See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/3436072
`
`IEEE 802.11n: Enhancements for higher throughput in wireless LANS
`
`Article  in  IEEE Wireless Communications · January 2006
`
`DOI: 10.1109/MWC.2005.1561948 · Source: IEEE Xplore
`
`CITATIONS
`318
`
`1 author:
`
`Yang Xiao
`University of Alabama
`
`591 PUBLICATIONS   21,317 CITATIONS   
`
`SEE PROFILE
`
`READS
`3,170
`
`All content following this page was uploaded by Yang Xiao on 09 April 2018.
`
`The user has requested enhancement of the downloaded file.
`
`Exhibit 1033
`Panasonic v. UNM
`IPR2024-00364
`Page 1 of 11
`
`

`

`XIAO LAYOUT 12/1/05 12:13 PM Page 82
`
`ACCEPTED FROM OPEN CALL
`
`IEEE 802.11N: ENHANCEMENTS FOR
`HIGHER THROUGHPUT IN WIRELESS LANS
`
`YANG XIAO, THE UNIVERSITY OF MEMPHIS
`
`ABSTRACT
`This article introduces a new standardization
`effort, IEEE 802.11n, an amendment to IEEE
`802.11 standards that it is capable of much high-
`er throughputs, with a maximum throughput of
`at least 100 Mb/s, as measured at the medium
`access control data service access point. The
`IEEE 802.11n will provide both physical layer
`and MAC enhancements. In this article we intro-
`duce some PHY proposals and study the funda-
`mental issue of MAC inefficiency. We propose
`several MAC enhancements via various frame
`aggregation mechanisms that overcome the theo-
`retical throughput limit and reach higher
`throughput. We classify frame aggregation mech-
`anisms into many different and orthogonal
`aspects, such as distributed vs. centrally con-
`trolled, ad hoc vs. infrastructure, uplink vs.
`downlink, single-destination vs. multi-destina-
`tion, PHY-level vs. MAC-level, single-rate vs.
`multirate, immediate ACK vs. delayed ACK, and
`no spacing vs. SIFS spacing.
`INTRODUCTION
`Wireless local area networks (WLANs) are
`becoming more popular and increasingly relied
`on. The IEEE 802.11 WLAN is accepted as a
`complementary technology to high-speed IEEE
`802.3 (Ethernet) for portable and mobile devices.
`One reason for such success is that it keeps
`increasing data transmission rates while main-
`taining a relatively low price. The IEEE 802.11,
`802.11b, and 802.11a/g specifications provide up
`to 2 Mb/s, 11 Mb/s, and 54 Mb/s data rates [1,
`2], respectively. Furthermore, the IEEE 802.11
`Working Group is pursuing IEEE 802.11n, an
`amendment for higher throughput and higher
`speed enhancements. Different from the goal of
`IEEE 802.11b/.11a/.11g, i.e., to provide higher-
`speed data rates with different physical layer
`(PHY) specifications, IEEE 802.11n aims at
`higher throughput instead of higher data rates
`with PHY and medium access control (MAC)
`enhancements.
`The IEEE 802.11 MAC employs a mandatory
`contention-based channel access function called
`the distributed coordination function (DCF) and
`an optional centrally controlled channel access
`
`function, the point coordination function (PCF)
`[1]. The DCF adopts carrier sense multiple
`access with collision avoidance (CSMA/CA) with
`binary exponential backoff, and the PCF adopts
`a polling mechanism. To support MAC-level
`quality of service (QoS), the IEEE 802.11 Work-
`ing Group is currently working on the standard-
`ization of IEEE 802.11e. The IEEE 802.11e
`MAC employs a channel access function called
`the hybrid coordination function (HCF), which
`includes contention-based channel access,
`enhanced distributed channel access (EDCA),
`and contention-free centrally controlled channel
`access, HCF controlled channel access (HCCA).
