Chapter 13
EMI and EMC Issues
Abstract
This chapter discusses electromagnetic interference and electromagnetic compatibility. Switch regulators produce radio frequency signals because they switch at high frequency. Ringing and harmonics can extend to several hundred megahertz. Methods of reducing signals at source are discussed, with the side effects that this suppression causes. Filtering and screening are also discussed.
Keywords
electromagnetic interference
electromagnetic compatibility
screening
filtering
harmonic frequencies
ringing
transient
The first two questions regarding electromagnetic interference (EMI) and electromagnetic compatibility (EMC) are: what is the difference between EMI and EMC? And which standards apply? Subsequent questions relate to how equipment can be made to meet the standards. Of course, meeting the standards often costs money (filter components, screening, and suppressors) so the aim is to just meet the standards with a small safety margin.
EMI is electromagnetic interference. This is the amount of radiation emitted by some equipment when it is operating. EMI is caused by emissions in the radio spectrum, which not only interfere with radio systems but also can cause other equipment to malfunction. One example is interference from portable radio transmitters like CB radios and cell phones; when used near a gasoline station, the pump can indicate the wrong amount being delivered. An often seen warning notice at a gasoline station says “using a radio transmitter can cause a fire,” but in reality the most likely effect is to cause an error in the fuel measurement.
So what is EMC? This is electromagnetic compatibility, and is a measure of how good a system is at rejecting interference from others. Medical systems have a high immunity requirement because the consequences of a failure are death or injury. Any system connected to the AC mains power line must be immune to transient surges; the degree of immunity depends on the application. Power meters connected to lines where they enter a building are subject to the highest potential surges, so they have very high immunity requirements. Internal lighting and domestic appliances have very much lower immunity requirements.
Before we look at EMI and EMC standards, and the design techniques used to meet them, it is important that we understand signals. Fourier analysis shows that any signal that is not a pure sine wave can be considered as a fundamental signal plus higher frequency harmonics, which are a multiple of the fundamental frequency. For example, a square wave with a 50/50 duty cycle has a fundamental signal at the switching frequency plus a third harmonic of 1/3 amplitude, plus a fifth harmonic at 1/5 amplitude, plus a seventh harmonic at 1/7 amplitude, etc. If the signal is not 50/50 duty cycle, or if the switching edges have some slope (as all practical signals do), then there will be both odd and even harmonics present and the amplitude of harmonics will be less predictable. Typically, this is like the signal across a MOSFET switch in an LED driver circuit.
13.1. EMI Standards
13.1.1. AC Mains–Connected LED Drivers
Any LED driver connected to AC mains supply has to meet the limit specified in harmonic current emissions standard IEC/EN 61000-3-2. Within this standard there are several classes and the one related to lighting is Class C. The harmonic emission limits specified in IEC/EN 61000-3-2, Ed. 2: 2000, up to the 40th harmonic, are listed in Table 13.1.
Table 13.1
Harmonic Limits for EN 61000-6-3 Class C
| Harmonic Order “N” | Maximum Current, Class C (% of Fundamental Current) |
| 2 | 2 |
| 3 | 30 × power factor |
| 4–40 (even) | Not specified |
| 5 | 10 |
| 7 | 7 |
| 9 | 5 |
| 11–39 (odd) | 3 |
Conducted emission limits in the 150 kHz–30 MHz frequency range are specified in the standard IEC/EN 61000-6-3.
13.1.2. General Requirements for All Equipment
All LED drivers have to meet the radiated emissions standards. The standard is IEC/EN 61000-6-3, which covers the frequency range 30 MHz–1 GHz. This standard uses limits previously set by CISPR22 in the USA and by the European Norm EN55022. The limits given in CISPR22 and EN55022 standards were intended for computers and communications-related equipment, but these have been adopted as generic limits for all electronic products, including LED lighting.
