Characterizing Latency in Touch and Button-Equipped Interactive Systems

Abstract

We present a low cost method to measure and characterize the end-to-end latency when using a touch system (tap latency) or an input device equipped with a physical button. Our method relies on a vibration sensor attached to a finger and a photo-diode to detect the screen response. Both are connected to a micro-controller connected to a host computer using a low-latency USB communication protocol in order to combine software and hardware probes to help determine where the latency comes from. We present the operating principle of our method before investigating the main sources of latency in several systems. We show that most of the latency originates from the display side. Our method can help application designers characterize and troubleshoot latency on a wide range of interactive systems.

Using High Frequency Accelerometer and Mouse to Compensate for End-to-end Latency in Indirect Interaction

“…A wireless mouse controller gave 102.9 ± 3.3 ms for the end-to-end latency of their system. Input latency varies even more on touch based devices, varying from 50 to 200ms depending on the hardware [10,16,30].…”

“…We measured the end-to-end latency in our application using the method described by Casiez et al [10]. We performed 190 measures for each artificial latency and found 40.5 ms (SD 7.0 ms) for 0 ms artificial latency, 74.1 ms (SD 6.8 ms) for 33.3 ms artificial latency and 108.5 ms (SD 7.6 ms) for 66.6 ms artificial latency, showing our artificial latency mechanics worked as expected.…”

“…This end-to-end latency can be split into hardware latency (on input side: sensing, transmission and on output side : display) and software latency (system, toolkits, frameworks and applications). End-to-end latency is still affecting most interactive systems [9,27,30], from desktop computers (where latency is between 50 and 90 ms [9]) to touch-based devices (typically, latency on touch-screens is between 60 and 200 ms [10,30]). And while modern input controllers (e.g.…”

Profiling Distributed Virtual Environments by Tracing Causality

“…The ability to unify heterogeneous data sources is important for distributed systems, both because some components may not support modification, and also because it is necessary to instrument the real world. Kämäräinen et al [14] and Casiez et al [4] also unified sensor data with internal samples, but their implementations were more tightly coupled than ours.…”

“…Kämäräinen et al unified samples from these sensors with those from internal software events by synchronizing their platform's clock with that of the mobile device using a protocol similar to NTP. Casiez et al [4] combined hardware probes and with software time-stamping in a similar manner, but relied on a low latency USB interface for their sensors, rather than clock synchronisation.…”