`
`AUGUST1989
`
`VOLUME 37
`
`NUMBER8
`
`(ISSN 0090-6778,)
`
`A PUBLICATION OF THE IEEE COMMUNICATIONS SOCIETY
`
`
`”APERS
`
`
`
`
`
`Coding
`1
`flijélarallel Viterbi Algorithm Implementation: Breaking the ACS—Bottleneck .................... G. Fertweis and H. Meyr
`‘
`Fading/Equalization
`
`791
`, ng in Adaptive Hybrids .................................. W. A. Set/111nm, C. R. Johnson, J11, and C. E. Rohrs
`
`ability of Error for Selection Diversity as a Function of Dwell Time ............... J. H. Barnard and C. K. Pauw 800
`‘ Modulation/Detection
`
`Al Digital Receiver Architecture for Bandwidth Efficient Transmission at High Data Rates .......................
`............................................................... G. Asc/wid, M. 061116! J. Sta/11, andH. Meyr
`Networks
`
`ghput Performance of an Unslotted Direct-Sequence SSMA Packet Radio Network ............................
`......................................................................... J. S. Storey and F. A. Tobagi
`
`.pti‘cal Communication
`;_
`'6 Division Multiple—Access Techniquesin Optical Fiber Networks—Part I: Fundamental Principles ..... J. A. Salehi
`
`6 Division Multiple-Access Techniques in Optical Fiber Networks—Part 11: Systems Performance Analysis ........
`.. ................................ _.......................................... J. A. Sale/11 andC A. Bracket!
`gnal Processing
`
`cking in Successive PCM/ADPCM Transcoders ................ M. Bonnet, O. Mam/11', and M. Jaidane—Saia’ane
`
`_ pread Spectrum
`. Elficient Technique for Evaluating Direct- Sequence Spread- Spectrum Multiple- Access Communications ............
`
`............................................................................................. J. S. Lehnert
`
`ity Coding for FH/MFSK Systems with Fading and Jamming—Part II: Selective Diversity ......................
`7 .................................................................... G L.Sr1'1'ber J W Mark and]. F Blake
`Sw1tchmg/Routmg
`
`785
`
`804
`
`814
`
`824
`
`834
`
`843
`
`851
`
`859
`
`'1
`
`
`
`;
`
`.
`1
`
`1
`‘1
`1
`
`
`
`‘RES‘P-ONDENCE
`
`Coding
`‘
`'
`‘ Maximum Likelihood Decoding Algorithm for Generalized Tail Biting Convolutional Codes Including
`' Codes .............................................................. Q. Wang and V. K. Blzargava
`g/E ualization
`asaDecision—Directed Equalizer Converged? .
`« adulation/Detection
`
`. .R. A. Kennedy, G.P11lford, B. D. 0. Anderson, andR. .R. Bitmead
`
`.
`
`'
`
`‘
`
`on “Pulse Shape, Excess Bandwidth and Timing Error Sensitivity in PRS Systems" .............. Y.Ta121'k
`
`Page 1
`
`rvation Multiple Access for Local Wireless Communications ...........................................
`................................... D. J. Goodman, R. A. Valenzuela, K. T. Gaylla1d, andB.Ramam111thi
`tc'hing/Routing
`ail Dynamic Routing in Double Ring Networks ............................ G. I. Stassinopoulos and H. Kukutos
`
`Ericsson Exhibit 1008
`
`Ericsson Exhibit 1008
`Page 1
`
`
`
`IEEE COMMUNICATIONS SOCIETY — IEEE
`"
`®
`
`The field of interest of the IEEE Communications Society consists of all telecommunications including telephone. telegraphy. f2 'simile. and pointito-point television. by electromagn
`propagation including radio: Wire; aerial. underground. coaxial. and submarine cables: waveguides. Communication satellites. and laser
`in marine. aeronautical. space and fixed station serv
`
`repeaters. radio relaying. signal storage. and regeneration: telecommunication error detection and correction: multiplexing and carrier techniques: communication switching systems: '
`communications: and communication theory.
`In addition to the above. this TRANSACTIONS contains papers pertaining to analog and digital signal processing and modulation. audio and video encoding techniques. the theory and 085m,
`
`
`transmitters. receivers. and repeaters for communications via optical and sonic media. the design and analysis of computer communication systems. and the development of communicatii
`
`software. Contributions of theory enhancing the understanding of communication systems and techniques are included. as are discussions of the social implications of the developingiil
`
`communication technology. All members of the IEEE are eligible for membership in the Society upon payment of the annual Society membership fee of $12.00. Members may receive-
`«
`TRANSACTIONS or the IEEE JOURNAL ON SELECTED AREAS iN COMMUNICATIONS upon payment of an additional $10.00 ($22.00 totall. or both publications upon payment of an additional 52'
`($32.0010tal). For information on joining. Write to the IEEE at the address below.