The emission levels to meet EN55022/CISPR22 Class B are 30 dBμV/m in the frequency range 30–200 MHz. From 200 MHz to 1 GHz the emission level increases to 37 dBμV/m. These are the signal levels measured at a range of 10 m from the equipment under test (EUT). Since the signal power is proportional to 1/R2; for example, at 1 m from the EUT the emission limit will be 20 dB higher (100 times the power), at 50 and 57 dBμV/m, respectively.
13.2. Good EMI Design Techniques
It is important to look at the circuit diagram and determine where the possible sources of EMI are located. This should happen before the printed circuit board (PCB) is designed. The center point for EMI sources must be the MOSFET switch. This turns on very quickly and so has sharp edges with high frequency content. When looking at the circuit schematic, consider the effect of high frequencies (1–200 MHz).
At very high frequencies, components do not behave the same as they would at low frequencies. An inductor that was thought to block AC signals suddenly behaves like a capacitor that passes AC signals very easily. Similarly, a capacitor thought to have a low impedance characteristic behaves like an inductor at very high frequency; a good example of this is an electrolytic capacitor. So, check the component datasheets and look at the frequency response curves showing impedance versus frequency; see where the resonant frequency is—you will be surprised!
13.2.1. Buck Circuit Example
Let us take a look at a simple buck circuit, to see where the EMI can arise. Fig. 13.1 shows a typical buck circuit. As a reminder, the integrated circuit at the heart of this buck circuit is a PWM controller. Internally, a clock signal triggers a latch, causing the gate drive output to be activated; the voltage on the Gate pin rises to 7.5 V in this case. The MOSFET Q1 turns on and the current increases at a fairly constant rate, due to the inductance of L1. When the voltage on the CS pin is raised above 250 mV, due to current in R2, the internal latch is reset and the gate drive output is disabled. The MOSFET Q1 turns off but current continues to flow in the LED and the flywheel diode D1 due to the energy stored in the inductor L1. When used in a buck circuit, this IC maintains an almost constant current in the LED.
Figure 13.1 Buck Circuit.
For EMI analysis, we should consider what happens when the MOSFET Q1 turns on. The drain voltage of Q1 will fall to a very low voltage relative to ground, just a small voltage due to the current flowing in the drain–source channel of Q1 and in the current sense resistor R2. When Q1 turns off, the flywheel diode D1 conducts, so the drain voltage of Q1 rises very fast and is clamped to the positive supply rail. Thus, when switching, the MOSFET drain voltage will have a rectangular waveform that is almost equal in amplitude to the supply voltage. The fast rising and falling edges create EMI with a broad spectrum of harmonics.
Current flows are shown in Fig. 13.2. Analysis shows that the gate current flows from ground, through VDD supply capacitor C4, through the IC and out of the gate drive pin, through the gate and current sense resistor and back to ground. Analysis of the LED current shows a path from ground, through the decoupling capacitors C1 and C2, through the LED and inductor, through Q1 and the current sense resistor R2, and back to ground. Both currents have fast rising and falling edges, which can produce high frequency EMI.
Figure 13.2 Buck Circuit Current Flows.
What is not shown in Fig. 13.2 is the current that flows through the flywheel diode D1. There is a forward current through D1 when Q1 is off, due to the energy stored in the inductor, which keeps the LED current flowing. There is also a momentary reverse current that flows in D1 when Q1 first turns on. This reverse current flows for a short time (typically 75 ns or less) and creates a current spike through the current sense resistor, R2. A small part of this current is due to the junction capacitance, but the main part is reverse recovery current.
Reverse recovery current occurs when a diode junction that is conducting in the forward direction is subject to a sudden reverse polarity. When a reverse polarity is applied, the free electrons in the junction area take some time to be swept away. Only after the free electrons have been removed is a depletion region created inside the silicon, which blocks further electron flow. The time taken to remove the free electrons in the reverse recovery time is Trr.
The choice of capacitors in the circuit is important. Capacitor C2 must have low impedance at high frequency, for handling the high frequency current in the power switching circuit. The capacitor dielectric could be ceramic for low voltage supplies, or metalized plastic film, such as polyester in high voltage circuits.
