Fuji X-Pro2 with Canon TS-E 45mm f/2.8 (via speedbooster adapter) at ISO 4000
Welcome back! Let’s dive right in, as I’m very eager to talk about the AF performance. It takes a lot of shooting before it’s even remotely possible to talk about it with any weight whatsoever, but it will make or break the whole idea of shooting weddings with Fuji so it’s of paramount importance.
It also takes a lot of testing with different lenses to account for differences amongst them. So, here are some thoughts after 20.000 shots with 4 native XF lenses.
First, a disclaimer. X-Pro2 has a standard performance mode and a high performance mode. The latter increases EVF refresh speed and AF speed, but eats battery more quickly. Everything in this review was tested in the high performance mode. I don’t feel standard mode is for pros, unless they happen to be running out of fresh batteries.
Now let’s start by breaking down the three important parameters which greatly influence the AF speed (besides a few obvious ones, like the amount of light and contrast on the scene).
First, type of lens attached. Second, AF frame size. Third, the black box that is the hybrid AF system of the X-Pro2.
Let’s go through each one in turn to see what they mean for AF performance.
(1)
Different lenses focus differently, that should be no surprise. With Fujinon lenses, the variation is very expressed.
I should now point out I only tried the XF lenses (which is the Fuji equivalent of Canon’s premium L-series), and of that, three out of four are fast primes: 16/1.4, 18/2, 35/1.4 (with 1,5x crop factor that gives 24/1.4, 28/2, 50/1.4). The final lens is 10-24/4 OIS (eq. 15-36/4).
One quick note is to mention that the XF 23mm f/1.4 and 35mm f/1.4 have their newer f/2 counterparts which by all reviews have much, much faster focusing speed (they’re also smaller, lighter, cheaper and weather resistant). Those lenses simply aren’t relevant for weddings (at least not for me) since I need all the shallowness of DOF and aperture I can get, but it’s good to bear in mind.
More generally, the wider the lens, the faster the focus. Specifically, the 18/2 is very fast, the fastest among those I tried. It’s a very light lens with pancake design so that’s not surprising. Even in very low light, focusing is surprisingly fast, although not as precise as I’d like.
The 35/1.4 is a different side of the spectrum. Focus speed is very variable. It’s a pretty relative thing to say, without some numbers (which would probably mean nothing to you), but I’ll give you some real world scenarios.
For relatively stationary subjects with enough contrast and in good light, it’s actually speedy enough that you wouldn’t notice it focusing. More specifically, while initial focus acquisition can lag a bit, focus re-acquisition (when subject is nearly at the same distance as on previous focusing attempt) is practically instantaneous.
In good light but with tricky conditions (like shooting against the light or focusing on an area with low contrast), the focus is terrible. Camera tends to hunt aimlessly without locking on to anything, usually leaving focus at worst possible point (closest focusing distance) so that everything is hyper-blurry in the EVF and it takes the longest time to drive it back closer to infinity. I’ve been close to missing some key shots because of this. Re-trying with a slightly different framing usually helps.
Fuji X-Pro2 with Fujinon XF 10-24mm f/4 OIS at ISO 200
One detail that goes to Fuji’s advantage is the fact that it’s sensor is using all the available light that fast lenses (like f/1.4) afford them. For comparison, in Canon’s case of 5Dmk3, the camera throws away all the light after f/2.8 because that’s the way AF sensor is constructed. So while you may think that your fast f/1.4 prime will make it easier for a DSLR to focus in low light, it’s not the case — extra light is thrown away, and all that remains is the extra shallow DOF with heavy glass elements for focusing.
It needs to be said, however, that dedicated AF sensors on DSLRs admit more light in the first place (because they lack the color filter array) so it’s not as if Fuji has an unfair advantage. My point is just that, in regard to focusing speed, faster lenses pay off with Fuji, while with DSLRs they don’t.
Please note that all of this low-light talk is without one very important parameter — AF assist beam from the external flash. 5Dmk3 is nearly useless without it in deep darkness and Fuji is somewhat better.
The reason why I didn’t do any tests with the AF assist beam is twofold. While X-Pro2 has AF assist in-camera, it’s mostly handicapped by the lens hoods and/or big lenses, it’s white and irritating for your subjects plus it’s range is mostly useless. So I chose to simply turn it off.
Second reason is that the pro flash I’m waiting for (the Fuji EF-X500) still hasn’t been released. It’s supposed to have an AF assist beam so we’ll see how that works out. My only real concern is that the AF assist will again be a white irritating LED, which is a deal-breaker for candid photography.
Even so, the X-Pro2 low light focusing performance is surprisingly good when you get the hang of it.
Let’s talk about AF frame sizes. If you’re used to DSLRs, you’re used to having one fixed focus point size and you’re probably not even thinking about it. (Yes, on many cameras you can expand it, but that simply includes neighbouring AF points in the process.)
On the X-Pro2 you can actually pick the AF point (or frame) size in 5 steps. The idea is to use as large AF frame as possible to cover your subject, increasing the amount of contrasty stuff in the frame to speed up the focusing (hence using a bigger part of sensor).
You’ll use small AF frames for precise work like subjects who are smaller in the frame where you want to avoid the AF picking up on the background, or shooting through a fence or leaves for example. The bigger AF frames come in handy when your subject is closer (big in the frame), especially if it’s relatively uniform or low-contrast, like focusing on a bride’s white dress against a white wall. You can also use it when you simply want faster focusing and can afford it, for example a couple leaving the church shot from afar or simply shooting in very low light where any extra photons count.
I mostly use middle 3 sizes of the AF frame at focal lengths I prefer.
(3)
Ah, the mysterious hybrid AF system. Basically, the X-Pro2 uses a combination of contrast focus detect (CDAF) and phase detection focusing (PDAF). It decides what and when to use, and as much as you might like not having to think about it, it makes it less predictable (i.e. you expect fast focus acquisition via PDAF, but then the lens starts hunting, a telltale of CDAF).
I’ve read somewhere that CDAF is used in 2/3 of all cases and always for outer focus points (PDAF only covers the central area, around 40% of the frame). Also, PDAF is supposedly used to gauge rough focus, and then CDAF takes over to fine tune the exact focus.
Fuji X-Pro2 with Fujinon XF 10-24mm f/4 OIS at ISO 200
I’m saying all of this because in my DSLR days I would never consider a camera without PDAF. It was a deal-breaker. So when I saw that X-Pro2 has both focus modes I was happy — I’d just use PDAF and never use CDAF. Right? Well… now it turns out things work a bit differently. They still work well, but I just wanted to make it clear — you’re not the one making the choice between CDAF and PDAF (unless you select one of the outer points that work only with CDAF).
Phase detection AF in DSLR and mirrorless systems
As you may have noticed, PDAF in DSLRs is somewhat unreliable. What you may not have realized is that the scattering it exhibits is inherent to the system as such, even with perfectly calibrated lens. There are better and worse cameras in this respect, but all exhibit at least some focus distance scattering — 5Dmk3 was especially bad at this.
Knowing that, I tended to shoot each important shot at least three times with refocusing each time in between, to make sure at least one shot on average is in perfect focus. This caused me to have way too many redundant shots when I got back home.
Due to way it works, the Fuji hybrid AF system simply shines in this regard. When it confirms focus, it means business! 90% of shots are in perfect focus with no scattering at all. Combined with instant image review in the viewfinder where you can see whether your subject turned out okay and check your exposure at the same time, this was a whole new outlook on shooting. You instantly know whether you get your shot. Wow!
These features liberated me from now unnecessary multiple shots, which meant both more pleasant experience for my subjects (less redundant clicking) and more pleasant experience for me when culling the images back in the studio. I instantly halved the number of images I shot per wedding, while improving the technical quality of shots!
What’s more, when I occasionally return to 5Dmk3 for various reasons, I get annoyed by the lack of a bright EVF with instant image review. Not to mention live histogram. It’s funny how quickly you can get spoilt.
