Loop latency (feedback delay)

Loop latency is the total time from a neural event to the moment its consequence appears in the sensory feedback the user perceives. It accumulates across the pipeline: acquisition and buffering, the group delay of causal filters, feature computation, the decode step, and rendering of the effector. In closed-loop BCIs this end-to-end delay commonly runs from tens of milliseconds up to roughly a hundred or more, depending on window lengths and smoothing.

From a control standpoint delay is costly because it erodes phase margin: a controller cannot react to what it cannot yet see, so increasing loop delay forces the user to lower effective gain and adopt slower, more cautious, sometimes move-and-wait control, and beyond a threshold it produces overshoot and oscillation. This creates a direct tension with denoising, since the averaging and smoothing that suppress decoder noise also add lag. Good closed-loop design treats latency as a first-class budget to be minimized and traded explicitly against smoothing, not as an afterthought.