High-speed Interconnects Research Articles

EPEEC: comprehensive SPICE-compatible reluctance extraction for high-speed interconnects above lossy multilayer substrates

With continuous advances in radio-frequency (RF) mixed-signal very large scale integration (VLSI) technology, the creation of eddy currents in lossy multilayer substrates has made the already complicated interconnect analysis and modeling issue more challenging. To account for substrate losses, traditional electromagnetic methods are often computationally prohibitive for today's VLSI geometries. In this paper, an accurate and efficient interconnect modeling approach-the eddy-current-aware partial equivalent element circuit (EPEEC)-is proposed. Based on complex image theory, it extends the traditional partial equivalent element circuit (PEEC) model to simultaneously take multilayer substrate eddy-current losses and frequency-dependent effects into consideration. To accommodate even larger scale on-chip interconnect networks, EPEEC develops a new simulation program with integrated circuit emphasis (SPICE)-compatible reluctance extraction algorithm by applying sparsification in the inverse inductance domain with an extended window algorithm. Compared with several industry standard inductance and full-wave solvers, such as FastHenry and Sonnet, EPEEC demonstrates within 1.5% accuracy while providing over 100/spl times/ speedup.

Crosstalk Bounded Uncorrelated Jitter (BUJ) for High-Speed Interconnects

Bounded uncorrelated jitter (BUJ), a subcomponent of total jitter, is commonly caused by crosstalk coupling from adjacent interconnects on printed circuit boards (PCB). However, the characteristics of BUJ are still not well understood. Neither a mathematical model of jitter, nor an algorithm to generate histograms for BUJ has been developed to this date. Such a model and algorithm would empower designers to predict BUJ to achieve total jitter budget without lengthy simulations and measurements. In this paper, we first review the characteristics of a crosstalk pulse induced by an aggressor signal on a quiet trace. Then, by applying the superposition principle, a jitter model to calculate the time difference between the distortion-free and the distorted edge crossings of the victim signal was developed. This model is also extended to calculate the worst case timing difference, BUJ/sub p-p/. In addition, an algorithm to generate the histogram distribution of BUJ is also developed. The developed algorithm has fast execution times of 10-20 s, compared to simulation and measurement times of 10-30 min.

Beyond the Bench: Virtual School

In today’s world of high-speed interconnection, technology in the classroom helps keep students interested and engaged in the learning process. Taking advantage of this favorable avenue of instructional opportunity, the Community Education and Outreach Program (COEP) of the NIEHS Center in Molecular Toxicology at Vanderbilt University, in conjunction with the university’s Center for Science Outreach (CSO), has developed an innovative interactive videoconference teaching program known as “Virtual Scientist in the Classroom.” The program creates a direct connection between Vanderbilt University faculty and students all over the country, allowing university researchers to lecture on environmental health topics related to the work they are performing in their own laboratories. “Through the center’s involvement with outreach and education, we are able to provide reliable, up-to-date, and cutting-edge science to classrooms throughout Tennessee and the U.S.,” says Bradley Hawkins, the COEP director. “In addition, the students are able to interact with our researchers in a manner that was not available just a few years ago.” The program relies on volunteer faculty with diverse research interests—neuroscience, diabetes mellitus, biomedical engineering, physics, molecular toxicology, and chemistry, for example—who create their own presentations and conduct the sessions in real time in the CSO virtual learning studio (all presentations are taped and archived for future multi-classroom sessions). The topics for presentations to date have included how medicines are developed, how chemicals damage DNA, and the importance of micronutrients. The format of each session is left to the discretion of the expert presenters, and may include anything from PowerPoint slides to movie clips, live virtual tours of lab facilities, even real-time electrocardiograms. One physics professor presented the theory of relativity in character as Albert Einstein. The sessions of 30–45 minutes can be presented to one school at a time or to multiple site audiences. Scientist–student interaction is a main component of the sessions; questions and feedback from students are expected and encouraged. By using a communications bridge capable of connecting to multiple sites within a videoconference session, the researchers open up the world of scientific discovery to students in classrooms all over the state of Tennessee and beyond. When the program was created in 1999, it primarily reached out to middle and high school students in Tennessee, but has grown to include videoconferences to children in 75 schools in 20 states. Typically four to six sessions on varied topics are offered each semester. Teachers can find complete descriptions of each session online at http://www.vanderbilt.edu/cso/ and can register there for each session. The sessions are free for Tennessee students, although a charge is applied for out-of-state schools. Once teachers have registered, they can download supplemental lesson material and will receive e-mailed confirmation and detailed instructions for participation. In a new feature, researchers answer questions that arise after each session, and the 1- to 2-minute video response also is archived on the site. The continued commitment and enthusiasm of the contributing faculty members is a cornerstone of the program, and helps keep the videoconference sessions relevant and timely. “I believe that as researchers we need to take an active role in helping to educate and inform the public about issues related to adverse health outcomes upon exposure to poisons, to educate the public about sources of poisons in food and air, and the mechanisms by which they affect our health,” says Michael Aschner, a professor of pediatrics and pharmacology who has presented on the subject of chemical insults to the brain. Hopefully, he says, educational outreach programs can help bridge the gap between public understanding and public perception of toxicology.

