[08amczb]

I've done some testing
The KP was set to AF-C, Sel L (start from center, than select from all), no AF hold. I selected the center point and did high speed burst and some photos after slowdown. The buffer fills very quickly, and it writes to a samsung pro 64Gb card after finishing the burst a minute approximetly. So the buffer clears slowly. Every lens correction was off, I cropped the center of the images and only my standard noise reduction settings was applied. All image taken with wide open aperture.
Here are the images per test case for evaluation:

My wife coming approximetly with 1m/s towards me and I'am using the DA* 50-135 @ 135:
AF-C Test Pentax KP DA* 50-135 Seres 1 | Flickr

My wife coming approximetly with 1m/s towards me and I'am using the DA* 50-135 @ 135:
AF-C Test Pentax KP DA* 50-135 Series 2 | Flickr

My wife coming approximetly with 3m/s towards me and I'am using the DA* 50-135 @ 135:
AF-C Test Pentax KP DA* 50-135 Series 3 | Flickr

My wife coming approximetly with 1m/s towards me and I'am using the DA 55-300 PLM @ 300:
AF-C Test Pentax KP DA 55-300 PLM Series 1 | Flickr

My wife coming approximetly with 3m/s towards me and I'am using the DA 55-300 PLM @ 190:
AF-C Test Pentax KP DA 55-300 PLM Series 2 | Flickr

--------------------------------------------------------------------------------------------------------------------------------
Last week I realized my mistake: the camera was left in FPS priority mode, now I set up my camera again.
The interesting, common settings are the following:
-AF-C
-First image action Focus priority
-Action in AF.C Cont. Focus priority
-AF Hold off
-Back button focus on (shutter button focus off)
-All images were taken with wide open aperture
-Images aren't cropped or sharpened, only my default nose reduction and CA correction applied.

I made the following new tests (see the detaild settings in the album descriptions):
DA*50-135
AF-C Test KP 50-135 Spot | Flickr
AF-C Test KP 50-135 Spot Fast | Flickr
AF-C Test KP 50-135 Sel | Flickr

DA55-300:
AF-C Test KP 55-300 PLM Spot | Flickr
AF-C Test KP 55-300 PLM Sel Camera Move | Flickr
This is what I was waiting for a long time:
AF-C Test KP 55-300 PLM 3D Tracking | Flickr

DA50:
AF-C Test KP DA50 Spot | Flickr
AF-C Test KP DA50 Sel | Flickr
And here is what is beyond the capabilities: my wife did some realy fast direction changes, she was realy unredictable.
AF-C Test KP DA50 Sel Erratic Movement | Flickr

The FPS drop by setting focus priority is minimal, but noticable for me.
My conclusion is now the Pentax AF system is up to date and it's competitive across different brands according to reviews I read. The 55-300PLM is realy good, much better in any way then the 55-300WR wich I have too. This is not the D5 or D500 from Nikon, I think it can be fooled by more subjects or more unpredictable movement, but now its a decent AF system. I think it will solve the DPreview bicycle test too, I am waiting for their opinions. I am satisfied with it.

---

Last edited by 08amczb; 03-11-2017 at 09:50 AM. Reason: Additional tests

[ogl]

Tell us your conclusion...

[08amczb]

The KP is my second DSLR, so I don't know what to expect. My conclusion is I'am a little disappointed, but I think it's usable for a wedding for example. I think at this speeds the lenses wasn't limiting the system, there are images where the lens caught up after missfocusing. Maybe I can do better also.

[bobbotron]

I can't seem to see the exif data for these photos, what shutter speed are these taken at?

[08amczb]

I've didn't remove the exif data and I can see it if I click on an image without signing in.
I used AV mode in bright sunlight, auto ISO set to fast, so It came out 1/400 or faster where I checked it. The blur is not motion blur. My wife walked slowly.

