“With its high power conversion efficiency, switch-mode power supplies are widely used in modern Electronic systems. However, a side effect of the increase in the number of switch-mode power supplies is the generation of switching noise. These noises are generally called electromagnetic interference (EMI), EMI noise, or simply noise. For example, the switching current on the input side of a typical buck converter is a pulsating current and is rich in harmonic components. Turning on and off the power Transistor quickly will cause a sudden interruption of the current, leading to high-frequency voltage oscillations and spikes.
“
Authors: Henry Zhang and Sam Young
Introduction
With its high power conversion efficiency, switch-mode power supplies are widely used in modern electronic systems. However, a side effect of the increase in the number of switch-mode power supplies is the generation of switching noise. These noises are generally called electromagnetic interference (EMI), EMI noise, or simply noise. For example, the switching current on the input side of a typical buck converter is a pulsating current and is rich in harmonic components. Turning on and off the power transistor quickly will cause a sudden interruption of the current, leading to high-frequency voltage oscillations and spikes.
The problem is that high-frequency noise can couple with other devices in the system, degrading the performance of sensitive analog or digital signal circuits. Therefore, the industry has produced many standards to set acceptable EMI limits. In order to meet these limits of switch-mode power supplies, we must first quantify its EMI performance, and if necessary, add appropriate input EMI filtering to attenuate EMI noise. Unfortunately, EMI analysis and filter design are often a difficult task for engineers, requiring repeated design, construction, testing and redesign, which is very time-consuming-this is still on the premise of having suitable test equipment. In order to speed up the EMI filter design process to meet EMI specifications, this article describes how to easily and quickly estimate EMI noise analysis and filter design, and use ADI’s LTpowerCAD®The program is pre-built.
Different types of EMI: radiated and conducted noise, common mode and differential mode
EMI noise is mainly divided into two types: radiation type and conduction type. In switch mode power supplies, radiated EMI is generally caused by rapid changes on the switch node and high dv/dt noise. Electromagnetic radiation industry standards generally cover the 30 MHz to 1 GHz frequency band. At these frequencies, the radiated EMI generated by the switching regulator mainly comes from switching voltage oscillations and spikes. This noise largely depends on the layout of the PCB board. Because these noises are determined by circuit parasitic parameters, it is almost impossible for an engineer to accurately predict how much radiated EMI a switch-mode power supply will transmit on the “calculation paper” except for ensuring good PCB layout practices. To quantify its radiated EMI noise level, we must test the circuit board in a well-designed EMI laboratory.
Conducted EMI is caused by rapid changes in the conducted input current of the switching regulator, including common mode (CM) and differential mode (DM) noise. The frequency range covered by the industry standard limits for conducted EMI noise is generally smaller than the range of radiated noise, ranging from 150 kHz to 30 MHz. Figure 1 shows the common conduction paths of common mode and differential mode noise of a DC-DC power supply (that is, a DUT in an EMI laboratory).
In order to quantify the conducted EMI at the input, we need to set up a line impedance stabilization network (LISN) at the input of the regulator during testing to provide standard input source impedance. Measure common mode conducted noise between each input line and ground. Common mode noise is generated on the high dv/dt switching node through the parasitic PCB capacitance C of the devicePGround, and then transfer to the power input LISN. Like radiated EMI, high-frequency switching node noise and parasitic capacitance are difficult to accurately model and predict.
Differential mode (DM) noise is measured differentially between two input lines. DM conduction noise is generated from the high di/dt, pulsed input current of the switch mode power supply. Fortunately, unlike other EMI types, the pulse input current generated in the input capacitor and LISN circuit and the resulting relatively low frequency EMI can be predicted by software such as LTpowerCAD with high accuracy. This is also the focus of this article.
Figure 1. Conceptual overview of the implementation of LISN-based measurements on differential-mode and common-mode conducted EMI of switch-mode power supplies.
