[UnknownVT]

I tested with a flashlight).
At the point where the AF would actually work:
ISO 6400, 1/20s, F1.4 on a Sigma 50 F1.4 HSM
I moved the flashlight closer and closer until it reliably focused multiple times. At this point, my metering system was pretty much giving up. However, I had to know that my k-x likes to go from complete OOF to almost AF, then one more hit of half-shutter for it to stutter into place with the Sigma 50 F1.4.

That seem such a good idea, I thought I'd give a try in a totally dark room -
but I found most of my flashlights were too bright for the indoor distances,
and I was getting no where near not being able to focus.

So I found a flashlight that was rated at 0.2 lumens lowest level output (this is very dim) and placed it at varying distances until I could not focus, and moved it closer - I took the shots when it seemed the K-x hexagonal would stutter, but if I held still long enough it would lock focus - so I think I am pretty much on the limit. (FWIW - flashlight was about 44", 1.12m from the target)

Now this was not without difficulties - even though I used a target with high contrast defined pattern - it was basically so dark in my viewfinder that I had a very hard time being able to place my central focusing bracket over the central + of the target.

Just so that you can laugh this literally was one of my shots:

EXIF attached - it's off center because I could not even use my usual trick of imagining where the diagonal would cross - because the corners of my viewfinder frame were pitch black so I could not make them out!

Any difficulties aside I did manage to get a few shots that were acceptable -

EXIF attached - ISO5000, f/3.5, 0.3 secs; 18mm (viewfinder exp info flashing, meaning below metering limit).

Metering segments -


May not seem as low as your readings -
but don't forget I am using a f/3.5 lens that gives up - 2 2/3 stops compared to a f/1.4 lens at the sensor.

---

Last edited by UnknownVT; 09-24-2012 at 10:42 PM.

[twitch]

The last straw for me on the K-5 came shooting my kids playing team sports this year. The centre AF cross point is simply too big to put it over my kid and only my kid (I'm being realistic here, I'm talking about when I have clear line of sight of their whole body, no obstructions between lens and subject). As a result I got many many sharp photos of trees/fenses/other players in the background. That's before I even mention the (total lack of) tracking in AF-C mode or the focus speed of my DA*50-135.

[UnknownVT]

dSLRs are supposed to be precision instruments -
and we are right to expect them to behave that way.

However it does mean we have to find out the limits and not make assumptions -
because it does behave in a precise repeatable way -
so if we are getting repetitious failures -
then we are doing something that is repetitious.

The Spot focus area - although is not a pin-point - it is not that big either.

There are ways to determine the size of the area in our own individual cameras -
and it is relatively simple to do.

That way we will know for certain how big that area is
and can use the precision instrument with more precision.

I did this less than an hour ago and it took a few trial and errors -
but the actual experiment took only a few minutes.

I outlined the spot focus area in red -

the size of the area was determined by the following photos:
Top edge

yes, laugh - since the target was mounted on something, the top edge could not be determined with the camera the right way up in the normal position - so I turned it upside down that's why the target was slightly off center......

Bottom edge:


Left edge:


Right edge:


This is for my copy of the K-x - of course your camera may be different - but I can't see the spot focus area being that different.

Now here's the catch - hence some of my trial and error - at first I thought it'd be a good idea to use something small and precise - like a pin head -
but it didn't work - the "target" was too small - the camera kept focusing on the background of leaves on trees.

So the spot focusing is not infallible - if there is a lot of other detail in the area - and the main object is very small the camera will focus on the other more prominent detail in the focus area. My Left edge picture is a good example - the camera chose to focus on the further wall despite the fact there was a closer target - but looking at the picture one can now see why, the further wall just has so much more detail - hence the "choice". Less obvious but even more telling is the Bottom edge picture - the main target is much more out of focus than either the Top edge or Right edge photos why? - well, I think it's because the camera has chosen to focus on that further wall again despite this time there is a lot more of the other stonework in the spot focus area - but looking at the photo more carefully it kind of makes some sense since the other stonework has very little detail to focus on - hence the choice that even I didn't expect - but the camera is doing its precision thing repeatably - the variable is my expectation and assumption.

I have used my K-x for nearly 3 years and over 59,000 shots and I do not often have out of focus shots - even with faster moving targets, and very, very low light. I think it helps understanding the limitations of our precision tools.

