Diode Bridge Failure Question

[johnnyw]

We have been using diode bridges (six diodes in a single package) that have been failing occasionally. Line side fuses blow just as current (load) is applied to the bridge. The load is very small (significantly under the rating of fuse and diode bridge) as it is only a pre-charge of a capacitor bank through a resistor. The fuses don't blow when the line side is energized and the load side is open, but only after the load is connected. Upon inspection during fuse replacement, there is always a diode in the bridge that has failed. We suspect that since the current draw is much smaller than the bridge and fuse ratings, that the diode bridge is actually damaged by and overvoltage condition which happens when the circuit was interrupted last, although we have not had the oppurtunity to catch any waveforms on a scope yet. There is adequate overcurrent protection via the fuses, but there is no overvoltage protection of the diode bridge, and the circuit is interrupted by vacuum relays always at very low currents (charging current to a capacitor bank with no load applied). We have purchased MOVs to install across the output of the diode bridges. Has anyone had a similar experience or is there another possible theory for the failures? Thanks for any insight.

 

[jraef]

Do you have a precharge resistor in the circuit? Without it the cap charging current inrush (dI/dt) may be too great for the diaode.

"Venditori de oleum-vipera non vigere excordis populi"

 

[Skogsgurra]

Since you mentioned a precharge resistor, I do not think that charging current is the problem. Diodes can take a lot of impulse (milliseconds) overload with no problems.

A very fundamental question is what voltage rating the diode bridge has. Remember that when the capacitors charge you will see capacitor voltage plus AC peak voltage across the bridge diodes. On the other hand, if voltage rating was a problem you would see failures also in operation and not only at switch-on.

Vacuum relays break current very fast and are known to produce severe overvoltage if there is any significant inductance in the current path. Where is the relay contact in the circuit?

A snubber or a MOV will probably solve your problem, but having a look at it with a fast digital scope is recommended.

 

[johnnyw]

The diode bridge is rated for ~30 A DC and the peak reverse voltage is 1600 V DC. The circuit operates at 600 V DC. The fuses are rated at 15 A AC ( applied line side of diode bridge) while the largest load is only 12 A DC. The pre-charge current is 6 A DC peak (100 ohm pre-charge resistor) and I have verified it is in the circuit prior to a bridge failure & blown fuses. The two relays (supply & return) are down stream of the diode bridge and the circuit inductance is in the transfomer winding that feeds the diode bridge. The relays are Kilovac K41A3 which I have not used before. They SHOULD only open under no load conditions, which is only charging current at 0.5 A peak.

An interesting side note is that the diode bridge is less expensive than either the fuses or the MOV!

 

[Skogsgurra]

I think that you should provide an alternate path for the current (however low it may be) when you open the relay contacts. Trying to stop secondary current in a transformer always produces an inductive overvoltage. A normal relay opens so slowly that an arc can form between the contacts and this arc limits the voltage. A vacuum relay doesn't form an arc so the voltage goes towards infinity. A MOV across the transformer secondary will probably cure the problem.

It is good that you do not blow a fortune every time the bridge fails. But a MOV doesn't need replacing so it will be cheaper - and less irritating - in the long run.

 

[buzzp]

I agree with Skogs. Except MOVs do fail. Each spike they absorb weakens the capacity. If this is something that is switching quite often you should consider other types of suppression methods for reducing the arc. Check out

http://www.pandbrelays.com

they have some good app notes to let you know what your up against. App number is 13C3236 amongst others.

 

[itsmoked]

I have used MOV's many times for these types of problems.

Keep in mind that MOV's last far, far, longer when they have a resistance to work with. They have very low resistance when triggered which is great for keeping the peak voltage down but lousy for actually dissipating the rogue energy. When deciding where to locate them in a circuit find or create a resistance for the "energy" to dissipate in. So put an appropriate resistor in series with the MOV and it will last forever.

Note that the resistor will increase the snubbing voltage but with your large PIV diodes and significantly lower working voltage this should do well.

BTW burning MOVs give off a really cool pink smoke, then end up looking like a burned marshmallow on legs.

