End USB Noise Without Stopping the Bus

As universal serial bus (USB) 1.1 becomes stable ground for developers, the introduction of USB 2.0 makes the ground uncertain again. The new specification calls for operation up to 40 times faster than USB 1.1, resulting in speeds of 480 Mb/s. With the major goal of USB still being to replace the traditional ports of the PC, it now has the speed to guarantee that it will be around for a while.

The USB primarily is a PC bus that can be readily applied to any device that would use an RS-232 or parallel port. Just about every PC on the market supports at least one USB port. USB is more flexible and more internally complicated than the interfaces that it replaces.

USB 1.1 supported two bus speeds: 1.5 Mb/s and 12 Mb/s. Low-speed applications are interactive devices such as a keyboard, a mouse, and other low-end peripherals. Full speed is designed more for phone, audio, and compressed video.

Now with USB 2.0 high-speed performance, bandwidth-hungry applications such as video and storage devices are easily connected to a PC. USB 2.0 can support higher speed applications such as high-density storage devices, high-resolution scanners, and printers.

As with 1.1, the host still is in charge, allowing peripherals to be simpler and cost effective. Table 1 (see below) describes the relationship for the range of data traffic workloads that USB can handle.

With the increase in speed comes the growing concern for EMI, where high-speed signal lines are notorious for generating noise. USB 1.1 and 2.0 use differential signaling because it is less prone to generating noise.

Ideally, the magnetic flux produced by the differential signal currents has a cancellation effect and, for that reason, is not a large contributor of EMI noise. However, harmonic noise from IC drivers or peripheral circuits can radiate onto the signal lines and be transmitted through the USB cable, which can act like an antenna.

Signal coupling by parasitic inductance or capacitance can cause common-mode currents. The common-mode currents are the dominant source of EMI problems for high-speed interfaces. As shown in Figure 1, all common-mode current flows in the same direction so the magnetic flux is not canceled as with differential-mode currents.

USB 2.0 presents new challenges to the design community because of its higher frequency band. The typical USB 1.1 EMI filtering solution consisted of ferrite beads on the power and ground along with a ferrite bead on each signal line to reduce the level of noise.

In USB 1.1, the signal frequency band and the noise frequency band are far enough apart that the ferrite bead can reduce the noise and not affect the signal. The higher speed of USB 2.0 makes this solution ineffective because the signal and noise frequency bands are close together or possibly occupy the same frequency band. To use the same ferrite-bead filtering solution for USB 2.0 would suppress both the noise and signal frequencies.

An effective way to provide the required noise attenuation includes a common-mode choke coil that reduces the noise while maintaining the integrity of the signal. As displayed in Figure 1, the magnetic flux created by the common-mode current increases and produces an impedance while the magnetic flux from the differential-mode current cancels, resulting in no impedance.

The common-mode choke coil must have two key features to make it an effective solution: a high common-mode impedance and a high magnetic coupling factor. The effective common-mode impedance provides a higher level of common-mode noise attenuation.

The test setup shown in Figure 2 was constructed to evaluate the effect of the common-mode choke coil. Testing was performed in an anechoic chamber with the USB Controllers separated by a distance of 2 meters. The noise radiation was measured using the standard 3-meter test method. Initial readings were measured and recorded.

A 90-W common-mode choke coil (Murata part No. DLW21SN900SQ2) was tested. The data presented in Figure 3 shows that the 90-W common-mode choke coil reduced the noise level by 6 dB at the peak levels. A 260-W common-mode choke coil was tested next and reduced noise levels by 9 dB at the peak levels. The signal integrity (eye pattern) was minimally affected by the insertion of either filter.

This is a case where more may be too much. The common-mode impedance should not be too high; otherwise, it might distort an end-of-packet (EOP) waveform associated with USB 1.1 due to an inductive voltage effect.

Put simply, the EOP signal in both USB 1.1 and 2.0 allows the receiver to know when the packet is complete. If the receiver does not receive the EOP, then the data will run together and become corrupted. Selection of an effective common-mode impedance value is key to preventing EOP waveform distortion in USB 1.1. USB 1.1 and USB 2.0 each use different encoding of the EOP signal. However, since USB 2.0 initially connects as USB 1.1, it will need to comply to the USB 1.1 EOP signal standards.

As illustrated in Figure 3, the 90-W choke coil has caused minimal distortion of the signal as compared to the 260-W part. Even though the 260-W part creates greater noise suppression, it distorts the EOP signal of USB 1.1. The 90-W coil has a high coupling coefficient to provide minimal distortion of the differential signal while providing common-mode noise suppression. The common-mode impedance is such that the EOP of USB 1.1 is not distorted. Once the USB 2.0 connection is established, the distortion of the EOP signal is not a concern.

USB 2.0 offers greater speed improvement to meet higher capacity needs of consumers. With the increase in speed, the design emphasis moves more in the direction of EMI filtering. The EMI solution for USB 2.0 isn’t as simple as for USB 1.1; the ultimate solution is a common-mode choke coil that can reduce the noise without attenuating the signal.

Selection of the correct common-mode choke coil is the key to reducing noise, maintaining high-speed signal quality, and preserving the EOP signal. After testing and reviewing the data, the 90-W common-mode choke coil proves to be the most effective choice. Since USB 2.0 is projected to become the standard interface for most PCs, the need for effective and efficient filtering of EMI should be addressed and planned for early in the design phase.

About the Authors

Jonathan Davis is a product engineer for the EMI filter and chip inductor group at Murata Electronics. He recently joined Murata after completing his bachelor’s degree in computer engineering technology at Southern Polytechnic State University.

Gerry Hubers is the market segment manager at Murata Electronics North America. Since joining Murata in 1990, Mr. Hubers held positions in product management and product marketing. He has a technical background in electronics engineering from Radio College of Canada.

Deryl Kimbro serves as a product manager for Murata Electronics North America. Mr. Kimbro earned a B.S. in electrical engineering technology from Southern College of Technology in 1995 and has been employed by Murata since 1997.

Shuji Mikami joined the company in 1987 as an EMI engineer and today is group product manager for Murata Electronics North America. He became the EMI product manager in 1990 in Murata Europe and moved to North America in 1999. Mr. Mikami received his bachelor’s degree in electrical engineering from Fukui University of Japan.

Murata Electronics North America, 2200 Lake Park Dr., Smyrna, GA 30080-7883, 770-436-1300

Return to EE Home Page

Published by EE-Evaluation Engineering
All contents © 2002 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.

June 2002