One complaint always arrives with the same shape: the classic Shelly EM measures fine under load, and at rest it returns numbers that make no sense — power factor on the floor and a huge reactive figure for almost no consumption.
It is not an installation fault. It is the physical limit of the clamp, and there is a way to prove it in two minutes.
If yours is the opposite — reading too high — that usually is wiring, and it is covered in the Shelly EM is reading higher than expected.
The problem zone: between 1 and 1.5 A
On a 50 A clamp, that range is 2-3% of full scale. There, the measurement noise is comparable to the signal it is trying to measure.
And the lower the current, the worse the ratio. It is not that the unit fails below some threshold: the relative error simply grows without limit as you approach zero.
The cause: a fixed offset, not a phase error
The clamp’s electronics introduce an almost constant offset in the reactive measurement, one that barely depends on how much current is actually flowing.
At high load that offset is negligible against the signal. At low load, it is the signal.
Telling it apart from a wiring fault
Here is the check that settles it, and it is simple: note the reactive power at rest, then note it again with a heavy load running.
A real case:
| At rest | Under load | |
|---|---|---|
| Current | ~1.4 A | ~10 A |
| Reactive | −247 var | −318 var |
| Power factor | 0.61 | 0.99 |
Read those numbers carefully: the current went up sevenfold and the reactive power rose only 29%.
That is the proof. If this were a phase error — clamp on the wrong phase, or fitted backwards — the reactive would scale with the current. Staying nearly flat means it is not proportional to anything: it is a fixed offset.
Your installation is fine. The clamp is the limit.
Why your own equipment makes it worse
What tends to stay on at rest — router, NAS, server, chargers — runs on switched-mode supplies. Those do not draw a smooth current: they take it in pulses, full of harmonics.
The classic EM’s electronics measure that kind of current poorly, and especially so when it is also small. The two things it handles worst arrive together.
And no, it cannot be calibrated
This is where people lose an afternoon: the classic EM has no calibration. Not in its web interface, not in its API. That feature arrived with the Pro and second-generation models.
You can reflash it with Tasmota or ESPHome and adjust the gain, but that does not touch the reactive offset, which is the actual problem. And it voids the warranty. Not worth it.
So what to do
It depends what you are measuring:
- If you are metering the house, stay as you are. Across the range that matters — when something is genuinely running — the classic EM measures well. The at-rest error does not distort your bill.
- If you need accuracy at low loads too, the Pro EM-50 does calibrate and measures to class B.
- If what you care about is one appliance, a metering plug such as the Plug S Gen3 gives better resolution at a few watts than any clamp. A 50 A clamp will never be the tool for measuring 20 W.
From the console you can read the history and check for yourself whether the reactive follows the current or stays flat — this article’s diagnosis, without standing at the board.
The console and the app work with Gen2, Gen3 and Gen4.
How to use both, section by section: mobile app manual and web console manual.