[licht96]

Thank you as well from my side, Kobie.
I'm a little bit with sebberry, struggling more or less with inconsistent AF.
I shoot manly persons and since I'm using the more advanced (AUTO) systems it seems to be not so accurate than my K3II (but this is a little unfair since I manly use Spot AF-S on that old camera).
But exactly THAT was the reason buying the K3III: more trust in subject recognition, more reliable AF-C, more trust in AUTO function... let's say more close to use the camera as a Mirrorless one.
But my main issue is (I sound like a scratched vinyl ;-)) still the different behavior of the AF points in AF-S (!) over the frame... I use the AF joystick only in Live View since I get bad results with my workhorse 16-50 2.8 PLM and D FA 50 1.4 at wide aperture.
And I transfer that behavior in AF-S to AF-C: may be the camera decides with the RGB sensor the correct point of all the 101...but may be it's like in AF-S: when choosen correctly an outer point: out of focus (field curvature of the lens (someone mentioned that) , or whatever...).

---

Last edited by licht96; 04-09-2022 at 09:45 AM.

[sebberry]

I think a few people have commented on focus differences across the different focus points. Seems to be worse with some lenses than others. I do see this with my K-5 as well.

We need to get more people to test the accuracy of the other focus points to see if it's a problem with a few copies of the camera or just the design of it.

When the camera gets a sharp photo, the results are beautiful. But it's frustrating getting there and I'm feeling a little buyer's remorse.

[snappyhoffy]

Excellent research and explanations, I too will have to keep this thread for reference, Thank you

Maybe it could be made a Sticky Thread if the Moderators would allow.

[c.a.m]

@Kobie, thanks for the extensive effort that you've put into understanding the various AF issues and approaches on the K-3 Mark III. It's a complex beast, and, as others have noted, the Operating Manual is sorely lacking detailed explanations for the AF modes, settings and use cases. Seems that it's up to the users to explore, test, and share info such as yours.

I noted a couple of questions in your excellent presentation. First, the missing AF-point indication in the Playback display, as shown in one of your samples. On my K-3 III, the red AF point appears in the Playback display only if the AF back button has been pressed at the time of shutter release (if the user is using the AF back button only). So, in AF.S or AF.C, the user must continue to keep the AF button pressed while taking the shot. Of course, this isn't an issue if AF is actuated by the shutter-button half-press. I don't know whether other conditions could preclude the AF Point's display, even if the AF button had been pressed.

An area that I think is worthy of additional investigation is the tuning of the frame burst rate to the subject's speed or motion while in AF.C mode. That's an interesting observation. I recall that a Ricoh Imaging designer (in the K-3 III Challengers series) mentioned that a faster continuous-shooting rate improves the focus prediction accuracy -- there is less movement between shots, so the position prediction can be better. The Ricoh guy didn't comment specifically on subject speed or the nature of the movement, so I think your finding sheds light on choosing an appropriate burst rate under the conditions you mention.

Again, thanks for the details that you've presented in this thread. Among your other conclusions, I especially appreciated your observations concerning the 'busy' scenes, low contrast between the subject and background, and movement that might distract or 'fool' the AF system.

- Craig

---

Last edited by c.a.m; 04-09-2022 at 04:14 PM.

[Kobie]

@Kobie, thanks for the extensive effort that you've put into understanding the various AF issues and approaches on the K-3 Mark III. It's a complex beast, and, as others have noted, the Operating Manual is sorely lacking detailed explanations for the AF modes, settings and use cases. Seems that it's up to the users to explore, test, and share info such as yours.

I noted a couple of questions in your excellent presentation. First, the missing AF-point indication in the Playback display, as shown in one of your samples. On my K-3 III, the red AF point appears in the Playback display only if the AF back button has been pressed at the time of shutter release (if the user is using the AF back button only). So, in AF.S or AF.C, the user must continue to keep the AF button pressed while taking the shot. Of course, this isn't an issue if AF is actuated by the shutter-button half-press. I don't know whether other conditions could preclude the AF Point's display, even if the AF button had been pressed.