[C_Jones]

F/6.3 does not provide the depth of field a higher number F stop would. 800 shutter speed may not be required for walking, something between 125 and 400 would be more accomodating. 1600 ISO could be brought down by adjusting the F stop and shutter speed. The focus should be attempted by focusing on an area of the subject which has contrast (lack of glare/reflection). Those findings are per image

https://www.flickr.com/photos/08amczb/33117146651/in/album-
72157679132878431/

---------- Post added 03-04-17 at 12:25 PM ----------

The F/2.8 is providing the minimal depth of field in this image, more depth (higher number F stop) would provide more detail. 400 shutter speed is good, 320 probably could have covered a walking subject also. 200 ISO is a nice ISO. These findings are per
IMG170304_0008 | Benjámin Czétényi | Flickr

In both images I looked at, the lens was at longest focal length, 300 and 135 respectively. In some cases, when the lens is at its longest focal length, the detail will be improved by using an F stop (higher F stop number) which will provide more depth than when at a shorter focal length on that same lens.

---------- Post added 03-04-17 at 12:43 PM ----------

I know that this may not be relevant to your test, but a burst should not be necessary in most cases for a slowly walking subject, single shot mode should be enough. Also, if you shoot weddings a flash is going to accommodate the ability for you to use a higher F stop number, which would provide more depth/detail capability. Single shot mode for a wedding should be suitable for most scenarios/situations.

I want to note that the AF area setting (AF Point) set to "Center" (no expansion) has been effective for moving subjects for me during shooting.

[biz-engineer]

AFC tracking on slow moving subjects does not work well. Tracking require prediction (Simplified: derivative from on time to another time, allow to predict where the subject will be at the third point in time and so on). Contrary to expectations, a walking person is hard to track accurately, especially hard with slow aperture lens because the amount of defocus that the sensor sees between 100mS time intervals is sunk into noise, while the same AF system with f4 lens is able to track a flying duck at 60mph.

Try a Pentax camera with a dummy subject moving at different speeds , from slow to very fast:
- slow: AFC focuses on the first frame and then fail to focus on the 2 or 3 subsequent frames (because it failed to measure a significant gradient of positions between frame #1 and #2, misses focus on frame #3 then catchup on frame #4)
- medium speed subject: approach you hear from the lens barrel, you hear a zzzzzzzzzzzzziiiipp. Interframe gradient is measured and the subject position on the next frame is predicted correctly, the servo works linearly.
- very fast subject: camera subject prediction is good (software calculates an accurate subject position for the next frame), but the lens AF motor can't spin fast enough to keep the subject in focus.

The problem of the Pentax AF is design for engineers , not so much for photographer. Perception of Pentax AF performance is very bad because customers don't have the time or wish to put efforts into making Pentax AF works for them. It is easier (but more costly) to switch brands.

---

Last edited by biz-engineer; 03-04-2017 at 11:14 AM.

[08amczb]

Jones, these arn't artistic images. Wide open apertures were intented, but I didn't set up the camera for the ideal ISO and shutter speed, because I only tested the AF system.

AFC tracking on slow moving subjects does not work well. Tracking require prediction (Simplified: derivative from on time to another time, allow to predict where the subject will be at the third point in time and so on). Contrary to expectations, a walking person is hard to track accurately, especially hard with slow aperture lens because the amount of defocus that the sensor sees between 100mS time intervals is sunk into noise, while the same AF system with f4 lens is able to track a flying duck at 60mph. AF tracking is something that require understanding, with any brand of camera, Pentax AF being the least engineered.

Tracking a bride slowly coming into the church is a must, I think the engeneers at Ricoh also know this. My wife was walking at a constant speed and wasn't too far, nor too near, so the AF system should predict her movement.

Altough I did a mistake: I left the "Action in AF.C Cont." at FPS priority. This gives me some hope. Sorry, I discovered this now. The KP has much more AF settings, then the K-30 had...

[SeaRefractor]

Every lens correction was off

I'm most certain you checked this, but when I found that my focus was off on all of my lenses I corrected the diopter on the viewfinder. Again, I'm sure you did do that, but as others could be new to DSLRs in general, it bears reminding to check the diopter for their benefit too.

Thank you for your tests. Any chance you also checked AFC in terms of video to see if that's improved over the K70?

[08amczb]

The diopter doesn't affect the autofocus system, but yes that was the first thing I set up on my camera.

