Completely Dejected (auto focus K5)

All those things are much brighter than 0.5EV John, especially the concert. The spotlights used will allow quite short exposures at surprising low ISO values.

For the concerts I've shot I would be using ISO400 with shutter speeds around 1/125sec with no problem, albeit it with lenses of f2.8 or faster.

If that wiki article is right, EV 0 is an ambient light level for which you'd want an exposure of 30 seconds at F5.6, at ISO 100.

That sounds pretty dark. But with the high ISO performance of the K-5, that's only 2s at ISO 1600, and only 1s at F4.

All of a sudden, it doesn't sound that unreasonable to be wanting to take pictures. Factor in almost anything that steadies the camera a bit, and you'd be a bit disappointed if the AF locked quite happily, but was still significantly OOF.

The fact is that as the kit gets more advanced, people want to do more ambitious things with it, so are quite justifiably disappointed when they find they can't do what the spec suggests they ought to be able to.

I would like to know whether (as Falk hints at in the paper) the issue is Pentax-only, though.

I'd speculate that, just as they've fixed the KxxD stop-down metering problem, and the K-7 shutter blur problem, they'll find a way of fixing this in the next dSLR, but (despite the suggestion that it's a deterministic error that could be fixed in firmware) won't retrofit the fix to the K-5.

Oh Dear, Oh Dear!
I read the whole article - the term "pseudo-science" come to mind.
I am not going to say that there is not a problem with the K5 - but I do not have one, and I don't do that sort of photography where the problem would get in the way. What I will say is that the article is flawed in some respects. One of these leapt out at me straight away, and the others sort of niggled so that I had to refer to other works.

My concerns start with section 2.2.
Terms like "brightness" and luminance" are used incorrectly.
The "common conversion formula" for brightness is actually the one normally used for the LUMA value of a colour image signal.
Brightness is a non-scientific term.
To determine how much light from each of the three lamps actually fell on the test target - known as Illuminance - one measures in LUX, using a device like an incident light meter (and that is what it really is).
To determine how much light is reflected from the test target (the Luminance of that target) one would normally use a very sophisticated colour spotmeter (look up Minolta CS100 and shudder at the price) which can measure the LUMINANCE (in Candelas per square metre) of a light source (whether a reflector or an emitter).

Now look at it like this - if the experimenter cannot use the correct terms, and either measure them correctly or use an acceptable alternative, then what other mistakes are there?

In terms of Autofocus, there is one really big one:
All our current autofocus systems rely on contrast.
The experiments - as conducted - do not seem to have taken contrast into account. so if the reflected light from the targets cannot be accurately measured, how does one know what the contrast was between the black and coloured areas of the target - and if this cannot be properly controlled, how can any comparisons be made?
Further, the target was lit from an angle - close up - and there would have been a fall-off in the illuminance value across the target. This could very easily affect the AF sensors.
If this were an undergraduate project, I would probably give it a good 3rd, but no more.
Geoff

Just Google 'AF in low light', and pretty soon you'll see every camera marque mentioned adversely.

As has been said, do the AF tests in daylight.

As to taking shots in low light, you'll have to spend many times more than the cost of a K5 to getting anything like half acceptable results.

What makes you think Canon is any better?

Why are people assuming that I am attempting to take shots in virtual darkeness? These tests took place in my smallish kitchen with 300Watts of tungsten lighting.

Even the girlfriend's Lumix pees all over my K5 when it comes to focusing.

I remember at college doing still life in the studio and having to suffer the embarrassment of my Pentax hunting back and forth before settling for an out of focus shot while all the Canon fanboys laughed while their cameras snapped to sharp focus in about a millisecond.

problem solved then,
ask your girlfriend if you can use her "Lumix "

And I've shot weddings in candle light, with cannon owners ridiculing my equipment all day long only to see them STFU and put away thier over priced toys when I projected my images at the reception.

BTW... I usually use manual evrything.... maybe get your eyes check, check the diopter setting on the camera and use manual or live view in low light. or as stated above, find a camera better suited for you.

Sounds like you have a faulty specimen then. My K-5 focuses fine in less than a second in such conditions (and with only 120 watts of tungsten illumination, in a dark corner of the room as just tested, EV is 5). Newspaper print is pin sharp handheld at 100%, slightly sharper than using Live View and much sharper than I can get using manual focus.

100mm F2.8 DFA WR, handheld at 3200 ISO, 1/100 wide open, AF fine adjust set at -10.

Haha no way could you shoot an Asian wedding using manual focus, it would be physically impossible.

Dave I have considered it may be faulty but there are hundreds of references to users with the same problem. So it must be a common fault.

If I shoot a portrait I usually want the iris to be sharp. At the moment it's the end of the nose, no matter how good your eyes are you will not tell the differene until you have taken the pic.

very interesting read that Thanks

Yes, good post.

This gave me a good opportunity to pin down the exact AF fine focus setting for my new 100mm lens, it ended up as -8 and then, under normal domestic tungsten lights, using a tape measure on the floor and a very small adjacent target, it focuses perfectly every time. The only lens I can't quite get enough focus adjust on is an old manual M/200 (for CIF use). Seems to focus pretty consistently too indoors and out, so I am puzzled by your issues. I don't have any AF lenses wider than F2.8 though.

I always use centre spot focusing, which might be relevant.