`To provide better QoS, especially for multi-
`media applications, increasing data rates is also
`highly desirable. The rationale is the same as
`Ethernet, which dramatically increases data rates
`from 10/100 Mb/s to 10 Gb/s. Data-rate-intensive
`applications exist such as multimedia conferenc-
`ing, MPEG video streaming, consumer applica-
`tions, network storage, file transfer, and
`simultaneous transmission of multiple HDTV
`signals, audio, and online gaming. Furthermore,
`there is a great demand for higher-capacity
`WLAN networks in the market in such areas ass
`hotspots, service providers, and wireless back-
`haul. Therefore, increasing data rates is crucial,
`and the IEEE 802.11 Working Group was seek-
`ing higher data rates over 100 Mb/s for IEEE
`802.11a extension [2]. However, we proved that
`a theoretical throughput limit exists due to MAC
`and PHY overhead [2]. In other words, the theo-
`retical throughput limit, about 75 Mb/s for IEEE
`802.11a with a payload size of 1500 bytes [2],
`upper bounds any obtained throughput even
`when the data rate goes infinitely high. There-
`fore, increasing transmission rate cannot help a
`lot. Both reducing overhead and pursuing higher
`data rates are therefore necessary and important
`[2]. In July 2002 the IEEE 802.11 High Through-
`put Study Group (HTSG) was established,
`emphasizing higher throughput for data rates
`over 100 Mb/s in WLANs. The first official
`meeting of the IEEE 802.11n Task Group took
`place in September 2003, replacing the IEEE
`High Throughput Study Group (HTSG). The
`scope of IEEE 802.11n is to define an amend-
`ment to the IEEE 802.11 standards to enable
`much higher throughputs, with a maximum
`
`MAC
`frame 1
`
`header
`MAC
`header
`
`PHY
`
`L1
`
`(d) MAC level version
`
`MAC
`frame 1
`
`MAC
`frame 2
`
`header
`
`PHY
`
`(e) MAC level version 2: sin
`
`Payload
`
`L1
`
`MAC
`header
`
`header
`
`PHY
`
`(f) MAC level version 3: sin
`
`A single PHY frame
`
`The authors
`introduce a new
`standardization
`effort, IEEE 802.11n,
`an amendment to
`IEEE 802.11
`standards that it is
`capable of much
`higher throughputs,
`with a maximum
`throughput of at
`least 100Mb/s,
`as measured at the
`medium access
`control (MAC) data
`service access point.
`
`82
`
`1536-1284/05/$20.00 © 2005 IEEE
`
`IEEE Wireless Communications • December 2005
`
`Exhibit 1033
`Panasonic v. UNM
`IPR2024-00364
`Page 2 of 11
`
`

`

`We showed that a
`theoretical through-
`put upper limit exists
`for IEEE 802.11
`protocols. Therefore
`increasing
`transmission rate
`cannot help much.
`Both reducing
`overhead and
`pursuing higher data
`rates are therefore
`necessary and
`important
`
`XIAO LAYOUT 12/1/05 12:13 PM Page 83
`
`throughput of at least 100 Mb/s, as measured at
`the MAC data service access point (SAP). IEEE
`802.11n will provide both PHY and MAC
`enhancements. Note that even though IEEE
`802.11e also provides some mechanisms for effi-
`cient MAC enhancements such as Direct Link
`Protocol and Block Acknowledgment Protocol,
`its major goal is still to provide QoS services,
`whereas the goal of IEEE 802.11n is to provide
`higher throughput via PHY and MAC enhance-
`ments.
`In this article we first introduce the history
`and current status of IEEE 802.11n, as well as
`some PHY proposals. Then we study the theo-
`retical throughput limit and provide an overhead
`analysis for IEEE 802.11, and compare this
`aspect with HIPERLAN/2. Finally, we propose
`several MAC enhancements via various frame
`aggregation mechanisms. We adopt the original
`IEEE 802.11 MAC in this article, but the mech-
`anisms can easily be applied to IEEE 802.11e.
