|
Sags, swells, and brownouts are slowly-varying changes in voltage, sometimes over periods of hours. They are sometimes called ‘voltage variations’ and some people don’t consider them to have anything to do with EMC. A brownout is another name for a sag, and is a term most often heard in the USA and Canada. Sags and brownouts can go right down to zero volts. Swells are simply slow increases in voltage, as opposed to surges and transients, which are fast increases.
Typical mains supplies are specified at ±6% voltage tolerance (sometimes ±10%) and it is not unusual to experience these ranges daily. Safety tests apply ±15% for long periods of time, so voltages of these levels are clearly not impossible. It is often claimed that European mains supplies don’t suffer from the traditional ‘brownout’ where the voltage falls by a large amount for minutes or even hours – but 60% nominal voltage for an hour or two every weekday is known to occur in parts of Spain – and 50% nominal for 8 hours has been measured by the author in the UK.
A sagging supply voltage is a particular problem for all AC motors and some types of DC motor because they can stall, overheat, and damage their insulation. This can lead to an increased risk of fire, fumes and electric shock, never mind the damage to the equipment. In industrial plant AC motors are usually protected by undervoltage trips which are usually incorporated with other protective functions into motor control circuit breakers devices (MCCBs).
EN 61000-4-11:1994 suggests testing with 40% supply voltage for 1 second and also 0% supply for 1 second with the supply voltage ramped down and then up again over a period of 2 seconds in each case. It does not suggest testing for swells despite the fact that these can damage surge suppression devices.
Surge suppression devices are often designed to operate as close to the maximum expected mains voltage as possible, to protect the equipment better. But their leakage at voltages above nominal can cause overheating and damage if the higher voltages last for more than a few seconds. Maybe EN 61000-4-11 assumes that safety testing would discover such problems.
Most (if not all) harmonised generic and product standards don’t specify any testing at all for sags, brownouts or swells. Testing to EN 61000-4-11 may be a good way to improve product reliability in the field. It may also be a good idea to test for anticipated swells, although EN 61000-4-11 doesn’t recommend any test levels.
Figure 6K shows the voltage versus time profile for a typical sag or brownout test to 50% nominal voltage (in this case). It can easily be done manually by simply using a variable transformer (‘Variac’), although a test laboratory would be more likely to use a programmable mains synthesiser with enough short-term capacity to handle the product’s inrush current.
In real life brownouts can last for minutes at a time (8 hours at 50% voltage was once seen in the UK) so where motors or other vulnerable parts are concerned it may be best to test for a longer time than merely 1 second to ensure reliable operation in the field.
Swell testing can also use a variable transformer and manual control. A test laboratory would be more likely to use a programmable mains synthesiser with enough short-term capacity to handle the product’s inrush current.
Dropouts are short-term 100% reductions in supply voltage (effectively 100% dips). Figure 6L shows a dropout 60ms long (3 mains cycles). Like dips, they are caused by load switching and fault clearance. Also like dips, they can be caused by switching between mains and alternative supply in uninterruptible power supplies or emergency power back-up systems.
EN 61000-4-11 refers to a Unipede study of public mains supplies which recorded 12 dropouts per year with durations lasting between 100 and 500ms, and suggests testing with dropouts lasting for 0.5, 1, 5, 10 supply cycles.
But EN 50082-1:1997 (EN 61000-6-1 from 2004) and EN 61000-6-2:1999 (replaces EN 50082-2:1995 in April 2002) don’t specify any dropout testing at all. It is clearly possible to declare EMC conformity without testing for dropouts, but your products might not be as reliable in the field as you would like. It is easy to make your own dropout tester using the method described for the dips tester but setting the alternative supply to zero.
Interruptions are supply voltage dropouts (= 100% dips) that last for a long time: seconds, minutes, or hours. They are really power outages. As for dips and interruptions, it is easy to make your own test equipment.
EN 61000-4-11 refers to a Unipede study of public mains supplies that showed...
