Hi,

We have a three phase powered Viper-22A based SMPS design in production which shows repeated failue at some sites particulary during lighting events. The components failing are the same in 99% of the cases:

1. Viper22A switcher IC with pins 1 and 2 (VSS pins) short-circuited with pin 4 (Vdd pin)
2. TL431 voltage reference IC setup at 13.5V (short from anode to cathode). Sometimes, the plastic package even breaks.
3. PC817 optocoupler IC that serves SMPS feedback.
4. 5V microcontroller IC (ATSAMC21G18A-AUT)

We have a common mode pi filter at the input after three bridge rectifier with a common mode choke of 82mH with 10nF X2 caps to protective earth at both sides of the common mode choke. The SMPS is configured for class 1 operation with secondary ground connected directly to protective earth.

My assumptions:
1. Even though there is a common mode filter section present at the input, some high frequencies in the tens or hundreds of MHz range passes through the filter taking advantage of the common mode choke's parasitic capacitance and the Y2 caps' parasitic inductances and rush through the smps transformer's primary through secondary parasitic capacitance (in the range of several pf) through secondary positive rail through the above said components to earth, destroying them.

2. Some switching circuit parasitic ringings coupled to secondary causes the issue. Since we have not provided an LC filter at the secondary, this ringing becomes so dominant with such high frequency that the electrolytic bulk caps at the secondary fail to filter them due to their high ESR/ESL. Also we have not provided ceramic capacitors at the secondary except for the 5V side which again is limited to supply pins of the 5V ICs.

I have attached our schematics. I would love to get your feedbacks on what we could be doing wrong with suggestions on how to recreate the issue in a lab and ideas to eradicate the issue.

 

Do you mean "lighting events" or "lightning events" (or even "lightening events")?
What is the input voltage? You say it's three phase, but is that 230V phase-to-neutral, 208V phase-to-phase or what?

 

Do you mean "lighting events" or "lightning events" (or even "lightening events")?
What is the input voltage? You say it's three phase, but is that 230V phase-to-neutral, 208V phase-to-phase or what?

I mean 'lightning' events (giant sparks of electricity in the atmosphere between clouds, the air, or the ground)

The input voltage is 230V phase-to-neutral. i.e, ~400V phase to phase

 

So that is 2 phase supply. not 3ph to the bridge rectifier?

 

So that is 2 phase supply. not 3ph to the bridge rectifier?

This is an smps that is powered over a three phase supply. That is, the input bridge rectifier consists of three half bridges, each for R, Y and B. Kindly refer the schematics.

 

I see three different symbols for commons, and another symbol for chassis common. It looks to me that the first problem is excess voltage. And excess voltage is able to do real damage. And if there is any lack of a neutral connection the voltages will be much higher.

 

That's just an ordinary 6-pulse rectifier. Output is 563V peak for a 230V supply. That's a bit too much for an IRF830. The various ground symbols seem sensible - triangle is circuit 0V, there are protective and functional earths.
Power Integrations makes flyback converters that work at rectified three-phase voltages.

 

My advice would be to add a suitably rated varistor in parallel with each of the X2 capacitors (C1, C5, C15); you could also consider adding varistors between each phase and PE (protective earth).

 

My advice would be to add a suitably rated varistor in parallel with each of the X2 capacitors (C1, C5, C15); you could also consider adding varistors between each phase and PE (protective earth).

Yes. This is something we have included in the updated design. But I have had some really bad experiences with varistors (275Vac rated connected connected phase to earth) that they themselves pose a threat where they catch fire burning out the entire product. So if there is a solution without varistors, I would go with that.

These products are often installed at very remote locations in India such as at the centre of a large paddy field or so where lightning induced voltages are really harsh where they seldom get many paths to earth to settle down. If in such environments these kind of failures should be expected, I would expect three phase energy meters in those locations to also fail which they don't. I teared down a three phase energy meter as well and found the circuitry to be much of the same with following configurations:

1. NO VARISTORS at all!
2. 5W fusible wirewound resistors at the power entry point for each phase to which a 10nF and 1nF Y1 ceramic caps in parallel are connected to neutral forming an RC filter from each phase to neutral. 10nF works best for low frequency and 1nF for higher frequency noises. This arrangement should be for differential and common mode filterings.
2. Four half bridges for the input power rectifier for R,Y,B and neutral following the RC filters.
3. Common mode choke on the rectified DC bus without any Y2 caps to earth or neutral (No earth connections are made available to the energy meter)
3. Secondary ground connected to neutral through an inductor for phase voltages measurement.
4. Each phase connected to ADC through resistor divider.
5. Star grounding at secondary.

Connecting secondary ground to neutral through the inductor is how they block common mode currents from phases to neutral (which is earthed somewhere nearby by the utility company) is my understanding. Do you think this is worth trying?

I strongly believe the issue we are facing is from common mode noise currents passing through the smps transformer to secondary ground (which is at present directly connected to earth) through the damaged components in the feedback loop and the microcontroller.

 

Fire can be a problem with varistors when they become leaky with age, and the resultant power/heat dissipation causes a fire. Varistors are available with an integral thermal fuse which prevents this hazard – I have seen constructions where a thermal fuse is heat-shrinked to the varistor body.

Rather than 275Vac rated varistors, fitting those having a 300Vac rating will reduce the age leakage issue and will still provide acceptable peak/impulse voltage clamping.

 

A further observation (which is not clear from the circuit diagram) is that there should be some overcurrent protection on each phase, which will also protect any fitted varistors from an overcurrent event.

 

A further observation (which is not clear from the circuit diagram) is that there should be some overcurrent protection on each phase, which will also protect any fitted varistors from an overcurrent event.

