24-90mm Focus Shift (Diglloyd)

[ramarren]

Advertisement (gone after registration)

Dear me, no insult intended, merely drawing your attention to the contradictions in what you have said. It can't be rude to state the facts, surely?

 

And as I have said, my points of reference are peer group cameras that generally, usually, in fact pretty much always manage to AF accurately when used within their design envelope whereas my SL and 90mm lens' default mode is not to do so. That is why it is, as I understand the current state of play, acknowledged as a known fault which is under investigation for a solution. My expectations appear, therefore, to be in line with Leica's expectations of their own product. The fact that they are higher than yours is a matter of personal choice on both our parts!

 

 

The way you say things is what makes the expression rude, Tim. And there are no contradictions in what I've said.

 

Your experience may be different from mine, but what I have said has been 100% consistent as it reflects MY experience, not yours. 

 

My Nikon F6 generally focuses very accurately but occasionally misses. My Nikon D750 ... the same. The Olympus E-M1 ... the same. The Olympus E-5 ... the same. I could go through all the AF cameras I've had between today and 1998 when I first bought an AF camera and, while the latest ones have become generally more accurate and faster, they still all misfocus now and then. The Leica SL and SL-24-90mm lens is right in the same ballpark: most of the time it's right on the money with perfect focus, occasionally it misses. In ALL cases and for ALL cameras, IF the AF misses, I tweak the setting manually. 

 

However, I don't give a damn about comparing the SL to an A7 or a D750 or anything else. I am interested in how the SL performs for my photography, not how it compares to other cameras. 

 

I just put the 24-90 on 90mm and f/4, set it on zone single auto focus, and took a walk around the block shooting 100 exposures at all distances, from minimum focus to more than 500 feet away. All but two of 100 exposures hit exactly the focus that I wanted; the two that didn't were simply the usual issue of the subject being more complex than the AF system could evaluate properly and needing either spot AF focus or a manual focus tweak.

 

I can't call that anything but completely satisfactory performance.

Edited March 6, 2016 by ramarren

[CheshireCat]

I just put the 24-90 on 90mm and f/4, set it on zone single auto focus, and took a walk around the block shooting 100 exposures

 

 

This is not the proper test for aperture focus shift, as the lens at 90mm f/4 is wide open.

 

Proper test for aperture focus shift is the following:

- Camera on tripod.

- Lens focal length set to 90mm.

- Subject at about 5 meters, in the center of the frame.

- Manual focus on subject (just once).

- Take different shots from f/4 to f/11, stopping down 1 stop each time. Do not refocus !

- Compare the results. Each subsequent shot must be sharper in the center than all the previous ones (until diffraction kicks in).

 

Please let us know the results.

Edited March 6, 2016 by CheshireCat

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[tashley]

Cheshire, I agree with you. Focus shift, as it is commonly used (your case #2) means the behaviour of a lens which when focussed at one aperture, shifts to another plane of focus when its aperture is changed. Problematic in lenses which are focussed at one aperture and which subsequently stop down to shooting aperture.

 

As I've said (somewhere!) above I don't think that's the problem with the 24-90 because it seems to miss accurate focus at F4 (when shot at the long end) though i have not specifically tested for the effect.

 

I also don't think the problem I and others are seeing is caused by the so called focus point bug because the effect of mis focus is seen on planar targets.

 

There is something else going on. Not sure what, but I assume Leica is aware of it and working on it.

 

It strikes me as being just possible that what is happening is that the lens does have some slight focus shift (even F4 lenses can have this) and that there are algorithms designed to ensure that the point of focus stays within the DOF when changing apertures from the focussing aperture, and that this algorithm has some bugs or errors? What do you think?

Edited March 6, 2016 by tashley

[cpclee]

You have every right to be upset and to return the camera if something as crucial as AF fails to work.  Point is though that I and others have not been able to reproduce the problem (beyond the UI flaw I described) and it's not clear if this is a general flaw affecting all units and would surface only under very specific situations or is just a defect affecting a subset of the units.  It could also be usage error but we all want to rule that out.

 

But bear in mind that some of the issues reported clearly arose out of confusion.  An example is diglloyd's point about the crosshair shifting when magnification is on.  As I mentioned the CDF area / crosshair size should technically enlarge proportionally with magnification so Leica's UI design very mislead, and diglloyd clearly got confused by it.  There isn't a common single root cause to all the issues reported here.

