Overvoltage when bayonet connector LED bulb is plugged in to live mains?

[treez]

The below shows the input filter of a 10W mains connected smps LED driver.
When the bulb is plugged into a live mains bayonet connector, -and if this occurs at the mains peak, then the voltage on the primary DC capacitor is 500V. The voltage on the primary DC Bus just after the mains rectifier bridge peaks up to nearly 900V.

obviously, these transient overvoltages would normally be taken car of by the MOV at the input...however, that will surely severly reduce the life expectancy of the MOV?
Also, this overvoltage may well occur every single time that the light is switched on?

I believe that it is the repeated surge through the MOV which wears that MOV out quickly?
(LTspice simulation of switch on also attached)

 

[mtwieg]

Of course, whenever a MOV does its job its lifetime is shorted a bit. That's how they are, and that's why they should only be used for rare events (large transients from the AC line). They wouldn't help at all with overshoot of the input LC filter (since the MOVs are on the input of the filter, right?) anyways. For that you should probably use TVS diodes. The energy of that event should be pretty low, so I doubt you'll need a big device.

 

[treez]

supposing that he primary side dc bus (film) caps are 500V rated...then would you settle for a tvs breakover voltage of 450v and a mov voltage of 500v?.....presuming that both tvs and mov are needed together.?

 

[mtwieg]

The MOV standoff voltage should simply be a bit more than the largest AC line voltage under normal operating conditions. The post-filter TVS diode is more ambiguous, and depends on how much current the diode actually has to conduct to suppress the transient to <500V. So it depends on what specific device you select. Also keep in mind the ratings of the diode bridge (which should probably be after the line filter as well).

 

[treez]

if the mov is just above the peak line voltage, then it will probably end up breaking over before the tvs gets chance to, and then the tvs would be a waste of time?..as it would never see 'action,.

 

[schmitt trigger]

As important as the clamping device itself, are snubbers to limit the voltage risetime dv/dt to a safer level.

The successful implementation of voltage clamping I have witnessed, also use a snubber to limit the dv/dt.

 

[chuckey]

I am a bit thick Treez, how does C7 get 500V across it, according to my theory it gets 1.4 X 230V ~ 322 V. thanks
Frank

 

[mtwieg]

if the mov is just above the peak line voltage, then it will probably end up breaking over before the tvs gets chance to, and then the tvs would be a waste of time?..as it would never see 'action,.

The TVS will clamp the ringing of the filter when it's connected to the AC line. The MOV acts on transients existing on the AC line which spike much higher than 230Vac, but it does nothing when you initially connect to the AC line. That transient only exists at the output of the filter, not the input.

I am a bit thick Treez, how does C7 get 500V across it, according to my theory it gets 1.4 X 230V ~ 322 V. thanks
Frank

He's referring to when you connect to the AC line when the AC is already at its peak. The output of the filter can then, in theory, ring up to twice that peak voltage.

 

[FvM]

500V is true, at least for a lossless inductor and switching at input voltage peak, by working of the LC filter.

In my view, the voltage level isn't too dramatic and won't necessarily need clamping.

 

[treez]

The MOV acts on transients existing on the AC line which spike much higher than 230Vac, but it does nothing when you initially connect to the AC line. That transient only exists at the output of the filter, not the input.

so the MOV goes at the AC terminals, but the TVS goes after the input LC filter?

In my view, the voltage level isn't too dramatic and won't necessarily need clamping.

But what if the dc bus cap is only 450v or 500v rated? (it will be a film cap, not 'lytic)

 

[mtwieg]

so the MOV goes at the AC terminals, but the TVS goes after the input LC filter?

Correct, and that's the most common implementation.

 

[FvM]

Standard film capacitor voltages are 400 and 630 VDC. Film capacitors have good surge capability, so 400 V maybe still safe.

 

[treez]

Hello,

If the following offline, 10W LED driver circuit is connected to the mains at the instant when the mains is at its peak (325V), then the voltage on the primary DC bus rings up to 665V. This is above the 630V rating of the primary side DC bus capacitors.