Figure 2 shows a typical differential mode EMI noise graph of a switch-mode step-down power supply (without input EMI filter). The most significant EMI spikes appear at the switching frequency of the power supply, and the additional spikes appear at harmonic frequencies. In the EMI graph shown in Figure 2, the peak values of these spikes exceed the CISPR 22 EMI limit. To meet these standards, an EMI filter is needed to attenuate differential mode EMI.
Figure 2. Typical differential mode EMI diagram of a switch-mode step-down power supply without an input EMI filter.
Differential Mode Conducted EMI Filter
Figure 3 shows a typical differential mode conducted EMI noise filter on the input side of a switch-mode power supply. In this example, we have the local input capacitance C of the power supplyINA simple first-order low-pass L is added between (EMI noise source side) and input source (LISN receiver side)fCfThe internet. This matches the test setup of a standard EMI laboratory, where the LISN network is connected to the filter capacitor C of the LC filterf One side. Using a spectrum analyzer to measure the differential signal on the LISN resistor R2 can quantify the DM conducted EMI noise.
Figure 4 shows the attenuation gain of the LC filter. At very low frequencies, the Inductor has a low impedance, which is essentially a short circuit, while the impedance of a capacitor is high, which is essentially an open circuit. The corresponding gain of the LC filter is 1 (0 dB), so that the DC signal can be transmitted without attenuation. As the frequency increases, at LfCfA gain spike appears at the resonant frequency of. When the frequency is higher than the resonance frequency, the filter gain is attenuated at a rate of C40 dB/10 octave. At relatively high frequencies, the filter gain is basically determined by the parasitic parameters of the components: such as the ESR and ESL of the filter capacitor, and the parallel capacitance of the filter inductor.
Since the attenuation ability of this LC filter increases rapidly with the increase in frequency, the amplitude of the first few low-frequency noise harmonics determines the size of the EMI filter to a large extent. Among them, the basic power switching frequency (fSW) is The noise component is the most important target. Therefore, we can focus on reducing the low frequency gain of the EMI filter to meet industry standards.
Figure 3. Differential mode EMI noise filter (from node B to node A).
Figure 4. Typical single LC EMI filter insertion gain versus frequency.
LTpowerCAD can easily predict the filter performance of the power supply
LTpowerCAD is an auxiliary tool for power supply design, which can be downloaded and installed at analog.com/LTpowerCAD. The program can provide engineers with design assistance, allowing them to design and optimize the entire power supply parameters in a few simple steps.
LTpowerCAD guides the user through the entire power supply selection and design process. Users can start to input power supply specifications. On this basis, LTpowerCAD selects the appropriate solution, and then helps users select power stage components, optimize power supply efficiency, design loop compensation, and load transient response.
In this article, we are going to introduce LTpowerCAD’s input EMI filter design tool, which enables engineers to quickly estimate differential mode conducted EMI and determine which filter components are needed to meet EMI standards. The filter tool of LTpowerCAD can help engineers estimate the parameters of the filter before the actual board and test, thereby significantly shortening the design time and reducing the design cost.
Implement EMI filter design using LTpowerCAD
Overview
Now let’s take a look at a DM EMI filter design example. Figure 5 shows the LTpowerCAD schematic design page, showing the selection of power components using the LTC3833 step-down converter controller. In this example, the converter uses 12 V input and 5 V/10 A output. Switching frequency fSWIt is 1MHz. Before designing an EMI filter, we should first design a buck converter by selecting the switching frequency, power stage inductance, capacitance, and FET.
Figure 5. LTpowerCAD schematic design page and integrated EMI tool icon.
After selecting the power stage components, as shown in Figure 6, we can click the EMI design icon to open the DM EMI filter tool window. This EMI design window shows the power input capacitance CINB/CINC The input filter network LfCf between and the input terminal LISN. In addition, there is a backup damping circuit, such as the network C on the LISN sidedA/RdA, The network C on the side of the power input capacitordB/RdB , And filter resistance LfSpare damping resistor Rf inP. The right side of Figure 6 is the estimated conducted EMI noise graph and the selected EMI standard limits.
Figure 6. LTpowerCAD conducted DM EMI filter design window (Lf = 0, no filter).
