EOS R6 focus guide inaccuracy

Started 11 months ago | Questions thread
ProfHankD Veteran Member • Posts: 8,823
Re: EOS R6 focus guide inaccuracy

HHaapala wrote:

ProfHankD wrote:

HHaapala wrote:


I recently bought a PK-EOS adapter, and Im now using a Pentax-M 50 1.7 on my EOS R6 with that and the EF-RF adapter. The PK-EF adapter has a focus confirmation chip, meaning it gives the R6 the ability to activate the manual focus guide feature. This uses the on-sensor AF system to help with manual focusing.

But its inaccurate. The focus guide seems to work normally, but when it says focus is nailed its actually backfocused. I have no idea how this is possible, since the focus guide is on-sensor so there is no way a misalignment or such could cause this. It simply just shows up green meaning 100% perfect focus when you're actually focused behind what you're supposed to. It even clashes with the cameras own peaking. Peaking works normally, and when peaking is on top of focus guide, guide says its not hitting focus.

Very interesting! That really shouldn't be possible with main-sensor PDAF.

My guess is that the focus chip is somehow identifying as an EF lens that needs a correction for focus data it returns to the camera. Is there even a micro focus adjustment on the camera?

Nope, there is no microadjustment. Microadjustment specific to this "lens" would fix the issue, sure, but that indeed is not an option on R-series cameras because as you said, it shouldnt be physically possible. Maybe i could see if external software can do microadjustment on a per-lens basis?

Ok. Let's assume nothing is defective....

The dual pixel phase detection is actually a lot trickier than one might think in that the SNR between halves of a pixel is terrible and the angular separation is pretty sloppy too -- at least compared with masked pixels or a separate PDAF sensor. Canon's algorithms for processing the dual pixel data are not public, nor do any of the patents I've seen make it clear how they're doing it. It is quite possible that their processing is dependent on lens parameters obtained from that focus confirm chip.

What lens does the focus confirm chip identify as?

I would guess the issue is bad parameterization of the dual-pixel processing algorithm. For example, if the lens suffers from fairly serious SA (which many normals do) or field curvature, the correct focus point could be dramatically missed if their processing essentially targets a relatively wide aperture because the marginal rays don't come into focus at the same point as the rays closer to on-axis. Basically, the question is really how different is the lens you're using from the one the chip IDs?

 ProfHankD's gear list:ProfHankD's gear list
Canon PowerShot SX530 Olympus TG-860 Sony a7R II Canon EOS 5D Mark IV Sony a6500 +32 more
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow