Fujifilm just released firmware 3.0 for the GFX 50S. One of the new features is what Fuji calls Focus Bracketing. In my opinion, it is misnamed, as it is optimized for making exposures for stacking. However, I’ll use their terminology here. It works a lot like the Focus Shift Shooting feature in the Nikon D850. You tell the camera the number of exposures to make, the step size between exposures (a number from 1 to 10), and the delay between each exposure, with 0 telling the camera to make them as fast as possible, which in the case of the GFX with EFCS on is about one per second.
The feature differs from the D850’s FSS in several interesting ways:
Focus Bracketing on the GFX is a drive mode. Once set to that mode, every depression of the shutter release will result in a bracket series with the last settings. The only way to stop the bracketing is to switch to another drive mode.
The D850 forces the use of the electronic shutter. The GFX does not.
At the end of a Focus Bracketing series, the GFX returns the lens to the original focus position. The D850 leaves the lens set to the last focus position of the series. This is good news and bad news. The way the D850 does it makes it easy to see how far the focus shifted during the series without going into the playback mode. But it makes it harder to repeat a series with the same starting position and, say, a different f-stop or step size.
There are several obvious questions about the new Focus Bracketing feature.
- Does the step size change with f-stop?
- Does the step size change with distance?
- Does the step size change with focal length?
- What is the smallest step size in terms of the focussed object plane?
- What is the smallest step size in terms of the focussed image plane?
- What is the smallest step size in terms of the image plane circle of confusion (CoC)?
When you’re done reading this post, you won’t know the answer to all those questions, but you’ll find that all but one are answered explicitly, and you’ll have enough information to be able to calculate the other one.
I started out with my checkerboard ramp about 2 meters from the camera, with the Fuji 120 mm f/4 macro mounted. I lit the ramp with a single Wescott 1×2 foot panel in a Westcott softbox.
I focused the camera on a spot near to the front of the ramp, set the number of exposures to 32, the delay to 0, and the step size to 1. I made sequences of exposures at f/4, f/5.6, f/8, f/11 and f/16. I bought the images into a Matlab program that I’ve written to identify the focused plane and spit the results out to Excel. When plotted, the object plane versus exposure looks like this:
The camera moves the focal plane away from the starting position faster if the f-stop is set narrower. This is what you’d like to see.
If we look at each step in the series, rather than the cumulative effect, this is what we see:
The f/16 line has a lot of variation. I don’t think this is because the camera has much variation, but rather that the program has a hard time determining the focal plane accurately when the depth of field (DOF) is as great as it is at f/16.
Looking at the average step size for each f-stop:
It goes up and goes up faster as you stop down further.
Looking at the movement of the image plane (with respect to the sensor) and changing the units of the vertical axis from millimeters to micrometers:
So it looks like — just as we saw with the D850 — the camera is smart enough to adjust the step size to give a constant CoC change as you stop down. This is exactly what you’d want it to do. Because it works in image space, the D850 is insensitive to focal length or subject distance. So is the GFX.
Note that the minimum step size is small enough for accurate focus bracketing. The pixel pitch of the GFX is about 5.3 um, and thus the minimum step size CoC is less than 40% of the pitch.
What happens when you change the step size?
I set the aperture to f/4, and made series with the step size set to 1, 4, and 8.
Except for the first step, the results are very linear. Looking at the average CoCs:
You can see that you can approximately double the selected step size to get the CoC in micrometers.
I think this is going to be very useful. You probably noticed up at the top of this post that I said this wasn’t optimized for focus bracketing, and now I’ll tell you why. It goes in only once direction. For focus bracketing, you’d want some mechanism that made exposures with the focal plane on both sides of the original focus point.
It could also use some enhangements for focus stacking, such as the ability to set the stap point, the step size, and the end point, and have the camera detirmine the number of steps. But it’s a great start.