An area that I think is worthy of additional investigation is the tuning of the frame burst rate to the subject's speed or motion while in AF.C mode. That's an interesting observation. I recall that a Ricoh Imaging designer (in the K-3 III Challengers series) mentioned that a faster continuous-shooting rate improves the focus prediction accuracy -- there is less movement between shots, so the position prediction can be better. The Ricoh guy didn't comment specifically on subject speed or the nature of the movement, so I think your finding sheds light on choosing an appropriate burst rate under the conditions you mention.

Again, thanks for the details that you've presented in this thread. Among your other conclusions, I especially appreciated your observations concerning the 'busy' scenes, low contrast between the subject and background, and movement that might distract or 'fool' the AF system.

- Craig

Very good point about the use of BBF, if the button isn't being pressed at the time of capture, no AF point will be selected.
I'll be doing quadruple checking on my findings regarding the burst rate and correlation of AF tracking but in the meantime, think of my current findings like this...
Say the subject is travelling along the Z-axis (coming towards you) at 60 mph to your eyes. The camera sees it as say ft or Meters per second. The camera doesn't have a multiple calculation to differentiate between distance traveled with 11 fps vs 7 fps. It only knows predicted distance of subject travel period.
Now say you're pretty close to the subject (whether zoomed in, or physically) and you're shooting at 7 fps but every other shot is out of focus. You switch to full high speed 11 fps and it's keeping up no problem.
The next subject travelling at 30 mph and you're still in high speed 11 fps. The predictive algorithm is calculating where the expected location will be for the subject as it travels towards you just like it did before, but this time, the subject is moving slowly. The camera doesn't know if the movement is from you, or the subject, it only knows the approximate distance change of what you're trying to target. So with a slower moving target with you in full high speed, you'll end up with a higher number of shots that have fallen in the gaps of the AF points or AF points that are all over the place because of incremental change of distance on a slower moving subject plus your own framing changes/movement. I know it sounds odd and doesn't make sense, but from my experiences of trying everything I could think of, this is an actual thing.
The Pentax presentation wasn't wrong or misleading, on a fast moving/erratic subject, 11 fps will absolutely get the keepers and be more accurate since it can better predict incrementally for extreme and unpredictable movement. The thing they never mentioned are subjects and scenes for using 11 fps burst rate. It was a blanket statement that the predictability has been improved by increasing the speed of the burst rate. It doesn't mean that it's the be all end all of getting the best chance of keepers when tracking subjects though.
It's like cars now days. Pretty much every car now has EBA (Emergency Braking Assistance). You wouldn't want that to be how the car brakes all the time though right? It still makes the car stop, but is only necessary when in an emergency situation when you need the full pressure of the brakes to engage right away to stop as soon as possible. Driving in traffic with full high speed braking wouldn't exactly be ideal right? You wouldn't be able to judge when the car would stop so you'd always be stopping before you intended to. However, on the highway and something happens, that's when the full high speed braking really comes in handy to safely drop the cars speed before hitting the vehicle in front of you.
Same idea with the AF and different burst rates. Each is ideal for a given situation, and neither of them are ideal for all situations.

[Larrymc]

On my K-3 III, the red AF point appears in the Playback display only if the AF back button has been pressed at the time of shutter release (if the user is using the AF back button only). So, in AF.S or AF.C, the user must continue to keep the AF button pressed while taking the shot. Of course, this isn't an issue if AF is actuated by the shutter-button half-press. I don't know whether other conditions could preclude the AF Point's display, even if the AF button had been pressed.- Craig

Craig, the K-3 Mk l was the same way, you had to keep the AF back button pressed when taking the shot for the AF point in focus to be registered in the EXIF Maker Notes Data. I discovered this while trying to determine why some images were registering AF points in focus and some weren't even though good focus was achieved on all images. I think the Mk III functions the same way, if using the AF back button, the button has to be held down until the shot is complete,

Larry

[SPLewis]

As we know, the bells and whistles that are imbedded into the K-3 III when it comes to AF can be, well, are confusing.
"Do I use subject recognition? Zone, Full Auto, SEL-EX (S, M, L?), SEL, Spot, AF.S, AF.C" and the list goes on and on and on.
Through the past year of concentrated shooting to figure out all the ins and outs of this new to Pentax Marvel of AF, I'd like to help others understand what the camera is doing when set to certain settings and when you should use one setting over another.
Keep in mind that my findings are not endorsed nor has any info been provided by Pentax. This is solely my own trial and error over the many firmware updates that have been released.
Your mileage may vary which I will attempt to also explain why in this post.
Note I will be focusing on the latest firmware as of this posting (1.41). From my experiences, previous Firmware's (although improvements end over end) were spotty regarding reliability if you used any sort of tracking AF setting.