The capacitor C3 mounted across the LED terminals provides a bypass path to carry high frequency signals, which are created when the MOSFET Q1 turns on charges the capacitance of the inductor windings. The inductors’ winding capacitance is simply due to insulated wires being wound over each other in a coil. Some inductors have more self-capacitance than others, due to differences in construction. It must be of low impedance and be able to carry high frequency signals. A typical value is 100 nF. The voltage rating of capacitor C3 should be high enough to withstand the supply voltage because there is a chance that the LED could be disconnected or become open circuit.
The VDD capacitor C4 should be a ceramic dielectric type, value typically 2.2 μF. This can be a low voltage type; a 16 V rating is commonly used. Remember that the actual effective capacitance with a 7.5 V VDD supply will be lower than the nominal capacitance value, due to the bias voltage effect described in Chapter 11.
We briefly mentioned the inductor L1. We discussed the interwinding capacitance, which affects performance by causing current spikes when Q1 turns on. But the magnetic field must be considered too; a shielded inductor or a toroidal construction should be used to minimize radiating magnetic fields. A rod inductor makes a good antenna!
When considering input filters, we need to raise the impedance of the current path from the buck circuit into the power source by adding an inductor L2; this is shown in Fig. 13.3. A small capacitor C5 on the power source side of L2 shunts any small signals that manage to pass from the switching circuit and through L2. Basically adding L2 and C5 creates a low-pass filter to attenuate (reduce) and high frequency signals from the MOSFET Q1. Voltage spikes and noise on the supply rail are prevented from reaching the buck circuit by the attenuation due to L2, C1, and C2.
Figure 13.3 Buck Circuit With Filter.
If a filter has been added, but emissions are still too high, consider placing a resistor in series with the MOSFET gate. A value in the range 10–100 Ω is likely to be sufficient. The resistor slows down the MOSFET gate-charging rate when switching on and switching off, so the high voltage switching is slowed; the waveform now has sloped edges with fewer high frequency harmonics. However, slowing down the switching may also introduce more losses.
Using an oscilloscope to look at the voltage waveform across the current sense resistor in a high voltage buck circuit is revealing, see Fig. 13.4. First there is the turn-on spike, “A,” due to inductor parasitic capacitance and the flywheel diode reverse recovery period, but then there is immediately some oscillation afterward which is usually in the 100–300 MHz frequency range. After the oscillation, the current rises linearly, “B,” due to the inductor. When the current sense threshold is reached, the MOSFET turns off, giving a sharp drop in the voltage across the current sense resistor. However, the MOSFET turns on again briefly, giving a final pulse of current, “C,” through the sense resistor.
Figure 13.4 Current Sense Waveform.
Some analysis of the circuit shows that the high frequency oscillation, after the initial turn-on pulse, is due to parasitic components resonating. Fig. 13.5 shows a circuit with the parasitic components added. Parasitic inductors are created by PCB tracks. A parasitic capacitor exists at the drain–gate junction of the MOSFET. The circuit is a series resonant LC circuit, which can be damped by a series gate resistor, Rg.
Figure 13.5 Gate Circuit.
The short pulse of current, shown as “C” in Fig. 13.4, which occurs after the controller circuit has turned off the MOSFET, is caused by the MOSFET turning back on. Parasitic capacitance from drain to gate passes sufficient current when the MOSFET drain suddenly switches from almost 0 V up to a high voltage (may be 400 V). This current recharges the gate above the threshold voltage, allowing the MOSFET to turn on (at least partially).
The gate recharging effect is more prevalent in high voltage applications, where the MOSFET driver may not present a very low impedance at the gate. This effect will be made worst by a resistor in series with the gate because the impedance will be even higher. One solution is to add diode in parallel with the gate resistor, but this is only effective if the gate drive circuit has very low impedance and can clamp the gate to ground. An alternative solution is to add a PNP transistor across the gate resistor, with the emitter connected to the gate, the base connected to the gate driver, and the collector connected to ground (0 V). These two solutions are shown in Fig. 13.6.
Figure 13.6 MOSFET Gate Drive Circuit.
