No abstract

No abstract

The growing demand for high-speed, low-cost, and low-overhead I/Os in today's electronic systems, has been addressed by three general categories of interconnects: electrical, optical, and wireless. The electrical interconnects are the oldest and have improved the most, where bit rates in excess of 20Gb/s are achieved over a pair of conductors [1]. At such high bit rates, these serial links must handle transmission line loss, dispersion, impedance mismatches, and electromagnetic crosstalk among multiple lines requiring sophisticated designs, often needing equalization, with their own cost and overhead limitations [2]. Optical fibers as interconnects do not suffer from similar bandwidth limitations or cross-talk issues. However, they require additional electrical-to-optical (EO) and optical-toelectrical (OE) conversion devices for generation and detection of optical signals [3], which impose serious constraints on power consumption, cost, and footprint of optical interconnects. Wireless connection at millimeter-wave frequencies can also be used for short distance connections [4]. While they provide the most versatility and are a promising option, additional development is still necessary to scale them to a highly parallel system with multiple channels running concurrently.This paper presents a 12.5+12.5Gb/s full-duplex plastic waveguide interconnect solution based on millimeter-wave signal transmission. The plastic waveguide is simply a long solid piece of plastic that provides a very simple, versatile, flexible, and low-cost transmission medium that has the main advantages of optical fiber in isolation and bandwidth, without the need for costly EO and OE. The dielectric waveguide does not need to be connected electrically like the wire or aligned to micron-level accuracy like optical fibers. It can be bent and twisted without significant impact on the signal. Compared to the wireless link discussed earlier, it offers additional signal isolation and confinement. Thus, it can be extended over much longer distances due to the low attenuation in the waveguide (as opposed to free space) and multiple independent lines can be run in parallel to increase the bandwidth.In our proposed plastic waveguide link, the TXs and RXs are fully integrated in CMOS, and the waveguide couplers can be fabricated in a conventional resin package without additional cost. In our existing setting there are a transmitter and a receiver operating at different carrier frequencies on each side of the waveguide, making it possible to realize a full-duplex solution. Because of the smaller fractional bandwidth for the millimeter-wave transmission, no equalization circuit is required. Figure 8.5.1 presents a diagram of our proposed solution. It consists of a pair of transceivers A and B, and a plastic waveguide. Transceiver A contains a 57GHz RX and an 80GHz TX, and Transceiver B contains an 80GHz RX and a 57GHz TX. This combination allows for a bi-directional full-duplex transmission. An alternative is to place both TX on one side and both RX on the...

One of the main properties of today's distributed and parallel systems, such as mobile ad-hoc networks and grids, is their heterogeneity in the available resources. Further, many applications of such systems are subject to Time/Utility Function (TUF) time constraints for jobs, unavoidable variability in job characteristics and arrivals, and statistical assurance requirements on timeliness behaviors. In this paper, we propose an exact analytical solution for performance evaluation of dynamic policies used for routing of TUF-constrained Firm Real-Time (FRT) jobs among parallel single-processor queues with arbitrary processing rates and capacities. The analytical method can be used for the evaluation of the compliance of some important statistical assurance requirements. Furthermore, we present a utility-aware dynamic routing policy to improve the expected accrued utility of the parallel system. The policy called Maximum Expected Utility (MEU) behaves based on the information gathered from the analytical solution. MEU is compared with some well-known Dynamic Routing (DR) policies for different TUF shapes and both cases of homogeneous and heterogeneous processors of a two-queue system. The comparisons show the efficiency of MEU for the former case and its good behavior in most situations for the latter case.

A new LSI configuration is proposed for multigigabit optical interconnections with a view to lowcost, low-power-dissipation, and precise-adjustment-free interconnection modules. This configuration is based on (a) minimization of additional functions for the optical transmitter and receiver and (b) realization of fully digital automatic timing adjustment. Test fabrication results of the transmitter and receiver LSIs, and modules show the effect of this new configuration, that is, a 2.8-GbiUs operation at,a received power of -9 dBm through a 100-m optical fiber. I N T R O D U C T I O NOptical transmission technology, mainly developed for long-distance communications, is currently being evolved to short-distance applications such as local area networks [ 11- [3]. In particular, optical interconnection systems are expected to eliminate board-to-board interconnection bottlenecks for future Gbit/s electronic equipment.A long-distance transmission system is required to have various functions such as automatic power control (APC) and automatic temperature control (ATC) for a transmitter or automatic offset canceling (AOC) and automatic gain control (AGC) for a receiver. These functions maintain a stable long-span transmission performance with a high sensitivity and a wide dynamic range by compensating the large fiber loss and characteristic variations of the LD and PD. The system also needs many off-chip components, such as large capacitors for phase compensation of DC feedback and for clock timing extraction, and for analog PLL timing adjustment circuit.The most important consideration in putting an optical interconnection to practical use is how to realize low-cost, low-power-dissipation, multi-channel, and preciseadjustment-free interconnection modules. NEW CIRCUIT CONFIGURATION(a) Minimization of additional functions Figure 1 shows the block diagram of an optical interconnection system with an AOC function in addition to basic functions such as E/O and O/E converters, an amplifier and a decision circuit. Dc-coupling with the AOC function is adopted because an ac-coupling capacitor and a dc-clamp circuit are difficult to monolithically integrate on a receiver LSI. The AOC circuit consists of two peak hold circuits and a DC amplifier, and it automatically adjusts the reference level to the post-amplifier, in order to equalize peak levels of balanced outputs (Q and Q-inverse) from the post-amplifier and to maximize an eye opening for a next stage decision circuit. This AOC circuit can compensate various offsets such as a zero-level offset caused by LD and PD variations and amplitude changes caused by a connection's loss as shown in Fig. 2 (a) and (b), respectively. Therefore, the AOC function is indispensable for systems with loss variations, APC and AGC functions can be omitted. Bit rate converters such as a multiplexer (MUX) and a demultiplexer (DMUX) are also necessary to eliminate board-to-board interconnection bottlenecks. (b) Automatic timing adjustment circuitA decision block with an automatic timing adjustment c...