Inrush current limiting, PTC or NTC?

[reyntjensm]

Hello everyone,

I have a 2 * 1000uF capacitor bank( 2 in parallel) with an ESR of 120mOhm. The circuit will be powered from 230VAC grid trough a bridge rectifier. I'm trying to limit the inrush current but i have no idea what is the best way to go. I know it's possible with a NTC or a PTC but what is the best solution? On what does it depend? The circuit will draw a constant current around 16 amps. I'm trying to simulate this in LTspice but i can't find any ready made symbol files to do so. What is the best way to simulate this in LTspice? Can anybody help me out with this?

Thank you all!

[Vovk_Z]

NTC resistors are made to withstand inrush currents.

[eblc1388]

... The circuit will draw a constant current around 16 amps.

With that high current the NTC will waste a lot of power as heat during normal operation. A better option would be a series resistor and then bypassing it with a relay contact after the capacitors have nearly charged up.

[Circlotron]

... The circuit will draw a constant current around 16 amps.

With that high current the NTC will waste a lot of power as heat during normal operation. A better option would be a series resistor and then bypassing it with a relay contact after the capacitors have nearly charged up.

Yep. An even better way perhaps is to use an NTC bypassed with a relay.

[Siwastaja]

Inrush current limiter needs considerable design, modeling and calculation. You just don't put an NTC there and call it a day.

Optimum solution is a fixed-value resistor coupled with thermal fuse (or possibly PTC) in series, this combination bypassed with a relay, with control logic.

NTC is a cheap alternative suitable for small supplies which are "almost OK" without any surge limitation at all, and is based on the fact you can accept that the inrush current is just reduced, not eliminated, and you can accept high inrush current under some edge cases, or you do rigorous analysis comparing the capacitor discharge time constant (based on the applied load) to the thermal time constant of the NTC cooldown.

[trobbins]

Have you worked through the many NTC inrush manufacturer design guides, as they have your design and selection advise all nicely set out.

[eblc1388]

Yep. An even better way perhaps is to use an NTC bypassed with a relay.

No, it is not. Using a NTC is:
a. more expensive then a power resistor
b. less reliable than a power resistor
c. offer no inrush protection if power is off and on in quick succession as the NTC needs time to cool down.

[Siwastaja]

If the rectified DC voltage is, for example, 320V, and the load (say, a PFC boost controller, for example) goes into Under-Voltage Lockout at, say, 80V, then the quick discharging of the caps stop at 80V, the capacitor voltage stays there for long enough to the NTC cool down, and if the supply is re-plugged while the NTC is still hot, the inrush current is 80V/320V = 25% less than it would be to empty capacitor bank. Not a big difference, but at least some limitation...

Add the hot resistance of the NTC, and it limits the inrush current even if it's still hot, compared to not using one at all.

This hot resistance, obviously, adds power loss and heating all the time is device is on, which totally sucks. So you need to think how to thermally insulate the NTC so it keeps hot at lower power, and not destroy nearby electrolytic capacitors.

[Circlotron]

Yep. An even better way perhaps is to use an NTC bypassed with a relay.

No, it is not. Using a NTC is:
a. more expensive then a power resistor
b. less reliable than a power resistor
c. offer no inrush protection if power is off and on in quick succession as the NTC needs time to cool down.

Seeing the NTC is bypassed by the relay, if the psu had been running long enough it would have cooled down and be ready if the power went off-on.

[Siwastaja]

Yep. An even better way perhaps is to use an NTC bypassed with a relay.

No, it is not. Using a NTC is:
a. more expensive then a power resistor
b. less reliable than a power resistor
c. offer no inrush protection if power is off and on in quick succession as the NTC needs time to cool down.

Seeing the NTC is bypassed by the relay, if the psu had been running long enough it would have cooled down and be ready if the power went off-on.

You can use NTC there (and add even more complex control logic to catch one more edge case you create with it); or you can choose the NTC so that it never heats up during the precharge, only utilizing its cold resistance; or you can use a potato there. It still makes absolutely no sense to use NTC with the relay. The point of the NTC is that it's a cheap replacement for the relay + control logic as it's "self-controlling". If you do have the relay + control logic, you can just use a fixed-value resistor, and choose optimum value for it. Adding protection for failed relay (or relay control logic) is always a good idea, though; that would be a thermal fuse in series.

Such controlled precharge tends to cost at least a $2-3 in BOM so it isn't feasible in sub-$100 supplies. Small supplies do without inrush limiting at all, or, use NTC. The range of supplies using NTC is quite small, in the end, they are some 50-100W supplies. This power level is ubiquitous, though, so you see a lot of NTCs. And because they sell in millions, they have the resources to choose the NTC properly, design the thermal insulation and placement for it, and model and verify the thermals of the complete product.

[akimpowerscr]

Yep. An even better way perhaps is to use an NTC bypassed with a relay.

No, it is not. Using a NTC is:
a. more expensive then a power resistor
b. less reliable than a power resistor
c. offer no inrush protection if power is off and on in quick succession as the NTC needs time to cool down.

+1

NTC resistor is not a good solution.

The NTC solution is an inexpensive solution used in poor quality SMPS power supplies.

There is one essential point that the OP did not indicate: what kind of rush current is it?

Indeed, the inrush current can have 2 different causes:
- the charge of a capacitor
- the saturation of a transformer on power-up.*

These two types of inrush current can have different solutions.

[Siwastaja]

NTC doesn't necessarily indicate poor quality.

It's a low-cost solution that fits some particular cases just right, and if the design is done properly, it isn't a bad or unreliable, and not even colossally inefficient (i.e., most of the losses are elsewhere).

It's not optimal from efficiency viewpoint though, but neither is energizing a relay coil all the time. For really high-efficiency products, a lot of thinking needs to go in a product, one part of that is optimizing the precharge circuit.

[opampsmoker]

Good inrush limiter is use an isolated 1W SMPS module, ....say feeding off a low voltage dc bit...and then get this to turn on the fet that shorts out the inrush resistor.......this is good because no logic is needed, as the low voltage bit would power up after the pfc anyway...so your inrush has happened (thru the inrush resistor), and then you can switch out the inrush resistor as i describe.

