Would need some advice to design a MOV based transient protection circuit

[Lefuneste]

Hi there, I would greatly appreciate some advice on a small upgrade project I'd like to make on an audio appliance I've been repairing for my local Repair Cafe, where I act as the main electronics guru. Here's a description of my issue and how I intend to address it. I would greatly appreciate some feedback from anyone having dealt with such issues.

About 6 weeks ago, one guy came to the Repair Cafe to have us look at a badly fried Yamaha Stagepas 600i. This appliance consists of a self contained switching power supply, a Class D stereo amplifier, a DSP and a mixer, and is provided with two matched speakers. It is a reasonably good quality device, quite recent and expensive too, and it was purchased by a local school in order to organize parties and fairs.

On the last time this system was used, a group of stupid kids came several times and unplugged the device from the main while it was amplifying a signal. On the third time they did this, the main board blew up quite badly.
The main board is a complex system consisting of a SMPS power supply : on the primary I have a switching controller, 2 switching MOSFET and a transformer, on the secondary 6 voltage lines with rectifying diodes, LVR and filtering caps. After the secondary we have the class D stereo amplifier sharing the same PCB. A self contained solution, not really what we think of a reliable and serviceable piece of equipment to say the least !! DSP and mixer boards connect to this main board.

After the "unplug event" occurred, the primary side of the SMPS just completely busted into flames. One of the power MOSFET was shorted, the shunt resistor network going to this MOSFET (6 low values 1W resistors in parallel/series configuration) was wholly carbonized (see pictures), and the shunt sense line PCB trace going to the switching controller just vaporized, along with few diodes and resistors on the gate control line. I replaced all these SMD components along with the switching regulator, a 12V regulator and the optocoupler for good measures, and hard wired a current sense line back to the controller. When I started the board again I immediately fried the same MOSFET, but without damaging other components. After thorough investigation using a light bulb in series to limit current and a variac, I came to the conclusion that the transformer was also damaged, unwrapped it and discovered a completely fried primary (the picture only give a partial idea of the damages found on the primary wraps). I've ordered the part from Yamaha so the board will likely be back to life shortly.

What brings me here though is the analysis I did of the failure mode, and the fact that, in my opinion, this product has a major design flaw which I'd like to correct.

Before this main power supply/amplifier board, there is a big filtering PCB with several large inductors. It is a typical RLC filter required to filter backward reinjection of switching noise back into the mains. This filtering board has one typical slow fuse as an overcurrent protection, which did not trip when the main board fried. This filter board is directly supplying the main power supply/amplifier board with two AC Live and Neutral wires. My analysis of the failure is that when the kids unplugged the mains cable, the filter inductors were loaded with high current, as the system was amplifying a signal. Without an incoming supply, the filter inductors violently discharged into the lowest impedance path possible, downstream toward the SMPS supply and into the transformer, with a spike of several kV.

So in my opinion there should be a protection device between the filter and the main power board, in order to protect against such an event. There are only a couple of high power fuses (ceramic case resistors) on the main power board, but no MOVs, neon gas or spark gaps of any kind at the entry of this power board. I would like to install a protection device of some sort to avoid such a catastrophic failure in the future and was thinking on installing a MOV across Live and Neutral. The problem I see in this solution is that if such an event occurs again, and the mains cable is unplugged while the system is operating, and if the MOV trips, the energy from the filter inductors will find no way to escape the circuit, as the mains cable will be unplugged. I believe that these protection devices need a low impedance path to release the energy once they are shorted. Does anyone have an idea on how to design a protection circuit and how to install between the filter board and the main power board ? I think that it is also possible to install a pair of MOVs at the exit of the filter board, where I have a connection to the Earth of the chassis.  One for the Live and one for the Neutral. But then the chassis would get to a high value voltage for a short time in case one or both of the MOVs do trip... Not sure this is the way to design this. Any idea about this ? I can provide the service manual and schematics for the whole system if required, it's too big to be added here.

Thanks for your help !!

[Seekonk]

It is a nice story but I think you are totally wrong.  In unplugging the speakers it is more likely they shorted the output.

[Lefuneste]

Not at all. They unplugged the mains cable and the damage is on the primary side of the SMPS. The amplifier is in perfect working order.

[Richard Head]

If they just unplugged the mains supply why on earth is there a transformer winding short?
Maybe the transformer just flashed over internally and blew the primary MOSFETs etc.

