[Anvh]

The first shot with the tree just doesnt work, does the AF need to focus on the tree or on the wall there, the AF doesnt know...

The sign should be a nice target but manual focus there.

[Williunck]

If you want critically accurate focus, use manual focus with an aftermarket focus aid.

I see criticaly accurate focus in sports illustrated all the time, there is no way that those guys at the NFL game with 3 cameras around their neck and a 400mm on a monopod have time to manual focus.

My K5 is fairly reliable at focusing in AF-s mode, it does not compare to my 20d in AF- C mode, and the 20d is what, 6 years old? I have also noticed with my K5 that it sometimes indeed has trouble locking on to its target even in good light using AF select. I am careful to pay attention to this and sometimes indeed have to manual focus before i take the shot.
Some of you come and assume Zafar has no idea what he is doing, but that doesn't explain why he gets better results with the older camera. Anyway, like someone said above, hopefully your problem is fixed when you get the camera back Zafar, I would say as a k5 owner that 50% accuracy is not up to par, assuming you are using proper shutter speeds for handholding and all that. I suppose it could be an SR problem, as suggested above try shooting with SR off to rule that out.
Good Luck

[vonBaloney]

I see criticaly accurate focus in sports illustrated all the time, there is no way that those guys at the NFL game with 3 cameras around their neck and a 400mm on a monopod have time to manual focus.

There were no in-focus sports shots for all those decades before autofocus?

[Williunck]

There were no in-focus sports shots for all those decades before autofocus?

Lol did I say that? My point is some professionals depend on it to be accurate, and yes I am sure it fails them from time to time, but on the street in good light with time to compose a shot it should be spot on more often then not.

[Fontan]

I think there is one and only solution to unreliable AF - Leica!

[elliott]

Lol did I say that? My point is some professionals depend on it to be accurate, and yes I am sure it fails them from time to time, but on the street in good light with time to compose a shot it should be spot on more often then not.

They depend more on the really high FPS, hoping that maybe 1 out of 20 will be good. You don't see the thousands they throw away, you just see the handful they print.

[Zafar Iqbal]

The first shot with the tree just doesnt work, does the AF need to focus on the tree or on the wall there, the AF doesnt know...

I focused for the tree in all three photos. The AF system isn't *that* bad (when it works).

Pic 1: Focus @ tree. Micro Adjustment set to 0.
Pic 2: Focus is at nearest half of the road. I used -10 @ micro adjustment. A minus value shifts the focus back. If I by any chance had focused on the building, the shift would not plant the focus in front of it.
Pic 3: The focus is in front of the tree. Again fits well with a micro adjustment value @ +10, which shifts the focus towards the front.

The sign should be a nice target but manual focus there.

Not sure what you mean. The only purpose with those photos was to illustrate, that the recomposing method does not lead to subject being severely out of focus, as has been claimed with photos like this one:

Now the fun part - I attempted some controlled testing.

AF-S, Center Point, used a tripod. 2 sec timer - I barely took a breath in the meanwhile! I also always refocused from MFT or infinity between each shot.

Initial micro adjustment setting: -10 (because it worked earlier). Check image descriptions and basic info (hover the mouse over the displayed image and select "i" icon from the popup).

Natural light except for the last 3 photos, which are a mix.

2012.07.04c - Zafar Iqbal's Photos | SmugMug

There are a number of f8 shots taken as reference pics to show print quality better.

Pic 7: I centered the K and the focus gets thrown more off (compare with pic 5).

Pic 13 vs 14: some red vs. not so much/any red = not in focus vs. in focus.

Pic 20 and onwards were shot a couple of hours later.

Pic 21: Micro Adjustment set to -6. much better (perhaps the original -10 setting wasn't too good to begin with).

Pic 26: turned on ceiling lights. Focus actually got a bit closer to K. Compare with pic 24. There is no difference other than the light being turned on.

Pic 27: Woar! turned on florescent light under my kitchen drawer thingies. These improved the focusing severely.

I only realized *yesterday* that many reds throw the focus way off (relatively speaking), which is why I used this color a lot in these test shots. If I hadn't done so then I'm sure all shots, as they otherwise are, would to most of you look like my K-r doesn't suffer from anything extraordinary. But it does! None of the photos without red or sufficient amount of red in the center leads to front focusing. It however happens too often when I'm out shooting. It's not all related to AF getting confused to the degree some of you seem to think and I hope these recent photos will help you to see something is off.

Not all artificial light is good for the cam though. The light from my monitor improves focusing, but the light from my energy conserving bulb does not/throws it off.

---

Last edited by Zafar Iqbal; 10-29-2013 at 02:04 PM.

[bossa]

I probably missed this info as I didn't read very carefully (got the flu)

DId you try a different lens?

[twitch]

Zack, did you know the centre AF cross point is not a point at all, it's a cross, a large + shaped thing that takes up the entire spot metering area in the middle of the VF marked by the ( ) symbols. Did you know that it is looking for contrast along those cross lines. AF is not an "area" and it is not a "point", it does not bullseye focus on where the cross lines intersect, it does not look around the entire ( ) area for what to focus on either. It looks for contrast along those 2 lines that make up the cross only (vertical lines of contrast on the - part of the + and horizontal lines of contrast on the | part of the cross.

