[Medex]

Hello, pentaxians,

Yesterday I was asked by my facebook friend to do "image in focus repeatability test", using new Pentax-D FA* 50/1,4 on the Pentax K-1 camera.
I always focus with central focusing point and recompose after, but my friend is using non-central points too.
I have done lens focus microadjustment +6 after I bouhgt the lens, and focus with central point is accurate in most situations (about 70 percent).
But I was completely disappointed with non-central point focusing ability - absolutely all images are slightly out of focus. I tried different focusint points on the edge of focusing area, I shot different objects, but every time they were slightly out of focus (not so much, but not tack sharp as using central point). The accuracy of central point was about 70 percent.
Then I tried another lens - FA* 85/1,4. And about 70 percent of shots were in good focus using non-central focusing point.
This "test" was done handheld, at the evening, using 400W halogen lighting, ISO400. 10 shots series were done, every time I defocused the lens manualy.

What would be your thoughts and suggestions?
My friens said that he has such a problem with 24-70/2,8 lens, that focused pretty good 2 years ago.
Could the firmware updates be associated with focus points decalibration? I have heard that after update some people complain about the need to remake focus microadjustments of the same lenses.

---

Last edited by Medex; 09-21-2018 at 12:14 AM.

[eccs19]

Are you doing these tests wide open? I don't know the characteristics of this lens, but maybe it's not so much missing focus, but maybe it's just not as sharp off center wide open. Not uncommon with a lot of lenses, so perhaps this is the issue, generally not sharp in the area of question vs missed focus..??

[CarlJF]

For one thing, only the central points are f2.8. All the others are f5.6 points. They're OK for general shooting but expecting them to give accurate and reproducible results with a f1.4 lens is a stretch... The're certainly not the best way to get accurate focus in thin DOF situation with a large aperture lens. So, there's nothing wrong with your results. If you close down to f5.6, the rated aperture of non central focus points, I'm pretty sure everything will be sharp.

[Medex]

Are you doing these tests wide open? I don't know the characteristics of this lens, but maybe it's not so much missing focus, but maybe it's just not as sharp off center wide open. Not uncommon with a lot of lenses, so perhaps this is the issue, generally not sharp in the area of question vs missed focus..??

The new DFA* 50/1,4 is highly corected star lens. In case of uneven sharpness across the frame this lens would be total fiasco for the Ricoh. Could not believe it could be the case.

---------- Post added 09-21-18 at 04:04 PM ----------

For one thing, only the central points are f2.8. All the others are f5.6 points. They're OK for general shooting but expecting them to give accurate and reproducible results with a f1.4 lens is a stretch... The're certainly not the best way to get accurate focus in thin DOF situation with a large aperture lens. So, there's nothing wrong with your results. If you close down to f5.6, the rated aperture of non central focus points, I'm pretty sure everything will be sharp.

Where from you know about focus point sensitivities? Is this information true for all pentax cameras/sensors? The information in PF Camera section says that K-1 has 25 cross type focus points. Are they different in sensitivity compared to central focus point?
Just curious...

---

Last edited by Medex; 09-21-2018 at 06:05 AM.

[CarlJF]

Where from you know about focus point sensitivities? Is this information true for all pentax cameras/sensors? The information in PF Camera section says that K-1 has 25 cross type focus points. Are they different in sensitivity compared to central focus point?
Just curious...

It's certainly written somewhere in the K-1 manual. I don't have a K-1, thus I can't give you the exact page number but here's a quote from the K-1 review on PF, p. 18:

As on the K-3, the center point and the two points directly above/below it are able to take better advantage of faster lenses (up to F2.8).

The default for all other points is f5.6. True not only for Pentax but for Canon and Nikon as well, focus points are f5.6 unless specified otherwise. There's even f8 points on some camera, not very accurate but still working with slow lens with TC.

They have the same sensitivity in terms of light level, but they don't have the same accuracy. The f5.6 points will be in focus according to the DOF the image will have a 5.6. If you use 1.4, it means the focus might be off because the DOF is much thinner, but the focus would have been somewhere in the plane of focus at 5.6... The same is true for the 2.8 points. They're significantly more accurate than the 5.6 ones, but might not be 100% accurate when using a 1.4 aperture.

So, what you experienced is a fundamental limitation of the PDAF system as implemented today in DSLR. This is where CDAF, in Liveview, or manual focus (particularly in LV with focus peaking) will give you better results since they will work based on the DOF of the final picture.

[Franc]

start using a tripod while doing test. Handheld will fluctuate the points

[Medex]

start using a tripod while doing test. Handheld will fluctuate the points

ok. I did test on the tripod with IBIS off and on. ISO 100, exp. 1/200, f1.4. The same result. About 50 percent in good focus. Much better than yesterday handheld.
Then I changed the location where there is more light. ISO 100, exp. 1/800, f1.4. On the tripod with IBIS on. Result is little better but not so good as using central point (90 percent in very good focus).
Yesterday I shot ISO400, 1/400, F1.4. This means 2x less light was available.

