In a AC/DC 400Vac three phase 100kW power converter with half controlled SCR bridge we have a few case of 20 to 30kV overvoltage. Any other members knows the possible cause of this problem? We see that the AC source have some impact on this problem but we searche others cause.

would have to see schematic, but to get 30KV overvoltage it pretty much has to be a fast current change over an inductance.

I am not a professional. But, it sounds like you have polarity issues going on. The half controlled SCR bridge might need to be reverse biased. It sounds like your motor is inducing current as it is operating, which could account for the overvoltage. This is one of the principles my power consumption neutralization technology capitalizes on. Be careful or you will burn things up. Everything has to be sized for the current involved. Comforting huh?

SCR switching may require an RC circuit across the SCR device to reduce this type of effect. Also, what exactly did you use to make these measurements? High voltages at high frequencies require properly compensated high voltage probes to make a reasonably accurate measurement. In some cases a differential high voltage probe may be required.

Many thanks at all for your answer.
Ken, sure the reason is V=L*dV/dt.
Steven,reverse biasing of SCR,motor capacitance to the PE,are reduced through a three fase inductance.
George, we have snubbers across the SCR.Unfortunately we don't have the possibilities to measure but we deduce the HVLE from the descharge between point at 10mm distance.

Excuse me for two mistake: V=L*dI/dt not dV; fase is phase: all mistakes are due to the short time.

Graziano, Don't you have a separate SCR for each phase? The inductance of each phase might be affecting the inductance of another phase, because of the corroborative electro-magnetic induction effects each phase has on the other due to field warpage within the motor. I don't know for sure..just an idea.

Please tell me the value of the RC components, as well as the distance the RC components are from the SCR's?
Long lead wires or traced are to be avoided, for the same reasons

We have three separate diode+SCR modules, one for each phases, connected through low inductance bus bars.
RC components are short connected across the SCR.
Many thanks at all for any consideratios but the problem is very hard to solve. We have simulated all the parasitic elements by P-SPACE that shows over voltages of 2-3kV not 20-30kV that we deduce from air distances of 10mm (air dielectric strengh is 3kV/mm).

Sure, what are the values of the RC comp?

Snubbers value are R=22ohms C=100nF.