In February 2025, at a friend's birthday party, I ran a test that finally made me understand autofocus.
I had a Nikon Z7 II with the Nikkor Z 50mm f/1.8, and a Canon R6 Mark II with the RF 50mm f/1.8. Same focal length, same aperture, same scene — a dimly lit living room full of moving kids and adults. I shot the entire evening with both cameras, switching between them every few minutes.
When I reviewed the images, the numbers were clear:
- Canon R6 II: about 87% of shots in focus. Maybe 75% critically sharp.
- Nikon Z7 II: about 70% in focus. Only 30-40% critically sharp.
The Nikon lens was sharper optically — the few shots that nailed focus were stunning. But the Canon system delivered consistently. And consistency is what matters when you're trying to capture a fleeting expression at a birthday party.
What surprised me most was the lens factor. The Nikon 50mm f/1.8 has ultrasonic AF (SWM), traditionally considered faster than stepping motors. The Canon 50mm f/1.8 STM uses a stepping motor. By all conventional logic, the Nikon should have had the advantage. It didn't.
Not all phase detection is equal
Every modern mirrorless camera uses "phase detection autofocus." Marketing materials love to advertise "693 AF points" or "1053 AF points." But these numbers hide a fundamental difference in how the technology actually works.
Sony and Nikon use a traditional approach: they scatter dedicated autofocus pixels across the sensor. These pixels are masked so they only see light from one side of the lens — by comparing pairs of them, the camera can calculate how far the lens needs to move to achieve focus. It's fast and proven. But those dedicated pixels have a cost — they don't contribute to the final image. When Sony says "693 phase detection points," it means 693 pixels that exist purely for focusing. The camera has to interpolate the missing image data at those positions.
Out of millions of total pixels, 693 is a tiny fraction — about 0.01%. You'll never see the difference in a photo. But it means the camera has a relatively sparse grid of focus information to work with.
Canon did something different in 2013. Their Dual Pixel AF (DPAF) splits every single pixel on the sensor in half. Each half sees light from a slightly different angle through the lens. During autofocus, the camera compares the two halves to calculate focus distance. During capture, both halves combine to record the full image. No pixel is wasted.
This means on a Canon sensor, every pixel is an autofocus pixel. Instead of 693 scattered focus points, you effectively have tens of millions. The camera has a far denser map of focus information covering the entire frame.
Where it matters
The density advantage shows up most in challenging situations:
Fast-moving subjects in unpredictable directions. With more AF data points, Canon's system can track subjects with less guesswork. At the birthday party, kids running around a dim room — this is where DPAF has a clear edge.
Advanced subject recognition. When the camera tries to identify and track eyes, faces, or animals, having focus data from every pixel gives the algorithms much more to work with. In my testing, Canon's advanced AF modes (eye tracking, subject recognition) simply worked more reliably than Nikon's. On the Nikon, I often had to fall back to simple spot AF to get consistent results.
Low light autofocus accuracy. This is interesting because it's not straightforward. Nikon's Starlight mode can detect down to -8.5 EV — more sensitive than Canon's -5 EV on paper. In practice, Nikon's system was better at detecting that something was there in the dark, but Canon was more accurate in nailing the focus when it locked on. The Nikon would sometimes confirm focus on a slightly wrong plane.
Where it doesn't matter
Landscape photography on a tripod. When you're composing a shot carefully, using live view, and focusing on a mountain that isn't moving — autofocus technology is almost entirely irrelevant. Manual focus works fine. Even contrast detection works fine. The whole DPAF vs scattered pixel debate becomes academic.
Controlled studio work. Same story. Plenty of light, stationary subjects, time to adjust — autofocus speed and tracking don't matter.
Video on a gimbal. Here autofocus matters again, but differently. Continuous AF smoothness is more important than speed, and honestly both Canon and Nikon (and Sony) have gotten good enough for most video work.
The long exposure plot twist
Here's something nobody talks about: DPAF has a side effect for long exposures.
Because every pixel on the sensor is doing double duty — measuring focus and recording light — the sensor generates more heat during continuous operation. More heat means more hot pixels. More hot pixels mean more colored noise dots in shadows during long exposures.
I wrote about discovering this in my main camera comparison post. It was the Canon R5's biggest weakness for the kind of photography I do. The same technology behind Canon's great autofocus was working against me during 30-second to 3-minute exposures.
Canon's long exposure noise is manageable for shorter exposures or when you don't push shadows hard. Capture One handles the hot pixel removal fairly well. But compared to the Nikon Z8, which produced virtually clean shadows even at extreme push levels — it's a real difference.
What I wish someone had told me
Before I spent months buying and selling cameras, I wish someone had said this:
Figure out what you actually shoot, then optimize for that.
If 80% of your photography is landscape on a tripod, optimizing for autofocus performance is backwards. You should optimize for sensor quality, dynamic range, lens selection, and filter compatibility. The autofocus in any modern camera — even a budget Nikon Z5 — is more than good enough for composed landscape shots.
If 80% of your photography is events, portraits, or sports — then yes, Canon's autofocus advantage is real. It's not marketing. It's a fundamentally different technology that produces measurably better results in challenging conditions. Sony's autofocus too is noticeably better than Nikon I assume, as I see is widely praised online, despite being technologically similar to Nikon.
The problem is that most of us shoot a mix. We want amazing landscape files and also sharp birthday photos. The camera industry wants you to believe one system can be the best at everything. It can't.
I eventually stopped chasing autofocus and focused on the parts that mattered more for my photography: sensor quality, colors, and choosing lenses based on practical field use. But I'm glad I understand the difference now, because it helps me know when the autofocus does matter — and when I can safely ignore it.