[reyntjensm]

Thank you all for your feedback.
It's a one of a kind prototype so i'm not worried about heat/energy loss.
So i think a NTC would be the best option. Any idea's on how to simulate this in LTspice?

[jonpaul]

NTC inrush is very old and reliable solution, used in many well designed supplies, including Tektronix 2465.

NTC: e.g. Cold = 100 ohms, hot (from current) = 1 ohm

PTC is the opposite of what is needed, e.g. cold = 1 ohm, hot = 100 ohm.


Relays and such add cost and reduce reliability.

See the many fine app notes from the NTC manufacturers.

https://www.ametherm.com/inrush-current/selecting-an-inrush-current-limiter.html

If properly selected and designed, the losses during operation are reasonable.


Jon

[Siwastaja]

Properly designed relay doesn't reduce reliability any more than a properly designed NTC.

Cost difference is true. NTC is a good idea for a 50-100W, low cost, reasonably efficient power supply, nothing shady in it.

I say it once more but do remember that NTC inrush limiter typically does not limit inrush properly in some very real corner cases, it's used to limit average/typical inrush current but if worst-case behavior is important (for example, to protect switches, mechanical or semiconductor), NTC is usually out of question.

[bdunham7]

Thank you all for your feedback.
It's a one of a kind prototype so i'm not worried about heat/energy loss.
So i think a NTC would be the best option. Any idea's on how to simulate this in LTspice?

Is the output of this circuit loaded as soon as it is powered up?  Is there an NTC available that can handle a 16 amp load like that?  You may need to parallel several devices and your inrush is still likely to be in the hundreds of amperes.

[Siwastaja]

Sanity check:

I think I have never seen NTC used in a multi-kW supply with 2000uF input capacitance after the rectifier. Such large supplies use relay for precharge.

It is possible yes but I'm 99% sure the solution will suck, i.e. require excessively large NTCs, possibly many paralleled, which then dissipate a lot of heat during the normal 16A load and need to be located far away from the capacitors.

I also think the safety of the OP's project should be questioned. If you are not very much sure what you are doing, and not qualified to do it, don't do it. Mains is dangerous. Just a reminder, feel free to ignore.

If it's a one-off hobby project or prototype, why spend time trying to specify NTC? The first time I did a surge limiter as a kid with a large 2kVA 24V transformer with some tens of thousands of uF at the output, I used a 150W halogen light bulb bypassed by simple mechanical switch in series with the primary. Worked a treat and provided an interesting current-limitation mode where the light intensity showed the level of output current! I used a standard 6A resetable fuse, properly mains rated, so that if I accidentally forgot the bypass switch on, this fuse would pop instead of the 16A house fuse.

[coppercone2]

Put a relay and timer on it with a simple circuit for that big a capacitance.

[akimpowerscr]

Hello everyone,

I have a 2 * 1000uF capacitor bank( 2 in parallel) with an ESR of 120mOhm. The circuit will be powered from 230VAC grid trough a bridge rectifier. I'm trying to limit the inrush current but i have no idea what is the best way to go. I know it's possible with a NTC or a PTC but what is the best solution? On what does it depend? .....
...
...
...
Thank you all!

The funny thing is that this is almost the same problem that happened in the power supply of the Philips chassis KM1 multistandart color televisions but the capacitor was only 470µF

https://onedrive.live.com/?authkey=%21ALsyL1wEiE%2Dmono&cid=D3F55CD5F0B2818B&id=D3F55CD5F0B2818B%216002&parId=D3F55CD5F0B2818B%216100&o=OneUp

The inrush current protection was done by a NTC.

I repaired dozens of these televisions which all still had the same problem: BY127 diodes (later replaced by BY227) really exploded.

It was enough to turn off the television and turn it on again a few seconds later and it was a disaster ...

So, those who say that CTN protection is reliable make me laugh. 

That's not a reliable protection, only a cheap solution.

And, with 100% certainty, totally unsuitable for a circuit that consumes 16A !!!!

[reyntjensm]

Thank you all for your feedback.
It's a one of a kind prototype so i'm not worried about heat/energy loss.
So i think a NTC would be the best option. Any idea's on how to simulate this in LTspice?

Is the output of this circuit loaded as soon as it is powered up?  Is there an NTC available that can handle a 16 amp load like that?  You may need to parallel several devices and your inrush is still likely to be in the hundreds of amperes.

The output isn't going to be loaded as soon as is powers up.

Sanity check:

I think I have never seen NTC used in a multi-kW supply with 2000uF input capacitance after the rectifier. Such large supplies use relay for precharge.

It is possible yes but I'm 99% sure the solution will suck, i.e. require excessively large NTCs, possibly many paralleled, which then dissipate a lot of heat during the normal 16A load and need to be located far away from the capacitors.

I also think the safety of the OP's project should be questioned. If you are not very much sure what you are doing, and not qualified to do it, don't do it. Mains is dangerous. Just a reminder, feel free to ignore.

If it's a one-off hobby project or prototype, why spend time trying to specify NTC? The first time I did a surge limiter as a kid with a large 2kVA 24V transformer with some tens of thousands of uF at the output, I used a 150W halogen light bulb bypassed by simple mechanical switch in series with the primary. Worked a treat and provided an interesting current-limitation mode where the light intensity showed the level of output current! I used a standard 6A resetable fuse, properly mains rated, so that if I accidentally forgot the bypass switch on, this fuse would pop instead of the 16A house fuse.

I've found a NTC that could be used, check the attachment. Everything has to be installed on a PCB and inside a box. I'm unable to install a 150W lamp inside the box....

[jbb]

“Max steady state current at 25C: 15A”

Nope, too small, sorry. That will likely mean something like “operating at maximum temperature of 185C I’m free 25C air”.

If you assume ambient goes up a bit and you’ve got a box, the air temperature inside would be, oh, 60C or more. So you’d have to derate it to like 10A or less.

And the PTC itself would be a destructive monster, leaning heat into everything nearby. I would worry about it wrecking it’s own solder joints which could then get very hot and literally set the board on fire. (Note: flame retardant does not mean non-flammable, and if you keep pumping electricity in it will keep burning).

[bdunham7]

The output isn't going to be loaded as soon as is powers up.

Then you can easily use a relay and power resistor to limit the inrush to arbitrarily low levels. 

