射频前端模组：自屏蔽工艺的量化建模和测量结果

很多射频前端芯片，都有银色的外观，并且有金属的光泽。这些金属光泽来自于芯片表面喷镀的金属屏蔽层。自屏蔽技术有哪些好处？它会取代传统的机械屏蔽罩吗？

Abstract

A reliable numerical modelling for shielding evaluation of on-packageconformal shields based on the integrated circuit (IC)-stripline method ispresented. As a pilot test, the effects of the number of grounding pads and thethickness of a conformal shield on the shielding performance are investigated.For validation, the conformal shield is fabricated and measured by both theIC-stripline method and TEM cell method.

Introduction

As modern mobile and wearable devices have become smaller, slimmer, and lighter, the shielding technology for preventing electromagnetic interference (EMI) in the devices is also changing.A traditional on-board shielding technique using small metallic cans makes itdifficult to realise smaller and thinner devices because the metallic cansresult in a high profile and large footprint, as depicted in Fig. 1. To overcome the limitation of theon-board shielding, on-package shielding technology named conformal shieldingis recently receiving great interest [1]. The conformal shielding is a novelmethod to reduce EMI on molded system in package (SiP) modules by coating thesurface of the module with a thin metal layer, which leads to a low profile andsmall footprint.

Fig 1 Advance in shielding technology formodern mobile devices

For shielding analysis of conformalshields, some research has been conducted by numerical modelling andmeasurement. In [2], the gigahertz transverse electromagnetic(GTEM) cell measurement has been proposed and the measured result was comparedwith the simulation result by a 3D commercial software. However, the differenceof shielding effectiveness (SE) between them is >20 dB at 3 GHz. Anear-field measurement using a loop probe has been proposed to evaluate thenear-field SE for conformal-shielded SiP with different shielding materials andthicknesses [3]. The simulation result has been obtainedfrom 3D commercial software, but the deviation from measured result exceeded10 dB at the frequency below 1 GHz. As a result, an accurate andreliable numerical modelling of a conformal shielding structure including ashielding measurement apparatus has not been introduced.

In this Letter, a reliable numericalmodelling for shielding evaluation of on-package conformal shields based on the integrated circuit (IC)-stripline method (IEC 61967-8) is presented. Theeffects of the number of grounding (GND) pads and the thickness of a conformalshield on the shielding performance are investigated by numerical simulation.To validate the simulation results, the conformal shield is fabricated on atest board by spraying technique and measured by both the IC-stripline methodand TEM cell method (IEC 61967-2). Good agreement between the simulation andmeasurement is achieved.

Design of test boards

To evaluate the SE of conformalshields by simulation and measurement, test boards with a conformal shield aredesigned and fabricated. In Fig. 2, a rectangular patch source 10 ×10 mm in size, which represents an IC die on a package, is located in thecentre of a test board. Multiple GND pads are located around the patch sourceon the test board. The patch source is molded by a dielectric material ()30 × 30 mm in size and 400 μm in thickness. A conformal shield made by anAg paste is coated on the molded source and in contact with the GND pads forgrounding. The conductivity of Ag paste is .Two different thicknesses (10 and 25 μm) for the conformal shield areconsidered in this Letter.

Fig 2Configuration of conformal shield ontest board

Fig. 3 shows the fabricated test boardswith and without conformal shield. The test boards have three different numbersof GND pads (12, 20, and 40 pads). The size of each test board is 100 ×100 mm, which is identical to that of TEM cell opening, for TEM cellmeasurement. A rectangular ring pad along the edge of the test boards ispositioned for a contact with the TEM cell case. Each test board is 1 mmin thickness and consists of four layers of FR-4 substrate.For the IC-stripline measurement, two pads to be connected with the ends of theIC-stripline are patterned on both sides of the conformal shield. On the bottomsurface, three ultra-small surface mount coaxial connectors are mounted forconnecting the coaxial cables.