`IEEE COMMUNICATIONS SOCIETY
`TRANSACTIONS Editorial Board 1988
`
`F. E. FROEHLICH.S€Illfll‘
`in“
`
`Dep. Inform. Sci. Univ.
`
`Pittsburgh. PA 15260
`
`I. JACOBS. Senior Advisor-
`Dep. Elec. Eng.. Virginia Tech.
`Blacksburg. VA 24061
`
`E. BIGLIERI. Dutu Commun.
`& Mod.
`Elec. Eng. Dep.. UCLA
`Los Angeles. CA 90024
`K. BRAYER. Fnding/
`Multipath Chan.
`The MITRE Corp.
`Bedford. MA 01730
`E. R. BYRNE. Commun. Software
`Bellcore. Rm. 4D7376
`Piscataway. NJ 08854
`C. CHAMZAS. Digital Cum/nun.
`AT&T Bell Labs.. Rm. 36-215
`Holmdel. NJ 07733
`D. C. COLL. CATV
`Dep. Syst. Comput. Eng.. Carleton Univ.
`Ottawa. Ont.. Canada KIS 586
`V. CUPERMAN. Speech Processing
`Elec. Eng. Dep.. Simon Fraser Univ.
`Burnaby B. C.. Canada VSA 156
`D. DIVSALAR. Coding Theory & App].
`JPL
`Pasadena. CA 91109
`N. FARVARDIN. Qua/1L. Speech/Image
`Dep. Elec. Eng.
`Univ. Maryland
`College Park. MD 20742
`M. J. FERGUSON. Electronic Pub.
`& Protocols
`INRS Telecommun.
`Verdun. P.Q.. Canada H3E 1H6
`C. GEORGHIADES. Sync/1. and
`Opt. Detect.
`Dep. Elec. Eng.. Texas A&M Univ.
`College Station, TX 77843
`E. GERANIOTIS. Spread Spectrum
`Dep. Elec. Eng.. Univ. Maryland
`College Park. MD 20742
`
`S. D. PERSONICK. Director QI'PI/ltlit'utions V. B. LAWRENCE. Elli-ll)I‘-il]r(llilfl- D. G. DAUT. Pub/it (tll()]1.\' Editor
`Bell Commun. Res.
`AT&T Bell Labs.. Rm. 1P—246
`Dep. Elec. Comput. Eng.. Rutgers Univ.
`Rm. NVC—2Z463A. Box 7020
`Middletown. NJ 07748
`Piscataway. NJ 08854
`Red Bank. NJ 07701
`Editors
`S. LAM. Netii'zu‘k Protocols
`Dep. Comput. Sci.. Univ. Texas
`Austin. TX 78712
`A. LAZAR. Vt]f('(’/[)l]ll] Networks
`Elec. Eng. Dep.. Columbia Univ.
`New York. NY 10027
`P. MCLANE. Mod. Theory tuid
`Nonlinear C/iun.
`Dep. Elec. Eng.. Queen‘s Univ.
`Kingston. Ont.. Canada K7L 3N6
`U. MENGAIJ.
`'vntli.
`tind Tl’t‘ll.
`
`IET. Univ. Pisa. Via Diotisalvi 2
`56100 Pisa. Italy
`A. NETRAVAIJ. llnttge flattening
`AT&T Bell Labs.. Rm. 3D—406
`Murray Hill. NJ 07974
`P. R. PRUCNAL. th’ltill'ul'l’ Netu'tu'k.\
`Dep. Elec. Eng.. Princeton Univ.
`Princeton. NJ 08544
`I. RUBIN. Queueiiie/Bulier II/Ittiiu_t’eiiieiit
`Dep. Elec. Eng.. UCLA
`Los Angelcs. CA 90024
`E. G. SABLE. Coniniun. Suite/ting
`AT&T Bell Labs.. Rm. 2134631
`Holmdel. NJ 0773}
`Associate Etillll]',\
`J. G. KAPPEL
`D. VLACK
`K. SABNANI. Prom-01x
`AT&T Bell Labs.. Rm. 3D7434
`Murray Hill. NJ 07974
`H. SARI. Channel Equaliuititui
`Lab. d‘Electron. et de Phys. Appl.