`IEEE 802.11N
`In this section we provide an up-to-date survey
`on the efforts to produce the IEEE 802.11n
`standard, including its history and current status.
`The IEEE 802.11n standard process has three
`phases: phase 1 is the preparation stage from
`January to September 2002; phase 2 was the of
`IEEE 802.11 HTSG from September 2002 to
`September 2003; phase 3 is the IEEE 802.11n
`Task Group (TGn), which began in September
`2003 and is expected to finish in March 2007.
`PHASE 1: PREPARATION
`The first formal presentation at IEEE 802 meet-
`ings about 802.11a higher data rate extension
`was at the IEEE 802.11 interim meeting at Dal-
`las, Texas, in January 2002 [2]. In this presenta-
`tion Jones et al. described the high demand for
`data rates over 100 Mb/s through IEEE 802.11,
`and described some potential approaches to
`achieve higher data rates: modulation and cod-
`ing enhancements, spatial diversity techniques,
`spatial multiplexing, and double bandwidth solu-
`tions with underlying IEEE 802.11a waveforms.
`Later on at the meeting, a straw poll for a call
`for interest on 802.11a higher rate extension was
`conducted in the IEEE 802.11 working group,
`and received tremendous interest among hun-
`dreds of committee members.
`At the St. Louis IEEE 802 plenary meeting in
`March 2002, we provided a throughput analysis
`for higher data rates over 100 Mb/s [2]. Tzannes
`et al. proposed a bit-loading (BL) approach [2].
`One of the drawbacks is that BL may require
`feedback from the receiver to the transmitter,
`and the communication from the receiver to the
`transmitter must happen faster than channel
`changes. Hori et al. compared four different
`potential approaches: the double clock rate
`approach, the double subcarrier number
`approach, the 4096-quadrature amplitude modu-
`lation (QAM)-orthogonal frequency-division
`multiplexing (OFDM) approach, and the
`OFDM/space-division multiplexing (SDM) (mul-
`ticarrier multiple-input multiple-output
`[MIMO]) system approach [2]. Coffey suggested
`some criteria for higher data rates, and that
`
`higher data rates should emphasize throughput
`with consideration for backward compatibility
`rather than data rate [2].
`At the Sydney, Australia, IEEE 802 interim
`meeting in May 2002, we showed that a theoreti-
`cal throughput upper limit exists for IEEE
`802.11 protocols [2]. Therefore, increasing trans-
`mission rate cannot help much. Both reducing
`overhead and pursuing higher data rates are
`therefore necessary and important [3]. The
`IEEE 802.11 HTSG was established in July 2002
`emphasizing higher throughput for higher data
`rates over 100 Mb/s WLANs.
`PHASE 2: IEEE 802.11 HTSG
`The IEEE 802.11 HTSG began operations in
`September 2002 and ended in September 2003.
`During this phase, a Project Authorization
`Request (PAR) and Five Criteria for Standards
`Development were established.
`The scope of the MAC and PHY enhance-
`ments assume a baseline specification to support
`higher throughput. The amendment seeks to
`improve the peak throughput to at least 100
`Mb/s, measured at the MAC data SAP. This rep-
`resents an improvement of at least four times
`the throughput obtainable using existing 802.11
`systems. The highest throughput mode shall
`achieve a spectral efficiency of at least 3 b/s/Hz.
`The Task Group (IEEE 802.11n) will undertake
`the following steps:
`• Identify and define usage models, channel
`models, and related MAC and application
`assumptions.
`• Identify and define evaluation metrics that
`characterize the important aspects of a partic-
`ular usage model.
`Initial usage models include hotspot, enter-
`prise, and residential. Evaluation metrics include
`throughput, range, aggregate network capacity,
`power consumption (peak and average), spectral
`flexibility, cost/complexity flexibility, backward
`compatibility, and coexistence.
