|Elliott Sound Products||Inrush Current Mitigation|
Inrush current explained very simply is the current drawn by a piece of electrically operated equipment when power is first applied. It can occur with AC or DC powered equipment, and can happen even with low supply voltages. There is now a second 'installment' on this topic - see Soft Start Circuits, and it covers some areas in greater detail than found here. The two share some information, but the second installment has more oscilloscope captures of some of the less obvious approaches. Also, see Project 39, which is a well used and very popular inrush limiter, used by hundreds of ESP customers.
By definition, inrush current is greater than the normal operating current of the equipment, and the ratio can vary from a few percent up to many times the operating current. A circuit that normally draws 1A from the mains may easily draw 50 to 100 times that when power is applied, depending on the supply voltage, wiring and other factors. With AC powered equipment, the highest possible inrush current also depends on the exact time the load is switched on.
In some cases, it's best to apply power when the mains is at its maximum value (peak of RMS = nominal voltage × 1.414), and with others it's far better to apply power as the AC waveform passes through zero volts. Iron core transformers are at their best behaviour when the mains is switched on at the peak of the waveform, while electronic loads (rectifier followed by a filter capacitor for example) prefer to be switched on when the AC waveform is at zero volts.
This is a surprisingly complex topic, and one that is becoming more important than ever before. More and more household and industrial products are using switchmode power supplies, and they range from just a few Watts to many hundreds of Watts. Almost all of these supplies draw a significant over-current when power is applied, and almost no-one gives any useful information in their documentation.
Please note that the descriptions and calculations presented here are for 230V 50Hz mains. This is the nominal value for Australia and Europe, as well as many other countries. The US and Canada, along with a few other countries, use 120V 60Hz. This is not a problem - all formulae can be recalculated using whatever voltage is appropriate. For most examples, the frequency is (more or less) immaterial.
It is not possible to provide a lot of detail for every example, so in many cases a considerable amount of testing or background knowledge may be needed before you will be able to make use of the information here. In addition, component suppliers do not always provide information in the same way, and some info included by one supplier is omitted by others. This can make selection a challenge at times.
|WARNING: This article describes circuitry that is directly connected to the AC mains, and contact with any part of the circuit may result in death or serious injury. By reading past this point, you explicitly accept all responsibility for any such death or injury, and hold Elliott Sound Products harmless against litigation or prosecution even if errors or omissions in this warning or the article itself contribute in any way to death or injury. All mains wiring should be performed by suitably qualified persons, and it may be an offence in your country to perform such wiring unless so qualified. Severe penalties may apply.|
While this is not an article that describes any construction, it does involve measurements that are hazardous, and that may require specialised equipment to ensure safety. If you do not have the required equipment, please do not attempt any of the measurements shown. Never connect oscilloscope probes to the mains, as destruction of the 'scope is probable. Under no circumstances should an oscilloscope be operated without a safety earth/ ground connection via its mains lead.
All current measurements were taken using the Project 139A and/or Project 139 current monitors, which ensure that no direct connection to the mains is needed. Switching at the zero-crossing and peak AC waveform was done using a specialised test unit that I designed and built specifically for assessing inrush current on a variety of products.
Inrush current is also sometimes known as surge current, and as noted above is always higher than the normal operating current of the equipment. The ratio of inrush current to normal full-load current can range from 5 to 100 times greater. A piece of equipment that draws 1A at normal full load may briefly draw between 5 and 100A when power is first applied.
This current surge can cause component damage and/or failure within the equipment itself, blown fuses, tripped circuit breakers, and may severely limit the number of devices connected to a common power source. The following loads will (or may) all have a significant inrush current, albeit for very different reasons ...
The list above covers a great many products, and with modern electronics infiltrating almost every household and industrial item used it actually covers just about every product available. Few modern products are exempt from inrush current - at least to a degree. Some of the most basic items we use do not have an issue with inrush current at all - most are products that use heating coils made from nichrome (nickel-chromium resistance wire) or similar. The current variation between cold and full temperature is generally quite small. This applies to fan assisted, column and most radiant heaters, toasters and electric water heating elements. Apart from these few products, almost everything else will have a significant inrush current.
In some cases, we can ignore the inrush current because it is comparatively small, and/or extremely brief. A few products may draw only double their normal running current for a few mains cycles, while others can draw 10, 50 or 100 times the normal current, but for a very short time (often only a few milliseconds). Some products can draw many times their normal current for an extended period - electric motors with a heavy starting load or power supplies with extremely large capacitor banks being a couple of examples.
Although they are being banned (either by decree or stealth) all over the world, there are still many incandescent lamps in use, and this will not stop any time soon. Most traditional filament (incandescent) lamps are known to fail at the instant of turn-on. This is for two reasons - the filament is cold so has much lower resistance than normal, and the thermal shock can cause a fracture.
When power is applied, there is a high current 'surge', along with thermal shock and rapid expansion of the tungsten. This doesn't affect the lamp initially, but as the filament ages it becomes thinner and more brittle, until one day it just breaks when turned on. For very large lamps used for theatrical lighting (amongst other things), the solution is to preheat the filament - just enough power is applied to keep the filament at a dull red. Full power is almost never applied instantly - it is ramped up so the lamp appears to come up to full brightness very fast, but this is a simple trick that works because the response of our eyes is quite slow.
The cold resistance of a tungsten filament is typically between 1/12 to 1/16 of the resistance when hot. Based on this, it might be expected that the initial inrush current for a cold filament would be 12 to 16 times the current at rated power. The actual initial inrush current is generally limited to some smaller value by external circuit impedance, and is also affected by the position on the AC waveform at which the voltage is applied.
I measured the cold resistance of a 100W reflector lamp at 41 ohms, and at 230V (assuming the power figure is accurate) the resistance will be around 530 ohms - a ratio of 12.9:1 and comfortably within the rule of thumb above.
The time for the initial inrush current to decay to the rated current is determined almost entirely by the thermal mass of the filament, and ranges from about 0.05 seconds in 15W lamps to about 0.4 seconds in 1500W lamps.  This varies with the rated voltage too - a 12V 50W lamp has a much thicker (and therefore more robust) filament than a 230V 50W lamp for example. If incandescent lamps are always either faded up with a dimmer or use some kind of current limiter, they will typically last at least twice as long as those that are just turned on normally.
Traditional (iron-core ballast and starter) fluorescent lamps also draw a higher current during the switch-on cycle. During the startup process, there are filaments at each end of the tube that are heated, and this draws more current than normal operation. Contrary to what you might hear sometimes, this startup current is typically only between 1.25 and 1.5 times the normal current, and it is not better to leave fluorescent lamps on than to switch them off when you leave the room. However, constant switching will reduce the life of the tube, so there is a compromise that depends on the application.
Power factor correction (PFC) capacitors are used in parallel with many fluorescent lamp ballasts, especially those designed for industrial use. These are necessary to minimise the excess current drawn by a passably linear but reactive load. When power is turned on, the inrush current may be very high - typically up to 30 Amps or more depending on the exact point in the main cycle when power is applied! This is many times the operating current of the PFC capacitor (as determined by the capacitance, voltage and frequency).