Okay, on we go. The continuous autofocusing (AF-C) is selectable via a rotary physical switch on the front, as I noted before. In short, AF-C in single point mode works well enough that I use it when I have to. On Canon this was something I didn’t give much thought (the button was always under my finger), but here you have to commit to it by reaching for the physical switch. The focusing reliability is variable, and turning it on/off isn’t so quick that I could do it in a split-second. It has to be a deliberate decision and you also need a bit of foresight to do it in time. As I mentioned before, on Canon I can temporarily engage it by holding the DOF button (reassigned to servo focus) which is just under my finger. It takes milliseconds to press it and let it go (to return to single AF drive).
AF-C reliability can vary from 80-90% of shots in perfect focus to 20-30% of shots in focus. It depends on the lens, lighting, amount of contrast in the scene, size of the AF frame, your skill to keep the point on your moving subjects and finally speed and direction your subjects are moving (Fuji suggests to use this mode for subject moving towards and away from you, not laterally).
All of this makes me somewhat wary of using this mode, especially if I need to use it in backlit condition, which are Fuji’s kryptonite.
It’s not like Canon is perfect in this regard. Most often I used 50/1.2L tracking subjects as they walk down the aisle, often backlit. The keeper rate was roughly 40-50%, but it was consistently 40-50%. I could be pretty sure I’d have 4-5 usable shots out of 10, and on one of those the couple would have a nice smile and open eyes.
With X-Pro2 it’s much more stressful for more than a few reasons.
One thing that’s endlessly confusing (until you get used to it) is the hunting AF-C tends to perform. On the face of it, it looks as the focus is often hopelessly lost, periodically hunting around and missing the subject completely. But as I read more and verified the findings, it turns out that this hunting is deliberate and won’t actually influence the sharpness of your shot! How’s that for an ironic twist?
In other words, when you trigger the shutter during this hunting in AF-C mode, the photo gets taken immediately and it’s magically in-focus!
The explanation is that the camera wants to verify the perfect focus by driving the lens around what it thinks is the correct focal plane and checking for contrast changes, but quickly returning to the best point as you depress the shutter.
Fuji X-Pro2 with Canon TS-E 45mm f/2.8 at ISO 200
However, there is one occasion where AF-C gets hung up on a really low contract object, when it starts to look for focus incredibly slowly. It literally takes 2-3 seconds on the XF 35/1.4 to go through the whole focus range — an eternity in real world shooting conditions.
The real trouble isn’t the low contrast object; no AF would lock focus on it. The trouble is that when you realize you’re doing something wrong and quickly reframe the shot so that AF point is over a contrasty object, the slow hunting still continues seemingly forever, unless you stop and re-initiate the AF… but even when you do, it’s usually so out of focus that it still takes some time to lock proper focus.
This happens rarely, but when it does, you’re toast and the shot is guaranteed to be long gone by the time you finish the negotiations with the AF.
Anyway, even if you pick a fast focusing lens and have decent conditions, there are two other tricky parameters to consider.
First, keeping the focus point on your subject. It’s not tricky as such, but what I always used to do on Canon is move the focusing point while tracking to shoot the subjects so that the framing was always perfect (especially useful for lateral subject movement). On Fuji the joystick for moving the focus point only works if you’re not tracking a subject, so you have to stop shooting/focusing to move the point in order to get desired framing. That’s a huge disadvantage.
Second, the focus frame size. Remember how you always want to keep it as large as possible? When a couple walks towards or away from you, this ideal size changes. And it’s very, very tricky requiring a lot of dexterity and quite a bit of time to resize the AF frame on the fly, while tracking and shooting. (Actually, it takes less than 1 second when you get in the groove, but if the couple is walking at normal pace down the aisle, forget about it, you’ll be missing shots.) You’ll need a few of these changes and it’s just too fiddly and stressful to even attempt it.
It’s better just to stick with one size, even though it won’t be ideal. Worse still, interrupting the AF-C mid-way means it will take it more time to re-acquire focus once you resume the tracking.
I should probably at least mention in passing something Fuji calls “focus modes” (a term I usually reserve for single/continuous/manual focusing). The modes are single, zone, and wide. They relate to different areas of viewfinder where camera tries to find focus and more importantly, detect subjects and their movement by itself. Any of those can be used in single or continuos focus mode, and you can choose different sizes of active areas. Zone is recommended for tracking predictable objects, while wide is supposed to track objects with unpredictable movement (although it uses only CDAF, paradoxically).
