In the past EMC Engineers have relied on metallic enclosures to contain electromagnetic fields and meet radiated emissions limits in military and consumer products. Modern commercial electronics products typically use molded plastic enclosures since they are considered to be aesthetically more pleasing than a metal enclosure, but also to save weight and cost.
With correct PCB layout, differential signaling and common mode filtering on cables, it is sometimes possible to meet commercial EMI requirements without employing any shielding in the enclosure. However with the increased complexity, component density and speed of logic, designers are frequently coating the plastic enclosure with a thin conductive layer to provide a level of shielding. In addition, metal shields may be placed directly over noisy and sensitive components on the PCB, to further reduce emissions and improve immunity. A conductive coating in principle can be very effective. In practice, the seam between the two halves of a clam-shell type enclosure or between the enclosure and the PCB reference plane limits the shielding effectiveness. This is due to poor electrical contact at the interface, caused by inadequate pressure, low contact surface area and gaps due to unevenness in the formed parts or the coating.
In a high density compact electronics system, such as a cell phone, it may be necessary to place solid metal EMI enclosures over noisy components to reduce emissions, or over sensitive components to improve immunity. This can be particularly important when multiple radio communications systems are closely located and radio frequency interference (RFI) must be minimized. The shielding performance of metal enclosures also strongly depends on electrical contact to the PCB. The enclosure typically includes a number of tabs to connect to the PCB and there can be gaps between successive tabs. Furthermore, the enclosure may be perforated, typically on the top surface, to provide ventilation and this may compromise the shielding performance, especially at high frequencies.
The relative shielding effectiveness of various PCB shield strategies will be investigated in this article by applying 3D electromagnetic field simulation, based on the time-domain 3D Transmission-Line Matrix (TLM) solver. Solving the EM fields in the time-domain enables the system impulse response to be extracted from a single computation. Fourier transform can subsequently be applied to yield the broadband peak radiated field or emissions. Shielding effectiveness can be calculated by comparing the radiation with and without the shield present.
We will first calculate the shielding of a conductively coated plastic enclosure and explore the degradation in performance with increasing seam impedance. We will then investigate the use of component shielding in a GSM cell phone application to isolate two sensitive PCB components from the antenna fields. Finally we will model a graphics PCB used in an automotive display system, where a metal cover is placed on one side of the board to shield noisy digital circuits.
Conductively Coated Clam Shell Enclosure
For this first application a plastic enclosure 8cm wide, 12cm long and 6cm high is coated with a conductive Nickel film of thickness 0.001 inch (0.0254 mm). For thin conductive coatings, it is important to assess the magnetic field shielding effectiveness, since it is possible that the skin depth of the surface current is larger than the conductive film thickness. The skin effect causes the effective resistance of the conductor to increase with the frequency of the current. At 1 MHz in Nickel, the skin depth is about 0.12 µm. The skin depth (δ) is inversely proportional to the square root of frequency (f) and conductivity (σ). Increasing frequency results in smaller skin depths.
δ = 1 / √ πfµσ
The frequency-dependent diffusion of current through the thin conductive coating is represented accurately in the TLM model by a special thin panel boundary condition. It is not necessary to use volume mesh cells to capture the film thickness so this speeds up the calculation and reduces the computer memory required to solve the problem.
Figure 1: 3D TLM model of Conductive Coated Enclosure with Transmit Loop
In reality, the enclosure contains a groove to hold a conductive gasket which makes electrical contact between the two mating halves of the enclosure. This is modeled by an equivalent conductive seam model in the TLM electromagnetic simulation. The model allows for the transfer impedance of the joint to be varied and the impact on shielding performance assessed. The two halves are screwed together in all 4 corners with conductive screws and it is assumed that there is good electrical contact at these points.
Due to the thin conductive coating and skin depth effect, the magnetic field shielding effectiveness is the primary concern for this study. To assess the magnetic shielding, a 20cm radius transmitter loop is located 5cm away from one of the walls and a similar receiver loop placed at the geometric center of the enclosure. The transmitter loop is driven with a 1V source and series 1 Ohm load and the receiver loop is terminated in a 1 ohm load. The mutual inductive coupling between the loops with the enclosure removed is first solved to obtain a reference result. The enclosure is then inserted and the fields re-calculated. The magnetic shielding effectiveness is determined by normalizing the results, or subtracting dB.
Shielding (dB) = Reference Result (dB) – Shielded result (dB)
Figure 2: Magnetic Shielding Effectiveness Plotted Against Frequency
Results are provided for seam transfer impedance values of 0, 1, 10, 100, 1000 milli Ohm-m. The results show a progressive reduction in shielding performance with increasing seam transfer impedance. The voltage developed across the seam (V) is proportional to the surface current flowing over the seam (Js) and the transfer impedance (Zt).
