How to choose the value of a NTC inrush limiter ?

[PinheadBE]

Hi,

I have a SMPS MeanWell RS-150 (datasheet attached) with a 40 A inrush current, and 2 A typical current.

I have already blown a power-meter because of that inrush current.

So, I would like to limit that inrush current by inserting a NTC in the mains input line.

For the voltage, I have to choose 400 V, OK. (We have 230 VAC here)

But for the resistance value ?  How do I choose it ?   And how to insure that the NTC itsef will support the current spike ?

Any help welcome !

[liaifat85]

Have you tried the formulae mentioned here?
https://www.vishay.com/docs/33001/seltherm.pdf

[srb1954]

Have you tried the formulae mentioned here?
https://www.vishay.com/docs/33001/seltherm.pdf

That document appears to be primarily for temperature sensing applications of NTC thermistors, not for surge protection.

[srb1954]

Hi,

I have a SMPS MeanWell RS-150 (datasheet attached) with a 40 A inrush current, and 2 A typical current.

I have already blown a power-meter because of that inrush current.

So, I would like to limit that inrush current by inserting a NTC in the mains input line.

For the voltage, I have to choose 400 V, OK. (We have 230 VAC here)

But for the resistance value ?  How do I choose it ?   And how to insure that the NTC itsef will support the current spike ?

Check the manufacturer's datasheet e.g. Ketema Rodan Surge-Gard devices which may provide selection criteria or use the parametric selection guides on the likes of the Mouser website to narrow down the choices.

Select a device starting with:
1. The NTC device current rating should exceed the maximum steady state operating current drawn by your power supply.
2. The cold resistance resistance value should be sufficient to limit the peak inrush current to your desired value. If you want to restrict the peak inrush current to less than 40A then you will need  a cold resistance of at least 8 on 230V RMS.
3. Select a device meeting, or exceeding, the above parameters and determine its actual resistance when operating at the normal steady state current.
4. Determine if the voltage drop across the NTC and its power dissipation are acceptable for your application. NTCs normally dissipate a bit of power and run moderately hot.
5. Determine if the NTC has sufficient surge energy rating to handle the total energy absorbed in the power-on surge. This may be either specified as the maximum allowable capacitive load or the maximum energy in Joules. You may need to measure the actual circuit in operation to determine the energy figure.
6. Repeat with a larger device if the energy rating is insufficient.

[PinheadBE]

Thanks srb for your answer 

I still have a question:

(...)
5. Determine if the NTC has sufficient surge energy rating to handle the total energy absorbed in the power-on surge. This may be either specified as the maximum allowable capacitive load or the maximum energy in Joules. You may need to measure the actual circuit in operation to determine the energy figure.

How do I calculate the energy in Joules ?   Is it proportionate to the surge current and the time it lasts ? If so, unfortunately, the DS does not mention it....   I suppose that time is a kind of curve with a parabolic descent, and the energy is the integration of that curve ?
Maybe, by being conservative, I could choose a default value for the NTC ?

[Vovk_Z]

But for the resistance value ?  How do I choose it ?   And how to insure that the NTC itsef will support the current spike ?

I would go to the good NTC datasheet. They typically have some minimum info about how to use. Manufactures often type rated current and max. load capacitance.
And I would use experience too, e.g. PC ATX 300-450 W power supplies use 5 R 7-9 mm diameter NTC resistor. So, for 150 W power supply you may use a bit (proportionally) higher resistance. I guess you may use 5-20 R 7-9 mm diameter NTC. Diameter depends on how often you want to turn it on and off.

I may add that 5 R is pretty common default value for 100-450 W power supply. You may take it as a starting point.

I have a SMPS MeanWell RS-150 (datasheet attached) with a 40 A inrush current, and 2 A typical current.

150 W rated power supply will consume 1 A current but not 2 A (that's for 115 VAC I guess).

[PinheadBE]

I have a SMPS MeanWell RS-150 (datasheet attached) with a 40 A inrush current, and 2 A typical current.

150 W rated power supply will consume 1 A current but not 2 A (that's for 115 VAC I guess).

The DS states 2 A for 230 VAC and 3 A for 115 VAC.   I don't know how they calculate that, especially with a claimed 74 % (worst case) efficiency. 
Whatever.

For my inrush question, I'll try with a 5 ohm NTC as a starting point as suggested, once I'd figured how to install it in my case so that it can dissipate without burning anything.... 

[shapirus]

How do I calculate the energy in Joules ?   Is it proportionate to the surge current and the time it lasts ?

It's proportional to power and time. To get a precise value, you need to calculate a definite integral of P(t)dt on an interval between two time points. I think it should be possible to simulate and plot the power dissipated by the NTC using spice, and, if it can't calculate the area under the P(t) graph directly, it should be possible to export the data points to a text file and perform the calculation using a simple script, or, if the graph is linear (enough), then it becomes trivial.

[Vovk_Z]

150 W rated power supply will consume 1 A current but not 2 A (that's for 115 VAC I guess).

The DS states 2 A for 230 VAC and 3 A for 115 VAC.   I don't know how they calculate that, especially with a claimed 74 % (worst case) efficiency. 

I'm not arguing about datasheet statements. I just try to help you so I say which exact current this power supply will consume at 230 VAC.

(BTW I have several MeanWell HDR-150 at my hands at the moment, and I have measured their consumed current. It is about 1.0-1.1 A at 220 VAC at output rated power).