These two parts will give you a robust solution that will be more reliable, generate a lot less heat and easily withstand repeated power cycles.  It will also limit the inrush to about 15 amperes maximum, which you can change if you wish by selecting a different resistor.  Your NTC will be roasting hot, will not reliably limit the inrush to any specific value and any momentary power interruption will give you that huge surge that you are trying to avoid. 

https://www.mouser.com/ProductDetail/TE-Connectivity-PB/1649341-2?qs=sGAEpiMZZMtGt%252Bn33CgIPwDNQq27pAXuXZwiRFgJ10g%3D

https://www.mouser.com/ProductDetail/Bourns/WS5M22R0J?qs=R%252BZOTiLH6ThNZ0lJxIR7uw%3D%3D

[bdunham7]

NTC inrush is very old and reliable solution, used in many well designed supplies, including Tektronix 2465.
Relays and such add cost and reduce reliability.

Yes- they work very well in designs where they aren't really needed--the one you cited is a perfect example of this.

In the OPs use case, the relay/resistor combo will cost less and likely be more reliable than any reasonable NTC "solution".  It will also actually do the job of reliably limiting inrush current.

[akimpowerscr]

Let's not forget that there is also the possibility of using a triac instead of a relay to short circuit the inrush current limiting resistor.

But with 16A of consumption, it will take a rather powerful triac with radiator and the power loss will be about 30W

[Ice-Tea]

Another option is to use a FET to limit inrush current protection. Keeping inside the SOA, especially during the current-limiting phase takes some juggling, though...

[akimpowerscr]

Using a FET in a linear configuration is never recommended.
So this is not a good solution.

On the other hand, by placing the inrush current limiting resistor on the DC side, we can use an FET to short circuit this resistor.

[Ice-Tea]

Using a FET in a linear configuration is never recommended.
So this is not a good solution.

Bold statement.

[akimpowerscr]

Using a FET in a linear configuration is never recommended.
So this is not a good solution.

Bold statement.

Power Mosfets have generally been designed to operate in switching, not linear mode.

It will be necessary to charge a capacitor of 2000µF to more than 300V, it is not negligible at all for o MOSFET.

See Microsemi paper about all the precautions which it is necessary to take for a use in linear mode of MOSFET.

Why take risks, when a simple wirewound power resistor can do the same safely and quite more reliably.

Better use of MOSFET is to short circuit the power resistor when the capacitor is fully charged.

[Siwastaja]

Using a FET in a linear configuration is never recommended.
So this is not a good solution.

Bold statement.

It never stops amusing me why people post completely random claims off the top of their heads like this.

Heck, whole FET series exist marketed for linear configuration. MOSFETs are common as salt in class AB audio amps since some 1980's.

It's true though that many are primarily meant for switching applications and linear SOA specification might be nonexistent, or require further research and derating.

In any case, there is no free lunch when it comes to precharge. It always requires understanding and proper design modeling and math.

[Ice-Tea]

It will be necessary to charge a capacitor of 2000µF to more than 300V, it is not negligible at all for o MOSFET.

Which is exactly why I stated that keeping it in the SOA takes some juggling. The Microsemi document is a 6-pager stating exactly the same thing.

Why take risks, when a simple wirewound power resistor can do the same safely and quite more reliably.

Can you show me one of those "simple" wirewound resistors that will do the job? Either you'll be charging those caps all days (which would be equally trivial for a FET based solution) or you will have to consider SOA of the resistor, again: just like for a FET.

[Siwastaja]

SOA of the resistor is typically easy IF you can guarantee your bypass switch AND its control logic always work. That part is non-trivial. So you need to consider the protection added for the resistor, and the SOA of this combination, in case the control logic keeps the resistor in-circuit and the load is on.

Precharge is one of those awkward things you wouldn't want to spend time on, yet there is no free lunch.

If controlling the load is completely in your hands, you can do all sorts of interesting stuff, for example I have done a thing where a "soft" power switch precharges the DC bus through a relatively large value resistor, a microcontroller boots through this resistor, and it's up to the software to measure the bus is at high enough voltage to bypass the resistor + soft power switch combo. Only after the bypass is enabled, then the microcontroller turns on the actual loads (it's there doing that anyway).

[akimpowerscr]

A wirewound resistor behaves like a fuse and its limiting characteristic is therefore given by his I²dt like a fuse. (NB:dynamic behaviour with pulse)

The concept of SOA does not apply to a wirewound resistor.

[Siwastaja]

A wirewound resistor behaves like a fuse

This is just ridiculous.

Yes, of course a wirewound resistor finally blows open-circuit, like a fuse.

It's well capable of starting a fire before that happens, though!

A fuse does not start a fire as a normal part of fusing. That's the difference, and that's why any wirewound resistor isn't a fuse.

There is a thing called "FUSIBLE RESISTOR" which conforms to your description "behaves like a fuse". Just getting a "wirewound resistor" doesn't mean you'll get a fusible resistor. Get one RATED for that, if you want it. Nevertheless, even fusible resistors have the tendency of causing serious damage around, and the common practice is to consider placement carefully, then use glass fiber - silicone sleeving to thermally insulate the part.

The concept of SOA applies to the complete system or combination of parts, in this case the resistor, its protection, and the surrounding components sensitive to heat.

You clearly lack the ability to properly design such circuit, let alone give advice about it.

[akimpowerscr]

A wirewound resistor behaves like a fuse

This is just ridiculous.

Yes, of course a wirewound resistor finally blows open-circuit, like a fuse.

It's well capable of starting a fire before that happens, though!

A fuse does not start a fire as a normal part of fusing. That's the difference, and that's why any wirewound resistor isn't a fuse.

There is a thing called "FUSIBLE RESISTOR" which conforms to your description "behaves like a fuse". Just getting a "wirewound resistor" doesn't mean you'll get a fusible resistor. Get one RATED for that, if you want it. Nevertheless, even fusible resistors have the tendency of causing serious damage around, and the common practice is to consider placement carefully, then use glass fiber - silicone sleeving to thermally insulate the part.

The concept of SOA applies to the complete system or combination of parts, in this case the resistor, its protection, and the surrounding components sensitive to heat.

You clearly lack the ability to properly design such circuit, let alone give advice about it.

You should be more reserved before expressing personal opinions on my skills so as not to make a fool of yourself.

