Inrush current limiting methods

[MrSlack]

I'm looking for a simple and cheap way of limiting inrush current. I've designed and half built a variable 30v 1.5A low ripple linear supply (using discrete parts only) that works very well. However the inrush current charging the filter caps (4x 4700u low ESR as that kills the ripple dead) is pretty high on initial turn on. Not sure how high but it just blew a 1N5401 to pieces which was rated for 3A and smoked my ammeter shunt (10A) but it was a shitty ammeter so I wouldn't take that as a reasonable reading. I'm using a rather overrated transformer nicked from a HiFi amp. I think it is about 120va but haven't measured it.

I looked at various solutions for this such as introducing a MOSFET based soft start, an NTC thermistor but I don't have any bits lying around to do this. Does anyone know any preferably very simple methods to do this with standard jelly bean parts? My own shitty hack idea came up with a small low power circuit with a relay that has a default state of a switched in resistor in series to the caps to current limit. When a threshold voltage is reached where the inrush will be low, the relay will turn on and short the series resistor then turn on the PSU output. This feels awful to me though.

[Richard Crowley]

Just yesterday, I watched this video by EEVblog member "MikesElectricalStuff".  I thought it was one of the best technology videos I have ever seen on YouTube, and it included some VERY APPROPRIATE discussion of exactly the kind of current limiting you are asking about.  EXTREMELY  HIGHLY RECOMMENDED!!!...

[MrSlack]

Perfect - thanks!

Will watch this now

[Tomorokoshi]

Just yesterday, I watched this video by EEVblog member "MikesElectricalStuff".  I thought it was one of the best technology videos I have ever seen on YouTube, and it included some VERY APPROPRIATE discussion of exactly the kind of current limiting you are asking about.  EXTREMELY  HIGHLY RECOMMENDED!!!...

Very nice. Thanks.

[mij59]

For a 1.5A power supply you don't need four 4700 uF filter  capacitors, one is more than enough.

What transformer are you using ?

[MrSlack]

Fair points on the cap values. I actually worked this out at about 2am last night and now have two 2200u 63v units in it (less stress). The ripple I was compensating for with lots of caps was due to a poorly designed feedback loop. I fixed this and moved the core of the design to an op amp and bandgap reference as well. By the time I've finished I'm going to have invented an LM723 again...

No idea what the transformer is rated at but it's pretty high. It's huge and weighs about 1.5kg. Unloaded I'm getting 40v peak to peak. It can shift 5A at 40v without even getting warm though.

This isn't the final transformer. There's a much smaller and more reasonable toroidal one back ordered on RS. I only had this one lying around that could shift more than an amp without catching fire.

[tron9000]

WARNING!

EXTREMELY  HIGHLY RECOMMENDED!!!...

[station240]

Don't have anything with a PFC circuit in it lying around (eg some PC PSUs), as that gets you the NTC, PWM controller IC and mosfet you need.

Basic idea is to PWM control the mosfet, using the capacitor rail voltage as the input, eg higher voltage = higher duty cycle.

Oh and 4x 4700uF is crazy, I've got a device with 8 of those caps (higher voltage though), it takes 3 phase 415V input.

[CarlG]

You could try some variation on this theme. C2 charges at a rate appr determined by C1 and R2.

C1 acts as a Miller cap. When Ugs = Ugs(th), voltage over R1 is appr constant, i e current through R2 is constant, charging C1 at a constant rate, resulting in a constant du/dt at the output.

The C4/C1 ratio needs to carefully consider the ramp-up time of Uin and Ugs(th) of M2: with C1 = 100n (as in the attachment) there's a turn-on voltage bump at the output when the Uin is applied. Changing C1 to 47n removes this bump (for the typical Usg(th) in the simulation). The circuit is quite useful although it has some limitations.

The power dissipation of course also needs to be considered. The energy "lost" in the switch (at turn on) is equal to the energy stored in the output capacitor (no load at output) so naturally M2 must be properly cooled

[CaveMannDave]

Greetings:

A salvaged 24V SPST NO relay (or, lower voltage, with an appropriate divider), salvaged middling-high value cap,@ 60V or so (for time delay), connect with smallish (500 ohm?) resistor to output of supply.
Contacts bridge salvaged high-watt, lowish ohm resistor in series w/transformer primary.

Relay waits 'till output voltage is up to cut in full voltage.
Old school, but performs.

HTH

Cheers,

Dave

[timb]

Wouldn't a simple PTC thermistor do the job? Pick a low value like 1ohm, put it before or after the rectifier, as the current starts to rush in during turn on the thermistor instantly heats up, raising the resistance and lowering the current. Once it's settled the thermistor cools and resistance goes back to normal.

This is the standard way I've always seen it done in power supplies.

[MarkF]

Take a look at this Soft-Start Circuit For Power Amps

[MrSlack]

Thanks for the replies everyone.

The winner is:

Wouldn't a simple PTC thermistor do the job? Pick a low value like 1ohm, put it before or after the rectifier, as the current starts to rush in during turn on the thermistor instantly heats up, raising the resistance and lowering the current. Once it's settled the thermistor cools and resistance goes back to normal.

This is the standard way I've always seen it done in power supplies.

This is the simplest and cheapest way of achieving this. Ordered from Farnell. Thanks - appreciated

[dom0]

Make sure you order a NTC, because a PTC won't work as well for this.
The NTC has a high resistance when cold, limiting current, thereby heating up, reducing resistance.
The PTC has a low resistance when cold and doesn't limit current much. They are usually used as poly-fuses or thermal switches.