Okay! Let's begin...
We'll start by looking at the camera itself to show the correlation of AF point vs intended subject over a variety of scenes.
Full Auto with a busy scene
Here are all the settings that were used. Notice the AF red dot that shows where the camera placed focus.


Now let's zoom in and see EXACTLY where it focused in the scene itself.

So what the heck happened here?! Why is it out of focus when it clearly showed an AF point was used?

Well, there's a few things at play here.
For one, the AF points on the screen or in the OVF don't actually mean the subject is in focus.
They only show which AF point was active during capture (if any).

The 2nd thing is using the Type-1 AF setting (which links both the RGBir metering sensor and the AF system together).
I personally leave this setting on, but it's important to know what to expect when using it.
On the surface it sounds incredible! 307,000 pixels to assist the 101 point AF system for improved accuracy and speed in tracking? It's a no brainer right?
Well, not so fast. Yes it does track very well and very fast, however, 307,000 pixels don't divide that well into 101 AF points.
There's gaps between the 101 AF points which many times is the reason you'll get images that show no AF points were used at all, or you'll get a sharp image but the AF point is just off the target.

I think what's happening is that the RGBir sensor is telling the AF system where to focus but there's no actual focus point at that location so it picks the next closest actual AF point.
In other words, the RGBir sensor is picking AF locations that fall in the "cracks" between the AF sensors 101 points.

Now the 3rd thing is how busy is the actual scene? The subject recognition uses color, shape, and a database of pre-programmed things to recognize.
If the scene has many shapes and/or colors (especially if they're similar to the subject itself) it will make it more difficult for the camera to isolate the subject for focusing. This is especially true if using the "Full Auto" setting that uses all 101 AF points which is shown in the above picture (for this scene, Zone would be a better choice (if you wanted the eye detect AF to be more reliable).

Zone AF with a busy scene
Just like before, take a good look at the screen and note the AF point location


Now let's zoom in and see how it looks!

Well that's weird! Why is this one so much more in focus?!
By using Zone, you're limiting the area the camera will scan for moving subjects and things to focus on.
This is a great setting to use for complicated scenes with a subject large enough to fill most of or all of the Zone area.
Don't forget, you can also move the Zone around with the joystick just like using Sel AF.

Here's an example of using Zone with a subject on the smaller side of things.


And let's zoom in...

Well that totally failed. So, ummm, what happened this time?
Color information is too similar between both the subject and the background so it picked an AF point that it believed to still be the correct subject.

For better or worse, I made a video regarding the "PENTAX" folder that's in the SD cards, what some of the data contains and how it correlates to the subject recognition.
Surprise!!!! Clouds is one of the subjects! So if you're experiencing issues with maintaining focus on a subject when using Full Auto or Zone with a cloudy background, it's probably because the camera is recognizing more than one subject at a time and doesn't know exactly which one to stay latched to.
Don't forget that both Auto and Zone AF allow for the recognition of eyes in the scene (as long as subject recognition is turned on).

So what settings should be used in AF.C to capture certain subjects since it seems to be really picky about everything?
It's not as complicated as it seems to be on the surface.

Larger subjects are tracked beautifully with no issues at all (Motorsports, birds in flight that take up at least a full 1/4 of the image area).

But wait!!! What about the shooting speed? Isn't a faster burst speed better?
Well, I've done a lot of playing around and analyzing this, and these are my findings...
The burst rate to use is actually determined by the speed of the subject you're trying to capture.
I know, say what? What does the burst rate have to do with the cameras focusing system?
Well, from all my field testing, it actually does, but not in the "traditional" sense.
The camera has a new predictive algorithm for focus targeting estimation and that's an important thing to understand.
For example, if you're shooting a car coming towards you at 100 mph (160 Km/h) and it's taking up approx 3/4 of the image area and you're shooting with continuous MED (7 fps), it's very possible for the camera to under/over predict the in focus location for the next shot. so the full 11 fps would get you more keepers as the camera better matches the subject speed with the "next frame" focus location. However, if the car is travelling at 40 mph (64 Km/h) and you're in full 11 fps burst, the AF location target will be a bit all over the map since the vehicle just isn't travelling fast enough for the cameras settings, so 7 fps would actually be the benefit here.
The same holds true for birds in flight. Faster smaller song birds require faster settings. Larger/slower birds require slower settings.