You don't know me, you don't know my qualifications, or my professional experience, so I wonder what criteria you allow yourself to judge me on.

In addition, it is not fair play and against the rules of this forum.

The main component of a wire-wound resistor is a wire, as is the case with a fuse, and during a short-lived strong current pulse, the heat transfer coefficient between the wire and the body of the resistance and his thermal resistance can be overlooked.

With a high peak current, the wire melts; like a fuse would.

For high impulse currents, it is therefore the i²dt which is the upper physical limit of destruction of the resistance.

This defines what is the maximum peak current that the resistor can withstand and for how long without being destroyed, being clear that these are only very short pulses.

This is what happen when charging a capacitor....High initial peak current who drops to 37% of his value after T (time constant) and to 13% after 2T.

The number of joules that will be dissipated in the resistor during the capacitor recharge sequence can easily be calculated.

It is enough to check if this dissipated power will not raise the temperature of the resistance beyond the permit, taking into account the thermal inertia of the resistance.

For the application of OP, he could choose a 33R 50W resistor.

NB: If the temperature of the wirewound resistor exceeds the allowable limit temperature, the resistor will not instantly be destroyed but its service life will be reduced.

This behavior is different from that of a semiconductor which suffers from other phenomena such as the second breakdown and instantaneous destruction and that is why the notion of SOA does not apply to a wirewound resistance

[bdunham7]

For the application of OP, he could choose a 33R 50W resistor.

That seems very large. What is your math on that? And why would you reduce the inrush current to less than the operating current?

[akimpowerscr]

Indeed, it is 3.3R 50W, not 33R 

Thank you for this correction.

[ejeffrey]

Sure you could use a triac or solid state relay but relays for this application are cheap and plentiful.  Also due to the resistor bypassing the contacts it will never see large voltages or have problems with arcing.

[bdunham7]

Indeed, it is 3.3R 50W, not 33R 

Thank you for this correction.

Well, now now you have a maximum surge of over 100 amperes, which won't do much good for protecting something like the on/off switch contacts.  After all, there must be some reason to limit the inrush, right?  The power utility certainly won't care.  But I was also referring to the power rating, which seems excessive and inadequate (or at least incomplete) at the same time.  In an earlier post I had proposed a specific 22R 5W resistor that I think would be adequate--have a look at the datasheet. 

[Gyro]

A wirewound resistor behaves like a fuse and its limiting characteristic is therefore given by his I²dt like a fuse. (NB:dynamic behaviour with pulse)

The concept of SOA does not apply to a wirewound resistor.

Every power dissipating device has a SOA. They may (are) not all be the same shape, but the fundamental power dissipation (separated out as  Voltage and Current) vs time is universal.

For instance, your short pulse wirewound resistor element fusing, before the heat has time to dissipate into the substrate, is virtually identical to the short term bond wire failure limit in a transistor.

[akimpowerscr]

Indeed, it is 3.3R 50W, not 33R 

Thank you for this correction.

Well, now now you have a maximum surge of over 100 amperes, which won't do much good for protecting something like the on/off switch contacts.  After all, there must be some reason to limit the inrush, right?  The power utility certainly won't care.  But I was also referring to the power rating, which seems excessive and inadequate (or at least incomplete) at the same time.  In an earlier post I had proposed a specific 22R 5W resistor that I think would be adequate--have a look at the datasheet.

For a power supply that provides 16A, the mains circuit breaker is at least 25A, the switch also and the diode bridge 35 or 50A.
None of these components will be damaged by a short time peak current of 100A
Regarding the 50W resistor, the choice of this power is guided by the pulse energy that the resistor can withstand.
For a resistance of 3.3R, the maximum for a resistance of 50W is 120 joules.
The energy dissipated in the resistor when charging a 2000 µF capacitor at 320V is:
E = 1/2 CV² = 102 Joules. So, it is compatible

https://docs.rs-online.com/e1d8/0900766b81360537.pdf

[akimpowerscr]

A wirewound resistor behaves like a fuse and its limiting characteristic is therefore given by his I²dt like a fuse. (NB:dynamic behaviour with pulse)

The concept of SOA does not apply to a wirewound resistor.

Every power dissipating device has a SOA. They may (are) not all be the same shape, but the fundamental power dissipation (separated out as  Voltage and Current) vs time is universal.

For instance, your short pulse wirewound resistor element fusing, before the heat has time to dissipate into the substrate, is virtually identical to the short term bond wire failure limit in a transistor.

You are completely wrong, SOA only applies to transistors, IGBT's and MOSFETs, not to thyristors, triacs and diodes, nor, of course, to wirewound resistors.

Consult the thyristors, diodes and triacs datasheets and you will never find any especification of SOA, but well of the i²dt of the semiconductor.

[T3sl4co1l]

So this might be a good point to remind people that AC-input PFC stages can offer input current limiting.

Not so much the FET-bridge configuration (though the savings on FWB losses is handy), but one with a SEPIC or flyback configuration -- assuming you don't mind the excess voltage.  Which may necessitate 800V+ devices -- SiC would be an excellent candidate then.

A SEPIC inductor for that power level will certainly be a custom part.

Tim

[bdunham7]

None of these components will be damaged by a short time peak current of 100A
Regarding the 50W resistor, the choice of this power is guided by the pulse energy that the resistor can withstand.

If you have a power switch, the peak will occur right as the contacts are engaging.  If they are rated for that type of use, great.  Not all switches are.

As for your resistor pulse specs, I don't think those specs mean what you think.  They don't list the test conditions, but the one I listed does mention a specific 50uS pulse test.  It doesn't make sense that this would be the thermal limitation because if you look at the graph, somewhat higher resistances (like 22R) have pulse energy limitations of less than 10 Joules.  This is more likely driven by voltage or dV/dt considerations, not thermal limits.

[Gyro]

A wirewound resistor behaves like a fuse and its limiting characteristic is therefore given by his I²dt like a fuse. (NB:dynamic behaviour with pulse)

The concept of SOA does not apply to a wirewound resistor.

Every power dissipating device has a SOA. They may (are) not all be the same shape, but the fundamental power dissipation (separated out as  Voltage and Current) vs time is universal.

For instance, your short pulse wirewound resistor element fusing, before the heat has time to dissipate into the substrate, is virtually identical to the short term bond wire failure limit in a transistor.