When the gate driver output is being pulled to ground and current flows out of the MOSFET gate, the PNP transistor’s base–emitter junction is then forward biased. The PNP transistor will turn on and the MOSFET gate will be pulled down to ground with low impedance; any current through the parasitic drain–gate capacitor will be shunted to ground. The collector–emitter voltage can drop to about 0.2 V at this time.
In some cases, the PNP transistor’s base–emitter junction is reverse biased and could break down like a Zener diode when the gate driver output goes high. Some people add a signal diode in series with the base, to prevent reverse bias. However, many LED drivers have limited gate drive (the HV9910, e.g., only has 7.5 V drive and the current is limited to about 0.2 A), so these circuits will not have a problem. If an LED driver has a higher voltage/higher current gate drive output, it is unlikely to require the PNP transistor circuit.
PCB layout is also important to limit EMI radiation. The paths for the switching currents must be kept short and compact. If components cannot be colocated, so the path length is a little longer than desired, a return path should be placed alongside to ensure that the magnetic field from the current loop is minimized.
It is important to realize that if a ground plane is beneath a conductor carrying high frequency signals, the ground plane will carry the return current along the same path. The reason for this effect is that magnetic forces produced by the signal current provide a lower impedance for the return current. These magnetic forces fall rapidly with distance from the conductor, so the lowest impedance path will be that closest to the signal path (i.e. directly underneath). It is therefore important that the ground plane is continuous along this path. If there is a break in the ground plane below the conductor, the current’s return path may follow a circuitous route and then the ground plane will become an antenna!
Using the circuit schematic of Fig. 13.3, we can look at the PCB design. In Fig. 13.7, the track layout of the bottom layer is shown.
Figure 13.7 Printed Circuit Board (PCB) Bottom Layer.
Notice how the ground connection goes from C5 to C1 and then on to C2 before reaching the ground plane. By avoiding the direct connection between the ground plane and C1, the current flow is steered in the direction we want. High frequency signals are taken from the grounded side of C2, which is low impedance at high frequency. The capacitors C5 and C1 hold up the input voltage during the cusps of the rectified AC input and are not intended to supply the high frequency current pulses needed for the LED load.
Fig. 13.8 shows the tracks on the PCB component side. The positive supply from the bridge rectifier BR1 flows to C5 and onto filter inductor L2. From the other side of L2, it passes to C1 and then to C2. Notice that the C2 connection is a node where current also returns from the cathode of D1. Thus the high frequency flywheel current loop is from Q1 drain, through diode D1, and back to ground via C2; this is in a small area to keep the impedance low and EMI radiation to a minimum. As with the ground connection, the high frequency current path is kept away from the low frequency current flowing into capacitors C5 and C1.
Figure 13.8 PCB Top Layer.
Fig. 13.9 shows both sides of the circuit board overlaid. Notice that the earth plane is below the drain area of Q1 and inductor L1. Both Q1 and L1 have high frequency, high voltage switching, and a ground plane below helps to reduce the radiation from this area by screening underneath and making the node low impedance. Of course the parasitic capacitive coupling adds to the switching losses, but this cannot be avoided.
Figure 13.9 PCB Top and Bottom Layers.
13.2.2. Ćuk Circuit Example
A Ćuk circuit is a boost–buck converter that performs well in a DC input application. An example of a Ćuk circuit is given in Fig. 13.10.
Figure 13.10 Basic Ćuk Circuit.
As with the buck circuit already described, or any other switching circuit, the aim in PCB design is to keep the switching currents flowing in as small a loop as possible. An earth plane under the main switching elements will also help reduce radiation.
Radiation couples easily into free space when the impedance of the signal source is similar (the impedance of free space is 377 Ω). Dipole antennas radiate and receive signals easily because their metallic elements are resonant at the transmit frequency and thus high impedance at the ends of the elements. Similarly, if the circuit area containing the high voltage switching signals is high impedance, it will radiate interference. An earth plane under the circuit lowers the impedance and reduces radiation. The PCB designer should take this into account when designing the circuit board.
