Commercial Tenant Sub-Metering Bridge

note

This reference application is intended to provide inspiration and help you get started quickly. It uses specific hardware choices that may not match your own implementation. Focus on the sections most relevant to your use case. If you'd like to discuss your project and whether it's a good fit for Blues, feel free to reach out.

Building with an AI assistant? Point it at this accelerator's firmware — firmware/tenant_sub_meter/ — so it can learn how the device works and help you adapt it to your hardware.

This project is a cellular energy monitoring retrofit for multi-tenant commercial buildings. Four Rogowski coil sensors with per-channel active integrator circuits on a panel, a Blues Notecard Cell+WiFi, and a Notecarrier CX with its onboard STM32 host, and a landlord gains per-tenant estimated interval energy (Wh), 15-minute blocked-average demand (W), and a per-channel fault bitmask every hour, on a data channel the tenants can neither see nor interfere with, without running a single network cable through the building.

1. Project Overview

The problem. A landlord of a multi-tenant commercial building (a flex-industrial park, a shared professional building, an office strip) wants to allocate electricity costs to each tenant based on their actual consumption rather than splitting the utility bill by square footage or some other proxy. The fair approach is sub-metering: install a watt-hour meter on each tenant's branch circuit and read it monthly.

The problem is the cost and the complexity. Commercial sub-meters require a wired or networked path to get their data off the panel and into the billing system. An older building with four or eight tenants has no data infrastructure of any kind in the electrical room, and retrofitting it means negotiating with tenants, coordinating with the utility, and running cable through occupied ceilings. Most landlords give up and keep allocating by square footage.

This project takes a different path. Instead of installing meters with network cards, it clips Rogowski coil sensors onto each tenant's existing feed conductors (no wire cuts, no conduit, no utility coordination) and uses a Notecard to move the data to the cloud. A voltage transducer provides a coherent voltage reference so the device computes active power (W) and estimated interval energy (Wh) from sequential interleaved V×I measurements, rather than an estimated figure derived from an assumed voltage and fixed power factor. The system installs in an afternoon and costs a fraction of a traditional per-tenant metered retrofit.

What this project delivers, and what it does not. This design provides estimated tenant load allocation data useful for proportional charge-back, load profiling, and identifying unusually high-consumption tenants. It does not replace a utility-certified sub-meter. Readings are derived from periodic active-power snapshots (one ~200 milliseconds V×I mean every five minutes) rather than continuous integration, and the sequential (non-simultaneous) V/I sampling introduces a small systematic error at low power factors. For internal bill-back purposes the approach is accurate enough to support proportional allocation between tenants. For billing applications requiring certified metering accuracy, for example, sub-metering subject to utility-tariff accuracy regulations, a dedicated simultaneous-sampling energy-metering IC or certified sub-meter interface should replace this approach.

Why Notecard, and why cellular specifically. The "why cellular" argument in most IoT applications is about location: the asset is remote, or on a rooftop, or in a place where running Ethernet is impractical. This application is different. The panel room is usually indoors, and there is WiFi somewhere nearby. The problem is whose WiFi it is.

New to Blues?

Learn what we do in this 3-minute video that walks you through how we simplify wireless connectivity.

Learn what we do in this 3-minute video that walks you through how we simplify wireless connectivity.

In a multi-tenant building, the WiFi access points are almost always under tenant control. A retail tenant pays for their own ISP and runs their own router; a professional office tenant does the same. The building owner typically has no WiFi at all, or if they do, tenants have access to it. Routing billing telemetry through any network the tenants can reach is a political and legal problem: a savvy tenant can see what's being transmitted, disrupt the connection, or argue that the data was tampered with. The landlord needs a data channel that is:

1. 100% landlord-controlled: the tenant cannot touch it
2. Invisible to tenants: the tenant cannot see or intercept it
3. Operationally simple: no IT setup, no passwords, no tenant cooperation

A cellular Notecard meets all three criteria. It's a SIM-bearing module that calls home to the Blues Notehub cloud service over the cellular network, a connection the tenant has no access to and no visibility into, exactly the same way the building's alarm system uses its own GSM dialout. There is no form to fill out, no AP to pair to, and no IT ticket to raise with any tenant. The data channel is as landlord-owned as the panel itself.

This is not a niche edge case. Billing disputes are among the most contentious issues in commercial tenancy, and the architecture matters: a meter whose data travels over the tenant's network is a meter whose readings a clever tenant can plausibly dispute. A cellular Notecard is the only architecture that eliminates that dispute by design.

Deployment scenario. A small, weatherproof or panel-rated enclosure mounted inside or alongside the building's electrical panel. Four Rogowski coil sensors — one per tenant feed — with the coil lead running to the enclosure via a short cable gland. A voltage transducer taps the same panel bus for a building voltage reference. Line power sourced from a spare breaker on the same panel. No rented rack space, no network closet, no IT involvement from anyone.

2. System Architecture

Device-side responsibilities. All four channels share one STM32L433 host on the Notecarrier CX, and that host wakes on a five-minute cadence (default) to do the same thing every cycle: read each active tenant channel by interleaving ADC samples between the voltage transducer output and the Rogowski integrator output, compute active power (W) via a V×I cross-product, and accumulate estimated interval energy (Wh), 15-minute blocked-average demand (W), and a per-channel hardware-fault bitmask into a RAM struct that survives each sleep, because the host sleeps in STM32 STOP2, which retains RAM. Once per hour the host emits a summary Note carrying the accumulated energy, peak demand, and fault flags. Configuration lives in environment variables distributed by Notehub (sensor sensitivity, voltage scaling, all of it) so the landlord can adjust calibration in the field without a reflash.

Notecard responsibilities. The Notecard is the cellular muscle and the sleep gate. It queues Notes locally, wakes the cellular radio on the configured hub.set outbound cadence (default 60 min), and flushes the entire queue to Notehub in a single cellular session. Environment-variable changes pushed from Notehub come back on its inbound cadence (default 120 min), and card.attn sleep mode holds the ATTN pin low between samples while the host sleeps in STM32 STOP2 (a few µA), then raises it to wake the host for the next sample.

Notehub responsibilities. Once Notes arrive at Notehub over the Notecard's embedded global SIM, they're stored durably and routed downstream. Notehub is also the sole monthly aggregation path: a billing or property-management platform sums each tenant's t*_wh values across all hourly meter_summary.qo events for a device over any billing period using the Notehub Event Query API. The device never emits a month-end Note and never tries to compute a billing total; the hourly event stream in Notehub is the complete, authoritative record.

Fleets group devices per property so a property manager with multiple buildings can push sensor calibration settings per-building without touching individual devices. A Notehub HTTP or MQTT route can forward each meter_summary.qo Note directly to the property-management or billing platform as it arrives.

Routing to the cloud (high level only). Notehub supports HTTP, MQTT, AWS, Azure, GCP, Snowflake, and other destinations; route setup is project-specific. See the Notehub routing docs; this project ships no specific downstream endpoint.

