Strobing

Strobing is the perception of multiple copies of an image at the same time in VR. Strobing along with smearing is part of judder. These effects reduce the visual quality of a virtual reality display, and Michael Abrash, then at Valve, suggested that they may also contribute to eye fatigue or motion sickness, a point he described as speculative. Strobing becomes more apparent when smearing is eliminated through the use of a low persistence display (smearing no longer hides strobing).

Strobing can be reduced by increasing the refresh rate of the display, because a higher refresh rate shortens the distance an image moves between successive frames. Removing it completely this way requires very high rates: in tests on an experimental 480 Hz monitor, Blur Busters found that even 480 frames per second at 480 Hz did not remove all stroboscopic effects. In vision science and display research the same kind of artifact is described as "multiple imaging", a form of temporal aliasing, and when it occurs during eye movements across a flickering light source or a low-duty-cycle display it is called the phantom array effect.

Reviewed 27 September 2026. Claims checked against the cited Abrash blog posts, papers, Oculus patent, Road to VR DK2 specs and Blur Busters article. About review dates.

How it works

A display does not show continuous motion. Each pixel can change only once per frame, so a moving object is drawn at a series of discrete positions rather than sliding smoothly. Abrash, writing on his Valve blog in 2013, explained that if an image moves more than about 5 to 10 arc minutes between successive updates, a viewer may see several copies of it at once. He added that whether strobing actually appears, and at what separation, depends heavily on the content.

The motion that matters is the motion of the image relative to the eye. On a head-mounted display the eye often moves relative to the screen, for example while it holds a virtual object steady during a head turn, so large frame-to-frame jumps happen easily. Abrash calculated that at 60 Hz successive frames of an image land 5 arc minutes apart when the eyes move at only 5 degrees per second relative to the image, and 10 arc minutes apart at 10 degrees per second. He compared this with a leisurely head turn, which he put in the region of 100 degrees per second. Strobing is more prominent in images that contain high spatial frequencies (fine detail and sharp edges).

Laboratory work on moving bars points in the same direction. In a 1995 study using a cathode ray oscilloscope, Bex, Edgar and Smith found that multiple images appeared once a fixed spatial displacement between image updates was exceeded, and that this critical displacement was approximately the same at every drift speed. The authors concluded that multiple imaging occurs when the visual system's motion energy detection fails, and that it could be reduced by raising the image update rate until the displacement per update stays below that limit, or by filtering out high spatial frequencies. Abrash cited this paper as the reason his team at Valve expected to run into strobing before it built any low-persistence prototypes.

Abrash offered two partial explanations for why an image that the eye is following can strobe rather than simply blur. Variation in latency, errors in prediction and inaccurate tracking all place edges at slightly different spots on the retina from frame to frame. The eye's temporal integration period may also not match the persistence time, so a single lit frame can be registered more than once; he labeled this second idea a theory. He also noted that the edges of wider features can flicker (an on and off effect) rather than strobe (a multiple-copies effect), because the eye detects flicker more readily at lower intensities.

Relationship to judder and persistence

Abrash used "judder" for the combination of smearing and strobing that appears on VR and AR headsets when the eye moves relative to the display. While a pixel is lit (the persistence time) it sweeps across the retina and produces a smear. When the next frame arrives, the image jumps back to its correct position, and that jump produces strobing.

On full-persistence displays, where pixels stay lit for the whole frame, strobing is usually not very visible, because smearing hides most of it and because it is less prominent in images without much fine detail. The slow LCD switching times of the early Oculus Rift development kits blurred images further, which masked strobing but lost even more detail.

Cutting persistence removes the smear. Abrash estimated that persistence below roughly 2 ms should almost completely eliminate the smear component of judder, and said experimental prototypes confirmed this. Without the smear, strobing that had been invisible becomes an issue on low-persistence displays. An image the eye is tracking does not strobe, because its pixels land on the same place on the retina each frame. The rest of the scene can strobe, since the eye is moving relative to those pixels. Abrash thought this background strobing might often go unnoticed, but called that a hypothesis his team had not tested. Perfect tracking and a consistent frame rate are assumed here; tracking error or a variable frame rate can misregister the tracked image enough to make it strobe as well. Replying to a reader on the same post, Abrash wrote that on a low-persistence display frames cannot be repeated without strobing, and that any significant variation in frame time would cause variable problems with flicker and strobing.

Mark Rejhon of Blur Busters described the same trade-off from the display side: on impulse-driven displays such as low-persistence VR headsets, the stroboscopic effect becomes more visible as resolution and brightness go up and persistence goes down.

Saccades and visual instability

Strobing is most obvious during saccades, the fast eye movements that shift gaze from one point to another. Abrash reported that when a viewer looks at a row of thin green vertical bars on a low-persistence display and saccades left or right, multiple copies of each line appear. He attributed this in part to the failure of saccadic masking: normally the image of the world smears across the retina during a saccade and the brain suppresses it, but short, sharp low-persistence images can get through. At saccadic eye velocities of several hundred degrees per second, strobing is what would be expected if that suppression fails.