At the end of the day, the K-3 III is what I call a "deliberate" camera. You need to think about what you're shooting to apply the correct settings to maximize the keeper rate. Gone are the simple, meh, spot, single shot, I'll get something generalist approach. Now don't get me wrong, I know some of you are doing that and it's working for you, I'm not knocking it. I'm just saying that for more intense split second shooting, you need to apply some thought into the settings and base those settings directly to the subjects that you'll be shooting.
It took me quite a while to figure out what was going wrong because everything I did with my K-30 and K-3 just did not work so well with the K-3 III (to the point where I almost sent the camera back to get checked). But I figured there must be something I'm missing here so I just kept testing EVERYTHING to figure it out.

So, in a nutshell, smaller subjects erratic movement/faster motion = 11 fps.
Larger subjects with more consistent/slower movement = 7 fps.

Busy scene taking advantage of subject recognition: Zone AF
Isolated subject taking advantage of subject recognition: Full Auto AF (all 101 AF points).

Now, you can also use SEL S/M/L which I found does a great job of sticking with the subject under most conditions. It also allows you to start focusing from the center point before the camera will begin tracking.
Conditions where SEL M/L don't work seems to mainly be shooting subjects around water. I strongly think this (again) has to do with color information and the fact that the camera is a deliberate system as I explained before. When I'm shooting swans, ducks, geese out in the water, I switch to AF.S so the system won't try "finding" things that are moving.
It seems (to me anyway) that AF.C specifically "scans" for movement which causes AF issues such as jumping from the subject to something else moving. This is especially the case if the intended subject is barely moving but something else in the scene is moving more.
AF.S is how it's always been, so there's nothing new (except for Zone AF, the subject recognition and eye detection) that we aren't already used to.

Here's some more examples with the settings that were used.
Remember, where I've zoomed in is where the AF target is located.












Notice this one shows No AF target but it's still sharp



I probably missed a couple of things here and there (I'm currently under the Covid weather) but I hope this post helps at least one person out when coming to terms with the whizbangetry of the K-3 Mark III with all it's doohickies and whatchamacallits and how it all comes together into one amazing thingamabobber.

Thanks for doing this for us, Kobie!

[kodai84]

The OP did a really good analysis & helpful writeup of this which is very interesting to me, even though I don't own anything near a K-3iii, but this is all way to complicated for my simple mind. I still miss the split-prism in my old manual-focus ME-Super even though I doubt that would help me with birds in flight, and mostly stick with single center-point AF on my K50.

[gatorguy]

(I) mostly stick with single center-point AF on my K50.

Shush, don't repeat this, it's a secret...
I often stick with center-point AF on my K3III as well.

[gatorguy]

FWIW the experience of using the Pentax K3III to capture action seems not all unlike that of a Nikon Z9 shooter:

"I experimented with a series of autofocus modes, quickly ruling-out the use of face detection because the player with the ball was rarely the face that the camera would choose to focus on. I also tried 3D Tracking with the subject-recognition modes turned off, and found that, while it followed the chosen subject well, it didn't react quite fast enough to lock onto a subject that I was racing to keep the camera pointed at....In the end I reverted to using a small 'dynamic' AF point (a small AF point but with some consideration given to the area immediately surrounding it, if your subject is no longer precisely under the point)."

"To a degree, I felt like the Seattle Seawolves, who fought to stay in the game, scored two very good tries, but ultimately showed there's work still to do (with the Z9).

"Ultimately, I don't feel I succeeded in capturing the key moments of the game: I was pointing at the action as a first try was scored, only to find I'd set the camera to track the goal post, instead of the player. I got the focus right for another, only to end up with a context-free photo of a mass of bodies under which I have to assure you the ball was grounded. For the final try of the match I managed to zoom out and keep pace with the Seawolves' winger, only to find that, even fully backed-out, I was still too close to capture the crucial grounding."

