Electrical · Power Transmission & Distribution
HVAC Cycling & Inrush Current Audit
Capture what a motor actually does in the first few hundred milliseconds after it starts. Measure the inrush current spike on a refrigerator or window air conditioner, compare it to the steady-state running load, and measure the duty cycle over a few hours. This is the project that teaches you why grid equipment is sized for events most people never see.
Safety: Use an AC line splitter so the clamp goes around an already-separated conductor. Never cut or strip a live cord, and never open your electrical panel.
Major
Electrical
Focus area
Power Transmission & Distribution
Total hours
12-18 hours
Why this project matters
Motor starting is the reason a lot of grid equipment is sized the way it is. A compressor that runs happily at 6 amps can pull 40 or more for a few hundred milliseconds at startup, and every device between that motor and the generator has to ride through it without tripping, sagging, or cooking.
This is also where protective device coordination comes from. A breaker has to be slow enough to ignore a legitimate motor start and fast enough to clear an actual fault. If you have measured a real inrush curve, time-current curves stop being an abstraction.
On the distribution side, inrush is why you see voltage flicker when a large air conditioner kicks on down the street, and why utilities care about how many units are on one transformer.
How it works
An induction motor at rest looks almost like a short circuit. There is no back-EMF yet because the rotor is not turning, so the only thing limiting current is the winding resistance and leakage reactance. That is the locked-rotor condition, and it is why the first moment of a motor start draws so much.
As the rotor accelerates it generates back-EMF, which opposes the applied voltage and drops the current. Within a few hundred milliseconds you are at normal running current. The whole event is over before a person notices anything beyond maybe a light dimming.
Inrush is also not the same every time. An AC waveform crosses zero 120 times a second, and where in that cycle the contacts close changes the DC offset of the resulting transient. Close at a voltage peak and you get one magnitude, close at a zero crossing and you get another. This is why you capture it several times and report a range.
Catching this needs speed. An RMS calculation averages over a window, typically 100 to 200 milliseconds, which is the same order as the entire event you are trying to see. The average smears the spike into a small bump and you conclude, wrongly, that inrush is overblown. Log raw samples at a few kilohertz and the spike appears exactly as it should.
Your sensor also has a ceiling. An SCT-013-030 saturates at 30 amps. A window AC or a fridge compressor can exceed that at startup, and when it does, your trace flattens into a plateau at the limit. That plateau is not data. It is your sensor running out of range.
What you'll need
- ESP32 development board (fast ADC, needed for this one)
- SCT-013-050 or SCT-013-100 clamp-on current sensor (the 30 A version will clip)
- AC line splitter with a separated conductor loop
- Burden resistor and capacitor sized for your sensor
- 3.5mm audio jack breakout board
- Breadboard, jumper wires, and a USB cable
Steps
- Set up the line splitter and clamp exactly as in the load profiler project, with no cord ever opened.
- Switch your firmware from RMS averaging to raw sample logging. You need the individual samples, not a smoothed value.
- Verify your actual sampling rate by timestamping a known number of samples. The rate you asked for and the rate you get are rarely the same.
- Confirm your sensor's current ceiling. If a start event flatlines at a constant value, you are clipping and need a higher-range sensor.
- Trigger a compressor start, capturing from just before the start through about one second after.
- Measure the first-cycle peak, then measure the steady-state running current once it settles.
- Repeat the capture at least five times. Inrush varies with where in the AC cycle the switch closes, so one capture tells you very little.
- Switch back to slower logging and record on and off cycles for three or more hours.
- Calculate duty cycle, average demand over the window, and the ratio of inrush peak to running current.
- Write up why a device that runs at 6 amps needs a circuit, a breaker, and a transformer that can tolerate many times that for a fraction of a second.
What usually goes wrong
Using an RMS sketch and concluding that inrush is a myth.
Do this instead: Standard energy monitor firmware computes RMS over a window of roughly 100 to 200 milliseconds, which is about as long as the entire inrush event. The averaging flattens the spike. Log raw ADC samples at a few kilohertz and analyze afterward.
Using a 30 A sensor and seeing a flat plateau at the top of the spike.
Do this instead: That plateau means you are clipping, not that the current held steady. The SCT-013-030 saturates at 30 amps and a compressor start can go well past it. Use the 50 A or 100 A version for this project, and keep the 30 A for the load profiler where currents are lower.
Capturing one start and reporting it as the inrush value.
Do this instead: Inrush magnitude depends on where in the AC cycle the contacts close, so it varies meaningfully start to start. Capture at least five and report a range with your maximum, the way a real test report would.
Trusting the sampling rate you asked for.
Do this instead: Loop overhead, serial writes, and SD card latency all eat into your real rate. Timestamp a known number of samples and compute the actual rate. If you claim 4 kHz and you are really getting 400 Hz, every conclusion downstream is wrong.
Forcing the compressor to start repeatedly by cycling power.
Do this instead: Short-cycling a compressor can damage it, since it needs refrigerant pressures to equalize before restarting. Wait at least five minutes between starts, or just let the appliance cycle on its own thermostat and catch the starts as they come.
From the field
Inrush is the kind of thing that stays theoretical until you watch it on a trace. A candidate who has actually captured a motor start talks about equipment sizing completely differently from one who has only read about locked-rotor current.
It also connects directly to the complaints a utility fields. When a neighbor says their lights flicker every time their air conditioner comes on, that is this phenomenon, showing up as a voltage dip across the impedance between them and the transformer. Measuring it yourself is how it stops being a formula and starts being a thing in the world.
Your Engineering Sister, PE (Electrical: Power), distribution planning
What to photograph for your portfolio
- The raw current waveform of a motor start, with the inrush peak and the settling tail visible
- Inrush peak and steady-state running current side by side on the same axis
- A multi-hour duty cycle chart showing the compressor turning on and off
- Your sampling rate math, showing why the rate you chose was fast enough
- The test setup with the clamp on the line splitter
Resume bullet starters
Copy one, then swap in your own numbers.
Captured motor inrush transients at [insert your sampling rate] samples per second, measuring a locked-rotor peak of [insert your result] against a steady-state running current of [insert your result]
Characterized compressor duty cycle over [number] hours and calculated an average demand of [insert your result] against a nameplate rating of [insert your result]
Quantified the inrush-to-running current ratio at [insert your result] and related it to utility equipment sizing and protective device coordination
If this is on your resume, expect to be asked
Your compressor runs at 6 amps steady state but you measured a 42 amp peak at startup. Why doesn't the 20 amp breaker on that circuit trip?
What they're listening for: An understanding that breakers are time-current devices, not instantaneous thresholds. The thermal element needs sustained overcurrent to act, and a few hundred milliseconds at 42 amps does not carry enough energy to trip it. A strong answer connects this to why coordination curves exist at all.
Why did your inrush measurement vary between captures?
What they're listening for: Point-on-wave switching. Where in the AC cycle the contacts close sets the DC offset of the transient. If they say the measurement was noisy without knowing why, they recorded data without understanding it.
How did you pick your sampling rate?
What they're listening for: Reasoning backward from the duration of the event. Anyone who says 'the default' has not engaged with the measurement problem.
Take it further
- Add a voltage channel and capture the voltage sag that happens during inrush, then estimate the source impedance from the sag.
- Compare a conventional single-speed compressor against an inverter-driven or soft-start unit and quantify the difference in peak.
- Compute what the duty cycle implies for the appliance's annual energy use and compare it to the EnergyGuide label.
- Plot your measured inrush against a published breaker time-current curve and show where it lands.
- Measure several motor loads and see whether the inrush-to-running ratio holds across them.