Here we see that Four Thirds lenses behave natively with E-M1, they only support PDAF. CDAF can not be used effectively with AF drives inside those lenses. Micro Four Thirds lenses are focused also by their native CDAF except for continuous AF where E-M1 mixes both systems. Single AF is already fast enough with CDAF. Contrast AF is here similar to E-M5 or E-P5.
Notably Olympus has not implemented PDAF for video shooting. I would guess the reason is that they have been lacking resources to write and test code for E-M1. They wanted first to make sure that both lens series work well enough for stills photography. Video should be on agenda for the next body. Please remember: I would guess. Olympus has not given me any explanations or promises.
Choosing the AF points on screen
There are 37 focusing points for PDAF. They cover the image area centrally up to half way to the edges. These points are used also for CDAF and they are surrounded by 44 more CDAF points. These 81 points cover almost the whole image area.
With PDAF lenses (FT lenses) the total AF area is always show with black corner marks in viewfinder.
All 37 PDAF focusing points active. Camera will choose which one to use primarily.
You can choose also a three by three group of focusing points. This group can be moved across PDAF area, even partly outside of it but only those points which remain inside will be used.
Only a single point, the center point chosen as active. You can move the point around with control dials or four by four control. With PDAF on sensor system every AF point is as sensitive and accurate. So there is no more need for the typical preference of center point as in DSLRs.
Another option is small PDAF AF points with center point active here. Normal size point is more accurate and easier to use in most situations. This size is better when you need to focus on a tiny detail or normal size point contains details which are too far apart in depth.
The area for CDAF focus points is the same as in E-M5. Inside this area you can, as above, use a group or two sizes of single points with mFT lenses.
How does E-M1 PDAF work?
Olympus has a patent on phase detecting AF, PDAF, which was filed last year. Actually they filed three patents and this system is based on the simplest of those. It is not as demanding on image quality, sensor manufacturing and driving as what we see in Canon´s new 70D sensor and it is a lot easier to correct than the other systems used by other companies. The problem is, you can´t use an element in sensor for both AF and image at the same time. Either you have to alternate usage (Canon) or use only for AF and correct later (the others). A PDAF element is like a dead pixel for the actal image, it must be compensated through software calculation in image processor. Olympus has placed PDAF elements so that every AF element is always surrounded by enough ”real” pixels in every direction. Thus PDAF elements should have no practical negative effect on image quality. The new True Pic processor interpolates an image pixel value for each PDAF pixel and saves it into RAW file. Thus Adobe and other companies making RAW converters do not need to take care of this.
Those 37 PDAF focusing points contain PDAF elements. Olympus is not telling how many PDAF elements there are all together on sensor or inside any target. Illustrations below show schematically how those elements are arranged and used. Illustrations give just the idea, and are also not meant to be used as any basis for calculations for the actual amount of PDAF sensors.
Each PDAF element is shown here as a half circle. Each of them recognizes only light coming from its side, i.e. either from the right side or left side of the lens. Like said, each element is surrounded by a field of normal image forming elements not to affect image quality.
While these "loosely" situated PDAF elements are good for image quality they would be bad for AF accuracy if they could not interract correctly. That´s why they are placed so that every focusing point is actually a point in a ”virtual” line. They are ”pulled together” into horizontal lines in software (not actually on sensor) and calculated as shown here. Some other cameras may have actual lines or just one line of PDAF elements, but this is better if other variables are similar.
These kind of horizontal phase detection lines see only structures of contrast which go across it, i.e. it does not detect horizontal structures. Most subjects contain structures in all directions, just avoid trying to focus on simple horizontal lines. One benefit of on-sensor PDAF is immunity to lens speed or lens vignetting. DSLR systems have always a treshold aperture, they do not work with slower lenses, and some focus points won´t work if lens vignettes too much.
(Basically there could also be up and down elements but no manufacturer has those yet in their on-sensor PDAF system. In DSLRs PDAF sensors which contain also diagonal lines are called cross type sensors. They can detect all directions but they also need usually a fast lens like f/2.8 or better to work at all, while normal elements have a treshold aperture of f/5.6.)
E-M1 in practise
According to Olympus: Compared to Olympus´ top of the line DSLR, the E-5, PDAF is now faster as it can focus up to 6,5 fps instead of E-5´s maximum 5 fps. Also instead of 11 focusing points there are now 37. Single AF speed is same as before because it is limited by the AF motor in lens. I have used E-5 but there was no demo E-5 available at Olympus Finland while I had E-M1. Based on what I saw now and looking at my earlier E-5 images and thinking about those shooting situations, while I can´t proof anything, I have nothing to say against what Olympus claims. Seems reasonable enough.