[MrSlack]

Yeah got a few NTC ones - don't worry

[sarepairman2]

I used a 555 timer to trigger a relay (from washing machine) to short out a power resistor.

I think its a pretty good method of doing things.

Remember that a thermal device will not like power cycling. A relay based system has a much faster recovery time in the event of a interruption.

An active circuit will offer even faster protection (triac, mosfet, etc) but you can argue that it is a less robust solution (putting the relay-resistor INFRONT of your protection circuits (mov, etc) will really harden your system much more then a semiconductor or whatever. It's hard to beat a thick ass wire wound power resistor). Like if you expect a spike/surge to happen after a power interruption (this might be the case if you are near a power plant)

[MrSlack]

This is a one off so incidental cost is around $8 for a few NTC thermistors. This is version one of at least three revs.

The digital half of the PSU (ICL71xx ADC) has it's own shitty Chinese switch mode so a future version of this will probably control and monitor precharging.

Next ver is a 20A 30v linear - should be large and dangerous and fun

[MarkF]

Take a look at



done by Peter Oaks.  He walks you through his whole design so you don't need to re-invent the wheel.

[MrSlack]

Thanks for this. I've already reinvented that wheel almost entirely already!

[macboy]

Wouldn't a simple PTC thermistor do the job? Pick a low value like 1ohm, put it before or after the rectifier, as the current starts to rush in during turn on the thermistor instantly heats up, raising the resistance and lowering the current. Once it's settled the thermistor cools and resistance goes back to normal.

This is the standard way I've always seen it done in power supplies.

No, the "standard" way, as you will find in most any AC-DC SMPS, is to use a NTC. The high cold resistance limits the current to a low value, then as the device draws power, the NTC heats up to its operating temperature, dropping in resistance to a point that effectively all but removes itself from the circuit.

I have once seen a PTC in an inrush limiter, but it did not work as you described. It was in circuit at turn-on , with the cold resistance providing the inrush limiting. After some short time, the PTC was shunted (by a triac in this case, could be relay instead) to remove it from the circuit. The reason to use the PTC here is due to self-protection in case the triac control circuit fails to remove it from the circuit. Critically, it was the cold resistance of the PTC that limits the inrush current. If you depend on the high inrush current to heat up the PTC in order to limit inrush current... that's not going to help.

[Siwastaja]

Macboy has it right.

NTC is a self-controlled, relatively cheap & poor inrush limiter, but a single-part solution.

If you see a PTC as an inrush limiter, it's completely different:

PTC is used as a part of a proper active precharge circuit. The PTC is bypassed by a relay or a semiconductor. PTC acts as a self-resetting fuse which opens (except for small holding current to keep the PTC hot) if there is a problem in the precharge logic, i.e., it gets stuck. Slow-blow fuse in series with a resistor, or simply a so-called fusible power resistor, work too.

Active precharge circuit is highly recommended, but NTC can work in a mid power design and is often used in power supplies ranging at between maybe 20 - 200 W. Small enough supply and you won't need anything; high enough, and NTC can't work properly at all in corner cases, and dissipates too much power in normal use.

NTC actually requires that the switcher has proper UVLO that prevents the filter caps from running too low too quickly. It easily takes about 5-20 seconds for the NTC to cool down, and the capacitors will first discharge faster no matter what, but if the switcher IC has UVLO, the discharging may slow down at maybe 2/3 of the full voltage (or 1/3 autoselecting 110/220V supplies), and of course the inrush current will be much smaller then even though there still is quite a lot of voltage difference. So as you can see, NTC can only be used if the inrush current is somewhat manageable even without the limiter. It's far from good solution, it's always a compromise. In some usual corner cases, it can only limit the inrush current to maybe 30-50% compared to the same circuit with the NTC removed altogether!

[Richard Crowley]

It's only a 1.5A power supply.  It seems to me like a large power resistor between the rectifier and the capacitors would limit turn-on current while being no significant barrier to proper steady-state operation.  KISS.

[timb]

Wouldn't a simple PTC thermistor do the job? Pick a low value like 1ohm, put it before or after the rectifier, as the current starts to rush in during turn on the thermistor instantly heats up, raising the resistance and lowering the current. Once it's settled the thermistor cools and resistance goes back to normal.

This is the standard way I've always seen it done in power supplies.

No, the "standard" way, as you will find in most any AC-DC SMPS, is to use a NTC. The high cold resistance limits the current to a low value, then as the device draws power, the NTC heats up to its operating temperature, dropping in resistance to a point that effectively all but removes itself from the circuit.

I have once seen a PTC in an inrush limiter, but it did not work as you described. It was in circuit at turn-on , with the cold resistance providing the inrush limiting. After some short time, the PTC was shunted (by a triac in this case, could be relay instead) to remove it from the circuit. The reason to use the PTC here is due to self-protection in case the triac control circuit fails to remove it from the circuit. Critically, it was the cold resistance of the PTC that limits the inrush current. If you depend on the high inrush current to heat up the PTC in order to limit inrush current... that's not going to help.

Okay, then I guess I had it backwards. I can never remember which unless I look it up. Thinking about it, I suppose it makes sense... The NTC starts at a high resistance, limiting initial current, then heats up, lowering resistance and allowing full current flow.