Interruptions lasting: 0.5-1s 1-3s >3s
Number of interruptions/year: 24 5 0
(How did they manage to find no interruptions lasting more than 3 seconds? This doesn’t agree with many people’s experiences of UK mains supplies.)
EN 61000-4-11 suggests testing with interruptions lasting 0.5, 1, 5, 10, 25, and 50 supply cycles. But EN 50082-1:1997 (replaced by EN 61000-6-1 in 2004) and EN 61000-6-2:1999 (replaces EN 50082-2:1995 in April 2002) both specify testing with interruptions of 5 seconds duration only (= 250 cycles).
Equipment which has been designed to pass a 5 second power outage test might fail a 1 second outage test, so testing with all the durations suggested by EN 61000-4-11 might be a good idea to increase product reliability.
Waveform distortion is an increasing problem due to the proliferation of electronic loads on the power networks. 4% is getting to be the typical total harmonic distortion of European mains supplies, and this reach 8% in a few years. Most of the distortion consists of ‘flat-topping’, which means that the mains’ peak voltage can be much lower than Ö2 times its RMS value.
In parts of China the mains can be an actual square wave. As Figure 6M shows, parts of Israel also suffer from significant waveform distortion and this is probably not untypical of a number of countries around the world. So although a meter might read 230Vrms, rectifier-capacitor power supplies could give as little as 75% output before any supply voltage tolerances are considered.
Privately-generated mains supplies can be much worse than the normal mains, although in the ’Internet Hotels’ which now host so many of our web sites the mains supplies can be very much better (they need to be - a typical ‘Cisco hotel’ has an availability specification of 99.9999% and a power consumption of around 10MW).
An oil exploration platform had a 230V mains supply with a tolerance of ±100% caused by the effects of starting and stopping its huge drilling motor on its diesel generator. The voltage variations would last for several seconds each. Emergency 230/400V mains generators can have output quality much worse than a typical public mains supply, and much worse than the suggested test limits in EN 61000-4-11. So if your product is likely to be operated on a private mains supply, it is best to always find out what it is likely to be exposed to!
It is often thought that electro-mechanical components such as relays, contactors and solenoids are immune from EMC considerations. Certainly the designers of so-called hard-wired safety systems seem to believe this untruth. In fact dips, sags, brownouts, dropouts and interruptions can cause energised relays and contactors to ‘drop out’ and solenoids to lose power and move.
Although most designers would presumably consider the effects of a power outage (interruption) they would probably not consider a momentary dip in the supply that might cause some types of relays to drop out while other types remain energised. This situation is made more complicated because the ‘pull-in’ (must operate) and ‘drop out’ (minimum hold-in) voltages of relays and contactors rise significantly as they age.
Sometimes relays, contactors, solenoids, etc. are ‘held in’ at a reduced voltage, to save power consumption, and so might not pull back in again when the supply recovers. So we can see that ‘relay logic’ might not recover to its original state after the dip, dropout, sag, or interruption; and during the event itself they might operate in an unpredicted manner. Few designers of industrial automation seems to design or test for these problems, even though dips and dropouts are considered by experts to be a a major cause of lost production world-wide. (Few also seem aware of the fact that supply surges can arc unpredictably across mechanical contacts, also causing unpredictable operation of ‘relay logic’ systems.)
Figure 6N shows a typical problem for electronic circuits – the logic supply voltage going out of tolerance due to a momentary dip or dropout. The example is of a 40% dip for 40ms (two cycles of the supply), and shows how the unregulated DC rail droops to below the 5V regulator’s drop-out voltage during the dip. The 5V supply to the logic ICs thus falls out of regulation and can fall outside of the tolerance band required for correct logical operation of the ICs.
During this out of tolerance period anything can happen to logic systems and any software they are running - it is a very unpredictable situation and the results can be very damaging to product functionality. Operational malfunctions may be expected and it is not unknown for areas of memory to be wiped or over-written with garbage.