Fuji released a whole booklet describing various shooting conditions and when to use which mode with which settings. I read it, tried it for a few days and just quit. Single AF point + continuos tracking worked best, as described above. The other modes might be fun to try if you’re a hobbyist, but not for pro work in my oppinion. They work great when they do work, but they’re way too unpredictable.
There is yet another way of focusing a reasonable moving object, using the ordinary AF-S mode with a particular setting. The technique is called shutter-mash and it’s a kind of a “squeeze and pray” workaround.
There’s an option to set the camera for release or focus priority in AF-S mode; you should set it to focus for this to work. Then, you simply set the point over your subject and depress the shutter fully in one swift motion. Camera will delay triggering the shutter until it’s reasonably certain it’s in focus and then it will take the photo. This sounds like a catastrophe waiting to happen, but for some reason it works really well. Of course, the delay will be noticeable (sometimes), but the hit rate is really high. So it’s all the question of whether you can deal with a bit of uncertainty about the delay.
I found it works wonders for most situations I encounter during the wedding day, with a keeper rate of around 60-70% in decent light.
So here’s how I use both modes for moving subjects: if I can predict the subject will be moving a lot (like walking down the isle, leaving the church, entering the venue), I switch to AF-C. If I’m caught by surprise with no time to switch on AF-C (or no desire to have to turn it off a second later), I’ll use the shutter mash technique.
Fuji X-Pro2 with Canon TS-E 45mm f/2.8 at ISO 200
The X-Pro2 has pretty interesting face detection modes. Besides recognizing faces, it can detect eyes and you can even choose whether to prefer closer or farther eye when focusing. It looks great on paper, especially for weddings where we tend to shoot people above else. And it works in average conditions… but it’s a double edged sword.
Pattern recognition at this point is still far behind what we as humans do, so the detection simply isn’t reliable. Have someone turn their head a bit or go further away from the camera, and bang, the magic is gone. Same goes for anything but the completely static scenes. And the trouble is that when you start to depend on the camera to detect faces and focus on them, you stop moving your focus points (which is the whole point), so when the camera fails, you’re usually off by a mile!
“Why not simply switch it off when it doesn’t work?”, you may wonder. Because it’s really unpredictable. It can detect the face 5 out of 5 times, and then fail for next 5 times because your subject turned their head 30 degrees or walked 3 steps away from you. In photojournalism, chances for great shots don’t come by twice, so getting the shot first time is imperative.
This is of course just a matter of pattern recognition software, so I hope future updates can turn things around. At least for casual shoots.
Manual focus
Now we’re getting into the interesting territory. Anyone who’s used a DSLR without specialized focusing screens realizes that manual focus through the viewfinder is a gamble. There are rare exceptions, like tilt-shift lenses, but that’s it.
X-Pro2 shines when it comes to manual focus. It offers just about every focusing aid you can think of, including some unique ones like the ERF mode, covered in first part.
I quickly settled on my favourite workflow when it comes to MF. I use the EVF and focus peaking for live action, and simple 100% magnification for slower paced shots (like sessions), when the subject is stationary enough to allow the zoomed in view to be useful.
Focus peaking is quick and dirty, simply ideal for cases where you need quick focusing and when the scene contains enough details to actually cause the right things to peak as you hit them. It’s amazing for working with tilt-shift lenses since the plane of focus is so obvious and localized.
Zooming into the scene is even more precise, but it takes one or two extra steps (clicking the rear dial) so it’s not as suitable for live action, especially since you need to move the focus point so the camera knows where you want to zoom in.
If your image has a lot of vertical lines, this tends to work great, but for anything else it’s pretty useless. B&W image throws away any color information that may help you when color works to your advantage (like a red flower in front of green bushes) and even when you line up the image stripes you’re not 100% sure when to stop fiddling with focus.
Lastly, the B&W portion of the image messes up with my framing. It’s hard to think of a harmonious composition when half of the image is black & white.
Even so, what doesn’t work for me might work for you, depending on the subjects you like, so give it a try. It’s easy enough to switch between MF assist modes by holding down the rear dial for 1 second.
