[grahame]

My 2nd K-1 arrived a few days ago and I wasn't very happy with the camera's AF performance. In the past few days, I was busying testing and adjusting AF accuracy for most of my lenses -- they did accurately on my first K-1.

But I noticed another issue on this K-1, which I have seem before on K5, but didn't see on K5II and K5IIs: basically, camera back focuses quite a lot at portrait position and/or front focuses when camera is at landscape position.

I set camera single center point AF-S, aperture wide-open, shake reduction off, ISO 3200 to bring up shutter speed, mount camera on tripod or hand-held. I focused to the cover of the book standing vertical at the same hight at camera, which aligned to the long line between 16 and 17. Take a picture and turn the camera to vertical position, refocus and take another one. Repeat 5-7 times for each lens I have tested.

I tried FA43/1.9, DA55/1.4, and Sigma 85/1.4. All combinations have same issue: auto-focusing points in portrait position were significantly further away from camera than same setting shooting at landscaping position. To exclude the possibility of focusing target texture pattern affects AF accuracy, I also rotated the book and got same result. And the book cover gives enough sharp edge high contrast image to focus on

Anyone noticed similar problem before, on K1 and on other cameras?

here are two new photos I just took.

---

Last edited by grahame; 06-28-2016 at 06:40 PM.

[Adam]

The shooting distance can make a difference. However, in this case it's possible that the test target is simply giving you inconsistent results, since it's quite small.

Horizontal vs vertical should matter since the central 5x5 point grid consists of cross-type points.

Have you observed any AF inaccuracies in daily shooting?

[grahame]

Adam: The test target (book cover) is pretty big for the lenses I have tested. Camera is about 3-4 feet way, close to min focus distance for each lens, and camera did not move during testing for each lens. with 55 or 85mm lens, the texture on the book cover is pretty big. The book covered almost half of the frame for 85mm lens.
I used only the center AF point, which I know should be the best and is cross type, and should be sensitive to both vertical and horizon lines similarly. That is why I am quite surprised to see this result. And after I rotate the book and focus to the same position, so the vertical liens camera used before became horizon lines, I got same results.

And it is very repeatable. The pattern of BF at portrait and FB at landscape position repeats every time of repeated testing.

The camera can not focus very reliably in real photographing, so I had to adjust almost all my lenses. Most lenses I have didn't need adjust on the 1st K-1. It was during testing and adjusting, I noticed this issue.

[grahame]

I wonder if it is sort of normal, and I didn't see it on other cameras only because I didn't test in this way.
I have requested return but want to ask around before send it back.

[rawr]

Testing at close to minimum focus distance for each lens is probably not good procedure. 5 - 10 times focal length in mm is usually what is recommended - or simply the distance that you shoot the most at in real world usage.

Before you return your 2nd K-1 for a 3rd, you should probably look into AF testing procedures a little more.

[grahame]

Testing at close to minimum focus distance for each lens is probably not good procedure. 5 - 10 times focal length in mm is usually what is recommended

True for AF adjusting. But there should be no reason a camera can not focus accurately at close to its minimum distance -and I have left enough space to their limits. So to show the issue I was talking above, shooting at close distance is not a bad idea.
And my first K-1 did great, no AF issue as I could find, under same testing condition. unfortunately I broke it and had to buy another one.

[rawr]

What about if you change your test target image to one that features the same pattern when displayed vertically and horizontally? Like a spreadsheet grid, or graph paper.

[grahame]

I downloaded such one this afternoon and just printed it and tested a few mins ago, still same result. I posted these two photos up in the original post.

test target image to one that features the same pattern when displayed vertically and horizontally

And because my k5 also has such problem (I installed a split screen and used it for MF lenses eventually), I wonder if it is an quite common issue that we just didn't notice before.