3. Technical Summary

Four independent sub-meter channels, each with isolated analog inputs and read in turn as a ~200 ms burst of 2000 sequentially interleaved V/I ADC sample pairs, feed the Notecarrier CX's STM32L4 host running accumulation windows across configurable billing periods. Per-tenant energy (Wh), peak demand (W), and hardware-fault status flow upward via Notecard on a configurable outbound cadence. On-device queuing, sleep gating, and fault-aware transmission tolerate carrier dropouts and cloud-side downtime without data loss. No external flash or EEPROM required; state lives in host RAM, which STOP2 retains between wakes.

Here is a sample Note this device emits:

{
  "t1_wh":       482.3,
  "t1_demand_w": 3120.0,
  "t2_wh":       214.1,
  "t2_demand_w": 1480.5,
  "t3_wh":       673.8,
  "t3_demand_w": 4220.0,
  "t4_wh":       149.2,
  "t4_demand_w":  980.0,
  "fault_mask":     0
}

4. Hardware Requirements

Prototype / Bench Evaluation BOM

Use this BOM for firmware bring-up, calibration, and evaluation. The ZMPT101B module is not agency-listed for permanent installation inside a commercial panel enclosure; replace it with an agency-listed isolated voltage transducer before deploying to a live panel — see §11 Limitations and Next Steps for the full production voltage-sensing path, including hardware requirements and firmware adaptation Notes.

All Blues hardware ships with an active SIM including 500 MB of data and 10 years of service — no activation fees, no monthly commitment.

5. Wiring and Assembly

All host I/O lands on the Notecarrier CX dual 16-pin header, and a short jumper between ATTN and D5 on that header lets the Notecard wake the sleeping host. The Notecard MBGLW seats into the M.2 slot on the carrier. Mojo (bench-only) sits inline between the 5 V supply output and the Notecarrier CX +VBAT pad.

warning

Cellular antenna — critical for metal-panel deployments. Connect the external cellular antenna's u.FL lead to the Notecard's primary antenna port and route the cable through a cable gland to outside the metal panel cabinet or NEMA enclosure. Steel panels and enclosures act as Faraday cages; a rubber-duck or pigtail antenna left inside a closed metal panel will not establish a reliable cellular session. This is the single most common field bring-up failure in panel-mounted deployments.

Power chain

AC/DC 5 V supply (IRM-10-5)
  └─ +5 V out ──► Mojo BAT             (bench validation — Mojo measures supply current)
                    └─ Mojo LOAD ──► Notecarrier CX +VBAT   (3.3 V regulated on-board)

For production deployment (no Mojo): connect the AC/DC supply 5 V output directly to Notecarrier CX +VBAT.

warning

Do not connect the supply output simultaneously to both +VBAT and +VUSB/+5V on the Notecarrier CX. The +VBAT pad is the correct power input for this supply voltage and load profile; feeding two supply rails at once risks back-feeding one regulator through the other.

Active Miller integrator (replicate for each of the four current channels)

Each Rogowski coil outputs a voltage proportional to dI/dt (the derivative of the primary current). The active Miller integrator converts this dI/dt signal back to a signal proportional to instantaneous current, which can then be presented to the ADC for RMS and V×I computation.

Rogowski coil output (bare lead)
  │
  ├── R_in (10 kΩ) ──────────────────────────► op-amp IN− (inverting)
  │                                                │
  │                         R_f (1 MΩ)             │
  │                   ┌────────────────────────────┤
  │                   │                            │
  │                   │     C_f (10 nF)            │
  │                   └──┤├────────────────────────┘
  │
  │   op-amp IN+ (non-inverting) ──────────────── 1.65 V bias node
  │
  └── op-amp OUT ─────────────────────────────── to ADC pin (A0–A3)

Bias node: +3V3 ── 100 kΩ ──┬── 100 kΩ ── GND
                            ├── 10 µF ── GND
                            └── op-amp IN+ of this channel

Use one section of the MCP6004 (U1A–U1D) per channel. All four op-amps share Vdd = +3V3 and GND.

Rogowski coil polarity: if a loaded channel shows near-zero or zero t*_wh accumulation in Notehub despite a known load, the coil lead polarity may be reversed. The firmware clamps negative watts to zero before accumulating energy, so reversed polarity appears as zero/near-zero billed energy — not negative values — in Notehub. The negative-watt condition is visible on the debug serial output (look for a negative watt figure in the [sample] T*: line) during bench bring-up. Swap the coil's two output leads at the R_in input of the integrator to correct polarity.

ZMPT101B voltage sensor module (bench/prototype — one module, building voltage reference)

This design uses a conditioned ZMPT101B module (VCC / GND / VOUT pinout) — not a bare transformer secondary. The module's PCB carries the ZMPT101B transformer, an onboard burden resistor (sets the output scaling), and a calibration trim pot. Connect it as follows:

Building 120 V line ──► ZMPT101B module AC primary (IN1 / IN2 screw terminals,
                         fused tap per local code)

Module power and signal:
  VCC  ──► +5V (from AC/DC supply before the Notecarrier CX regulator)
  GND  ──► GND

  VOUT ──► 10 µF (series AC coupling) ──┬── ADC A4
                                        │
               +3V3 ── 100 kΩ ──────────┤
                                        │
                GND ── 100 kΩ ──────────┘

The module outputs an AC signal riding on a DC level set by its onboard circuitry (typically near half its supply voltage — approximately 2.5 V when powered at 5 V). VOUT is AC-coupled through the 10 µF series capacitor to ADC A4; the external 100 kΩ / 100 kΩ voltage divider then sets the DC operating point of the ADC input at 1.65 V, independent of the module's internal reference. Firmware removes any residual DC offset in software (v_dc = mean(v_buf[])), so the exact bias voltage does not need to be trimmed.

The module's onboard burden resistor and trim pot set the output peak-to-peak swing. At 120 V primary with the factory burden, the output is typically ≈0.1 V RMS at the VOUT pin; volt_scale = 1200 (default) is the corresponding calibration constant (120 V ÷ 0.1 V). Measure with a precision AC voltmeter at a known line voltage and trim volt_scale in Notehub for best accuracy.