 

[Skogsgurra]

Regarding life and death of MOVs

While it is true that a MOV can only take a limited number of surges when run close to its specification it is also true that they can take an unlimited number of surges when run with less surge energy. The numbers for a lifetime of one million surges or infinite number of surges do not differ much. A 250 V rms varistor that can stand 1 million surges at 25 A 20 us can take an infinite number at 15 A. And going up one size changes the numbers to 50 A and 25 A. So the problem with reduced life-time can be easily solved. Either by adding a resistor that limits the surge current as itsmoked recommends - or simply choosing the next bigger varistor. I would choose the bigger one. Adding a resistor means one more component and I strive to keep population down.

You know why I love this site? Of course you know: You learn new things all the time. I've never seen that cool pink smoke. I shall start looking for a MOV that I can abuse. I shall even set up a camera to photograph it. Now, can I have a tip on how to trigger the camera to best catch the smoke?

 

[buzzp]

MOV's do make a lot of smoke when they go. They can do some damage as well. We had a competitor who decided they shall use MOV's. This is fine, except they encapsulated it in epoxy. This made a bomb, of sorts, which our customer did not appreciate.
Can not say that I have seen 'pink' smoke but definately smoke and all that was left was two wire leads.

Skogs, ideally you would have a digital video camera then you might capture some cool images if you set it on the highest frame rate. If you do this, you have to post the pics so we can all be entertained.

 

[dpc]

A little off-topic, but the discussion of MOV failures reminds of what a lightning (surge) arrester engineer told us once:

"It's very simple product with no moving parts. If it has moving parts, you'd better get out of the way."

 

[Skogsgurra]

This may be off topic. I checked what my Minolta Z3 can do. It seems to be able to do "progressive capturing". Looks like I can have a fairly high resolution movie as well. That will probably do it. Now I have to find MOVs a pair of goggles and a safe place. Stay tuned!

This is science at its best.

 

[RalphChristie]

Skogs:

As long as you are not the one to trigger the trigger...

 

[itsmoked]

skogsgurra,

It was a long time ago that me and two colleagues witnessed the "pink smoke" event. Me thinks that it was a slow (relatively), energy limited overvoltage situation, if that helps. Otherwise you get the kabloowee form of failure. This is why you see most MOV's in products shrink-tubed. Never epoxied..er well rarely epoxied.

 

[buzzp]

Can't wait to hear about the results. Are you going to hit it with a short burst or slowly bring up the voltage to failure, or both?

 

[gcaudill]

Is there possibly a high frequency component of your supply voltage to your bridge that is making the capacitor look like a short compared to the 100ohm pre-charge resistor? The inductance of the transformer may be low enough to allow the high frequency to pass, and the MOV's may not be fast enough to catch the transient. A clamp-on amp meter would most likely not account for this current either.

 

[johnnyw]

As a final update, I've uncovered the root cause of the problem which ended up not being related to opening of downstream relays, but opening of the transformer primary with no load. After reading several MOV application notes, I overlooked the common problem of interrupting transformer primary line voltage at a zero crossing (magnetizing current at peak) which causes the flux in the core to collaspe and a resultant spike on the unloaded secondary. I ended up placing MOVs after the fuses (before the diode bridge). I found inexpensive MOVs (<$1 per) from Littlefuse which I could mount directly into my fuse holders. So far, no more diode bridge failures

 

[mc5w]

Each diode has to have a resistance-capacitance snubber across it to limit rate of rise of voltage when the diode switches off.

You also need to have a drive isolation transformer or a line inductance to limit the current that flows during commutation overlap which acts as amomentary short circuit - the diodes do not swith off instantaneously. You get 6 very short time short circuits per cycle.

Your diode bridges do not have enough reverse voltage rating. You need at least 4 times the phase to phase supply voltage in order to have enough margin between that the diodes can take and what a surge limiter clamps.

Also, rectifiers try to turn spikes into useful direct current. If you do not have a fixed bleeder resistor, switched bleeder resistor, or better yet both your filter capacitor will eventually blow up at no load. The first spike will not blow up the capacitor but if you get a series of spikes that are charging your filter capacitor and there is no minimum load, one of the spikes in that series of spikes will blow the capacitor. A lot of electronic fluorescent ballasts have this defect.