Circuit-Level Monitoring and Smart Panels
Why this matters
More customers now have a monitor in their panel, or a whole panel with metering built into each breaker, and they will show you its app before you have your tools out. The data can save an hour: it shows when a load ran, how often, and whether its pattern changed. It can also send you after the wrong appliance, because a number on a phone looks precise whether or not it is. This card explains what these devices actually measure, where their accuracy runs out, and the configuration errors that produce confident nonsense, so the data shortens a diagnosis instead of steering it.
Reading the data is not working in the panel
Two different activities get confused here. Reading the app, the history and the alerts is something anyone can do from a phone. Installing a monitor, moving a sensor, relabelling circuits inside the panel or checking a reading at the conductors is work inside an energized panelboard, and it is for a qualified electrician: the panel is de-energized and locked and tagged where the work allows, and proved dead with a meter proved on a known live source before and after, per NFPA 70E-2021, 120.5, with work practices at 29 CFR 1910.333(b)(2). Many monitor sensors go on conductors at the service, where there is no breaker upstream in the house to open, so this is not a job for a tech from another trade with a screwdriver. Everything below is about reading the data.
What the device actually measures
Nearly every circuit monitor builds its numbers from two measurements.
Current, by a current transformer, or CT: a ring or a hinged clamp that goes around one conductor and produces a small signal proportional to the current through it. The CT does not touch copper. It measures the current in the one wire it surrounds, and nothing else.
Voltage, through a connection to the bus or a breaker. Better monitors sample voltage and current together, many times per cycle, which lets them compute real power in watts: the power actually doing work. Simpler monitors measure current only and assume a voltage, which gives apparent power, volts times amps, and treats it as watts.
From power over time the device computes energy in kilowatt-hours, which is what the utility bills. See related: Reading Energy Data Without Misleading the Customer, which owns the kW versus kWh distinction and what can be claimed from it.
Smart breakers and smart panels do the same thing with the sensing built into each breaker. The physics is identical; what changes is that every circuit gets its own sensor, so there is no question which conductor the CT was put on, and the labels are set in the app rather than written in a panel schedule.
The geometry that halves a reading
A 240 V load, such as a dryer, a water heater, a well pump or a heat pump, is fed by two hot conductors, one from each leg of the service. The same current flows out on one and back on the other.
LEG A (120 V to neutral)
panel --------[CT]---------+
| 240 V load
panel ---------------------+ (dryer element)
LEG B (120 V to neutral)
one CT sees the current in one leg only
A monitor with one CT on one leg can get the power right only if it is told the circuit is 240 V, so it multiplies the leg current by 240. If it is configured as a 120 V circuit, it multiplies by 120 and reports exactly half the true power. If it has a CT on each leg and is also told the circuit is 240 V, it can count the same current twice and report double. A half or a double is the signature of this configuration error, and it is far more common than a faulty sensor.
Accuracy: what the specification means
Every CT has an accuracy specification, and before you use a reading you need two things from it: what the error is a percentage OF, and what kind of error it is.
The basis. A typical specification reads something like "plus or minus 1 percent of reading" over a stated range, say from about 10 percent of the CT's rated current up to its full rating, with a fixed error floor below that range. The numbers are illustrative; take the real ones from the datasheet for that sensor. What matters is the shape: inside the range the error shrinks with the reading, and below it a fixed floor dominates. On a CT rated for 200 A at the service, 10 percent is 20 A. A refrigerator idling at about 1 A sits far below that range, where a fixed floor of even half an amp is half the reading. So a whole-house CT can be trusted on a dryer cycle and cannot be trusted on standby loads at all. A 20 A circuit CT has a range starting near 2 A and is much better on small loads.
The character. Most CT error is a fixed gain error: that sensor reads a few tenths of a percent high or low, consistently. That has a useful consequence. Comparing the same circuit against itself, last month against this month, on the same CT, the gain error scales both figures equally, so the comparison is about as good as the percentage itself. Comparing one circuit against another means two different sensors, two different gain errors, and possibly two different ranges, so small differences between circuits mean little.