Many fluorescent tube lights are now using the relatively new T5 tubes, and these are specifically designed to use electronic ballasts. Even the older T8 tubes will give more light output with a high frequency electronic ballast, and we will eventually see the iron-core ballast disappear completely, The electronic versions can be made to be more efficient, but they won't last anywhere near as long. Some of the efficiency gained will be lost again when the ballast (or the entire fitting) has to be replaced because a $0.10 part has failed.
Many other lamps also have (often very) high inrush currents, but these will not be covered here.
This is such an important topic that some explanatory notes are essential. Why is it essential to know, you may well ask. Simply because so few modern loads are resistive, and power factor correction (PFC) is (or will be) used in a vast array of products. Many loads that currently have little or no PFC will be required to perform very much better in the future, and this has already happened with some categories of equipment. Many PFC circuits draw very high inrush current when switched on. If you want a more in-depth explanation of power factor, see Power Factor - The Reality.
Power factor is not well understood by many people, and even some engineers have great difficulty separating the causes of poor power factor. Simply stated, power factor is the ratio of 'real' power (in Watts) to 'apparent' (or imaginary) power (in Volt-Amps or VA). It is commonly believed (but only partially correct under some specific circumstances) that power factor is measured by determining the phase angle between the voltage and current (commonly known as CosΦ (Cosine Phi - the cosine of the phase angle). This is an engineering shorthand method, and does not apply with any load that distorts the current waveform (more on this shortly).
An inductive load such as an unloaded transformer will draw current from the mains, but will consume almost no power (note that a loaded transformer passes the load seen by the secondary back to the mains, and it's usually not inductive). Fluorescent lamps use a 'ballast' - an inductor that is in series with the tube. Similar arrangements are used with other types of discharge lighting as well. For the sake of simplicity, we will use a resistive load of 100 ohms in series with a 1H (1 Henry) inductor. Voltage is 230V at 50Hz, so the reactance of the inductor is 314 ohms. Total steady-state circuit current is shown in Figure 1, for both the inductive and capacitive sections. The inductive current lags the applied voltage by about 72°, and the capacitive current leads the voltage by 90° (voltage is not shown as it would make the graph too difficult to read).
Figure 1 - Test Circuit With Voltage And Current Waveforms
Without the capacitor (C1), the mains current is 698mA (700mA near enough) in this circuit - an apparent power of 161VA. However, the power consumed by the load (R1) is only 48.7W - 698mA through 100 ohms. Therefore, the power factor (PF) is ...
PF = P R / P A Where PR is real power and PA is apparent power
PF = 48.7 / 161 = 0.3
This is considered very poor, because the power company and your wiring must supply the full 700mA, but only a small fraction is being put to good use (about 213mA in fact), and only about 70V of the input voltage is available for the 100 ohm load. The majority of the current is out of phase with the voltage, and performs no work at all. This type of load is very common (all inductive loads in fact), and is easily fixed by cancelling the inductance with a parallel capacitor. Scams that claim that a silly power factor correction capacitor will make "motors run cooler" are obviously false - the inductive current is not changed!
For the above circuit, the capacitance needed is about 9µF and it will draw around 650mA (again at 230V, 50Hz). Because the capacitive and inductive currents are almost exactly 180° out of phase with each other, the reactive parts cancel as shown by the graph. As a result, the generator only needs to supply the 48.7W used by the load, and the supply current falls to 213mA - exactly the value needed to produce 48.7W in a 100 ohm load (ignoring losses). The current we measured in the inductor (698 mA) does not change when the capacitor is added. The difference is that the majority is supplied by the capacitor and not the mains.
One of the greatest problems with the idea of power factor is that many of the claims do not appear to make sense. The above example being a case in point - it seems unlikely that adding a capacitor to draw more current will actually cause it to fall. To understand what is going on requires a good understanding of reactive loads, phase shift and phase cancellation - even though some of it might seem nonsensical, it's all established science and it does work. For example, a leading phase angle implies that the current occurs before the voltage that causes it to flow, and while this might seem impossible, it is what happens in practice. It usually only takes a few cycles to set up the steady state conditions where this occurs.
Inrush Current: The capacitor will normally be discharged when the mains is off, but when power is applied, the cap appears to be close to a dead short at the instant of switch-on. The inrush current is limited only by the mains wiring resistance and the ESR (equivalent series resistance) of the capacitor. Fluorescent lamps also require a starting current to heat the filaments, and this adds to the inrush current.
Where some of the old-timers (and the not-so-old as well) get their knickers in a twist is when the load current is distorted. It has been argued (wrongly) on many a forum that the voltage and current are in phase, so power factor is not an issue. This is completely wrong - those who argue thus have forgotten that the CosΦ method is shorthand, and only applies when both voltage and current are sinewaves. It has also been argued (and again wrongly) that the capacitance following the bridge rectifier creates a leading power factor. It doesn't ! By definition, a reactive load returns the 'unused' current back to the mains supply, but this cannot happen because of the diodes. Non-linear circuits have a poor power factor because the current waveform is distorted, not because of phase shift.
Note in the graph below that there actually is a 'displacement', with the maximum current peak occurring slightly before the voltage peak. However, this is not a leading power factor as many might claim, it's just a small displacement in an otherwise distorted waveform and doesn't mean anything even slightly interesting.
Figure 2 - Non-Linear Test Circuit With Voltage And Current Waveforms
Figure 2 shows the test circuit and waveforms for a non-linear load. These are extremely common now, being used for countless small power supplies, computer supplies, etc. Most power supplies below 500W will use this general scheme. The load won't be a simple resistor, but rather a switchmode power supply used to power the equipment. Note that current flow starts just before the AC waveform peak to 'top-up' the partially discharged filter cap (C1). Input current ceases just after the peak voltage, as the cap is fully charged and discharges much slower than the rate-of-change of the mains voltage.
The power provided by the above circuit is 48W - as close as I could get to the previous example. Input current is 454mA, so apparent power is a little over 104VA. Power factor (calculated the same was as above) is therefore 0.46 - again, not a good result. Most power companies prefer the PF to be 0.8 or better (1 is ideal).
The big problem we have with this circuit is that adding a capacitor does no good at all, nor does adding an inductor. Adding both (called a passive PFC circuit) will improve things a little, essentially by acting as a filter to reduce the current waveform distortion. Passive PFC circuits are physically large and expensive, because they require bulky components. The above circuit can have a considerably improved power factor (perhaps as high as 0.8 without becoming too unwieldy), but the inductance and capacitance needed will still be quite large. In a simulation, I was able to achieve a PF of 0.83 by adding a 1.5µF capacitor and a 100mH inductor, but these are neither cheap nor small. The inductor will also be quite heavy.
Because of the severe waveform distortion (which the power companies hate), many new switching power supplies (especially those over 500W) use active PFC. This requires special circuitry within the supply itself, and if well done can achieve a PF of at least 0.95 - I've seen some that are even better. This is not without penalty though - there is more circuitry and therefore more to go wrong, and the cost is higher. Efficiency is usually slightly lower because of the additional circuitry needed - no circuit is 100% efficient.
It is expected that all switchmode power supplies above perhaps 20W or so will eventually require basic PFC circuitry to achieve at least 0.6 or so without serious current waveform distortion. As most people are well aware, the cost of power is increasing all the time, and anything that increases distribution costs (such as poor power factors) will be passed on to the consumer. The effects of the PFC circuits on inrush current are described further below.