`94450 LimeiI»Brevannes
`France
`M. SIDI. Commun. Netii'urlm
`Dep. Elec. Eng.. Technion-I.l.T.
`Haifa. 32000. Israel
`
`
`
`
`
`
`
`
`J. A. SILVESTER. Computer Commun.
`Dep. E1cc.Eng..U.S.C.
`Los Angeles. CA 90089
`H. TAKAGI. Queueing and Network Pctfurm.‘
`IBM Japan Tokyo Res. Lab.
`5719. Sanbanacho
`Chiyoda-ku. Tokyo 102. Japan
`
`D. TAYLOR. Sig. Design. Mod. and Detect.
`Dep. Elec. Comput. Eng.. McMaster Univ.
`Hamilton. Ont.. Canada L88 4L7
`J. UHRAN, Analog Commun.
`Dep. Elec. Eng.. Univ. Notre Dame
`South Bend. IN 46556
`G. UNGERBOECK. Sig. Prtu .-Si(ll'tl;{(‘ Merlin
`IBM Zurich Res. Lab.
`Saumersti'asse 4
`8803 Ruschlikon. Switzerland
`
`L.-F. WEI. Coding Theory & App].
`AT&T Bell Labs.. Rm. HOH-R13
`Holmdel. NJ 07733
`S. G. WILSON. Coding Theory & App].
`Dep. Elec. Eng. 84 Commun. Syst. Lab.
`Univ. Virginia
`Charlottesville. VA 22901
`J. W. WONG. Wide Area Networki
`Comput. Sci. Dep.. Univ. Waterloo
`Waterloo. Ont.. Canada N2L 361
`W. W. WU. Satellite Cont/nun. & Coding
`INTELSAT
`Washington. DC 20008
`T.—Y. YAN. Multiple Access Strti/egies
`Jet Propulsion Lab.. 238/420
`Pasadena. CA 91101
`Consultants
`T. OSATAKE. Ftir Eastern
`47-5 Nogata-(rchome
`Nakanoeku. Tokyo. 165. Japan
`L. H. ZETTERBERG. Europe
`Royal Inst. Technol.
`S-100 44 Stockholm. Sweden
`
`D. C. GLOGE. Uptii'ul Trims. SYN].
`AT&T Bell Labs.. Rm. 3F-632
`Holindel. NJ 07733
`M. S. GOODMAN. ()ptt'rn] Cont/nun.
`Bellcore. Rm. 2L-241
`445 South Street
`Morristown. NJ 0796071910
`1. GOPAL. Network Protot'uh
`[BM T. J. Watson Res. Cen.
`Yorktown Heights. NY 10598
`B. G. HASKELL. Intttge Commun.
`AT&T Bell Labs.. Rm. 40538
`Holmdel. NJ 07733
`R. ILTIS. Spread Spectrum
`Dep. Elec. Comput. Eng.. Univ. California
`Santa Barbara. CA 93106
`J. JAFFE. Routing and Suite/ting
`IBM T.J. Watson Res. Cen.
`P.O. Box 218
`Yorktown. NY 10598
`M. C. JERUCHIM. CAD C(ruunun. Syxt.
`General Electric Co.. Box 8048
`Philadelphia. PA 19101
`M. J. JOINDOT. Rtidiu Commun.
`CNET/LAB/MER
`Route de Tregastel. BP 40
`22301 Lannion. France
`M. KAVEHRAD. Mobile & L<7('[ll
`Commun.
`Dep. Elec. Eng.
`Univ. Ottawa
`Ottawa. Ont.. Canada KIN 6N5
`D. KAZAKns. Rmidmn Arrest <91. Dist.
`Commun. Sys‘t.
`Elec. Eng. Dep.. Virginia Univ.
`Charlottesville. VA 22901
`K. KUMMERLE. Loetil Al‘eti Netit’orkt
`IBM Zurich Res. Lab.
`Saumerstrasse 4
`8803 Ruschlikon. Switzerland
`.
`.
`.
`(Information for Ant/wry. CmnSot‘ Butu'd of Cover/unit. Departure/its. mid Cunt/nitteex uppeur on Cover III.)
`THE INSTITUTE OF ELECTRICAL AND ELECTRONICS ENGINEERS. INC.
