On 01/01/2016 08:24 PM, Rafael wrote: >> There are a lot of good reasons to tie one leg the transformer to ground >> besides to establish the safety ground and neutral as is common on the US. >> Intermittent faults to ground, with an ungrounded system, can cause the >> secondary of the transformer to fly way above absolute ground causing >> connected devices, or the transformer to suffer from insulation >> breakdowns. That's the extreme, but it can happen. > > Strongly agree! Do not float transformer secondary and have a PC > connected to it! This is not about an isolation transformer, it's about > stepdown transformer which is way different. Primary side of this step > down transformer needs to float in this case. In ideal situation, CT on > the primary side would be neutral and connected to local ground but it > is better to not use it; transformer is not ideal.
For good measure; I did not say that you should not ground the PC. Actually, I noted that it already is, but it is grounded at a _different_ point in the chain of connections. There is a reason why the PC /should/ be grounded and that has been underlined in the replies here. The question that seems to be popping up is _where_ to have the ground connected. Short explanation: You must ensure that a grounding connection does not carry current and is located at such a point where a symmetric coupling can be achieved. A bit longer explanation: See attached image. Outputs of the step-down transformer A and B are capacitively coupled to L1 and L2 through four parasitic capacitors. Connecting A to ground when C1A not equal C2A or connecting B to ground when C1B not equal C2B will create a current in the ground connection. CxA and CxB form capacitive voltage dividers, which will create a virtual zero point equal to neutral only if both capacitors are equal. Otherwise, you will have a potential differential and therefore a current through the ground connection from the secondary(*). An RCD measures the currents IL1 and IL2 where it must be ensured that IL1 + IL2 < threshold, where threshold depends on the particular RCD. When you cannot guarantee that the ground-current is zero, because the capacitive couplings are not equal, then you risk tripping the RCD. If you have no RCD, then you are creating an unbalanced circuit that will come to haunt you at another time (any imbalance in L1 and L2 will propagate through the system wreaking havoc). If you look at the DC side of the system, then you have a symmetric coupling from 0V (DC) to both A and B through Cf. The coupling stems from the rectifier system, which couples both A and B through the same circuit (which is why Cf is drawn twice). Therefore, connecting the output of the PSU's 0V (DC) line to ground will carry nearly(**) no current. This will protect the whole circuit from inadvertent potentials and calm the noise in the system. The /problem/ with having 0V (DC) at ground are again the rise of the dreaded ground-loops. If you have multiple PSUs all connected to ground and connecting the 0V (DC) lines through the signal return connections, then you have two distinct paths (a loop). That is where you need the galvanic isolator(s) to break the ground-loop. (*) JT has done a measurement that show a different voltage from A-to-ground and B-to-ground. Thereby establishing that the parasitic capacitances are not equal. (**) the parasitic capacitances are in series and the differentials are guaranteed less than 90 degrees shifted. The effective leakage is normally in the order of micro-amps and a few orders of magnitude away from tripping an RCD. >> One of way too many references on the web. >> http://ecmweb.com/bonding-amp-grounding/basics-bonding-and-grounding-transformers > I hope that nobody gets zapped before this highly charged and perhaps > the long(est) thread is over. Always have a bleeder resistor over your caps. Running a bleeder resistor over an email-thread helps too ;-) And then, coffee will help too. -- Greetings Bertho (disclaimers are disclaimed)
------------------------------------------------------------------------------
_______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users