[jolepp]

I have suffered from AF FF with my K-5 and Tamron 17-50 1:2.8 in low light, which usually happens to be from tungsten bulbs or fluorescents with similar color temperature. I have tried some simple tests to maybe figure out something about the problem. I thought I might start a thread to share test ideas and results without cluttering the main discussion thread(s); I'd like to see others to follow suit and post their tests and results.

I plan to start with a series of posts where each post includes a pair of test images under the same conditions, one with LV contrast AF and one with VF phase AF; the first post introduces the simple test arrangement used in those to follow.

I'd hope that sharing tests/results helps to pin down the problematic conditions and thus gives a better handle on the problem; we'll see about that, though

---

Last edited by jolepp; 01-18-2011 at 11:08 AM. Reason: typo

[jolepp]

A stack of three AA batteries is used as a focus target. These have been spaced so that the center AF (selected by AF switch) should have no problem locking on the center one. This arrangement is very convenient in that given a spacing in depth to suit DOF it allows one to immediately see if a focus problem exists, and if so, whether it is FF or BF. Attachment #1 shows the arrangement of the batteries (the paper under the batteries has a 7mm grid). Attachment #2 demonstrates the general test arrangement.

The tests have been carried out with the K-5 and tamron 17-50mm 1:2.8 @f=2.8/50mm with 0.50 m distance from the image sensor plane to the battery in the middle. The middle battery is intented to lie on the optical axis of the camera/lens. The "width" direction of the graduated paper grid in attachment #1 is aligned to the sensor plane, while the "depth" direction is at a 90 deg. angle to that. Other lights were switched off while it was dark outside to bring out the effect of the specific light source used. Other settings: ISO100, matrix metring, AWB, AFS, AF fine adjustment 0. [edit: firmware 1.02]

I forced the camera to refocus for every shot simply by moving my hand in front of the objective and letting AF to try to focus on that.The 2s timer was used to minimize shake/blur resulting from handling the camera.

light sources: tungsten = 40W tungsten spot, led = 'white' led bicycle light

The images in the following posts are intented to be identical center crops resized for posting. Each pair of shots is from the from the same tripod position and in the same light as closely as I could make it (by simply not disturbing anything between shots). Shutter times are based on camera metering (M-mode, exposure meter at middle/0); some of these would have needed a longer shutter time to be optimally exposed. The reported EVs have simply been calculated from the parameters used for each shot.

[jolepp]

Light source: tungsten @8.6EV (1/50s f=2.8), direct light.

#1 : contrast AF : camera WB 3730K
#2 : phase AF : camera WB 3739K

In comparision #2 is slightly front focused.

[jolepp]

Light source: led @8.6EV (1/50s, f=2.8), direct light

#1 : contrast AF, camera WB 6948K
#2 : phase AF, camera WB 6970K

In comparision #2 is slightly back focused.

---

Last edited by jolepp; 03-03-2017 at 02:41 AM.

[jolepp]

Light source: tungsten @EV 5.2 (1/5s, f=2.8), reflected from an A4 sheet and room surfaces.

#1 : contrast AF : camera WB 3210K
#2 : phase AF : camera WB 3218K

In comparision #2 is front focused.

[jolepp]

light source: led @EV 5.2 (1/5s, f=2.8), diffused with paper (bicycle light wrapped in a A4 sheet, sitting on the edge of the book on which batteries are placed)

#1 : contrast AF : camera WB 7302 K
#2 : phase AF : camera WB 7280 K

In comparison #2 is slightly front focused.

[jolepp]

In an attempt to demonstrate the difference light color temperature makes as such I did one a shot with contrast AF and tungsten light (#1), moved the body switch to MF without disturbing the focus set on the lens (I hope) and then took a shot (#2) with flash (YN-560 bounced from the white ceiling) as the only source of light.

#1 : 'well-lit' tungsten : camera WB : 4304K
#2 : flash : camera WB : 6285K

It is hard to spot any difference in focus here.

Here the flash was adjusted to a power level that gave a nice enough exposure, nominal parameters were the same (1/80s, f=2.8, ISO 100).

---

Last edited by jolepp; 03-03-2017 at 02:41 AM.