`Officers
`HARRY W. MERGLER. Vice President. Edut‘utionttl Artivitiet
`EDWARD C. BERTNOLLI. Viee Previzlent. I’I’Q/Z’J'Sltnltll Activities
`TIMOTHY N. TRICK. Vice President. Publication Activities
`MICHAEL J. WHITELAW. Vii-e Preiident. Regional Activities
`H. TROY NAGLE. JR.. Viee President. Tt’l'll]ll( [ll A(']il‘lll(’.\'
`RALPH W. WYNDRUM. JR.. Diret tor. Dll'inUH III—Community]lions Tee/iiiolog)‘ Division
`Headquarters Stall"
`ERIC HERZ. Executive Director ttnd General Munuger
`.Sltlfl'Dll‘t’t‘llll'. Publishing Services
`DAVID L. STAIGER.
`THOMAS W. BARTLETT. Controller
`RUDOLF A. STAMPFL. Still/'Diret’lttr. Edut‘titimitt] Activities
`DONALD CHRISTIANSEN. Edilor. IEEE Spertrtnn
`CHARLES F. STEWART. JR.. Stufl‘ Director. Administrative Seri'it'ei‘
`IRVING ENGELSON. Stnfl‘Dii-ettor. Tee/mica] At'lil‘ltieA
`DONALD L. SUPPERS. Sltifi‘Diret’IU/u Field Services
`LEO FANNING. SltllliDll'l’CiUF. PI‘O/essiotiti/ Activities
`THOMAS C. WHITE. Stnfl‘Di‘ret-ttn: Public Infln-nintimz
`ANDREW SALEM. Std/fDiret‘tor. Standards
`Publications Department
`Pllblit'ullolm Managers: ANN H. BURGMEVER. GAIL S. FERENC. CAROLYNE TAMNEY WALSH
`Axrueinte Editor: BILL COL/\CCHIO
`
`EMERSON W. POOH. Preside/it
`CARLETON A. BAYLESS. President-Elect
`GEORGE F. ABBOTT. Executive Vice President
`HUGO RUCHARDT. Secretary
`WALLACE S. READ. Treasurer
`
`
`IEEE TRANSACTIONS ON COMMUNICATIONS is published month1y by The Institute of Electrical and Electronics Engineers. Inc. Responsibility for the contents rests upon the authors and no
`1"
`.Vt
`the IEEE. the Society/Council. or its members. IEEE Headquarters: 345 East 47th Street. New York. NY 10017-2394 U.S.A. NY Telephone: 212-705 + extension: Information-7900
`..
`'1'
`-
`Manager—7910. Controller-7748. Educational Services—7860. Publishing Servicese7560. Standardsv7960. Technical Services-7890. IEEE Service Center 201-98170060. Professional
`
`Washington Office 202785-0017. Telecopiers: NY (Headquarters) 212—75274929. NY (Pub]ications/Standards) 2116816005. NY Telex: 236-41 1 (international messages only). IEEE Service _
`(for orders. subscriptions. address changes. Educational Activities. Region/Section/Student Services]: 445 Hoes Lane. PO. Box 1331. Piscataway. NJ 08855-1331. NJ Telephone:2 310
`
`l9th Street. NW. Suite 608. Washington. DC 20036. Washington Telephone: _ --78570017. Price/Publication Informationtlmii
`IEEE Washington Oflice (for U.S. professional activities): 1|l|
`copies: IEEE members $10.00 (first copy only]. nonmembers $20.00 per copy. Available in microfiche and microfilm. Copyright and Reprint Permission: Abstracting is permitted with credil
`source. Libraries are permitted to photocopy beyond the limits of U.S. Copyright law for private use of patrons; 1 l those postel977 articles that carry a code at the bottom ofthe first page. illll'l
`the per-copy fee indicated in the code is paid through the Copyright Clearance Center. 29 Congress St.. Salem. MA 0197012) pie-1978 articles Without fee. Instructors are permitted to i
`.
`isolated articles for noncommercial classroom use without fee. For other copying. reprint. or republication permission. write to Copyrights and Permissions Department. PiiblishingS
`IEEE Headquarters. All rights reserved. Copyright © 1989 by The Institute of Electrical and Electronics Engineers. Inc. Printed in U.S.A. Second»class postage paid at New Y0rk.N
`additional mailing offices. Postmaster: Send address changes to IEEE TRANSACTIONS ON COMMUNICATIONS. IEEE. 445 Hoes Lane. P.O. Box 1331. Piscataway. NJ 08855-1331.
`
`Ericsson Exhibit 1008
`
`Page 2
`
`
`
`Pa“1
`'
`
`7 En
`'
`
`-
`
`1'.“
`"
`
`I
`
`i 0
`‘
`In
`hit!
