Abstract: Session ITT-3

ITT-3.1	Design and Implementation of an Efficient Point Slicing Algorithm for the V.34 Modem Standard Oguz Tanrikulu (DSP Software Engineering, Inc) The V.34 modem standard uses QAM constellations with 4 to 1664 points depending upon the results of the line probing where the telephone channel is characterized to connect at the maximum possible data rate. In the receiver, the decisions errors due to the channel noise are prevented by using the Viterbi subset decoder which uses a point slicing algorithm that determines the closest valid point on the constellation for a given constellation size. In this paper we present a point slicing algorithm where the complexity is nearly independent of the QAM constellation size.
ITT-3.2	Buffer Control Technique for Transmission Frequency Recovery of CBR Connections over ATM Networks Gianmarco Panza (CEFRIEL), Silvio Cucchi, Daniele Meli (ALCATEL) The transmission of audio-video coded signal in real time applications over ATM networks requires sophisticated techniques for synchronization and buffer control. The presence of ATM cell delay variation (CDV) represents the major jitter source affecting the reconstruction of the time reference signal associated to the actual real-time service. In this paper is presented a buffer control technique and the related implementation aspects, based on both measure and utilization of CDV statistic or alternatively making use of buffer occupation statistic. The system allows setting target jitter attenuation in a way to have pre-established buffer underflow and overflow probabilities and its optimal utilization. The presented technique can be further extended to any asynchronous network context and is particularly suitable for high demanding professional audio-video applications.
ITT-3.3	A Personal and Inter-Vehicle Cordless Communications System William Jacklin (PrairieComm, Inc.), Stuart Collar, Scott Stratmoen, Brian Fitzpatrick, Jeffrey Stone (Northrop Grumman Corporation) The Northrop Grumman Cordless Communications System (CCS) is a state-of-the-art, wireless communications system targeted for tomorrow’s armored battlefield. Through a single, hand-held or body-mounted Personal Communications Unit (PCU), the CCS provides the soldier with access to a digital, full-duplex wireless intercom, as well as the capability to simultaneously communicate over a point-to-point digital data link. In addition, through the use of a single Universal Adapter Interface (UAI), the soldier can remotely access the existing analog or digital wired vehicle intercom and the attached wide area tactical radios. Moreover, using a novel protocol and system architecture, the CCS is able to reconfigure itself automatically and seamlessly into various centralized and distributed network configurations as the operational scenario changes. At the core of the CCS is a 20 MIPS, fixed-point Digital Signal Processor, which implements the protocol, control and signal processing algorithms required to achieve robust personal and inter-vehicle communications.
ITT-3.4	THe Impacts of Errors and Delays on the Performance of MPEG2 Video Communications Raouf Boutaba (ECE Department, University of Toronto, Canada), Ahmed Mehaoua (University of Cambridge, Hitashi Europe Telecom Lab., UK) Transmission of MPEG2-encoded video is one of the most demanding applications in terms of network resources and QoS requirement. It needs high bandwidth with stringent transmission delays. It can not tolerate large variations on delaysand it requires low error and data loss rates. Therefore, in order to design efficient integrated video communication systems over ATM networks, we propose in this paper to analyze the effects of errors and delays on both video signal and network performance.
ITT-3.5	CLOCK COMPENSATION IN A DATA/FAX RELAY SYSTEM Ehsan Daeipour (DSP Software Engineering, Inc.) In Data/Fax relay applications, one challenge is to overcome problems associated with having independent clocks in the system. Due to slight differences in the clock frequency of the relay system compared to the far modems (fax machines), the relay system is unable to transmit with the same rate that it receives data. Eventually, the system either starves for data or loses data due to under/over flow of its buffers. The clock difference problem in a Demod/Remod system and a method to compensate for it is addressed in this paper. The transmitter clock in each side of the relay system is tuned to match with the clock rate at the other side. This clock tuning is done via a closed loop feedback system. The bit error rate measurements show a significant improvement in the performance of the Demod/Remod system when the clock difference is compensated.
ITT-3.6	Space-Time Processing TDMA Wireless Testbed Hemanth T Sampath, Arogyaswami J Paulraj (Stanford University) The paper describes the architecture of the Stanford University (SU) testbed. The testbed was developed to evaluate space-time processing (STP) algorithms for diversity, co-channel interference and intersymbol interference (ISI) mitigation, array gain and space-time coding. It operates in both uplink and downlink modes and uses a hybrid (combining a real and simulated) channel environment. A description of transmit and receive schemes implemented on the testbed is presented.
ITT-3.7	GSM EFR Implementation for TRAU Application on DSP16000 Junchen Du, George Warner, Erik J Vallow, Penny e Breyer, Tom l Hollenbach (Lucent Technologies) An implementation of GSM EFR using traditional single MAC DSPs (DSP1600) takes 24 MIPS to run a single speech channel. Lucent DSP16000 is a dual MAC high performance DSP. Analysis has shown DSP16000 code can run GSM EFR at 10 MIPS per channel with the same program space as DSP1600. The code has been re-structured to minimize delays and RAM usage for multi-channel TRAU applications. For a 100 MIPS DSP16210 running 6 EFR speech channels, DSP16210 code takes 15.7K words of RAM with 3.2ms maximum delay for the encoder and 0.5ms for the decoder. Without re-structuring, the numbers are 27K, 15.4ms and 1.5ms respectively. This makes the DSP16000 very attractive for TRAU applications. The initial implementation runs at 12.8 MIPS per channel with 19.3K words of RAM, 4.7ms maximum encoding delay and 0.5ms maximum decoding delay for 6 speech channels.
ITT-3.8	Design of a Synchronization Scheme for a Bandwidth-on-Demand Multiplexer-Demultiplexer Pair Based on Wavelet Packet Tree Filter Banks Michael Sablatash, John H Lodge (Communications Research Centre) For a spectrum-efficient, bandwidth-on-demand multiplexer-demultiplexer for a multiuser, multicarrier communication system based on wavelet packet trees for downstream transmission described in recent publications by the authors, an innovative scheme for synchronization is proposed which uses a unique sync word at the root of the tree. An example filterbank tree with 8 leaves is used throughout the paper. For this example the creation of the unique 32-bit sync word is described. The multirate signal processing and construction and properties of 25 32x32 matrices required in the multiplexer to find locations of sync words at input ports to cancel the effects of data and and insert the 32-bit sync word in a window in the correct location at the root of the tree is described. The ratio of maximum to minimum absolute values of eigenvalues of the matrices is used to ensure well- conditioning, since they must be inverted in the processing.

< ITT-2

ITT-4 >

Last Update:  February 4, 1999         Ingo Höntsch