I tried a mirrorless. Have you?

Roland Karlsson wrote:

MarBa wrote:

Nikon was able to do this! I see no reason why this would not be possible.

Of course, Nikon could do it. Pentax can also do it. May point was mainly that I cannot do it. Because the protocol provided do not contain enough information and manipulation of the camera.

NOTE though - that it is not easy for Nikon/Pentax. The limitations and problems I mentioned do exist. And I assume that not everything works well with the Nikon solution.

Another thought. Nikon has a MUCH better AF than Pentax. K3-III is the first Pentax camera that comes even close. This neeeds to be taken inteo account. Making it automatic is much easier if the AF is good and reliable.

I don't think it makes it easier, it makes it more reliable. If AF with a lens is not reliable, that won't change after fine tuning anyway ... You want a better "mean focus" if possible, even if distribution around it is too big.

It's even one of the reasons to automatize: if you take more samples, you'll get a better tuning in average. But taking 10 pics per fine tuning value, it's so long and boring ...

They managed to get astrotracer without GPS...

That was nice. I never liked the GPS based one. Mainly because it needed a compass, and that is not a reliable tool.

Calibration of lenses seem easier than that.

It is different problems.

There can be multiple correction numbers stored for zooms, different AF points or even apertures.

It could. Do Nikon have that?

That's another issue. They can automatize current AF fine tuning system, even if it does not handle zooms as we would like. It's independent in my opinion.

LV can focus really well with contrast focus. So it is NOT a problem to find out if things are sharp. And then, when things are in focus using LV, you just need to read out what the AF module is reading and use this information to calculate a proper correction.

Yes, LV is very good. Ordinary phase focus on my K-1 is not. The focus varies from try to try. Not much. But more than a step in the calibration.

Again, you calibrate the AF system. That does not improve it, it just fine tunes the AF for a specific lens by a constant margin. I wouldn't fine tune my Sigma 8-16, it never focuses correctly through OVF, whatever happens, so it would be useless. The "unreliability" is just too big. But I don't have this issue with other lenses.

But what you calibrate is what the AF system does when it focuses a lens, so the fact that it's good or bad is irrelevant.

As an easier method, they could just use the "good focus" information, that we see in viewfinder. That's the AF system saying "not focused" or "good focus" or "not very good focus" (blink). If they have access to a continuous value they can use that (and repeat multiple times to average). If they have only 3 states, they can still use that to deduce a value (and again, repeat several times ...). The advantage if automatized is that this method could be blazing fast, even if configured to take many samples for each value (just move out of focus / refocus / save focus indicator).

It's a pity they didn't publish this stuff in the API. If they don't implement the feature, they could at least publish what's needed to make it on our side.

I can't be worse than when you do this manually by trial and error and have just ONE number per lens! The current implemention is terrible if you think about it.

Fully agree. Manually correcting lenses (if you are picky) is a hell.

The worst is to spend hours on it to then realize your results are not conclusive for any reason