[sandilands]

That seem such a good idea, I thought I'd give a try in a totally dark room -
but I found most of my flashlights were too bright for the indoor distances,
and I was getting no where near not being able to focus.

This has been both the oddest and most informative thread I've read in a while. It has also revealed to me how little I understand about the way AF works in any of my cameras. More to the point this thread is helping me to learn more about how to use the cameras I have instead of bitching that I don't have the camera I think I need.

@JinDesu and @UnknownVT, please elaborate more about how you conducted your tests. First, how do you obtain the interesting diagrams showing the light values recorded by each af sensor?
By trial and error I've been using small LED flashlights for some time as AF assists (giving one to the model to shine on her own face so I can focus), but how have you used them in these experiments?
I have also wrestled with the fact that in extreme low light my eyes can detect a high contrast spot to focus, but they're not good enough to actually manually focus successfully. What have you done to both select the AF point and let the camera achieve focus lock at these light levels?

This shot is typical of some of the work I am doing. The "image" on the model's back is made with essentially a semi transparent GOBO mounted on a fresnel lens and focused with a 2x magnifying glass (think overhead projector but with a speedlight). It alone is a nightmare to achieve focus. To achieve decent contrast most shots like this are made in next to complete darkness. I try to achieve DOF from the spot on her back lit by the projection to the tip of her nose. Not too successful as you can see. My fastest lens (55* f/1.4) not surprisingly is the worst at achieving AF in situations like this - so I am usually more successful with the 18-135.

I think conducting the experiments you have done will be useful as I attempt to perfect this technique.



BTW the image on her back is Aries the Ram, not a buffalo leaving a pile of Ka-Ka.

[JinDesu]

That seem such a good idea, I thought I'd give a try in a totally dark room -
but I found most of my flashlights were too bright for the indoor distances,
and I was getting no where near not being able to focus..

I actually unscrewed my flashlight's head to the point where it was as dim as possible (due to the diffusion). It was still stronger than I liked, and it had some hotspots, but I was making do at the time. I was also shooting a darker and less contrasty subject (that orange barrel), which was near impossible for me to make out if it was in focus or not in the viewfinder. I'd definitely love to redo such a test with a properly diffused light source that I can shift distances (without me casting a shadow).

@JinDesu and @UnknownVT, please elaborate more about how you conducted your tests. First, how do you obtain the interesting diagrams showing the light values recorded by each af sensor?
By trial and error I've been using small LED flashlights for some time as AF assists (giving one to the model to shine on her own face so I can focus), but how have you used them in these experiments?
I have also wrestled with the fact that in extreme low light my eyes can detect a high contrast spot to focus, but they're not good enough to actually manually focus successfully. What have you done to both select the AF point and let the camera achieve focus lock at these light levels?

The program I used was PhotoME - it's an exif viewer. Inside, it'll list the light levels as recorded on the sensor, and by clicking it, you can see that graphic.

In doing the experiment, I was sitting in my backyard at night. It was too dark for even my adjusted eyes to see. I took my flashlight, screwed the head loose until it was as diffuse as could be, then set it a distance away from the barrel. I then tried to lock focus (if I get 3 in a row, I'm good). If it didn't lock, I'd move the flashlight closer and closer - increasing the light on the subject.

The flaws in my experiment are: A not very contrasty subject used, the flashlight not being diffused enough - causing hotspots that might have influenced the AF (although I don't think it did), and being a bit rushed (there were mosquitos). I could retry this experiment in a dark room this weekend. I'm sure the result will be similar (and seeing UnknownVT's subsequent post, I'm very sure it'll be similar).

I have not used LEDs to lock focus. I've used a big flashlight once, while I was camping, but I typically don't shoot in such dim conditions. Low-light dinners are about the darkest I've shot in. I'll see if I can get some values off my dimmer pictures, but I suspect they are all about 0.5-1LV in the center.

If you are trying to center your DOF so it covers the projection and her nose, I really have no tips (besides using a flashlight, or manual focus). It seems the most reasonable spot to aim for is the ear - since that is about equidistant from the back plane and the nose plane. Since it's practically black in the shot, I don't know how it actually looked in real life. I can say though, the ear isn't too contrasty, but the hairline above the ear is.