While incandescent lamps have always been a common source of (fairly modest by modern standards) inrush current, up until fairly recently only motors and transformers were the other sources of very high inrush currents. A 500VA transformer is hardly a monster, but is easily capable of an instantaneous current of over 50A if the external circuit will allow it. Even relatively small electric motors can draw very high instantaneous currents, and also draw a higher than normal current during the time taken for them to come up to full speed.
This is a real issue for power transformers used for amplifiers and power supplies, but it is far worse when large distribution and sub-station transformers are involved. At the voltage and power levels involved, simple techniques that are quite effective with small transformers cannot be applied without significant additional cost and complexity. Ultimately it comes down to the design of the transformers, which is decidedly non-trivial for distribution and sub-station units. To minimise losses (which can become very expensive), these transformers must be as efficient as possible, which tends to make the problems worse.
There are several added complications with electric motors that would fill a sensible sized article by themselves, so I will concentrate on transformers. Some of the factors for motors are almost identical, but others are too complex to explain for the purposes of this article. As a result, I will concentrate on transformers, because these are very near and dear to the hearts of DIY people everywhere.
I have described a transformer soft start circuit (see Project 39), and this is specifically designed to limit the inrush current of a large transformer. It is recommended for any tranny of 500VA or more, as these draw a very heavy inrush current. In common with anything that draws much more current at switch-on than during normal running, the maximum inrush is determined by (amongst other things) the point on the AC waveform where power is applied.
When we switch on an appliance, in 99% of cases it's just a simple switch, and there is no control over the point where power is connected. It may connect as the mains waveform passes through zero, it may connect at the very peak of the voltage waveform. Mostly, it will be somewhere between these two extremes, and the first partial (or half) cycle could be positive or negative. AC circuits (including power supplies with full-wave rectifiers before the main circuitry) don't care about the polarity, but they do care about the instantaneous voltage.
Transformers and other inductive circuits behave in a manner that is not intuitive. Should the power be applied at zero volts (the zero crossing point), this is the very worst case. As the voltage increases the core saturates, and peak current is limited by one thing only - circuit resistance. Since a 500VA toroidal transformer will have a typical primary resistance of around 4 ohms (usually less than 2 ohms for 120V countries), the worst case peak current is determined by ...
I P = V Peak / R
I P = 325 / 4 = 81A
External circuit resistance can be added into the formula, but in total it is unlikely to be more than 1 ohm in most cases, so the worst case peak current is still around 65A. Consider that a 500VA transformer at full load will draw a little under 2.2A, so inrush current may be up to 30 times the normal full load current. This is significantly worse than a typical incandescent lamp. Note that the transformer winding can never draw more current than is determined by Ohm's law - it will usually be less, but the formula above is for the worst possible situation. The situation would be different if there were a way to prevent saturation, such as using a core that is many times larger than necessary, but this is clearly not an option due to size and cost.
Figure 3A - Measured Transformer Inrush Current (5A/ Division)
Figure 3A is two captures combined into one, and shows the inrush current waveform captured when power is applied at both the mains zero crossing point and at the peak. The transformer is a single phase, 200VA E-I type, with a primary resistance of 10.5 Ohms. Absolute worst case current is simply the peak value of the mains voltage (325V or 170V), divided by the circuit resistance. This includes the transformer winding, cables, switch resistance, and the effective resistance of the mains feed. The latter is usually less than 1 Ohm, and allowing an extra Ohm for other wiring, this transformer could conceivably draw a peak of about 28A. My inrush tester also has some residual resistance, primarily due to the TRIAC that's used for switching. Although it's bypassed with a relay, there is a time delay before the relay contacts close and this reduces the measured inrush current slightly. Peak switching quite obviously reduces the inrush current dramatically, from a measured 19A down to 4A.
In the above, worst case inrush has been based on the peak value of the AC waveform, and in theory this is correct. However, a more realistic peak inrush current figure is obtained if the RMS voltage is used. When working out something like inrush current, there are many things we don't know that affect the final value, including information about the steel used. Using the RMS voltage will usually give a final value that's closer to measured results. Not especially scientific I know, but for small transformers (up to 1kVA or so) the answer is likely to be closer to reality and not quite as scary.
It is always better to close the switch at the peak of the input AC line voltage. Since the inductor's current is initially zero (as it was before power was applied), switching at the AC peak puts the applied voltage and the inductor's current immediately (very close to) being in quadrature (i.e. at 90° phase displacement) with each other. This minimises the inrush current, as can be seen clearly in Figure 3A. Normally, we don't build mains switches to do this (it's possible, but not simple), so random switching is normal, and is always better than zero-voltage switching that maximises inrush every time the transformer is turned on. Peak switching SSRs (solid state relays) are (or perhaps were) made, but it's unlikely that you'll be able to buy one for a sensible price.
Note that inrush current is unidirectional - all pulses are one polarity until the inrush 'event' has settled and normal operation is attained. This typically takes between 10 and 100 cycles, depending on the transformer. Some very large transformers as used in electrical sub-stations (for example) may take a lot longer to reach normal operation. Although you might expect otherwise, the DC 'event' occurs both with zero-voltage and peak switching.
When the power is connected to a transformer at the very peak of the AC voltage waveform, this is (surprisingly) a much better alternative. Inrush current will usually be quite low, generally less than 1/4 of the worst case value. Without additional relatively complex circuitry, it is not possible to choose when power is applied, so any provision for inrush current must assume the highest possible value - that which is limited only by the winding (and external) resistance.
Note that the following graph shows the capacitive inrush only, and does not include the inrush current caused by the transformer. The reason for this is simple - it is extremely difficult to simulate transformer inrush - as shown in Figures 3A and 3C it is easy to measure though if one has the equipment. The 'ideal' transformer shown doesn't saturate, a real one does! Without a suitable test system it also differs significantly each time power is applied because there is no predictable time within a mains cycle where the power is connected or disconnected. Inrush current may vary from the nominal full load current of the transformer, up to a value limited only by the winding resistance of the primary and external wiring.
This is a complex area, and is not one that is adequately covered for the most part. The basics of inrush current are generally explained well enough, but the effects when a heavy load is present at the same time are mainly covered in passing only, with the transformer and capacitive inrush most often covered separately. In reality, they are almost always present at the same time, which makes everything far more complex. The effects are easy to measure, but are a great deal harder to simulate or prove with a few maths formulae.
Things become far more complicated when the secondary feeds a rectifier, followed by a large bank of filter capacitors. Worst case inrush current is still limited by the winding (and other) resistances, but the capacitor bank appears to be a short circuit at the output of the transformer. Depending on the size of the capacitors, the apparent short circuit may last for some time. During this period, the transformer will be grossly overloaded, but this is of little consequence. Transformers can withstand huge overloads for a short period with no damage, and they will normally last (almost) forever even when subjected to such abuse many times a day.
Figure 3B - Transformer Feeding A Rectifier And Filter Capacitors
The optimum switching point on the mains waveform is at the zero-crossing for a capacitor bank, and this would appear to be in direct conflict with the transformer's requirements for minimum inrush. This can only ever apply if you have a source of ideal transformers, which of course only exist in theory (and simulators). In reality and as seen below in Figure 3C, the transformer inrush is dominant - the 'ideal' point on the AC waveform to apply power is still at the AC mains peak, something you would not expect. Lacking a sensible way to ensure that power is only ever applied at the voltage peak, the use of an inrush mitigation circuit is the only real alternative for transformer-based power supplies. This can be a thermistor (with reservations) or a high power resistor with a bypass circuit. See Project 39 for details of a tried and proven inrush current limiter that is very effective.