V = Js x Zt
If the seam impedance is zero, in other words perfect electrical contact between the two halves of the enclosure, the seam voltage will be zero and the shielding will be purely based on the inherent ability of the conductive film to attenuate the fields. From the curve in the graph plot, we can observe that the conductive film provides approximately 30dB shielding at 100 KHz. The shielding effectiveness improves with increasing frequency and this is due to the skin depth effect. At high frequencies the skin depth is smaller than the film thickness and the current is confined to the external surfaces of the enclosure.
The field plot in Figure 3 shows the magnetic field vectors at 1 MHz with a 10 milli Ohm-m seam transfer impedance. The magnetic field is mainly coupling through the seam and this is limiting the shielding performance of the enclosure.
Figure 3: Magnetic Field at 1 MHz with 10 milli Ohm-m Seam Transfer Impedance
The TLM simulation requires approximately 10 minutes run time on a core 2 Duo T9600 based laptop. The model uses 10,500 mesh cells and requires only 13 MB of computer RAM.
RFI Shielding in a GSM Cell Phone Application
The next application is a cell phone with a dual-band Printed Inverted F Antenna (PIFA) antenna, tuned for the GSM frequencies 850 and 1900 MHz, typically used in North America. The model is used to investigate the isolation of two sensitive electronics components located nearby to the antenna element. Component A is located approximately 10mm away from the PIFA antenna element and component B is directly under the element. Wire traces are used to model nets at the component locations and the induced voltage and current monitored. The wires are arranged diagonally to ensure that different polarizations of the field are captured.
Figure 4: Cell Phone Model with Component Shields Present and Removed
Simulation is used to predict the reduction in coupling when metal shields are placed over the components. The PIFA antenna is essentially a folded monopole, with an inductive stub used to compensate for the capacitance between the radiating element and PCB reference plane. The near field impedance is relatively high, so mutual capacitance between the antenna element and victim traces could be the coupling mechanism of concern. The metal covers serve as electric field shields and shunt the RF current to the reference plane.
The covers are not perfect shields, due to the use of 1mm diameter round perforations to provide ventilation for cooling of the internal electronics. There are also small gaps between the metal tabs used to make contact to the PCB reference plane. The results in Figure 5 plot the coupling to the two components when a constant 1 Amp (0dB) current is driven into the PIFA antenna.
Figure 5: Coupling Between PIFA and Components With and Without Shields
With no shields present, the received current is approximately 48dB down at the antenna resonances of 850 and 1900 MHz. The metal enclosures provide around 38dB to 44dB shielding at 850 MHz, increasing the isolation to 86dB (component A) and 92dB (component B). The shielding effectiveness reduces to 28 dB at 1900 MHz, but this still improves the isolation to 76dB (both components). It is not surprising that the shielding is less for higher frequencies since the ventilation holes and spaces between contact tabs become electrically larger.
The surface current density is plotted in Figure 6 at the antenna resonant frequencies. Notice that the current prefers to flow along the sharp metal edges of the antenna element and corners of the metal cans, indicated by the orange-red coloring. This is a well known effect for high frequency currents. The electric field will be strong at the metal edge discontinuities, so it is possible that there is capacitive coupling from the edges of the antenna element to the edges of the metal enclosures.
The GSM cell phone simulation requires 15 minutes run time on a core 2 Duo T9600 based laptop and uses 25 MB RAM. This produces the shielding results of the entire spectrum from DC to 2.6 GHz.
PCB Shielding in an Automotive Display System
The final example is concerned with the shielding of a graphics PCB used in an automotive display cluster. The PCB is approximately 10 x 6 cm and has multiple layers. For the electromagnetic analysis we focus our attention on the emissions generated by the DRAM clock net, which is routed on one of the outer layers. The net is essentially a microstrip conductor surrounded by a reference plane structure and this is intended to provide return paths for the high frequency currents and thereby reduce the emissions. Nevertheless, some field will inevitably “escape” and lead to radiation from the PCB. To contain the fields, a metal shield of size 7cm x 5cm is placed over the PCB. It is critical that the shield does not short out components and traces on the PCB, so contact can only be made to the reference plane at certain locations. For this design, contact is made at the 4 corners of the shield and also the middle points along the two longer edges. Therefore, we do not expect the shield to be perfect, but we would certainly hope for some level of shielding across the frequency band of interest.
Figure 7: Automotive Display System Graphics PCB Model With and Without Shield
In reality the DRAM clock signal has a certain frequency and rise/fall time which generates a spectrum of discrete frequencies including the fundamental and harmonics. We could drive the model with this transient signal, but it is often more useful to excite the net with a pseudo-impulse which contains all frequencies up to the limit of the model. This ensures that any narrowband peaks in radiated emissions are detected. The impulse response of the electric field observed at a point 1m above the PCB is shown in Figure 8. The response includes all the reflections and resonances associated with the PCB and shield structure.
