p.1 #1 · Tip: Use larger focus spot size for greater AF accuracy and consistency
In my testing yesterday of the 24-105 I found my landscape shots @ 24mm had much higher AF-S variability than my shots at 35mm-105mm. I theorized this may relate to using the "small" box for my flexible spot size - at wide focal lengths and/or subjects at far distance, there is a lower frequency of detail for the AF system to work with and perhaps that has a negative impact on AF accuracy/consistency, along with a lower chance of detail that is perpendicular to Sony's non-cross point AF array.
To prove/disprove this theory I took 90 AF-S shots with my A7rII locked-down on a tripod. All were AF-S flexible spot - 30 with the small flexible spot size, 30 with medium flexible spot size, and 30 with large flexible spot size.
Result: Moving from a small to medium spot size increased AF accuracy/consistency by just under 4X. Increasing it from medium to large spot size showed no further improvement.
Conclusion: Use at least the medium spot size or larger when shooting with a wide-angle focal length and/or subject that is smaller and/or distant. The amount of AF improvement vs small will depend on the focal length, subject distance, subject size, and subject type.
Methodology: For a quick way to calculate focus accuracy/precision without having to evaluate every photo, I used the "FocusPosition2" field in the EXIF, which reports the lens focus position on a scale of 0-255. For a given focal length (24mm in this case) and distance (infinity), I visually examined a few photos to determine what FocusPosition2 corresponds to best focus. I then examined the FocusPosition2 value for all 30 photos in each set and calculated the std. dev. Here are the results:
AF-S small box: Std. Dev 2.705
AF-S medium box: Std. Dev 0.673
AF-S large box: Std. Dev 0.712
Here is the scene I used - focus is on distant house near top of hill at center of frame: Scene Image (Full Size)
p.1 #2 · Tip: Use larger focus spot size for greater AF accuracy and consistency
snapsy wrote:
In my testing yesterday of the 24-105 I found my landscape shots @ 24mm had much higher AF-S variability than my shots at 35mm-105mm. I theorized this may relate to using the "small" box for my flexible spot size - at wide focal lengths and/or subjects at far distance, there is a lower frequency of detail for the AF system to work with and perhaps that has a negative impact on AF accuracy/consistency, along with a lower chance of detail that is perpendicular to Sony's non-cross point AF array.
To prove/disprove this theory I took 90 AF-S shots with my A7rII locked-down on a tripod. All were AF-S flexible spot - 30 with the small flexible spot size, 30 with medium flexible spot size, and 30 with large flexible spot size.
Result: Moving from a small to medium spot size increased AF accuracy/consistency by just under 4X. Increasing it from medium to large spot size showed no further improvement.
Conclusion: Use at least the medium spot size or larger when shooting with a wide-angle focal length and/or subject that is smaller and/or distant. The amount of AF improvement vs small will depend on the focal length, subject distance, subject size, and subject type.
Methodology: For a quick way to calculate focus accuracy/precision without having to evaluate every photo, I used the "FocusPosition2" field in the EXIF, which reports the lens focus position on a scale of 0-255. For a given focal length (24mm in this case) and distance (infinity), I visually examined a few photos to determine what FocusPosition2 corresponds to best focus. I then examined the FocusPosition2 value for all 30 photos in each set and calculated the std. dev. Here are the results:
AF-S small box: Std. Dev 2.705
AF-S medium box: Std. Dev 0.673
AF-S large box: Std. Dev 0.712
Thanks for posting this. You tested the 24-105, but hasn't this been for other lenses, too? I've never had much luck with flexible spot-small and quit using it. (When I want small, I use focus-center, but obviously your can't move that one around.)
At the beginning of your post I thought you were going to say/test that flexible-small didn't work well at 24, but it still did ok beginning at 35. Did you happen to measure whether there was any improvement in focusing by going from a small spot to larger at those longer focal lengths?
(That's a slick way you found to measure AF efficacy -- nice!)
p.1 #3 · Tip: Use larger focus spot size for greater AF accuracy and consistency
I'm wondering if A9 users would have the same issue, with faster readout and more AF points and coverage?
[And a bit off topic: I'm also wondering if/when a 'cross-type' on-sensor PDAF implementation will emerge, and how other on-sensor PDAF implementations compare here, including Canon's 'every pixel is a PDAF sensor' Dual-Pixel AF]
p.1 #5 · Tip: Use larger focus spot size for greater AF accuracy and consistency
I will only use small focus spot for shots with large apertures (f1.4 - 2.0). The rest it’s either medium, large or wide area focus, depending on the type of subject. For landscapes on a tripod, i’d prefer to use manual focus.
p.1 #7 · Tip: Use larger focus spot size for greater AF accuracy and consistency
RobertHolloway wrote:
Hi Snapsy, I'm not trying to be smartass, just wondering why you'd use AF for a landscape shot, particularly if a tripod was involved.
I'm actually surprised that giving the camera a larger focus area yields more consistent AF. But heh...
Thanks, Rob
---------------------------------------------
normie610 wrote:
I will only use small focus spot for shots with large apertures (f1.4 - 2.0). The rest it’s either medium, large or wide area focus, depending on the type of subject. For landscapes on a tripod, i’d prefer to use manual focus.
I think he was using a tripod in this case precisely because he was making a test of the AF.
p.1 #8 · Tip: Use larger focus spot size for greater AF accuracy and consistency
Parariss wrote:
Thanks for posting this. You tested the 24-105, but hasn't this been for other lenses, too? I've never had much luck with flexible spot-small and quit using it. (When I want small, I use focus-center, but obviously your can't move that one around.)
