Lens Auto-Focus Speed versus Light Level
Ever wonder how your camera and lens focus speed changes with light levels? Everybody knows that focus gets slower in dim light, but by how much?
Focus speed is primarily a function of three things. First, the camera tells the lens what distance to focus. Second, the lens hardware (and firmware) has its built-in capability to respond to focus requests. Third, light levels dictate the quality of focus feedback the camera gets to work with. Bad light equals bad focus feedback.
There are, of course other variables that can influence focus speed, such as the ambient temperature and the subject contrast. My own tests are done in good conditions (around 70F to 80F) with a fully-charged battery and a pretty high-contrast target.
A fast-focus lens
Any focus speed measurement is only valid for a particular camera and lens combination. For the tests that follow, I used my Nikon D500. This camera seems to behave almost exactly like my D850 in regards to focus capability. I tested three lenses: my Sigma 70-200 f/2.8 Sports (at 200mm), my Nikkor 85mm f/1.4 AF-S, and my Nikkor 24-70 f/2.8 E VR (at 70mm).
For the Sigma Sport, I tried both the “standard” and “high speed” focus algorithms. Surprisingly, they seemed to perform the same at any given light level. My Sigma 150-600 is about 20% faster using its “high speed” algorithm.
I measure focus speed via how long it takes to change focus from minimum distance to infinity. For the Sigma 70-200, that means the range is a bit less than 4 feet to infinity. I use the back-button (AF-ON) focus technique while in continuous auto-focus mode and phase-detect focus. As I explained in a previous article, I use my D850 video at 120 fps to film the focus distance scale on the lens. Using this technique, I get a resolution of 0.0083 seconds per frame (120fps) to time the focus action.
The f-stop used in testing doesn’t matter, because the camera always focuses with the aperture wide open.
I measure light levels by looking at the EXIF data from the photograph, which I get using the “Exiftool” program. This program is free, and invaluable for retrieving a wealth of information about each photo.
I had intended to graph my measurement data, until I looked at the numbers. There was almost no change in how long it took to focus, up to the point of failure. In dim light, focus just got unreliable instead of slow. In really poor light, the camera would eventually go through the whole focus range a couple of times and then give up. Graphing this kind of timing behavior would be an exercise in futility.
I did all of my testing outdoors. After sunshine testing, I waited until sunset. I would take a new shot about every 5 minutes until near-darkness. I did the testing outdoors so that I could focus on a target that was about 200 yards away.
For my “bright” light testing, I’m using sun-lit shots in the later afternoon. This isn’t the brightest light you can get, but light levels beyond this don’t provide any improved focus speed. The EV (exposure value) corresponding to this light is around 14.3.
I tried light levels all the way down to EV 1.3, which is near-darkness. The only lens to (occasionally) succeed at this level was the Nikkor 24-70 f/2.8, and it started getting unreliable at EV 3.6. Tests that didn’t succeed are marked with a “---“.
Focus Time (Seconds) Versus EV Results
EV 70-200mm 85mm 24-70mm Notes
14.3 0.358 0.458 0.25 Bright Light
13.4 0.358 0.458 0.25
9.9 0.358 0.458 0.25
9.3 0.358 0.458 0.25
8.9 0.358 0.517 0.25
8.5 0.358 0.558 0.25
8.3 0.358 0.567 0.258
7.9 0.358 0.567 0.258
7.3 0.358 0.567 0.258
6.6 0.358 0.567 0.258
6.3 0.358 --- 0.267 85mm too slow now
5.6 0.367 --- 0.267
4.6 0.367 --- 0.267
3.6 0.383 --- 0.283 50% fail Sigma & 24-70
2.6 --- --- 0.909 70% fail 24-70
1.3 --- --- 0.909 80% fail 24-70
I was very surprised by these test results. I remember reading many years ago about cameras/lenses that would demonstrate a smooth exponential decrease in focus speed at decreasing light levels. Lower light levels always meant slower focus. I imagine that old camera technology (and very slow on-board computers) forced this type of behavior. Fast-reacting modern cameras can largely bypass this problem, and they remain very responsive in most levels of illumination.
The testing I did (both the D500 and D850) had very little change in focus speed up to the point that they would suddenly become unreliable at focusing. In very low light, the focusing would typically go through the whole focus range twice (near-far-near-far) and then quit. Sometimes final focus was okay, but mostly it would fail.
The Nikkor 85mm f/1.4 would get “near” to the correct focus distance, and then really slow down just before achieving final focus (in low light levels). I consider this a “fail”, since it would take about 2 seconds to finish focusing. This lens has always been notorious for being slow to focus (most f/1.4 lenses are).
I also noticed that my D850 video (used to capture the lens focus scale action) showed how pathetic its contrast-detect focus was in dim light. It would end up hunting back-and-forth, rarely locking on the target. I totally understand why cinematographers stick with manual focus.
Each lens/camera combination you own probably has its own focus behavior fingerprint. Knowing how your gear behaves before you go shoot something important might just save the day. There’s a reason why lenses have that manual focus ring on them.