High frequency emissions are caused by the fast rising and falling edges of the MOSFET drain voltage. As we discussed earlier, these can be reduced in amplitude by slowing down the switching of the MOSFET. Not only does this reduce high frequency emissions, it also reduces high frequency ringing that is caused by the drain–gate capacitance resonating with stray circuit inductance. In Fig. 13.10 notice that a resistor (R5) has been connected in series with the gate. Slowing the MOSFET switching speed reduces the efficiency of the LED driver circuit, but saves the cost of additional filters.
Despite the natural input and output filtering inherent in the Ćuk circuit, additional filtering of the input and output power connections is likely to be required. A modified circuit is shown in Fig. 13.11, with an added input filter needed to meet demanding automotive specifications.
Figure 13.11 Input Filter.
Capacitors C3 and C4 provide the current source for high frequency switching; these are ceramic capacitors and have very little high frequency ripple across them. Inductor L1 and capacitor C2 form a low pass filter to attenuate any ripple that appears across C3 and C4.
The circuit ground is not the same as the supply ground because two parallel resistors R1 and R3 break the ground connection. This means that a path for any high frequency signals is needed from the positive input to both the circuit ground and the supply ground. The path to circuit ground is provided by C20. This small value ceramic capacitor does not affect current sensing at the switching frequency. The path to supply ground is provided by C2, which is a high value ceramic capacitor.
Stray coupling from the LED driver circuit to ground can create a common mode signal that is present equally on positive and negative inputs. This means that a differential capacitor, like C2, has no effect since the voltage is the same on both sides of the capacitor and no current flows through it. For this type of signal, a common mode inductor (choke) is required.
A common mode inductor L4 has two windings on a common magnetic core. Differential currents produce opposing magnetic fields, so the result is no net inductance. Common mode currents produce magnetic fields that add together and thus have a high inductance. A common mode signal will present high impedance to common mode signals and thus reduce radiation.
Finally, a small value ceramic capacitor C11 is connected differentially across the power supply input to provide a low impedance path at the higher frequencies.
The output filter is highlighted in Fig. 13.12.
Figure 13.12 Output Filter.
The output filter may be required, especially if there is a considerable wire length between the driver circuit and the LED load. If the distance is very short, the only filter usually required is a differential capacitor (C10) across the load. Distances greater than 10 cm (4 in.) between driver circuit and LED load can cause common mode signals to be created, primarily due to stray coupling between the LED and ground. Thus we may require a common mode inductor, L5, and a second differential capacitor C23. Small value ceramic capacitors C21 and C22 provide a shunt path to circuit ground for high frequency signals developed across L3 and the parallel current sense resistors R8 and R12.
In addition to a ground plane on the circuit board to reduce the impedance of the switching circuits at high frequency, a screen over the components may be needed. The position of such a screen is shown in Fig. 13.13.
Figure 13.13 Area for Screening.
A metal screen over the switching area, with an earth plane underneath, provides an enclosure that stops EMI radiation. However, there will always be some leakage due to signals being carried outside the enclosure by connections to the remainder of the circuit. Even the PWM control wire will radiate unless a simple RC filter is added to it.
13.3. EMC Standards
The EMC performance is often automatically assured by the EMI precautions previously described. If radio frequency signals cannot get out of some equipment, they usually cannot get in either. However, electrostatic discharge (ESD) and surge immunity are two areas that are not taken into account in EMI practices.
People generate high electrostatic voltages during normal activities, such as walking across a carpet or opening a plastic envelope. A charged person touching electrical equipment can cause damage or malfunction. Thus equipment must be protected against high voltage discharge. Testing is carried out as specified in IEC/EN 61000-4-2 using an ESD gun, which can simulate the “human body model” (or HBM as it is often referred to). The standard voltage levels are 4 kV for a contact discharge and 8 kV for an air discharge.
Any equipment connected to the AC mains supply must withstand surge pulses, as specified in IEC/EN 61000-4-5. Each surge pulse has an open circuit rise time of 1.2 μs and a fall time of 50 μs. In domestic equipment, the peak surge voltage is 1 kV, which is added to the AC mains supply. In addition, 2 kV surges are applied between the inputs and ground (earth). The test pulses are positive and negative, and are applied at 0-, 90-, 180-, and 270-degree phases of the AC mains voltage.