`alg
`dI
`
`,
`
`.
`
`'
`.
`
`:
`~
`
`.
`
`‘
`
`'
`
`»
`
`i
`
`.
`
`«all
`
`-
`
`7V
`
`db‘a
`at
`
`. CH1
`I
`f
`"a
`. bat
`
`.
`0d
`
`11
`
`.Olu
`
`Ericsson Exhibit 1008
`Page 2
`
`
`
`~
`
`IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 31, NO. 8, AUGUST 1989
`
`15
`
`Duoknory
`Dicode
`Modified Duobmary
`
`R:Cosine Roll-off
`L : Linear Roll -off
`
`885
`show (and we quote) “the rate of decrease in @(e) as a function
`of excess bandwidth is more significant in 1 - D than in 1 +
`D and 1 - Oz.” This is also consistent with the other results
`our our paper, namely, the rates of increase in speed tolerance
`and eye width (slopes of Figs. 5 and 6) are more significant for
`1 - D than for the other schemes. However, Tanik is also
`correct in his interpretation of the plot of (unnormalized) @(e).
`In terms of the absolute performance measure @(e) [as well as
`in terms of speed tolerance and eye width] 1 + D is the best of
`,111. Thus, in summary, unnormalized and normalized @(e)
`IOW different aspects of timing jitter sensitivity as a function
`f excess bandwidth.
`
`Fig. I. P(0) for continuous filters.
`
`Duobinary
`Dicode
`Modified Duotinary : ----
`R:Cosine Roll-off
`L : L i n w Roll-off
`
`Packet Reservation Multiple Access for Local
`Wireless Communications
`
`D. J. GOODMAN, R. A. VALENZUELA, K. T. GAYLIARD, AND
`B. RAMAMURTHI
`
`Fig. 2. P(0) for modified filters.
`
`filters G ( f ) used in the system are cosine roll-off (C),
`modified cosine roll-off (MC), linear roll-off (L) and the
`filters are given in the above paper. ’
`modified linear roll-off (ML) filters. The spectra of these
`Figs. 1 and 2 show @(e) (with T = 1) for all PRS systems
`for normal and modified filtering, respectively. The best
`performance is obtained with the duobinary signaling for 0 <
`0.5 for the continuous filters and for 0 < 0.35 for the modified
`ones. Dicode gives the largest values for the low values of 0.
`This result is in contrast to the conclusions in the paper’ where
`it is stated that jitter performance of dicode is the best of all
`and that of the duobinary is the poorest. It seems that the
`normalization of @(e) curve to @(O) prior to plotting them
`(Appendix of the paper I ) has caused this misinterpretation.
`Actually, it is true the rate of decrease of normalized @(e)
`functions is highest for dicode signaling as stated in the paper.’
`However, @(e) should not be normalized for comparing the
`three PRS systems, since the samples of r ( t ) at optimum time
`instants take the same values for all techniques.
`
`Author’s Reply
`
`Abstract-Packet reservation multiple access (PRMA) allows a variety
`of information sources to share the same wireless access channel. Some of
`the sources, such as speech terminals, are classified as “periodic” and
`others, such as signaling, are classified as “random.” Packets from all
`sources contend for access to channel time slots. When a periodic
`information terminal succeeds in gaining access, it reserves subsequent
`time slots for uncontested transmission. Computer simulations and a
`listening’ test reveal that PRMA achieves a promising combination of
`voice quality and bandwidth efficiency.
`I. BACKGROUND
`Wireless access to public telecommunications networks is at
`present a topic of intense interest to researchers, developers,
`manufacturers and service providers throughout the world.
`There are healthy markets for the present generation of
`cellular mobile telephone services and residential cordless
`telephones. Plans for second generation products and services
`are advancing rapidly [ 11, 121, [3].
`Looking further into the future, we see several wireless
`access issues that remain to be resolved. Three important
`questions are as follows:
`1) how to create wireless private branch exchanges and local
`area networks that fill a gap between mobile telephony
`(serving a metropolitan area) and cordless telephones (serving
`a single residence),
`2) how to use the same resources to communicate efficiently
`voice, computer data, images and other types of information,
`and
`3) how to unify a variety of wireless access modes including
`cellular radio, cordless telephones, wireless private branch
`exchanges, wireless local area networks, dispatch services,
`and radio paging.
