World to skin, stage by stage, modeled. Five stages from the sensor's exposure to the skin's perception, each a modeled range. The bars are drawn at the upper bound of each range, because that is how the record forms its totals: 25 + 5 + 5 + 20 + 15 is about 70 milliseconds for a hazard ahead at 60 Hz, and 66 + 10 + 6 + 20 + 15 is 117, which the record rounds to about 120 for the ground at 15 Hz. The motor is the pacing item, so its rise and brake time are measured first. When the bench number exists, every modeled phrase on the site switches to measured through one config value. Source: docs/14.
Stage
Ahead, modeled
Ground, modeled
How it will be measured
Sensor exposure and integration
16 to 25 ms (60 Hz)
33 to 66 ms (15 Hz)
logic analyzer on the frame ready interrupt
Readout over I2C or UART
2 to 5 ms
5 to 10 ms
bus capture
Fusion and three laws on RP2350
2 to 5 ms
3 to 6 ms
GPIO toggle around the loop, oscilloscope
Driver write and LRA rise
10 to 20 ms
10 to 20 ms
accelerometer on the pod
Skin perception
5 to 15 ms
5 to 15 ms
reaction time psychophysics
Total, modeled
about 70 ms
about 120 ms
high speed camera plus pod accelerometer
The motor is the pacing item. Measure its rise and brake time first. Every "modeled" phrase on the site switches to "measured" through one config value when the bench number exists.
Source: docs/14-latency-budget.md, the engineering record. Every number on this page is read from that file. If a fact is not in the record, it is measured on the bench, or it stays off this site.