Pin-by-pin assignments

• A0 → Tenant 1 Rogowski integrator output (op-amp OUT direct)
• A1 → Tenant 2 Rogowski integrator output (op-amp OUT direct)
• A2 → Tenant 3 Rogowski integrator output (op-amp OUT direct)
• A3 → Tenant 4 Rogowski integrator output (op-amp OUT direct)
• A4 → ZMPT101B VOUT via 10 µF series AC coupling cap into the 1.65 V bias node
• +3V3 → MCP6004 Vdd, top of all half-rail bias dividers
• +5V → ZMPT101B module VCC (bench)
• GND → MCP6004 GND, bottom of all bias dividers, negative terminals of the four current-channel 10 µF bias-decoupling caps (A0–A3 bias nodes), Rogowski coil returns, ZMPT101B module GND (the voltage-channel 10 µF capacitor is the series AC-coupling cap between ZMPT101B VOUT and A4 — its negative lead connects toward A4 and the 1.65 V bias node, not to GND; see the ZMPT101B wiring diagram above)
• D5 → Notecarrier CX ATTN (jumper; see below)

ATTN-to-D5 jumper (host sleep)

Fit a short jumper between the ATTN and D5 pins on the Notecarrier CX header — both are on the same 16-pin header. The Notecard pulls ATTN low while the host sleeps and raises it when the next sample is due; the rising edge on D5 wakes the host from STOP2. Leave the Notecarrier CX EN pin unconnected: on the CX it enables the shared 3.3 V VIO rail (host, Qwiic, 3V3_OUT, and Notecard I/O), so driving it from ATTN would brown out the whole board rather than sleep the host. See Putting a Host to Sleep Between Sensor Readings for the full pattern.

Rogowski coil placement

Each Rogowski coil wraps around one conductor only of the tenant's single-phase branch-circuit feed — the hot leg. The coil must close fully (the two ends of the flexible Rogowski coil clip together); an open coil gap produces a non-integrable signal. This firmware is designed for single-phase circuits where all tenants derive from the same phase leg; see §11 for the multi-phase case.

warning

Electrical safety. Panel interiors operate at hazardous voltages. Installation must be performed by a licensed electrician following applicable electrical code, site lockout/tagout procedures, and the panel manufacturer's instructions. The Rogowski coils and voltage transducer are galvanically isolated from the mains; the hazard is the bus bars and termination points nearby during installation.

Before powering on for the first time:

• PRODUCT_UID set — edited tenant_sub_meter_helpers.h line 21 with your Notehub ProductUID (e.g., com.your-company:your-app).
• Core installed — Arduino core for STM32 added via Boards Manager. Search "STM32 MCU based boards".
• Blues Notecard library installed — installed via Library Manager, along with STM32duino Low Power and STM32duino RTC.
• ATTN→D5 jumper fitted — on the Notecarrier CX header; EN left unconnected.
• DIP switch set to HST — on the Notecarrier CX, before connecting USB.
• Compiled and uploaded — arduino-cli compile succeeds with no errors. arduino-cli upload completes successfully.
• Debug serial at 115200 baud — on the ST-LINK virtual COM port (debug jack), watch for [notecard] and [sample] lines (host is running).

If compilation fails with "PRODUCT_UID is not defined", you missed step 1. If upload fails with "board not found", check the DIP switch and USB cable.

6. Notehub Setup

Fast path to first event

1. Notehub — create a Notehub project, copy its ProductUID.
2. Wire the bench rig — Notecarrier CX + Notecard MBGLW + four Rogowski integrator circuits on A0–A3 + ZMPT101B on A4. Full pinout in §5.
3. Edit one line in firmware/tenant_sub_meter/tenant_sub_meter_helpers.h — replace #define PRODUCT_UID "" with your project's value.
4. Flash —
   COPY

   arduino-cli lib install "Blues Wireless Notecard" "STM32duino Low Power" "STM32duino RTC"
   arduino-cli compile -b STMicroelectronics:stm32:Blues:pnum=CYGNET,usb=none firmware/tenant_sub_meter/
   arduino-cli upload -b STMicroelectronics:stm32:Blues:pnum=CYGNET,usb=none -p /dev/cu.usbmodem* firmware/tenant_sub_meter/

   The usb=none board option disables the USB CDC stack, which must be off for the host to sleep (see §7.1). Full troubleshooting in §7.1. Connect the Notecarrier CX with USB and set the DIP switch to HST.
5. Watch serial — open a serial monitor at 115200 baud on the ST-LINK virtual COM port (the Notecarrier CX debug jack; see §7.1). You will see Notecard configuration, then channel sample lines. After the first sleep cycle, serial goes quiet (host is in STOP2) for 5 minutes — that silence is expected and healthy.
6. Check Notehub — open Events tab. A _session.qo appears within minutes. The first meter_summary.qo typically arrives in the initial session, shortly after the first sample cycle completes (within an hour).

Fleets, environment variables, and routing

Once your device is appearing in the Devices tab and events are flowing, configure the project for ongoing operation:

1. Create a Fleet per property. Fleets group devices for shared configuration. The natural unit here is one fleet per building — every panel in the same building typically operates at the same line voltage and uses the same Rogowski coil model, so fleet-level environment variables encode "this building's panels all use these sensor calibrations." Use Smart Fleets to auto-assign devices by serial number prefix for multi-property deployments.

2. Set environment variables. Navigate in Notehub to Fleet → Environment (or Device → Environment for per-device overrides). Edit the table below, then save. The device pulls them on its next inbound sync (default: every 2 hours) — no reflash, no truck roll.

   Variable	Default	Purpose	When to adjust
   sample_interval_sec	300	Seconds between measurements (minimum 60, max 86400).	Leave at 300 for most installs; smaller values consume more power and produce more frequent events.
   summary_interval_min	60	Minutes between hourly summaries (minimum 15, max 1440). The minimum of 15 matches the 15-minute demand window.	Leave at 60 for standard hourly billing. Values < 15 are silently clamped to 15.
   rogowski_amps_per_volt	400.0	Rogowski coil + integrator sensitivity in A/V. Derivation: rated primary amps ÷ integrator output V RMS at rated load.	Critical: Calibrate against a reference clamp meter at a known full load before billing. Incorrect values produce proportionally wrong energy readings. See §10 calibration note.
   volt_scale	1200.0	Voltage transducer calibration: line-voltage V RMS ÷ ADC-pin V RMS.	Default (1200.0) matches ZMPT101B at 120 V nominal. Measure line voltage with a precision AC voltmeter and trim for accuracy.
   num_tenants	4	Active current channels to sample (1–4).	Set to the actual number of tenants metered to skip unnecessary ADC reads.

3. Configure routes. Route meter_summary.qo to your time-series database or billing platform. This is the canonical energy record and the correct source for monthly aggregation — sum each tenant's t*_wh values across all hourly events over the billing period using the Notehub Event Query API. See the Notehub routing docs for destination types.

What you should see in Notehub

Within minutes of first power-on, the Events tab in your project should show a _session.qo (the cellular handshake). The firmware emits a summary note on first boot, so the first meter_summary.qo typically arrives in the initial session.

Example meter_summary.qo payload:

{
  "t1_wh":       482.3,
  "t1_demand_w": 3120.0,
  "t2_wh":       214.1,
  "t2_demand_w": 1480.5,
  "t3_wh":       673.8,
  "t3_demand_w": 4220.0,
  "t4_wh":       149.2,
  "t4_demand_w":  980.0,
  "fault_mask":  0
}

This arrives hourly. t*_wh is estimated interval energy in watt-hours per tenant over the past hour; sum these across all hourly events over a billing period to derive monthly kWh. t*_demand_w is the peak 15-minute average demand (in watts) in that hour. fault_mask = 0 means all channels passed hardware health checks.

