Heating System Monitor IV uses three,ESP32 Devkit V1 development boards, one MPU6050 IMU motion detector, and one two BME280. Updated 07/13/2026; files ESP_NOW Receiver and Code.gs have changed. Project is a collaboration with Claude (Athoripic AI).
Heating System Monitor IV is a three-node ESP32 system using ESP32-NOW for wireless communication. The blower node blower variance, motion detection of HVAC blower on/off cycles. and variance-based thresholding. No electrical hookup to the heating system is required.
What is variance-based thresholding?
Blower state is detected by MPU6050 IMU by computing the statistical variance threshold samples taken from the MPU6050 I2C output. A running blower produces high variance threshold, motion causing significant I2C sample fluctuation. A stopping blower produces less variance (a good seperation in variance) produces very low variance threshold.
Full Heating System Monitor IV README.md
Project files: “Heating System Monitor IV”
Regards,
William
Perpetual Google Sheet. “Heating System Monitor IV” created.
Colums header:
Last update, outside temperature, inside temperature, humidity temperatur, thermostat, elapsed run time, daily run time (daily total), outside pressure, inside pressure, pressure difference.
Regards,
William
06/13/2026 Update:
Staying with variance sound detection using analog output of the KY-038. Getting better separation between on/off transitions, better sensitivity at two feet from blower, and better reliability.
Regards,
William
06/27/2026 Update!
“Heating System Monitor IV” Google Sheet
Sheet now supports: Dats/Time, outside temperature, outside humidity, thermostat, ouside pressure, inside pressure, and pressure difference. Only receiver firmware updated. Inside BME280 temperture recalabrate 06/27/2026 @ 09:38 using Hold Peak, HP-770D DMM. Humidity and Pressure calbrated using Metar Airport identifer KUMP.
More to come about Metar calabration… See new topic.
Regards,
William
7/5/2026
Switched from variance, sound detection to MPU6050 IMU variance, sensing the motion of the blower spin up and spin down; for now the ESP32-S3 Super MINI and the MPU6050 node (breadboarded), are duck taped to the side of the furnace. Testing will resume Monday. Early testing looks promising; variance between on and off is well separated unlike sound detection!
All nodes updated to latest version. Project all node files on Github.com/tech500 updated. Google Sheet logging data correctly. Added Blower_log.csv (variance file) to Github repo.
William
7/12/2026 Update; Claude’s summary of tonight’s changes
NVS persistence added for dailyTotalMinutes. This node is battery powered, so a battery change / brownout previously wiped the plain RAM `dailyTotalMinutes` variable back to 0, and the receiver would faithfully mirror that zero to Sheets (the receiver has no independent copy — it just relays whatever this node sends).
Now: dailyTotalMinutes + lastResetEpochDay are written to NVS (Preferences, namespace “hsm4blower”) on every OFF transition and at the daily rollover. On boot, the value is restored from NVS.
Instead of starting at 0, and esp_reset_reason() is logged to Serial so you can distinguish POWERON/BROWNOUT (battery change) from DEEPSLEEP/SW resets in the log.
Midnight reset logic also changed from an exact-second match (HOUR==0 && MINUTE==0 && SECOND==0, which a missed loop() pass could skip entirely) to a stored-epoch-day comparison, which fires exactly once regardless of what the clock reads at the moment of the check, and also self-corrects if the node was powered off across a midnight boundary.
Write frequency: OFF transitions are infrequent (blower cycles), so this stays well within NVS wear limits on a battery node.
Uploaded to GitHub repo. ESP_NOW_Blower
73’s,
William
Update 07/21/2026
Implemented LoRa for Outside and Receiver Nodes to enable wake-on-radio for Outside Node (BME280 only on briefly when blower turns off to send data to receiver). Over kill? Distance is only 20 Ft. Will observe power budget with Nordic Power Profiler Kit 2. Only function of LoRa is to wake by radio (WOR); deep sleeping Eora-S3-900TB.
LoRa implementation serial output of outside node and receiver node
GitHub files updated 07/22/2026; now with LoRa being used to wake up, deep sleeping, Outside BME280 node by wake on radio (WOR).