It is not predictable (to me anyway) as to which object AF will focus on if it detects multiple points of contrast along the - and the | so you need to try and ensure if there are multiple choices that they are all on the same plane of focus, ie, the same distance from the sensor. Draw (in your mind) a big plus symbol covering your () symbols of your VF and look for vertical lines crossing the - and horizontal lines crossing the | before you squeeze the shutter. Is it unambiguous for the camera? Can you make it unambiguous?

It seems to me you don't understand this as many of the shots you claim to be OOF are in fact perfectly in focus, but just not on the area you wanted. The street scenes with trees in the background are unbelievably contrasty with multiple choices for the AF cross point.

That's why people here are suggesting test charts. Good test chart provide a completely unambiguous point for the AF to lock on to and can thereby separate user error from true BF of FF.

My conclusion looking at your test shots is you may or may not have an AF problem, frankly there is no way to tell becuase you don't seem to understand how the AF in your camera works.

[Anvh]

What i mean is that the tree is too small to be a reliable target since you also have some contrast going on in the background with the wall.
Like has been said the focus point is LARGE so keep that in mind, it's about as large as the circle in the middle of the viewfinder.

[Anvh]

That's why people here are suggesting test charts. Good test chart provide a completely unambiguous point for the AF to lock on to and can thereby separate user error from true BF of FF.

My conclusion looking at your test shots is you may or may not have an AF problem, frankly there is no way to tell becuase you don't seem to understand how the AF in your camera works.

He has done some test chart shots with the moire pattern on the monitor an they were perfect he said so that leaves user error.

[Zafar Iqbal]

DId you try a different lens?

If you mean with the recent test shots, then no. I tried again after reading your reply, using all three lenses, but on a single target:

Camera settings as before.

"Red" shots were now sharper and the difference between in focus (reds) and out of focus (none reds) was very small.

The main thing missing was natural light, and as I have been claiming, this camera including my K-5 shifts focus as as the light level changes (and very likely with other things as well, such as color - red in daylight!).


twitch:Thank you for your input. I appreciate your time with the reply, but your explanation isn't convincing. Sure, if the distance gets too large then things get complicated - but I have plenty of close enough shots that IMO show the problem clearly. But still, many of the distant (and closer) shots do not show back focusing, but front - front to what? I don't see the AF caught something closer as you supposedly do. I see the AF not doing what it should.

What i mean is that the tree is too small to be a reliable target

How about these then :P



A bit wider - should make things more complicated, yes?:


And finally a pic where you can get a better idea about the setup. I placed the bottle on the side. Chocolate bar beneath and have a detailed poster in the background. The cable is thinner than the tree (relatively on the photo, of cause ). I didn't have problem locking on to it. It was kind of harder to take the pic @ 200mm 1/10th f11 .. handheld :P



(The focus isn't competely on the cable because of the lamp in my room affecting AF and some minor shake probably. But they serve the purpose well anyways)

He has done some test chart shots with the moire pattern on the monitor an they were perfect he said so that leaves user error.

True and I just did again using Pentax 70mm Limited, Tamron 90mm Macro and Sigma 70-200mm. All of cause same lenses that I've been using so far. None of them indicated FF/BF issues as described on the web page. The result photos have been uploaded to the album as well.

I also did some quick classic chart shots. 3 (one with each lens) close to my energy conserving lamp and 3 close to my monitor. Notice the difference :P These are uploaded as well.

Oh, btw.. Micro Adjustment @ 0 with both moire and chart tests.

What a joy this is.

[twitch]

twitch:Thank you for your input. I appreciate your time with the reply, but your explanation isn't convincing. Sure, if the distance gets too large then things get complicated - but I have plenty of close enough shots that IMO show the problem clearly. But still, many of the distant (and closer) shots do not show back focusing, but front - front to what? I don't see the AF caught something closer as you supposedly do. I see the AF not doing what it should.



How about these then :P

I'm not trying to "convince you", I'm explaining how phase detect cross type AF points work in practice, you seem to think they work like some kind rifle scope where the bullet hits the scene at the intersection of the big plus sign.

Draw a big plus sign in the centre of your latest pic and count the number of contrast lines that intersect the two lines of the big plus sign. Are those intersections at different distances to the sensor? If so then you are leaving it up to luck as to whether PDAF will lock onto the object you want it to. You want to rule out luck when it comes to trying to isolate whether you have a BF/FF problem.

[Zafar Iqbal]

you seem to think they work like some kind rifle scope where the bullet hits the scene at the intersection of the big plus sign.

Your assumption is wrong. I can't see how you got that idea either.

Edit: also, you might be very good but I can not keep the camera so rock solid, that the angle will not shift from the time I focused till I took the photo. I dont mean just these photos but in general. Do not expect the exact center to represent *exactly* what was focused on.

[audiobomber]

I've found the chart at the bottom of this page invaluable in checking for focus error:
http://focustestchart.com/focus21.pdf