QWe are about to start producing 3phase industrial energy meters. It is a
multirange meter. But here we have some difficulties. The ADE7753 has odd
offset characteristic when changing applied voltage from 230V to 58V (voltage
on Ch2 is 750mV pp when 230V, and 180mV pp when 58V). It looks like that offset
curve is not linear - likely has two ranges. Most noticeable is when measuring
active energy with minimal currents, and changing voltage from 230 to 58V.
AThe issue is that rms reading is not only a measurement of the signal but also
of the noise.
RMS reading = Sqrt(rms_signal^2 + rms_noise^2)
i.e. the square root of the sum of the signal squared + noise squared.
When the signal amplitude is 3x the noise amplitude, then the quantities add in
proportion 9 to 1 (3^2 = 9) When the signal amplitude is 10x the noise
amplitude then the quantities add in proportion 100 to 1 (10^2 = 100)
The expression has the effect of reducing the effect of the smaller quantity
and exagerating the effect of the larger quantity in a non-linear fashion.
So if the offset is caused by noise then it will not scale linearly in the rms
measurement => noise will quickly become an insignificant part of the rms
reading as signal amplitude increases => equally, the noise will quickly
become a significant part of the rms reading as the signal amplitude decreases.
So as you change gain, the signal amplitude changes while the noise amplitude
stays roughly the same. Then the proportion of signal to noise in the rms
reading will change non-linearly.