Mains Transient protection circuit looks very bad......do you agree?

[Flyback]

Hello,
Our contractor has put forward a Mains Transient protection circuit (for our power factor corrected 40W offline LED driver) which looks so disastrous that please may I put it here for judgement? (it is as attached).

The idea is that any transient gets shunted by both the TVS and the controlled NFET , M3. The Current clamp based around M1 ends up taking some of the transient voltage, so that the LED driver and LEDs don’t get exposed to the transient overvoltage spike.
The problem I see is that the resistive divider which connects to the ZR431 has extremely high value resistors, and so noise tripping of the ZR431 is likely. Also, the NFET M3 has a 100k resistor connected Gate-Source. This is very high value, and so the NFET M3 may noise trip, and thus turn ON when it shouldn’t and disastrously shunt the high voltage DC Bus and get blown up.

What do you think? Should we kick this circuit out? In theory it seems fine, but in practice…ummmmm.

ZR431 datasheet:
https://www.diodes.com/assets/Datasheets/ZR431.pdf

(This PCB has very little room for components, and so we can’t fit more conventional Mains Transient protection circuitry on it. –This is why the attached circuit has been put forward.)

 

[Diver300]

Firstly, tell whoever drew the circuit that lines that join should never form a cross. I first thought that the Drain on M1 only connected to the Anode on D10. Not having physical space is no excuse for drawing the circuit ambiguously.

Is there capacitance on the DC circuit?

The only problem that I can see is that the 10 V, limited by D10, will be drained down when the ZR431 turns on. M3 needs 4 or 5 V to turn on. I don't know how low the output of the ZR431 can go, but it is clearly less than 2.5 V *. The 100 Ω resistor (R10) will drop nearly nothing, and there can only be about 0.6 V across B-E of Q2, so there won't be enough voltage to turn on M3. The C1/R10 time constant is 0.3 μs, so C1 won't help much.

Apart from that, I would have thought that the basic idea would be fine. I did something similar with load-dump protection, although mine only isolated the voltage, it didn't try to shunt it as well.

If you work out how much current would need to be injected into R3 or R12 to get the ZR431 or M3 to trigger, I think that it would need quite a lot of stray capacitance and some very sharp edges to do it.

* The diagram at the bottom left of page 6 of https://www.diodes.com/assets/Datasheets/ZR431.pdf implies that the output of the ZR431 will get to about 2 V when the input is high.

 

[Nigel Goodwin]

Firstly, tell whoever drew the circuit that lines that join should never form a cross.

Sorry, but that's how circuits have been drawn for MANY years now - the connections are signified by a dot at the junction.

 

[Diver300]

Sorry, but that's how circuits have been drawn for MANY years now - the connections are signified by a dot at the junction.

I know that, and I did work out what is happening, but it's a useful guide to avoid having a cross that is a junction.

DerStrom8's guide here https://www.electro-tech-online.com/threads/rules-for-drawing-readable-schematics.144863/ mentions it.

That is what the that part of the circuit diagram looked like on my laptop with a 1280 x 800 pixel display. I've zoomed that in 5x so that the two pixels that differentiate a cross from a join are visible. On one level zoom out, there is no difference at all on screen between a join and a cross

Newer laptops might well have more vertical resolution but they would probably have a lower-height screen, so the pixels would be a lot smaller and less visible. The .pdf varies the line - dot ratio depending on the zoom so I can't be sure how visible it would be.

 

[Nigel Goodwin]

That is what the that part of the circuit diagram looked like on my laptop with a 1280 x 800 pixel display.

Don't use a laptop then!

 

[Flyback]

thanks, i confess to being told off for doing cross connections whilst working in London last year. The manager was a 54 year old guy and maybe that was why i dont know.

Is there capacitance on the DC circuit?

Thanks, There is only 22nF (nanoFarads) of capacitance on the DC circuit (sorry i forgot to show it).

The only problem that I can see is that the 10 V, limited by D10, will be drained down when the ZR431 turns on.

Thanks, yes, that’s true. This is because C1 is only 3.3nF. However, since transients have such a short duration we didn’t want M3 to be able to stay on for long, since M3 is across the high voltage DC bus, and so we wanted its ON-time to be minimal (just enough to clobber the transient and no more).

When a spike happens, the NFET M3 does indeed take about 2us to turn ON, (due to its Cgs capacitance needing to charge up to Vgs[th]) but the TVS is supposed to fight the transient during this initial few microseconds.

