29. August 2016

In earlier times when power supplies contained linear regulated transformers, the world of electromagnetic compatibility (EMC) was still okay. Circuits were clearly defined, frequencies low, electromagnetic emissions negligible. With increasing efficiency and performance, and with the breakthrough of switch mode power supplies EMC problems arose. But what caused these problems and gave the engineers quite a headache?
Electromagnetic compatibility describes the interaction between an electric device and its environment. This interaction both defines the limit for maximum radiation and implies an insensitivity to a coupled radiation. The limit for the permissible emission of electromagnetic waves ensures that adjacent devices are not impaired in their function and that no user is at risk. The sensitivity to electromagnetic radiation defines a boundary for existing radiation at which the power supply still works satisfactorily. But what are those limits? The considerable number of electric devices in our environment and the constantly increasing number of wireless applications make it difficult to create test conditions living up to such a complex interaction. This is where the standards come in which, in accordance with the individual application of the unit and the corresponding country, do not only define the limits but also specify the test setup. The following provides a short overview of the various measurements. The scope of this article will not suffice to describe all measurements and tests and is therefore limited to circuit and fieldbound emissions only.

Electromagnetic interferences which may be caused by the DUT can be transferred either conducted or not conducted to the environment. To measure the corresponding values, the first challenge is to create a neutral and repeatable environment. This means that the test setup needs to be done environmentally isolated to prevent falsified readings. Ideally, this would be a place in the open, far away from possible sources of interference.

When measuring a conducted emission, the DUT is not directly connected to the mains but via LISN (Line Impedance Stabilization Network). LISN simulates a representative impedance for the mains which is important for obtaining repeatable measuring conditions complying with subsequent application scenarios. In addition, the LISN suppresses high frequency interferences on the mains which would falsify the measurements. LISN also provides a connection for a test receiver, to which high frequency noises of the DUT, disconnected from the mains voltage, are applied.

Measuring of a spatially radiated interference emission requires more efforts for obtaining a normative setup. The setup specifies the creation of conditions similar to an open area measurement. The 10 m open area test site serves as reference for the standards. This test site is - except for the antenna and an electrically conductive surface - free of electromagnetic obstacles. The specimen is located 10 m away from the antenna and 1 m above the conductive surface. The conductive surface serves as reference ground plane and reflects the electromagnetic radiation of the specimen. As no obstructions are permitted in the environment, the reference plane is the only reflector for the electromagnetic field of the specimen. Since the specimen does not necessarily radiate isotropically, i.e. it does not evenly radiate in all directions, the specimen must be scanned from all sides and rotated accordingly. Since the polarization of the electric field emitted by the specimen is unknown, the antenna has to detect both horizontally and vertically polarized fields, requiring a rotatable installation of the antenna. Such an open area test site is influenced by its environment. Not only the weather conditions play a crucial role but also various radio services preventing the detection of interferences in the respective frequency bands.

In order to obtain a setup which can be compared to an open area measurement but which is independent of environmental influences, absorber chambers or reverberation chambers or cells (GTEM & TEM) are required. They differ in size, maximum possible size of the specimen, and the permissible frequency range. This article describes the measurement in an absorber chamber.

The test setup in an absorber chamber is similar to the structure of an open area test site. An antenna with a defined distance to the specimen and the reference ground plane is arranged like for an open area test site. The test setup, however, is completely enclosed by a cage with high electrical conductivity to shield external noises. Since this cage acts like a reflector for the emitted radiation of the specimen, the readings are falsified. The generated reflections overlap with the direct radiation constructively and destructively, the measured field intensity is therefore either reinforced or nullified. Minima and maxima in turn are highly dependent on the spatial arrangement as well as on the frequency, and hardly predictable. To suppress these reflections, different absorbers are attached to the inside of the metal wall, matching the impedance of the field resistance with that of the open area (377 Ω). Due to the impedance matching, the energy is not reflected but converted into heat. Since the efficiency of the absorber is dependent on several factors, which is why the quality of the absorber chamber can not be predicted, a reference measurement is necessary. A reference radiator substitutes the DUT and emits a defined radiation. The theoretically expected field intensities, generated for open area test sites, are compared to the measured results and may only differ within defined limits (± 4 dB). If these criteria are met, the absorber chamber is approved.

Readings for both wired and conducted emission are displayed in an amplitude spectrum and may not exceed frequency-dependent limits. The readings can be determined with different detectors as average (AVG), peak (PK) or quasi peak (QPK). QPK is a weighted score which takes into account both the maximum amplitude and the frequency of occurring interferences and which is therefore less than or equal to PK. QPK and PK are equal only in continuous unattenuated signals. Since the QPK measurement is significantly more time consuming, it is quite common to carry out a PK measurement first and perform the QPK- test only for those frequencies whose PK readings exceed the predefined QPK limit.

