- Ed Dozier
Nikon DSLR Camera Focus Speed Comparisons
I thought it would be fun to see how Nikon has progressed with improving focus speed over the years. I picked cameras starting with the D50, introduced way back in June 2005. I tested a representative sample of their DSLRs up through the D850.
For the tests shown in this article, I chose my Sigma 70-200 f/2.8 Sport lens, since it’s about as fast at focusing as anything that I’ve got. My focus speed timing tests are usually done in bright sunlight, but for this series of tests I chose to try heavy overcast conditions. It turns out that focus speed didn’t seem affected by this lower light level (EV 14 to be exact), which is good news.
I bet you think that the D50 probably has Alzheimer’s disease by now, and likely has forgotten how to autofocus. You’d be very, very wrong on that point. You’d be right, on the other hand, to think that its sensor is total crap compared to anything modern.
For all testing, I chose the center focus point, since they’re always the most accurate and sensitive. I used slow-motion video from my D850 (120 fps) to monitor and time the focus action, except when testing the D850 itself, where I used the D500 with its 60 fps video. I start with the lens set on minimum focus (about 4 feet), and then I have it focus on a distant high-contrast target to force the lens to focus on infinity. The focus target here is “easy”; I didn’t try anything tricky to focus on.
I used the Sigma USB dock to program the lens focus algorithm on my 70-200 f/2.8 Sport lens. I selected the non-default algorithm called “High Speed AF” (or “Fast AF Priority”), which has the highest available speed of its three focus algorithms. It has proven to be totally accurate and reliable, so it makes zero sense to me to use a slower focus algorithm, such as the default “Standard AF”.
That’s right: that's a Nikon D50 behind that 70-200 lens
Sigma USB Dock: program lens focus speed behavior
The Sigma Optimization Pro program that is used with their USB Dock to program any of their modern (Global Vision) lenses is shown above. Besides focus speed behavior, it also lets you customize focus fine-tune calibration at various focal lengths and distances and also the optical stabilization characteristics. Companies like Nikon and Canon don’t have anything as sophisticated as this for their lenses.
This series of speed tests produced some real surprises, both good and bad. First, I’ll show a table of results, followed by an analysis of the character of the focus behavior.
Camera comparisons
If you study the tabulated results shown above, something should stand out to you: The Nikon D50 performed very, very well. In fact, the D50 beat the D7000, D7100, and D610 cameras. How is this possible? I believe the answer lies in the camera’s “focus algorithm” and not the camera raw focus horsepower.
Detailed Focusing Behavior Analysis
For each test, I carefully studied how the camera would focus the lens in the slow-motion video frame-by-frame. The key to fast focus has a lot more to do with getting straight to the distance goal than with raw speed. Again, each test started at the lens minimum distance and I had the lens focus on a distant target requiring an infinity setting.
A smart focus algorithm won’t cover the same ground more than once; a dumb focus algorithm will move forward, then backward, then forward to get to its goal (or the reverse for a near target).
To nobody’s surprise, the Nikon D500 and D850 focus algorithms are “smart”; they move straight to the focus goal, and they do so at considerable speed, basically skidding to a stop at infinity. These cameras don’t suffer from getting distracted or being lazy; they’re all business when it comes to the job at hand. The results are quite close to each other, and probably within experimental error of being a tie. They finished within 1/100 second of the same time and won the contest. I used the slower D500 video at 60 fps to analyze the D850, but for the other tests I used the D850 video at 120 fps. The focus hardware in these cameras is equivalent to the D5 hardware, so it should behave similarly.
The D50, surprisingly, is also quite smart; it went straight to the focus goal, although more sluggishly than the D500 or D850. As the lens got close to infinity, it just coasted at a slow pace to reach the final focus position with no overshoot.
The D610 could move the focus at a reasonably quick pace, but it stopped at 4 meters, back-tracked to 3.5 meters, and then went forward to stop at infinity. Had the camera not back-tracked, it would have had a fast focus time.
The D7000 focus looked very confused. It focused to 10 meters, back-tracked to 5 meters, went forward to 20 meters and stopped, then moved forward beyond infinity, then finally moved backward to get to the correct infinity setting. This extra motion caused the D7000 to finish in an embarrassing last place in the competition, behind the lowly D50.
The D7100 focus moved to 4 meters, back-tracked to 3.8 meters, and then went straight from there to the correct infinity setting. The individual focus motions were quite fast, but the stutter maneuver cost it precious time. I always had the “feeling” that this camera was faster than my D610, but this test proves it.
Summary
Contrary to the expected results, camera age and the processor speed aren’t reliable predictors of how fast a camera can focus a lens. The intelligence of the focus algorithm built into the camera firmware can make a huge difference in focus speed.
I remember when Sigma produced a firmware update for my 150-600 zoom (programmed with their USB Dock). The focus became both faster and more consistent. This indicates that firmware in both the camera and the lens factors into focus behavior. Life’s complicated.
Different lenses may produce different focus behaviors, which is why I decided to use a single lens in all of my testing. I chose a very fast lens (with a fast focus motor), so that it wouldn’t be the limiting factor in determining how fast a camera can focus a lens.
There are of course other factors involved in focus, such as low-light performance, the number of cross-type sensors, frame coverage and the sheer number of sensors. That kind of information is available in an uncountable number of sites on the internet. What you won’t find is information comparing camera focus algorithm cleverness, or the lack thereof.
Someday, camera manufacturers will probably get around to using artificial intelligence. I think one of the big beneficiaries of this technology will be smart focus algorithms combined with smart subject tracking. That will be a great day.
It’s very illuminating to study focus behavior using video. Your own eyes are just too slow to perceive what’s really going on. These test videos reminded me of Aesop’s The Tortoise and the Hare. Slow and steady can win the race, or at least get you the bronze medal.
Comments