Judder
More actions
Judder is visibly uneven or repeated-image motion in a head-mounted display. In virtual reality and augmented reality, it can appear when the viewer turns their head or follows a moving object. The term covers related temporal artifacts: discrete motion samples can produce smearing and strobing, while late or repeated application frames can make objects appear to vibrate or jump. These effects reduce visual stability and can contribute to discomfort associated with simulator sickness.[1][2]
Display persistence and repeated frames
Display refresh rate, application frame rate and display persistence describe different parts of image delivery. Refresh rate measures display updates, frame rate measures newly rendered images, and persistence measures how long an image emits light within each refresh interval. A high-refresh display can still show repeated content if the application or source video updates less often.[3]
During smooth pursuit, the eyes follow a target continuously while the displayed image remains fixed within each frame. Light from that fixed image moves across the retina, producing motion blur. A low-persistence display shortens the emission interval to reduce this blur. Short illumination does not supply missing motion samples: repeated images can still appear as distinct copies during eye movement. Low persistence and timely content updates therefore address different causes of degraded motion.[4]
| Artifact | Characteristic appearance or cause |
|---|---|
| Persistence blur | A tracked object smears while its image remains illuminated during eye movement.[3] |
| Stroboscopic or multiple-image motion | Discrete or repeated images appear at separate retinal positions rather than forming continuous motion.[1][5] |
| Missed-frame judder | An application misses its delivery deadline, so the compositor uses older content and object positions jump between updates.[2] |
How to reduce judder
Increasing refresh rate can reduce temporal artifacts, and shortening illumination reduces persistence blur. Neither change guarantees artifact-free motion when content updates are missing. The benefit depends on the display's emission pattern and the motion being viewed.[3]
Consistent frame delivery matters alongside the average frame rate. Runtime scheduling balances delivering a frame before its display deadline against producing it so early that it waits in a queue and increases motion-to-photon latency. Scene transitions, changing workloads and thermal throttling can disturb that timing.[6]
For a fixed-refresh application targeting a new frame every refresh, the approximate interval is:[2]
| Display refresh rate | Interval between refreshes |
|---|---|
| 72 Hz | 13.9 ms |
| 90 Hz | 11.1 ms |
| 120 Hz | 8.3 ms |
A higher target leaves less time to complete each frame. Reducing rendering cost, using dynamic resolution or foveated rendering, and choosing a refresh rate the application can sustain can help avoid missed frames. Profiling individual CPU/GPU frame times and stale-frame counts is more informative than relying on an average frame-rate number alone.[2][6]
Timewarp and judder
Timewarp adjusts an already rendered image using a newer estimate of head pose. In orientation-only timewarp, the adjustment compensates for head rotation after rendering. Asynchronous timewarp performs this work separately from application rendering, allowing the latest completed image to be warped for a display refresh even when the application misses a frame.[5] This reduces rotational judder and perceived latency without rendering the scene again.[7]
Rotational correction does not reconstruct the parallax caused by positional or translational head movement. Positional judder is particularly noticeable around nearby geometry. Moving and animated objects also remain at their old positions in a warped frame; their judder depends on their speed, number and projected size.[5]
Positional timewarp uses depth information to adjust for translation.[7] It can expose regions that were hidden in the original image, leaving the reprojection without image data for those regions.[5] Motion-aware techniques go further: Meta's Application SpaceWarp uses submitted depth and motion vectors to synthesize intermediate frames accounting for head and object movement. These are distinct capabilities from orientation-only timewarp, with application integration requirements.[7][2]
Reprojection works alongside runtime scheduling, which must still handle changing workloads and avoid consecutive stale frames.[6]
Research
A 2019 study by Saeed Mahmoudpour and Peter Schelkens examined judder and compression in 360-degree videos, using 30-fps source sequences viewed on a 90-Hz Oculus Rift. Nineteen participants remained after outlier rejection. Tracking faster objects reduced perceived quality more strongly in high-quality video, where compression artifacts were less dominant. The experiment concerned those videos and viewing conditions, rather than a universal threshold for every headset or interactive scene.[1]
A 2022 study by Chumin Zhao and colleagues measured and modeled three Vive headsets. It found benefits from short-pulse emission and demonstrated multiple-image artifacts when content updated more slowly than the display. Its analysis assumed ideal smooth pursuit and excluded system tracking/rendering latency; it was a display-performance study rather than a test of simulator-sickness rates.[3]
References
- ↑ 1.0 1.1 1.2 "Visual Quality Analysis of Judder Effect on Head Mounted Displays". 2019 27th European Signal Processing Conference (EUSIPCO). 2019. https://eurasip.org/Proceedings/Eusipco/eusipco2019/Proceedings/papers/1570532689.pdf. Retrieved 2026-10-18.
- ↑ 2.0 2.1 2.2 2.3 2.4 "Missed frames and frame recovery". Meta Horizon OS Developers. 2026-03-09. https://developers.meta.com/vr/documentation/unity/os-missed-frames/. Retrieved 2026-10-18.
- ↑ 3.0 3.1 3.2 3.3 "Spatiotemporal image quality of virtual reality head mounted displays". Scientific Reports. 2022-11-24. https://www.nature.com/articles/s41598-022-24345-9. Retrieved 2026-10-18.
- ↑ "A perceptual model of motion quality for rendering with adaptive refresh-rate and resolution". ACM Transactions on Graphics. 2020-07. https://www.cl.cam.ac.uk/~rkm38/pdfs/denes2020_motion_adaptive_rendering.pdf. Retrieved 2026-10-18.
- ↑ 5.0 5.1 5.2 5.3 "Asynchronous Timewarp Examined". Oculus Developer Blog. 2015-03-02. https://developers.meta.com/vr/blog/asynchronous-timewarp-examined/. Retrieved 2026-10-18.
- ↑ 6.0 6.1 6.2 "FrameSync". Meta Horizon OS Developers. 2026-05-13. https://developers.meta.com/vr/essentials/framesync/. Retrieved 2026-10-18.
- ↑ 7.0 7.1 7.2 "The compositor". Meta Horizon OS Developers. 2026-08-28. https://developers.meta.com/vr/essentials/the-compositor/. Retrieved 2026-10-18.