Safely powering up switch mode mains to dc converters with PFC input

[najrao]

I handle a number of server power supplies, telecom 'rectifiers' and other like sundries, varying in power from tens of watts to 4kW and more. Any PS over 100W has the ubiquitous '100 to 240V a.c. input, unity power factor with sine current waveform'. This is all very well but for the fact that they draw, and need to draw,  tens of amperes inrush current at start-up. Any hiccup in the PFC would blow the fuses, and more than likely, their switching mosfets and fast rectifiers.
I tend not to plug in any unknown PS directly, but with a current limiting tungsten filament lamp in series. This allows up to ten times the lamp's rated current through for the first half-cycle, decreasing rapidly. A 100W lamp on 230V mains suffices for PFC's up to 250W or so. For large  (4kW) rectifiers, I have had to use 1000W halogen lamp! If the PFC has a shorted mosfet,  the lamp simply lights up inviting you take a look inside. If the PFC is indeed good, the lamp lights up briefly and goes out; or more likely, the cycle repeats a good ten times until the PFC fills up the dc bus capacitors. Each time, a relay bypassing the inrush limiting resistor or thermistor, clicks.
A gradually increased a.c. input from a variac is useless here, as the PFC is held off till some 90V is reached, after which all hell breaks loose.
I would be interested in learning if there are any better ways of testing unknown rectifiers.

[David Hess]

I was thinking that PFC regulators were better behaved than that but of course the common PFC boost topology suffers from the same problem that any boost converter does where the output is immediately charged to the input voltage through the inductor and rectifier so there is no way to control the current when power is first applied short of increasing the input resistance which is what the series light bulb does.

This however points to a solution.  Using a variac to ramp up the input will prevent the initial surge current from charging the output capacitor and once the PFC regulator starts, it will actively control its own input current.  The problem then becomes the behavior of the PFC regulator at low input voltages.

What is wrong with the incandescent light bulb test?

What exactly are you trying to do?  Diagnose shorted PFC regulators or moderate their input surge current?

[Red Squirrel]

Interestingly a while back we had a fuse pop in our DMS100 switch, typically not a huge deal depending what the fuse is for, and often there is redundancy anyway.  I don't recall what that particular fuse was for, but the input of that device had a large capacitor.  Every time they tried to put a new fuse it would pop.  Turns out when you replace a fuse for that you need to put it in series with a light bulb first!   I'm not sure how this is done, physically though, like there must be a device for it I guess. Thought that as interesting.

For your case I would just build a jig with a light bulb socket and then a switch that bypasses the light bulb.  Could even have multiple light bulbs in series with their own switch.   It could have an regular power outlet as output, or terminal blocks or whatever is more convenient.

[najrao]

My apology for being late.
The variac  solution is next to useless  because the PFC generally picks up a little under the usual minimum of 85V, so you only recharge to 100V dc or so; the PFC then takes the huge inrush to build the voltage close to 400V dc. Still with the variac, the a.c. input voltage would not have risen much over the 100V mark in the time it takes to turn the variac up -- to 230V. At the lower input levels, the PFC actually draws even more current , being inversely proportional.  One is better off applying 230V straight.
The series lamp method is unsatisfactory in that at first your estimate of lamp wattage needed may be grossly in error. I tend to start with 60W, progressing to 100, 200, 500 and 1000. Some units take several seconds with a small lamp, with or without the PFC cycling intermittently. Once the dc bus is up to level, the current drops (all testing is without external dc load),  but even this is sometimes enough to glow the lamp/s because the input EMI filters have very large capacitors. If the smaller lamp simply remains alight, one does not know if it is safe to increase the current; if the PFC is faulty, as it very well may be, high currents will start a smoke bomb.
I was hoping to hear of a new, out-of-the -box method.
Thank you,

[David Hess]

Well, here are some of my thoughts:

1. The input capacitor should not be that large.  Its voltage has to follow the input voltage.  The PFC output capacitor is large but isolated from the input line once it has charged.

2. PFC converters are all current mode as far as I know so the input transistor is protected by cycle-by-cycle current limiting.

3. No PFC converter should be producing a surge current larger than the equivalent rectifier-capacitor filter.

4. If a non-boost PFC converter topology was used, then soft start capability could be included very easily.  The standard boost converter topology cannot do this without adding another switch to convert it into a buck-boost topology.

What about applying AC power at the zero crossing to reduce the startup surge of trying to charge the PFC output capacitor?

Alternatively, I might try adding an external NTC current limiter.

If you are looking for a setup for testing PFC converters and off-line converters in general under controlled conditions, then I might just go the brute force route and build a giant controlled AC current limiter.  This could be done linearly or with a switching output.  In the later case, it is just a better sine wave inverter for power applications.  They make AC power supplies for exactly this application.

[najrao]

Thank you Mr.Hess.
I agree with each of your points in  general.
1. There is ALWAYS a direct bypass to the PFC circuit, a surge rated diode charges the dc bus if the PFC doesn't. This itself, and the PFC if active, are inrush limited by a NTC thermistor, or a surge rated power resistor, subsequently shorted out once the bus voltage exceeds about 330V by a relay. So there IS some attempt to rein in the inrush. I ask if n external circuit used temporarily can be devised to moderate further. I have to test psu's with unknown history, and quite a few are blown.
2. Certainly, but more psu's fail due to the PFC mosfet/IGBT than any other cause.
3. see 1.
4. Never seen one.

I do have some 40A SCR relays which claim to be zero-voltage turn on. I shall experiment with one by and by.

[David Hess]

1. There is ALWAYS a direct bypass to the PFC circuit, a surge rated diode charges the dc bus if the PFC doesn't. This itself, and the PFC if active, are inrush limited by a NTC thermistor, or a surge rated power resistor, subsequently shorted out once the bus voltage exceeds about 330V by a relay. So there IS some attempt to rein in the inrush. I ask if n external circuit used temporarily can be devised to moderate further. I have to test psu's with unknown history, and quite a few are blown.

The direct bypass would not be needed if a different PFC topology was used.  Nobody does this of course for the reason sited below.

Why not add a large external NTC thermistor with bypass circuit to further lower the surge current?

Some thyristor circuits include a series inductor to limit dI/dT,  The inductor is deliberately designed to saturate lowering its impedance after a time which conveniently makes it small and inexpensive.  Something similar could be done to limit inrush current.

For testing power supplies, one of the large (and expensive) AC power supplies designed for off-line power supply design and testing would be the solution albeit an expensive one.

2. Certainly, but more psu's fail due to the PFC mosfet/IGBT than any other cause.

The power switch is the most expensive part so it is the least derated.  Still, they really should never fail but of course they do.

4. Never seen one.

I have not either.  But all three common switching converter topologies may be used for PFC.  I suspect the easiest way would be to change a boost topology to a buck-boost but since the surge current demands of a standard boost converter are no greater than the rectifier-capacitor input filter that it replaces, there has little need for such a thing.  The ubiquitous solution is to use an NTC thermistor to limit surge current.

I do have some 40A SCR relays which claim to be zero-voltage turn on. I shall experiment with one by and by.

Some solid state relays are zero crossing.  A TRIAC with a zero crossing optocoupler is the usual inexpensive way to implement this.