 

Cpclee and ramarren,

 

I think (hope?) that most of us here probably understand that the AF 'points' (whether represented by a crosshair or a small box) on modern cameras (whether using PDAF or CDAF or hybrid) are actually not points at all, but small areas. I also think that it's mostly understood that as soon as the photographer is working with an area and not a point, the chance arises that parts of the subject field which stand at differing distances from the sensor might both/all fall within that area and therefore provide room for ambiguity and therefore apparent error. For those of us who generally use AF rather than AF with MF tweak or pure MF, we need to be alive to this possibility and to use the AF system accordingly. This is why both of my examples have used planar targets and why, when testing for, for example, focus shift, we need to work with equipment such as a Spyder Lenscal. Even then we need to be aware of the possibility of field curvature in the optic, and of the fact that field curvature can be very complex and can change at differing distances and apertures.

 

It is, however, my experience with my other current equipment (Sony A7RII and Nikon D810) that modern systems have a very very high rate of accurate AF provided that the photographer understands their strengths and limitations and uses them accordingly.

 

It is also my firm opinion that, the SL is not particularly well suited to 'AF and then MF refinement before shot' because it has no split screen or prism and no ability to trigger magnification with a twist of the MF ring other than by button presses (I assumed it would have this but haven't yet found it. If I'm wrong, I stand corrected and would like to know how to achieve it.)

 

In other words, the marketing materials of the camera and the operational design of the camera (not to mention its price) taken into conjunction with the performance of its peer group ( which I'd mainly consider to be any high DSLR or mirrorless, most specifically the D5/1DX type camera but also D800 and Sony A7 series) strongly lead one to expect that it will have accurate AF when deployed correctly and under appropriate conditions. 

 

According to my pretty appropriate testing, my copy does not. One hears from some other people that theirs do not, either, including some fairly well-qualified testers. One also hears that Leica is aware of a focus bug and is working on correcting it. It seems therefore to me most likely to be the case that there is a bug and that if that bug is software based, it is universal and that, therefore, those who do not experience it are either lucky, shooting subject matter that doesn't trigger it, or simply not noticing it.

 

Nonetheless I will proceed with the hypothesis that there is a bug and that it does mean that, at least at the long end, a certain and significant proportion of images made within the normal shooting envelope of the camera will not be as well focussed as the lens is capable of and the punter has a reasonable right to expect.

 

If that hypothesis is correct, the question for each of us becomes 'should I put up with it until it is fixed?'

 

My personal answer to that is, 'no'. My reasons are that it doesn't work properly and that I need it to and that at the price I paid, I have a right to expect it to - as is indeed the fact, I believe, in consumer law. Hence my intention to return mine rather than wait for the outcome of an extensive beta program for new firmware, during which time my 'return window' might well close and during which time I will have a system that doesn't work properly.

 

Rammaren feels differently - seemingly unbothered by the problem, happy to use the camera differently and to await a fix. I get that. I drive a Tesla and it has satnav which is often so bad that it makes you laugh. I know of Tesla drivers who love their vehicles so much that they install TomTom systems or similar in order to make up for the failings of the built-in system. These things are highly personal and each of us is more or less in thrall to the gestalt of the brands to which we subscribe.

 

But I think it's important to separate these threads out. Can modern AF systems be misunderstood and therefore give unexpectedly poor results when used incorrectly? Yes. Can you test to ex-out that as a factor thereby determining not whether the system was correctly used but whether it correctly functions? Yes. Does the SL pass that latter test? For at least some of us, no. Does that matter? Yes or no, depending on what each of us expects for her/his money and how each of us uses the camera. But let's not muddy the waters by a mixed bag of implications that the 'problem' either doesn't exist or doesn't matter. Please.

[tashley]

You have every right to be upset and to return the camera if something as crucial as AF fails to work.  Point is though that I and others have not been able to reproduce the problem (beyond the UI flaw I described) and it's not clear if this is a general flaw affecting all units and would surface only under very specific situations or is just a defect affecting a subset of the units.  It could also be usage error but we all want to rule that out.

