[BruceBanner]

Yesterday a rather nice Pentaxian on one of the many FB Pentax communities put me onto something I wasn't aware of before... 'Focus Shift'. I mean I was kinda aware of it, in a way that I stopped relying on using FA (Fine Adjustments) and Focus Charts so much because it became apparent to me one of a few things would happen;

1) Set up the Focus Chart test equipment up on another day and results may differ (possibly due to not having extremely accurate and fixed points for the equipment to be housed in-conjunction with focus chart etc.
2) Over time perhaps the lens 'changed' and those values no longer hold true.
3) Different apertures require different values, why are we only calibrating to wide open apertures? (especially if this is not an aperture many shoot at).

Point No.3 is what this video is all about below, and I never knew it had a specific name 'til now, Focus Shift. Please watch, I think you'll find the video informative;

Youtu.be

Youtu.be

Ok so I thought I would put my recently acquired FA77 through the Focus Chart testing this morning and see what values I got through the following apertures, f1.8, f2.8, f4, f5.6 and f8. I had actually never done the focus chart with the FA77 yet, but I was happy to see that my 'out in the wild' adjustment of +3 for f1.8 was confirmed correct when actually doing the focus chart test, below are the rest of the outcomes;



As you can see I also tested HD DA 35 Macro as well, and perhaps true to what this fella was saying in the video that Macro lenses are superior to non Macro lenses in this regard, the variance for Focus Shift was far less. Maybe I shouldn't have sold my DFA 100 after all...


Just as this guy talks about in the video, there doesn't seem to be a brand of camera yet that allows for FA values to change accordingly throughout the aperture range. Perhaps however with Pentax we are a little graced with a somewhat workaround. Here's what I did...

You'll notice that with my FA77, +3 was fine for apertures ranging 1.8-2.8, and that +5 would be better for apertures 4-5.6, and lastly f8 requiring +8. You could do this;

Create a User Mode, call it 'FA77 f1.8-2.8' and set the AF Fine Adjustment (Custom Menu, Item 25 for the K-1) to being '3 Apply One' (and in this case for me set to +3). Then... create a second User Mode, call this one 'FA77 f4-5.6' and set the AF Fine Adjustments to being '2 Apply All' (and for me this time the setting would be +5). Once those User Modes are saved and stored you now at least have two user modes to go between that increase accuracy of the shot at varying apertures, you just need to toggle a User Mode if you drift out of the range too much.
It's a shame that you can't use '2 Apply All' and save that setting value per User Mode, but unfortunately it appears at least with my K-1 that during power downs and power ups it recalls only the last setting value used. Hence why you would use both '3 Apply One' and '2 Apply All' to get differing values for the same lens.
If shooting at f8 I would just have to manually increase the FA value for the second User Mode (ie '2 Apply All') or even just use Manual Focus (as things are a little more forgiving at this more stopped down aperture). Write down on the lens cap F8=+8 as a mental note or something.

Things are a little easier for my HD DA 35 Macro, I could indeed just set it to '+2' and have a single User Mode ('3 Apply One') as a 'compromise between wide open (2.8) to f8 (as it doesn't alter that much).

It's not the best workaround, I mean I guess if you really use a lot of different lenses and they all need varying degrees of different FA values then this idea goes out the window. Perhaps tho there is a theme to your lenses, that the more you stop down the more the FA value needs to change, perhaps that knowledge in itself might be handy in the field. Moving simply up from f1.8 to f8, bump the FA value up quickly also, and just don't sweat it too much etc.

I am curious however to see if other users are also discovering this issue with their own lenses, would be quite fascinating to gather large amounts of data on what FA values people are finding they are using on common lenses, perhaps there being a very common adjustment theme for specific lenses!

Cheers!

Bruce

[stevebrot]

For the most part, focus shift with modern (post mid-70s) lens designs are an artifact of pixel peeping. Some shift is almost always there, but is mitigated by increased DOF; however, magnification (pixel peeping) defeats DOF.

