The Ultimate Guide to Soft Start Design
As you have probably noticed, my knowledge base articles are free of advertising. Instead of distracting you with annoying ads, I kindly request your donation. If you find the contents of this page to be useful, please consider making a donation by clicking the Donate button below.
The Ultimate Guide to Soft Start Design
Anyone who has ever turned on the power to a piece of equipment containing a large transformer, such as a welder or a large audio power amplifier, will have noticed the authoritative growl emitted by the transformer and the dimming of the room lights as the power is applied. The growling hum is caused by the electromagnetic forces within the transformer, and subsides as the magnetic field builds within the transformer core. Until the magnetic field has been established, the primary winding will draw significant inrush current – often hundreds of ampere – as seen in the measurement below.
The measurement shows the inrush current of a typical 150 W DIY audio power amplifier. The power supply within the amplifier consists of a 1 kVA toroidal transformer followed by a rectifier and 6 × 15000 µF reservoir capacitors operating at a rail voltage of ±65 V. As seen in the graph, the inrush current reaches 130 A, and the primary current takes nearly half a second to settle to its steady-state value. The force induced by the inrush current will, effectively, attempt to straighten out the primary winding which causes the transformer to vibrate and hum. Similarly, the inrush current causes a significant voltage drop across the mains wiring in the building, thus, the dimming of the room lights.
While this hum may sound authoritative and powerful, the stretching of the primary winding reduces the lifespan of the transformer. The enamel and insulation on the primary winding will wear through over years of use, and the primary will short-circuit to the transformer core. This will blow the mains fuse and require replacement of the power transformer.
In addition to the wear and tear on the transformer, the inrush current also poses a significant challenge for the mains fuse. For smaller transformers, say 200 VA or below, one can often get away with simply oversizing the fuse a little. But for larger transformers this approach quickly becomes impractical and unsafe, as a 15-20 A slow-blow fuse will be required to survive the inrush current. Such a fuse will provide no protection in cases of over-current and very delayed protection against a short circuit. Thus, simply increasing the ampacity of the fuse is not a safe or practical solution in these cases. A far better and safer approach is to limit the inrush current of the power transformer by using a soft start circuit.
Resistor Based Soft Start Circuits
One approach to a soft start circuit is to limit the inrush current by adding a power resistor in series with the transformer primary. A schematic of such a resistor-based soft start circuit is shown below. The resistor is usually a parallel combination of power resistors, and relays are often used for the switches.
When the power is off, both relay contacts are open. The power is applied by closing the relay contacts of RL1. This allows current to flow into the transformer primary through the current-limiting resistor, R. After a short delay RL2 closes to bypass the resistor, thereby, applying the full mains voltage to the transformer. This circuit offers two variables for optimization: The resistance of the current limiting resistor and the delay between the closure of RL1 and that of RL2.
The measurement below shows the inrush current of the same 150 W power amplifier with a resistor-based soft start inserted in series with the power transformer. The inrush current limiting resistor, R, is a 34 Ω power resistor.
As seen in the measurement, the resistor limits the initial inrush current to approximately 3.5 A. However, a significant current spike results when RL2 closes and bypasses the current limiting resistor. This spike reaches over 70 A in magnitude. Although 70 A is lower than the 130 A observed without the inrush limiter there is certainly room for further optimization. To optimize this circuit either R must be decreased or the delay must be increased. This will allow the voltage across the reservoir capacitors in the power supply to develop further before RL2 is closed.
The measurement below shows the same circuit with R = 6.2 Ω. This time the inrush current at turn-on measures 21 A. The second current spike, which occurs as RL2 closes, measures 21 A as well. Thus, 6.2 Ω is the optimum resistor value for the chosen delay of 530 ms.
Drawbacks of the Resistor-Based Soft Start
The observant reader will note that 21 A into 6.2 Ω for half a mains cycle will result in:
dissipated in the resistor during the first mains half-cycle, which is orders of magnitude higher than the power rating of a typical power resistor. Needless to say, dissipating such high power will require the use of many resistors in parallel.