7. Firmware Design

Four-file implementation — all four must be present in the same sketch directory:

• firmware/tenant_sub_meter/tenant_sub_meter.ino — main sketch: setup() / loop() / runCycle(), state management, measurement orchestration, and sleep/wake lifecycle.
• firmware/tenant_sub_meter/tenant_sub_meter_helpers.h — shared types (TenantState, AppState, RuntimeConfig, ChannelMeasurement), constants, fault-flag definitions, and helper function declarations.
• firmware/tenant_sub_meter/tenant_sub_meter_helpers.cpp — implementations of all helper functions (measureChannel, fetchEnvOverrides, initNotecard, defineTemplates, sendSummary, and getEpochSec).
• firmware/tenant_sub_meter/cx_sleep.h — host sleep: the card.attn request, STOP2 entry, and ATTN-pin wake (cxSleepBegin, cxSleepUntilAttn).

7.1 Installing and flashing

Dependencies:

• Arduino core for STM32 — stm32duino/Arduino_Core_STM32. Install via the Arduino Boards Manager (search "STM32 MCU based boards") or add the index URL https://github.com/stm32duino/BoardManagerFiles/raw/main/package_stmicroelectronics_index.json under File → Preferences → Additional Boards Manager URLs. The Notecarrier CX's host uses the Cygnet board definition (see build settings below).
• Blues Wireless Notecard library — note-arduino. Install via the Arduino Library Manager, or arduino-cli lib install "Blues Wireless Notecard". Verify the latest release against the note-arduino releases page at install time.
• STM32duino Low Power library (and its dependency STM32duino RTC) — stm32duino/STM32LowPower. Install via the Library Manager. Provides the STOP2 entry and ATTN-pin wake used in cx_sleep.h.

Build settings: the Notecarrier CX's host is the same STM32L433 as the Blues Cygnet, so the board is Blues boards → Cygnet (pnum=CYGNET), with USB support (if available) set to None (usb=none). With the USB CDC stack enabled and no USB host attached (the Notecarrier CX USB-C switch in NC, or no cable), the STM32's USB wakeup interrupt fires continuously and the host never stays in STOP2. Debug output goes to the LPUART on the Notecarrier CX debug jack instead — an ST-LINK V3 exposes it as a virtual COM port; see Serial Logging With STLINK.

Flashing — Arduino IDE: Open tenant_sub_meter.ino, apply the build settings above, and click Upload. With the Notecarrier CX's DIP switch in HST, the host's DFU bootloader is reachable over the USB-C port (hold BOOT, tap RST); alternatively upload over SWD with an ST-LINK on the debug jack.

Flashing — arduino-cli (step by step):

First, confirm the board and port:

# List all recognized boards
arduino-cli board listall | grep -i cygnet
# Expected output: STMicroelectronics:stm32:Blues:pnum=CYGNET

Then set the DIP switch on the Notecarrier CX to HST and compile/upload. usb=none disables the USB CDC stack, which must be off for the host to sleep:

# Compile only (no upload yet)
arduino-cli compile \
  -b STMicroelectronics:stm32:Blues:pnum=CYGNET,usb=none \
  firmware/tenant_sub_meter/

# List USB ports connected to see your device
arduino-cli board list

# Upload (replace /dev/cu.usbmodem* with your actual port from board list)
arduino-cli upload \
  -b STMicroelectronics:stm32:Blues:pnum=CYGNET,usb=none \
  -p /dev/cu.usbmodem* \
  firmware/tenant_sub_meter/

# Watch the progress — when upload is complete, disconnect USB and flip DIP to NC

On Windows, ports are named COM3, COM4, etc.; on macOS / Linux, /dev/cu.usbmodem* or /dev/ttyUSB*.

After flashing, open a serial monitor at 115200 baud on the ST-LINK virtual COM port. On first power-on you will see the Notecard configuration exchange, then the first channel sample lines. After the first sleep, the serial output goes quiet for sample_interval_sec — that silence is expected; the host is in STOP2 and resumes loop() in place when the Notecard raises ATTN on the next wake.

7.2 Module map

7.3 Sensor reading strategy

The measureChannel() function captures 2000 interleaved voltage/current sample pairs from VOLTAGE_PIN (voltage transducer output) and the selected current pin (Rogowski integrator output). At the STM32L433's default 12-bit ADC rate, each analogRead() takes approximately 50 µs; one interleaved pair therefore takes ~100 µs, so 2000 pairs cover ~200 ms, approximately 12 full 60 Hz cycles. Voltage is sampled first within each pair, so any systematic ADC switching latency between channels is consistent across all pairs.

DC removal. Both signals are biased to 1.65 V (half-rail) by external resistor dividers. A single-pass mean calculation over all samples determines the per-channel DC offset, which is subtracted before computing RMS or cross-products. This removes the bias voltage and any residual ADC offset without a separate hardware AC-coupling capacitor in the signal path.

Active power computation. Active power P = mean(v[n] × i[n]), where v[n] and i[n] are the DC-removed voltage and current samples scaled to physical units. The V×I cross-product captures the actual phase relationship between voltage and current; no assumed power factor is needed.

Phase/accuracy limitation from sequential sampling. Because voltage and current are read sequentially (not simultaneously), there is a fixed ~50 µs time offset between the V sample and the I sample in each pair. At 60 Hz, one full cycle is 16.67 ms, so 50 µs corresponds to a consistent phase offset δ of approximately 1.08°. The relative error in active power is approximately tan(φ)·sin(δ) — where φ = arccos(PF) is the load power-factor angle — because the measured cross-product evaluates cos(φ − δ) rather than cos(φ). At power factors above 0.9 (typical commercial loads, φ ≈ 26°), this error is approximately 1% of the true active-power reading. At PF 0.7 (φ ≈ 46°) it reaches approximately 2%. For inductive loads (positive φ) the firmware systematically over-reads active power by this margin. For metering applications requiring certified accuracy (e.g., utility-tariff sub-metering subject to accuracy regulations), a dedicated simultaneous-sampling energy-metering IC should replace this approach.

A Rogowski coil installed with reversed lead polarity produces a consistently negative watts result; the firmware clamps negative values to zero and the issue is visible in the serial log as a zero-energy reading on the affected channel.

Energy estimation. Each sample_interval_sec (default 5 min) the firmware takes one ~200 ms active-power snapshot (mean V×I over 2000 sample pairs) and multiplies it by the full interval duration to produce an estimated interval energy in Wh. This is accurate for constant or slowly-varying loads, but may over- or under-state energy on bursting or cycling loads where power changes significantly within the 5-minute window. The t*_wh accumulator sums these interval estimates; it is not a continuous integral. This makes the design suitable for proportional load allocation (identifying which tenants use more energy than others) but not for applications requiring certified metering accuracy.

