Phase-detection and contrast-detection autofocus estimate focus in different ways. Phase detection can estimate the direction of focus error, while contrast detection evaluates image contrast as focus changes. Hybrid systems combine these approaches or integrate phase-detection measurements on the image sensor. Their evolution is best understood through named camera implementations rather than a general claim that one method simply replaced the other.
For a detailed explanation of the underlying phase-detection process, see phase-detection autofocus.
How the methods differ
Dedicated phase-detection autofocus historically used a separate AF module in many DSLRs. On-sensor phase detection allows focus measurements to be made through the imaging sensor, supporting autofocus during live view and video. Contrast detection can be accurate but may need to search through focus positions to determine the peak contrast. Modern systems combine sensor readout, subject detection, prediction and lens communication.
For a deeper look at modern algorithmic tracking and subject recognition, see autofocus algorithms.
Documented milestones
- 2013 – Canon EOS 70D: Canon announced Dual Pixel CMOS AF, an on-sensor phase-detection technology for Live View and video. Canon described the EOS 70D as the first EOS camera to feature the technology. See Canon’s 2 July 2013 announcement.
- 2017–2019 – Sony α9 firmware: Sony’s α9 firmware 5.00 in March 2019 added Real-time Tracking and updates to Eye AF and phase-detection behaviour. This is a documented firmware milestone, not the origin of all AI autofocus. See Sony α9 firmware history.
- 2020 – Canon EOS R5/R6: Canon announced Dual Pixel CMOS AF II and deep-learning-based subject tracking in the EOS R5 and R6. See Canon’s EOS R5/R6 announcement.
- 2021 – Nikon Z 9: Nikon announced a stacked CMOS sensor and EXPEED 7 processing with deep-learning subject detection for nine subject types and 3D tracking. See Nikon’s Z 9 announcement.
Why chronology does not prove superiority
These examples illustrate distinct implementations across manufacturers. They do not provide a matched autofocus performance test. AF results depend on lens motor, subject, lighting, camera settings, firmware and the definition of a successful focus hit.
For a current explanation of AI-assisted subject detection and tracking, see AI autofocus systems.
How to benchmark a modern AF system
- Use named camera bodies, lenses and firmware versions.
- Define the subject path, movement speed, lighting and occlusion conditions.
- Measure acquisition time, tracking hit rate, focus error and reacquisition time.
- Repeat across subject classes and lens combinations.
- Publish the sample size and failure definition.
Conclusion
Autofocus innovation has moved from separate phase-detection modules toward increasingly integrated on-sensor and subject-aware systems. A credible history identifies dated implementations, while claims about speed or accuracy require controlled tests.