Offline LED drivers sometimes blow up at power ON test

[Flyback]

10% of our LED drivers are blowing up on Mains Power ON test. Do you know why?
(Schematic and LTspice simulation attached.)

Please bare with me, this situation is a simple one to understand, but lengthy to explain….

We had very many 230VAC, 150W, Offline LED driver PCBs made to which we add (by hand wired Modification to PCB) two 100nF X2 capacitors to help them through EMC.
Aswell as this , we also (as a ‘hand’ modification to PCB) cut through some PCB copper which was connected to the 0.5W Buck converter’s switching node. We did this so it could pass common mode emissions testing. We cut the PCB copper with a Dremel Drill type thing.
Now, the problem is that since doing these two modifications, we now get about 10% or so of manufactured products blowing up at Mains Switch ON test.

What do you think is the cause of these blow-ups? The blow-ups never happened before these two modifications were done.

The blown-up units have the two 1206, 0R56 resistors blown open, the 1A mains input fuse is blown, and the 470uH inductor is badly blasted and goes open…and the downstream 220nF ceramic capacitor is often cracked.
We suspect that the blow-ups happen when the production test staff inadvertently switch the finished LED lamps ON at mains peak. This we think results in a large inrush current.
The thing is, the FET M15 should act somewhat as an inrush limiter. However, we believe that the act of cutting the PCB copper with the Dremel tool is actually ‘rubbing’ the PCB in such a way that the NFET M15 is getting induced ON (Vgs voltage getting induced to >4V). Then when we sometimes inadvertently apply mains at mains peak, there is no inrush limitation and the blow-up happens.

Do you agree that this could be our problem?

Schurter 3403.0166 Fuse datasheet:
https://www.farnell.com/datasheets/...96.333053046.1497723066-1679746183.1489787856

470uH (Wuerth 744045471) inductor datasheet..
https://katalog.we-online.com/pbs/datasheet/744045471.pdf

0R56, 1206 resistor (ERJ8BQFR56V) datasheet:
https://nl.farnell.com/panasonic-el...bqfr56v/resistor-1206-0-33w-1-0r56/dp/1717876

 

[chemelec]

Put a Cap across D108 to Slow down M15 Turn On Time. (1uF or 10uF)
Possibly also a Filter cap, After the diodes on the Rectified Output.
(+ of D24 & D25 to - of D26 & D27)
To Eliminate some AC Ripple

 

[alec_t]

According to the sim, those caps are prolonging the ringing on the +ve rail from ~30uS to ~100uS and are introducing an extra 12A or so RMS current pulse over that ~100uS at switch-on. I'd be surprised if that would tip the balance and cause things to fry, unless some component were already stressed, but who knows?

Edit:
Of the two resistors R63, R70, is it always the same one which fries?
Edit2:
I see from the sim that there is a -100V spike on the FET gate. Without those caps the spike is -400V! A 1uF cap from drain to gate tames the spike (but the suggested cap across D108 doesn't).

 

[Flyback]

Thanks, I like the idea of the capacitor across D108. I will have to think about quick “OFF/ON/OFF/ON” situations though, when the cap would already be charged up for the “second oncoming”.

Possibly also a Filter cap, After the diodes on the Rectified Output.
(+ of D24 & D25 to - of D26 & D27)
To Eliminate some AC Ripple

Thanks, will consider though will have to be limited in value since as you know it affects the power factor, and also the operation of our power factor corrected, non-SMPS based LED driver.

Of the two resistors R63, R70, is it always the same one which fries?

R63 and R70 both get blown up.

Actually, having thought about it, the NPN, Q5 would soon clamp the gate of M15 and limit inrush current. I am actually now thinking that maybe M15 is getting killed by the electrostatic situation of the dremelling and is just short from drain to source, and so not limiting inrush current at all.

Do you think that the attached surge current waveform would be likely to blow up the 1206, 0R56 resistor?
The power waveform is also shown.

Unfortunately the ERJ8BQFR56V datasheet doesn’t allure to its surge handling capability.

ERJ8BQFR56V datasheet (0R56,1206 resistor)
https://www.farnell.com/datasheets/21...183.1489787856

New simulation (ltspice) posted to show case of M15 being short from drain to source.

