USB Interfaces and EMC

   The exposed connector pins on universal serial bus (USB) receptacles may have an effect on system-level EMC performance. To control this effect, a pin shield is typically employed. Our goal in the following test was to assess the actual impact of a pin shield on USB connector receptacles and EMC. In this component level test, USB cables and connectors were excited using two different excitation techniques. We also explored the effect of shield termination.

Summary

   Cost-effective EMC design requires the focusing of EMC suppression features on dominant radiation mechanisms. In our opinion, second-order features, which either do not directly affect EMC performance or affect it in a minor way, generally should be considered only if they cost little or nothing. It is with this perspective that we evaluated the USB receptacle pin shields.

   Our test results show that the pin shield has, in fact, very little effect on the radiated EMC performance in a typical USB interconnect. The impact of this pin shield is minor compared to the shield termination technique. Radiation levels were highest when a USB cable was attached; direct radiation from the aperture in the panel cutout is approximately 30 dB lower without the cable. Aperture radiation is clearly a second-order effect compared to cable radiation. Shield termination of the USB receptacle to chassis showed a 20 dB improvement compared to shield termination via the PCB ground.

   Cable radiation is dominated by the method used to terminate the shield. If the USB receptacle is directly connected to chassis ground, the effect of the pin shield cannot be discerned (within a measurement error of several dB). If, however, the USB receptacle relies solely on a PCB connection for shield termination, the pin shield appears to alter the transfer impedance of the cable shield to chassis ground. It is not clear whether it improves or degrades the performance, but generally it has very little effect. We believe that the PCB layout for the USB receptacle connection to chassis will have a far greater impact on EMC performance than the pin shield itself. We make this assertion on the assumption that the inductance of the PCB planes, vias, and traces to chassis ground will be greater than the connector shield inductance.

Test Setup

   The general test setup is shown in Figure 1. The general strategy is to excite the USB cable and enclosure in a representative fashion, then use the mode-stirred technique to measure the radiated power. This approach yields repeatable results because the mode-stirred technique has the property of time-averaged uniform power density. This test technique is commonly used within AMP Inc., the National Institute of Standards and Technology, and the Naval Surface Warfare Center to assess the shielding performance of devices.

   The frequency range for this test was 50 MHz to 1 Ghz. Given the size of the mode-stirred chamber, sufficient modal density will not exist at frequencies below about 200 MHz. In the 50–200 MHz range, it is important for the tested device to be kept in the same location and orientation to make a valid relative comparison. For all testing herein, this practice of maintaining cable orientation between test configuration changes was strictly followed.

   The detailed setup is represented in Figure 2, which shows the signal excitation going to a radiator (patch antenna) and the hybrid coupler. In practice, either the antenna was driven or the signal conductors were driven, but not both. The antenna mimics the scenario whereby internal circuitry drives the USB PCB ground plane that in turn drives the USB cable shield, which then radiates.

   Initially the USB receptacle shields did not contact the chassis of the brass enclosure. Instead, the cable shield termination relied on the connection from the USB receptacle shield to the PCB ground. For this testing, the PCB ground plane was connected to chassis ground by using a 1-in.-wide right-angle bracket. In actual production hardware, it is unlikely that inductance of the connection between signal ground and chassis ground would be as low as that achieved in this test setup.

   An additional test was performed whereby the USB receptacle shield was connected to chassis with a short piece of solder wick (1/4-in.-wide braid, 1/8 in. long on two sides of the receptacle). In this way the cable shield is tied to chassis at the USB receptacle through a low-inductance connection (though not a circumferential shield termination).


   
The USB specification permits the connection of the USB shield to chassis ground at the host end only. We made no attempt to use discrete capacitors for shield termination, a technique sometimes used to connect signal ground to chassis ground at high frequencies but maintain low-frequency isolation. The D+/D– signals were excited differentially using a hybrid coupler. The power and return lines in the USB cable were terminated to ground using 50-W resistors.

   To mimic a pin shield, copper tape was soldered to the USB receptacle on three sides (the top and two sides, as shown in Figure 3). This essentially formed a box over the exposed pins. The connection from the USB receptacle to PCB ground was not altered by the addition of the pin shield.



   Photographs of the test setups were taken as part of the documentation process. These photographs and their descriptions are shown in Figures 3, 4, and 5.

Test Results

   The test results are shown in Figures 6–9. Figure 6 shows the effect of aperture radiation from the shielded enclosure when the USB cable is not attached. The excitation source was the patch antenna inside the brass enclosure that drives the PCB ground plane. The ground plane is connected to the USB receptacles, which are the likely RF source in the absence of a cable. We base this statement on the size of the apertures and the recognition that the USB receptacle acts to a minor degree as a conductor coupling RF energy from inside the resonant cavity to the outside. The data show that the added pin shield affects the data but does not reduce RF emissions overall. (In Figures 6–9, "closed recept shld" means that the pin shield was added, and "open recept shld" means there was no pin shield.)

    The data shown in Figure 7 is similar to that in Figure 6 in that a USB cable was not inserted into the USB receptacle. The difference is that instead of driving the patch antenna in the brass enclosure, the D+/D- signal pins in the receptacle were excited with the hybrid coupler. The intent was to mimic a USB signal, or noise on USB signals, exciting the USB receptacle. The test result is again similar; no appreciable improvement is achieved by adding the pin shield.

   The data shown in Figure 8 were recorded with the same test conditions as described for Figure 6, except that a USB cable was plugged into the USB receptacles. Note that the radiated levels have increased by some 30 dB with the same signal drive level. Clearly the addition of the cable (or EMI antenna) plays a major role in the radiated emission levels. In general, the added pin shield did not play a major role in reducing the radiated levels, with the exception of two resonances, at 496 MHz and 791 MHz. These resonances are suspicious in that at all other frequencies the pin shield makes very little difference. We believe that these two resonances are related to cavity resonances of the brass enclosure.

   The data shown in Figure 9 were recorded under the same test conditions as described for Figure 7, except again that a USB cable was plugged into the USB receptacles. The test results are likewise similar; the pin shield has a relatively minor effect compared to the addition of the cable. Comparing Figures 7 and 9 shows that the radiated power levels increased roughly 30 dB with the addition of the USB cable (the EMI antenna).

   The next step in the test sequence was to mimic a connection of the USB receptacle to chassis ground at the panel cutout for the connector. Basic EMC design principles would suggest that the decreased inductance using this shield termination process would greatly lower the radiated power. The data shown in Figure 10 were recorded under the exact same test conditions as those shown in Figure 8. In general the radiated power was reduced by roughly 20 dB, and the pin shield had very little effect on the radiated power.

Conclusion

   The common perception that a shielded connector is required to seal apertures caused by connectors in a chassis is an oversimplification of the radiated phenomena. The experiments we conducted clearly show that a pin shield does not reduce radiation from a USB interface. The purpose of these tests was to show that as designed, the USB interface does not provide lower radiated emissions when a more costly pin shield is added to the USB connector.

   Jim Nadolny has been with AMP Inc. (Harrisburg, PA) for the past six years as an EMC specialist with an emphasis on EMC modeling and research. Nadolny has an MSEE from the University of New Mexico and a BSEE from the University of Connecticut. Kieran Kelly is a test engineer at AMP Inc.'s corporate test lab, and specializes in signal transmission quality and shielding performance test systems, measurement procedures, and product enhancements. He graduated from Pennsylvania State University in 1991 with a BSEE, and is an IEEE EMC Society member. Kelly can be contacted at kpkelly@amp.com.

Back to July/August Table of Contents