Choose an EMI standard
When designing an EMI filter, you need to understand the design goal-that is, the EMI standard itself. LTpowerCAD has built-in schematic diagrams of CISPR 22 (commonly used in computers and communication equipment), CISPR 25 (commonly used in automotive devices) and MIL-STD-461G standards. You can directly select the standard you need from the EMI specification drop-down menu.
In the example in Figure 6, the value of the filter inductance is set to 0 to show the EMI results in the absence of an input filter. EMI spikes appear at the fundamental and harmonic frequencies, and both exceed the limits of the CISPR 25 standard shown, resulting in a red warning on the EMI and specification schematic Display.
Set EMI filter parameters
After selecting the required EMI standard, you can enter the required EMI margin, that is, the distance between the selected standard limit and the peak value of the fundamental frequency. Generally, a margin of 3 dB to 6 dB can be selected. Among these options, the LTpowerCAD program uses the given filter capacitance Cf and power supply operating conditions to calculate the recommended filter inductance L¬¬-sug., which is displayed in the yellow cell of LTpowerCAD. You can enter an inductance value slightly larger than the recommended value in the L cell to meet the EMI limit and the required margin.
In this example, Figure 7 shows that the design tool recommends using 0.669 µH filter inductance, and the user actually inputs 0.72 µH inductance to meet its requirements. Regarding the benefits of the filter, you can compare the results with the filter and the results without the filter. Open the Show EMI option without input filter, and you can view the filtering result superimposed on the gray schematic diagram without filter.
When selecting the filter capacitor CfAt this time, there is an important detail that needs attention. X5R, X7R and other types of dielectric material multilayer ceramic capacitors (MLCC), the capacitance value will decrease significantly as the DC bias voltage increases. Therefore, in addition to the nominal capacitance C(nom) of LTpowerCAD, the user should also apply a DC bias voltage (VINAOr VINB) Input the actual capacitance value.
For the derating curve, please refer to the data sheet provided by the capacitor supplier. If the selected MLCC capacitor comes from the LTpowerCAD library, the program will automatically estimate the derating associated with the DC bias voltage of the capacitor.
In addition, we also need to consider the change in the inductance of the input filter. When the inductance saturates with the DC current, a non-linear inductance will be produced. As the load current increases, the inductance value may decrease significantly, especially for ferrite bead type inductors. The user should enter the actual inductance to generate an accurate EMI prediction.
Figure 7. Choosing the filter inductance value to meet the EMI standard limit.
Check filter attenuation gain
In the EMI graph with input filter shown in Figure 7, the LC input filter resonates at 245 kHz (the frequency is lower than the switching frequency of the power supply), which produces a noise spike. Figure 8 shows the filter attenuation gain graph, which replaces the EMI results in the LTpowerCAD EMI window (click the filter attenuation tab) to show the filter’s resonance attenuation gain at 245 kHz.
In some cases, LC resonance peaks may cause noise to exceed EMI standards. In order to attenuate this resonance peak, a pair of additional damping components C can be addeddAAnd RdA , And filter capacitor Cfin parallel. In addition to displaying attenuation diagrams, LTpowerCAD also simplifies the process of selecting these components. Under normal circumstances, the selected value is the real filter Cf2 to 4 times the value of the damping capacitor CdA. LTpowerCAD will provide the recommended damping resistance RdAValue to reduce the resonance peak.
Figure 8. EMI filter attenuation gain (with and without damping at the LISN end).
Check filter impedance and power input impedance
When adding an input EMI filter before the switch mode power supply, the output impedance of the filter ZOFWill be related to the input impedance Z of the power supplyIN, The interaction may cause the circuit to oscillate. In order to avoid this unstable situation, the output impedance amplitude of the EMI filter ZOF, Should be much lower than the impedance amplitude Z of the power inputIN, , And leave enough margin. Figure 9 shows ZOFAnd ZINThe concept of and the stability margin between them.
In order to simplify the analysis process, an ideal power supply with a high feedback loop bandwidth can be used as a constant power load; that is, the input voltage VINThe value multiplied by the input current is constant. As the input voltage increases, the input current decreases. Therefore, an ideal power supply has a negative input impedance ZIN = C(VIN2)/PIN.
