Type A rcd . EICR coding ? etc

I was perhaps simplifying too much - all non-DC sensing RCDs are a current transformer followed by a peak detector, with varying degrees of low pass filter. So long as the core is large enough not to saturate, in effect this rounds off the more violent changes in current waveform and then measures the peak to peak variation. The rectified waveforms do not have the same RMS to peak relationship as a sinewave, so it is not quite as simple as a 30mA RMS RCD for sinewaves being a 60mA RMS one for half wave rectified waves, but it is close - actually is is the half wave rectified waveform that has the same peak-to peak voltage as the sine wave of 30mA - which is pretty close to 60mA, but not quite.

However, if there is a DC term large compared to the ripple, so that the core saturates, then we are in different territory- more like the last few waveforms in that chart - here the variation needs to be large enough relative to the DC to reduce the total magnetisation for the part of the cycle where they are in opposition, back into the linear (i.e. very much not saturated,but transformer like) part of the magnetic material curve. So if you have a 100mA DC, you need to add enough AC to it that when the AC part of the cycle is in opposition to the DC it can more than cancel the DC - so a result of needing 115mA of AC to overcome 100mA of DC bias is perfectly believable, but the exact amount will depend on the magnetic core properties one that is far from saturating will not be desensitised at all - but it will never be worse than a core that was just saturating on the crests of the original 30mA sinewave and no DC.

As regards the VSD, can you measure the input filter capacitances L-E ; these will contribute to a steady 'leak' that will fool the RCD.  Larger VSDs , especially those designed for TT countries (Japanese designers tend to remember this better than EU ones)have  options to configure the filters for lower leakage - essentially the filter capacitors are wired L-N and then N-E but not L-E directly.

It is also possible that the inrush of charging the DC bus is actually only an L-N current but so much larger than the on load current that the RCD mis-reads it (if the L and N windings are not quite symmetrical on the core, and they never are, then that imbalance can cause a trip on high currents - say the paths are balanced only to 99.9% -  a 100A inrush looks like a 100mA imbalance.) 

I was perhaps simplifying too much - all non-DC sensing RCDs are a current transformer followed by a peak detector, with varying degrees of low pass filter. So long as the core is large enough not to saturate, in effect this rounds off the more violent changes in current waveform and then measures the peak to peak variation. The rectified waveforms do not have the same RMS to peak relationship as a sinewave, so it is not quite as simple as a 30mA RMS RCD for sinewaves being a 60mA RMS one for half wave rectified waves, but it is close - actually is is the half wave rectified waveform that has the same peak-to peak voltage as the sine wave of 30mA - which is pretty close to 60mA, but not quite.

However, if there is a DC term large compared to the ripple, so that the core saturates, then we are in different territory- more like the last few waveforms in that chart - here the variation needs to be large enough relative to the DC to reduce the total magnetisation for the part of the cycle where they are in opposition, back into the linear (i.e. very much not saturated,but transformer like) part of the magnetic material curve. So if you have a 100mA DC, you need to add enough AC to it that when the AC part of the cycle is in opposition to the DC it can more than cancel the DC - so a result of needing 115mA of AC to overcome 100mA of DC bias is perfectly believable, but the exact amount will depend on the magnetic core properties one that is far from saturating will not be desensitised at all - but it will never be worse than a core that was just saturating on the crests of the original 30mA sinewave and no DC.

As regards the VSD, can you measure the input filter capacitances L-E ; these will contribute to a steady 'leak' that will fool the RCD.  Larger VSDs , especially those designed for TT countries (Japanese designers tend to remember this better than EU ones)have  options to configure the filters for lower leakage - essentially the filter capacitors are wired L-N and then N-E but not L-E directly.

It is also possible that the inrush of charging the DC bus is actually only an L-N current but so much larger than the on load current that the RCD mis-reads it (if the L and N windings are not quite symmetrical on the core, and they never are, then that imbalance can cause a trip on high currents - say the paths are balanced only to 99.9% -  a 100A inrush looks like a 100mA imbalance.)