Issues with ATM90E26 Energy Metering IC Calibration

[adharsh96]

Hello everyone,

I've been working on a project involving a PCB designed with the ATM90E26 single-phase energy metering IC with a current clamp. I'm using an Arduino code to read the values from the ic via SPI using this Arduino library. Following the application note, I've calibrated the IC to measure current and voltage accurately, and the values appear correct. Additionally, I calculated the PL_CONSTANT value based on my hardware characteristics to achieve 1000 pulses per kWh. However, I'm encountering significant differences in the energy readings, with errors ranging from 10% to 30% (it's not consistent) .

Here is my setup for measuring the error:
Load: A resistive load of 1000W
Reference Meter: A class 1 energy meter
Method: I let the setup run for 10 minutes and compared the energy measured by both the ATM90E26 and the reference meter.

Despite these, the energy readings from the ATM90E26 are consistently off.

Am I missing something in my calibration process? What are the best practices or steps or standard ways to accurately calibrate the ATM90E26 or any similar energy metering device?

Any advice would be greatly appreciated!
Thank you in advance!

Here are some references:
ATM90E26 Datasheet
ATM90E26 Application note
Additional reading

[TizianoHV]

Hi,
10 minutes may be too short for the class 1 energy meter. What values ​​are you getting? What resolution do they have? How often do they refresh the readings?

I would make sure to "synchronize" the two meters and do a longer test.

For longer tests, having to dissipate 1kW is not ideal. If you have a step-down transformer you can use it together with a 100W load and get a higher "equivalent" power. Horrible schematic below.

[coppice]

Hello everyone,

I've been working on a project involving a PCB designed with the ATM90E26 single-phase energy metering IC with a current clamp. I'm using an Arduino code to read the values from the ic via SPI using this Arduino library. Following the application note, I've calibrated the IC to measure current and voltage accurately, and the values appear correct. Additionally, I calculated the PL_CONSTANT value based on my hardware characteristics to achieve 1000 pulses per kWh. However, I'm encountering significant differences in the energy readings, with errors ranging from 10% to 30% (it's not consistent) .

Here is my setup for measuring the error:
Load: A resistive load of 1000W
Reference Meter: A class 1 energy meter
Method: I let the setup run for 10 minutes and compared the energy measured by both the ATM90E26 and the reference meter.

Despite these, the energy readings from the ATM90E26 are consistently off.

Am I missing something in my calibration process? What are the best practices or steps or standard ways to accurately calibrate the ATM90E26 or any similar energy metering device?

Any advice would be greatly appreciated!
Thank you in advance!

Here are some references:
ATM90E26 Datasheet
ATM90E26 Application note
Additional reading

Two likely issues:

1- You are only calibrating close to unity power factor, and simply calibrating the gain. What about the phase shift? Real loads can have horrible power factors. If you don't properly correct for phase alignment between the voltage and current channels you can see horrible results at 0.5PF, when they look great at 1.0PF. Use your existing load as a test at around 1.0PF and find another load at around 0.5PF. Between those you should be able to check the phase correction.

2 - How well have you investigated your design for things like crosstalk between the voltage and current channels? With a 1kW load you are strongly exciting both ADC channels. Are your highly inaccurate measurements at low power? If so, you may well have crosstalk of the voltage signal into the current ADC. This is a very common problem. At low loads you are looking for a very small signal in the current ADC, while a large voltage signal is present not far away. You need to keep the signal to the current ADC very well balanced. Check the layout of the ADC input filters. People are often sloppy there, and inject differential noise into the current ADC.

[adharsh96]

The reference meter I have has a 10Wh resolution, and I'm getting almost 160-170Wh(line voltage is not stable) on a 1kW load when running for 10 min.
Also, I'll do a longer test and see what values I get.

[adharsh96]

At this moment I'm not considering the phase shift, I'm trying to calibrate it and test it on resistive loads with unit power factor. Also, the phase angle is calibrated in a different step, planning to do that after getting this right.
And crosstalk or other noise doesn't seem to be an issue here because the current/voltage readings from the device are proper. I get reliable readings on loads greater than 10w with a 100A CT.
For some reason, the current/voltage and energy are calibrated separately on this IC

[Kleinstein]

With 160 Wh and only 10 Wh of resolution for the ref. meter the result will never be very accurate.

There is also the possiblity to have a problem with current sensor or votlage divider / power supply (e.g. current from the supply side interfering with a divider). So there ways to srew up the circuit PCB.
With some current clamps a DC background (e.g. from the grid or load) could be an issue to.