Some manufactures specify the maximum dissipated power for a short-term (typically five seconds) overload. The short-term overload power rating typically falls in the range of 5-10× the rated power for power resistors, thus, a 137 W resistor (or 14 ten-watt resistors in parallel) would be needed in the soft start circuit. The prevalence of DIY soft starts that use significantly fewer resistors would indicate that the resistors are operated significantly beyond their short-term overload specs, thus, potentially quite close to their destruction limit.
Unfortunately, very few manufacturers specify the destruction limit of their resistors. One noteworthy exception is the Vishay PR03-series of 3 W metal film power resistors. As seen in the data sheet figure below, one can expect the PR03-series to fail open if more than 235 W is dissipated in the 3 W resistor for longer than 500 ms. Thus, a minimum of six PR03-series resistors in parallel would be the minimum needed to survive the power dissipated in the resistors during the inrush current limiting.
Needless to say, the power dissipated in the inrush current limiting resistor imposes a severe limitation for this type of circuit. Ultimately, the energy required to establish the magnetic field in the transformer core and to charge the power supply capacitors needs to flow through the inrush current limiting resistor.
Thus, a possibly better approach is to optimize the soft start delay. The maximum delay can be determined once the inrush energy is known. For the 150 W audio power amp in this example, the inrush energy, i.e. the energy required to magnetize the power transformer and charge the power supply capacitors, is 222 J (see below for details). Thus, the longest acceptable delay for given power resistor can be calculated as:
Therefore, if three 10 W power resistors (such as the Vishay CW010-series) are used in parallel, and each resistor can handle up to 10× its power rating in a short-term overload, the delay in the soft start should be no longer than:
The delay should also not exceed the resistor manufacturer’s definition of ‘short-term’ (five seconds in case of the Vishay CV010-series).
Observant readers will note that, in above math, I assumed that all the inrush energy would be dissipated in the inrush current limiting resistors. As seen in the inrush current measurements above, some energy is clearly transferred once RL2 closes, thereby, creating the second spike in primary current.
I suggest that those who wish to implement a resistor-based soft start choose the resistors carefully. Then size the delay such that the power dissipated in the resistors is well below the destruction limit of the resistors. Select the resistance such that the first and second spike of primary current are roughly equal in magnitude. In a fully optimized design, this approach will result in a design margin of about 2×.
Needless to say, a fair amount of experimentation is required to properly design a resistor-based soft start. It begs the question, “is there a better way?”
NTC-Based Soft Start Circuits
A better soft start can be implemented by using a device that has been optimized for inrush current limiting (ICL) applications. These devices are essentially ruggedized ceramic resistors with a very high negative temperature coefficient (NTC). Such inrush limiters will exhibit their specified resistance (cold resistance) at room temperature, but drastically decrease in resistance at higher temperatures to near zero ohm at their maximum operating temperature. Thus, the inrush limiter will heat up and reduce in resistance as the inrush current passes through it. The main advantage of this, is that the inrush limiter will more closely mimic the action of an active current limiter. This allows the magnetic flux in the transformer and the voltage across the reservoir capacitors in the power supply to develop quickly.
Selecting the Inrush Limiter
The inrush limiter should be selected from the maximum allowed inrush current. Common household circuits are generally protected by a circuit breaker with an ampacity of 15-20 A. Thermally actuated circuit breakers will handle substantial short-term overload, thus, aiming for a maximum inrush current near, or slightly above, the ampacity of the circuit breaker is perfectly reasonable.
The actual value of the mains voltage varies. Even in countries with reliable power grids, ±5-10 % variation is common, and larger transient excursions are possible. The worst case inrush current will occur when the mains voltage is higher than its nominal value. Thus, if a 20 A RMS maximum inrush current is to be ensured in a system with 120 V mains, the minimum cold resistance is:
Similarly, in a 240 V system, assuming 10% high mains voltage, the minimum cold resistance can be calculated as:
Note that the maximum inrush current of 20 A was somewhat arbitrarily chosen, which leaves considerable wiggle room in the choice of NTC resistance. Thus, I would consider an NTC inrush limiter with a cold resistance of 10 Ω to be a good choice in a soft start circuit intended for use world wide.
