Latency appears as separation between nib and ink
Aliases: ink gap · following gap · ink lag
What it is
On the screen, inking latency is not an abstract millisecond. It is a gap: the nib is already somewhere, the ink is still behind. People read that gap as “the ink is dragging, not following”, not as a number. Parallax from glass thickness can also put ink off the nib, but that gap does not lengthen with speed and does not close on a hard stop—it is not the same seam.
Why it happens
Ink drawn at time t corresponds to the pen at t−Δ. The nib’s visual place is now; the ink front is the past; on the plane they separate by Δ times speed. At uniform speed the seam is a steady blank; it lengthens under acceleration; on deceleration or a stop it briefly closes, and ink may “catch up” to the nib. That breathing with speed is latency’s signature. A parallax seam comes from eye, nib, and emissive layer not being collinear; it changes with viewing angle, not hand speed, and remains after a stop. Treating the two seams as one picks the wrong tool: latency wants a shorter pipeline or (cautious) extrapolation; parallax wants geometric calibration. Occlusion also matters: an opaque nib hiding the join makes people underestimate delay; a transparent nib or a side view brings the seam back. Refresh tearing makes the seam’s edge flicker, which feels more “loose” than the same mean delay.
Studying it
Overlay a nib marker and the ink front on high-speed video, measure gap length per frame, and log speed. Have people write head-on and at an angle, and rate “is there a gap” on stops, uniform motion, and acceleration.
Independent variables: injected delay, writing speed, viewing angle, nib transparency. Dependent measures: gap–speed slope (latency should be near linear), whether the gap closes after a stop, correlation with viewing angle (parallax should correlate, latency should not).
A gap with near-zero slope is more like parallax. A gap that will not close on a stop is also like parallax.
Where it stops holding
In very slow writing the latency gap shrinks inside the nib width; people report “no delay” while measurement may still show tens of milliseconds. Projection, camera preview, or compressed mirroring can enlarge the seam with codec blocks. Predictive fill can close the gap and then paint the wrong place on a sharp turn—that is a barb, not a vanished seam. A coloured nib or a fat brush hides the join inside the footprint; the subjective gap shrinks, Δ does not.
Applying it
- When diagnosing following, watch whether the gap lengthens with speed and closes on a stop, to separate latency from parallax.
- Demo and accept with a thin or transparent nib, once head-on and once at an angle, so occlusion cannot hide the seam.
- Do not replace measuring Δ with “it looks glued”; glue may be prediction or a fat brush.
- How to check: a uniform fast stroke, measure the gap; double the speed and the gap should roughly double. On a hard stop it should close. Changing viewing angle should not move the length much; if it does, do geometric calibration before piling on prediction.