`
`A. GRAM1 AND S. PASUPATHY
`
`Paper approved by the Editor for Voice/Data Networks of the IEEE
`Communications Society. Manuscript received March 14, 1988; revised
`September 6, 1988. This paper was presented at the IEEE 38th Vehicular
`Technology Conference, Philadelphia, PA, June 1988. This work was
`We thank Y. Tanik for drawing our attention to the issue of
`normalization of @(e). As stated in the Appendix to our paper,’
`performed while the authors were at AT&T Bell Laboratories, Holmdel, NJ
`07733.
`we have indeed plotted the normalized measure @(0)/@(0) in
`D. J. Goodman is with the Department of Electrical and Computer
`our figures. This was done in order to study the effect of
`Engineering, Rutgers University, Piscataway, NJ 08855.
`excess bandwidth 0 on the rate of decrease in sensitivity to
`R. A. Valenzuela is with the Codex Corporation, Mansfield, MA 02048.
`timing phase jitter and also to show clearly that the relative
`K. T. Gayliard is with AT&T Bell Laboratories, Holmdel, NJ 07733.
`sensitivity stops decreasing after a certain value of 0 for certain
`B. Ramamurthi is with the Indian Institute of Technology, Madras, India.
`schemes. As mentioned in our paper, such normalized plots
`IEEE Log Number 8929108.
`OO90-6778/89/0800-0885$01 .OO 0 1989 IEEE
`
`Ericsson Exhibit 1008
`Page 3
`
`
`
`886
`
`This paper examines a key component of all of these issues,
`the multiple-access technique. Current and emerging systems
`use frequency division [ 11, [3] and time division [2] to provide
`many users with simultaneous access to the same wireless
`medium. Code division is another alternative that has received
`extensive attention [4], [5]. Here we explore a fourth one,
`packet contention.
`Packet contention techniques such as ALOHA and carrier
`sense multiple access [6] find widespread use in data commun-
`ications, including common control channel signaling in
`cellular mobile radio systems [7]. Among the principal merits
`of packet contention methods is their ability to serve a large
`number of terminals, each with a low average data rate and a
`high peak rate. While they function with little or no central
`coordination, packet contention techniques often make ineffi-
`cient use of the shared transmission medium. When too many
`terminals try to communicate at once, throughput goes down
`and transmission delay increases substantially. While recent
`studies [8], [9], [lo] indicate that packet contention schemes
`perform better in local radio environments than elsewhere,
`unpredictable, possibly long, delays have made packet conten-
`tion appear unattractive for voice transmission.
`Addressing this problem, this paper explores PRMA,
`packet reservation multiple access, a technique for transmit-
`ting, over short range radio channels, a mixture of voice
`packets and packets from other information sources. The
`PRMA protocol is organized around time frames with duration
`matched to the periodic rate of voice packets. In each frame,
`time slots are dynamically reserved for packets from active
`voice terminals. As a consequence, the terminals with reserva-
`tions share the channel in a manner closely resembling time
`division multiple access (TDMA). The throughput is high and
`the voice packet delay is constrained to meet a specific design
`limit. To enforce this constraint, terminals discard packets that
`encounter excess delay. Dropped packets are the main cause of
`speech impairment.
`PRMA is closely related to the reservation ALOHA
`protocol, R-ALOHA [ 1 13, [ 121. PRMA is distinguished from
`R-ALOHA by its response to network congestion and by its
`short round trip transmission time. In R-ALOHA, congestion
`causes long packet delays. In PRMA, information packets
`from periodic sources, such as speech, are discarded if they
`remain in the terminal beyond a certain time limit.
`In local wireless access systems, the roundtrip propagation
`time between terminals and base stations is on the order of a
`few tens of microseconds outdoors, and less than one
`microsecond indoors. Packet durations typically are 500- loo0
`ps. The short propagation times allow terminals to learn
`quickly the results of transmission attempts. In many cases, an
`acknowledgment message for the current time slot can arrive
`at the terminals before the beginning of the next time slot, or,
`at most, one slot later. In our studies, we have assumed
`immediate acknowledgments are possible. A one slot delay
`would have little effect on performance.
`In the configuration we have studied, PRMA makes
`efficient use of speech activity detectors to obtain a bandwidth
`efficiency improvement over time division multiple access.
`The control complexity of TDMA makes it a difficult matter to
`use speech detection to improve efficiency. PRMA, on the
`other hand, is simple to implement and gracefully accommo-
`dates many types of information.