[UnknownVT]

@JinDesu and @UnknownVT, please elaborate more about how you conducted your tests. First, how do you obtain the interesting diagrams showing the light values recorded by each af sensor?

For my low light experiment - simply set up anything you wish as a "target" (my case this time was a Paterson lens test chart/card) in a totally dark room.

Set a flashlight on a tripod or any other conveniently moveable mount - shine light on target and see if you can focus - if so move light further away (so the illumination is dimmer) until the camera will not focus - then inch the light closer until the camera will just focus.

All I was doing was to vary the amount of light reaching the target until I found the critical point where the camera will just focus to determine how low a light my K-x will still AF.

My flashlight was "rated" at 0.2 lumens (very dim for a light) and even then it had to be 44 inches or 112cm away from the target for the camera to only just be able to focus - so the K-x is actually capable of focusing in very, very low light - don't forget most cameras are rated using a f/1.4 lens - I am using a f/3.5 lens which means there is -2 2/3 stops less light reaching the focus sensor.

Exposure segments are part of the manufacturers notes in the EXIF data -
I am sure there may be other ways of showing it - but the only way I know of is to use PhotoME (PC only)
do a find on "seg" and the <Graphis> "link" is shown:


I see JinDesu has already replied - I much prefer his using a flashlight diffused, by unscrewing the reflector head - should have thought of that.......
another easy way to diffuse a flashlight and make it dimmer is to simply put layers of tissue paper over the head/lens - like one diffuses flashes!

[JinDesu]

One other thing to note is that - my metering system was giving up at -1LV, so it's definitely not a normal shooting environment. With the k-5ii's AF system being rated to -3EV, I can only imagine that it helps the accuracy and speed of AF when shooting at super dim levels like what we've experimented with.

When shooting my shot, it took maybe 2-3 seconds to lock focus. Normally my Sigma 50 is very quick, even in dim dinner situations. Under that low light condition, I had to go completely OOF, hit the shutter halfway to focus, let it stop unfocused, then press the shutter halfway again. Once it did that, the HSM slowly stuttered it's way into focus. Without being aware of my lens' and camera's AF quirks, I'd have given up at the first try and said it's impossible.

In addition, my Sigma 50 tends to be troublesome when I have the focus near where I want it to be, but not exactly. It makes the camera think it's actually in focus. So my solution is to set the lens at minimum focus before every shot, and it locks on extremely accurately. I attribute this to the 1.4 aperture and my k-x's 5.6 sensors. I assume the F2.8 sensors in the k-5ii will allow the camera to actually determine if it's near or in-focus.

And again - without being patient and figuring out these quirks, I'd have dumped the k-x and complained that Pentax sucks at AF.

[Ubuntu_user]

I have found in my experience with Pentax and other camera brands (including CaNikon) that my K-5's AF speed is not really any slower than any of the competition. If you put a comparable lens (say the same sigma lens) on a K-5 and and D700 e.g., the focusing speed is the same. However, that said, why we have more trouble with Pentax is because of it's larger AF points.

[UnknownVT]

Now this was not without difficulties - even though I used a target with high contrast defined pattern - it was basically so dark in my viewfinder that I had a very hard time being able to place my central focusing bracket over the central + of the target.
Just so that you can laugh this literally was one of my shots:
... - it's off center because I could not even use my usual trick of imagining where the diagonal would cross - because the corners of my viewfinder frame were pitch black so I could not make them out!

I actually unscrewed my flashlight's head to the point where it was as dim as possible (due to the diffusion). It was still stronger than I liked, and it had some hotspots, but I was making do at the time. I was also shooting a darker and less contrasty subject (that orange barrel), which was near impossible for me to make out if it was in focus or not in the viewfinder. I'd definitely love to redo such a test with a properly diffused light source that I can shift distances (without me casting a shadow).

After reading JinDesu's great idea of using a diffused flashlight - I thought I'd at least redeem my embarrassment a little by re-doing my experiment -
Totally dark room, target mounted on wall, camera on tripod, LED flashlight with good neutral tone with several layers of tissue to diffuse, handheld to vary distance (light over left shoulder)

Found distance that camera could not focus, then moved light closer until K-x would just focus:

EXIF attached:

ISO5000, f/3.5, 0.3secs (luckily the same exposure as before - so this is good collaboration)

exposure metering segments:

again very similar to previous segment read outs.