The need for speed [Internet system architecture

The exponential growth of the Internet has brought about a need for a proportional increase in the speed on interconnect. To deal with the need for greater speed, it is no longer feasible to make interconnect wider and faster because of the high pin count per device, tighter margins for skew on the bus, and reliable high-speed interconnect as a logical step. In this paper, the author argues that the growth of the Internet is changing the landscape of the system architecture. The Infinitband, Rapid Fabric, and advanced switching interconnect standards discussed are different in terms of choice of features and the trade-off between simplicity and flexibility. Interconnects that can best use the industry's economies of scale are likely to find greater acceptability and usage.

A study of a hybrid phase-pole macromodel for transient simulation of complex interconnects structures

An overview of standard macromodeling techniques (i.e., employ poles but no phase shifts) for transient simulation of high-speed interconnects is first presented. Then, the limitations of these standard macromodeling techniques (e.g., high model order and slow convergence) are discussed. In order to overcome these limitations, generalized method of characteristics (MoC) techniques include the physical phenomenology of phase shift (time delay) in addition to the system poles, thereby making it possible to model single and coupled transmission lines using far fewer terms than when standard macromodeling techniques are employed. Since MoC techniques incorporate the time delay into the model, causality is also guaranteed. In this paper, the MoC idea is extended by developing a hybrid phase-pole macromodel (HPPM) for the modeling of more complex interconnects with embedded discontinuities. Unlike other generalized MoC techniques that have only been applied to single and coupled transmission lines, the HPPM macromodel can be applied to larger portions of the system that contain multiple cascaded transmission lines and discontinuities. The HPPM parameters can be extracted from either measured or simulated transient data. Comparisons between a standard macromodel and the HPPM show that the HPPM has significant advantages in terms of reduced macromodel orders.

Worst case crosstalk noise for nonswitching victims in high-speed buses

Considering a RLC interconnect model, we determine switching patterns and switching times of multiple aggressors to generate the worst case crosstalk noise (WCN) for a quiet or a noisy victim. We consider the routing direction as it has a significant impact under the RLC model. When there are no timing window constraints, we show that the commonly used superposition algorithm results in 15% underestimation on average, and propose a new SS + AS algorithm that has virtually the same complexity as the superposition algorithm but has a much improved accuracy. On average, the SS + AS algorithm only underestimates WCN by 3% compared to time-consuming simulated annealing and genetic algorithm. We also show that applying a RC model to the high-speed interconnects in the International Technology Roadmap for Semiconductors 0.10 /spl mu/m technology virtually always underestimates WCN, and the underestimation can be up to 80%. Furthermore, we extend our algorithm to consider aggressor switching and victim sampling windows. We show that the extended SS + AS algorithm well approximates WCN with 2% underestimation on average. Although the RC model usually severely underestimates WCN with timing window constraints, it does overestimate when both the aggressor switching and the victim sampling windows are small enough. We conclude that the RLC model is needed for accurate modeling of WCN in design in the multigigahertz region.

Global passivity enforcement algorithm for macromodels of interconnect subnetworks characterized by tabulated data

With the continually increasing operating frequencies, complex high-speed interconnect and package modules require characterization based on measured/simulated data. Several algorithms were recently suggested for macromodeling such types of data to enable unified transient analysis in the presence of external network elements. One of the critical issues involved here is the passivity violations associated with the computed macromodel. To address this issue, a new passivity enforcement algorithm is presented in this paper. The proposed method adopts a global approach for passivity enforcement by ensuring that the passivity correction at a certain region does not introduce new passivity violations at other parts of the frequency spectrum. It also provides an error estimate for the response of the passivity corrected macromodel.

Piecewise linear model for transmission line with capacitive loading and ramp input

Transmission line effects become increasingly significant for on-chip high-speed interconnects. Efficient and accurate transmission line models are required for analysis and synthesis of such interconnects. In this paper, we first present an efficient model for the far-end response of a single transmission line considering ramp input and capacitive loading. Our model divides the time axis into a number of regions according to the time of flight and the input rising time, and then approximates the far-end response by piecewise linear (PWL) waveform in each region. We name the resulting model as the PWL model. Experiments show that the waveform from the PWL model differs from the SPICE simulation result with the average voltage difference less than 0.9% V/sub dd/, and the PWL model is at least 1000/spl times/ faster than SPICE simulation. We further apply the PWL model to calculate the delay, rising time, and oscillation amplitude of the coplanar waveguide structure, and achieve less than 10% average error compared to SPICE simulation. Combining the PWL model and decoupling technique, we analyze the far-end response of bus structures and obtain waveform almost perfectly matching the SPICE simulation result.