The inrush limiter must also be selected according to the expected load current. Unfortunately, a simple "bigger is better" approach is not helpful here. Inrush current limiters are designed to operate at high temperatures, thus, need to be operated near their maximum operating current. This operating point minimizes the resistance of the inrush current limiter once its inrush current limiting duties are no longer required.
Drawbacks of NTC-Based Soft Start Circuits
When used in audio amplifiers, one major drawback of the NTC-based inrush limiter is that it runs too cold. Consider the scenario of a typical Class AB or Class D power amplifier: The inrush limiter is needed when the amplifier is first turned on, but following the initial power-up, the amplifier does usually not draw enough current to heat up the NTC. Thus, the NTC will, effectively, present a resistance equal to its cold resistance in series with the power supply mains input. This will limit the transient performance of the amplifier by causing significant power supply sag on heavy transients.
Ironically, a significant disadvantage of inrush limiters in other applications, including Class A power amplifiers, is that the inrush limiters tend to run very hot in these applications. It is common for some of these devices to reach 200 ºC when operated at their maximum rated current. If not accounted for, this presents a fire hazard, and these components should be kept away from heat-sensitive components, wiring harnesses, etc. Also, due to their thermal mass, the inrush limiters will not provide effective inrush limiting following a brief power outage. During a short power outage, or brown-out, the inrush limiter will remain close to its nominal operating temperature and, thus, not present any significant resistance to limit the inrush current once the power is reapplied, which will result in a blown fuse.
The solution to these issues, is to bypass the inrush limiter once its services are no longer needed. This can be accomplished with a relay similar to the one used in the resistor-based soft start. The schematic for such a soft start is shown below.
There are two parameters to optimize in an NTC-based soft start: The cold resistance of the inrush limiting deice and the amount of time the NTC is engaged in the circuit (i.e. the delay from RL1 closes until RL2 closes).
The cold resistance of the inrush limiter should be chosen for the maximum allowed inrush current, as described above. The delay between the closure of RL1 and RL2 should then be adjusted such that the peak current following the closure of RL2 remains below the maximum allowed inrush current.
In addition, the inrush limiter must be able to handle the energy required to establish the magnetic field in the power transformer and to charge the power supply capacitors. Determining these will require a bit of math.
So How Much Energy are we Talking Anyway?
The energy required to charge a capacitor is easily calculated:
Thus, the energy required to charge the 6×15000 µF to ±65 V in my test amplifier (3×15000 µF per 65 V rail) is:
The energy required to magnetize the 1 kVA Plitron toroidal power transformer in my test amplifier will need to be determined experimentally.
The inductance of the transformer at startup and, thereby, the energy stored within the transformer core can be determined from a measurement of the maximum inrush current of the transformer (Ametherm, n.d.). It may seem somewhat counter-intuitive, but the highest inrush current will occur if the power to the transformer is turned on right as the mains voltage crosses zero volt. As the transformer current is proportional to the integral of the applied voltage, the inrush current will reach its peak value at the next zero crossing of the mains voltage half a mains cycle later.
I devised the circuit shown below to control the power to the power transformer while measuring the primary current using a current probe.
The MOC3163 TRIAC-output optocoupler features zero crossing detection, thus, will apply power to the transformer under test when the mains voltage crosses zero. I chose 180 Ω gate resistors to ensure the fastest triggering on 120 V mains. For use on 230/240 V mains, use 390 Ω.
The result of this measurement is shown below.
My mains voltage at the time of this measurement measured 121.7 V RMS. Thus, the inductive reactance during the inrush event can be determined as:
Once the reactance is known, the equivalent inrush inductance can be calculated as:
where fmains is the mains frequency. Once the inrush inductance is known, the energy required to magnetize the transformer core can be calculated as:
Thus, the total energy needed to magnetize the transformer and charge the supply capacitors is:
I performed the same measurement for a small handful of toroidal power transformers that I had on hand. The results are tabulated below. As seen in the table, even relatively small transformers store significant energy and draw significant inrush current as result.