Current-only monitors and motor loads
A monitor that measures current and assumes voltage overstates any load whose current and voltage are out of step, which is most motor loads. The ratio of real power to apparent power is the power factor, between 0 and 1. A resistive load such as a heating element is close to 1. A small motor running lightly loaded can sit near 0.7.
Say a 120 V pump draws 6 A at a power factor of 0.7. Apparent power is 120 times 6, which is 720 VA. Real power is 720 times 0.7, which is about 504 W. A current-only monitor shows roughly 720 W, which overstates real power by 720 divided by 504, about 1.43 times. On a monitor that samples voltage and current together, the same pump shows about 504 W. So if a customer's motor loads look surprisingly expensive, the first question is which kind of monitor it is.
Worked example: the dryer that seemed to have lost an element
A customer called an appliance tech about a dryer taking twice as long to dry. They had a circuit monitor and showed its history: the dryer circuit peaked at about 2.5 kW every cycle, and they had read that the dryer's heating element was rated 5.0 kW. Their conclusion was that the element was half dead.
Work the numbers before the dryer. A 5.0 kW element at 240 V draws 5,000 divided by 240, about 20.8 A. A monitor configured as a 120 V circuit, seeing 20.8 A on its one CT, reports 120 times 20.8, which is about 2.5 kW. So the reading the customer showed is exactly what a healthy element looks like on a monitor with this configuration error. The monitor's history also showed 2.5 kW on the first dryer cycles after it was installed, two years earlier, before the drying complaint began. A peak that never moved cannot be the cause of a symptom that started recently.
What the monitor did tell the tech, correctly: the element was cycling on and off normally, the cycles were longer than they used to be, and the peak had not changed. Long cycles at full element power point away from the element and toward airflow: a restricted exhaust vent makes the dryer run longer at normal heat. The tech unplugged the dryer before pulling it from the wall, checked the exhaust run, and found a crushed transition duct behind the machine and a vent hood packed with lint, cleared them, and the next cycle ran in a little over half the time.
The configuration fix went to the customer's electrician, who set the dryer circuit to 240 V in the app. The history then showed about 5.0 kW per cycle.
The failure mode. A tech who trusted the 2.5 kW figure would have replaced a good element, seen no change, and been back the next week.
What the data is good for, and what it is not
Trust it for timing and pattern. When a load ran, how long, how often, whether its cycle changed between two periods, and whether a load runs when it should not. These rest on the shape of the signal, not its absolute size, and a monitor gets shape right even when it has the size wrong.
Trust comparisons on the same circuit. This month against last, before a repair against after, on the same CT and the same configuration.
Do not trust absolute values on small loads, especially on CTs at the service. Do not trust a circuit label until it has been confirmed by switching that breaker off and watching the reading drop to zero. Do not trust short events: a monitor reporting one-minute averages cannot see a two-second inrush or a contact chattering, which is often exactly what a fault looks like.
And confirm before you condemn. Any finding that will lead to replacing a part gets confirmed with a meter at the load, by someone qualified to take that reading, before the part is ordered. The monitor suggests where to look. The meter decides.
How to tell a monitor is configured right
Three quick checks from the app alone, none of which requires touching the panel. First, a known resistive 240 V load, such as a water heater or a dryer on high heat, should read close to its nameplate wattage; half or double is the configuration error above. Second, the sum of the circuit readings should land close to the whole-house reading, with the gap explained by circuits that have no sensor; a sum well above the total suggests a load counted twice. Third, a circuit should never read negative unless it carries solar or a battery; a negative reading on an ordinary load means a CT installed backwards, which is a fix for the electrician.
References
- NFPA 70E-2021, 120.5, establishing an electrically safe work condition
- 29 CFR 1910.333(b)(2), deenergizing and lockout and tagging of electrical equipment
- Manufacturer datasheets for the monitor and its current transformers, accuracy and range specifications
- See related: Reading Energy Data Without Misleading the Customer
- See related: Using a Home Energy Monitor to Find a Problem Load