Because most designers only provide the minimum amount of energy storage on the unregulated side, many modern products are vulnerable to this problem. It is often found that products which are quite immune to their mains supply being switched on and off are vulnerable to dips and dropouts lasting just a few tens of milliseconds. It is difficult to manually switch mains faster than 200ms, while surveys (such as Unipede’s) show that shorter dips and dropouts are more commonplace.
Immunity to dips, flicker, and dropouts is usually easiest to achieve by making sure that the unregulated DC rails of a product’s power supply have long enough ‘hold-up’ times. This is usually achieved by using higher values of capacitance on the unregulated DC rail and also by increasing the unregulated voltage so as to allow for the increased DC ripple during the dips and dropouts. We need sufficient energy storage, given the power consumption of the product, to ride through the frequent short dips and dropouts. If non-essential functions can be switched off during a ‘hold up’ event, or microprocessors put into ‘sleep’ mode, it will help increase ‘hold-up’ time.
When dips or dropouts last so long that they become sags, brownouts, or interruptions and the unregulated energy storage cannot ‘hold-up’ any longer, it is important to ensure that before the logic (or other) DC supplies go out of tolerance the product ‘cleanly’ and safely ceases operation and shuts down. Then, when the event has passed, ‘cleanly’ and safely restoring operation (see later for more on the safety issues associated with power restoration). (In this context ‘cleanly’ means that no unintended operations occur and no memory is lost or corrupted.)
Figure 6P shows that because capacitive energy storage (in Joules) = ½CV2, for two identical CV products the energy storage of the higher voltage part is greater. CV product is closely related to physical volume and cost, so the two capacitors in the example will have the same physical size and cost despite the higher voltage one holding twice the energy. Note that to use a much higher unregulated voltage usually requires appropriate circuit design for the regulator so that the higher voltage can be efficiently utilised.
Continuous operation from a battery which is usually being charged up by mains-powered charger, as is normal for portable computers, is an excellent way of overcoming all dips, sags, dropouts; and interruptions lasting from minutes to hours (even months in some cases).
Most mains power supplies use unregulated DC rail voltages that exceed the dropout voltage of their regulators by the minimum amount possible, to improve efficiency and reduce size and cost. Unfortunately, this makes it difficult for the unregulated storage to provide much protection against dips and dropouts. (In the section on emissions of voltage fluctuations it is also shown that such designs make some products more likely to have emissions problems.)
Using unregulated DC rails with an increased voltage makes it easier to achieve significant energy storage to cope with dips and dropouts at reasonable cost, because capacitors store more energy per unit volume at higher voltages, as described earlier.
Operating the unregulated DC storage at a much higher voltage than is needed for circuit operation makes it possible for all the excess voltage to be used for DC ripple caused by dips and dropouts. When linear regulation is used higher unregulated voltages make efficiency worse, so this method is more suited to switch-mode regulators (which will probably need higher voltage ratings for their power devices).
Where an ‘Active PFC’ or similar boost converter precedes the unregulated DC storage capacitor (refer to the section on harmonic emissions) – designing the boost converter for a wide range of input voltages can ensure that the unregulated DC rail is kept fully charged even during quite severe dips and sags. A standard ‘universal input’ power supply (80 - 260V) used on 230V mains will be happy to operate continuously with only 80Vrms input, equivalent to a 65% dip or sag lasting for any length of time.
But ‘universal input’ power supplies won’t protect from dips and sags of more than 30% when used on 110V mains. To protect from larger dips and sags they would need to be re-designed for a 40 - 130V range. Alternatively they could be preceded by a selectable (110/230) voltage-doubler type rectifier circuit. In both these cases the benefit of ‘universal’ operation is lost unless the switch from 115 to 230V operation is done automatically.