|Manufacturer||Manufacturer P/N||Power Rating||Peak Inrush Current||Core Energy|
|Antek||AN-0010||5 VA||1.20 A||0.19 J|
|Antek||AN-0512||50 VA||13.8 A||2.23 J|
|Antek||AS-3222||300 VA||82.5 A||13.3 J|
|Antek||AS-4225||400 VA||110 A||17.8 J|
|Antek||AN-5225||500 VA||135 A||21.8 J|
|RS Components||177-945||530 VA||145 A||23.4 J|
|Plitron||N/A||1 kVA||200 A||32.4 J|
The importance of selecting an inrush limiter that can handle the inrush energy cannot be understated. These devices are known to explode on overload. Thankfully, the manufactures of these devices also include overload specs in the data sheets. The destruction limit is commonly twice the rated energy of the inrush limiter.
Use of Inrush Current Limiters with Switch Mode Power Supplies (SMPS)
Have you ever noticed the arcing that occurs when you plug your laptop or phone charger into a mains outlet? The arcing can be quite severe, in particular on 230/240 V mains. The high inrush current drawn by the capacitors in the power supply causes the arcing and can be limited by the use of NTCs.
To determine the correct energy rating for an inrush current limiter used in an SMPS, the input capacitance of the power supply must be known. As an example, I will use the Connex Electronic SMPS800RE.
The SMPS800RE has two 1200 µF input capacitors. When used on 120 V mains, the capacitors are in series and form a voltage doubler. When used on 240 V mains, the capacitors are simply in series with appropriate ballast resistors in parallel to ensure they share the charge. This is a common arrangement for 120/240 V switchable devices. The capacitors are charged to the peak voltage of the incoming AC waveform (or twice the peak value if used on the 120 V setting). For 10 % high mains voltage, this amounts to:
As the two supply capacitors are identical and connected in series, the total capacitance of the series combination will simply be half of their individual capacitance. Thus, the energy stored in the capacitors can be calculated as:
Thus, only a relatively small inrush current limiter is needed.
Note that many switch mode power supplies include inrush current limiters. Many of them, including the Connex SMPS800RE and the Mean Well SE-600, even bypass the inrush limiters once their services are no longer needed. This is evidenced by the two spikes observed in their inrush current, as seen in the measurement below of the inrush current of two Mean Well SE-600 connected in parallel.
About Those Switches
The final point of optimization of the soft start circuit is the two switches used to apply power and to bypass the inrush limiter. Traditionally relays have been used for this, but they do suffer from one major drawback: arcing. This is especially the case with inductive loads, such as transformers. Arcing can destroy the switch contacts, even in a heavy-duty relay. Have a look at the videos here for some spectacular examples.
Arcing is present in the relay contacts both when the contacts close and when they open, however, the arcing tends to be worse when the contacts open, as the current in the inductive transformer load is interrupted. Thus, I set up the experiment shown in the schematic below to quantify the amount of arcing in the relays I use.
The image below shows the current through the relay contacts as the contacts are opened.
It is interesting to note that it takes well over 2 ms for the relay coil to de-magnetize enough that the relay contacts start to open. However, note the switch bounce that occurs as the contacts open. This is guaranteed to result in arcing, which will reduce the service life of the relay.
Electronic switches, such as TRIACs, SCRs, and MOSFETs cannot arc, thus would be the perfect replacement for a relay. The main issue with solid-state switches is that they will fail over time from conducting current continuously. In addition, solid state switches also require a minimum of voltage across them in order for them to begin conducting. Thus, solid-state switches are not ideal for use as a mains switch, but they are excellent for use in the inrush current limiting path.
Therefore, the perfect switch combination for a soft start circuit would be to use a solid-state switch to engage the soft start (RL1 in above schematics) and to bypass the soft start using a relay (HP 1982a; HP 1982b). This way the solid-state switch handles the making and breaking of the connection to the inductive load, while the relay contacts carry the current for the vast majority of the operating life of the load. In such configuration, the relay contacts always switch very little current as the relay contacts only open and close while the solid-state switch is conducting. This eliminates the arcing in the relay contacts both when the contacts open and when they close.