`
`11. SCOPE OF THIS WORK
`We are concerned with a wireless packet communication
`network with a star topology. All terminals use a single
`channel to transmit information packets to a central base
`station. This upstream (terminal-to-base) channel is slotted,
`and after each time slot, the base station transmits a short
`acknowledgment packet in addition to a downstream informa-
`
`IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 31, NO. 8, AUGUST 1989
`
`tion packet. Downstream traffic can be transmitted in a
`separate channel (using a different frequency band). Or, it can
`time share a single channel with the upstream traffic. In either
`case, the base station schedules the downstream traffic
`avoiding all contention. In this paper, we concentrate on the
`problem of dispersed terminals competing for access to the
`upstream channel.
`We are interested in possible applications of this network in
`an indoor or other localized service area. In terms of radio
`transmission, two salient features of these environments are
`short round-trip propagation delays, and wide variations in
`path attenuations (near/far phenomenon). The short delays
`permit rapid acknowledgments of the results of packet
`transmissions. The near/far phenomenon admits the possibility
`of packet capture when two or more terminals transmit packets
`in the same time slot. In the absence of capture, all contending
`packets require retransmission. On the other hand, accurate
`detection of the strongest received packet could lead to
`substantial performance improvements [8], [9], [lo].
`To explore the capabilities of such a network for telephony,
`we have simulated on a computer the transmission of up to 50
`simultaneous conversations. To do so, we have created an
`elaborate statistical model of the patterns of talkspurts and
`silent intervals in conversational speech. In addition to
`artificial speech generated under the control of this model, we
`have also simulated the transmission of real speech. In a
`listening test, the simulated speech transmissions reveal the
`subjective effects of impairments caused by network conges-
`tion.
`Each terminal contains a sensitive voice activity detector, a
`32 kbit/s speech encoder, and a packet assembler. Packets
`consist of 64 bits of header and other non-speech material plus
`512 coded speech bits. Our study explores two important
`variables. One is the packet transmission protocol which can
`be ALOHA or PRMA, defined in detail in the next section of
`this paper. The other variable is the strength of the capture
`phenomenon. We have studied performance with no capture,
`partial capture, and perfect capture [ 131.
`A fundamental requirement in speech communication is
`prompt delivery of information. This is in contrast to packet
`data systems which respond to congestion and transmission
`impairments by delaying packets in queues. In our study,
`terminals discard speech packets that are not successfully
`transmitted within 32 ms. A figure of merit is the amount of
`voice traffic carried in the upstream channel without exceeding
`a specified probability of packet dropping.
`A transmission delay as long as 32 ms implies that echo
`control will be required when PRMA is used for access to the
`public telephone network. This is comparable to the delay
`budget of Pan-European mobile radio [2] and the delay budget
`of a statistical multiplexer used in a terrestrial packet speech
`network [ 141. In response to congestion, this multiplexer
`reduces the lengths of speech packets, rather than discard
`entire packets. Although packet length reduction leads to
`higher efficiency than packet dropping, it is a difficult matter
`to provide variable length packets in a system with dispersed
`terminals.
`
`111. PACKET RESERVATION MULTIPLE ACCESS
`At a speech terminal, the time slots are grouped in frames.
`Each slot in a frame is recognized as “reserved” or
`‘ ‘available” according to the acknowledgment message re-
`ceived from the base at the end of the slot. When a talkspurt
`begins, the terminal uses the ALOHA protocol to contend for
`an available slot. When it successfully transmits a speech
`packet, it reserves that slot in future frames and there are no
`subsequent collisions with packets from other terminals. At
`the end of the talkspurt, the terminal releases its reservation by
`leaving the reserved slot empty.
`
`Ericsson Exhibit 1008
`Page 4
`
`
`
`IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 31, NO. 8, AUGUST 1989
`
`A . Packet Categories
`The packet assembler distinguishes between two types of
`information packets: periodic information packets and random
`information packets. The packet category is communicated by
`means of one bit of the packet header. Speech packets are
`always labeled as “periodic.” Certain data packets, such as
`those involved in file transfers, can also be “periodic.” Other
`data packets, such as keyboard entries to a computer terminal,
`signaling messages and system control information, are
`labeled as ‘ ‘random. ’ ’
`B. Information Frames
`Each terminal organizes the transmission time slots in
`frames with N slots per frame. N is a system parameter
`common to all terminals. However, it is not necessary for all
`terminals to agree on which slot is the first in the frame. The
`terminal contains a frame reservation register, with one bit
`for each slot in the frame. It sets a bit to “0” when informed
`by the base station that the corresponding time slot is
`unreserved; otherwise it sets the bit to “1.”