Figure 3C - Inrush Waveform, Cap Input Filter (5A/ Division)
Figure 3C is again two oscilloscope captures in one. The yellow trace shows the inrush current (14.5A peak) when the mains is switched at zero, and the blue trace shows the inrush current (8.5A peak) with switching at 90° (peak mains voltage of 325V). The same transformer was used as for the Figure 3A capture, but with a full-wave rectifier (2 diodes), 10,000µF capacitor and a 16 ohm load, with ~38V DC output. It's obvious that peak voltage switching is still preferable, and it shows a much smaller inrush current than zero-voltage switching.
Perhaps unexpectedly, the presence of a load that appears to be close to a short circuit at switch-on actually tames the worst-case inrush current somewhat, and also minimises the unidirectional (DC) effect seen when an unloaded transformer is switched on. Although I don't have a mathematically proven explanation for this, there are two different effects ...
Firstly, the load damps the inductance of the transformer so it no longer behaves like a 'pure' inductance. Consider too that the core is saturated in one direction, so transformer action is impeded. A fully saturated core is not capable of providing magnetic coupling between the windings, so the efficient transfer of energy between primary and secondary can only exist when the core is pulled out of saturation by the AC input voltage. The capacitive load doesn't actually get much charge at all in the first half-cycle.
You can see in the above waveform that in the second half cycle, the current is higher than when the transformer is unloaded. This is because the cap is now charging. The steady state input power of the Figure 3C waveforms measured 120W and the power factor was calculated to be 0.83 - better than expected. Total system losses are about 30W - higher than I expected.
Note that these tests were performed using a 'conventional' E-I lamination transformer. All peak currents will be much higher with an equivalent toroidal, because of reduced winding resistance, better magnetic circuit and the extremely low leakage inductance that is typical of toroidal transformers. However, the general trends seen above will still be apparent.
As you can see, once a capacitor bank is connected to the secondary of a transformer (via a rectifier of course), it doesn't matter a great deal when power is applied. A fairly large inrush current will occur regardless of the exact point on the AC waveform where the switch closes. The previous examples show the possible combinations, and predictably, more capacitance and/or lower winding resistances mean higher peak current. The inrush current settles down quite quickly, and after 100ms it has all but disappeared as you can see from Figure 3C. Much of what remains after 4 cycles is normal load current (about 600mA).
This leads neatly into the next topic ...
A vast number of small appliances now use what is known as an 'off-line' switchmode power supply. This means that the mains voltage is rectified, smoothed (at least to a degree) with an electrolytic capacitor, then the DC is fed to the switching power supply itself. This type of power supply is found in everything from compact fluorescent lamps to DVD players, mobile (cell) phone chargers to TV receivers. They have become truly ubiquitous, and are used to run just about all mains powered appliances that need low voltage DC for operation.
Larger power supplies are also very common, used for PCs, some microwave ovens, high power lighting and numerous other tasks. Many of these now use active power factor correction, which makes them far more friendly to the electrical grid than those with no PFC at all. Many do not use PFC of any kind, and these always present a very unfriendly current waveform to the supply grid.
The majority of these power supplies (both with and without PFC) have high inrush current - often far greater than anything we have used before. Even little compact fluorescent lamps (CFLs) have such a high inrush current that people have been surprised that large numbers of them can't be used on a single switch (or circuit breaker). A typical CFL may be rated at 13W and draw around 95mA (assuming a PF of 0.6). In theory, it should be possible to have over 80 of these lamps on a single 8A lighting circuit, but even with as few as 20, it may be impossible to switch them all on at once without tripping the circuit breaker.
Predictably, the reason is inrush current. Some CFLs and other small power supplies with similar ratings use a series fusible resistor (typically around 10 ohms) in series with the mains, both as a (lame) attempt to limit inrush, and as a safety measure (a fusible resistor will act like a fuse if abused - or so we are led to believe). Even with a relatively small capacitor (22µF is not uncommon), the worst case inrush current may be as high as 30A, and that's allowing for wiring impedance.
Clearly, any power supply that draws up to 315 times the normal running current at switch-on is going to cause problems. Standard circuit breakers are rated for peak (inrush) currents of around 6 to 8 times the running current, so on that basis switching on just 2 CFLs at the same time and at the worst moment could theoretically trip an 8A breaker. This normally never happens, because there is enough wiring impedance (both resistance and reactance) to limit the current to a somewhat saner maximum. The fact does remain though that at least in theory, attempting to switch on just two or three CFLs at the same time could trip a standard 8A breaker.
Figure 4 - Off-Line Rectifier And Filter Capacitor
Figure 4 shows the typical rectifier circuit, along with the waveform. For the sake of being a little more realistic, the switch was closed 0.5ms after the zero-crossing point of the AC waveform, when the voltage has only risen to 51V. As you can see, the peak is still just under 11A, and is over 100 times greater than the RMS operating current. Note that if power is applied at the voltage peak and the capacitor and wiring were perfect (no internal resistance at all) the current would be equal to 32.5A as dictated by Ohm's law (325V peak / 10 ohms).
When a manufacturer has gone to all the trouble of including active power factor correction, you might expect that the inrush current will be minimal because there is no large capacitor following the rectifier (see Figure 5). Unfortunately, the PFC circuitry generally will not start until there is a reasonable charge in the bulk capacitor. This issue is addressed by the diode (D6) as shown, and it conducts fully when power is first applied - this diode is always used, but it forms a dual purpose here. Sometimes there may be another diode in parallel with L1. C1 is the filter cap for the PFC controller, and is coupled via D5 to prevent it from being discharged when the AC waveform falls towards zero volts.
Figure 5 - Simplified Active PFC Circuit 
The switch and inductor form a high frequency switched boost regulator, and the DC output is usually around 400V. The inductor has almost no effect at DC (or 100/120Hz) though, so the bulk (storage) capacitor C2 is charged directly from the mains, via L1 and D1. It is only after the MOSFET (Q1) starts switching at high speed that L1 starts to function normally - at low frequencies (100 or 120Hz) it does nothing at all. It is well beyond the scope of this article to explain switching boost regulators in any further detail, but suffice to say that this is a very common arrangement.
The value of C2 depends on a number of factors, but for even a small power supply of perhaps 150W or so, C2 will be around 150µF. Most manufacturers will use a negative temperature coefficient (NTC) thermistor to limit the inrush current, but it's not at all uncommon for them to get the value horribly wrong. One that I recently came across used a 4 ohm thermistor - completely useless, and I was able to measure 80A inrush peaks easily.
This type of power supply behaves very differently from what we expect with normal linear loads. The operating mains current depends on the voltage, but if voltage increases the current decreases ! This is not expected unless you are used to working with switchmode power supplies (SMPS). As a result, a 100W supply will draw 435mA at 230V, and 830mA at 120V when operating at maximum input power (output power will typically be around 10% less than input power due to circuit inefficiencies).