You are completely wrong, SOA only applies to transistors, IGBT's and MOSFETs, not to thyristors, triacs and diodes, nor, of course, to wirewound resistors.

Consult the thyristors, diodes and triacs datasheets and you will never find any especification of SOA, but well of the i²dt of the semiconductor.

I wasn't referring to it having a datasheet graph! I exists as a property nontheless. If pushing a component outside its normal ratings (such as a resistor surge then you need to characterize it for yourself (there is no rule of thumb).

Some resistors (eg fusible ones) give you operating graphs for 'safe' operation (fusing is not part of their everyday operating mode, you need to ensure reliable operation). The fusing characteristics plot of a fuse is effectively a SOA graph too. Fuse elements have temperature coefficients, so their characteristics are not simply dI/dt.

Many components (including Diodes and SCRs) have temperature related parameter changes which come into play depending on the magnitude and rise time of operating current. Just because the manufacturer doesn't provide you with a convenient SOA graph doesn't mean that you should not consider these factors during design.

You are completely disrespectful.

[akimpowerscr]

To say that you are wrong is not to be disrespectful.

We are on a technical forum and we are discussing technical topics.

You do not provide any technical document to support your theories.

You talk about the fusion of the wires of a transistor, whereas it is the semiconductor which melts well before the wires.

Datasheet of a fast fuse.....Where you see SOA in this document ?

You are only asked one thing: if you feel you are right, present evidence, documents, publications.

On the other hand, I have dozens of publications on thyristors, triacs and none cite the word SOA.

[trobbins]

Why aren't posters initially just addressing the OP's request?  If there is nothing commercially available, then by all means move on to other methods, or provide advise as to the risks, but at least start from a known base.

Many commercial NTC product ranges go up to 16A max continuous, and as pointed out, the NTC datasheets and guides identify the derating that is required for that max current selection.  So let's look for product ranges with higher max rating.

Littlefuse ST range is pretty easy to find, and has a 18, 20 and 30A max parts that can be assessed.

Although LTSpice can be used, I'd suggest the simpler PSUD2 application would be easier to give initial assessment.  As an example, the ST1R020B with 20A max rating has a 1 ohm cold rating, and PSUD2 indicates an initial pulse of 110Apk, settling to 50Apk in 2-3 half cycles, so fuse or CB discrimination assessment can be pursued.  The steady state dissipation would be circa 4W.  Unfortunately Littlefuse have scant additional tech data on that product, but TDK has B57464S0109M000, a 20A part which indicates a 1 ohm cold rating, and suitable for up to 2,500uF at 230Vac input.

Yes there is more to NTC assessment than this, but it indicates there are commercial parts that could be ok to deploy - so perhaps the thread discussion should have started by critically assessing a standard NTC solution, to eek out the implications for the OP, and once the implications are well appreciated THEN move on to how other options may present a better outcome based on the OP's specific requirement.

[Zero999]

Good inrush limiter is use an isolated 1W SMPS module, ....say feeding off a low voltage dc bit...and then get this to turn on the fet that shorts out the inrush resistor.......this is good because no logic is needed, as the low voltage bit would power up after the pfc anyway...so your inrush has happened (thru the inrush resistor), and then you can switch out the inrush resistor as i describe.

That's a sensible enough solution, but you might still need some sort of timer, as the DC:DC converter will probably work at a low output voltage, when there's still a decent inrush current. Also beware that many cheap DC:DC converter modules don't have enough isolation for mains use.

[bdunham7]

Why aren't posters initially just addressing the OP's request?

The OP's original quest included:

I'm trying to limit the inrush current but i have no idea what is the best way to go.

He did mention NTCs and PTCs, but he certainly didn't just ask someone to spec out an NTC.  I'd be curious as to what the rest of the story is, but he didn't provide it and asked for generic advice.  Unless we know what the exact desired characteristics are--if they are even known, there's not much else to say.  Some seem to think that a few hundred amps of inrush are no problem, I offered a much gentler and adaptable solution at a reasonable cost.

[eblc1388]

How do one connects up the relay with a 240Vac coil as recommended in Reply#22?

I always thought choosing the relay is non trivial. I personally would go for one with a coil voltage of 48V DC up to 250V DC. Of course there would be some timing circuits to make sure it will only operate seconds after power is ON.   

[Gyro]

To say that you are wrong is not to be disrespectful.

We are on a technical forum and we are discussing technical topics.

You do not provide any technical document to support your theories.

You talk about the fusion of the wires of a transistor, whereas it is the semiconductor which melts well before the wires.

Datasheet of a fast fuse.....Where you see SOA in this document ?

You are only asked one thing: if you feel you are right, present evidence, documents, publications.

On the other hand, I have dozens of publications on thyristors, triacs and none cite the word SOA.

It is disrespectful to tell somebody they are "completely wrong" without addressing the valid technical arguments that they have put forward. As you say, this is a technical forum.

With regard to fusing bond wires, please see the attached datasheet. You will see in Figure 5 that the upper horizontal portion of the curve is specifically labelled as a bond wire limitation. This is the case with most transistors, although the manufacturer often does not often call out the specific reason for that horizontal portion. It must surely be obvious that when a transistor is in saturation or running at high current with low V_CE, the die will be dissipating relatively little power, while the Emitter bond wire is carrying full current. It is not a die dissipation issue at this point. The distinction between switching devices designed for linear and switching purposes is the die construction and danger of current hogging hot-spots* and secondary breakdown, this is represented in other areas of the SOA graph. In all devices though, there is always a hard limit on the upper horizontal portion of the SOA which represents bond wire limitations. I would have hoped that you would have known this.

* By the way I hope that you know that slow rise time application of gate current in large wafer area 'puck' SCRs can also cause localized hot-spots leading to premature failure.

I really don't care how many publications you have (I have many myself). If you blindly say that because you don't see the magic letters SOA, a device doesn't have an area of voltage, current and, (in the case of pulse operation) duration, where it can safely be operated then you are not understanding. This area (accompanied by the letters SOA or not) clearly exists.