Abrash's team also found a stranger effect that they called visual instability. In a prototype game room textured with a grid of thin, bright green lines, team member Atman Binstock stood near a wall, looked into a corner and quickly shifted his gaze to the middle of the wall; the whole room seemed to shift or turn. Everyone who tried it felt the motion, although the room had not moved. When the team switched the prototype to full persistence, the instability disappeared, and they could produce a similar effect in the real world with a strobe light. Abrash noted that the effect was less pronounced or undetectable with normal game content.

A patent filed by Oculus VR in December 2015 and granted in April 2018 describes the same problem. It states that low persistence of 10% to 20% is typically used at moderate refresh rates of 75 to 120 Hz to avoid blur during head rotation, and that a side effect is strobing, in which multiple copies of a visual object are perceived simultaneously, most noticeably during saccades. During a saccade, according to the patent, copies of objects from different video frames are "imprinted" on the retina, and the brain can fail to match objects to themselves between frames.

History

Abrash's June and July 2013 posts on the Valve blog set out why judder appears on head-mounted displays and what new problems, strobing among them, low persistence would bring. Low persistence reached developers the following year. The Oculus Rift DK2, announced at GDC on 19 March 2014 for US$350 with shipping planned for July 2014, used a 1080p OLED display with low-persistence technology. Its listed refresh rates were 75, 72 and 60 Hz, with persistence of 2 ms, 3 ms or full.

Later consumer headsets pushed persistence lower. In 2022 measurements by researchers at the U.S. Food and Drug Administration's Center for Devices and Radiological Health, the HTC Vive and HTC Vive Pro had an emission duty cycle of 17% (their OLED panels are lit for that fraction of each frame), while the HTC Vive Pro 2 ran its LCD at 120 Hz with a 5% duty cycle, an emission time of 0.42 ms per frame. The study concluded that a duty cycle under 20% helps reduce motion blur, and that once a headset uses such short pulses, raising the refresh rate from 90 to 120 Hz does little more for smooth-pursuit blur. That study measured motion blur, not strobing; Abrash had described strobing as a fairly predictable consequence of low persistence.

Mitigation

No single fix removes strobing without cost. The main approaches discussed by Abrash, Oculus and Blur Busters are:

Approach	How it works	Drawbacks noted by the sources
Higher refresh rate and frame rate	Shortens inter-frame time, so an image moves less between frames. Doubling the frame rate halves the stroboscopic stepping distance for a given motion speed.	Abrash put the sweet spot for 1080p at a 90 degree field of view somewhere between 300 and 1000 Hz for judder generally, far beyond consumer parts in 2013. Blur Busters found 480 Hz still insufficient to remove all stroboscopic effects.
Rendered motion blur	Renders moving objects with motion blur so images in successive frames are not far enough apart to strobe.	Needs extra rendering, the motion of each object relative to the eye, and eye tracking; Abrash said it was unclear whether the benefits would outweigh the costs. Blur Busters noted that added blur goes beyond natural vision and some readers report headaches from it.
Eye-tracked blur	Uses eye tracking to blur only objects that are not moving with the eye's gaze.	Eye-tracker latency and rendering overhead can still leave lag artifacts.
Persistence switching during saccades	An eye tracker detects a fast saccade and the headset switches to a higher-persistence display mode for its duration, then returns to low persistence. In the described embodiments the low mode is 10 to 20 percent and the high mode at least 50 percent.	Patent filing by Oculus VR; the patent says using higher persistence only during the saccade reduces both strobing and blur.
Content design	Authoring content that avoids high-frequency detail, which Abrash suggested for the related visual instability effect.	Abrash said it was not clear exactly what would be needed to make this work well.

Aaron Nicholls, who worked on the prototypes with Abrash, wrote in a reply on the blog that the team had tried rendering pairs of subframes 1 ms apart within each 60 Hz frame, hoping that extra motion signals would reduce the visibility of strobing. He reported that motion looked better with evenly spaced frames, with 120 Hz looking better than back-to-back pairs of subframes at 60 Hz.

Research

Researchers at Facebook Reality Labs measured the effect directly. In a 2019 paper in the SID Symposium Digest of Technical Papers, T. Scott Murdison and colleagues reported a pair of psychophysical experiments quantifying flicker and eye movement-induced ghosting at the refresh rates and persistence levels used in head-mounted displays, and made general recommendations for HMD design.

Goettker, MacKenzie and Murdison followed this in 2020 with a high-speed, AR-like head-mounted display prototype, comparing the phantom array artifacts of low duty cycles with the motion blur of high duty cycles. Detection of phantom arrays varied widely between people, but low duty cycles consistently disturbed where saccades landed, and this error went with more mistakes on a perceptual discrimination task. High duty cycles made motion blur during head movements more likely to be seen, especially at lower refresh rates. The display's temporal settings had no measurable effect on compensatory eye movements such as the vestibulo-ocular reflex.

A 2022 follow-up by Xiuyun Wu and Murdison at Reality Labs tested reading in a saccadic rapid serial visual presentation task and found no effect of persistence on reading performance or eye movement metrics. Outside VR hardware, a 2023 study of flickering light sources found that phantom array visibility increased with saccade speed and was higher at a wide viewing angle; its authors wrote that reducing the phantom array effect is important in AR and VR head-mounted displays and that eye movement speed should be considered in their development.

See also

• Judder
• Smearing
• Low persistence
• Pixel persistence
• Refresh rate
• Saccade
• Motion-to-photon latency
• Simulator sickness