it's normal for 35 18g, then maybe I should just stop using the lens. But in the situation, it may be silly to buy another lens.Again, you're using targets and conditions known to cause focus inconsistencies. In some situations it's best to focus on a good spot/target then recompose.And if your point 3-4 are right, I do need some service for the camera.Not really, I said those issues can affect focus. Simply check the AF unit for dust...and try testing using better targets (black and white high contrast targets) in differing lighting conditions...to see if AF-Sub mirror adjustment is called forBesides, for the bad tweak they gave me alone, I almost cannot take any in focus picture now.Not using the targets you're currently showing us...and that is expected. Your blue and red targets  are very close to trying to focus on a blank white wall (as far as what the Arrays are seeing) and saying you have focus issues because it can't lock or is inconsistent when it does. That's expected and doesn't necessarily point to a focus problem. Same with your red and blue targets.Now try a newspaper on a wall compared to Liveview (in bright light). That's a better test to see if you actually do have a focus issue.As for the rotation thing: I tested later with vertical direction, and even 45 degree rotation with the blue paper, so long as I put the focus frame in the blue area, the behaviors are identical to the horizontal positioning.Again, the focus arrays have a vertical and horizontal component that looks for a contrast along the array. You are showing them a single tiny contrast (white to blue/red border) and not understanding how poor a focus contrast that is. Like trying to focus on a tiny black spot on a white wall. Tiny black spots are hard to focus on and that's expected.Targets like this are easy:And why they are used to actually see if there might be a need for service due to focus issues. You're, in a sense, asking the dealer to fix your car because it seems to pull to the left when you drive on the side of a hill. Not a perfect analogy but kind of demonstrates the issue.

liwei zhen wrote:.OOF on the red figure.These are taken after the tweak. f1.8, iso 800, around 1/160s, center point AF-S. handhold,Your camera instructions clarify subjects not parallel to the sensor can sometimes cause poor focus accuracy.In addition using AF-s if you move maybe just 1 inch after the camera has first focussed - the camera does not re focus.The pictures are OK, but not for critical testing of AF accuracy.

liwei zhen wrote:Sorry - once again this set up is not ideal for testing focus accuracy.It is vital to have a subject surface parallel to the sensor - the batteries are not suitable.The thin lines in the X in the circle could be too fine (as in thin) for AF to detect accurately.Leonard Shepherd Producing good quality photographs, or being good at sport or art, involves a little more than buying appropriate equipment. Practice, some learning and perhaps natural talent often play a bigger role than the equipment in your hands.

Mako2011 wrote:All Nikon DSLR camera instruction books caution geometric patterns may result in focus error.Not aligning exactly on the centre of the red circle could confuse the camera into focussing at the wrong distance.I would not rely on this type of target for a 100% reliable test.

Leonard Shepherd wrote:Mako2011 wrote:All Nikon DSLR camera instruction books caution geometric patterns may result in focus error.Not aligning exactly on the centre of the red circle could confuse the camera into focussing at the wrong distance.I would not rely on this type of target for a 100% reliable test.Hi, the manual does say "regular geometric patterns", and the example was a row of windows on the building, where the frame of the center focus point fall in one of the window.I thank what they wanted to say is that when shooting at those windows like pattern, all the intended focus points (if not single point) has a chance that fall into area with no contrast at all.I should think the board be a valid target, unless we focus very closely, make the focus point enclosed in pure black or white.

Leonard Shepherd wrote:liwei zhen wrote:Sorry - once again this set up is not ideal for testing focus accuracy.It is vital to have a subject surface parallel to the sensor - the batteries are not suitable.The thin lines in the X in the circle could be too fine (as in thin) for AF to detect accurately.Leonard Shepherd Producing good quality photographs, or being good at sport or art, involves a little more than buying appropriate equipment. Practice, some learning and perhaps natural talent often play a bigger role than the equipment in your hands.No, thanks. The targets could be rather bad. But two things: the batteries are not the targets. and the color pattern may have caused some stable difference here.I'm not using the targets again, but maybe later will try some better red + black patterns. Because it looks rather different from the red + white pattern.

Leonard Shepherd wrote:liwei zhen wrote:.OOF on the red figure.These are taken after the tweak. f1.8, iso 800, around 1/160s, center point AF-S. handhold,Your camera instructions clarify subjects not parallel to the sensor can sometimes cause poor focus accuracy.In addition using AF-s if you move maybe just 1 inch after the camera has first focussed - the camera does not re focus.The pictures are OK, but not for critical testing of AF accuracy.Thanks again. But don't totally agree. On one thing: in theory the camera should lock on a random place than my expectation, but still within the frame, right?The fact is the shift difference (I'm not talking which one is better here) is far beyond one paper.And the move thing: yes, there could be moves after AF lock, not as big as one inch I believe. while the shift difference is far bigger than that.Plus this result are very stable it self, or compare with parallel positioning of the cards.Those said, I don't believe this is a precise proof of shift difference too, but it leads to suspect and more test with the similar patterns. I didn't suspect the shift thing, until I see quite some real world pictures that didn't quite match my intention.

Mako2011 wrote:liwei zhen wrote:Thanks, the color/light condition affecting PDAF thing is nice to know. Still I wonder if it's common in DSLR world, that they may affect the focus that much, like the one shown with the eye of my boy, at f4.5.Yes, a known and common phenomenon.Run into Nikon today, and ask them to check the paper I printed---with the blue and red figures together, with a reference scale aside. They refused to test the paper in their environment, but I showed them with the paper on table and shooting from 45 degree---again, but the result is valid.It shows that under the sunlight, the red one has big front focus, the blue one basically okay, then under the light (don't know the type) of theirs, the red one relatively okay, the blue one with big back focus.It looks they don't experience this much (so I doubt how common the thing is), but they offer me one more D5300 + 3518g, one D3100 or 3200, and one 50mm f1.4 to test (whose chromatic aberration are heavy too, I think), I got basically the same result.I guess that's why I have to accept the fact that so big a shift difference is common, so I set your post as the answer this time.Still not clear about the reasoning though. Now I only wonder that with a FF 35mm f18g ED, what will become of the test result, or what if I do with a better body like D600/800. Not gonna purchase more of these though, as not deeming myself a DLSR user material.Thanks, I had my share of fun.

liwei zhen wrote:It is about 7 years since the number 1 tech guy at Canon USA said "Chuck Westfall “I recommend using a flat, detailed target parallel to the focal plane. After reading through the PDF linked from your message, it appears that the author has missed a major point, i.e., any individual focusing point in a digital SLR is much longer than the simple line he is using on his chart. The nature of the AF sensors used by EOS digital SLR’s as well as those from other manufacturers is that they perform most reliably when the entire length of the focusing area sees readable detail. This condition is not satisfied by a thin line on a piece of paper. It's OK to include an angled chart in a test photo. In fact, Canon Factory Service Centers always do this. But the test target is always separate from the angled chart, and parallel to the camera's focal plane”.A thin line scan can produce spurious resultsIt never ceases to amaze me why some Nikon users prefer to believe in fairy tales for testing AF instead of following the guidance from Nikon and Canon, and then holler without taking sensible tests to find out if the problem is them or the equipment.