In your fuse datasheet, I can clearly see a safe (as in continued operation of the equipment that it is protecting) operating area. It is the area within the Time Current curve where the fuse does not blow. Just because it doesn't have the magic letters SOA doesn't mean that it isn't the area where the equipment will continue to operate satisfactorily. A fuse is a special case - you are also interested in it fusing gracefully on overload. In the case of a resistor, you are more interested in not exceeding it's safe operating conditions leading to long term or catastrophic failure.

Please stop fixating on the whether or not the letters SOA appear in a datasheet and understand the underlying concepts of whether a component is operating at a point where it will safely continue to operate.

[trobbins]

 reyntjensm, it may be best if you elaborate on your power supply.  Is your mains actually 230V, and is their a nominal range that you see or need to cope with?  Is the load really a constant 16A, and is that a measured mains true-rms level (given you are direct rectifying it), or a DC level load current, and what tolerance are you expecting on the 16A?  Is the mains connection a fixed wired installation or are you using something like a 20A plug/socket?  Are you including an over-current protection device on the incoming mains in the device and what is it?  Same question for the upstream mains feed distribution board?

[akimpowerscr]

To say that you are wrong is not to be disrespectful.

We are on a technical forum and we are discussing technical topics.

You do not provide any technical document to support your theories.

You talk about the fusion of the wires of a transistor, whereas it is the semiconductor which melts well before the wires.

Datasheet of a fast fuse.....Where you see SOA in this document ?

You are only asked one thing: if you feel you are right, present evidence, documents, publications.

On the other hand, I have dozens of publications on thyristors, triacs and none cite the word SOA.

It is disrespectful to tell somebody they are "completely wrong" without addressing the valid technical arguments that they have put forward. As you say, this is a technical forum.

With regard to fusing bond wires, please see the attached datasheet. You will see in Figure 5 that the upper horizontal portion of the curve is specifically labelled as a bond wire limitation. This is the case with most transistors, although the manufacturer often does not often call out the specific reason for that horizontal portion. It must surely be obvious that when a transistor is in saturation or running at high current with low V_CE, the die will be dissipating relatively little power, while the Emitter bond wire is carrying full current. It is not a die dissipation issue at this point. The distinction between switching devices designed for linear and switching purposes is the die construction and danger of current hogging hot-spots* and secondary breakdown, this is represented in other areas of the SOA graph. In all devices though, there is always a hard limit on the upper horizontal portion of the SOA which represents bond wire limitations. I would have hoped that you would have known this.

* By the way I hope that you know that slow rise time application of gate current in large wafer area 'puck' SCRs can also cause localized hot-spots leading to premature failure.

I really don't care how many publications you have (I have many myself). If you blindly say that because you don't see the magic letters SOA, a device doesn't have an area of voltage, current and, (in the case of pulse operation) duration, where it can safely be operated then you are not understanding. This area (accompanied by the letters SOA or not) clearly exists.

In your fuse datasheet, I can clearly see a safe (as in continued operation of the equipment that it is protecting) operating area. It is the area within the Time Current curve where the fuse does not blow. Just because it doesn't have the magic letters SOA doesn't mean that it isn't the area where the equipment will continue to operate satisfactorily. A fuse is a special case - you are also interested in it fusing gracefully on overload. In the case of a resistor, you are more interested in not exceeding it's safe operating conditions leading to long term or catastrophic failure.

Please stop fixating on the whether or not the letters SOA appear in a datasheet and understand the underlying concepts of whether a component is operating at a point where it will safely continue to operate.

You persist in the error, yet very simple checks could show you it, but you refuse to do it.

A search on the Internet could allow you to verify that the SOA is always associated with bipolar transistors, Mosfets and IGBTs, and only them.

You will not find anything that combines the principle of SOA and a wirewound resistor, or with a fuse, contrary to your claims.

The SOA would however be an essential characteristic (if it existed for those components !) that one could not under any circumstances ignore in technical specifications.

As I told you before, no datasheet nor technical documentation of thyristors, fuses and wirewound resistors talks about SOA.

You invent theories which are incorrect and you persist in error.

There is no worse deaf than the one who does not want to hear 

[Siwastaja]

All properly rated resistors include what is equivalent to transistor SOA curves, the title just isn't SOA. It's usually titled "derating curve". The concept is the same: operating parameters inside which operating the part is safe*.

*) assuming the designer also knows how to design the part in safely, including but not limited to, cooling the part.

SOA isn't a trademark term, it's a concept. The English words where it comes from are: "safe operating area". Anything that has safe operating area (and, as a corollary, outside of this area is unsafe), can be called as having a safe operating area. You could think this is obvious.

Yes I agree you won't find the string "SOA" from resistor datasheets. This is irrelevant.

Wirewound resistors are easily capable of producing localized heat well over 500degC before the wire melts. This easily causes a fire if there are flammable parts nearby. This is the difference to an actual fuse. Suggesting a wirewound resistor would act as a fuse is one of the most dangerous advice seen for quite some time here.

Design with resistors so you can prove the thermals never exceeding the threshold of catastrophic damage. If unsure, add thermal fuses, or use a fusible resistor. Even then, you need to prove the protection actually protects; even this is nontrivial.

You have given horribly bad and dangerous advice, to which we have responded giving correct information. There is nothing more to say, and going back and forth doesn't help anybody. One more troublemaker to ignore. I hope the OP stays safe.

[Gyro]

There is no worse deaf than the one who does not want to hear 

Thanks, I was looking for a suitable phrase.

[akimpowerscr]

 

All properly rated resistors include what is equivalent to transistor SOA curves, the title just isn't SOA. It's usually titled "derating curve". The concept is the same: operating parameters inside which operating the part is safe*.

The behavior of a resistor is totally different from that of a bipolar transistor.
This is why there are different notions that apply to transistors and resistors.

SOA and "derating curves" are not the same concept.

A resistor may remaind safe even if it is reasonably overloaded.

Suggesting a wirewound resistor would act as a fuse is one of the most dangerous advice seen for quite some time here.

That's what happens when you apply to the wirewound resistor a pulse voltage of short duration and of high voltage.

Peak current is sufficient to melts the wire, but average power stay low, so the resistor stay cool.

This is simply what could happen when you use a resistor to limit charging current of a capacitor and when you choose to charge it in a short time with high current.

[bdunham7]

How do one connects up the relay with a 240Vac coil as recommended in Reply#22?