 

[alec_t]

There is only ~1mJ of energy in the R63 pulse over its ~30uS duration. If you know the thermal capacity of the resistor in situ you can calculate the temperature rise .

 

[Flyback]

Thanks, its the 400W pulse for 5us which makes me wonder if the 1206 can handle this.
I know it seems like only small energy, but when you have specific surge resistors, its amazing how such small bits of energy in tiny bits of time can kill it. The current pulse does go up to 27A, and i think the 1206 bond wires just literally get shocked off their connections and it dies?

I notice that the 1206 , SG73-RT resistor is surge rated but over any <10us interval the power goes to 1000W or more it will blow up, as its datasheet tells...
https://www.koaspeer.com/catimages/Products/SG73-RT/SG73-RT.pdf

...Our 1206 resistors are not stated as being surge proof, (ERJ8BQ type)...........
https://nl.farnell.com/panasonic-el...bqfr56v/resistor-1206-0-33w-1-0r56/dp/1717876

What i also notice is that our Schurter 3403.0166 1A mains fuse doesnt give any "time to break" data for currents above 10A....and i wonder if this is because it just breaks instantly if it gets more than 10A....(it does see more than 10A if FET M15 is short.)
Schurter 3403.0166 Fuse datasheet:
https://www.farnell.com/datasheets/...96.333053046.1497723066-1679746183.1489787856

 

[alec_t]

Interesting. I've never considered surge resistance of SMD resistors .... until now . But if the resistor is the weak link, how come the fuse, choke and caps also bite the dust?
What load is attached while the test is being made? Could something not shown in the sim be drawing excessive current? Faulty test jig?

 

[Flyback]

What load is attached while the test is being made? Looks like something not shown in the sim is drawing excessive current. Faulty test jig?

At startup the 150w led driver is disabled so theres no load.
The LNK302 0.5W buck bias supply starts up pretty quickly though.
There is no test jig as such, its just a test where the product is powered up.

 

[alec_t]

Divide and conquer. Might be worth sacrificing a board by replacing the 1206 resistors with something much beefier and see if your test guys can break that board. If they do, but the beefier resistors survived, then move on to the next suspect component and replace that.

 

[Flyback]

Thanks , it is very odd that fuse, choke and resistors are all blowing.
Suggests a really big current transient.

On the AC fuse note, the attached inrush current through the 1A Schurter 3403.0166 (1A) fuse is above 10A , and the fuse datasheet seems to suggest that anything above 10A is not allowed, even for microscopic instants of time. The datasheet does not list any break time for currents above 10A for the 1A fuse.

 

[alec_t]

The sim has a 50A spike through the "TVS". How realistic is that?

 

[Flyback]

Thanks, that TVS is only rated to Ipp max of 2.6A...that TVS current happenes (in the sim) with the M15 FET shorted from D to S.

 

[alec_t]

When the FET is not shorted, the 30A current surge as the input caps charge is causing a -600V spike on the FET gate. Is the FET rated to survive Vdg = 600V ? If not, and the FET breaks down, you could get that 50A current spike through the TVS (at least in the sim).

 

[Flyback]

Thanks, yes Vdg can be 600V. The FET is rated to more than that.

 

[Flyback]

Do you think the wuerth 744045471 470uH inductor can take the following (attached) surge current without blowing up?
This is the current surge it gets if the FET M15 has already failed short drain-source

744045471 470uH inductor datasheet
https://katalog.we-online.com/pbs/datasheet/744045471.pdf

 

[alec_t]

Those currents, both the +ve and the -ve bits, far exceed the rated current and the saturation current. Have you asked the manufacturer?
Has the FET ever failed drain-source short (you don't say so in post #1)?

 

[simonbramble]

have you tried a small amount of resistance in series with the input caps, C8 and C1. If you apply a step function to an LC tuned circuit, the output voltage will ring up to 2x the input step. This is what you are doing when you close the switch. the upstream inductance is ringing with the newly added capacitance and the input voltage is ringing above the voltage you expect it to be. The Transorbs are conducting the shooting a high current through the sense resistors.

This is my humble opinion after looking at the circuit for all of 2 minutes.

Series resistance will add damping to the LC circuit formed by L1, L2, C1, C8 and damp the spike, so the transorbs don't conduct.

Or... put some C across the sense resistors to slow down the turn on of the protection npn. Not sure if this would work, but yet another avenue to investigate...