Experiment record
AI Log Note: Six-Channel CT Monitoring Design for Dedicated Residential Circuits
A design-stage AI Log Note for monitoring four dedicated residential circuits with a six-channel CT energy-meter platform, leaving two spare channels while explicitly preserving unresolved Japanese 100/200 V voltage-reference and calibration questions.
Conclusion
A six-channel CT platform is a reasonable design candidate for the requested monitoring scope: four dedicated appliance circuits can occupy four channels while two channels remain available for later expansion.
The strongest prebuilt candidate identified in the earlier design work was the CircuitSetup Expandable 6 Channel ESP32 Energy Meter. Current official documentation confirms that the main board provides six CT current channels, uses two ATM90E32AS metering ICs, communicates with an ESP32 over SPI, and has maintained software paths including ESPHome/Home Assistant and EmonESP/MQTT.
However, this is a PLANNED design note, not an executed installation.
The retained project history does not establish that the board, CTs, or voltage-reference hardware were purchased, installed, commissioned, or calibrated. It also does not fully resolve the voltage-reference and phase-mapping arrangement required for accurate active-power measurement on the specific Japanese 100/200 V residential supply topology.
Accordingly, the correct reusable conclusion is:
The six-channel CircuitSetup architecture is suitable for further engineering evaluation, but the Japanese voltage-reference design, final CT selection, physical panel integration, and calibration must be verified before it can be treated as a completed power-monitoring system.
Why this note was created
A retrospective review of prior AI Experiment Log project conversations found several discussions that had not yet become public records.
Some were already covered by existing public experiments, including the SwitchBot environmental logger and hybrid-water-heater power investigations. The residential branch-circuit CT discussion remained distinct and reusable, but it had stopped at the design and parts-selection stage.
Publishing it as a design-stage AI Log Note preserves the useful architecture while avoiding a false claim that any electrical work or measurement was completed.
Observed design requirements
The retained project conversation directly established the following requirements:
- monitor a dedicated 200 V dishwasher branch;
- monitor a dedicated washing-machine branch;
- monitor a 100 V hybrid water-heater branch;
- monitor a 200 V, 30 A induction-cooktop branch;
- retain spare CT channels for later additions;
- send measurements over Wi-Fi to an existing Raspberry Pi, or potentially into a separately engineered Google Apps Script collection path;
- prefer a design built from commercially available modules rather than a completely custom analog front end.
These requirements are treated as OBSERVED inputs, not as evidence that the corresponding installation has been performed.
Verified CircuitSetup platform facts
The current official CircuitSetup documentation was checked again on 2026-08-16.
The vendor documents the following capabilities for the Expandable 6 Channel ESP32 Energy Meter:
- six current-transformer channels on the main board;
- two ATM90E32AS energy-metering ICs;
- SPI communication to a supported ESP32 development board;
- active, reactive, and apparent power, power factor, frequency, and related metering functions when the voltage/current sensing arrangement is correctly configured;
- current-transformer compatibility subject to the documented secondary-output limit;
- maintained ESPHome/Home Assistant and EmonESP-oriented software paths.
CircuitSetup states that a connected CT must not exceed 720 mV RMS or 33 mA secondary output at peak rated current.
This matters because the final CT cannot be selected only from the branch-circuit ampere rating. Conductor diameter, CT aperture, secondary type, burden-resistor arrangement, and the metering-board input limit must all agree.
Proposed six-channel allocation
The following allocation is a PROPOSED architecture, not an observed installation:
| Channel | Proposed load | State |
|---|---|---|
| CH1 | 200 V dishwasher branch | PROPOSED |
| CH2 | Washing-machine branch | PROPOSED |
| CH3 | 100 V hybrid water-heater branch | PROPOSED |
| CH4 | 200 V 30 A induction-cooktop branch | PROPOSED |
| CH5 | Reserve | PROPOSED |
| CH6 | Reserve | PROPOSED |
The value of the six-channel board is not merely that it can hold the four initial CTs. Keeping CH5 and CH6 unused at first leaves capacity for later measurement of another dedicated appliance, a generation circuit, or a comparison/reference channel without immediately replacing the metering front end.