Some products with linear power supplies have used triac-tap-switched mains transformers for many years. The triacs are switched automatically between a number of tappings on the primary of the transformer, to maintain the secondary AC voltage (and the unregulated DC voltage) within limits during dips, sags and swells. Hysteresis is used to prevent the triacs from ‘hunting’ between taps when the sensed voltage is close to a switching threshold. The unregulated DC needs to have enough hold-up for the response time of the tap-changer.
No boost converter or tap-changer can cope with a dropout or interruption – but at least they can help a little by making sure the unregulated DC capacitors are fully charged when the dropout begins. This won’t make any difference on formal EMC tests, but it may make a useful difference in real life, where dropouts can follow immediately on from a dip or a sag (brownout).
Designing equipment to cope with swells can just be a matter of rating components to cope with the overvoltage, and maybe with some extra heating too. Alternatively, it is possible to detect a dangerous overvoltage and shut down the equipment to protect it – although this should only be done where the user can cope with the resulting non-availability and no safety hazards are caused.
Surge protection devices (SPDs) are usually dimensioned to start conducting at just above normal mains tolerances. A swell which is higher than these tolerances could easily burn them out, since they aren’t rated for continuous dissipation. Some SPD types fail short-circuit and cause the equipment’s supply fuse to open, taking the equipment out of service. Some types fail open-circuit and can leave the equipment vulnerable to supply surges, with no warning. Overheating SPDs can cause fire and shock hazards if the designer has not considered the fact that they might glow red hot during a supply swell.
A number of ICs are now available for monitoring power supplies and effecting a controlled shut-down and reset when unregulated or regulated DC rails drop too low. These are usually sold as ‘brownout detectors’. Almost no equipment nowadays should rely on a simple resistor-capacitor-gate ‘power-on reset’ circuit, because they have always been inadequate for protecting circuits from the full range of perturbations of the AC supply.
Some power-monitoring devices monitor the unregulated rail so they are able to protect logic circuits before their DC rails goes out of tolerance by inhibiting read/write operations to memories of various types. This prevents memories from being overwritten with garbage during an undervoltage situation. In the case of brief events which don’t require a reset some devices can simply freeze RAM data momentarily enabling operation to continue as if no hiccup in the supply had ever happened.
Various grades of brownout detector are available, depending on the accuracy and tolerance of their voltage detecting functions. The more accurate and the tighter their tolerance, the more costly the tend to be. They are also available with a variety of functions to suit different applications.
Some circuits sample the mains voltage, usually to control heat or other parameters. These can often use a smoothing capacitor to ride out short-term disturbances. But some circuits take timing data from the mains, usually from the zero crossings. These can miscount, misfire, and go wrong in a number of ways – so their design must ensure that all foreseeable timing glitches are prevented or else don’t cause any malfunctions or damage.
The desired behaviour of equipment during supply dips, sags, swells, dropouts and interruptions depends on the application, and may be critical in safety-related applications. Some applications may need a controlled power-down, which may mean providing sufficient energy storage to manage the power down process. But some applications (e.g. medical life support) may not permit a power down at all, making their energy storage requirements very high.
Logic circuits should be prevented from causing malfunction by brownout detectors, but analogue circuits also need consideration. For example an audio amplifier must not produce loud instability, pops, clicks, or thumps during power down (or power up).
For motor drives the requirements depend strongly upon the application. They might be required to stop as quickly as possible, coast, or slow down in synchronism with a number of other motors controlled by other manufacturer’s equipment as a part of a large installation. When power is restored motors might need to ramp their speed up slowly or quickly, or they may not be allowed to rotate until manually commanded.
Since the EMC Directive does not cover safety, where errors and malfunctions in safety-related systems can be caused by the effects of electromagnetic disturbances on electronic circuits, these must be dealt with under safety directives such as the LVD or Machinery Directives.
Uninterruptible power supplies (UPSs) can be used to help products ‘ride through’ dips, sags, dropouts, and interruptions, but are not a universal panacea. Their ability to cope with all AC power EMC disturbances without passing them on to the ‘protected’ equipment should be investigated carefully, as should their reliability performance.