It seems that k-1 has just 3 autofocus points sensitive to f2.8. All other 22 cross type points are sensitive to f5.6. Maybe here is the answer why I got little better focus accuracy in better light using non central AF points and central one.

But in real life we shoot handheld in most situations.

So I did the shot in very well lit place. ISO 100, 1/8000, f1.4. Handheld. Non central cross type point. Result , - 40 percent in accurate focus.

---

Last edited by Medex; 09-21-2018 at 07:02 AM.

[CarlJF]

It seems that k-1 has just 3 autofocus points sensitive to f2.8. All other 22 cross type points are sensitive to f5.6. Maybe here is the answer why I got little better focus accuracy in better light using non central AF points and central one.

Light level has nothing to do with it. The points all have the same sensitivity to light, they're all rated at -3 EV. The diiference is simply because the 2.8 points are more precise than the 5.6 ones.

To make it simple, the way a phase detect autofocus system work is by splitting the image in two slits. When the two images are the same (in phase), it means the image is in focus. If the image in one slit is blurry relative to the other it means the focus is too far or too near. Now, the farther the slits are separated, the more accurate the system is. In a 2.8 point, the slit are a distance calibrated to give proper focus based on an aperture of 2.8. In a 5.6 point, the slits are nearer and calibrated for an aperture of 5.6. Why then not all points are calibrated for 2.8 ? It's because the larger distance between teh slits require a larger aperture to cover both. Thus, in a 2.8 point, you need at least a lens with an aperture of 2.8 to cover both slits. Since many lenses aren't 2.8 or larger, this would mean the AF system would not work with many lenses. Thus 5.6 is kind of the lowest common denominator since it probably covers 99% of lenses. The center points also have both pair of slits: 2.8 and 5.6. So, if a lens slower than 2.8 is used, they still work but fall back to the regular 5.6 slits and are then no better than any other points. On the other hand, if a lens with an aperture of 2.8 or larger is used, the center points will use their 2.8 slits to give more precise and accurate focus. And this is exactly what you observe with your 1.4 lens: the center 2.8 AF points gives you more precise and accurate focus than the other 5.6 ones. This is to be expected and what the 2.8 points have been designed to do.

[Medex]

Light level has nothing to do with it. The points all have the same sensitivity to light, they're all rated at -3 EV. The diiference is simply because the 2.8 points are more precise than the 5.6 ones.

To make it simple, the way a phase detect autofocus system work is by splitting the image in two slits. When the two images are the same (in phase), it means the image is in focus. If the image in one slit is blurry relative to the other it means the focus is too far or too near. Now, the farther the slits are separated, the more accurate the system is. In a 2.8 point, the slit are a distance calibrated to give proper focus based on an aperture of 2.8. In a 5.6 point, the slits are nearer and calibrated for an aperture of 5.6. Why then not all points are calibrated for 2.8 ? It's because the larger distance between teh slits require a larger aperture to cover both. Thus, in a 2.8 point, you need at least a lens with an aperture of 2.8 to cover both slits. Since many lenses aren't 2.8 or larger, this would mean the AF system would not work with many lenses. Thus 5.6 is kind of the lowest common denominator since it's probably cover 99% of lenses. The center points also have both pair of slits: 2.8 and 5.6. So, if a lens slower than 2.8 is used, they still work but fall back to the regular 5.6 slits and are then no better than any other points. On the other hand, if a lens with an aperture of 2.8 or larger is used, the center points will use their 2.8 slits to give more precise and accurate focus. And this is exactly what you observe with your 1.4 lens: the center 2.8 AF points gives you more precise and accurate focus than the other 5.6 ones. This is to be expected and what the 2.8 points have been designed to do.

Good knowledge and fantastic explanation. Thank you.

[normhead]

This is why I quickly revert to the inside points when focusing close in, and use manual focus and live view for any critical focus, on macros etc.

The outside points are fine for landscape or for things further away at higher F-stops. Once you get far away that you have a large hyperlocal DoF, the 5.6 points are good enough.

The points wide left and wide right are great for picking part of a scene to force your camera to a good hyperlocal setting, but often refuse to focus at all with wildlife or moving subjects.

[UncleVanya]

Light level has nothing to do with it. The points all have the same sensitivity to light, they're all rated at -3 EV. The diiference is simply because the 2.8 points are more precise than the 5.6 ones.