There are a few different types of solid state switches available. Traditionally, back-to-back SCRs have been used, though these days TRIACs are more common. More recently, thanks to the development of ultra-low on-resistance MOSFETs, power MOSFETs have been used to switch mains voltage. The big question is then which of the three solid-state switch types is optimal.
As the only advantage of back-to-back SCRs over TRIACs is ruggedness, I decided to not test back-to-back SCRs. The solid-state switch only conducts current for a fraction of a second before the relay bypasses the inrush limiter, thus, ruggedness is not needed. Furthermore, a TRIAC-based solid-state switch is simpler and more cost-effective than back-to-back SCRs.
Thus, to determine the optimum switch type, I built two circuits: One based on a TRIAC and one based on MOSFETs. The switch circuits controlled the primary voltage to a power transformer. The secondary of the transformer was loaded by a power resistor. I measured the current through the relay contacts as when the relay was disengaged. Ideally, this current should decrease to zero when the relay opens. If arcing occurs, the switch current will show significant spikes.
The schematic of the TRIAC-based inrush limiter is shown below followed by the measurement of the relay switch current as the relay is opened. I repeated the measurement until the relay contacts opened when the contact current was the highest.
To get a better sense of the time scale, I repeated the measurement with a longer time base on the oscilloscope. You can now clearly see the sinusoidal primary current. This time the relay turned off at the second half of the mains cycle.
As seen in the measurement bypassing the relay with a TRIAC practically eliminates the arcing of the relay contacts. The only drawback of this solution is that the TRIAC will need to be kept on for a short time after the relay is de-energized which will complicate the control logic slightly.
In the subsequent experiment, I swapped the TRIAC for a MOSFET pair driven by a photodiode output optocoupler. The schematic is shown below.
Analog designers may recognize the MOSFET pair as a transmission gate. Two MOSFETs are needed to prevent the AC current from conducting through the body diode of the MOSFET during one half of the mains cycle.
Unfortunately the MOSFET pair offered negligible improvement over a TRIAC-based switch.
Design Example I
I have mentioned my 150 W test amplifier multiple times during this article, thus, it seems reasonable to summarize this article by designing a soft start for use with this amplifier.
- Mains voltage: 120 V ±10 %
- Maximum inrush current: 20 A RMS
- Power transformer: 1 kVA Plitron toroidal
- Power supply capacitance: 3 × 15000 µF per rail
- Power supply voltage: ±65 V
As calculated above, these requirements result in the following requirements for the soft start components:
- Cold resistance of inrush limiter: 6.6 Ω
- Energy handling capability of inrush limiter: 222.4 J
As mentioned above, there is considerable wiggle room in the choice of the cold resistance of the inrush limiter. Ametherm makes a beefy 5 Ω type (P/N: MS32-5R020: 5 Ω, 250 J), which I opted to use. The resulting peak inrush current at high mains voltage is then:
The TRIAC must be selected so it can survive the peak mains voltage and the peak inrush current during the inrush event. TRIACs can generally handle peak currents multiple times larger than continuous currents. The maximum allowed non-repetitive peak current can be found in the data sheet for the TRIAC. Note that "non-repetitive", in this context, means that the TRIAC is allowed to cool to room temperature between inrush events.
The figure below shows the maximum number of current surges allowed for a STMicroelectronics T1035H TRIAC. I have indicated the number of surge pulses allowed for the 37.3 A peak current expected in the inrush limiter.
As noted in the figure, the TRIAC is capable of handling approximately 55 cycles of 37.3 A peak. 55 cycles at 60 Hz mains frequency corresponds to 55/60 = 0.92 seconds. Thus, the delay from the triggering of the TRIAC to the closure of the RL2 contacts should be no longer than 0.92 seconds.
Should a MOSFET switch be desired, the MOSFET chosen must be able to handle the 37.3 A peak inrush current. Thus, the selection process for the MOSFET will include some exploration of the safe area of operation in the data sheets of various MOSFETs.