Proposed data path
The most directly documented path is:
branch conductor
↓
current transformer
↓
CircuitSetup 6-channel metering board
↓ SPI
ESP32
↓ Wi-Fi
local collector / Home Assistant / MQTT
↓
long-term storage and analysis
For this project, an existing Raspberry Pi is a natural local collection target.
A Google Apps Script destination was also discussed in the original design conversation, but the CircuitSetup documentation reviewed here does not establish a native Google Apps Script upload path. Reaching Apps Script would therefore require a separately implemented and verified HTTP endpoint, relay, MQTT bridge, or equivalent integration.
That integration is not treated as completed by this note.
The unresolved Japanese 100/200 V issue
Current measurement alone is not enough to produce trustworthy active-power and energy values.
The meter also needs an appropriate voltage reference and correct phase/topology mapping. This is especially important when the monitored loads include both 100 V and 200 V circuits on a Japanese residential supply.
The retained conversation did not preserve enough verified detail to publish a final voltage-reference wiring design. The official CircuitSetup documentation describes use of AC-AC voltage-reference transformers and multiple voltage-reference capabilities, but a complete mapping for this specific Japanese installation still requires engineering verification.
Therefore this record deliberately leaves the following as UNKNOWN:
- the final voltage-reference transformer arrangement;
- how the meter’s voltage channels should be mapped to the actual service legs for every 100 V and 200 V branch;
- whether one or more isolated AC-AC references are required for the final configuration;
- the calibration constants required after installation.
This unresolved point is the main reason the design is not labelled deployment-ready.
CT selection remains open
No final CT model is asserted here.
The original discussion considered a prebuilt six-channel approach, but a defensible final CT BOM requires at least the following checks for each branch:
- breaker/branch maximum current;
- actual conductor outside diameter and available clearance;
- CT opening size;
- CT primary-current rating;
- secondary output type and maximum output;
- whether the CT has an internal burden resistor;
- compatibility with the selected CircuitSetup input configuration;
- calibration against a trusted reference instrument.
For example, the 30 A induction-cooktop branch needs a CT whose primary rating safely covers that current, while lower-current appliance branches may justify a smaller CT for better usable resolution. The exact model should be selected only after the physical conductor and branch ratings are confirmed.
Safety and publication boundary
This public note intentionally does not publish:
- distribution-board photographs;
- the exact breaker layout or conductor routing of a private residence;
- addresses or other residence-identifying details;
- private IP addresses or Wi-Fi information;
- account or device identifiers.
It also does not provide a claim that work was performed inside an energized panel.
Distribution equipment can expose hazardous voltages. This record is an architecture and evidence note, not an instruction to perform energized work. Any physical installation must follow applicable electrical safety and legal requirements.
Execution state
| Item | State |
|---|---|
| Six-channel architecture selected for evaluation | PROPOSED |
| CircuitSetup platform capabilities checked against official documentation | VERIFIED |
| Final CT models selected | UNKNOWN |
| Japanese 100/200 V voltage-reference design completed | UNKNOWN |
| Hardware purchase | NOT VERIFIED |
| Panel installation | NOT EXECUTED OR NOT VERIFIED |
| ESP32 firmware deployment | NOT EXECUTED OR NOT VERIFIED |
| Current/power measurements observed | NO |
| Accuracy calibrated against a reference meter | NO |
This state table is the most important part of the record: it prevents a parts-selection discussion from being mistaken for an experiment that has already produced measurement data.
Next verification steps
A future follow-up record should convert this design note into an executed experiment only after the following evidence exists:
- confirmed branch ratings and conductor dimensions;
- final CT model and burden/input compatibility for each channel;
- verified isolated voltage-reference design for the Japanese 100/200 V topology;
- enclosure and mounting design;
- ESP32 firmware/configuration actually deployed;
- Raspberry Pi or other receiver actually collecting timestamped measurements;
- calibration comparison against a trusted reference meter;
- repeat measurements demonstrating stable operation.
Until then, the correct lifecycle state remains PLANNED.
Sources
- CircuitSetup — Expandable 6 Channel ESP32 Energy Meter hardware/software documentation
- CircuitSetup — Expandable 6 Channel ESP32 Energy Meter product page
The language-independent factual record is stored as experiments/RESIDENTIAL-BRANCH-CT-DESIGN-001/experiment.json in the project repository.