High current rectifier and commutation spikes

[Joe_echo]

I have been tasked to build a rig for heating a synchronous rotor to allow pressing of the coils. 70vdc 500A.  We have a 3ph supply going through a large 40kva variac to a stepdown transformer 415v to 50v, to supply a rectifier circuit using high current diodes (1200vrrm 650A).

It doesn't need to be perfectly smooth dc but I am including 30mF of capacitors on the output from the rectifier to help absorb any surges.

My concern is the commutation spikes as the diode pairs switch over causing large voltage spikes. The rotor has an inductance of 50mH. Resonant frequency of the LC circuit would be around 5hz so well below the 300hz ripple on the output.
The diodes have a reverse recovery time of around 300ns.

I'm unsure how to calculate the potential spikes and if I require either a snubber circuit or mov to protect the diodes.

[moffy]

I have built something very similar but the three phase variac fed a stepdown transformer which when rectified gave a max of 20V DC @ 1500A. The diodes were hockey puk style and well heatsinked, we didn't need any snubbers at all, but they were very high voltage diodes for the application. They were feeding the field of some large DC traction motors. The field inductance provided all the filtering necessary, with your 50mH you don't need the capacitors as 3 phase rectified is already pretty smooth. Just operationally always turn on or off the supply with 0V on the variac and slowly wind it up or down to suit. We never had a single failure of the field diodes.

[mtwieg]

So you're heating the rotor by running large currents through the coils? I don't see why you would get spikes from the rectifier in the first place, should be no need for any smoothing capacitors... maybe I'm picturing the schematic wrong.

Why not just use AC directly? Heat is heat.

[Joe_echo]

From my understanding, during the commutation when all diodes are off the inductor (rotor) will continue trying to pull the 200A leading to large spikes. There is also the potential for large spikes as the diodes reverse recover.

I'm happy to be proven wrong though.

We will absolutely be ensuring the variac is ramped down to 0v before disconnecting.

The reason for not using AC is that there will be a big press pushing the coils down as it heats up and the last thing we want is eddy currents being induced into the press equipment and someone getting a belt off it.

[mtwieg]

From my understanding, during the commutation when all diodes are off the inductor (rotor) will continue trying to pull the 200A leading to large spikes. There is also the potential for large spikes as the diodes reverse recover.

I'm assuming by "commutation" you're referring to the three phase rectifier, not the rotor itself commutating (nothing spinning, right?).

I'm not familiar with any issues with any mechanism for a simple three phase rectifier to generate "large spikes", even if there's significant series inductance on the phases or in series with the diodes (so long as no connections suddenly go open circuit, that is). I also don't see how reverse recovery would be an issue. Maybe a schematic would clear things up...

The reason for not using AC is that there will be a big press pushing the coils down as it heats up and the last thing we want is eddy currents being induced into the press equipment and someone getting a belt off it.

True, eddy currents might be an issue depending on how the press is constructed. I'd be very curious to see a photo of such a setup.

[Joe_echo]

Yes, the switching of the diodes, there will be no rotating.

It was raised by one of our other engineers but he made me worry about it, as he has said the spikes can be quite severe. But he hasn't been able to give any further details on it when questioned. I think it might be needless worry and it may only affect higher frequency supplies where the pulse time is faster than the reverse recovery time of the diode.

The schematic would just show a the phase transformer feeding 6 diodes in a full wave rectifier and an inductor and resistor in place of the rotor windings. Nothing complex.

[mtwieg]

Spikes and reverse recovery generally only are issues when there's hard switching involved, which won't happen if the rectifier is made just from passive diodes (not SCRs or IGBTs or FETs).

And in the case that there were spikes, you'd be better off using RC snubbers across each diode than putting one big capacitor on the DC output (especially when the load is an inductor).

[Benta]

AoE3, X-Chapters has a discussion of the issue (just 2 pages).
Chapter 9x.6 "Transformer + Rectifier + Capacitor = Giant Spikes!"

[moffy]

We saw no real spikes on our setup, the only deviation from the 3 phase ripple was the reverse recovery time of the currently conducting diodes, which gives a small dip in the output on transitions.The switching is not very rapid, 300Hz, so you have a relatively slow transitioning to another pair. You also have 1200V diodes in a 70V system, you have an enormous margin for any suspected spikes.

[trobbins]

The vintage method to suppress any spike across a commutating diode from Tx secondary leakage inductance, or a load disturbance, was an RC across each diode and was generically 0.1u and 22R on a pcb and located on the heatsink with the puck/module/stud with shortest practical connection.  With 1200V diodes you have some play to use MOV's, but they could degrade if continuous energy into them was above their 'no degradation' limit.

As an aside, some vintage applications like telco rectifiers had to be quite aware of such commutation transients, not only extending into the dc distribution, but also exiting on the AC mains and causing nearby synchronous clocks to go out of sync/time.

[coppercone2]

Anyone wanna put a part # for a good diode that can do something like 6V at 2000 amps and you know it works?

I was always curious about putting a rectifier on my brazing transformer (but its single phase)

overkill is preferred

for like etching, I don't think I need DC

[johansen]

A 300 amp power supply i bought on ebay 13 years ago had no snubbers, just a current doubling topology with 300 amp 60v shotkey rectifiers in parallel (so 4 total 300 amp rated dual diodes)

I do not understand why people think there would be high voltage spiked generated. Electrical noise yes, but the voltage cannot exceed the dc bus or the other diode conducts.

Tim williams likely has opinions on this. I once saw 10v spike 4 nanoseconds show up on an induction heater and i still do not Believe it was the forward recovery of the silicon carbide diodes but rather the oscope probe bandwidth and the inductance of a to-220 diode.


Anyhow. High voltage diodes are just killing you with excess losses. You only need 70vdc use 200 volt diodes

[trobbins]

Industrial applications can expose semiconductors to unintended spikes.  Even the OP's application could likely induce a destructive spike if the load abruptly disconnected, and the load capacitance was connected with sufficient leakage inductance.  The variac and step-down will have leakage inductance, which may be even more noticeable with a capacitive filter if that causes load current to become 300Hz pulses that reach zero.

[Terry Bites]

Monster SCRs can take a serious beating.
Look out for second hand modules.