Skip to content
Shane Saxon edited this page Apr 1, 2019 · 5 revisions

The Mysterious World of Latency!!!

There is NO consensus as to what constitutes acceptable latency. At this time, we are NOT looking to start a 'Bruce Lee' style argument, but merely mention observations that have been reported to us.

Lower is better, but what is acceptable is specific to the situation. Such as band playing together(<20ms), or when the dancer appears to moves in step to the music (40-80ms). Human perception varies with the individual, the specific senses (sight, sound, touch...) and the environment (mobile, cinema, VR, AR, mixed reality). Now consider:

  • Sound travels at 343 m/s (sea-level) = 5.7 meters per video frame (59.94Hz).
  • VR studies (in Japan) found that with head movement we can discern 1ms.
  • Haptic (touch) is also about 1ms.
  • Reaction time ranges with age from about 1/4 sec to 2 secs.
  • You can test your hand eye coordination HERE!

We use a psuedo-state-machine approach to keep latency the lowest possible within the limits of the computers architecture (hardware and OS). Video tends to inherently have the greatest delay:

  • Camera's at 59.94Hz adds 1 frame = 16.7ms
  • Video input to GPU output without sync is 3-4frames = 50-67ms
  • Video input to GPU output with GPUdirect & SYNC is 2-3frames = 34-50ms
  • Screens add another frame + response delay for about 17-22ms
  • So (best case at 60P) is: 17ms(cam) + 34-67ms(PC) + 17ms(screen) = 67-100ms

Minimizing (Perceptual) Latency:

  • Hardware with SYNC brings video latency to minimum possible.
  • VR can anticipate head movement (based on physics.)
  • Software can run within the limits of the OS event handling jitter = 1-20ms typical.

Clone this wiki locally