Phase Detection vs. Contrast Detection Autofocus: How On-Sensor Microlenses Dictate Speed and Accuracy

For an autofocus system to lock onto a subject, the camera must answer two physical questions: which direction to move the lens elements, and exactly how far to drive them. The binding constraint is light. The system must have enough illumination to either measure the phase difference between two incoming beams or evaluate the contrast of adjacent pixels. When light drops below a critical threshold, the mathematical certainty of phase detection breaks down, forcing the camera to fall back on trial-and-error contrast evaluation. In 2026, the speed at which a camera answers those two questions dictates whether a fast-moving subject is captured sharply or lost to motion blur.[2][6]

Phase Detection Autofocus (PDAF) operates as a distance calculator, utilizing dedicated pixels on the image sensor to detect phase differences in incoming light. The system splits the incoming light through specialized microlenses, directing the beams to left-looking and right-looking pixel halves. By comparing the parallax between these two offset images, the camera's image signal processor calculates the exact distance to the subject in a single measurement step. Because PDAF determines both the focus error and the correction direction simultaneously, it achieves focus lock without requiring the lens to sweep through various positions.[1][2]

This predictive capability makes PDAF exceptionally fast. Advanced systems can achieve focus in 0.1 to 0.3 seconds, which is fast enough to track erratic movement in sports or wildlife photography. During continuous autofocus tracking, the system updates phase difference measurements 30 to 60 times per second, adjusting the lens actuator in real time. This speed advantage is particularly critical in dynamic environments, such as unmanned aerial vehicles or robotics, where a 13-megapixel sensor like the Sony IMX258 relies on PDAF to eliminate focus hunting on moving subjects.[6][7]

Contrast Detection Autofocus (CDAF), by comparison, operates as a sharpness checker rather than a distance calculator. It does not require specialized half-masked pixels; instead, it analyzes the contrast of the captured image directly from the standard sensor pixels. The system operates on the principle that an image is sharpest when the contrast between adjacent pixels is at its maximum. To find that peak, the camera must move the lens elements step by step, continually evaluating whether the contrast is increasing or decreasing.[1][5]

Because CDAF relies on trial-and-error adjustments, it must intentionally drive the lens past the point of perfect focus and then reverse direction—a process known as hunting. This makes CDAF inherently slower than PDAF. However, because it evaluates the actual image data hitting the sensor, it is highly accurate for static subjects. In controlled environments, such as medical imaging or document scanning kiosks, CDAF ensures maximum sharpness without the risk of front-focus or back-focus errors that can occasionally plague phase detection systems.[2][3]

The transition from digital single-lens reflex (DSLR) cameras to mirrorless architectures fundamentally changed how these systems are deployed. Traditional DSLRs relied on a separate autofocus module located in the floor of the camera, utilizing a secondary mirror that diverted light away from the main sensor. This split path reduced the light reaching the dedicated phase array by roughly 50 percent and introduced mechanical alignment errors. Mirrorless cameras embed the phase-detection pixels directly into the main image sensor, eliminating the secondary mirror and allowing autofocus coverage to extend across the entire frame.[4]

Removing the mirror box also granted optical engineers new freedoms. Nikon's Z mount, for instance, utilizes a 55-millimeter inner diameter and a remarkably short 16-millimeter flange focal distance. This geometry allows for lenses with maximum apertures as wide as f/0.95, flooding the sensor with light. Because the autofocus system reads the scene exactly the same way the exposure system does, mirrorless cameras can perform real-time face and eye detection, locking onto a subject's pupil and tracking it flawlessly even when obstacles temporarily obscure the view.[4]

As engineers at Vadzo Imaging note regarding embedded systems, "Selecting the wrong autofocus technology does not fail in the lab. It fails in the field." To maximize performance across all lighting conditions, modern cameras rely on Hybrid Autofocus. This approach combines the rapid acquisition of PDAF with the pinpoint accuracy of CDAF. For example, a camera like the Sony a6500 utilizes 425 phase detection points to handle the heavy lifting—rapidly driving the lens to the correct ballpark—before handing off to 169 contrast detection points for final fine-tuning. This synergy ensures that the camera inherits the strengths of both technologies, using phase detection for speed and contrast detection to guarantee the subject lands exactly in focus.[5][6]