Available to internal TestFlight testers. This note does not mean build 2020 is available through the public invitation. TestFlight shows the builds available to your account.
What changed
The phone-motion demo and regular sensor-backed training now share the new Default rep counter. It smooths the force curve with a 2-pole Butterworth filter at 1.4 Hz, then counts an excursion when the signal rises and falls far enough relative to the set’s own range. We call that second step the bathtub counter.
In the demo, the main rep estimate now uses the same strategy as the Range / bathtub → BW 1.4 Hz cell in Signal debug. You no longer need to open the comparison table to see that counter’s result.
Sensor-backed Train sets show the live estimate and save it when you log a set. In Freestyle, you can correct the count with the reps stepper, then choose Use estimate to return to automatic counting. Manual sets still use the count you enter.
Your recorded force curve, peak force, and impulse are unchanged by this filter. Older saved rep counts are left as they were. These are estimates: this release does not promise perfect counts or infer form quality from phone motion.
What to try
- In Train → Try a demo, record ten curls, then ten lateral raises. Try a comfortable slow pace and a faster pace.
- Open Inspect signal after a set. Compare the saved estimate with the bathtub count in the BW 1.4 Hz column. Moving the debug filter slider is a preview; it does not change the fixed counter used to save the set.
- With a Progressor, log repeated sensor-backed sets. Check that the count starts fresh for each set, and try a Freestyle correction.
The counter needs to observe a fall before accepting a rep. Stopping halfway through a cycle, or while the smoothed signal is still catching up, can leave the final rep uncounted. Tell us the actual count and the estimated count when you send feedback.