Another form of surge test is the fast transient burst, as specified in IEC/EN 61000-4-4. This comprises ±2 kV pulses with a rise time of 5 ns and 50% decay at 50 ns. These pulses are repeated at a 5 kHz rate (200 μs between pulses), for 15 ms. There are 75 pulses in each burst, and the bursts are repeated every 300 ms, for 1 min. Testing is usually carried out by first applying ±250 V bursts, then ±500 V, then ±1 kV, and then finally ±2 kV.
13.4. EMC Practices
Equipment connected to AC mains power lines must be surge tested. The surges are applied, which are added to the normal AC voltage, at times to coincide with different phases of the AC line. The source impedance of the surge test pulse generator is a nominal 50 Ω. The energy in surge pulses can be absorbed or reflected to limit its damaging effects in the EUT. Absorbing the energy in surge pulses is the most common method of preventing damage.
A varistor, which is a voltage-dependent resistor made from a metal oxide, is commonly used to absorb energy by clamping the voltage. In a varistor rated at 275 V AC, the clamping voltage is typically 710 V, although conduction begins at about 430 V. The amount of energy absorbed in a varistor depends on its physical size. A varistor is usually wire-ended and disc-shaped; the diameter of the disc is related to the maximum energy (usually given in Joules). For example, a 9-mm disc varistor from Epcos that is rated for 275 V AC has a transient energy rating of 21 J and a peak current rating of 1200 A.
Another energy absorbing device is a transient voltage suppressor (TVS or TransZorb). This device is a Zener diode made in silicon and has a stronger clamping action. These are available with either bidirectional or unidirectional breakdown. In AC systems a bidirectional breakdown is required, but in automotive and other DC applications, a unidirectional breakdown is sufficient. TransZorb devices are usually rated in peak power (watts); 600- and 1500-W devices are commonly available.
The oldest technology, and still sometimes used, is the gas discharge tube. This has a glass tube filled with inert gas and metal electrodes at either end. When the voltage across the electrodes is high enough, the gas ionizes and conducts to clamp the voltage.
A plastic film capacitor is often connected across the AC line (typically 100 nF, 275 V AC X2 rated). This not only helps to reduce EMI emissions and susceptibility, it also helps to absorb some of the energy in surge pulses. Surge suppressors take some time to respond to impulse voltages; so fast transients can sometimes pass with little loss and can cause damage.
Many systems have a large electrolytic capacitor across the power rails, after a bridge rectifier. This capacitor will absorb surge energy; however, electrolytic capacitor construction results in some inductance that will have high reactance to fast rising surge pulses. A plastic film capacitor connected in parallel with the electrolytic capacitor will help to absorb high frequency energy. A clamping device, such as a varistor directly across the AC line is still a good idea because it limits the surge before it reaches the bridge rectifier.
A fuse should be fitted to every piece of equipment powered from the AC mains supply. This provides a means of limiting energy into an antisurge device, like a varistor. When a high-energy surge causes the varistor to break down, the fuse will blow. Some people fit a high power wire-wound resistor between the AC line and the varistor, to limit the current from surges and prevent burnout of the varistor.
When laying out a PCB, the spacing between tracks should be carefully considered. The breakdown voltage of an air gap is about 1 kV/mm, so at the potentially high voltage input of a power supply, sufficient gap should be allowed. An air gap of 3.2 mm is the minimum to prevent breakdown and a potential fire hazard. On a PCB the gap between conductors is known as the creepage distance. The air gap from a live part of the circuit to any other parts of the enclosure is known as the clearance distance.
Integrated circuits that can be powered directly from the rectified AC supply usually have “no connect” or NC pins adjacent to the high voltage pin. This is designed to give a suitable “creepage” distance. Where no gap exists, a slot can be cut in the PCB, or the contact pins can be coated with a conformal coating or a resin to increase the insulation.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.