Fig 3Test boards with three different GNDpads

Shielding measurement setup

Fig. 4 shows two apparatuses for shieldingmeasurement based on IEC 61967 standard, which is for radiated emissionmeasurement from ICs. In the TEM cell measurement, electromagnetic emissionsfrom an IC are coupled to the septum of the TEM cell. The distance between theseptum and a device under test (DUT) is 45 mm. The other is the IC-stripline measurement, which allows a very near-field measurement from ICs due to a close distance between a stripline and DUT (typically 6.7 mm).The stripline length projected on the test board is 58 mm. The striplinewas made of a brass plate with 0.5 mm thickness. To measure the radiatedemissions from conformal shields, a vector network analyser (VNA) is used. Oneport of the VNA is connected to the patch source on the test board. The otherport of the VNA is connected to one end of the TEM cell or the IC-stripline.The other end of the TEM cell or the IC-stripline is terminated by 50 Ω. In thecase of TEM cell, we conducted the measurement up to 3 GHz due to thecavity resonance within the TEM cell. However, the IC-stripline measurement wasconducted up to 10 GHz because it has no cavity resonance due to an openstructure.

For numerical analysis, a 3D electromagnetic field simulation using CST Microwave Studio is employed [4]. For the TEM cell measurement, a totalnumber of meshes of about 48 million is required when the conformal shieldis 25 μm in thickness. However, this is difficult to process for normal performance personal computers. For the IC-stripline measurement, a totalnumber of meshes of about 10 million is needed for the same conformalshield. It is just about one-fifth of the TEM cell due to the smaller distancebetween the stripline and DUT. Consequently, for an efficient analysis, the IC-stripline method is employed in simulation.

Comparison of simulation and measurement

Fig. 5 shows a comparison of simulated andmeasured SEs of the conformal shield with 12 GND pads. The simulated result isin good agreement with the measured result. At the frequency below 3 GHz,the TEM cell measurement result corresponds with the simulated result betterthan that of the IC-stripline measurement. At the frequency above 3 GHz,however, the IC-stripline measurement also matches well with the simulatedresult. The SE gradually decreases as the frequency increases up to the firstresonant frequency around 3.4 GHz. However, above 3.4 GHz, the SE isalmost 0 dB, except for some resonant peaks or dips. Compared withFigs. 5a and b, the SEs are almost identical andthe shielding performance does not improve with a thicker conformal shield (25μm). The reason is that the skin depth of Ag is 9.2 μm at 0.1 GHz, whichis <10 μm. Thus, the shielding performance does not increase any more aslong as the thickness of a conformal shield is thicker than its skin depth.

Fig 4Two apparatuses for shieldingmeasurement

Fig 5Simulated and measured SEs ofconformal shield with 12 GND pads

Fig. 6 shows a comparison of simulated andmeasured SEs of conformal shield with 20 GND pads. Compared with Fig. 5, the shielding performance under thefirst resonance frequency is improved by about 10 dB due to the increasednumber of GND pads. However, the accuracy of the measured results by the IC-stripline is worse than that in the case of 12 GND pads.

Fig. 7 shows a comparison of simulated and measured SEs of conformal shield with 40 GND pads. The SE under the firstresonance frequency is improved by 5–10 dB more than that of 20 GND pads.The accuracy of the measured results by the IC-stripline is much worse thanthat in the case of 20 GND pads. Since the IC-stripline is an open structure,it is more likely to interfere with ambient noises as the conformal shield hasa higher SE. From the results in Figs. 5-7, it is found that the conformal shield needs to have at least 40 GND pads in order to accomplish the SE over 30 dB up to 3 GHz.

Fig 6Simulated and measured SEs ofconformal shield with 20 GND pads

Fig 7Simulated and measured SEs ofconformal shield with 40 GND pads

Conclusion

In this Letter, a reliable numerical modelling for shielding evaluation of on-package conformal shields based on the IC-stripline method has been presented. The effects of the number of GND pads and the thickness of a conformal shield on the shielding performance have been examined by numerical modelling and experimental measurements. The simulated results have been validated by both the IC-stripline and TEM cell measurements and good agreement was achieved.

致谢和引用

全文引用自Shielding evaluation of on‐package conformal shields by numerical modelling and experimental measurement - Park - 2017 - Electronics Letters - Wiley Online Library

感谢原作者的付出，所有版权归原作者所有。