Some of the recent switchmode supplies I've seen have active inrush limiting - an electronic soft-start built into the power supply. There are several ways this can be done, and some basic ideas are shown below in section 7. If done well, inrush current can be almost completely eliminated, and the mains current gently ramps up to the full load value with no evidence of a current 'surge'.
There is an expectation now that everything should work anywhere in the world without change, so universal power supplies (90 - 260V AC, 50/60Hz input range) are common. It is unfortunate that this may make the circuitry to reduce inrush current far more of a compromise than would otherwise be the case.
While thermistors are cheap and effective for inrush current suppression, they have a number of serious limitations, as discussed in the next section.
One simple choice for reducing inrush current to an acceptable value is to use a resistive component. This needs to present sufficient impedance at switch-on to prevent potentially damaging current surges, but must not waste power needlessly during normal operation. The amount of current drawn during the first few milliseconds should ideally be no more than perhaps double the normal running current, but some switchmode power supplies will refuse to start if the voltage fails to rise above a preset lower limit within a specific time period.
There are all kinds of reasons that may limit the range of choices for the start-up current, but most are limitations (either deliberate or otherwise) within the design of the power supply. Very simple supplies will try to start working as soon as any voltage is present, but may be completely unable to operate even after the limiting resistance is out of circuit if the inrush protection is not designed correctly.
Other more sophisticated designs will use protective circuits that prevent the power supply from operating if the input voltage fails to reach a preset minimum, and/or does not rise quickly enough. In such cases, it may be necessary to accept a higher than desirable inrush current. Things become more complicated when equipment is "universal" - having a power supply range of 90-260V AC at 50 or 60Hz.
An inrush limiter that works perfectly at 230V may prevent the supply from starting at 120V, but if set for 120V operation the inrush current at 230V (or above) becomes excessive. Ideally, this should signal that the power supply itself requires a redesign, but that may not be possible if the PFC integrated circuit used has limitations of its own.
Some of the latest switchmode power supplies use an active inrush limiting scheme, and I have seen several examples where there is no inrush current at all. The input current (relatively) slowly increases from zero up to full operating current, with the input current never exceeding the maximum loaded input current for the power supply. Active inrush limiting has only been seen so far on power supplies that also have active power factor correction, and the additional complexity is necessary to prevent start-up problems. One area where this is becoming common is LED lighting, where many lamps are likely to be wired into a single circuit.
NTC (negative temperature coefficient) thermistors (aka surge limiters) are a common way to reduce inrush. They are readily available from many manufacturers and suppliers, and are well established in this role.
There is a very wide choice of values and power ratings, and a thermistor is just a single component. Nothing else is needed ... at least in theory. Indeed, manufacturers make a point of explaining that their thermistor is the most economical choice, and that additional parts are not required. They may (or more likely may not) point out the many deficiencies of this simple approach.
Thermistors range in value from less than 1 ohm to over 200 ohms and have surge current ratings from around 1A up to 50A or more. It is the designer's job to pick the thermistor that limits the inrush current to an acceptable value, while ensuring that the power supply starts normally and the thermistor resistance falls to a sufficiently low value to minimise losses.
It is useful to look at the abridged specification for what might be considered a fairly typical NTC thermistor suitable for a power supply of around 150-300W depending on supply voltage (From Ametherm Inc. ).
|Max Steady State Current up to 25°C||2 A|
|Max Recommended Energy||10 J|
|Actual Energy Failure||30 J|
|Max Capacitance at 120V AC||700µF|
|Max Capacitance at 240V AC||135µF|
|Resistance at 100% Current||0.34 ohm|
|Resistance at 50% Current||0.6 ohm|
|Body Temperature at Maximum Current||124°C|
It is important to note that the resistance tolerance is very broad - this is common with all thermistors. Expecting close tolerance parts is not an option. The maximum capacitance values shown are for a traditional capacitor input filter following a bridge rectifier. Direct connection to mains is assumed. At rated current, the resistance is 0.34 ohm, so power dissipated is 1.36W which doesn't sound like much, but note the body temperature - 124°C. I would suggest that optimum operation is at 1A, where dissipation is only 0.6W and the temperature will be somewhat lower.
The good part is that the surge energy is specified - in the above case it's 10 Joules. This means that it can withstand 10W for one second, or 100W for 100ms. It can also theoretically handle 1kW for 10ms or 10kW for 1ms, and unless stated otherwise this should not cause failure. Although there is some butt-covering with the maximum capacitance specification, this is largely a guide for the end-user. Based on this I'd suggest that 1kW for 10ms would probably be quite alright, as it's still only 10 Joules. Be warned though - there are probably as many ways of specifying thermistors as there are manufacturers, and not all provide information in a user friendly manner.
While it is a fairly common suggestion (and used by some people), thermistors by themselves are completely useless in any equipment that draws a widely varying current during normal operation. Power amplifiers are a case in point - certainly the transformer and filter caps will cause a high surge current when the amp is switched on, but at low listening levels the thermistor has so little current through it that its resistance will be much higher than it should be. This can lead to power supply voltage modulation, and while that might lower the output transistor dissipation slightly, the thermistor is undergoing consistent stress - heating and cooling constantly whenever the amp is operating.
Thermistors should only be used by themselves where the protected equipment draws a relatively constant power after it has settled down after power is first applied. While very convenient, NTC thermistors have a number of limitations.
They dissipate power constantly while equipment is operating, and normally operate at a relatively high temperature (~125°C for the example shown in the table). This means that they must be kept well clear of temperature sensitive parts (semiconductors, capacitors, etc.). Because they run hot, this means they are dissipating power, and this adds to the heat load inside enclosures and lowers the overall efficiency of the product.
Because thermistors normally run hot for minimum resistance, they must have time to cool down again between the time power is removed then restored. This may not be feasible, because momentary power outages are fairly common worldwide. If the power is off for only a couple of seconds, the thermistor will not have had time to cool, and there is almost no inrush protection when the power is restored. Most NTC makers suggest that a cool-down period of 30 seconds to a couple of minutes is needed, depending on the size of the thermistor, surrounding air temperature, etc.
The use of thermistors is fine, but only if there is a bypass circuit that shorts them out after 150ms or so, and this is my recommendation for any audio equipment.
Thermistor makers like to point out that using an NTC thermistor is so much better than a resistor, because they are physically smaller for the same energy absorption. While this is certainly true, they are fairly wide tolerance devices and unsuited where the application may be subject to strict specifications. The best you can hope for is ±10%, available from some suppliers for some of the range.
Resistors (which will be wire-wound for this application) are a very viable alternative, but they must have some method of bypassing once the surge has passed and the circuit is operational. The alternatives for this are described below.
Resistor selection must be made on the basis of the maximum permissible current, but this is usually an unspecified value. To an extent, experienced engineers can estimate the allowable maximum for reliable operation over an extended period, but this is always a variable and may change if the resistor supplier changes the design.
Some wire wound resistors are capable of astonishing surge currents, while others of apparently equivalent size and value will be destroyed instantly the first time they are used. Nevertheless, resistors remain a commonly used and extremely reliable means of protecting against inrush current. If properly sized and perhaps used in parallel to obtain the power and value needed, there is no reason that an inrush protector using resistors cannot outlast the equipment it protects.