This assumes that the input voltage rises from 0V to 325V in 1us. I assume this is correct?

Schematic (offline led driver)
**broken link removed**

Attached is the schematic of just the input filter, and the simulation showing the overvoltage at switch-on.

 

[betwixt]

You seem to be ignoring the load presented by the remaining electronics. You are only observing the peak because of resonance in the inductor but it's Q factor, especially under load will be quite small.
You could also argue that switching on at peak voltage would blow the diodes because the inrush current to the capacitors would be excessive. In real life the over voltage/current condition if present at all would be insignificant.

Brian.

 

[FvM]

Yes, connected circuit elements will slightly reduce the resonance Q. But you have also the option to build a lossy LC filter, as in the Powerint example designs.

 

[treez]

You seem to be ignoring the load presented by the remaining electronics

yes, the reason why I ignore that is because, as you know, mains powered smps's often take a good second or so to start up, because the bias rail capacitor has to first slowly charge up via the high value resistors coming down from the primary DC bus rail..so I thought it was valid to ignore the load for the switch-on interval.

But you have also the option to build a lossy LC filter, as in the Powerint example designs.

As you can see I do have two 3.3 ohm resistors (plus 6 ohms of esr in the inductor L1) which add to the lossiness, but these still don't stop the overvoltage.
I admit that I could put in one of the powerint active lossy networks involving a fet that gets switched out just after start up, though I would rather avoid the expense.

 

[FvM]

As you can see I do have two 3.3 ohm resistors (plus 6 ohms of esr in the inductor L1) which add to the lossiness, but these still don't stop the overvoltage.

Calculation helps. The characteristic impedance of the 3.3m/200n LC filter (in effect while the rectifier is conducting) is around 100 ohm, with reversed bias rectifier above 200 ohm. An effective damping circuit must use a resistance with same order of magnitude. A C-RC parallel circuit as used by Powerint seems suitable.

 

[betwixt]

yes, the reason why I ignore that is because, as you know, mains powered smps's often take a good second or so to start up, because the bias rail capacitor has to first slowly charge up via the high value resistors coming down from the primary DC bus rail..so I thought it was valid to ignore the load for the switch-on interval.

True, but the capacitors are after the filter and will present a low impedance at switch-on and even the current drawn into the start-up circuit will present some load. I'm not saying you are technically incorrect, just that you are worrying about something that is inconsequential. 1uS is a millionth of a second and the parasitic elements and surge ratings of most components will tolerate it. To simulate in more life like conditions, add an input inductance to represent the length of wiring before the circuit and don't forget you are also assuming a perfect rising edge at peak AC cycle, most switches and statistics will make that improbable.

Brian.

 

[FvM]

IEC61000-4-11 (Voltage Dips, Short Interruptions and Voltage Variations Immunity Tests) has a test case that switches the input voltage at 90 or 270 degree phase angle, test generator rise time is specified between 1 and 5 µs. In so far the input voltage assumed by treez is quite reasonable.

 

[treez]

Ive added the wiring inductance, and the RC damper, the startup resistor, and that has now cut down the peak to 610V, so its below the 630V of the capacitor now.

I have used a rise time of 1us for the voltage to go from 0V to 325V. Is that pretty accurate?..since after all, the contacting is fairly instantaneous I would have thought?..i confess it will bounce a little but i'd have thought the contacting is instantaneous.

I think an interesting case would be if it was accidentally connected to a triac dimmer, whereby it is dimmed such that the mains snaps on at the mains peak in every cycle..i think this is why these offline lightbulb circuits seem to go so over-the-top as regards input filter damping. Ive never seen this kind of use of significant series resistance, or active filter damaping resistor_fet circuits in other "non-lightbulb" offline smps's.

I think that that must be it...ie...they are afraid the thing will be accidentally connected to a leading edge dimmer such as a triac dimmer?