Universal Serial Bus Signal Integrity Analysis for High - Speed Signaling

The Universal Serial Bus transmits differential data over twisted-pair cables originally designed for signaling at 12Mbit/s. Because of the recent specification of a 480Mbit/s signaling speed with the existing cabling, the high frequency behavior and temporal response of these cables is of interest. In this work, twoand threedimensional field solvers are used to characterize the field behavior within the cables. The results are used to perform SPICE simulations of complete time-domain USB packets, which indicate that existing cables support the higher bit rates well, and that the quality of signaling is not limited by the cable construction. The simulations match well with the limited measurement data available for 480Mbit/s signaling. 1. Universal Serial Bus Cables The Universal Serial Bus (USB) is a general-purpose technology for connecting peripherals to personal computers. A twisted pair carries differential signals at signal rates of 1.5, 12, and 480 Mbits/s. Additionally, transitions on one member of the pair are used to signify ends of packets at all signal rates and starts of packets at 480 Mbits/s. The presence of both commonand differentialmode signals makes impedance control involved. The presence of a power pair, which eliminates the need for power supplies for USB peripherals, further complicates the situation [1]. Figure 1 shows the USB cross section. The six conductors i.e. signal lines D and D, power pair Vbus and ground, drain, and shield are shown, along with the insulations polyvinyl chloride (PVC) and highdensity polyethylene (HDPE). The outer dimension of the shield is approximately 3mm. The power pair is 20 American wire gauge (AWG), the signal pair is 20-28 AWG, and the drain wire is 28 AWG. The signal lines D and D can be in an unbalanced configuration, owing to their twisted pair rotation along the cable. USB cabling has been shipping in high volume since 1998, and hence it is important to estimate signal integrity in existing USB cables, developed for 1.5 Mbit/s and 12 Mbit/s signaling in 1995, for the new specification of 480 Mbit/s adopted in 2000 [2]. This necessitates the use of electromagnetic and SPICE solvers. Not many signal integrity results for USB systems are available in the public literature, and it is expected that the simulation approach and results presented here will enable the signal integrity analysis of other existing and future high-speed hardware connector schemes [3]. 2. Electromagnetic and Circuit Simulation of USB Cables The simulation procedure uses twoand threedimensional signal integrity software [4] made available by Ansoft Corporation for use in graduate course teaching and projects [5] under the EleXIR (Electromagnetic Simulation Instruction and Research) effort by the second author. Ansoft’s two-dimensional transmission line solver was used to compute differentialand commonmode impedances as a function of frequency. Two models, a simplified single-conductor model, and a more complex stranded conductor model, were used. Figure 2 shows the impedances for the two models in a balanced configuration. Similar simulations were run for rotated signal pairs to obtain average impedances. Also of interest is the propagation loss and velocity in the cable. This is shown for two cable rotations and for both models in Figure 3, along with the conservative USB specification limit, which is easily achieved by the cable model. The propagation velocity was modeled using an average of velocities of eigenmodes weighted by the relative strength of the eigenmode in a typical USB signal packet. The power spectral density of a USB signal has been studied in the past [6]. The resulting velocity is plotted in Figure 5, and is close to measured velocities that are observed to be near 2.1 X 10 m/s. For a three-dimensional analysis of the USB cable, Ansoft’s Eminence software was used to simulate small sections of the cable including signal pair rotations that were linked together to obtain a SPICE-level representation. Figure 6 shows USB signal outputs from the simulation. Arrows 1 and 2 show transition errors that occur because the differential parts of the single ended transitions start before the slower common-mode components arrive. The differential power drives a voltage into the victim line until the common-mode signal arrives to correct the output. Another effect, shown by Arrow 3, is the upward shift of the first few byte of the packet, which occurs due to an impedance mismatch between common-mode terminations and the cable’s common-mode impedance. These effects are also visible in a measured USB output signal, as shown in Figure 7. 3. Conclusions The USB simulation based on electromagnetic models and SPICE simulation was able to predict the time-domain effects present in measured data. Also, the cable model studied was able to function with 480 Mbit/s signaling and this is an indicator that existing cabling can handle the faster bit rate. More extensive simulation would require more accurate cable models, and full-wave threedimensional electromagnetic simulation of an electrically long cable. It is expected that similar simulations will benefit the signal integrity studies of other high-speed connection schemes and will enhance public domain information.