At the beginning of your post I thought you were going to say/test that flexible-small didn't work well at 24, but it still did ok beginning at 35. Did you happen to measure whether there was any improvement in focusing by going from a small spot to larger at those longer focal lengths?
(That's a slick way you found to measure AF efficacy -- nice!)...Show more →
I didn't measure beyond 24mm for this test because I only noticed the inconsistency @ 24mm for the subject matter I'm working worth. I suspect the cutoff for the size affecting the AF consistency will be related to focal length, subject distance, subject content, etc... And I did notice this inconsistency @24mm on the 24-70GM as well.
johnctharp wrote:
I'm wondering if A9 users would have the same issue, with faster readout and more AF points and coverage?
[And a bit off topic: I'm also wondering if/when a 'cross-type' on-sensor PDAF implementation will emerge, and how other on-sensor PDAF implementations compare here, including Canon's 'every pixel is a PDAF sensor' Dual-Pixel AF]
I don't think a faster readout would help, unless the A9 performs multiple CDAF confirmation cycles as part of its AF-S logic. I'm not sure about dual-pixel but I know Olympus's on-chip PDAF are cross-point.
RobertHolloway wrote:
Hi Snapsy, I'm not trying to be smartass, just wondering why you'd use AF for a landscape shot, particularly if a tripod was involved.
I'm actually surprised that giving the camera a larger focus area yields more consistent AF. But heh...
Thanks, Rob
Naturally AF would be faster. Depending on the scene and technique (like the tip this thread is about) it would approach MF's consistency as well, esp. on MILCs like Sony which use a CDAF confirmation cycle for AF-S.
p.1 #9 · Tip: Use larger focus spot size for greater AF accuracy and consistency
snapsy wrote:
In my testing yesterday of the 24-105 I found my landscape shots @ 24mm had much higher AF-S variability than my shots at 35mm-105mm. I theorized this may relate to using the "small" box for my flexible spot size - at wide focal lengths and/or subjects at far distance, there is a lower frequency of detail for the AF system to work with and perhaps that has a negative impact on AF accuracy/consistency, along with a lower chance of detail that is perpendicular to Sony's non-cross point AF array.
To prove/disprove this theory I took 90 AF-S shots with my A7rII locked-down on a tripod. All were AF-S flexible spot - 30 with the small flexible spot size, 30 with medium flexible spot size, and 30 with large flexible spot size.
Result: Moving from a small to medium spot size increased AF accuracy/consistency by just under 4X. Increasing it from medium to large spot size showed no further improvement.
Conclusion: Use at least the medium spot size or larger when shooting with a wide-angle focal length and/or subject that is smaller and/or distant. The amount of AF improvement vs small will depend on the focal length, subject distance, subject size, and subject type.
Methodology: For a quick way to calculate focus accuracy/precision without having to evaluate every photo, I used the "FocusPosition2" field in the EXIF, which reports the lens focus position on a scale of 0-255. For a given focal length (24mm in this case) and distance (infinity), I visually examined a few photos to determine what FocusPosition2 corresponds to best focus. I then examined the FocusPosition2 value for all 30 photos in each set and calculated the std. dev. Here are the results:
AF-S small box: Std. Dev 2.705
AF-S medium box: Std. Dev 0.673
AF-S large box: Std. Dev 0.712
I feel the theory also applies across different systems. For the last 12 months I have been using a slightly wider AF area of d9 or d25 with AF-C specifically on Sigma WA Art series lenses of 35mm and less. This made a quite difference on bodies like the D750/D810 and D850.
p.1 #10 · Tip: Use larger focus spot size for greater AF accuracy and consistency
I've been complaining in another thread over my frustration with AF with the 12-24 @ 12mm. The camera seems to randomly grab focus anywhere in the massive DOF, generally leaving either the foreground or the background OOF. I have been using Small Flexible Spot, but maybe I could get a better compromise from Wide or Zone.
p.1 #11 · Tip: Use larger focus spot size for greater AF accuracy and consistency
sandycrane wrote:
I've been complaining in another thread over my frustration with AF with the 12-24 @ 12mm. The camera seems to randomly grab focus anywhere in the massive DOF, generally leaving either the foreground or the background OOF. I have been using Small Flexible Spot, but maybe I could get a better compromise from Wide or Zone.
Yes, and the issue is figuring out where to focus. I really need to test test test with the 12-24 and figure that out before I commit to purchasing it. I love the size of it, but after using it for a week, I'm scared to use it without a tethered computer to check focus.
p.1 #12 · Tip: Use larger focus spot size for greater AF accuracy and consistency
sandycrane wrote:
I've been complaining in another thread over my frustration with AF with the 12-24 @ 12mm. The camera seems to randomly grab focus anywhere in the massive DOF, generally leaving either the foreground or the background OOF. I have been using Small Flexible Spot, but maybe I could get a better compromise from Wide or Zone.
AF on a lens as wide as the 12-24 is arguably more of a liability than a benefit. It's nothing unique to the lens, the same is true of all ultrawides. Stop it down to 5.6 and use DMF to see the DoF.
p.1 #14 · Tip: Use larger focus spot size for greater AF accuracy and consistency
I've found this to be true with Batis 25/18 primes on my A7 too. Small flexible spot, despite turning green, very often misses focus, whilst med spot is usually good.