Demand window. Energy accumulated in demand_window_mwh and elapsed time in demand_window_sec grow across sample wakes. Once demand_window_sec reaches DEMAND_INTERVAL_SEC (900 s = 15 minutes), average watts are computed as demand_window_mwh × 3600 / (demand_window_sec × 1000) and compared against peak_demand_cw, the per-tenant centi-watt accumulator for the current summary period. If the new average is higher, peak_demand_cw is updated. Both window fields then reset to begin the next 15-minute interval. This is a true interval-average demand reading — not a transient snapshot — matching the most common North American utility demand-charge billing window. The t*_demand_w field in meter_summary.qo reports peak_demand_cw / 100 (converted to watts) for the summary period. The demand window deliberately straddles summary-period boundaries: peak_demand_cw is cleared after each confirmed sendSummary(), but demand_window_mwh and demand_window_sec continue uninterrupted so no 15-minute interval is silently truncated at a summary boundary.

Summary interval and demand window floor. summary_interval_min is clamped to a minimum of 15 in firmware (MIN_SUMMARY_INTERVAL_MIN), matching DEMAND_INTERVAL_SEC / 60. A summary period shorter than one demand window would always emit t*_demand_w = 0 because no complete 15-minute window could close within it; this is not a hardware fault, just a window-timing gap. Operators who need finer-grained reporting should use the t*_wh field and accumulate their own demand windows in the downstream system.

Per-channel fault bitmask. Each call to measureChannel() checks four conditions per channel nibble: FAULT_BIAS_RANGE (0x01) when the ADC DC offset falls outside the expected 1.40–1.90 V half-rail window; FAULT_SATURATED (0x02) when the centered signal's RMS exceeds the RMS-equivalent of 85 % of the 1.65 V half-rail peak amplitude (≈ 0.99 V RMS at the ADC pin), indicating that the waveform is approaching ADC rail clipping; bit 2 reserved (formerly FAULT_NO_SIGNAL, retired because a legitimately unloaded tenant circuit is indistinguishable from a disconnected Rogowski coil by current magnitude alone; see the commissioning diagnostic note below); and FAULT_VOLTAGE_REF (0x08) when the shared voltage reference path is suspect (bias out of range, signal saturated, or measured line RMS below VOLTAGE_MIN_V_RMS). Because the voltage reference is shared, FAULT_VOLTAGE_REF is propagated into every active channel's fault nibble simultaneously so downstream billing can identify periods where all tenant watt calculations are compromised. Fault flags are OR'd across all sample wakes in the summary period and packed into the fault_mask field of meter_summary.qo as a 4-bit nibble per tenant channel (T1 in bits 3:0, T4 in bits 15:12). A fault_mask of 0 means all channels passed all checks on every sample in the period. Downstream billing systems should reject or flag any summary where fault_mask != 0. Low RMS current (below 0.05 A) is logged to the debug serial port as a commissioning diagnostic only; it is never placed in fault_mask.

7.4 Event payload design

meter_summary.qo is registered as a template-backed notefile at first boot. Templates store Notes as fixed-length binary records, reducing on-wire size by 3–5× vs. free-form JSON.

meter_summary.qo — emitted hourly (canonical record)

Nine 4-byte floats: estimated interval energy (Wh), 15-minute blocked-average demand (W), and a combined fault bitmask per summary period. t*_wh is the sum of (W_sample × sample_interval_sec / 3600) across all periodic samples in the reporting period: estimated interval energy derived from power snapshots, not a continuous integral or kVAh estimate. t*_demand_w is the peak 15-minute blocked-average demand observed during the summary period: the highest average watts across any completed DEMAND_INTERVAL_SEC (900 s) window; see §7.3. fault_mask packs per-channel hardware-fault flags as a 4-bit nibble per tenant (T1 in bits 3:0; see §7.3 for bit definitions); 0 means all channels clean for the entire period.

{
  "t1_wh":       482.3,
  "t1_demand_w": 3120.0,
  "t2_wh":       214.1,
  "t2_demand_w": 1480.5,
  "t3_wh":       673.8,
  "t3_demand_w": 4220.0,
  "t4_wh":       149.2,
  "t4_demand_w":  980.0,
  "fault_mask":     0
}

To derive monthly per-tenant totals, sum t*_wh across all meter_summary.qo events for a device over the billing period using the Notehub Event Query API, then divide by 1000 to obtain estimated kWh. This Notehub-side aggregation is the sole monthly rollup path; no device-side monthly note is generated.

7.5 Low-power strategy

The panel installation is line-powered, but keeping the host asleep between samples reduces enclosure heat, reduces supply wear, and produces firmware that ports directly to battery variants without a rewrite. After each sample cycle, loop() calls cxSleepUntilAttn() (in cx_sleep.h), which sends a card.attn sleep request for sample_interval_sec seconds, waits for the Notecard to pull ATTN low, and puts the STM32 into STOP2, where it draws a few microamps with its RAM intact. When the interval elapses the Notecard raises ATTN; the rising edge on D5 wakes the host and execution resumes on the next line of loop(). Because the host resumes in place rather than cold-booting, the AppState struct simply lives in RAM — nothing is serialized to the Notecard, and setup() runs only once per power-up. If the sleep request fails or ATTN never goes low (jumper missing), loop() waits out the interval awake and tries again on the next cycle, so the sample cadence is preserved either way.

The Notecard sits in its own low-power idle state (~18 µA at 5 V for the Cell+WiFi variant) between cellular sessions. Summary notes queue locally and flush together in one hourly session; the radio is not touched on intermediate wakes.

7.6 Retry and error handling

Environment variable fetch failure. fetchEnvOverrides() calls env.get and checks for a NULL response (Notecard not yet synced) before touching the RuntimeConfig struct. If the response is absent or malformed, firmware defaults are kept intact. Numeric bounds checks guard each variable; for example, rogowski_amps_per_volt is only applied if it falls in (0, 10000], preventing a zero or negative value from corrupting subsequent power calculations.

card.time not yet valid. getEpochSec() returns 0 if the Notecard has not yet established a time sync. The first_summary_sent latch prevents the firmware from re-triggering sendSummary() on every pre-sync wake: once sendSummary() succeeds, the latch is set and subsequent triggers require now > 0 and an elapsed summary_interval_min. If the very first sendSummary() succeeded while now was still 0, last_summary_epoch is left at 0, which would permanently block subsequent hourly summaries. On the first wake where now > 0, the firmware detects this condition (first_summary_sent && last_summary_epoch == 0) and seeds last_summary_epoch = now so the normal elapsed-time gate takes over from that point.

note.add / template failure: the most critical failure path. sendSummary() is gated on state.notecard_configured == 1. If initNotecard() or defineTemplates() fails on a given wake, the configured flag stays clear, all note emission is skipped for that cycle, and accumulators are preserved without ever reaching the send path. When notecard_configured is set, sendSummary() uses notecard.requestAndResponse() rather than the fire-and-forget sendRequest(), so the Notecard's acknowledgement (or error) is explicitly checked:

J *rsp = notecard.requestAndResponse(req);
bool ok = rsp && !notecard.responseError(rsp);
notecard.deleteResponse(rsp);
return ok;

Accumulators are only reset if sendSummary() returns true. If the Notecard returns an error (queue full, I²C fault, or any other condition), the per-tenant accum_wh_milli and peak_demand_cw fields are preserved in AppState. On the next wake cycle the firmware retries sendSummary() with the accumulated totals, which now include an additional sample period's energy. This retry loop continues until a successful acknowledgement is received. Because STOP2 retains RAM, the data survives every sleep during the retry window; only a power cycle or reset clears it.