None of the high-end cameras today with their overall excellent AF are as easy to succeed with as marketing would like to tell you it is. Practice, and more practice is needed. It won't all come automatically without real dedication by the photographer to understand what the system does, how it does it, and what the various settings accomplish in addressing your own personal use-cases. In that regard the Pentax K3III is no different than the Z9. Learn to work with them and they can perform wonders. Otherwise most any far less expensive camera with a decent resolution and set to auto will work just as well in most cases.

[BigMackCam]

FWIW the experience of using the Pentax K3III to capture action seems not all unlike that of a Nikon Z9 shooter:

"I experimented with a series of autofocus modes, quickly ruling-out the use of face detection because the player with the ball was rarely the face that the camera would choose to focus on. I also tried 3D Tracking with the subject-recognition modes turned off, and found that, while it followed the chosen subject well, it didn't react quite fast enough to lock onto a subject that I was racing to keep the camera pointed at....In the end I reverted to using a small 'dynamic' AF point (a small AF point but with some consideration given to the area immediately surrounding it, if your subject is no longer precisely under the point)."

"To a degree, I felt like the Seattle Seawolves, who fought to stay in the game, scored two very good tries, but ultimately showed there's work still to do (with the Z9).

"Ultimately, I don't feel I succeeded in capturing the key moments of the game: I was pointing at the action as a first try was scored, only to find I'd set the camera to track the goal post, instead of the player. I got the focus right for another, only to end up with a context-free photo of a mass of bodies under which I have to assure you the ball was grounded. For the final try of the match I managed to zoom out and keep pace with the Seawolves' winger, only to find that, even fully backed-out, I was still too close to capture the crucial grounding."

None of the high-end cameras today with their overall excellent AF are as easy to succeed with as marketing would like to tell you it is. Practice, and more practice is needed. It won't all come automatically without real dedication by the photographer to understand what the system does, how it does it, and what the various settings accomplish in addressing your own personal use-cases. In that regard the Pentax K3III is no different than the Z9. Learn to work with them and they can perform wonders. Otherwise most any far less expensive camera with a decent resolution and set to auto will work just as well in most cases.

No such thing as a free lunch, right? The camera can only do so much.

I'm grappling with Topaz AI image processing software at the moment. In the right circumstances, with the right images, it can perform wonders (especially the denoise and sharpening tools) - but rarely without a good deal of intervention from the user. Some knowledge of the software and skill in using it is required to get (fairly) consistent, satisfying results. Folks who simply accept the automatic settings will, undoubtedly, be disappointed with it...

[vimlr007]

I did some testing of my Pentax-D FA* 70-200mm F2.8 - normally I use on a K1 mkii -- but I really wanted to test out the APS-C AF on my new K3 mkiii -- the dogs were running FAST towards me and I was using a mono-pod. After the first shot just about all were crisp and sharp.

I will post the pre-processed picture later -- as I am interested to know what the focal points were ... the post LR processed picture is currently third in this month's photo competition.


Domesticated ? - PentaxForums.com

[c.a.m]

After the first shot just about all were crisp and sharp.

Excellent! Great example image, too.

Have you updated the firmware to v. 1.41? Any comments on the AF performance compared to earlier FW versions?

- Craig

[vimlr007]

I had forgotten that the dogs were so off center -- the the AF.C kept them in focus as they raced towards me.

The focus points used are not held in the EXIF are they ?

---------- Post added 13th Apr 2022 at 08:54 ----------

Excellent! Great example image, too.

Have you updated the firmware to v. 1.41? Any comments on the AF performance compared to earlier FW versions?

- Craig

I have updated to v1.41 but not had a chance to do more similar tests -- trying - but labs are not always co-operative !

[sebberry]

I had forgotten that the dogs were so off center -- the the AF.C kept them in focus as they raced towards me.

The focus points used are not held in the EXIF are they ?

---------- Post added 13th Apr 2022 at 08:54 ----------


I have updated to v1.41 but not had a chance to do more similar tests -- trying - but labs are not always co-operative !

Fast moving dogs at 1/320 sec on a focus point not in the center.

And yet when I can't get a docile duck in focus I'm told that the off-center focus points aren't very accurate and I should crank up the shutter speed.

I give up