[Vovk_Z]

To get a precise value, you need to calculate a definite integral of P(t)dt on an interval between two time points.

That's superfluous I guess (that's not a rocket science. Or is it?). Good datasheet plus a couple of experiments can give and answer too, "accurate enough for Australia". 

[shapirus]

That's superfluous I guess (that's not a rocket science). Good datasheet plus a couple of experiments can give and answer too, "accurate enough for Australia". 

Well it's always good to have a general understanding of the phenomenon and its associated math. You don't always need to make precise calculations of course. But it allows to understand what degree of approximation is going to be fine for your specific practical scenario.

[PinheadBE]

I'm not arguing about datasheet statements. I just try to help you so I say which exact current this power supply will consume at 230 VAC.

(BTW I have several MeanWell HDR-150 at my hands at the moment, and I have measured their consumed current. It is about 1.0-1.1 A at 220 VAC at output rated power).

Of course, I was not thinking you were arguing      I find it also a bit over-estimated given the efficiency....
I was about to use a 4 A Slow blow fuse, expecting it to be strong enough to handle the remaining surge, after I'd placed the NTC.
Experimentation will tell.   I have enough spare fuses 

Well it's always good to have a general understanding of the phenomenon and its associated math. You don't always need to make precise calculations of course. But it allows to understand what degree of approximation is going to be fine for your specific practical scenario.

Agreed.   Your explanation was not far from what I thought it would be, and I always appreciate to understand the how's and why's, but calculating it in practice is too much for me       

Thanks to both of you !

[shapirus]

but calculating it in practice is too much for me       

The bare minimum one (not even necessarily an electr(on)ics hobbyist) should know is that P (Watts) = W (Joules) / t (seconds), where P is power, W is energy (or work), t is time. Therefore, Joules = Watts * seconds, from which you can infer much of what can be required for a practical application.
Integration, either numeric or analytical, comes into play when either of the variables is not constant or at least linear

[Vovk_Z]

I find it also a bit over-estimated given the efficiency....

RS-150 is a bit less efficient compared to HDR-150 (90% eff.), so we may add 0.1 A or 10% more to 1.0-1.1 A of HDR-150 consumption. So you'll have about 1.1-1.2 A consumption I guess. That's a bit too much for NTC rated for 1 A, so you have to look at 2 A rated or more.

[Vovk_Z]

once I'd figured how to install it in my case so that it can dissipate without burning anything.... 

That's a good question because NTC have ability not only burn something nearby but they can disintegrate into parts when they broke. So, they are both easy to implement and can be dangerous at the same time.

[Kalvin]

Hi,

I have a SMPS MeanWell RS-150 (datasheet attached) with a 40 A inrush current, and 2 A typical current.

I have already blown a power-meter because of that inrush current.

So, I would like to limit that inrush current by inserting a NTC in the mains input line.

For the voltage, I have to choose 400 V, OK. (We have 230 VAC here)

You should perform the inrush calculations using the peak value of the mains voltage: 230VAC_RMS * SQRT(2) = 325VAC_Pk.

400V NTC may be sufficient, but I would select the next higher available voltage rating.

[PinheadBE]

Those figures are from the DS 

[srb1954]

Thanks srb for your answer 

I still have a question:

(...)
5. Determine if the NTC has sufficient surge energy rating to handle the total energy absorbed in the power-on surge. This may be either specified as the maximum allowable capacitive load or the maximum energy in Joules. You may need to measure the actual circuit in operation to determine the energy figure.

How do I calculate the energy in Joules ?   Is it proportionate to the surge current and the time it lasts ? If so, unfortunately, the DS does not mention it....   I suppose that time is a kind of curve with a parabolic descent, and the energy is the integration of that curve ?
Maybe, by being conservative, I could choose a default value for the NTC ?

It may be easier to build up a circuit and measure it. You will need a scope to measure the voltage across the NTC and the current through it. You can then use the scope waveform maths to multiply the current and voltage waveforms to obtain the power waveform and then integrate that to obtain the total energy in Joules.

If you don't have a suitable scope you could possibly simulate the circuit in SPICE but it will be complicated getting a suitable SPICE model for the NTC that incorporates both the thermal properties (thermal mass and dissipation constant) as well as its electrical properties. I am not aware of any suitable models being available but I haven't done any significant research on that topic.

[Siwastaja]

Also remember that the NTC isn't a true inrush current limiter. If you quickly re-apply power, before the NTC has cooled down, but with enough load that power supply capacitors (that cause the inrush to begin with) have discharged, then inrush current is limited only by the hot resistance of the NTC. If that suffices, then you could have used fixed resistor instead to begin with. With NTC, the common strategy is to accept that "sometimes" the inrush current is only limited to the hot resistance, this would be the critical worst case design parameter, and then the cold resistance would, sometimes, maybe, offer even lower inrush current, on average case, but that's only a bonus.

Best would be to design the upstream to handle that inrush current. Meanwell designers have designed their product to certain level of inrush current and 99.99% of applications are fine with it. Maybe a better idea instead of trying to reduce inrush current, is to fix your use case to fit the typical - if at all possible, unless your application is truly special. On the other hand, if your application truly is sensitive to high inrush current for whatever reason, you might want a better limitation strategy than NTC - for example, a resistor, protected by thermal fuse, bypassed by relay, plus control logic.

If it killed a "power meter", it must mean this meter was just shit and totally unsuitable for the job.