As noted for resistors, a bypass scheme must be used to remove the series resistance from the circuit after the surge current has passed. The humble relay is a popular choice, because they are extremely reliable and are available for almost any application known. The voltage across relay contacts is negligible when they are closed, so contact power loss is close to zero. There is a small current needed for the relay coil though, but for equipment rated at less than 1kW the relay coil should consume no more than about 1W.
Another alternative is a so-called 'solid-state relay' (SSR). These are usually more sensitive than traditional relays (less energising power is needed), but they dissipate some power across the TRIAC or SCR switching component (typically around 1-2W for each amp of continuous current). Cost is usually significantly higher than traditional relays, but they are used in some cases because they are often seen as being more convenient.
It is also possible to make a solid state relay using a TRIAC or SCR directly controlled by a suitable opto-coupler. This is what's inside a 'real' SSR anyway, but by making it from discrete parts gives much greater flexibility. The general bypass schemes used are shown below, but other alternatives are possible.
Figure 6 - Resistor/ Thermistor Bypassing
Many of the main complaints against NTC thermistors are completely eliminated if the thermistor is bypassed shortly after power is applied. The thermistor gets to do its job, and they are fully specified for the instantaneous power dissipation (unlike resistors). Once the circuit is operating normally, the relay shorts out the thermistor, so it is allowed to cool and adds no heat into the enclosure. This means that it is ready immediately after power is removed - no cooling time is needed at all.
It is very important that the relay (or other device) removes the short from the thermistor or resistor very quickly after power is removed. If not, a momentary power outage will cause all equipment to draw a very large surge current when power resumes. The bypass circuit ideally needs to disconnect within a few milliseconds, and certainly well before the power supply 'hold-up' time expires.
Many power supplies are designed to continue functioning and providing output for up to 500ms or so after mains power is removed. This is intended to guard against data loss (for example) during a momentary power outage. General purpose supplies may function properly only over a few missing cycles before the regulated DC voltages start to sag. Hold-up time also depends on the load - a lightly loaded supply will maintain voltage for much longer than one operating at maximum output current.
With a proper bypass arrangement, resistors and thermistors are both equally suitable for circuit protection from inrush current surges. Thermistors have an advantage in that they will fall to a low resistance state even if the bypass system fails to operate, so if there is a fault they will not usually be subjected to massive power dissipation and possibly destroyed.
Resistors do not have this fail-safe advantage, so it may be necessary to add a thermal fuse to protect against fire. Consider a 10 ohm resistor effectively connected directly across the 230V mains. If the bypass relay doesn't work, power dissipation may be as much as 5kW. Current will be close to 23A, so the fuse (if fitted !) should blow, but the resistor may fail first. Higher resistance values are worse - the current is not high enough to cause the fuse to blow straight away, but the resistors will get exceptionally hot and may set the PCB on fire. I generally suggest a soft-start resistance of around 33 ohms in series with the power supply. This is typically in-circuit for about 100ms, after which it is bypassed by a relay (see project 39 for an example).
Electronic power supplies are becoming more common every day, but a great many have extraordinarily high inrush current. In all cases the peak input current at switch-on is created as the main filter capacitor charges. There might be an NTC thermistor or current limiting resistor in the circuit, but neither is particularly useful at maintaining the peak current to a manageable value. This is not generally a problem where the appliance is a one-off, such as an amplifier, DVD player or even a PC, because it's not normal to have a very large number of devices on the same circuit.
With lighting (CFL, fluorescent tube with electronic ballast or LED) it's a very different matter. For example, a 50W ceiling lamp is expected to draw around 220mA at 230V. This assumes a unity power factor, but the actual current may be up to 440mA (power factor of 0.5). It's unlikely that the power factor will be taken into account, so based on the rated power and the common use of a 16A circuit in commercial premises, an electrician could easily be fooled into thinking that you could safely have maybe 50 (or more) fittings on a single circuit (a total of 2,500W, drawing just under 11A). However, unless all the lamps have a very effective inrush limiter and power factor correction, the peak current when turned on will trip the circuit breaker every time someone tries to turn on the lights. Without power factor correction, the total current may be as high as 22A - the breaker will trip due to continuous overload. Where the power supply is rated at more than perhaps 25W, some form of active inrush protection system is essential.
We need to examine the worst-case inrush current, and then figure out how it can be limited to a safe value. 'Safe' in this context means that the circuit breaker won't trip when lights are turned on, only when there is a fault. In general, it should be possible to ensure that inrush current is no more than 4-10 times the nominal operating current, with the inrush duration limited to a single half cycle (10ms at 50Hz, 8.3ms at 60Hz). This keeps the inrush current below the trip threshold for most typical breakers. It will not be possible to load the circuit to its maximum though - the maximum operating load might be as low as half the circuit breaker's current rating.
The risetime of the mains (commonly called dv/dt - delta voltage/ delta time, ΔV/Δt) depends on how the mains is switched. Normal mains switches of all kinds create extremely fast risetimes, but the dv/dt may be tamed somewhat by the building wiring. At (or near) the zero-voltage point, the dv/dt is only about 100mV/µs, but if switched anywhere else during a half-cycle, the dv/dt can easily be several hundred volts/µs.
The magnitude of the impulse depends on the exact time between a zero-crossing and the switching point. Worst case is at 90° after zero-crossing, where the mains is at its peak voltage. At other phase angles, the risetime doesn't change, but the amplitude of the transient is lower.
Figure 7 - Off-Line SMPS Input Test Circuit
Figure 7 is very similar to the circuit shown in Figure 4, but is a new circuit for this specific test. Note that the load shown will normally be a DC/DC converter that powers the circuitry and this applies to all the following diagrams.
The load is 50W, with a 230V supply, and the 10 ohm input resistor is a lumped component that includes the mains impedance, diode forward resistance, capacitor ESR and any current limiting resistance (or thermistor) that may be fitted. 10 ohms is not an unreasonable figure, and even if that were used as a physical component its dissipation would be about 1.7W with the normal distorted current waveform created by the diode bridge and filter capacitor.
If power is applied at the zero-crossing of the AC waveform (zero volts, green trace), the peak current is a passably friendly 8A, compared to the RMS operating current of 415mA for 50W output. Remember that this power supply example does not include power factor correction so current is higher than expected. So, inrush current will be about 20 times the operating current - not wonderful, but it might be acceptable. The magnitude of the inrush current is almost directly proportional to the capacitance, which in turn is determined by the output power. For example, a 50W supply will typically use a 100µF capacitor while a 100W supply will need 220µF (and so on). The value used also depends on the supply voltage, with more capacitance being needed for 120V operation than 230V.
While zero-voltage switching does cause a significant inrush current, things rapidly become serious when the mains happens to be switched at the very peak of the AC waveform (red trace).
Inrush current is now over 30A, just because the switch was closed at the AC waveform peak rather than the zero-crossing. In use, the current will always be somewhere between the two currents measured, depending on the exact moment the switch is closed. 30A is over 72 times the operating current, and as few as 5 loads using this power supply switched on at once will cause intermittent circuit breaker tripping. Should the series resistance be less than the 10 ohms shown, then the peak current will be proportionally greater - up to 100A is not out of the question! Larger capacitance values cause the inrush event to last longer, but do not increase the magnitude of the current because that's limited by the series resistance.
There is a hint in the above as to one method of limiting the inrush current - arrange for some electronic switching to ensure that the power supply is not connected to the mains unless the voltage is close to zero. Zero-voltage switching is still not ideal, but is far better than random.