I have not seen much focus shift in the lenses on my shelf with the exception of my MC Rokkor-PF (Minolta) 58/1.4. It has visible shift in the split-image finder from f/1.4 to f/2.8 amounting to about a centimeter at 1 meter distance! Other than that, it is a fine lens.


Steve

[stevebrot]

I have not seen much focus shift in the lenses on my shelf with the exception of my MC Rokkor-PF (Minolta) 58/1.4.

FWIW, I just tested my FA 77/1.8 in the range f/1.8 to f/8 with focus established at f/1.8 using split image (KatzEye) on my K-3 with no obvious shift. A more rigorous test would require something like the Reikan or LensAlign systems to allow statistical evaluation of focus accuracy to detect focus shift.

Edit: Repeated using magnified live view with same result.


Steve

[pschlute]

ou'll notice that with my FA77, +3 was fine for apertures ranging 1.8-2.8, and that +5 would be better for apertures 4-5.6, and lastly f8 requiring +8.

Bruce, I have heard of focus shift. But one thing needs to be clarified. I don't know how far you were from your focus chart when doing these tests but I assume 10 feet. At f1.8 the DOF would be 49% in front of subject and 51% behind. At f8 that changes to 44% and 56% respectively. This change in DOF would give the appearance of the focus point moving back (back-focus). How are you eliminating this effect from your calculations ?

[BruceBanner]

FWIW, I just tested my FA 77/1.8 in the range f/1.8 to f/8 with focus established at f/1.8 using split image (KatzEye) on my K-3 with no obvious shift. A more rigorous test would require something like the Reikan or LensAlign systems to allow statistical evaluation of focus accuracy to detect focus shift.

Edit: Repeated using magnified live view with same result.


Steve

Bruce, I have heard of focus shift. But one thing needs to be clarified. I don't know how far you were from your focus chart when doing these tests but I assume 10 feet. At f1.8 the DOF would be 49% in front of subject and 51% behind. At f8 that changes to 44% and 56% respectively. This change in DOF would give the appearance of the focus point moving back (back-focus). How are you eliminating this effect from your calculations ?

Thanks both for the replies.

First of all, when watching his video a couple of things stuck out of me that didn't make a whole lot of sense. Perhaps he addressed this in the video and I missed it, but all the while when he was doing the Focus Charts, it was all in Live View mode, and we're using the OVF for AF FA... right? I think he just did the LV stuff for the video to illustrate, but not as a meaning to do the testing with LV?

The other thing, pointed out by a friend of mine, was that if shooting at say F8 or F5.6, the aperture is wide open anyway, so if you FA for wide open, that should be all that matters, how exactly in that split second when the shutter closes and the aperture stops down does that affect focus so dramatically has escaped us both...

Nonetheless, I try not to be biased and swayed by anything said, I repeated my focus chart testing in a manner I thought would be akin to how he shows his results in the video (this point here

What is focus shift and why every photographer should be aware of it? - YouTube

Here he shows what I'm assuming is no FA on that lens, just each shot taken at the same distance to the chart with only varying apertures throughout (but of course refocusing each time per shot, and I'm also hoping he was doing it with OVF and PDAF rather than LV like he kinda illustrates )

So I set out the same way, I got my FA77 as close to the focus chart as possible that centre AF point would get a lock on (green square), and not a red square (oof, too close).

The thing with doing the test at f1.8 is it's easy to see any front or back focusing, and still relatively ok to see at apertures such as f2.8 and even f4. It's when it gets to f5.6-8 that I am having to look quite far up/down the chart values to see which numbers are more in focus than the others (like the top '8' vs bottom '8').
Nonetheless I did repeat shootings at those apertures, refocusing each time, taking the time to study the effects, adjust as necessary, and the figures I presented above seem to be an accurate representation. So to answer your question pschlute, I didn't vary any of my distances, I was flying by the seat of my pants here, I got no instructions on how to test this so I just assumed it was a regular focus chart test done with differing apertures, that's all

But what I found interesting (and backs up Steve saying that the Focus Shift is mitigated by DoF) is that afterwards, I went outside, put a bottle on the wall, and shot at it with f5.6 and f8, and playing around with FA from 0 to +10... yeah makes very little if any difference, bottle was in focus the whole time at an acceptable level. I mean the oof blur ring (or whatever it's called) could be shifting up and down, but really in a regular outdoor shot... who can really tell or care...