Segment reading says K-x is capable of focusing as low as -0.9LV at the sensor.
My exposure of ISO5000, f/3.5, 0.3secs
for ISO100 (standard ISO) 5 2/3 stops difference from ISO5000
=
ISO100, f/3.5, 15.25secs

For f/1.4, 2 2/3 stops difference from f/3.5
=
ISO100, f/1.4, 2.42secs

this is about EV -0.3EV @ ISO100 = -0.3LV = 0.21 foot-candles = 2.2Lux
- the reason why it is short of -1LV spec is simply because I am using a f/3.5 lens
which is -2 2/3 stops dimmer than a f/1.4 lens (used for rating) -
so for my case - IF I had a f/1.4 lens in theory I should be able to focus as low as -3LV!!!

[JinDesu]

Not so sure about focusing down to -3LV, because my k-x with the Sigma 1.4 was definitely struggling at -1LV. Not sure how to approach the subject since technically, the AF sensors are seeing less light from your lens than mine.

[UnknownVT]

Thats not necessarily true.
You can choose center point AF, put the center point deadnuts in the middle of your target and still not get your subject in focus due to the center AF point being so gigantic. Yes, there are work arounds like MF, but why spend the extra $$$ for AF lenses and camera if they aren't terribly reliable or precise?

However, that said, why we have more trouble with Pentax is because of it's larger AF points.

This is more than once that Pentax focus sensors are said to be overly large - perhaps there is something in this?

However my own measurements - on a K-x gives this:

details in Post #33
which doesn't seem that overly-large to me -
can people with large center focusing sensors please show us how large they are please?

[cali92rs]

which doesn't seem that overly-large to me -
can people with large center focusing sensors please show us how large they are please?

I have no idea how the k-x focuses, but for a k-5, see attached pic. Anything within the ( ) that has some contrast will cause the camera to say focus has been achieved. So what ends up happening is you place the center focusing point, which is small, right over someone's eye, but the camera instead focuses on some tree or fence in the background because it was within the ( ).

[UnknownVT]

I have no idea how the k-x focuses, but for a k-5, see attached pic. Anything within the ( ) that has some contrast will cause the camera to say focus has been achieved. So what ends up happening is you place the center focusing point, which is small, right over someone's eye, but the camera instead focuses on some tree or fence in the background because it was within the ( ).

The ( ) area seems reasonably small to me -
so if one places that over an eye - how does a fence get in, I would have thought the rest of the face would be in the ( ) area?

If however the eye area is so small that there can be a fence in the background within the ( ) area -
then it is a very small "target" -
and I absolutely understand why - as I said in Post #33

Now here's the catch - hence some of my trial and error - at first I thought it'd be a good idea to use something small and precise - like a pin head -
but it didn't work - the "target" was too small - the camera kept focusing on the background of leaves on trees.

So the spot focusing is not infallible - if there is a lot of other detail in the area - and the main object is very small the camera will focus on the other more prominent detail in the focus area. My Left edge picture is a good example - the camera chose to focus on the further wall despite the fact there was a closer target - but looking at the picture one can now see why, the further wall just has so much more detail - hence the "choice". Less obvious but even more telling is the Bottom edge picture - the main target is much more out of focus than either the Top edge or Right edge photos why? - well, I think it's because the camera has chosen to focus on that further wall again despite this time there is a lot more of the other stonework in the spot focus area - but looking at the photo more carefully it kind of makes some sense since the other stonework has very little detail to focus on - hence the choice that even I didn't expect - but the camera is doing its precision thing repeatably - the variable is my expectation and assumption.

[cali92rs]

The ( ) area seems reasonably small to me -
so if one places that over an eye - how does a fence get in, I would have thought the rest of the face would be in the ( ) area?

If however the eye area is so small that there can be a fence in the background within the ( ) area -
then it is a very small "target" -
and I absolutely understand why - as I said in Post #33

If you are close enough to your subject, it is small...otherwise, it is not.

I guess this is just like the SDM phenomenon...if people aren't personally effected, it is acceptable, or they believe it doesnt exist. However, there are people that have sold their cameras because of the lack of precision in the K-5 AF.

[KungPOW]

I used to think my miss focus shots were my fault with my K10D.

Then I used a D200. Far less missed focus shots.

Then I used a D300. The D300 locks onto the subject like a cruise missile onto Baghdad.