It has been known for the reliability of electronics to be reduced because the UPSs they were run from had lower reliability than the mains supply they were supposed to be improving on. Having said that, properly dimensioned and reliable UPSs which use full-time double-conversion can be a great help in preventing mains power quality problems from disturbing the operation of equipment.
Continuous double-conversion types are preferable as they don’t cause dips and dropouts by switching between their mains and alternative power sources. Many low-cost UPSs don’t use continuous double conversion and may not be able to respond to short dips and dropouts – indeed they may actually cause dips and dropouts by their mode of operation. These may be able to be used if the equipment’s power supply has sufficient hold-up time.
Some models of UPS provide good isolation from conducted RF interference, transients, surges, etc. – but not all of them do – so check the manufacturer’s specification carefully! Also make sure that the UPS will withstand the types of dips, sags, swells, and dropouts that you anticipate for the application. Some UPSs do not themselves have low levels of harmonic emissions into their mains supply, and some might supply distorted output waveforms (including rotary types such as motor-generator sets).
Serious energy storage may be needed if equipment is to keep functioning during a long power outage. Where loads are light, batteries or ‘supercapacitors’ may provide sufficient energy for weeks, but at the other extreme serious power users may need rooms full of batteries just to keep them going for the 30 seconds or so it takes their diesel-powered generators to get up to speed.
One way of reducing energy storage needs and getting increased operational time from your energy storage is to shut down non-essential functions during an outage. Some lighting, heating, chilling, and some visual displays may be able to be treated as non-essential.
Some types of UPSs only support and preserve the mains waveform, rather than take over from the mains. A paper mill in South Africa uses a superconducting energy ring in such a UPS to provide enough storage just for dips and dropouts. At a cost US$10 million it cost less than the web-breaks they were getting every week.
Motor-generator sets with alternative power sources for the motor are the original continuous double-conversion UPS. Motor generator sets are also very useful for preventing harmonic emissions from getting into the public mains supply. If they have enough rotational inertia they can also help reduce the load current fluctuations that would otherwise cause emissions of voltage fluctuations and flicker.
Automatically controlled motorised variable transformers (variacs), ferroresonant constant voltage transformers (CVTs), etc. are all systematic ways of dealing with sags and swells.
EN 61000-3-3 assumes a typical value for mains supply impedance at 50Hz of (0.24 + j0.15) W for the phase conductors and (0.16 + j 0.1)W for the neutral. When testing single-phase equipment these two are combined into one overall source impedance of (0.4 = j 0.25)W, and it is worth noting that j0.25W at 50Hz corresponds to an inductance of 796mH. When testing balanced three-phase equipment which draws negligible neutral current the neutral impedance is neglected and the overall phase-to-phase impedance is (0.48 + j0.3)W.
When equipment draws a fluctuating current from its mains supply these impedances in the supply convert the fluctuating currents into fluctuating mains voltages, as shown by Figure 6Q. These can cause problems for other equipment and also cause lighting ‘flicker’. Lightning flicker is very annoying to most people and can reduce their working efficiency. In some cases flicker can cause actual illnesses (especially those related to stress and/or eyestrain) and it may be able to encourage epileptic fits in susceptible people.
EN 61000-3-3 is now mandatory under the EMC Directive for all equipment that consumes up to 16A/phase from the public 230V supply. Test laboratories use an instrument called a ‘flickermeter’ to measure to this standard. EN 61000-3-3 limits the emissions of voltage fluctuations and flicker and also limits the inrush current at equipment switch-on. The inrush current requirement arises because after a power outage the very heavy inrush currents typical of electronic equipment (as they recharge their DC storage capacitors directly from their bridge rectifiers) can cause an overcurrent trip to occur in the high-voltage distribution, making restarting the network very difficult.