Parametric macromodels of differential drivers and receivers

This paper addresses the modeling of differential drivers and receivers for the analog simulation of high-speed interconnection systems. The proposed models are based on mathematical expressions, whose parameters can be estimated from the transient responses of the modeled devices. The advantages of this macromodeling approach are: improved accuracy with respect to models based on simplified equivalent circuits of devices; improved numerical efficiency with respect to detailed transistor-level models of devices; hiding of the internal structure of devices; straightforward circuit interpretation; or implementations in analog mixed-signal simulators. The proposed methodology is demonstrated on example devices and is applied to the prediction of transient waveforms and eye diagrams of a typical low-voltage differential signaling (LVDS) data link.

Twisted Differential Transmission Line Structure for Global Interconnect in Si LSI

This paper investigates the twisted differential transmission line structure to achieve high-speed signal transmission and electromagnetic interference (EMI) noise reduction of global interconnects in Si LSIs. The differential transmission line in Si LSIs can transmit a 12 Gbps pulse signal and has the capability of reducing EMI noise. The proposed twisted-diagonal-pair line provides high-speed, low-EMI-noise, high-crosstalk-robustness and high-density global interconnects in Si LSIs.

Effects of the Parasitics on the Time Response of RCE-PDs

Resonant cavity enhanced photodetectors (RCE-PDs) are promising candidates for applications in high-speed optical communications and interconnections. However, because parasitics effects of these high-speed photodetectors can significantly degrade the performance of the photodetector, then they must be carefully considered. Here, an accurate model for the time response of the RCE-PDs that includes the effects of their parasitics is presented. The effects of an inductor that may be added in series with the load resistor are also studied and it is shown that the external inductor can improve the performance of the photodetector because it compensates some of the degradations resulting from capacitive parasitics. The effects of the parasitics have been investigated for different dimensions of the photodetectors, different values of both the load resistance and the added inductor and also for different multiplication gains for the case of RCE-avalanche photodetectors.

Finite-difference time-domain macromodel for simulation of electromagnetic interference at high-speed interconnects

This paper presents an efficient and systematic approach for transient analysis of hybrid interconnect systems. The approach uses the finite-difference time-domain (FDTD) method to fully characterize the subnetwork of the high-speed interconnects in the form of admittance parameters. It then uses the admittance parameters to construct a macromodel of the subnetwork by the vector fitting method (VFM). The resulting macromodel is ready to be synthesized into a SPICE-compatible circuit simulator to efficiently expedite the transient analysis of the hybrid system with interconnects and linear/nonlinear lumped circuit elements. Numerical examples show the validity of the method.

Two Dimensional Electric Field Measurement in Microstrip-Line Circuits by Coaxial-Type Probe

A detailed investigation of the electromagnetic field distributions in high frequency printed circuits and high-speed interconnects is very useful for physical understanding, studies of electromagnetic coupling effects for EMC and EMI and for optimization of electromagnetic circuit designs. The aim of this paper is to show how to measure the electric field distributions in electromagnetic circuits. An electromagnetic analysis for microstrip-line circuits is carried out by using a finite-difference time domain technique and its measurement is carried out by using a small probe antenna. The measurement results are in fairly good agreement with those of the numerical analysis using the FDTD method. Thus, the measurement system offers a valid means for predictions in the theoretical analysis of more complicated discontinuity problems.

REAL TIME COMPUTING IN A HIGH PERFORMANCE CLUSTER.

Nowadays, cluster computing is a common strategy to achieve high performance computing and resolve computationally intensive tasks such as flight simulation, medical imaging, etc. Two powerful tools have been used to facilitate the development of cluster oriented applications, namely the message passing interface (MPI) and the parallel virtual machine (PVM). However, there are situations in which the application execution is time constrained, e.g., complex real-time image processing, but neither MPI or PVM take time constraints in consideration. Therefore, this paper proposes a cluster integration methodology that is amenable to real-time analysis and implementation, based on Linux with improved temporal resolution and a high-speed TDMA interconnection bus built on top of SCI.The main contribution of the paper is a worst-case response time analysis, its characterization using randomly generated task sets and its verification using a system simulator. The proposed analysis can be applied to a class of distributed real-time systems in which a set of nodes, running tasks under fixed priorities, communicate over a TDMA bus, but it is overly pessimistic in the presence of a significant percentage of communicating tasks.