Template registration failure. defineTemplates() uses requestAndResponse() and returns false if the Notecard returns an error. The notecard_configured flag is only set to 1 if both initNotecard() and defineTemplates() return true; a failure leaves the flag clear so the next wake automatically retries the full initNotecard() + defineTemplates() sequence. Because all note emission is gated on notecard_configured, no Note can be queued before the template exists; the firmware defers emission until setup completes, accumulating energy across the retry wakes without loss. If template registration is suspected to have failed after initial deployment (e.g., after a factory reset of the Notecard), the deterministic recovery procedure is:

1. Wait for any pending notes to complete an outbound sync (so queued data is not lost).
2. Issue {"req":"card.restore","mode":"factory"} via the Notecard In-Browser Terminal or the USB serial REPL. This resets the Notecard to factory defaults, including its registered templates.
3. Power-cycle the device. AppState lives in host RAM, so the power cycle clears notecard_configured to 0 and the firmware runs the full initNotecard() + defineTemplates() sequence again.

note

card.restore with mode:"factory" also clears any Notes queued in the Notecard's local store. Complete step 1 before restoring if preserving in-flight data is important.

7.7 Key code snippet 1: template registration

Only meter_summary.qo is registered as a binary-packed template. 14.1 is the note-c type hint for a 4-byte IEEE-754 float (TFLOAT32). defineTemplates() uses requestAndResponse() and returns a bool; notecard_configured is only latched after a successful response.

bool defineTemplates(void) {
    J *req = notecard.newRequest("note.template");
    if (!req) return false;
    JAddStringToObject(req, "file", SUMMARY_NOTEFILE);
    J *body = JAddObjectToObject(req, "body");
    char key[28];
    for (int t = 1; t <= 4; t++) {
        snprintf(key, sizeof(key), "t%d_wh",       t);
        JAddNumberToObject(body, key, 14.1);  // estimated interval energy, TFLOAT32
        snprintf(key, sizeof(key), "t%d_demand_w", t);
        JAddNumberToObject(body, key, 14.1);  // peak 15-min demand-window average, TFLOAT32
    }
    // fault_mask: 4-bit nibble per channel, packed uint16 as TFLOAT32.
    // 0 = all channels clean; non-zero = at least one FAULT_* flag set.
    JAddNumberToObject(body, "fault_mask", 14.1);
    J *rsp = notecard.requestAndResponse(req);
    bool ok = rsp && !notecard.responseError(rsp);
    notecard.deleteResponse(rsp);
    return ok;
}

7.8 Key code snippet 2: sleeping until the Notecard raises ATTN

cxSleepUntilAttn() asks the Notecard to hold ATTN low for exactly cfg.sample_interval_sec seconds, confirms the pin has gone low, and puts the STM32 into STOP2. The host stays alive, so the card.attn call is a request whose result is checked rather than a fire-and-forget command. When the Notecard raises ATTN, loop() continues with the next sample and the AppState struct — accumulated estimated Wh, peak demand, demand-window progress, and summary epoch — is exactly as it was left, because it never left RAM.

// In setup() — arm the ATTN wake once
cxSleepBegin();

// In loop() — one sample/report cycle, then sleep until ATTN rises
runCycle();

if (!cxSleepUntilAttn(notecard, cfg.sample_interval_sec, NULL, &dbgSerial)) {
    // The Notecard didn't take the sleep request, or ATTN never went low
    // (check the ATTN -> D5 jumper).  Keep the sample cadence and try again.
    dbgSerial.println("[sleep] ATTN sleep failed — waiting out the interval awake");
    delay(cfg.sample_interval_sec * 1000UL);
}

8. Build and Flash

Prerequisites: Arduino IDE (or arduino-cli on the command line) with the following installed via Boards Manager and Library Manager:

• Boards: STM32 MCU based boards (the stm32duino/Arduino_Core_STM32 core). Add the index URL https://github.com/stm32duino/BoardManagerFiles/raw/main/package_stmicroelectronics_index.json under File → Preferences → Additional Boards Manager URLs if it is not already listed.
• Libraries: Blues Wireless Notecard (from note-arduino), STM32duino Low Power, and its dependency STM32duino RTC.

Steps:

1. Clone or download the repo and open firmware/tenant_sub_meter/tenant_sub_meter.ino in the Arduino IDE. All four files in the firmware/tenant_sub_meter/ directory must be present in the same sketch folder.

2. Set your Notehub ProductUID in firmware/tenant_sub_meter/tenant_sub_meter_helpers.h by replacing the empty string on line 21:

   COPY

   #define PRODUCT_UID "com.your-company:your-app"

   Find this value in Notehub under Project Settings → ProductUID, or see Finding a ProductUID.

3. Set the DIP switch on the Notecarrier CX to HST so USB serial routes to the host MCU for programming.

4. Via Arduino IDE: Select Tools → Board → Blues boards → Cygnet, set USB support (if available) to None, put the host in DFU mode (hold BOOT, tap RST), then click Upload; see the build settings in §7.1. Alternatively upload over SWD with an ST-LINK on the debug jack.

5. Via command line (arduino-cli):

   COPY

   # Compile (usb=none disables USB CDC, which must be off for the host to sleep)
   arduino-cli compile \
     -b STMicroelectronics:stm32:Blues:pnum=CYGNET,usb=none \
     firmware/tenant_sub_meter/
   
   # Find the port (replace below with the value reported)
   arduino-cli board list
   
   # Upload
   arduino-cli upload \
     -b STMicroelectronics:stm32:Blues:pnum=CYGNET,usb=none \
     -p /dev/cu.usbmodem* \
     firmware/tenant_sub_meter/

   On Windows the port appears as COM3, COM4, etc.; on macOS or Linux it is typically /dev/cu.usbmodem* or /dev/ttyUSB*.

6. Open a serial monitor at 115200 baud on the ST-LINK virtual COM port (the Notecarrier CX debug jack; see §7.1) to verify the sketch boots. On first power-on you should see the Notecard configuration exchange (hub.set, template registration), followed by the first channel sample lines. After the initial cycle the host enters STOP2 for sample_interval_sec and the serial output goes quiet — that silence is expected; the Notecard raises ATTN to wake the host for the next sample and loop() resumes in place.

7. Once the device appears in Notehub and the first meter_summary.qo Note arrives, flip the DIP switch back to NC for production operation.