Perhaps this is something that needs to be cared about with only certain lenses and only at those very wide apertures, after f4 etc, who cares, the shot/DoF will take care of itself etc.

[BigMackCam]

The thing with doing the test at f1.8 is it's easy to see any front or back focusing, and still relatively ok to see at apertures such as f2.8 and even f4. It's when it gets to f5.6-8 that I am having to look quite far up/down the chart values to see which numbers are more in focus than the others (like the top '8' vs bottom '8').

Unless you're shooting subjects at very short range with faster apertures, depth of field is going to be enough that the effects of focus shifting aren't really significant. As @stevebrot said, you'll only become aware of the shift if you pixel peep. The fact that you have to look carefully to see where the precise point of focus is at f/5.6 - f/8 means it's simply not an issue. It's easy to become obsessed about getting absolute focus accuracy, when for all practical purposes, "good enough" is all that's needed much of the time. Also, bear in mind that AF isn't 100% accurate all of the time, yet most of us accept that as a limitation of the AF system. So long as our photos still look good, the margin of error is acceptable.

I tend to calibrate my AF fine adjustment for each lens based on wide open aperture, such that it gives me accurate results at reasonably close (not "macro" close) distances. Stating the obvious, the fastest aperture gives the shallowest depth-of-field, so any focus inaccuracy is most noticeable there...

[pschlute]

Bruce. The point of focus shift is that the optics have a different focal point when the lens is wide open say 1.4 compared to say f8 FOR THE SAME POSITION OF THE FOCUS RING. Thus a lens focussed wide open but the image captured at f8 will not have the correct point of focus in the resultant image. As has been said this only applies to fast lenses and in most cases is negligible. It also gets disguised by the dof

My point to you about methodology is the fact that distribution of dof changes as you stop down in addition to the range of dof. So I would suggest that simply looking at your test picture is not an accurate method of determining if focus shift has happened at all. You claim to need an extra AF-FA at f8 but I suggest this may simply be the dof distribution change you are seeing

---

Last edited by pschlute; 10-01-2018 at 04:44 AM.

[UncleVanya]

I'm not sure that apply one and apply all are added together. Can someone connect on that aspect?

[BruceBanner]

Bruce. The point of focus shift is that the optics have a different focal point when the lens is wide open say 1.4 compared to say f8 FOR THE SAME POSITION OF THE FOCUS RING. Thus a lens focussed wide open but the image captured at f8 will not have the correct point of focus in the resultant image. As has been said this only applies to fast lenses and in most cases is negligible. It also gets disguised by the dof

My point to you about methodology is the fact that distribution of dof changes as you stop down in addition to the range of dof. So I would suggest that simply looking at your test picture is not an accurate method of determining if focus shift has happened at all. You claim to need an extra AF-FA at f8 but I suggest this may simply be the dof distribution change you are seeing

Gotcha (maybe... ). So perhaps another way to test this would be to set the focus charts up again, wide open at f1.8 with the FA77, use AF to get the focus on teh chart, then flick the switch to MF, take the shot (without touching the focus ring, then change the aperture and repeat throughout those main aperture ranges and see the results? i.e. just don't continue to refocus each time.

[BigMackCam]

I'm not sure that apply one and apply all are added together. Can someone connect on that aspect?

They're not added together.

Interestingly, though, if you get into debug mode on, say, a K-5, you can dial in an overall offset. Apply One or Apply All will then be added to that overall offset. I did this with my K-5 to bring it in line with my K-3 and K-3II, so I end up using the same AF fine adjustment settings on all three cameras...

[BruceBanner]

I'm not sure that apply one and apply all are added together. Can someone connect on that aspect?

Never said they were...