In the case of momentary power outages caused by arc-suppression protection devices on high-voltage overhead power lines when a lightning strike occurs, an overcurrent trip when the power is re-applied can extend what was a two-second outage to one that could last several hours.
Because so much of the load on the low-voltage power network is now electronic apparatus, high-voltage overcurrent tripping due to inrush is now a serious problem and the inrush currents of apparatus need to be controlled. Unfortunately, the issues of EN 61000-3-3 which exist at the time of writing do not mention the reason for its inrush current limit, making it easy for designers and test laboratories to accidentally overlook this requirement.
The requirements of EN 61000-3-3 are very complex because they are based on human perceptions of lighting flicker. The limits for voltage fluctuations and flicker follow a curve of amplitude versus rate (figure 4 in EN 61000-3-3) intended to correspond to the human perception of lighting flicker on a 60W filament light bulb, and are shown in Figure 6R. The peak human sensitivity to flicker is between 5 and 20Hz (300 to 1200 fluctuations per minute) so these rates have the toughest limits. At lower or higher rates greater levels of voltage fluctuation are permitted. For instance: at a rate of once per minute, the permitted voltage fluctuation level is approximately 10 times higher than the limit for 1000 fluctuations per minute.
For voltage fluctuations that occur every 1.25 minutes, the steady-state limits for a step-change are 3% of the nominal voltage, corresponding to a change in resistive load current of 15A every two minutes or so. This allows typical domestic hotplates, heaters, etc., to be switched on and off without suppression as long as they don’t switch more often than once every 1.25 minutes.
The peak limit for a step-change fluctuation every 1.25 minutes is 4%, one-third higher than for steady-state value, equivalent to a resistive load current change of about 20A. The steady-state limit must not be exceeded for more than 200ms.
During an initial switch-on inrush situation, or for load-switching events that occur less than once per hour, the values of the steady-state and peak voltage fluctuations are allowed to be 33% higher than the above. It may be that future versions of EN 61000-3-3 will permit equipment that delays its switch-on inrush current by more than 10ms to have even higher emissions, since delayed inrush places less of a burden on a mains network which is being restarted after a power interruption.
The basic limit curve (figure 4 in EN 61000-3-3) assumes a simple step change in mains voltage (whether an increase or a decrease) – but different voltage fluctuation waveshapes cause different flicker perceptions. Some waveforms will measure lower or higher values for their peak voltage fluctuations than others, even if their instantaneous peak values are the same. A complex mathematical transformation is required to determine whether a non-step type of waveform complies with the limits. This transformation is conducted automatically by the digital signal processing in ‘flickermeters’ (for which the defining standard is EN 61000-4-15).
Section 4.2.3 of EN 61000-3-3 gives some guidelines to designers for estimating the effects of waveshape on peak voltage fluctuation for a few commonly-encountered shapes – but only for fluctuations that occur less than once per second. This makes it easier to estimate the likely voltage fluctuation emissions from an equipment calculation, simulation, or by simple measurements using standard laboratory equipment (oscilloscopes, for example). The accuracy of these estimates is claimed to be no better than ±10%, so results which are within 20% of a limit should be checked with a flickermeter on the actual equipment to make sure it complies.
If you are using a test supply with a total harmonic distortion of under 10% and a supply impedance the same as that specified by EN 61000-3-3 you can measure the voltage fluctuation directly with an oscilloscope. If instead you measure the load current fluctuation with another supply impedance you would need to transform it mathematically into the voltage fluctuation that could be expected using the standard impedance. But beware – the load current fluctuations will themselves depend upon the supply impedance, so if measuring load current it is best to make sure your supply impedance is close (both in resistance and inductance) to the standard supply impedance. Synthesised sources of mains voltage are now available, either with programmed impedances (or zero) achieved by feedback techniques or with the standard impedance. As time goes on more manufacturers are entering this market and the cost of these sources is falling. Figure 6S shows an example of such a source. |