[Seekonk]

I'm sure you will carry that theory to your death just because you invented it.  Just hope no one else takes it seriously.  The FET is usually protected by a high speed diode connected to a capacitor with a resistor in parallel.  A variation of that has the capacitor connected to the FET with the diode and resistor in parallel.  Be interested in the condition of those components.  Stories of what happened leading up to a failure are quite unreliable.  This could just be a random unrelated failure.

[Lefuneste]

Hi Richard and thanks for your help. Yes this is also quite possible, it can also be a defect on isolation of the transformer primary wraps which may have been shorting spontaneously or triggered one way or another I agree. I am nevertheless suspicious about the filter stage as it has huge coils and to my understanding these coils can load up quite a lot of energy which can be discharged into a high voltage surge. This filter board is not in the pictures but the combined inductance of the coils is quite big.

[electronic_eel]

I guess that the transformer failure is a subsequent fault due to overcurrent after the FETs died.

The FET most probably died due to overvoltage. I don't know about the exact details of the SMPS, but it probably has an active PFC. That is required for devices above a certain power level, IIRC 75W. Devices without active PFC have a largeish capacitor after the rectifier, that should absorb smaller to medium pulses. But this one probably hasn't, so a voltage spike from the input can kill the active pfc circuit.

The spike could have been created by contact chatter in the mains plug during slowly unplugging it.

A MOV after the chokes would probably have prevented that.

[Richard Head]

I guess that the transformer failure is a subsequent fault due to overcurrent after the FETs died.

I don't believe that. I have never seen a transformer primary damaged by shorted MOSFETs on the primary.

[Lefuneste]

After having studied the matter a bit further, I will be adding 3 MOVs in a standard one-phase configuration at the output of the filter board. Hot to Ground, Neutral to Ground and Hot to Neutral.
Those interested can read these documents or do some research one "system-generated voltage surge".

http://www.littelfuse.com/~/media/electronics_technical/application_notes/esd/littelfuse_overview_of_electromagnetic_and_lighting_induced_voltage_transient_application_note.pdf

http://encyclopedia2.thefreedictionary.com/Voltage+Surge

http://www.electronics-tutorials.ws/resistor/varistor.html

[T3sl4co1l]

If it were the inductors, why didn't it happen any time it was turned off?

While I can't speak to the underlying events of the failure (who knows, maybe it was a very close coincidence that a lightning surge blew it up the moment it was plugged), it's unlikely the filter inductors did what you propose: they are intended for common mode filtering, and so they store very little energy, especially due to load currents.  The direction is also wrong: the switched end of the inductors is where the peak voltage occurs; the other end is as happy as can be.

It is plausible that a turn-on surge could've occurred, where the filter inductors (and other stray inductance) absorbs a huge current spike (>>100A), which in turn pushes the DC link voltage well above where it's supposed to be.  But this is also unlikely, because the DC filter cap stores far more energy than those inductors do, even under such surge conditions.

Anyway, this is no different from ordinary turn-on conditions either.

The only condition that quick plugging does make a concern for is, if there is an inrush prevention NTC resistor somewhere: if it's already hot (from operating), it will have a low resistance and thus draw a large current surge, whereas under normal conditions (long delays between on-off cycles), it cools down and has a nice large resistance which absorbs the inrush surge.  This still shouldn't cause problems for the circuit, but maybe it really wasn't designed all that well?

Tim

[Seekonk]

There are many potential failure modes possible.  The one suggested is the least likely.  I just can't see that much energy
being stored.  That defice must already have mov's installed and the filter capacitor should be able to absorb any spike.

More likely could be a latch up condition of either the class D amplifier or the switching regulator. Most regulators and
digital electronics have a soft start or POR provision.  Repowering electronics after the supply has dropped to around half
voltage could cause latch up. Think of all the electronics that have an instruction to remove power and wait a minute before
turning on.  Multiple power on is a hard thing to protect from unless an on delay timer is installed.

I have several nice Xantrax, Lambda etc power supplies that would have coust thousands of dollars each. These were obtained
at a reasonable price because each one had failed.  I haven't felt the need to redesign them. It is good to think like an
electron when analyzing a failure.  Just remember that jumping to conclusions is the biggest pitfall in decision analysis.


I would certainly add a light bulb to your test bench. I have three lamps of different wattage in parallel.  These feed a AC
socket in series.  Then the appropriate lamp(s) are screwed in depending on the device under test.  Lamps have a PTC that can
increase the resistance seven times from cold to hot. I have had this setup on my benches since the mid 60's after seeing an
EICO kit ad for a transistor amp.  It showed a lamp and said this was the only piece of test equipment that was needed.  This
has saved countless semiconductors over the years. It is insane to just put new parts in, turn on and hope for the best.