energy meter ICs and dynamic range

[int1]

I am trying to understand the specs for energy meter ICs. Microchip, Cirrus and Analog Devices for example have utility meter ICs
that claim 1000:1 dynamic range at 0.1% error.

This means 1000W @ +-0.1W or 1W @+-0.001W
So does that really mean the meter can actually detect between 0.001W to 1000W i.e a dynamic range of 1000000:1.
The Cirrus IC claims 4000:1 0.1% error

That is impossible given the noise performance shown in the datasheets (75~80dB).

https://www.microchip.com/wwwproducts/Devices.aspx?dDocName=en028191
https://www.analog.com/en/analog-to...rgy-measurement/ade7753/products/product.html
https://www.cirrus.com/en/products/cs5490.html

 

[FvM]

That is impossible given the noise performance shown in the datasheets (75~80dB).

You don't consider the small bandwidth of the energy measurement which is probably in a mHz range,

 

[int1]

You don't consider the small bandwidth of the energy measurement which is probably in a mHz range,

The problem is to calculate real power the instantaneous voltage and current must be multiplied and then averaged or low pass filtered.
Any inherent noise in the V or I samples will show up in the average power calculation.

The only way I can see to band limiting is put V and I samples through a kind of comb filter where only the fundamental and harmonics are extracted. But I cannot find any references to an applications doing this. ADC input bandwidths are typically 1kHz to 5kHz range which covers all the important harmonics of 50/60Hz line.

 

[FvM]

Dynamic range in the energy meter specification means current dynamic. Voltage will be nominal mains voltage plus/minus expectable variations. As a result, all current spectral components except a small band around the fundamental will be filtered out when multiplying current and voltage and integrating over full 50/60 Hz periods. Low-pass filtering will further reduce the effective noise bandwidth of instantaneous power measurement to a few Hz. In the energy integral, you finally have mHz noise bandwidth.

 

[int1]

Ok I finally got it, the voltage measurement is relatively spectral pure with a very high S/N ratio. The power meter then effectively becomes a product detector or synchronous demodulator. Low filtering the 'mixer' output limits the current channel bandwidth, increasing S/N ratio and achieving the > 120dB range needed.

Adam

 

[Jibby Benjamen]

It is not to detect from .001w but in a 1000W it could be +- 0.1W over 1000W.