I always thought choosing the relay is non trivial. I personally would go for one with a coil voltage of 48V DC up to 250V DC. Of course there would be some timing circuits to make sure it will only operate seconds after power is ON.

By simply having the resistor on the AC input after the switch and fuse, placing the contacts of the relay across the resistor and the coil of the relay across the AC input line after the resistor.  Admittedly this is a rather simple and crude way of doing it, but I was trying to show the OP how easy it can be.  In this case, I don't have enough specs on the relay to tell exactly how nicely it will work--but if the pull-in voltage is 180VAC and the actuation time is 10mS, for example, it will work OK according to my back-of-an-actual-envelope chicken scratching.  There may be a 'bump' in inrush current when the relay engages, but the simple circuit has some advantages--it latches on independently of load, it works quickly,  and it can be power-cycled repeatedly and will still work each time. And, of course, adjustments to component values might be useful once we have the complete picture of the desired operating characteristics.

And, in case anyone notices, it has not escaped my attention that if the load is connected when the power is turned on or cycled, the resistor will fry quite quickly.  This is why I asked the OP about the load.  Obviously if there is any question about the load remaining, the relay should be DPST and only connect the load when engaged, or some other method of disengaging the load during the precharge should be used. 

[akimpowerscr]

And, in case anyone notices, it has not escaped my attention that if the load is connected when the power is turned on or cycled, the resistor will fry quite quickly.

This is why to choose a lowest possible value of resistor is a good choice.

With 3R3 and 16A, voltage drop on the resistance is 52.8V and you may short-circuit the resistance after a short delay (let's say 1s for exemple) without damaging the components.

NB: suggested relay
Finder 62.32 with coil 240Vac, 2 poles in =

http://www.farnell.com/datasheets/2237785.pdf

[bdunham7]

This is why to choose a lowest possible value of resistor is a good choice.

With 3R3 and 16A, voltage drop on the resistance is 52.8V and you may short-circuit the resistance after a short delay (let's say 1s for exemple) without damaging the components.

Without knowing the load or the desired limit for inrush, I can't say. What if the load draws more current at lower voltages? I think we have enough information to give ideas, but not enough for definite prescriptions.  Simplicity, robustness, size and cost are all competing objectives.  The fact that this is a one-off and not production means that simplicity is a big factor--the OP's project is not inrush current control, he has some other objective.  If it has to be very simple to implement and also needs to be capable of starting or restarting into a load, perhaps an NTC might be the answer despite the other problems. 

If you use timers and control logic the possibilities are endless.  For example, you could use my 22R example, a 4.7uF capacitor in series and a 5-second startup timer, along with load disconnect via a separate relay.  But that is a lot more stuff for just one function that may not be all that important.

[akimpowerscr]

Since the load is not connected when power is applied, a very simple solution would be to short circuit the current limiting resistor only when the load is connected.

An auxiliary contact of the load switch could supply power to the short-circuit relay.
The circuit would remain very simple, no need for a timing circuit in this case.

[reyntjensm]

Thank you all for the feedback. I think i'm going for the relay solution( since this is more easy to model in LTspice  ).
According to my simulations, a 10ohm resistor would be a good choice to keep the inrush current reasonable low.

But i'm still interested in the NTC solution. Somebody suggested a NTC with a 1ohm cold resistance. I think this would not help to limit the inrush current? I should need a NTC with around 10 ohm cold resistance right? I've been looking at different NTC manufacturers but i couldn't find any with a 10ohm cold resistance who could handle 20A continuously( so i have some margin). Most of the NTC's have a very low cold resistance in the current range of 20A. Is this correct? I guess i don't have a good understanding of how the datasheets define a NTC. So a relay it is! haha

[trobbins]

reyntjensm, you have provided no background information as to why you may think in-rush limiting is needed, and if needed then what level is acceptable.  You have made this thread to assist your deliberations - you need to provide the thread with sufficient details to make any advise focussed and relevant and safe.

Perhaps you have set up your 'equipment' and tested it, and tripped out some mains connection device ?

You can do an LTSpice sim in two parts for an NTC - part (a) uses a 1 ohm NTC, part (b) uses a circa 0.05 ohm part.  But that is just the start of appreciating what to do and assess next.

[reyntjensm]

reyntjensm, you have provided no background information as to why you may think in-rush limiting is needed, and if needed then what level is acceptable.  You have made this thread to assist your deliberations - you need to provide the thread with sufficient details to make any advise focussed and relevant and safe.

Perhaps you have set up your 'equipment' and tested it, and tripped out some mains connection device ?

You can do an LTSpice sim in two parts for an NTC - part (a) uses a 1 ohm NTC, part (b) uses a circa 0.05 ohm part.  But that is just the start of appreciating what to do and assess next.

I did a simulation in LTspice as you suggested. But i would like to have a detailed simulation of the NTC. Now i'm not going to bother anymore, a relay it is. I didn't want to say any more about the project since i already asked a few questions about this project on the forum. Some people here liked to link all my questions together and call me a complete idiot with no knowledge. I'm working on a PCB to use in a research project. So it's one of a kind with not that many requirements. I do need to make a good simulation for everything so they can include this in the research project. The device will be protected by a slow fuse. According to the LTspice simulation my inrush current was going to be well over 100A, this would affect my bridge rectifier and maybe blow the safety devices in the grid. I want to limit the inrush current to 20A to be safe. Since i have enough time to charge the capacitors a relay with power resistors looks to be the easiest and most reliable solution.

[Vovk_Z]

Somewhere here was a similar topic (with a similar questions "What to use - NTC or resistor" or "What peak power a powerful wired resistor can withstand"). So, in short, I may say for you: use either NTC or powerful (>50-100W) resistor. With a soft-start (with a relay).
If a load is 16A constant then it seems to me that a fuse (I mean 'a protection') has to be somewhere about 25 A.
A 25 A fuse may withstand transients up to 50-100 Amps peak at least, or more. So there is no need to limit a current to 16 A, but it is much reasonable and efficient to limit current somewhere near 50 Amperes.
So you need 230V/50A = 4-5 Ohm additional resistance (NTC or not NTC).
When we say 'NTC' it doesn't mean the one NTC, but it rather means (in this situation) something like several series-connected low-Ohms (0.5R) large size NTC. For example, 10 x 0.5 R NTC. The same with a powerful 50-100 W resistor. You may use one, or it may be several series-connected ones. 50-100 W rated resistors are more expensive (but you may possibly need only one 4R7 100W).
50-100 W rated wired resistors have datasheets and there you will find an info about how large power or voltage it can withstand.