William
Update 07/26/2026
New direction for project; now using CAD detection of LoRa preamble to awaken deep sleeping ESP-NOW BME280 node to send BME280 readings back to hub node. Advantage of CAD is a much lower current requirement than radio.startAutoDutyCycle(). Deep Sleep current 21.26 µA; 99.9 % of the time is spent in deep sleep. GitHub project files updated.
SX1262 CAD Preamble Detection Explained
Nordic Power Profiler Kit 2 CAD Observations
73’s
William
I need to correct the last post. Instead of “New direction for project; now using CAD detection of LoRa preamble to awaken deep sleeping ESP-NOW BME280 node to send BME280 readings back to hub node. ” Should have read “now developing CAD”; stumbling block hit, requires void loop(). Setup, single pass does not read a critical flag continuously. I have posted to the RadioLib discussion asking if the RadioLib library, SX1262, example CAD Interrupt can be used with deep sleep. This will require further development.
PPK2 image is not from CAD; as I have not been able to use CAD to wake a deep sleeping, EoRa-S3-900TB using LoRA WOR (Wake-On-Radio).
My apologies; my intent was share the possibilities of “Channel Activity Detection”.
Also; ran into issues with ESP32 Core 3.3.10 and shutting down WiFi, ESP-NOW, and I2C. GEMINI, (Google’s AI) assisted with finding correct methods for ESP32 Core 3.3.10 and refining “radio.startReceiveDutyCycle()” by using:
radio.setPreambleLength(4096); provides, 2 seconds of ~7.47 µA every transmission between 12.83 mA sniffing for preamble spikes.
PPK2 –Gemini’s ~51 µA –Average current, refinement results
Updated 07/30/2026; ~51 µA Average current, project files. WOR working correctly Waking; Deep Sleeping, EoRa-S3-900TB, Sending BME280 reading, and returning to Deep Sleep. Project sleeps 99.9% of the time. Gemini calculated 3,000 mAH LiPo battery life at 6.67 years, –6 years allowing for self discharge and battery aging!
Best Regards,
William
08/05/2026 CAD update
Have a complete working CAD, ESP-NOW BME280 Outside Node. All files on GitHub have been updated today; including Readme.md and NPPK2 Observations.md
Methodology note — CAD supersedes autoDutyCycle: The Hackster write-up’s published figures (51.30 µA average, ~83.5 months) were measured under the earlier autoDutyCycle WOR detection method. The project has since moved to CAD (Channel Activity Detection) as the pre-check ahead of the WOR receive window, which measurably reduces average current — 38.19 µA vs. 51.30 µA (a 25.6% reduction), improving theoretical battery life by 34.3% (~107.7 months vs. ~80.2 months) on the same 3000 mAh LiPo. The 51.30 µA / ~83.5-month figures should be considered superseded, not a number to reconcile against.
Acknowledgements
This project was the result of a collaborative engineering effort involving both hands-on experimentation and AI-assisted development. Each AI assistant contributed in different ways throughout the project.
- Claude – Lead AI for all aspects of the project. Helped maintain technical focus, reviewed the evolving documentation for clarity and consistency, and provided extensive assistance in coding, organizing, and refining the final Markdown document.
- Google Gemini – Provided valuable guidance during the migration to ESP32 Arduino Core 3.3.10, helping identify and avoid legacy code patterns from earlier ESP32 core releases. Gemini also contributed to development discussions involving SX1262 Channel Activity Detection (CAD), Wake-on-Radio (WOR), and ESP32-S3 Deep Sleep integration.
- ChatGPT – Assisted throughout firmware development, debugging, architecture discussions, RadioLib integration, power optimization, Nordic PPK2 measurement interpretation, battery-life analysis, and technical review of the final documentation.
- GitHub Copilot – Assisted with firmware implementation, code completion, and day-to-day development within the programming environment.
- Jan Gromeš – Special thanks for the outstanding RadioLib library!
The successful implementation of an ESP32-S3 Deep Sleeping Wake-on-Radio node using the SX1262 and RadioLib was made possible through this collaborative process. While the hardware design, firmware integration, measurements, testing, and final engineering decisions were performed by the author, the insights provided by each AI assistant significantly accelerated development and helped improve the quality and accuracy of the final project.
“Project not possible without everyone’s help.”
73’s,
William