One interesting point is that a few wide angle lenses, like 14/2.8, 16/1.4 and 23/1.4 have a clutch mechanism on the lens itself which needs to be disengaged to allow MF (but even then you need to set MF on the camera; it takes a bit of getting used to, so read the manual). What this gives you is a linearly accessible focus range with definite stops on each end — something I sorely miss on the rest of focus-by-wire lenses (basically all other lenses). Actually, I shouldn’t say linear, but non-accelerating.
Now that I mention it, the focus-by-wire system that most of XF lenses use features focus drive acceleration. This simply means that a quick sharp turn will move the focus more than a slow, longer turn, which tends to be more precise and appropriate for focus fine tuning. While this sounds great on paper (it works great for your mouse, after all), I really don’t like it.
Maybe it’s the years of using MF lenses with mechanical (direct) focusing talking here, but it’s really confusing to add focus ring turning speed to the equation of already too many variables. If I could manage to always rotate the focusing ring with exact same velocity, it wouldn’t matter, but I’m not a machine. When I’m using manual focus, in the heat of the action I might turn the ring really quickly, jumping from 1m to infinity in the blink of an eye, missing the subject by a mile. Or I could turn the focus slowly to focus something near me… and turn… and turn… and turn. You get the picture.
Finally, there’s an AF+MF mode. It’s similar to Canon’s FTM (full time manual) override that their USM lenses feature, but not exactly the same. This is probably a good time to make it clear that in most cases, when X-Pro2 is set to AF, turning the focusing ring does absolutely nothing. The only exception is after you half-press the shutter and the AF+MF function is turned on. Only then will the focus ring actually move the focus. After you correct the focus, you can take the photo. If you let go of the shutter button before taking the photo, you need to do everything over again.
Fuji X-Pro2 with Canon TS-E 45mm f/2.8 at ISO 200
One of the Fuji’s mythical qualities is their willingness to upgrade their cameras to seemingly no end with new firmware versions. While Canon usually sticks to fixing a few bugs and that’s it (with rare exceptions), Fuji was known to overhaul the whole camera a few times (just look at the X-T1 and it’s firmware v4.30).
With that in mind, I’ll say that the X-Pro2 firmware I tested it for this review is v1.02, with the firmware v2.00 expected soon. Among other things, it promises to bring better AF algorithms with parallel processing for CDAF, while PDAF should now be used in 2/3 of the cases (instead of 1/3). This will probably mean a lot and will bring the camera to the level of X-T2, which supposedly has much better AF performance (I’ve yet to try it).
When the firmware v2.00 is out, I’ll test it and publish the fourth part of this review.
Fear and loathing of the AF microadjustments
One thing I truly hated on DSLRs, yet found no way around it, are the dreaded AF microadjustments. Let me just set up the scene.
I have 4 Canon camera bodies and 8 lenses, each one with their own tolerances in production. That means that most of the lenses don’t give proper focus on most of the bodies.
The way it’s done is time consuming and unreliable (depending on many factors like illumination levels, white balance and even temperature). Even with semi-automated systems like FoCal, it’s still unreliable for many lenses and manual corrections with a lot of testing are needed, just like periodical re-calibrations. This is one part of my life a s photographer I truly dread(ed).
It all begins (and ends) with the fact that the AF sensor on DSLR is not in the same place as the main imaging sensor. The point of the calibration is to correct this discrepancy. (This is a simplified view, but it’s enough for our purposes)
On mirrorless cameras in general, the problem is magically solved — the main sensor is used for AF, so when it thinks image is in focus, it really is in focus. As simple as that. Any possible manufacturing tolerance has been accounted for by the virtue of sampling focus where the photo will actually be imaged. No more calibration!
While you may think it’s not a big deal, it’s a really big deal as you gather more bodies and lenses, use and abuse them, take them around the world etc. Tolerances aren’t fixed, and it’s incredibly liberating to just buy new gear and use it without tinkering with focusing charts, tripods, lighting etc.
Fuji X-Pro2 at 30 sec with Fujinon XF 16mm f/1.4 at ISO 3200
Review: Part 1 / Part 2 / Part 3