A strong candidate appears to be the ON Semiconductor FCP099N60E (600 V, 37 A, 99 mΩ). However, one issue with MOSFETs is that their channel resistance increases with temperature. This is a problem when significant power is dissipated in the device. As shown in the figure below, the channel resistance of the FCP099N60E MOSFET is 2.5× higher at a die temperature of 150 ºC than it is at room temperature.
Thus, the voltage across the MOSFET will be significant as it conducts the inrush current. As seen from the equation below, 9.2 V develops across the MOSFET during the inrush event.
Thus, the MOSFET is operated very close to its SOA limit, as seen in the figure below.
Granted, this is a pessimistic estimate, but selecting a MOSFET that is less capable than the FCP099N60E seems ill-advised.
As mentioned previously, the TRIAC and MOSFET-based switches provide very similar performance. In addition, they are similar in circuit complexity. Thus, it seems reasonable to compare the two switch types on the basis of cost as well.
The cost of the TRIAC-based switch is tabulated below.
|Description||Quantity||Manufacturer||Manufacturer P/N||Price Each||Extended Price|
|390 Ω, 250 mW resistor||2||KOA Speer||MF1/4DCT52R3900F||$0.23||$0.46|
|TRIAC Driver||1||ON Semi||MOC3163TVM||$1.90||$1.90|
The following table shows the cost of the MOSFET-based solution.
|Description||Quantity||Manufacturer||Manufacturer P/N||Price Each||Extended Price|
Basically, the MOSFET provides a less rugged solution, that is well over three times the cost of a TRIAC-based solution. As the two solutions perform identically in practice, I chose to implement the TRIAC-based solution.
Thus, I implemented a soft start with above mentioned components using my Intelligent Soft Start as a test platform. I adjusted the soft start delay until the peak inrush current and the current peak, which occurs when RL2 closes, were approximately equal. The resulting inrush current is shown below.
As expected, the inrush current reaches about 34 A peak at the nominal mains voltage of 120 V RMS. The inrush delay is approximately 180 ms, thereby, allowing the amplifier to be ready nearly instantly after power-up, which makes for a positive user experience.
Design Example II
Many readers of these pages will likely be interested in my recommendations for inrush limiting components for typical chipamps powered by my Power-686, hence, this design example.
- Mains voltage: 240 V ±10 % (international mains compatible)
- Maximum inrush current: 20 A RMS
- Power transformer: Antek AS-4225 (2×25 VAC @ 400 VA)
- Power supply capacitance: 2 × 22000 µF per rail
- Power supply voltage: ±35 V
Following the equations above, these requirements result in the following:
- Cold resistance of inrush limiter: 13.2 Ω
- Energy stored in power supply capacitors: 53.9 J
- Energy stored in power transformer: 17.8 J
- Energy handling capability of inrush limiter: 71.7 J
Chosen inrush limiter: Ametherm P/N: SL32-10015 (10 Ω, capable of handling 150 J). Should a smaller component be desired, the Ametherm P/N: SL22-10008 (10 Ω, 90 J capable) will work as well.
Similar to the previous example, the TRIAC (or MOSFET) should be chosen such that they allow a peak current of 37.3 A during the inrush event. Hence, my recommendations for these components remain the same.
Heat Sinking (or not)
Power is only dissipated in the TRIAC or MOSFET for a fraction of a second during the inrush event. Thus, the power dissipation in the semiconductor device is limited by the semiconductor die itself. A heat sink is, therefore, not necessary. In fact, the thermal impedance of the TRIAC or MOSFET package will prevent the dissipated energy from reaching the heat sink until well after the relay has engaged and the power dissipation in the TRIAC or MOSFET has fallen to zero.
Ametherm (n.d.) Transformer Inrush Current Protection. Downloaded from: https://www.ametherm.com/inrush-current/transformer-inrush-current.html
HP (1982a) 14570A Op/Service Manual. Downloaded from: Artek Manuals: HP 14570A Op/Service Manual.
HP (1982b) An Inproved ac Power Switch. HP Journal 12/1982, 34-40. Downloaded from: http://hparchive.com/Journals/HPJ-1982-12.pdf
Did you find this content useful? If so, please consider making a donation by clicking the Donate button below.