 

But bear in mind that some of the issues reported clearly arose out of confusion.  An example is diglloyd's point about the crosshair shifting when magnification is on.  As I mentioned the CDF area / crosshair size should technically enlarge proportionally with magnification so Leica's UI design very mislead, and diglloyd clearly got confused by it.  There isn't a common single root cause to all the issues reported here.

 

We'll no doubt find out soon enough! Either some of us have affected units or none of us do or all of us do, and there may well be more than one factor at play. I am pretty confident that mine is not as it should be but clearly I can't speak for other people's units. 

 

I appreciate your concern - but honestly, I'm not upset, just bored with having to try to bottom stuff out that should just work but.... c'est la vie..

Edited March 6, 2016 by tashley

[CheshireCat]

It strikes me as being just possible that what is happening is that the lens does have some slight focus shift (even F4 lenses can have this) and that there are algorithms designed to ensure that the point of focus stays within the DOF when changing apertures from the focussing aperture, and that this algorithm has some bugs or errors? What do you think?

 

Indeed, even f/4 lenses can have noticable focus shift, and even if this lens has 4 asphericals that should help with this issue, it is a zoom and cannot be optimized for all focal lengths.

However, given Leica's tight tolerances, the focus shift could be predicted by means of raytracing in order to create a "focus shift compensation table" in the firmware. This "compensation table" can be used by a simple (read: not compute-intensive) algorithm in the firmware that takes as input the following parameters:

- FL

- Aperture

- Focus distance (wide open)

And optimizes the latter for the actual shooting aperture just after the shutter is pressed (and the lens is stopped down to take the shot). This needs to be correctly timed.

 

Nevertheless, the real problem here is that... there are multiple problems here

 

If this lens/camera system really cannot reliably achieve correct focus wide open, then all engineering sophistications I described above are just science-fiction.

And if that is really the case, my very personal opinion is that the system cannot be called "professional".

Edited March 6, 2016 by CheshireCat

[talt03]

Advertisement (gone after registration)

This is not the proper test for aperture focus shift, as the lens at 90mm f/4 is wide open.

 

Proper test for aperture focus shift is the following:

- Camera on tripod.

- Lens focal length set to 90mm.

- Subject at about 5 meters, in the center of the frame.

- Manual focus on subject (just once).

- Take different shots from f/4 to f/11, stopping down 1 stop each time. Do not refocus !

- Compare the results. Each subsequent shot must be sharper in the center than all the previous ones (until diffraction kicks in).

 

Please let us know the results.

Has anyone tried this on SL with a Leica 90mm f/4 prime? How different is this focus shift between the SL 24-90 and the 90 prime?

And how about doing the same test with another pro camera with a similar lens? Do we have to assume that the other brand pro cameras have better AF if only the SL's AF has a problem?

Edited March 7, 2016 by talt03

[tashley]

Has anyone tried this on SL with a Leica 90mm f/4 prime? How different is this focus shift between the SL 24-90 and the 90 prime?

And how about doing the same test with another pro camera with a similar lens? Do we have to assume that the other brand pro cameras have better AF if only the SL's AF has a problem?

 

I've tried it on my D810 with 24-70 VR @70mm and that combination focusses perfectly, pretty much all the time. On the same large planar subject (at about a 230 metre distance) I've used the SL/24-90 @90mm and it is much more miss than hit. It's an interesting test because it shows that the actual pixel size of the subject is very similar between the two cameras despite the differing focal lengths due to the Nikon's higher pixel count.

 

I then tried it with the Sony A7RII and 24-70mm F4. This lens is pretty good for the size and price but its main weak spot is anything at 90mm and F4. Nonetheless, it focusses very nearly as accurately in AF as it does in MF even though the F4 shots are always a little soft. When I have returned my SL I will purchase the new 24-70mm F2.8 G Master lens for the Sony, which I hear is very sharp and which at 70mm will give a slightly larger actual subject @70mm than the Leica @ 90mm because of the extra pixels. Call it digital zoom.

Edited March 7, 2016 by tashley

[talt03]

This means we don't even have to look at pro cameras to have better AF hit rate than the SL system.

 

Unless someone comes up with the opposite result using his own SL sytem compared to this SL system, there seems to be a bigger probability of some bug or failure in the SL.

 

It can be either in the body or in the lens and either mechanical or in software/firmware.