`C. Contention
`To begin to send periodic information, a terminal uses the
`slotted ALOHA [6] protocol to contend with other terminals
`for an unreserved time slot. If the terminal does not success-
`fully transmit the first packet of a talkspurt in the first
`unreserved time slot, it retransmits the packet with probability
`q in subsequent unreserved slots. It continues to do so until the
`base station acknowledges successful reception of the packet.
`The permission probability q is a design variable.
`D. Reservations
`At the end of each upstream transmission, the base station
`broadcasts the outcome in an acknowledgment packet. When
`the base station acknowledges accurate reception of a periodic
`information packet, the terminal that sent the packet reserves
`that time slot for future transmissions. All terminals then
`refrain from using that slot in future frames. The terminal with
`the reservation thus has uncontested use of the time slot.
`When a terminal stops sending periodic information in the
`reserved slot, this event is broadcast by the base station in the
`acknowledgment packet. All terminals are then free to contend
`for that slot in future frames.
`E. Packet Loss
`While it is contending for unreserved time slots, the
`terminal holds packets in a first-in first-out-buffer. If the
`packets are speech, the buffer size is limited according to the
`delay constraint imposed upon the network. In our study, the
`buffer holds 32 ms of speech. When a new speech packet
`arrives at a full buffer, the buffer discards the oldest packet.
`The number of lost packets and their temporal distribution
`strongly affect the quality of the received speech. With
`PRMA, all packet losses occur at the beginnings of talkspurts.
`It has been observed that this “front end clipping” is less
`harmful to subjective speech quality than other types of packet
`loss [15]. Our listening test supports this observation.
`
`F. Random In formation Packets
`A terminal transmits random information packets in unre-
`served time slots. In the event of a collision, packets are
`retransmitted with probability r. This probability could differ
`from q, the permission probability for periodic information
`packets. By setting q > r, the system would give priority to
`periodic over random information. When a random packet is
`successfully transmitted, the terminal does not obtain a time
`slot reservation. If it has other packets to send, it must contend
`for subsequent unreserved time slots.
`The buffer size for random information packets can be quite
`
`887
`
`long. If it is, the effect of network congestion on random
`information is long packet delay, rather than packet loss as
`with periodic information packets.
`G. ALOHA
`One of our aims is to compare PRMA to conventional,
`nonreservation slotted ALOHA. In conventional ALOHA,
`contention takes place as in PRMA. However, all slots are
`unreserved, and all periodic information packets must contend
`with transmissions from other terminals. It is known that
`ALOHA benefits from packet capture. An interesting question
`is whether a strong capture mechanism also enhances the
`performance of PRMA.
`IV. COMPUTER SIMULATION
`We have performed computer simulations to investigate
`PRMA performance and to compare PRMA with slotted
`ALOHA. The simulated network carries conversational
`speech coded at 32 kbits/s. The channel rate is 720 kbits/s
`which is a conservative (low) estimate of what an indoor
`channel can support [16].
`A . Transmission Format
`After obtaining a reservation, a periodic information termi-
`nal transmits one packet per frame. Therefore, the frame
`repetition rate must equal l / T p , the rate at which the terminal
`generates packets. In our simulations Tp = 16 ms and there
`are 62.5 frameds. With 32 kbit/s speech coding, there are 512
`speech bits per packet. In addition, 64 bits are allocated for
`header information and other purposes. Therefore, each slot
`contains 576 bits. This packet size is typical of those
`considered in general packet voice studies [17], [18]. It is
`employed in an experimental (wired) packet voice network
`r141.
`With the frame duration 16 ms, the 720 kbit/s channel
`transmits 11 520 bits per frame. Therefore, there are 11 520/
`576 = 20 slots per frame. With a delay limit of 32 ms, a
`packet is dropped after waiting two frames (40 time slots) for a
`reservation.
`B. Packet Collisions
`When two or more packets contend for the same time slot,
`the ability of the base to detect the strongest packet depends on
`the channel characteristics, the transmission technique (modu-
`lation and coding), and on the locations of the active terminals.
`Our simulation study employs a simple capture model, in
`which the base station is at the center of a service area and the
`terminals are uniformly spaced between the cell center and the
`perimeter of the service area.
`We identify three levels of capture: no capture, partial
`capture, and perfect capture. With no capture, the base
`station is unable to detect any packet when there are two or
`more simultaneous transmissions. All colliding packets must
`be retransmitted.
`With partial capture, the ability of the base station to detect
`the strongest packet depends on the relative positions of the
`two active terminals that are nearest the