Should measurements prove that the limits are not met, countermeasures must be taken, for example the implementation of a filter. Especially for mass production or products with limited space the filter must be tailored to the device to meet the desired efficiency and performance. In addition to experience, it might be helpful to examine the readings which often provide information about the origin of the interference. Figure 3 shows measurements of a switch mode power supply with flyback converter. It explains some relationships between the time pattern of the power switch's drain-source voltage (1) and the readings of the conducted (2) and fieldbound (3) interference level..

Noise sources differ in bandwidth of the radiated interference and can be grouped as narrow-band or broadband interference. A sinusoidal signal with a fixed frequency would appear as a single line in an amplitude spectrum and is regarded as narrow-banded. If, for example, a decaying sinusoidal signal as shown in figure 3 (1 b and 1 c) is concerned, this signal echoes at the frequency of the fundamental oscillation (b = 2,2 MHz, c = 450 kHz) as a cluster of several spectral lines. Such a measurement cluster for conducted radiation is displayed in figure 3 (2), the corresponding areas are marked (b and c) for easy identification. The steep switching edges (see 1 a) are considered as broadband interferences. Via the pulse width and repetition rate (in this case the repetition rate is equal to the switching frequency), and by using a Fourier transformation the corresponding amplitude spectrum can be determined. This spectrum shows an undulation which peaks at the switching frequency and odd multiples, see figure 3 (2 d). If the switching edges are very steep, this undulating process is reflected in the entire measured spectrum. In these examples, the noise level could be quite easily linked to the timing, as it was a known source. Since a switch mode power supply, however, contains many components which can be considered potential sources of interference, a clear identification is not that simple. The measurement of conducted interferences as displayed in figure 3 (3) shows a distinct interference in area (e) of approx. 100 MHz. Such high-frequency interferences may point to faulty switching of the power switch, to reverse recovery currents of a diode or to resonant systems.

To tackle these problems analytically is barely successful since parasitic effects may cause the interference. It should be empirically tested which component causes the problem. This is very time consuming since an interaction between different components cannot be excluded. Consequently, the detection of interferences is often more complicated than the actual compensation.

There are various possibilities to reduce the noise emission. One of it is the short-circuiting of these interferences. For this purpose, a low frequency-related impedance is connected in parallel to the source. Typically a capacitor or a small network is chosen. An example can be found in the Y-capacitor which compensates the coupling transferred via the parasitic capacity between primary and secondary winding (see figure 4). In the fieldbound radiation a signal can be short-circuited by a screen, consisting, for example, of a metal foil, a heat sink or the like.

The second alternative to compensate a noise emission is to introduce a high impedance to prevent spreading. This can be done with simple resistors or inductors. A good example of this correlation is the filter at the entrance. It serves the purpose of protecting the electric circuit against high-frequency signals of power electronics. Simultaneously, the energy from the mains is to be passed efficiently to the power electronics. Since the mains voltage is 50 Hz compared to the interference of 15 kHz and therefore small, inductors are used because they come with low impedance at low frequencies and increase only at high frequencies. This is important because the filter is located on the conductive path and the best tradeoff from good filter effect and energy balance.

The third option to compensate an interference is to fight the cause. An obvious solution at first sight but not always the best. Considering the switching behavior of the power switch, the noise emission can be reduced by flattening the switching edges. Since the power loss of the power switch is increased, this process is somewhat limited. If the source of the interference turns out to be a resonant system, the cause of the interference can be suppressed by attenuating the overall performance of the system, or by shifting the system's resonance frequency by adding more components. Unfortunately the engineers are facing limitations again since parasitic effects within the components can create these phenomena. The only thing left to do is to replace the components. The usage of a transformer often leads to these phenomena, since the design, shape factor and materials considerably impact the electromagnetic behavior. Sometimes small changes of the windings are responsible for serious differences. The same applies to the layout. As mentioned before, large conductor loops or electrostatic coupling between two parallel tracks should be avoided. This is not always possible since the power density of the switch mode power supplies increases constantly, reducing the available space on the PCB.

Friwo AG published this content on 29 August 2016 and is solely responsible for the information contained herein.
Distributed by Public, unedited and unaltered, on 08 September 2016 09:42:03 UTC.

Original documenthttp://www.friwo.de/en/news/news/emc-for-switch-mode-power-supplies/91b072bf22d95b4dad457efab542a02d/

Public permalinkhttp://www.publicnow.com/view/72C97922416C1A2B712A9F3F8762F614794850D5