 

If this appears for all SL systems, it may force Leica to start analyzing and adding a test for this oof defect in their final assembly and possibly recall units if the root cause were found to be mechanical.

Edited March 7, 2016 by talt03

[tashley]

I think the SL AF is very fast and accurate at most focal lengths and I'm sure they will fix it at 90mm.

[CheshireCat]

Do we have to assume that the other brand pro cameras have better AF if only the SL's AF has a problem?

 

Focus-shift is caused by a particular lens design and seen only when the lens is correctly focused wide-open and then stopped down.

 

If the AF is not even correctly focusing wide-open, then the problem is in the camera firmware (or hardware, depending on the AF technology used).

The SL has contrast-based AF, which is old technology and considered slower than more modern phase-based AF in other competing cameras.

I don't understand how Leica could claim their AF is "faster than any SLR system", but the "great haste, great waste" saying is the first thing that comes to mind.

Edited March 7, 2016 by CheshireCat

[jaapv]

This means we don't even have to look at pro cameras to have better AF hit rate than the SL system.

 

Unless someone comes up with the opposite result using his own SL sytem compared to this SL system, there seems to be a bigger probability of some bug or failure in the SL.

 

It can be either in the body or in the lens and either mechanical or in software/firmware.

 

If this appears for all SL systems, it may force Leica to start analyzing and adding a test for this oof defect in their final assembly and possibly recall units if the root cause were found to be mechanical.

Don't worry, Leica does not need to be forced to take issues like this seriously.

[cpclee]

That's not quite true. Phase detection has been around much longer and was already widely used in the film days.  Contrast detection using the image capturing sensor is a more recent development and only got comparably fast in the last few years.  

 

Mirrorless cameras are going to use contrast detection as the default method for AF.  Some companies add phase detection on top of that but the phase detection only switches on in certain modes (Eg. AF tracking).  The main advantage of phase detection is the camera knows which way to turn the focus without needing to evaluate contrast changes in both directions.

 

 

Focus-shift is caused by a particular lens design and seen only when the lens is correctly focused wide-open and then stopped down.

 

If the AF is not even correctly focusing wide-open, then the problem is in the camera firmware (or hardware, depending on the AF technology used).

The SL has contrast-based AF, which is old technology and considered slower than more modern phase-based AF in other competing cameras.

I don't understand how Leica could claim their AF is "faster than any SLR system", but the "great haste, great waste" saying is the first thing that comes to mind.

[cpclee]

Wikipedia has a nice entry that explains the pros and cons.  On contrast detection:  

 

"Contrast AF does not have this inherit design limitation on accuracy as it only needs a minimal object contrast to work with. Once this is available, it can work with high accuracy regardless of the speed of a lens; in fact, for as long as this condition is met, it can even work with the lens stopped down. Also, since contrast AF continues to work in stopped-down mode rather than only in open-aperture mode, it is immune to aperture-based focus shift errors phase-detection AF systems suffer since they cannot work in stopped-down mode. Thereby, contrast AF makes arbitrary fine-focus adjustments by the user unnecessary. Also, contrast AF is immune to focusing errors due to surfaces with repeating patterns and they can work over the whole frame, not just near the center of the frame, as phase-detection AF does. The down-side, however, is that contrast AF is a closed-loop iterative process of shifting the focus back and forth in rapid succession. Compared to phase-detection AF, contrast AF is slow, since the speed of the focus iteration process is mechanically limited and this measurement method does not provide any directional information. Combining both measurement methods, the phase-detection AF can assist a contrast AF system to be fast and accurate at the same time, to compensate aperture-based focus-shift errors, and to continue to work with lenses stopped down, as, for example, in stopped-down measuring or video mode.

Recent developments towards mirrorless cameras seek to integrate the phase-detection AF sensors into the image sensor itself. Typically, these phase-detection sensors are not as accurate as the more sophisticated stand-alone sensors, but since the fine focussing is now carried out through contrast focusing, the phase-detection AF sensors are only need to provide coarse directional information in order to speed up the contrast auto-focusing process."

 

I have only really used one-shot AF on the SL but in that mode it is extremely fast.  Focus is generally achieved instantly.  The native SL lenses are designed to maximize the speed of contrast detection.  The 24-90 was designed so that it has only one very light lens element to move.