An easy way to ensure zero voltage switching is to use a 'solid state relay' (aka SSR) . Many of the common SSRs are already designed for zero-crossing switching, and they do not activate unless the voltage across the relay is below around 30 volts or so. Because of this, they are completely unsuited for use with transformers, because transformer inrush current is at its very worst if the power is applied at zero volts. Never use a zero crossing SSR with transformers!
It's relatively simple to incorporate an SSR (either packaged or discrete) into an electronic power supply, and if done properly this will ensure that the inrush current is limited to around 20 times the normal operating current, but this is still a significant inrush event and limits the number of appliances using the power supply on a single circuit. There are other issues when using any form of SSR as a switch for electronic power supplies, which may make this technique more difficult to implement that it might seem at first. The main problem is that SCRs and TRIACs don't conduct at all unless there is enough current, and this can cause continual spike currents to be generated because the switching is so fast. This is similar to the problem seen when CFLs are dimmed using a standard TRIAC dimmer (see CFLs - Dimming for more info and waveforms)  . Provided a TRIAC or SSR is provided with a continuous gate current after it's first triggered there should be no major issues.
Figure 8 - Zero Voltage Switching Using TRIAC
Zero voltage switching is easily accomplished using discrete parts, such as the MOC3043 zero-voltage switching optocoupler and a suitable TRIAC (as shown above). No special circuitry is needed, because the MOC3043 has the sensing and switching circuits built-in. While this technique can (and does) work, it carries a risk for any power supply that only draws current at the peak of the AC waveform. The zero-voltage sense circuit will try to turn the TRIAC on, but nothing will happen because there's no current drawn until the input voltage exceeds the stored voltage in the capacitor.
This means that the circuit might not work properly, and the same applies to a SSR that incorporates zero voltage switching. Ideally, such an arrangement should be bypassed once the power supply inrush event is over. This adds even more complexity, and it's not really very effective anyway. Trying to find useful info on this method isn't easy, because there's not a great deal available on the Net.
As with the following MOSFET circuit, the risetime of the voltage waveform (dv/dt) must not be so fast that it causes the TRIAC or SCR(s) to conduct (static dv/dt). The mains EMI filter needs to be designed to keep the risetime below the critical limit. This can range from as low as 50V/µs up to several hundred volts/µs, depending on the device. As always, it better to err on the safe side, and it's not that difficult to limit the risetime to around 50V/µs. This will probably happen automatically simply due to the distributed capacitance, resistance and inductance of the mains wiring. The TRIAC used must have a static dv/dt rating that's greater than the actual dv/dt so it doesn't self-trigger. A resistor should always be used between the gate and T1 (aka MT1) to maximise the static dv/dt performance (this resistor may be included in some TRIAC packages).
One thing that is very obvious is that the exercise is not trivial. While it might be imagined that you could just give up and use an NTC thermistor, it should be very clear by now that this is rarely a workable solution in real life. Apart from anything else, there will always be a significant amount of excess heat within the enclosure that must be disposed of, and this alone can be a daunting prospect for a compact power supply.
There are several schemes to use MOSFETs as the current limiter. These can be used in linear or switched modes, and there are quite a few variations on the theme. Linear mode is the easiest to implement, but the MOSFET has very high dissipation for the first few half-cycles. Switching mode causes much lower dissipation in the soft-start MOSFET, but requires more circuitry. As power supply design becomes more sophisticated with dedicated ICs, the added complexity isn't as great as it might have been just a few years ago, but it's still not as straightforward as we might hope.
The greatest advantage of using a MOSFET is that the start-up inrush current can be made to be no greater than the normal operating current, so there is effectively no inrush current at all. The current waveform simply increases smoothly over a few cycles then settles at the running current with no high current peaks at all. Look at Figure 11 as an example.
Figure 9 - Linear MOSFET Inrush Limiter
The biggest problem with the linear scheme is that peak power dissipation in the MOSFET can easily reach several hundred Watts. While it's not difficult to get rid of the heat (it only lasts for about 200 milliseconds or less), the stress on the MOSFET may be high, which may lead to premature failure. However, it is still fairly easy to ensure that the MOSFET remains within its safe operating area (SOA), and it is by far the easiest scheme to implement. The arrangement shown above reaches a peak dissipation of about 120W, and the average over the 170ms turn-on period is under 30W. This is not at all stressful, and as seen below, inrush current is all but eliminated.
The very narrow spike just after switch-on is caused by the EMI filter's input capacitor. While the peak current can be rather high (8-10A is not uncommon), it only lasts for a few microseconds. The same thing is visible in the oscilloscope capture shown in Figure 12. X-Class caps are supposedly capable of withstanding this surge easily, but I have seen some that have degraded in use and show less capacitance than the marked value (allowing for tolerance). It's not known if the degradation was due to switch-on current surges or a 'dirty' mains supply, as the affected units were from an industrial complex.
Figure 10 - Linear MOSFET Inrush Limiter Waveform
There is one point that is extremely important, but is also likely to be unexpected. When the power is applied via a switch, the dv/dt (rate-of-change of voltage vs. time, aka Δv/Δt) is extremely high. The input filter and MOSFET drive circuit must be configured so that the drain-to-gate capacitance of the MOSFET doesn't cause spontaneous conduction. This generally means that at least two mechanisms must be in place so the MOSFET is never forced into unexpected conduction because of the extremely fast voltage rise when the switch is closed.
The first line of defence is to limit the maximum risetime of the applied voltage, and the second is to ensure the gate has a low impedance path to the source (via a large capacitor for example) so the instantaneous current that flows in the drain-gate capacitance cannot raise the gate above the conduction threshold. Parasitic inductance must be kept to an absolute minimum, and the capacitor must be located as close to the gate and source pins as possible. While it's probably not well known that MOSFETs will switch on due to high dv/dt between the drain and source, it's very real - it can happen even when the gate is connected to the source via any impedance! .
Any capacitively coupled energy is absorbed by C3 in Figure 9, which is very large compared to the drain-gate capacitance. It will easily absorb any current spike without the voltage changing appreciably. Local inductance between the gate and source must be kept very low indeed, or problems may still occur. This means very short tracks on the PCB from the MOSFET to the capacitor.
The circuitry needed for a proper switched MOSFET inrush limiter is relatively complex, but easily within the capability of a fairly straightforward IC. One may already exist, but if so the details are not available (at least nothing that I could find). The requirement for avoiding spontaneous conduction due to high dv/dt is just as important with a high-speed switching limiter as with a linear version. The mains current waveform during the inrush period is similar to that shown above. The PWM controller starts off with narrow pulses and increases the pulse width over a period of perhaps 100ms, after which it applies a continuous gate current for Q1. Once the inrush period has elapsed, Q1 remains fully on, so losses are minimal.
Figure 11 - Switched MOSFET Inrush Limiter
Figure 12 shows the measured inrush current for a LED lighting power supply fitted with active current limiting. As you can see, it is very effective, and inrush current is virtually non-existent. The very short spike is the point where power was applied, and is simply the EMI filtering capacitor (typically 100nF, X2 Class) charging. This spike is high current (~8A) but is so short that it will never cause a problem. Power was applied at the peak of the input waveform (~325V for nominal 230V mains), using a purpose-built tester that I designed and built. It allows me to select zero-crossing or 90° (peak) switching. The start-up waveform doesn't change, but the sharp spike disappears when the mains is switched at the zero-crossing.