Whatever current gets drawn through the 100R, 50 times more current will go through Emitter-collector of the PNP, so that makes sure that the NFET M3 does actually get turned ON.
I am just wondering if we should have put the 100R resistor in the emitter connection of the PNP? –This way the turn-ON of the PNP will not be so dramatic, because the 100R placed there, would kind of give some negative feedback. –This (I think) would help the ZR431 to better control the M3 based shunt regulator to regulate (clamp) the high voltage DC bus to 413V?

If you work out how much current would need to be injected into R3 or R12 to get the ZR431 or M3 to trigger, I think that it would need quite a lot of stray capacitance and some very sharp edges to do it.

Thanks, but sorry I didn’t understand the meaning of this.
----------------------------------------------------
The attached shows the circuit complete with the MOV and the 22nF DC bus capacitor.

 

[Diver300]

If you work out how much current would need to be injected into R3 or R12 to get the ZR431 or M3 to trigger, I think that it would need quite a lot of stray capacitance and some very sharp edges to do it.

Thanks, but sorry I didn’t understand the meaning of this.

What I meant is that in normal use, even taking the worse case that the supply voltage is near the peak, the voltage across R3 is quite a bit less than the voltage needed to turn on the ZR431. For the the voltage to get to 2.5 V, there has to be a significant increase in current from somewhere. That could briefly come from stray capacitance.
My point was that with 100 kΩ to ground, there would have to be a few microamps flowing though the stray capacitance. You could work out what value of stray capacitance and what rate of change of voltage is needed on the other side to raise the voltage enough for the ZR431 to turn on. I think that you would find that it is difficult for that turn-on to happen, especially if it has to last 2 μs.

The bigger danger is dampness or contamination on the 8.2 MΩ resistors.

 

[Flyback]

What I meant is that in normal use, even taking the worse case that the supply voltage is near the peak, the voltage across R3 is quite a bit less than the voltage needed to turn on the ZR431. For the the voltage to get to 2.5 V, there has to be a significant increase in current from somewhere. That could briefly come from stray capacitance.

Thanks, but i am under the impression that the voltage at the reference input of the ZR431 gets up to 2.5V due to the actual transient itself raising the voltage of the high voltage DC Bus.

By the way, this lamp is mounted outdoors, on top of a pole.

The bigger danger is dampness or contamination on the 8.2 MΩ resistors.

Thanks, i am just wondering if you have absolutely hit the nail on the head here. This coudl be the "killer" of this circuit....it would make the ZR431 false trigger and mean M3 gets turned on when it shouldnt turn ON...and blows up.


The lamp enclosure is actually open to the elements, -it is not *totally* sealed. Well, there are gaskets around most of the mechanical interfaces…however, the wire entrance hole for live and neutral to get to the PCB, comes up through the heatsink on which the PCB rests……this hole is not sealed as it would be expensive to seal it. So moist external air can first come in through the gap between the lamp and the vertical mounting pole….this cannot be totally sealed since the pole belongs to the customer and is obvioulsy a separate part to the lamp itself.

So what I am saying is, moist air can make its insidious way in to the PCB, and could settle on the 8.2Meg resistors, and thereby reduce their resistance, and thereby make the ZR431 false trip…..which would mean M3 shunting the high voltage DC rail, and thereby blowing up M3 and indeed R4. The PCB hangs upside down, so water can't easily collect over the PCB, but a film of moisture from the atmosphere could still settle on the 8MEG2 resistors and reduce their resistance?

 

[Diver300]

My 9 - 5 job is now in electronics for cars. We're not allowed resistors > 100 kΩ, or circuit points that would be affected by <100 kΩ of leakage, for that reason. The "Nissan Micra boot opening by itself" problem is caused by a circuit that will trigger at 20 μA on a 12 V circuit, so around 500 kΩ of stray leakage.

I think that your design could be problematic. Also you've got DC on the high value resistors, so any moisture could lead to dendrite growth and failure in minutes.

 

[Flyback]

Also you've got DC on the high value resistors, so any moisture could lead to dendrite growth and failure in minutes.

Thanks, though we do have 100's of units out there in pub gardens (for 2-3 yrs) with this circuit , and they havent failed in big numbers so far, so its hard to see this dendrite growth problem being one for us?

 

[kubeek]

I doubt that M3 will have very long life in such configuration, as the first surge will likely overheat it. A better solution would be a varistor across the mains, then a series-pass transistor that limits its output voltage to about 400V.

But since you say you have already hundreds of units installed, have you actually tested this unit on surges? What voltage level did the circuit survive? (I think that line to line would be tested with 9uF and 2ohm coupling IIRC)