[bdunham7]

I want to limit the inrush current to 20A to be safe. Since i have enough time to charge the capacitors a relay with power resistors looks to be the easiest and most reliable solution.

Typical use cases don't require limiting the inrush to that low off a level--only 125% of operating current--and there are 5-ohm NTCs with 20A continuous ratings.  A 50 amp or so inrush is usually no big deal.  However, since you specify here that you want 20A and have no problem with a start delay, the relay is the obvious choice.  There isn't an NTC solution that is going to accomplish anywhere near that. 

[reyntjensm]

Somewhere here was a similar topic (with a similar questions "What to use - NTC or resistor" or "What peak power a powerful wired resistor can withstand"). So, in short, I may say for you: use either NTC or powerful (>50-100W) resistor. With a soft-start (with a relay).
If a load is 16A constant then it seems to me that a fuse (I mean 'a protection') has to be somewhere about 25 A.
A 25 A fuse may withstand transients up to 50-100 Amps peak at least, or more. So there is no need to limit a current to 16 A, but it is much reasonable and efficient to limit current somewhere near 50 Amperes.
So you need 230V/50A = 4-5 Ohm additional resistance (NTC or not NTC).
When we say 'NTC' it doesn't mean the one NTC, but it rather means (in this situation) something like several series-connected low-Ohms (0.5R) large size NTC. For example, 10 x 0.5 R NTC. The same with a powerful 50-100 W resistor. You may use one, or it may be several series-connected ones. 50-100 W rated resistors are more expensive (but you may possibly need only one 4R7 100W).
50-100 W rated wired resistors have datasheets and there you will find an info about how large power or voltage it can withstand.

It's a common thing to connect multiple NTC's in series? I thought about this also but i don't know if this is a good option.

[trobbins]

You indicate in-rush is an issue for you due to blowing a main feed fuse, and diode selection.  You really should take time to work through what a fuse, or a cb, or a diode can handle first and relate that to what your mains feed has that could trip you off-line.  Any diode bridge that can handle 20A is likely to handle a peak of well over 200-300A, but you need to know enough electronics to go through that - there are on-line links on how to interpret diode selection and derating.  Similarly there are fuses and CB's that can handle at least 10x rated current for the first pulse, and on-line links on how apps like PSUD2 can help you select a fuse or CB.  LTSpice will show you that the caps charge pretty much in one first half cycle.  You need to go through a learning curve, and ask questions on what details you don't understand.

You don't need to divulge what your load is, but you do need to be able to characterise it in terms of when the load is applied and with what characteristic (eg. resistive) and whether it is switched on-line or somehow ramped, and what the tolerances are.

NTC's can be connected in series - it is simple, but they need to be separated to allow them to operate with thermal independence, and the surroundings need to provide an ambient air temperature for all parts.

[bdunham7]

It's a common thing to connect multiple NTC's in series? I thought about this also but i don't know if this is a good option.

You can, but then you are generating more heat and you have more parts.  This one would be a robust choice in most cases, but won't reliably limit your inrush surge to anywhere near 20A.

https://www.ametherm.com/datasheets/ms355r025

Have you built something or are you just modeling in LT Spice?  Keep in mind that there are things that don't show up in a simple model, such as the resistance and inductance of your power cord and wiring, the impedance of your circuit breaker in your panel, etc, etc.  If you just model the 340Vpk input and 60mR of ESR you mentioned earlier, you get over 5000 amps of surge.  That won't happen in real life.  Your 2000uF of capacitance is actually large enough to certainly justify inrush limiting, but I've seen non-PFC PSUs with 300-500uF that just plug in and switch on.  If you use the back of old envelopes instead of LT Spice, you'll see that the entire charge at 340V is just 0.68 Coulombs and the energy is 115 Joules.  That's less than 1 amp for 1 second and a bit more than 100 watts for 1 second.  Now compare that to the continuous load your circuit has on it.

[akimpowerscr]

If this is just a prototype for your own use, you could replace the relay with a simple 16A switch to short-circuit the resistor.

A well-defined sequence must then be followed: you must switch it "on" before connecting the load.

[Zero999]

The last time I did this, was for some LED drivers, which kept blowing the plug top fuse. I didn't bother with a relay or NTC resistor, an ordinary ceramic resistor was sufficient to limit the inrush current to a low enough level to avoid blowing the fuse, yet the low voltage drop resulted in minimal excess power loss. If power loss is an issue, for minimal fuss, just use an off the shelf timer relay to bypass the current limiting resistor, after a one to three seconds. The item linked below is a bit overkill, but is reasonably priced and will do the job.


https://www.automation24.co.uk/analogue-multifunction-timer-relays-selec-600xu-a-1-cu?previewPriceListId=1&refID=adwords_shopping_UK&gclid=Cj0KCQiAqo3-BRDoARIsAE5vnaKI8HJDOPeff6zTWk756Gq3rTuttv0Sx2oy5efgThje9SafrMpP3ikaAt9WEALw_wcB

[cakyol]

I would like to add an angle here which I do not think has been addressed.
The angle is, regardless of whether an NTC or a PTC is used, what happens if the bypass device (typically a relay), does NOT engage ?

With safety in mind:

- Using an NTC will continue working, but heat buildup will cause problems.
- Using a fixed resistor will also continue working, maybe at much less power and ALSO cause a LOT of heat build up.

The two above are NOT safe.

Using a PTC, for the general case should be the best.

If the bypass device works as usual, there will be no problem.
Otherwise, if it breaks, then the eventual rise in resistance of the PTC will lower the power consumption and will be the SAFEST option in my opinion.

If you do not agree, please comment.

Thanks everyone

[Circlotron]

There are also fusible resistors available that will go open circuit if overheated.

[Vovk_Z]

- Using an NTC will continue working, but heat buildup will cause problems.
- Using a fixed resistor will also continue working, maybe at much less power and ALSO cause a LOT of heat build up.

Using NTC is a common practice. When chosen right they won't cause problems. And NTC are usually designed to withstand larger transients than fixed resistors.