Persistence

Persistence is the length of time each pixel on a display remains lit during a single frame. In virtual reality, persistence is one of the main factors that decides how sharp the virtual world looks while the user moves their head. High persistence, where a pixel stays lit for most of the frame, can cause blurring and smearing of the image during motion. A low persistence of about 3 ms or lower is generally required to create a convincing sense of presence in VR. As pixel density increases, lower persistence is needed to avoid blurring.

Full persistence versus low persistence

A "full persistence" display keeps each pixel illuminated for the entire duration of a frame. At a Refresh rate of 60 frames per second, a full persistence pixel stays lit for about 16.7 ms, the full length of the frame. Most ordinary monitors, televisions, and phone screens work this way, holding the same image steady until the next frame replaces it. This sample-and-hold behaviour is fine for a screen that sits still in front of the viewer, but it becomes a problem when the display is strapped to a moving head.

A "low persistence" display lights each pixel for only a short part of the frame, often just 2 to 3 ms, and then goes dark for the rest of the frame. Because the image is shown as a brief flash rather than held for the whole frame, the eye does not have time to smear it. The dark gap between flashes is not consciously noticed when the refresh rate is high enough.

Why low persistence matters in VR

The reason persistence matters so much in VR comes from the way human eyes track motion. When a person turns their head while wearing a headset, the whole virtual scene sweeps across the display, and the eyes rotate smoothly to follow points in the scene. The display, by contrast, only updates in discrete steps, one frame at a time. If a pixel is held lit for the full frame on a full persistence panel, the eye keeps moving while that static image stays in place, so the single frame is dragged across the retina and is perceived as a smear. This effect, sometimes called eye-tracking motion blur or hold-type blur, makes text and edges look soft and unstable exactly when the user is looking around, which is most of the time in VR.

Low persistence fixes this by showing each frame for such a short moment that the eye barely moves during the flash, so the image lands cleanly on the retina instead of being spread across it. The world then stays sharp during head motion. Michael Abrash, who led Valve's VR research before joining Oculus, described persistence as one of the central problems standing between VR and true presence, because blur and judder that follow head movement break the illusion that the virtual world is real and stable. The amount of blur scales directly with persistence: at a motion of 1000 pixels per second, roughly 1 ms of persistence produces about 1 pixel of motion blur, so cutting persistence from 16.7 ms to around 2 ms removes most of the smearing without needing an extremely high frame rate. Low persistence can deliver blur reduction comparable to what a much faster full persistence panel, on the order of 500 Hz, would provide.

How low persistence is achieved

Low persistence is produced by illuminating the display for only a few milliseconds of each frame and keeping it dark the rest of the time. The display still receives a fresh image every frame; the difference is how long that image is allowed to emit light. There are two common ways to time this illumination.

• Global illumination transmits the full image to the panel, waits until every row has been loaded, and then lights the entire screen at once for a brief pulse. The whole frame appears and disappears together, which keeps every part of the image in sync but makes the screen darker because it is lit for only a small slice of the frame.
• Rolling illumination (also called a rolling scan) lights narrow horizontal bands of the screen in sequence as the image data arrives, so the lit region tracks down the panel from top to bottom rather than flashing the whole screen together.

The light source itself depends on the panel technology. OLED panels can switch their own pixels on and off very quickly, which makes them well suited to the sharp, brief flashes that low persistence needs; early VR headsets adopted OLED partly for this reason. LCD panels do not emit their own light, so they achieve low persistence through backlight strobing, rapidly turning the backlight on and off, an approach related to black frame insertion on monitors and televisions. In both cases the effective persistence is set by the duty cycle, the fraction of each frame during which light is actually emitted. VR panels commonly keep this duty cycle below 20 percent to control motion blur.

History

The importance of low persistence for VR was recognised at roughly the same time by two groups: Oculus, in work connected to the display-motion research at Blur Busters, and Valve, through the VR research led by Michael Abrash. Abrash's public writing on judder and persistence helped explain why holding an image for the full frame was incompatible with comfortable head-tracked VR.

Oculus shipped the first widely used low-persistence VR display on the Oculus Rift DK2, announced at the Game Developers Conference in March 2014 and shipped to developers from July 2014. The DK2 replaced the LCD panel of the first development kit with a low-persistence AMOLED panel, derived from a Samsung phone display, running at 1920 by 1080 total (960 by 1080 per eye) and 75 Hz. Contemporary coverage described the change as killing the motion blur and judder that had been among the biggest causes of discomfort in earlier prototypes, and it was treated as a major comfort improvement for the medium. Low persistence quickly became standard across consumer VR. The original Oculus Rift CV1 was reported to hold about 2 ms of persistence, and later headsets pushed it lower still, with the HTC Vive around 1.9 ms and the Valve Index near 0.33 ms.

Trade-offs

Low persistence improves motion clarity but comes with costs. Because the display is lit for only a small fraction of each frame, far less total light reaches the eye, so a low-persistence panel is dimmer than the same panel run at full persistence. Manufacturers compensate by driving the emitters brighter during the short flash, which raises power and thermal demands.

The bigger constraint is flicker. Flashing the screen on and off many times a second can be seen as flicker if the rate is too low, and on a global-illumination panel the whole screen brightens and dims together, which makes any flicker more noticeable. Avoiding visible flicker requires a high Refresh rate, high enough that the flashes blend together above the eye's flicker fusion threshold. This is one reason VR headsets target refresh rates of 90 Hz and above rather than the 60 Hz common on desktop monitors, and it ties persistence closely to refresh rate: the brief illumination removes motion blur, while the high refresh rate keeps that illumination from looking like flicker. Persistence is distinct from Latency, which measures the delay between head movement and the matching update on screen; both must be kept low for comfortable VR, but they address different parts of the motion-to-photons pipeline.