If compilation fails with a PRODUCT_UID is not defined warning, revisit step 2. If upload fails with a "board not found" error, check the DIP switch position and USB cable.

9. Data Flow

Collected. Every sample_interval_sec (default 5 minutes): 2000 sequential interleaved V+I ADC sample pairs per active tenant channel. The V×I cross-product yields active power (W); RMS current (A) is computed as an intermediate value and is available to callers but is not transmitted in the current firmware. Adding per-channel RMS current to the summary Note is a straightforward future extension. See §11.

Accumulated on-device. Between summaries: per-tenant estimated Wh (sum of W_sample × sample_interval_sec / 3600); per-tenant peak 15-minute blocked-average demand W (highest completed DEMAND_INTERVAL_SEC window average; see §7.3); per-tenant hardware-fault flags (OR of FAULT_* bits across all wakes in the period).

Transmitted — hourly (canonical record). meter_summary.qo: one Note per summary_interval_min (default 24 per day). Template-encoded, queued and flushed in the Notecard's periodic outbound sync. Carries t1_wh–t4_wh (estimated interval energy in Wh), t1_demand_w–t4_demand_w (peak 15-minute blocked-average demand W), and fault_mask (combined per-channel fault bitmask) for each tenant. Note is not sync:true; it batches with the periodic radio window. These Notes are the authoritative energy record and the sole input for Notehub-side monthly aggregation.

Routed. Notes flow through Notehub to whatever downstream the project routes specify. A typical deployment routes meter_summary.qo to a time-series database as the primary billing and analytics record. Monthly per-tenant totals are derived by summing t*_wh across the billing period's hourly events in Notehub — either via the Event Query API or a downstream aggregation pipeline. Notehub stores every Note durably; no device-side aggregate is needed to reconstruct any monthly total.

10. Validation and Testing

Expected cadence in steady state. A correctly-operating unit generates one meter_summary.qo event per hour. The first _session.qo should appear in Notehub within two minutes of first power-on; the first meter_summary.qo typically arrives in the initial session shortly after the first sample cycle completes.

Calibration — before deployment. Each Rogowski coil and integrator combination has a unique sensitivity. Before using readings for billing:

1. Load one tenant circuit to a known stable draw (e.g., 10 A measured by a calibrated clamp meter).
2. Wait one hour for a meter_summary.qo to arrive in Notehub.
3. Calculate expected energy: loaded_amps × hours_run × 120_V = Wh_expected. For 10 A @ 1 hour at 120 V ≈ 1200 Wh.
4. Compare t*_wh in that hour's summary against expected. The ratio is your per-coil correction factor.
5. Update rogowski_amps_per_volt in Notehub Fleet Environment: (default_400 × expected_wh) / reported_wh. Push the change; device fetches it on next inbound sync (default: 2 hours).
6. Repeat for a second load point (e.g., 20 A) to confirm linearity. Repeat for each active tenant channel.

Skipping calibration leaves you with readings that may be off by ±10–30% depending on coil model and integrator tolerance.

Using Mojo to validate power behavior. The Blues Mojo coulomb counter connects inline between the 5 V supply output and the Notecarrier CX +VBAT input and reports cumulative mAh to the Notecard over its Qwiic connector. Published current figures from the Notecard low-power design guide and the MBGLW datasheet:

The dominant energy cost is the once-per-hour cellular session. Because the supply is line-powered, absolute mAh is less important than the shape of the current trace:

• Healthy trace: near-zero baseline, brief (<1 second) micro-spikes every 5 minutes as the host wakes and samples, one 10–20 seconds burst at 80–150 mA once per hour.
• Host not sleeping: continuous 15–30 mA baseline. The host is not entering STOP2 — check the ATTN→D5 jumper (see §5) and confirm the firmware was built with the usb=none board option (USB support set to None).
• Cellular sessions running long (60+ s): correctly-spaced hourly bursts but each running long. Usually a weak-signal site; check antenna placement or move closer to a window.

Troubleshooting

If a problem is not on this list, visit the Blues community forum; it's the fastest path to a second pair of eyes on a specific Notecard + sensor bring-up issue.

11. Limitations and Next Steps

This is a reference design for proportional tenant allocation, not a certified utility sub-meter, and that distinction is deliberate. The list below makes explicit where the design's measurement model, the agency-listing of the bench-grade voltage transducer, and the single-phase assumption draw the line, then points at the production paths for the cases that need more.

Simplified for this Proof-of-Concept

The simplifications below are deliberate scope choices; each marks where the measurement model, the bench-grade voltage transducer, or the single-phase assumption draws the line for a proportional allocation bridge rather than a certified meter.

Allocation-grade estimation, not certified metering. This design measures estimated interval energy by taking one ~200 milliseconds active-power snapshot per sample_interval_sec and multiplying it by the full interval duration. The firmware does not continuously integrate power between wakes. This approach is accurate for constant or slowly-varying loads but may over- or under-state energy on bursting or cycling loads. For internal bill-back between tenants in the same building it is an appropriate allocation method. It is not suitable for applications where accuracy is subject to regulatory oversight, for example, utility-tariff sub-metering subject to accuracy standards — without replacing the measurement front end with a dedicated simultaneous-sampling energy-metering IC or a certified pulse-output sub-meter interface.

Single-phase voltage reference shared across all channels. The voltage transducer provides one voltage waveform used as the reference for all four current channels. This is accurate when all tenant circuits derive from the same phase leg of the building supply — common in small commercial buildings with a single-phase 120 V service. In a split-phase (120/240 V) building where tenants may be on different legs, or in a three-phase building where tenant feeds come from different phases, the phase relationship between the shared voltage reference and a tenant's actual line voltage introduces a power-factor error in the real-power calculation. For buildings with known multi-phase distribution, a dedicated voltage sensor per phase (and per-phase V×I pairs in the firmware) is required for accurate real-power metering across all tenants.

Rogowski coil and integrator calibration. The rogowski_amps_per_volt default is a starting point derived from the integrator component values and a typical coil's mutual inductance; the actual value varies by coil model, production tolerance, and board layout. Calibration against a reference clamp meter at a known load is required before the readings are used for billing. The volt_scale default likewise assumes the ZMPT101B module's standard burden; different burden resistor values require a different constant. Both are tunable via Notehub environment variables without a reflash.

t*_demand_w matches the 15-minute demand-charge window but energy is still estimated. The t*_demand_w field reports the peak 15-minute blocked-average demand (average watts over the highest completed 900-second window in the summary period). This matches the most common North American utility demand-charge billing interval and is a useful load-profile indicator and a reasonable input for internal bill-back proportioning. However, the underlying energy is still derived from periodic power snapshots, not a continuous integral. For 30-minute demand tariffs, change DEMAND_INTERVAL_SEC to 1800 in the firmware and reflash.