[Tommot1965]

did FF get any worse at lower EV than 5.2?, or didn't you go any lower ?

[jolepp]

did FF get any worse at lower EV than 5.2?, or didn't you go any lower ?

I didn't go lower with this batch. The EV values should be taken with a largish grain of salt anyway: they have been calculated from the parameters that result from setting the camera meter (matrix mode) to 0 with my improvised lights. Since this resulted in some apparent underexposure these are good as upper limits only.

[Tommot1965]

I didn't go lower with this batch. The EV values should be taken with a largish grain of salt anyway: they have been calculated from the parameters that result from setting the camera meter (matrix mode) to 0 with my improvised lights. Since this resulted in some apparent underexposure these are good as upper limits only.

no worries jolepp, thanks for doing the test...

[betaPhoto]

First of all, thanks for doing such a thorough test!

It seems as though EV1, EV2, EV3 & EV4 is where the problem is. Below EV1 the focus assist light comes on (and focus is good) and above EV4 the focus is not off by much. It might be good to concentrate on EV2 (red in the chart below) and EV3 (blue in the chart below).

Pondering the problem, my first theory is the focus sensors stop working on some color wavelengths between EV1 & EV6 and eventually all color wavelengths.

As the light dims and wavelengths drop off the focus sensor, it plays havoc with focus accuracy. Under certain low white balance situations (maybe below 5000k or so) the problem is worse because those are the wavelengths dropping off first. If there is some bright white in the frame it can also make matters worse because the focus assist light won't come on as early.

My second theory is, not all focus sensors drop off color wavelengths at exactly the same EV value. The hardware has tolerances and some are better than others. Therefore some K-5's may perform differently than others at these crucial cut-off points. Furthermore, then getting the firmware to adjust for the fall-off and focus change is difficult, because it's never exactly the same.

-----------------

I would like to see multiple tests with different K-5's done in the EV2 & EV3 area (as well as EV 8 or so for accuracy comparison). Also, the focus distance is important. I think between 3 and 5 feet (1-1.5 meters) is where the problem is more pronounced.

It would be interesting to see if some K-5's actually handle these low light settings accurately!

[Smeggypants]

First of all, thanks for doing such a thorough test!

It seems as though EV1, EV2, EV3 & EV4 is where the problem is. Below EV1 the focus assist light comes on (and focus is good) and above EV4 the focus is not off by much.

This is pretty much my findings.

EV4 is OK here. At EV2 the AF is unusable. My K20D is fine and focusses accurately at EV2 and below

[Michaelina2]

So... how is this a Pentax issue and NOT a non-branded lens issue?

All you seem to be saying is... if a non-branded lens (in this case, Tamron) is paired to a K-5 expect results like yours.

Is it possible your results simply reflect the 'no longer being good enough' generic design of this particular non-brand lens?

Just curious... M

[peterh337]

Does a K5 focus differently according to which lens is attached to it?

I would have thought the AF was a closed-loop system where the camera is using the sensors to detect the focus and drives the focus actuator to achieve it. In live view, the camera will be using a max-contrast AF method and that does the same thing.

If a lens achieves good focus at say F8 but not at say F4 the cause of that must be something fairly subtle.

It would be pretty interesting to do any tests in both normal and live-view modes because they use completely different methods for detecting the optimum focus point.

[betaPhoto]

So... how is this a Pentax issue and NOT a non-branded lens issue?

All you seem to be saying is... if a non-branded lens (in this case, Tamron) is paired to a K-5 expect results like yours.

Is it possible your results simply reflect the 'no longer being good enough' generic design of this particular non-brand lens?

Just curious... M

Post #14 from "Relevant news re malfunctioning AF under artificial light" shows a test with the FA50-1.4 with the same FF problems.

https://www.pentaxforums.com/forums/pentax-k-5-forum/128562-relevant-news-re-...ial-light.html

There are other tests showing the same thing as well, just because this is a Tamron doesn't in any way mean other Pentax lenses don't FF.

Some have postulated Pentax SDM and other HSM lenses don't FF under lowlight, but I haven't seen any tests showing FF with a non-SDM lens and then no FF with a SDM lens (both under the same EV2-3 lowlight conditions of course)...