I am very confused by why you want to "Earth" the DC mid-point. Perhaps you would like to explain?

Hi Mike, thank you for the explanation.

So if I use an isolation transformer at the input, I could ground the DC link mid-point? In that case, I would have L(+) and L(-), instead of L and N.

Yes, that will work.
Can  I ask what you intend to do with the grounded centre ?  Clearly as drawn there, any inequality in the capacitors will mean that the above and below sides of the DC bus cease to be equal, and any attempt to draw or inject current from or to that point is also likely to lead to centering imbalance issues.

Of course if the isolating transformer has a fully floating secondary, as you have drawn, so you are at equally liberty to ground any one point on the secondary side, and the voltages of the rest of the design will work as before and the voltages simply fly around that. However in terms of understanding circuit operation, ,and not having EMC problems from accidental antennas with large dV/dt some points are more sensible to earth than others ;-)

Mike.

mapj1 said:

Can  I ask what you intend to do with the grounded centre ?

I am trying to measure common-mode voltage and most of the papers i found [1] [2] [3] [4] use grounded mid-point so I am trying to replicate the same measurements. I assume they all use rectified three-phase voltage or variation to get a DC link, while I use rectified single phase.

When I tried to measure a common-mode with un-grounded converter (as shown in my initial post) I measured a 50 Hz signal on the oscilloscope superimpose to the common-mode voltage weveform.

I am hoping that by properly grounding the inverter (i.e., DC link mid-point) I will get the same results as in the papers (no superimposed AC signal). Because I don't see any other difference in my set-up

Ah well, in the lab when initial testing the inverter part in isolation the DC bus comes from a regulated DC supply  with all sorts of safety interlocks and fast shutdown, and is much more complex than the off-line version in the final product. Similarly when testing the off-line PSU part, initially the inverter may be replaced by another adjustable  load that is easier to manage.

Once it gets off the bench and is being integrated then pairs of probes and differential measurements are your friend.

By CVM I assume you mean the movement of the motor star-point - assuming a star load and not a delta, relative to some notional DC bus centre. Ideally it will be well centered over a full cycle but will precess at the cycle frequency and probably in reality be a few volts off-centre.

In your shoes I'd create that load centre from 3 equal loads on the phases,  and then AC couple it to one side of a diff-amp, and put the bus voltage, top, centre or bottom, to another probe, and then subtract.

Please take care, experimentally  this is not easy stuff to do safely - hence the first reference you gave - the IEEE paper - being about modelling it, not measuring it.

LT spice or something more powerful is your friend for initial design verification, it saves a lot of red faces.

Mike.