Runtime sensor-fault detection is best-effort, not certified. fault_mask currently reflects three categories of hardware condition: FAULT_BIAS_RANGE (ADC DC offset outside the expected 1.40–1.90 V half-rail window), FAULT_SATURATED (centered-signal RMS approaching the ADC rail, trips when RMS exceeds the equivalent of 85 % of the 1.65 V half-rail peak amplitude, ≈ 0.99 V RMS at the ADC pin), and FAULT_VOLTAGE_REF (shared voltage-reference path bias, saturation, or implausibly low line voltage, propagated simultaneously into every active channel's nibble). Low or absent current is intentionally excluded from fault_mask: a legitimately unloaded tenant circuit produces the same near-zero current reading as a disconnected Rogowski coil, so treating it as a hard fault would quarantine valid zero-usage billing intervals. These checks can catch gross hardware faults but cannot distinguish a genuinely zero-load tenant from a silently failed sensor. A production deployment that requires more rigorous fault coverage should add a commissioning-time baseline capture and cross-check fault_mask against historical baselines in the downstream system.

Single-phase circuits only. Each Rogowski coil wraps one conductor. A split-phase (120/240 V) tenant feed requires two coils per tenant; a three-phase feed requires three. This design uses A0–A4, leaving one remaining on-board analog input (A5) before an external analog multiplexer or ADC is required.

Electrical safety and code compliance. This design is a proof-of-concept. Panel installations in commercial buildings must comply with applicable electrical codes (NEC in the United States, CEC in Canada, etc.), be performed by licensed electricians, and in some jurisdictions require a permit. Sub-metering equipment used for tenant billing may also be subject to accuracy and certification requirements under local utility or metering regulations. Confirm local requirements before deploying.

No local display. The firmware provides no real-time readout at the panel. A small OLED on I²C or a status LED driven by a firmware threshold check would be a straightforward addition.

Mojo is bench equipment. The firmware does not read the Mojo's coulomb-counter register over Qwiic. Adding a runtime mAh field to the summary Note is a straightforward extension if fleet-level power telemetry is valuable.

Production Next Steps

Taking the bridge toward a production rollout means calibrating each installed coil, replacing the bench-grade voltage front end with an agency-listed transducer, and wiring the hourly stream into a billing platform. The following extensions are the natural progression.

Per-channel calibration workflow comes first: at commissioning, capture a reference reading alongside a calibrated clamp meter to establish the correct rogowski_amps_per_volt for each installed coil, then push the trimmed value as a per-device environment variable in Notehub.

Production voltage sensing (future revision). The ZMPT101B module used in this design is bench/prototype only; it is not agency-listed for permanent installation inside a commercial panel enclosure. Before deploying to a live panel, replace it with an agency-listed isolated AC voltage transducer. Two substitution paths are available depending on the device chosen:

(a) AC waveform output (e.g., a Yokogawa or Verivolt panel-mount PT with a 0–3 V AC output): wiring-only drop-in. The existing measureChannel() waveform-RMS algorithm works unchanged. Retain the 10 µF series AC-coupling capacitor and the 100 kΩ / 100 kΩ half-rail bias divider on A4 exactly as wired for the ZMPT101B. Calibrate volt_scale to the new transducer's rated output sensitivity (line V RMS ÷ ADC-pin V RMS at rated input). This is the simplest production upgrade path.

(b) 4–20 mA DC output (e.g., NK Technologies VTR5-420-24L-DIN, UL Listed, 120 VAC input, 24 VDC loop-powered, DIN-rail mount): requires the following additions and changes not in the current BOM and firmware:

• 24 VDC loop supply: the transducer is loop-powered and requires a dedicated 24 VDC supply (e.g., MeanWell IRM-10-24 or equivalent DIN-rail 24 V module); the 5 V AC/DC supply already in the BOM cannot power it.
• 150 Ω ±0.1% precision burden resistor: wire from ADC A4 to GND. This converts the 4–20 mA loop current to 0.60–3.00 V DC, staying within the host's 3.3 V ADC limit. Remove the 10 µF AC-coupling capacitor and the 100 kΩ / 100 kΩ half-rail bias divider from A4; the signal is DC and needs neither AC coupling nor a bias network.
• Loop wiring: connect the 24 V supply (+) → transducer VIN; transducer IOUT → burden resistor (+) → ADC A4; burden resistor (−) → GND → 24 V supply (−). Verify transducer polarity against the device datasheet.
• Firmware adaptation: measureChannel() must be updated to replace the 2000-sample AC waveform loop with a single DC ADC read, compute instantaneous line voltage as (adc_v - 0.60) / (3.00 - 0.60) × V_rated_full_scale (where 0.60 V corresponds to 4 mA and 3.00 V to 20 mA), and pass that DC voltage value through the power and fault checks in place of v_rms_adc. The voltage-path bias and saturation checks must also be adapted — DC offset is no longer meaningful on A4, and the plausibility check should compare the derived line-voltage value directly against VOLTAGE_MIN_V_RMS. Implement the 4–20 mA path as a #define-gated alternative to the existing waveform path so both options can be compiled and bench-tested without modifying the core algorithm.

Per-phase voltage sensing extends the design to buildings with multi-phase distribution: replace the single voltage transducer with one per phase and extend the firmware to pair each current channel with its corresponding phase voltage.

Notecard Outboard DFU lets firmware be updated across the entire fleet without a truck roll to each panel.

Integrate with a billing platform by configuring a Notehub HTTP route that POSTs each meter_summary.qo to the analytics and billing database. Configure the downstream system to reject or quarantine any meter_summary.qo where fault_mask != 0 and fall back to manual estimation for the affected period. Monthly totals are derived by querying the Notehub Event Query API and summing t*_wh for each device over the billing period.

Capture a commissioning baseline on first installation by recording the ADC DC offset for each channel with all loads off and storing the values in Notehub device metadata. The downstream system can compare fault_mask flags against these baselines to distinguish genuine zero-load readings from sensor faults.

12. Summary

The landlord who used to allocate the utility bill by square footage now allocates by actual use. Four Rogowski coils with active per-channel integrators and a voltage transducer turn the panel into a landlord-controlled, tenant-invisible metering bridge: sampling every five minutes, computing real power on-device, and delivering compact hourly summary Notes through Notehub with estimated interval energy (Wh), 15-minute blocked-average demand (W), and a per-channel fault bitmask per tenant. Monthly per-tenant totals come from querying that hourly stream in Notehub; the device never has to compute a billing figure. The readings are allocation-grade estimates accurate for proportional charge-back inside the building, not utility-tariff revenue grade — which for most landlords is exactly the right trade-off, because the alternative is either no allocation at all or a full certified sub-meter retrofit at multiples of the cost. And the deciding architecture isn't the hardware, it's the network: cellular is the only data channel that belongs entirely to the landlord and is structurally inaccessible to tenants. Any WiFi-based path puts billing data on a network the tenant can reach, and that's a dispute waiting to happen. The Notecard's landlord-owned SIM resolves the dispute by design: the panel stays dumb, the tenants stay unaware, and the allocation data arrives every hour without anyone having to think about it.