Figure 8: Typical Impulse Response from the Time-Domain TLM Analysis
The radiated field is monitored at a single point 2cm above the metal shield (near field probe) and at multiple points scattered around the PCB on a 1m radius (far-field probes). The field is also scanned continuously on a 1m radius to determine the peak radiated emissions.
Figure 9: Shielding Effectiveness Observed 2cm above the PCB/Shield
The graph plot in Figure 9 shows the shielding effectiveness as observed by the probe 2cm above the metal shield. The metal enclosure provides good shielding at low frequencies, and this is due to the observation point being located in the “shadow” of the electromagnetic field. For other components placed in this location we can expect very good isolation. The shielding steadily reduces with increasing frequency and in fact negative shielding is seen at 2.1 GHz. Negative shielding can occur when one or more half wavelengths match one or more physical dimensions of the structure. Reflections back and forth between opposing boundaries generate standing waves, producing cavity resonances and build up of field strength.
Figure 10: Shielding Effectiveness Observed 1m Away from the PCB/Shield
The shielding derived from the 1m emissions scan is not as effective. This is due to radiation from the air gaps formed between the shield and PCB reference plane. The distant 1m observation points are in the path of the radiated field. The air gaps can essentially be considered to be slot antennas that will radiate very efficiently when the wavelength is comparable to the slot length.
The surface current density and peak electric field distribution is plotted in Figure 11 at 867 MHz for the cases without and with the shield present. 867 MHz is chosen because the DRAM clock net exhibits a resonance around this frequency and shielding of the radiated fields is important. The field plot clearly shows very little field escaping beyond the shield. The scale is from -100dB to 0 dB. The deep blue regions are -100dB down on the peak electric field.
The peak electric field distribution is plotted in Figure 12 at 2.1 GHz for the cases with and without the shield present. At this frequency the air gaps between successive electrical contact points are just the right length to resonate and radiate electromagnetic waves. Comparing the two field plots it is clearly seen that the shield actually increases the emissions at this particular frequency (negative shielding effectiveness). Notice the high field strength in the PCB/shield gaps and propagation of the fields beyond the shield.
Figure 12: Electric field at 2.1 GHz: without Shield (top), with Shield (bottom)
The PCB/shield simulation requires a 2 hour run time on a dual quad-core computer and uses 275 MB RAM. This produces the shielding results over the entire spectrum from DC to 5 GHz.
We have shown through 3 application examples how electromagnetic modeling can be effectively used to assess the performance of PCB shields. In all cases, the simulation run times and computer memory requirements are quite reasonable and this enables multiple iterations to be solved quickly to determine trends in the results. The ability to display the surface currents and fields can provide greater insight and verification of the dominant coupling mechanisms. There is tremendous value in simulating EMC problems early in design and revealing potential issues before manufacturing and testing. In the applications considered here, it has been shown that a PCB shield can be effective over certain bands, but it can have the opposite effect and increase emissions for certain frequencies. It is important for EMC Engineers to understand the limitations of proposed solutions when making decisions in product design reviews.
|Dr. David P. Johns is the VP of Engineering and Support for CST of America and is based in CST’s Boston MA location. He received his PhD in Electromagnetic Analysis from Nottingham University (UK) in 1996 for developing a new 3D frequency-domain Transmission-Line Matrix (TLM) method for solving electromagnetic fields. He contributed to the development of CST’s 3D time-domain TLM code MICROSTRIPES and in particular efficient techniques for modeling current diffusion, apertures and wires. David has over 20 years of electromagnetic simulation experience and specializes in the modeling of real world EMC/EMI problems. He is a regular speaker at IEEE EMC conferences and chapter meetings and recently the co-chair of the IEEE EMC Symposium Workshop “How to simplify real-world complex systems into realistic, solvable, accurate models.”
|Scott Mee received his BSEE in 1998 from Michigan Technological University (MTU) with focus areas in RF communications and electromagnetics. Since his graduation he has been working for Johnson Controls in EMC test development, A2LA/AEMCLRP accreditations, EMC design and simulation. Currently he is the Global Manager of the EMC expert team in the Automotive Electronics group at Johnson Controls. Scott is an IEEE EMC Society member and has been a contributing author to numerous technical papers and presentations on EMC. He served as a co-chair of the technical paper committee for the 2008 IEEE EMC symposium and an automotive EMC special session chair for the 2007 IEEE EMC Symposium. Scott is a NARTE certified EMC Engineer and his interests include EMC design, simulation, pre-compliance testing and product debugging.