What I said (or tried to say but failed lol), is that with for example the FA77, if you have a User Mode 1, and you set the AF FA option to being No. 3 (Apply One), give it a value (pretend +3) then that value is stored and recalled for that lens whenever No. 3 is selected. But if in User Mode 2 you don't have No. 3 selected in the AF FA option, and now instead use Option 2 (Apply All), you can give that a new different value (such as +5 in my case), and now the lens attached (in this case the FA77) can use those values so long as you're in that (User) mode. As soon as you rotate the User Mode dial back to User Mode 1, it will drop back to +3 because that user mode is set to use No. 3 (Apply One) of the AF FA option. Get me? There is no adding here, no 3+5=8 or anything, just a way to make one lens attached to the camera have different AF FA values (at the turn of a dial).

[pschlute]

Gotcha (maybe... ). So perhaps another way to test this would be to set the focus charts up again, wide open at f1.8 with the FA77, use AF to get the focus on teh chart, then flick the switch to MF, take the shot (without touching the focus ring, then change the aperture and repeat throughout those main aperture ranges and see the results? i.e. just don't continue to refocus each time.

You must not allow the camera to refocus at all if you are trying to diagnose focus shift. To do so would defeat the object of the test

Your proposed method will still not deal with the issue of dof distribution

Why do you suspect focus shift may be an issue ?

[UncleVanya]

The problem was that the selected numbers looked like the added to 5 but one was the user mode and not an adjustment... Lol. Bad reading on my part.

[CarlJF]

Lets not forget that we work with photography equipment, not some kind of high precision astronomical or scientific devices. So, yes, there is some variation and they are normal and expected. Although there can be focus shift when closing the aperture, in practice this doesn't matter for photography. Because the increase in DOF will more than cancel any slight shift the focus may have (unless the lens is defective, which is totally different matter). So, basically, just enter the adjustment for the widest aperture and don't overthink or overwork it, unless doing it as some kind of scientific experiment. It should work good enough for more than 99% of photography needs.


If you need extremely accurate focusing, the best is still to work in manual focus. No matter how far you go in defining precise adjustment, autofocus will still be less accurate for fundamental reasons.

[stevebrot]

First of all, when watching his video a couple of things stuck out of me that didn't make a whole lot of sense. Perhaps he addressed this in the video and I missed it, but all the while when he was doing the Focus Charts, it was all in Live View mode, and we're using the OVF for AF FA... right? I think he just did the LV stuff for the video to illustrate, but not as a meaning to do the testing with LV?

Optical evaluation for focus shift requires accurate and precise fine focus technique. That is why live view is used; it is the gold standard. Using PDAF fine adjust to evaluate focus shift just adds another set of variables. My opinion (based on reads of over a hundred threads on AF fine adjust) is that most people over-adjust or adjust based on spurious or inadequate data. That is why I added the qualifier that use of one of the available focus evaluation software packages might be in order when trying to detect subtle focus shift, even with manual focus under live view.* That being said and out of the way, a +8 adjust from wide open is anything but subtle. I did not see that kind of shift with my lens. My unsolicited advice is to redo using live view.

FWIW, focus shift results from uncorrected spherical aberration and is most evident at close distances. While once not unusual and part of the lore of measurebating film photographers when I was young, it was usually not an issue under field conditions. Ansel Adams covers the subject in the 1980 edition of "The Camera" with suggestions for refocus stopped down if there is any doubt. His solution was in regards to a particular class of large format lenses and involved refocus on specular reflections using a magnifying loupe on the view camera ground glass; think of it as magnified live view. Focus shift used to be part of photo magazine lens tests, but as optics improved, that aspect of testing was dropped. Now days, the term is more frequently applied to focus stacking with Nikon than to lens aberrations. Significant focus shift is more of a novelty of the past than a serious and ongoing concern.

Roger Cicala dedicates two paragraphs of his article on using fast lenses to the subject. His short discussion is concise and to the point. In fact, the whole article is excellent for putting the "must have fast primes" discussion in perspective and may shed light on why many of the experienced (older) users on this site take a fairly dim view of the current call for ultra-fast wide angles and such.

Lens Rentals | How to Shoot With Wide Aperture Lenses



Steve

* I have seen example scattergrams from PDAF evaluation and they are fairly alarming. PDAF, while quite adequate, remains a fairly blunt tool for fine focus.