Precision Bluetooth Timing on ESP32-S3: Causal Clock Discipline and Hardware Event Capture
How does internal causal clock quality compare with an independently observed relative PPS result?
Overview
The completed 660-second ESP32-S3 experiment measured 166.590 ns relative-PPS RMS across 657 eligible records, with three invalid and zero nominal missing. A fresh N30 second-board transfer measured 150.954 ns RMS across 48 eligible of 64 records.
What this paper covers
- Separate internal prediction RMS from delivered relative-PPS error.
- Keep calibration-selected data distinct from fresh transfer validation.
- Retain output tails and every unavailable record when assessing a timing target.
Research scope
The first window retains a 1,591 ns maximum. The later transfer retains 16 invalid records and a 362 ns maximum. An observed 49 ns internal display RMS is a different metric. The sub-100 ns downstream target, RF endpoint delay, physical output and absolute UTC remain unqualified.
Read it in context
This is the dated 22 October follow-up. The 24 October W12 baseline is a separate later population with different source versions and calibration; it does not overwrite this paper’s measurements.
Cite this edition
Jonathan Yantis. Precision Bluetooth Timing on ESP32-S3: Causal Clock Discipline and Hardware Event Capture. 22 October 2026. Research draft. Original PDF.
Summary permalink: https://ptob.dev/research/esp32-s3-causal-clock-discipline/
Jonathan Yantis leads protocol design, hardware development and research. GPT 6.1 contributes implementation, analysis and documentation. This page summarizes the linked edition; retain its original evidence dates and measurement references.