I'am waiting for the firmware upgrade wich enables sensor shift stabilization and I'am realy not a video guy. I might try it.

---------- Post added 03-04-17 at 08:05 PM ----------

Any chance you also checked AFC in terms of video to see if that's improved over the K70?

I think the KP doesn't has phase detectors on the sensor, so I don't expect any improvement over the K-70 regarding video AF-C.

[luftfluss]

One of the more charming AF-C tests I've seen...

[Tatouzou]

I think KP should AF at least as good as K3, which I have and my son have.

Our experience with K3 is that the AF-C tracking in X and Y axis is not good.
We both tried AF-C tracking modes (sel9, sel27) and gave up, the results were not satisfactory.

But we experienced that Z-axis AF-C was fast and accurate, even whith rather fast moving targets. Hence the method described below:

We obtain best results by using a single point AF. This point doesnt need to be the center point, any of the 25 cross focus point will work. But the 2 outmost AF points should be avoided for moving targets.

What we do is set the camera to AF on back button press only (=disable AF via shutter release), and AF-C single point.
Choose the single focus point according to the framing you wish.
Track the subject in AF-C by panning the camera while pressing the AF back button.
Check the AF lock (green hexagon) and continue panning to track the subject.
Press shutter release.
If you are in burst mode, use M or L burst, focus priority, shoot short bursts (2 to 5 shots).



This method can also be used as a substitute to AF-S: just release the back button once the AF is locked on target; no need to switch between AF-S and AF-C, we thus use it all the time.

This method produced good results with HD DA 55-300 (the screwdriven version) on african wildlife in Tanzania, and also even at full aperture with DA 18-135, DA 17-70 f4,with fast primes like FA31, FA77 (my son has both), DA70, DA35 f2.4, and also DA35 macro f2.8, FA50 f1.7, DA40-XS (I have all these lenses). as long as you concentrate on panning the target and dont change the zooming too often while tracking.

I shot the night Corso Carnavalesque in Nice last week with DA 35macro at f2.8 and AF nailed focus on most shots. My son did the same using FA31 and FA77 at f2/f2.5, and got even more keepers thanks to better technique.
I shot the dailight "Fête des citrons" in Menton (a city on French riviera close to Monaco) using DA 18-135 fully open all the time to get maximum subject separation and it also nailed focus all the time.

DA 18-135 is one of the fastest APS-C focusing lenses in Pentaxland.
On the other hand, the screwdriven DA version of HD DA 55-300 is not the fastest focusing lens, but web reviewers rate the new PLM version as the most efficient Pentax APS-C lens in AF-C.
Whereas DA50-135 f2.8 is known as being slow to focus (my son had it, he loved the IQ but he sold it because he considered the lens was too big and prefers shooting with primes) .

AF is more tricky when using DA21 or DA15, IMO because of field curvature and very short focus throw. We consider that only the central point focuses accurately whith these lenses wide open, thus we use the old fashioned focus using central point, hold the focus (=stop AF-C) by releasing the back button, then recompose and fire the shutter.

[08amczb]

Tatouzou, thanks for the recommended settings.
I hope the focus priority also improves the results, and I hope some x-y tracking capability from the KP. I know the 50-135 is a slow focusing lens, but I hope it can track someone marching towards me. I'am waiting for some DPreview like bicycle tests done by more experienced reviewers than me.

[bobbotron]

I've didn't remove the exif data and I can see it if I click on an image without signing in.
I used AV mode in bright sunlight, auto ISO set to fast, so It came out 1/400 or faster where I checked it. The blur is not motion blur. My wife walked slowly.

Funny, it was on my phone where I couldn't see it. It appears in browser!

[apete]

I hope the focus priority also improves the results, and I hope some x-y tracking capability from the KP. I know the 50-135 is a slow focusing lens, but I hope it can track someone marching towards me. I'am waiting for some DPreview like bicycle tests done by more experienced reviewers than me.

I have only K5ii which is much older generation but - from what I have observed - once focus priority is selected FPS is always reduced dramatically. This setting may be a game changer. Would you be able to repeat the test in focus priority mode?

Though still I think your effort was one of the best of all I have seen. Thanks for that!