To facilitate the design of the input filter, LTpowerCAD displays the output impedance Z of the filter in the impedance diagram shown in Figure 10.OFAnd power input impedance ZIN . Note that the power supply input impedance is a function of input voltage and input power. The worst case (ie the lowest impedance level) is at VINMinimum, PINOccurs at maximum.
As shown in Figure 10, the output impedance of the EMI filter is at the filter inductance LfAnd power input capacitance CINA peak point appears at the resulting resonance frequency. In a good design, the amplitude of this peak should be lower than the worst case ZINA peak point appears at the resulting resonance frequency. In a good design, the amplitude of this peak should be lower than the worst case CdBAnd resistance RdB, And the power input capacitance CINin parallel. This CINThe side damping network can effectively reduce the filter ZOUTPeak. The LTpowerCAD EMI tool provides recommended CdBAnd RdB parameter.
Figure 9. Check whether the output impedance of the EMI filter and the input impedance of the power supply are stable.
Figure 10. LTpowerCAD EMI filter impedance diagram (with and without damping).
Accuracy of LTpowerCAD EMI filter tool
In order to verify the accuracy of the LTpowerCAD EMI filter tool, we conducted actual test comparisons based on real circuit boards in the EMI laboratory. Figure 11 shows the difference between the real test results using the modified LTC3851 step-down power demo board (using 750 kHz, 12 V input voltage, 1.5 V output voltage, and 10 A load current) and the predicted EMI noise of LTpowerCAD Comparison. As shown in Figure 11, the measured EMI data and the low-frequency noise peaks of the EMI data simulated by LTpowerCAD roughly match. The actual test peak value is a few dB lower than the simulated EMI peak value.
In the higher frequency range, the magnitude of the mismatch is greater, but as mentioned earlier, since the size of the DM conducted EMI filter is mainly determined by the low-frequency noise peak, the errors in these high-frequency bands are not important. Many high-frequency errors are caused by the accuracy of inductance and capacitance parasitic models, including PCB layout parasitic values; for now, PC-based design tools cannot achieve this accuracy.
Figure 11. EMI measured by real board experiment and EMI estimated by LTpowerCAD (12 VINTo 1.5 VOUT/10 A step-down example).
It is worth emphasizing that the LTpowerCAD filter tool is an estimation tool that can be used to provide the initial design of EMI filters. To obtain truly accurate EMI data, users also need to build prototypes of power supply circuit boards and conduct real experimental tests.
Summarize
The systems used in many industries require stricter suppression of electromagnetic noise interference. Therefore, the industry has issued many clear standards for EMI noise. At the same time, the number of switch-mode power supplies continues to increase, and installation locations are closer to sensitive circuits. The switch mode power supply is the main source of EMI in the system, so in many cases it is necessary to quantify its noise output and reduce it. The problem is that the design and testing of EMI filters is often a process of trial and error, which consumes time and design costs.
In order to solve this problem, the LTpowerCAD tool allows designers to use computer-based predictive simulations before implementing real designs and tests, which greatly saves time and costs. The easy-to-use EMI filter tool can estimate the differential mode conducted EMI filter parameters, including an optional damping network to reduce EMI to a greater extent while maintaining a stable power supply. The experimental test results verify the accuracy of the prediction results using LTpowerCAD.
author
Henry Zhang
Henry Zhang is ADI’s Power by Linear™ applications manager. He received a bachelor’s degree in electrical engineering from Zhejiang University, China in 1994, and a master’s degree and doctorate in electrical engineering from Virginia Polytechnic Institute and State University (Black Fort) in 1998 and 2001, respectively.
Sam Young
Sam Young is a senior application engineer at ADI, providing μModule® regulator products and LTpowerCAD product support. He has more than 10 years of experience in helping to develop voltage regulator designs for instrumentation and communication products. Sam likes outdoor sports and will do something/things by himself in his free time when he is raising his son.
The Links: MG75N2YS40 6DI75MB-050