There is some recent news about EM1x bird AF settings, both on how it works, and using a 25 point grid. This has come from several sources, including a couple of YouTube videos, so I thought it might be useful for EM1x fans to put it all together in this short post.
EM1x firmware 2 bird AF – how it works
I was recently alerted by PV Hiker at Fred Miranda.com to this video from Dave Etchells of Imaging Resource. It was done nearly 2 years ago, but I clearly missed it. In my view Dave Etchells is one of the best technical analysts of cameras and photographic technology on the internet. Unlike just about every other internet pundit he is trusted so much by the developers that they talk to him directly, and that way he gets unparalleled technical insights into camera systems. I really trust his technical ability, and what he says about the the Olympus bird AF is fascinating.
Dave describes intelligent subject detection from around 12:30 minutes into the video, and at 14:10m he says “I’m told the processing is done in the Truepic processors themselves, it’s not using the phase detection at all, what it’s doing is just looking at the data itself, and analysing the objects within it”
From this it seems clear that the intelligent bird detection (IBD) process runs a separate pattern recognition algorithm purely in the Truepic processors to identify the bird. The pattern recognition algorithm identifies which focus points to use and then hands over to the phase detect focusing system to obtain physical focus. That might explain why you can get a perfect green box around the bird but the subsequent image can often be out of focus.
By the way Dave also inter alia explains at 13:55m why train recognition was included in the first release of intelligent subject recognition, a subject choice that was completely baffling to most European and American photographers.
EM1x firmware 2 bird AF – what size grid?
All the internet “how to” guides on intelligent bird detection say that you must use the “all points” grid but I’ve never seen this stated by Olympus themselves. I wondered if a smaller focus area would help the focus accuracy as it does in normal CAF, so I tested the bird AF with the 5 by 5 grid, and this works also.
If you start with a 5 by 5 grid, the bird recognition box will operate outside that grid and will identify a bird anywhere on the screen. That is explained if indeed the bird recognition is a separate process from the bird focus.
All my prior experience with the EM1x and E-M1ii, has been that small grids result in better hit rates than the all-points pattern. This suggests that the bird AF focus accuracy might be improved with a 5 by 5 box if you use the bird detection AI to identify the subject, and then move the bird back into the 5 by 5 box to focus if it falls outside. I have not been able to try this, but have just come across this excellent video from Andre Boss, who has been using the E-M1x and IBD for over 2 months apparently. Andre reports at 2:48m into the video that he has had improved accuracy by reducing the grid to 5×5, so this method clearly works. As soon as the weather improves, I’ll be out to test this. He also suggests reducing sensitivity to -1 or -2 for stationary or slow moving birds. Andre is an excellent pro wildlife photographer, and has won many awards – he is really worth following.
All great input, adds to my prior experience, and a good basis for further tests.