AF microadjustment

Errors will still occur, since it is using a separate system to determine if the image is in focus or not. The prediction of amount of focus shift is only really to speed it up and to know which direction to move it. With an AF lens it still usually checks focus after moving it.

I suppose the key point is that the phase AF mechanism tells the camera that the 'split image' derived from two areas of the incoming beam is now aligned as one superimposed image once the lens is moved to a certain point. The error is that the actual image on the sensor isn't exactly focussed when the phase AF gets the best match. This can be due to the lens aperture being small so that the baseline for the analysis is very short, or slight curvature of field. The AF sensors are also sensitive to different wavelengths, leading to the situation where focus may be different under different lighting.

Not sure that answers the question, but I'd expect the manual focus with green hexagon to match the AF, it is still based on the same measurement.

I agree with this, and I've been checking AF accuracy with my old 18-55 AL kit lens, and my far older 200M/F4 manual focus lens, on my new K5. The autofocus for both (using the hexagon on the manual) seemed to be accurate on my 100DS but that could simply be because of its lower sensor resolution compared to K5.

On the K5, the 18-55 is hard to evaluate but is just about in focus or perhaps has a slight front focus bias, fine adjustment would be 0 to -5 after further testing, if any. The 200M manual lens needs -10 to be set on AF fine adjustment "apply all", to get the focus hexagon anywhere near the truth, else it 'front focuses' badly. Ideally, it would need at least -12 but that's out of the adjustment range so I'll have to live with that.

Can you (or anyone) confirm that if I leave -10 set on "apply all", and subsequently set -5 on the "apply one" for the 18-55AF lens, the K5 will use -10 for the manual lens(es) and -5 for the 18-55 autofocus lens? The manual explains that setting "apply all" overrides "apply one", but doesn't mention what happens if you then reapply "apply one". If I still had my 50-200 AF I could test it easier, but I broke it...

Most of the MF lenses I've tried on the K-5 require beyond -10
Bit of a pain

It would be useful to know if "apply all" gets overridden?

To be honest, I think it's expecting a bit too much for decades old lenses to give a reliable result with electronics that are designed and optimised for DA optics.

In exactly the same way, exposure cannot be guaranteed to be accurate either.

I think I will stick to using my eyes. Perhaps a 2z eyepiece magnifier would be an alternative for anyone who finds finding focus difficult.

However, don't forget the importance of setting the dioptre correctly on the eyepiece. Point the camera at a blank wall or sky and adjust till the lines etched on the focusing screen are crisp. Don't try to set it whilst looking at a subject - the eye tries to adapt and the end result is no good.

This is one of the most interesting and informative threads we have had in a while.

There's a 31 page thread on the other forum about AF problems
under tungsten light link and a White Paper by Lumo Labs on the problem. link

He states that Contrast AF is twice as good as phase detect AF
but that would be with the phase detect set correctly.