[thighslapper]

I have just taken several dozen shots at 90/4 with OIS on and AF spot focus versus MF of orchids (some all white) and various other things in the conservatory at 5-7m and can find NO DIFFERENCE WHATSOEVER on 100% crops of the DNG's in LR. 

 

Shots taken at infinity and 60m few days ago at also show no difference either. I have been careful to eliminate any other potential variation. 

 

THERE IS a mismatch between the green spot focus cross and the actual point of focus ....... basically the real point on an unmagnified subject is the bottom end of the cross ....... which MAY under some circumstances be an issue.

 

If there is a software issue you would really expect it to be uniform and afflict everyone to the same extent ....... always assuming lens variation and user error is minimal .... 

 

That leaves the odd results using 'abnormal' targets like inclined pages of type and other test images ...... where the camera seems to sulk and misbehave.

 

I cannot reconcile this with the real world performance or reports of big differences such as Tims......... like most I have had no problems at all in everyday photography ...... and that included a day of portraiture done mostly at 90/4 where the only duds out of a 1000 or so images were due to camera shake and not using a fast enough shutter speed.

 

I have never used anything except spot focus ..... I suppose it's just a habit from RF M's ....... so I can't comment on the other modes on the SL.

Edited March 7, 2016 by thighslapper

[CheshireCat]

That's not quite true. Phase detection has been around much longer and was already widely used in the film days.  Contrast detection using the image capturing sensor is a more recent development and only got comparably fast in the last few years. 

 

I was obviously talking about on-sensor phase-detection AF.

 

Mirrorless cameras are going to use contrast detection as the default method for AF.  Some companies add phase detection on top of that but the phase detection only switches on in certain modes (Eg. AF tracking).

 

The A7R2 uses phase-detection for normal focusing (not only tracking).

It is actually a hybrid system, and my educated guess is it also uses contrast-detection for fine-tuning in the final step.

Edited March 7, 2016 by CheshireCat

[CheshireCat]

Wikipedia has a nice entry that explains the pros and cons.  On contrast detection:  

 

"Also, since contrast AF continues to work in stopped-down mode rather than only in open-aperture mode, it is immune to aperture-based focus shift errors phase-detection AF systems suffer since they cannot work in stopped-down mode."

 

That article is not accurate. The A7R2 can use PDAF up to f/8.

Edited March 7, 2016 by CheshireCat

[Jared]

About a month ago I had a portrait session where I was using the 24-90 at a distance of roughly 2m taking head shots for my daughter's play mostly at 90mm with the lens wide open. Several hundred images across fifty different actors/actresses. All but a couple achieved perfect focus on the nearest eye. I was using point autofocus and shifting it where required for the different compositions. I don't have an a/b comparison to manually focused images, but there was clearly no problem. I was already aware of the issue with the focus point not being exactly where expected and accounted for that. I did not run into Tim's issue and would have expected to if it were consistent in its behavior.

 

- Jared

[tashley]

Then maybe I have a dud? 

 

I went outside this afternoon and selected a planar target at over 200 meters away and shot at 90mm, five frames each at F5.6 and F8 and every one was notably soft (earlier I had done the same test but only at F4 and with several AF shots versus several MF shots -  the MF shots were perfect, the AF shots were soft).

 

Shutter speeds were easily high enough, everything was in order. I then shot my A7RII also at 5.6 and f8 at 70mm in AF and the difference between it and the Leica was instantly visible. The Sony was waay sharper.

 

So: I am pretty damned sure this isn't the 'crosshair mismatch' problem (the target is several times larger than the crosshair and is planar) and I'm pretty sure it isn't a focus shift issue because it's the same at F4 thru F8. It is quite simply focussing quite far forward of where it should.

 

Some of us have this issue and some don't. That either means that there's a unit-specific bug, which would imply that it isn't firmware related (if we're all running the same FW) or it's a widespread bug and some people can't tell a soft file when they see one. I doubt it's the latter.

 

So my current best guess is that some units have a hardware error.

 

We'll see.

Edited March 7, 2016 by tashley

[Advertisement]

Hallo,

Look here: Leica L-Mount

[k-hawinkler]

Hi Tim, do you have the option to try another SL camera/ lens combination?