Zero Sequenec Current & Zero Sequence Voltage Derivation in Numerical Relays

In a numerical relay, the phase currents are sampled regularly, the residual current is simply calculated as i1 + i2 + i3, with signs +/-. So there is no need for a physical summing by residual connection and its input to a relay. I would, however, investigate whether i1 - i3 were sampled simultaneously.

However, I understand that residual current is less sensitive than a dedicated neutral or earth current CT [where possible] because the transient and saturation errors of three CTs limit sensitivity and speed, noting that they will usually handle different current levels and "point on wave" instants.

Remember that switching transients on power transformers differ greatly between zero voltage and peak voltage connection, causing flux doubling or worse, depending on residual flux and are always different between phases. This also affects errors of iron core CTs.

For voltage measurement and calculation of residual voltage, there does not seem to be any similar error.

I would not assume all numerical relays are the same. Electronically, one can have independent A/D for each phase, with synchronised sampling or one A/D switched between phases or three synchronised "sample and hold" analog circuits with a common A/D, apart from different samples per cycle of AC and A/D bit size. Once the numbers are digitised, the processing that can be done will depend on the processor power and, finally, how well the algorithms discriminate real faults from transients.