Thank you for your note on thermal fuse enabled varistors. Regarding over current protection, we have 5W fusible wirewound resistors on each phase (R48, R60, R76).

What do you think about connecting secondary ground to earth through an inductor to prevent high frequency common mode currents from flowing through the secondary?

Also do you think switching transient ringing at the secondary could be the problem? During normal operation I have noticed around 200mV ringing at 70MHz.

Also switching waveform across viper22a's internal fet drain to source is heavily distorted though peak switching transient voltage is effectively clamped by the TVs diode clamp circuit keeping it below 750V. Waveform Image attached.

 

What is clear to me is that if there are ever phase to phase voltages of 400 volts (RMS, or average), that the DC voltage from that triple bridge rectifier will be in excess of 440 volts. And using capacitors rated to 450 volts in a circuit operating at 440 volts is a risky action. The failure mode of all the devices damaged would be expected from an over-voltage incident.
One simple revision that can reduce the voltage can work if the mains common connection is available.
That will be to revise the rectifier circuit into three half-wave rectifiers relative to the common, so that the resulting DC voltage will be much less, (1.414 X 230 volts.
Removing the lower six diodes, ), and connecting the neutral source to the negative end of C21. can accomplish that.

An alternative that would be more reliable and possibly less expensive, will be to use a step down transformer with a 400 volt primary connected between phases, and a center tapped 24 volt secondary . Of course the transformer must be adequately insulated for that sort of application. Such a transformer, of adequate quality, may not be available in some areas.

 

What is clear to me is that if there are ever phase to phase voltages of 400 volts (RMS, or average), that the DC voltage from that triple bridge rectifier will be in excess of 440 volts. And using capacitors rated to 450 volts in a circuit operating at 440 volts is a risky action. The failure mode of all the devices damaged would be expected from an over-voltage incident.
One simple revision that can reduce the voltage can work if the mains common connection is available.
That will be to revise the rectifier circuit into three half-wave rectifiers relative to the common, so that the resulting DC voltage will be much less, (1.414 X 230 volts.
Removing the lower six diodes, ), and connecting the neutral source to the negative end of C21. can accomplish that.

An alternative that would be more reliable and possibly less expensive, will be to use a step down transformer with a 400 volt primary connected between phases, and a center tapped 24 volt secondary . Of course the transformer must be adequately insulated for that sort of application. Such a transformer, of adequate quality, may not be available in some areas.

If you are mentioning about input bulk capacitances C21, C22, C23, C25 which are 450V rated, they are connected in series to achieve 900VDC voltage rating with resistors in parallel to ensure that each capacitor in series charges to the same voltage (half the rectified DC voltage). If you are mentioning C24, the input DC voltage is linearly regulated down to 318VDC (ideally) before applying to it. Also, we have not had a single failure instance of these capacitors. The failing components are always those in the path of a common mode noise current. ie, .viper22a, pc817, tl431 and the microcontroller. Also at sites with external SPD (surge protection devices) installed, the devices are working reliably. So this must definitely be an EMI issue.

 

If you are mentioning about input bulk capacitances C21, C22, C23, C25 which are 450V rated, they are connected in series to achieve 900VDC voltage rating with resistors in parallel to ensure that each capacitor in series charges to the same voltage (half the rectified DC voltage). If you are mentioning C24, the input DC voltage is linearly regulated down to 318VDC (ideally) before applying to it. Also, we have not had a single failure instance of these capacitors. The failing components are always those in the path of a common mode noise current. ie, .viper22a, pc817, tl431 and the microcontroller. Also at sites with external SPD (surge protection devices) installed, the devices are working reliably. So this must definitely be an EMI issue.

The solution is easy then, good low impedance EARTH grounding and SPD to that ground to shunt the harmful energy away externally of your power supply.

 

But that's not a long term solution. Just can't label "THIS DEVICE MUST BE PROTECTED WITH AN SPD FOR RELIABLE OPERATION" on the package!

 

Where the primary switching peak voltage significantly exceeds the normal rectified dc mains, you are likely to experience EMC compliance issues – with the fast switching high voltage acting as an rf transmitter.

Although limited to 750V with the TV diode clamp, you could consider adding a diode/resistor/capacitor snubber network connected between pins 1 & 3 of the transformer to reduce this voltage further (and importantly, increase the waveform rise-time) – see the typical arrangement of these snubber components in figures 1 - 3 in the link below.

https://audioxpress.com/article/repairing-switching-mode-power-supplies

 

Another observation of the circuit, capacitor C26 (2n2F, 2kV) is bridging the reinforced isolation barrier between primary and secondary circuits – this must be a Y1 rated safety capacitor meeting IEC 60384-14; the voltage rating of 2kV would suggest otherwise.

 

The solution is easy then, good low impedance EARTH grounding and SPD to that ground to shunt the harmful energy away externally of your power supply.

NO!! I was not mentioning those capacitors filing .I m describing the voltage cross the input to the switching regulator, which can be quite bit above the 400 volts stated. That voltage can easily exceed 450 volts if the input is 400 volts between phases.
What I suggested was changing the supply to lower voltage arrangement that used the neutral connection.
That is clearly stated in the second paragraph of post #13.
If byou do not understand a half wave rectifier perhaps someone else will explain it. That arrangement will require neutral connection.

 

NO!! I was not mentioning those capacitors filing .I m describing the voltage cross the input to the switching regulator, which can be quite bit above the 400 volts stated. That voltage can easily exceed 450 volts if the input is 400 volts between phases.
What I suggested was changing the supply to lower voltage arrangement that used the neutral connection.
That is clearly stated in the second paragraph of post #13.
If byou do not understand a half wave rectifier perhaps someone else will explain it. That arrangement will require neutral connection.

I think you quoted the wrong text. My response to the OP was not about capacitors.