To make it simple, the way a phase detect autofocus system work is by splitting the image in two slits. When the two images are the same (in phase), it means the image is in focus. If the image in one slit is blurry relative to the other it means the focus is too far or too near. Now, the farther the slits are separated, the more accurate the system is. In a 2.8 point, the slit are a distance calibrated to give proper focus based on an aperture of 2.8. In a 5.6 point, the slits are nearer and calibrated for an aperture of 5.6. Why then not all points are calibrated for 2.8 ? It's because the larger distance between teh slits require a larger aperture to cover both. Thus, in a 2.8 point, you need at least a lens with an aperture of 2.8 to cover both slits. Since many lenses aren't 2.8 or larger, this would mean the AF system would not work with many lenses. Thus 5.6 is kind of the lowest common denominator since it probably covers 99% of lenses. The center points also have both pair of slits: 2.8 and 5.6. So, if a lens slower than 2.8 is used, they still work but fall back to the regular 5.6 slits and are then no better than any other points. On the other hand, if a lens with an aperture of 2.8 or larger is used, the center points will use their 2.8 slits to give more precise and accurate focus. And this is exactly what you observe with your 1.4 lens: the center 2.8 AF points gives you more precise and accurate focus than the other 5.6 ones. This is to be expected and what the 2.8 points have been designed to do.

Are you sure? I use center point AF exclusively and I have never noticed any issues with slower lenses. I am aware of the lack of precision in the non-central AF points and how this relates to aperture but can't fathom how your explanation can be true if my f5.6 lens can focus using the central point. None of the data I have read says what you have posted. I do see that when you get to f/8 then some of the focus points stop working but I think this is related to the overall brightness sensitivity of those sensors. I'm honestly not an expert here so I could be barking up the wrong tree. I have read a few pages of this article on DPReview that seems to confirm what I understand: Autofocus System Design: Nikon FX SLR (DF, D1-D5, D600-D850) Talk Forum: Digital Photography Review The problem with having all f/2.8 sensors is that the farther you are away from the center of the frame the more likely the sensor would not function. I think I understand that this is due to the submirror not being able to see the entire exit pupil of a large aperture lens when viewed from a more offset position but I could be wrong. In any case the topic is very complicated there are some diagrams I can link that show that at various f/stop only certain sensors are active Aperture and Focus Points | Autofocus Points | Peachpit

Would love to find a better set of explanatory articles but this is the best I have been able to find.

[BrianR]

Would love to find a better set of explanatory articles but this is the best I have been able to find.

Here's an article on Canon's system with the layout of the AF module (which illustrates some of what Carl has said) that might help:

Canon EOS DSLR Autofocus Explained

And since no ones linked it yet, the "Make the most of your PENTAX camera!" article shows which points on recent cameras are the extra sensitive type:

How to optimize focusing accuracy with large-aperture lenses / Beautiful Photo-life | RICOH IMAGING

I can't recall off hand when the first f/2.8 points started to appear on Pentax dslr's, but the k5ii(s) has one in the centre point.

[UncleVanya]

Here's an article on Canon's system with the layout of the AF module (which illustrates some of what Carl has said) that might help:

Canon EOS DSLR Autofocus Explained

That article doesn't agree with a lot of others. The question is how narrow is too narrow - the long article I linked seemed to imply that an f/2.8 sensor works to about f/5.6 accurately - but that f/8 is a bridge too far. I'm struggling with this so any help is welcome.

[CarlJF]

Are you sure? I use center point AF exclusively and I have never noticed any issues with slower lenses. I am aware of the lack of precision in the non-central AF points and how this relates to aperture but can't fathom how your explanation can be true if my f5.6 lens can focus using the central point.

Well, have you really read my explanation:

The center points also have both pair of slits: 2.8 and 5.6. So, if a lens slower than 2.8 is used, they still work but fall back to the regular 5.6 slits...

So, your 5.6 lens works just fine with the central focus point. But they do not take advantage of the higher precision offered by these points.

None of the data I have read says what you have posted. I do see that when you get to f/8 then some of the focus points stop working but I think this is related to the overall brightness sensitivity of those sensors. I'm honestly not an expert here so I could be barking up the wrong tree. I have read a few pages of this article on DPReview that seems to confirm what I understand: Autofocus System Design: Nikon FX SLR (DF, D1-D5, D600-D850) Talk Forum: Digital Photography Review

This says exactly the same thing as I explained above, but in three pages instead of a few sentences. The goal here wasn'.t to explain in depth all the technical intrecacies of a modern PDAF system. It was rather to explain in a concise an simple way that focusing points aren't all equals by design. Some are 2.8, some are 5.6, and some are f8 (although not on Pentax bodies). The f2.8 ones are more accurate by design and it's normal and expected that the OP get more shots in focus with these f2.8 points than the 5.6 ones.

In any case the topic is very complicated

Yes it is. There are plenty of articles on the web and a few thread on PF discussing this topic in depth, if people want to know how it works under the hood. But for this specific thread, there's no need for the OP to go to know more than there are 2.8 points and 5.6 points and that 2.8 are more precise and accurate than 5.6 when using a large aperture lens.

[UncleVanya]

Thus, in a 2.8 point, you need at least a lens with an aperture of 2.8 to cover both slits. Since many lenses aren't 2.8 or larger, this would mean the AF system would not work with many lenses. Thus 5.6 is kind of the lowest common denominator since it probably covers 99% of lenses. The center points also have both pair of slits: 2.8 and 5.6.

Well, have you really read my explanation:

I did but I missed the highlighted text. This explains it. I'm no longer totally lost. Thanks.