First I went to shoot some birds.
I am no nature photographer, so... Anyway my idea was to catch birds in flight. It was noon and nobody was flying. At least they were moving and I had some time to see how well a small focusing spot kept a small object in focus. No problem here. Above is the whole frame and below 100% crop. These are shot as jpegs. Never mind sharpening settings and all, I had no time to try various options for those or be pricky at image quality at all. I wanted just to see how focusing works. The lens is Zuiko D. 150mm F/2.0 FT lens with Zuiko 1.4X tele extender. Total focal length is 215mm which has the same angle of view as 430mm lens in a 35mm camera. Handheld, 1/3200s, f/5.6.
Sometime later when I happened to be a little higher, there was a sudden rush of movement for a second. I turned around, got the birds in frame, an immediate focus lock and after a few frames they were just swimming again. This here is the first frame. My focusing point in use happened to be chosen from below center and there was no time to change anything. The crop is here 50%. Same lens/extender combination, handheld, 1/640s, f/4.5. This one would have been absolutely impossible to get with E-M5.
As there was nothing much happening I tried various things about how focusing system works like following the birds through tree foliage etc. The crop is 100%. Same lens/extender, handheld, 1/160s (!), f/4.0.
While not a bird, these wings look like having done lots of flying. Same lens/extender, handheld, 1/800s, f/4.0
Next I went to an overpass to try following incoming cars. Here the speed limit is 80 km/h. I tried both the above combination and M.Zuiko 75-300mm f/4.8-6.7 II mFT lens. This one is with the latter at 156mm focal length. Handheld 1/500-1/1000s, lens wide open f/5.7. Of these 15 images I would call three as unsharp because of bad focus. I followed many cars and noticed only afterwards as I saw the images on computer screen that the results varied more than I realised when shooting. I think one reason is that I used only the small center focus point. If the focus point sees only the hood without any details there is nothing to focus at, and if the photographer misses pointing the spot at the car altogether that doesn´t help either with consistent results... Well among inconsistency I did not see any trend among those two lenses. There were good and bad series. I guess the best result should be obtained with a focus group and with tracking on. I will come back to this test later when I get another chance to try E-M1.
100% crops form the first and the last image.
Lets start with these three consecutive frames (6 fps). I am following here the boy with ball and then suddenly those two guys run in between. What does E-M1 do?
As you can see E-M1 tries to refocus at those two boys (but didn´t quite catch them) and then snaps back where it should be. This behaviour is common to all AF systems. The tendency to keep focus on object or change it for another object coming into focus point is controlled by a latency setting as shown below. Here it was at Normal. This series was shot with M.Zuiko 75-300mm f/4.8-6.7 II mFT lens wide open at f/5.3, 1/2000s. A totally impossible situation with the same lens on E-M5. It could never get the right focus lock back this fast.
This is the menu to set suitable amount of C-AF lock. Off means that E-M1 tries to change focus to a new object instantly and at High it is most reluctant to do so. This is something every user must test to find which setting is best for one´s type of photography.
Below two more examples with each already mentioned lenses from this junior football game. There were hits and misses, sharp and unsharp with perfect ones with a good enough percentage to see that E-M1 is definitely in DSLR class rather than mirrorless class with continuos AF what ever the lens type.
The first two images with Zuiko D. 150mm plus 1.4X extender and the last two with M.Zuiko 75-300mm.
On-sensor PDAF does not remove the fact that there remains one uncertainty in focusing. The lens. With DSLRs there´s another one and that is the camera body. Image sensor and AF sensor might not and many times are not aligned absolutely right. With on-sensor PDAF we get rid of this problem. PDAF system measures the amount of phase difference and tells the lens to shift focus a certain amount. This translates into a certain amount of rotations for electric focusing motor in lens. This mechanical calibration can be slightly off in any lens. That´s why E-M1 includes data for every FT and mFT lens, and this data can be adjusted if any single lens is not exact enough. I had no time to check this feature in E-M1 but it is familiar to anyone who has done this adjustment in Olympus E-5.
I think E-M1 is now in a good DSLR class regarding C-AF. It is not in the top class. Where more exactly and comparable to which DSLR? I really don´t know. It would need more time be able to test various settings at various situations and doing a comparison with some well known benchmarks. Anyway, everyone coming from Olympus E-5 or any other FT DSLR should be happy, and any comparison to any other mirrorless camera is pointless.