Another important consideration concerning cavity resonance is the effect on the shielding performance of the enclosure. Since energy inside the cavity is amplified at the resonant modes, it is likely that the shielding effectiveness (SE) would be lowest at these frequencies. As Fig. 4 shows, the SE drops significantly at each cavity resonance (due to the dimensions of the cavity).

The methods for evaluating the effectiveness of PCB shields can be broken into three categories: compliance testing, functional testing, and indirect testing. Compliance testing involves evaluation of the final product given industry-standard test methods and acceptance levels, such as spurious emissions, susceptibility, or electrostatic discharge (ESD). Functional testing involves evaluation according to performance requirements set by the manufacturer. Intercavity shielding, radiation from the antenna back into the receiver (Rx), and phase noise are examples of parameters that would be considered.

The last category, indirect testing, is used when describing shielding products. Methods such MIL G 83528B, ASTM D 4935, and ASTM D 991 are guidelines followed by shielding manufacturers to evaluate their products. These methods can characterize the constituent properties of a shield, but cannot predict the performance in a specific application.

Typical surface-mount-technology (SMT) components suffer minimal levels of radiation, although these levels are not negligible. Predicting the radiation response, particularly in the near field, of typical SMT devices is a formidable task. A first step is to consider the coupling between two microstrip lines that would be used to connect components on the PCB. Short transmission lines, even when close together, exhibit very little coupling. This characteristic can help to develop an inefficient radiator that can be used as a test vehicle (Fig. 5).

Since the dielectric substrate thickness on the outer layers of the PCB is usually very thin, it is desirable to have the radiating element be a separate component, instead of integrating it onto the PCB. The height of a separate radiator can be readily adjusted, while maintaining the height sufficiently lower than the inside height of the shield. By using the remaining dimensions of the transmission-line radiating element (height, length, width, and termination impedance), the response of the radiator can be optimized for testing shields on PCBs.

The special test fixture developed for testing shields on PCBs (Fig. 6) includes a custom-designed SMA connector that, using a solder pre-form, completely seals the launch point to the ground plane of the PCB. The connectors then feed the radiating elements that are attached to the opposite side of the PCB. The shields would then be centered and attached over each of the elements.

The mode-stirred reverberation chamber technique (Fig. 7) is an excellent EMI test method because of its high dynamic range and repeatability.^4,5 In this technique, the radiation characteristic from the device under test (DUT) is compared to that of a reference horn antenna. Measurements are first performed on the horn antenna, then the DUT is substituted for the horn and tested. Of course, only one radiator/cavity can be tested at a time. The radiator is first measured without a shield over it, then the shield is attached and the device is re-measured. The SE is calculated as the difference between the received power levels (in decibels) before and after the shield is applied.

The frequency range of such a test is determined by the room dimensions, the test equipment, and the antenna bandwidth. The main limitation is usually the lower frequency boundary, determined by the room size and antenna used. A frequency range of 1 to 13 GHz was used for the tests in this article.

Practically, EMI enclosure cannot be made to be a complete Faraday shield. Gaps due to perforations in the shields, incomplete shields, breaks in the shielding gasket, spaces between grounding vias, and relief areas in the ground plane are necessary to manufacture the complete PCB. But as long as the size of the aperture is much smaller than the wavelength of the highest frequency of interest, it should not cause an appreciable amount of leakage.

The effects of apertures much smaller than a wavelength at the highest operating frequency of interest has been studied to great lengths.^9-11 For perforated screens, the formula of Fig. 8 has been used to show the frequency relationship between the size of the aperture and SE. Although this frequency dependence represents an accurate relationship for the far-field response of a large array of holes, an offset factor in SE can deviate quite a bit from real-life applications. To overcome this, a relationship for the far-field (plane-wave) response was derived using empirical data taken from thin copper sheets perforated using the traditional hexagonal pattern. Initially, 1-mm-diameter holes were placed on a 1.7-mm hexagonal grid, such that about 1552 holes fell within the annulus of a typical ASTM D 4935 test cell. The SE was obtained using S₂₁ measurements with and without the sheets in place (Fig. 8).

A simple model was generated to represent this test pattern, using the formula of Fig. 8 and a correction offset factor. The next three test patterns were generated by doubling the spacing between the holes each time. This way, about four times fewer holes fell within the annulus of the coaxial cell for each pattern. To generate the modeled data, the original model was changed by a factor of 4 each time, which yielded 12-dB offsets. As Fig. 8 shows, the frequency relationship follows what would be expected for plane-wave excitation of perforated thin sheets.