Hi AP
To answer your questions:
A. I have attached the shunt connection figure as shown in user guide:
This figure shows that the shunt's ground terminal is connected to AGND_ADC1 ( which is separated fom AGND_ADC or GNDiso by just a ferrite bead). So, if you are using a shunt sensor, the phase will be connected to the ground of the ADE7912. If you have noticed that the recommendation is different, in any other section of the document, please feel free to point it out here and we will get it corrected.
B.I believe, by 'offset' you mean the DC offset of the ADCs. As there is no high pass filter (HPF) in this chip, the DC offset isn't removed from the signal chain. However, you must be able to remove it externally.
But coming back to the attenuation at 1MHz, you might need to rethink about using the 5 KHz anti-aliasing filter instead of 50KHz. Pondering about the same made me realize that the 26dB of attenuation might not be sufficent enough. The sigma-delta modulator has a noise-shaping profile as shown below( Refer to page 18-19 of the datasheet):
The digital LPF within the chip must help you with removing the higher frequencies to get a high resolution output from the signal obtained from the output of the ADC stage. But at the ADC level itself, it is the input anti-aliasing filter that you use ( on the current and voltage inputs of the chip) which will help attenuate the noise at higher frequencies. So, it is essential to have the filter corner set close to 5KHz so that when the noise of the modulator starts to pick-up, the filter would attenuate it comfortably.
Regarding your point about having no phase distortion on frequencies prior to 2.5KHz, I have a few more questions/suggestions:
1. The voltage and the current channels will have identical phase response as you will choose identical filter corners. So, if you are computing energy measurements, that shouldn't be a problem, I would think.
2. If you are doing some FFT analysis on current and voltage channels separately, would it be possible for you to compensate for the filter's response.
Also, kindly note that the chip has a sinc filter and a LPF (low pass filter) in its data path, which would introduce some delay as well. I did not know if you have taken this into account so just wanted to point it out.
Thanks!
Regards
hmani