Figure 12 - Active Inrush Limiting In Action
The above image is a direct capture from a digital oscilloscope, and has not been altered other than resizing and cropping. As you can see, the technique is 100% effective. The inrush current is suppressed so effectively that the worst-case peak current is only fractionally more than the running current. This is an important part of electronic power supply design that it can be expected to become standard for any application where large numbers of supplies can be turned on simultaneously.
We have discovered above that the dv/dt of the switched mains can cause problems both for MOSFETs and TRIACs (spontaneous conduction), and that even the EMI filter can create a large current spike. If designers were to use a zero-crossing SSR or TRIAC, coupled with a MOSFET based 'true' soft-start circuit, it becomes fairly easy to ensure that there can never be a current impulse at the moment of switch-on. It would no longer make any difference if the mains were switched on at the peak or anywhere else on the waveform, because a zero-crossing SSR always applies current only when the voltage is close to zero, and the MOSFET ramps up the current in a controlled and entirely predictable manner.
This approach would provide the best possible result, with no components being subjected to high impulse current. It's even possible to use this arrangement with transformers, because although zero voltage causes the worst case inrush current, the active MOSFET circuit can provide a smooth voltage increase so saturation effects can be eliminated completely. However, to do so means that we must include the MOSFET soft-starter circuit within a bridge rectifier so that it works with AC. This is a design exercise in itself, and still cannot address all issues.
Figure 13 - Combination Active Inrush Limiter
This idea isn't as 'over-the-top' as it might seem at first. There will never be a high dv/dt waveform applied to the MOSFET circuit, and this simplifies the design. For a manufacturer there is only a small additional cost, and it can be based on a dedicated controller IC that does all the hard work. Even the simple act of switching on a large bank of lights (for example) places stresses on the switch contacts. If the switching is done electronically using a zero-volt switching relay as shown above, small-gauge wiring can be used from a master controller to the switch and the switch only handles low current.
If incorporated into individual power supplies, the TRIAC and MOSFET can be comparatively low power, and inrush current is kept to an absolute minimum. Somehow I doubt that extremely cost-sensitive manufacturing would ever consider such an approach though, because it would inevitably add cost to the power supply. Unfortunately, there are suppliers, distributors and manufacturers who don't even know that inrush current is a problem. They certainly won't change anything to fix an issue they either don't know exists, or choose to ignore in the hope that no-one notices.
For a great many small appliances, inrush protection should be mandatory, simply because overall power consumption is falling, so people think (not unreasonably, I might add) they can use more appliances on the same circuit than was previously possible. As this article has shown, you may actually end up with far fewer than you might imagine. As explained here, getting a good inrush suppression system is actually quite difficult, and involves aspects of electronic devices that people generally don't think of because they are so obscure - especially dv/dt.
Limited ability to connect multiple devices at once is especially troublesome with lighting - the inrush current is much higher for most modern 'equivalents' to traditional incandescent lamps or fluorescent tubes, regardless of whether the replacements use CFL or LED technology. Even the iron cored transformer of old for halogen lighting has given way to an electronic equivalent, which may be more efficient, but can never last as long. Indeed, a great many of the small efficiency gains that are mandated upon us by government decree can vanish in an instant if the 'new, improved' replacement device fails prematurely. Yes, I know this is a separate topic, but it's important here too. For what it's worth, the inrush current for 'electronic transformers' is generally fairly modest, and is almost entirely due to the lamp itself.
Until manufacturers strive to minimise surge (inrush) current, installers will continue to have issues - especially if there is no inrush information provided in any of the documentation. It's actually very uncommon for any manufacturer to provide this info, even though it can (and does) cause some fairly major headaches when an installer gets caught out. The problem affects everyone - the manufacturer and/or distributor gets a bad name, installers have to change their wiring, and customers are inconvenienced.
At the very least, details of inrush current should be provided with documentation. Installers need to know how many 'things' can be connected to a single circuit breaker, even if they are grouped into individually switched banks. This applies especially for lighting equipment, because the lower power demands of many modern lights can easily mislead people into thinking that potentially very large numbers can be used on a single circuit breaker. While there might not be a problem if lights are switched on in some kind of sequence, restoration of service after a power outage will cause all lights to try to come on at once.
The circuit breaker may trip every time someone tries to reset it, until some of the individual banks are turned off with their individual switches. This is untenable in the workplace - unless there are regular power failures so everyone learns and knows the proper sequence, it could easily take some time before anyone figures out what needs to be done. Momentary interruptions will simply trip the circuit breaker when power resumes, and this is quite unacceptable. Doubly so because someone will try to reset the breaker over and over again, until by chance it manages to stay on. I know of two installations where this has occurred, exactly as described. The only fix was to rewire some of the lights onto an additional circuit breaker, and to change the circuit breakers to delayed action (D-Curve) types.
All manufacturers of appliances that use switchmode power supplies need to provide useful information to installers or users, so that people know that these new products may behave differently from what might be expected. Even traditionally resistive loads like electric stoves are sometimes using a SMPS to power induction cooking systems, so there are very few things that are not affected. Even electric hot water systems are now available using a heat-pump (an air conditioner in reverse), and this will also have a significant inrush current.
Imagine the load on the electricity grid if there is a short service interruption, and tens of thousands of high inrush current appliances all try to come back on simultaneously. As we've seen above, a 50:1 ratio is not uncommon, so if a fully loaded electrical substation suffers a momentary break in supply, how can it possibly cope with a 50 times overload when it tries to come back on-line?
The answer, of course, is that it probably can't, unless the total load is significantly less than the rated capacity of the substation. Likewise, possibly hundreds of switchboard circuit breakers that are close to their limit will drop out. All this because no-one will tell installers and users that these items draw 20, 40 or 50 times the normal current when power is applied. In fact, it may only be because hundreds of switchboard breakers trip that the substation will be able to be reconnected, because the peak load is reduced. I've not heard of this happening (yet), but it's inevitable as electronic loads become the standard and with higher power ratings.
I expect that most manufacturers will eventually get it right, but based on what I've seen so far it's likely to take a while. It's not hard to imagine how this problem has come about - I'm fairly sure that they simply haven't thought about the likely consequences, and any inrush current limiting is simply to protect the product itself. There appears to be little thought for the installation or the grid with many products. One only needs to remember where most products are manufactured now, and that the supplier is often selected on price alone. Unfortunately, it should come as no surprise that problems exist .
It is worthwhile mentioning that many of the latest (as of 2017, with some earlier) LED power supplies do incorporate an active soft-start function, and I've now tested quite a few that simply ramp up their input current over a number of mains cycles, with the current finally settling at the normal operating level. This is fairly recent, with most of the older lighting products failing rather dismally. This is obviously an issue that caught out a lot of installers (I know of several from immediate contacts in the industry), but the problems that used to be common are now a thing of the past. LED lighting has come a long way in a short time, and continued improvements are to be expected for some time to come.
|Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott (Elliott Sound Products), and is Copyright © 2010 - all rights reserved. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws. The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference. Commercial use is prohibited without express written authorisation from Rod Elliott.|