Design of a High-Speed Optical Interconnect for Scalable Shared-Memory Multiprocessors

The paper proposes a highly connected optical interconnect based architecture that maximizes the channel availability for future scalable parallel computers, such as distributed shared memory (DSM) multiprocessors and cluster networks. As the system size increases, various messages (requests, responses and acknowledgments) increase in the network resulting in contention. This results in increasing the remote memory access latency and significantly affects the performance of these parallel computers. As a solution, we propose an architecture called RAPID (reconfigurable and scalable all-photonic interconnect for distributed-shared memory), that provides low remote memory access latency by providing fast and efficient unicast, multicast and broadcast capabilities using a combination of aggressively designed WDM, TDM and SDM techniques. We evaluated RAPID based on network characteristics and by simulation using synthetic traffic workloads and compared it against other networks such as electrical ring, torus, mesh and hypercube networks. We found that RAPID outperforms all networks and satisfies most of the requirements of parallel computer design such as low latency, high bandwidth, high connectivity, and easy scalability.

Evaluating InfiniBand performance with PCI Express

The InfiniBand architecture is an industry standard that offers low latency and high bandwidth as well as advanced features such as remote direct memory access (RDMA), atomic operations, multicast, and quality of service. InfiniBand products can achieve a latency of several microseconds for small messages and a bandwidth of 700 to 900 Mbytes/s. As a result, it is becoming increasingly popular as a high-speed interconnect technology for building high-performance clusters. The Peripheral Component Interconnect (PCI) has been the standard local-I/O-bus technology for the last 10 years. However, more applications require lower latency and higher bandwidth than what a PCI bus can provide. As an extension, PCI-X offers higher peak performance and efficiency. InfiniBand host channel adapters (HCAs) with PCI Express achieve 20 to 30 percent lower latency for small messages compared with HCAs using 64-bit, 133-MHz PCI-X interfaces. PCI Express also improves performance at the MPI level, achieving a latency of 4.1μs for small messages. It can also improve MPI collective communication and bandwidth-bound MPI application performance.

Twisted differential line structure on high-speed printed circuit boards to reduce crosstalk and radiated emission

Differential signaling has become a popular choice for high-speed digital interconnection schemes on printed circuit boards (PCBs), offering superior immunity to crosstalk and external noise. However, conventional differential lines on PCBs still have unsolved problems, such as crosstalk and radiated emission. When more than two differential pairs run in parallel, a line is coupled to the line adjacent to it because all the lines are parallel in a fixed order. Accordingly, the two lines that constitute a differential pair are subject to the differential-mode crosstalk that cannot be canceled out by virtue of the differential signaling. To overcome this, we propose a twisted differential line (TDL) structure on a high-speed multilayer PCB by using a concept similar to a twisted pair in a cable interconnection. It has been successfully demonstrated by measurement and simulation that the TDL is subject to much lower crosstalk and achieves a 13-dB suppression of radiated emission, even when supporting a 3-Gb/s data rate.

D-SPTF

Distributed Shortest-Positioning Time First (D-SPTF) is a request distribution protocol for decentralized systems of storage servers. D-SPTF exploits high-speed interconnects to dynamically select which server, among those with a replica, should service each read request. In doing so, it simultaneously balances load, exploits the aggregate cache capacity, and reduces positioning times for cache misses. For network latencies expected in storage clusters (e.g., 10--200μs), D-SPTF performs as well as would a hypothetical centralized system with the same collection of CPU, cache, and disk resources. Compared to popular decentralized approaches, D-SPTF achieves up to 65% higher throughput and adapts more cleanly to heterogenous server capabilities.

D-SPTF

Distributed Shortest-Positioning Time First (D-SPTF) is a request distribution protocol for decentralized systems of storage servers. D-SPTF exploits high-speed interconnects to dynamically select which server, among those with a replica, should service each read request. In doing so, it simultaneously balances load, exploits the aggregate cache capacity, and reduces positioning times for cache misses. For network latencies expected in storage clusters (e.g., 10--200μs), D-SPTF performs as well as would a hypothetical centralized system with the same collection of CPU, cache, and disk resources. Compared to popular decentralized approaches, D-SPTF achieves up to 65% higher throughput and adapts more cleanly to heterogenous server capabilities.

D-SPTF

Distributed Shortest-Positioning Time First (D-SPTF) is a request distribution protocol for decentralized systems of storage servers. D-SPTF exploits high-speed interconnects to dynamically select which server, among those with a replica, should service each read request. In doing so, it simultaneously balances load, exploits the aggregate cache capacity, and reduces positioning times for cache misses. For network latencies expected in storage clusters (e.g., 10--200μs), D-SPTF performs as well as would a hypothetical centralized system with the same collection of CPU, cache, and disk resources. Compared to popular decentralized approaches, D-SPTF achieves up to 65% higher throughput and adapts more cleanly to heterogenous server capabilities.