`The Five Criteria for Standards Development
`are:
`• Broad market potential: It shall have a broad
`market potential; that is, broad sets of appli-
`cability, multiple vendors and numerous users,
`and balanced costs (LAN vs. attached sta-
`tions).
`• Compatibility: Keeping the MAC SAP inter-
`face the same as for the existing 802.11 stan-
`dards is required for compatibility. New
`enhancements shall be defined in a format
`and structure consistent with existing 802.11
`standards.
`• Distinct identity: Each IEEE 802 standard
`shall have a distinct identity from other IEEE
`802 standards.
`• Technical feasibility: Those introduced in
`phase 1 and later parts of this subsection can
`provide technical feasibility. Furthermore,
`there are currently reliable WLAN solutions.
`• Economic feasibility: Economic feasibility
`includes known cost factors, reasonable cost
`for performance, and consideration of installa-
`tion costs.
`Next, we introduce some additional proposals
`for technical feasibility. In [4] the authors pro-
`posed exploring space diversity through multiple
`
`IEEE Wireless Communications • December 2005
`
`83
`
`Exhibit 1033
`Panasonic v. UNM
`IPR2024-00364
`Page 3 of 11
`
`

`

`XIAO LAYOUT 12/1/05 12:13 PM Page 84
`
`coding and modulation functions to provide
`improved performance. Coded modulation
`schemes combine with BL to encode all informa-
`tion bits. The advantages include:
`• It does not have a preset maximum data rate.
`• It is optimally suited to a multipath channel.
`• It is based on mature and widely understood
`technology, such as asymmetric digital sub-
`scriber line (ADSL).
`• It requires a relatively small standardization
`effort.
`In [6] the author proposed a MIMO-OFDM
`solution. In [7] the authors claimed that 250
`Mb/s data rate is achievable. In [8] the authors
`showed some experimental results based on
`MIMO-OFDM for high-throughput WLANs. In
`[9] the author proposed using smart antennas to
`improve SNR, increase coverage/range and data
`rate, reduce interference and multipath, and
`increase network capacity and battery life.
`PHASE 3: IEEE 802.11N TGN
`The first official meeting of the IEEE 802.11n
`Task Group took place September 2003 in Sin-
`gapore. The IEEE 802.11n standard is planned
`to be published in March 2007. The TGn will
`further go through the following steps: establish-
`ing the proposal selection process and criteria,
`call for proposals, combination of proposals, sev-
`eral letter ballots, standard approval, and finally
`standard publication. We will see more contribu-
`tions in future IEEE 802.11n meetings. So far,
`most contributions in IEEE 802.11n meetings
`focus on PHY enhancements. Instead, this arti-
`cle serves a good purpose in discussing MAC
`enhancements.
`
`THEORETICAL LIMIT AND OVERHEAD
`ANALYSIS OF THE IEEE 802.11 MAC
`The achievable maximum throughput (MT) can
`be met when the system is in the best case sce-
`nario:
`• The channel is ideal, without errors.
`• At any transmission cycle, there is one and
`only one active station that always has a frame
`to send, and other stations can only accept
`frames and provide acknowledgments (ACKs).
`The throughput upper limit (TUL) [2] is defined
`as the maximum throughput when the raw data
`rate goes infinitely high.
`As indicated in [3], overhead is the major fun-
`damental issue for inefficient MAC, and it
`includes headers (MAC header, frame check
`sequence [FCS], and PHY header), interframe
`spaces (IFSs), backoff time, and ACKs. Define
`overhead as the difference between data rate and
`throughput, and define normalized overhead as
`overhead divided by data rate. We further assume
`that all higher data rates are compatible with
`IEEE 802.11a. Let Tslot, TSIFS, TDIFS, and CWmin
`denote a slot time, a short IFS (SIFS) time, a dif-
`ferentiated IFS (DIFS) time, and the minimum
`backoff contention window size, respectively. Let
`TP and TPHY denote transmission times of a phys-
`ical preamble and a PHY header, respectively.
`Let TDATA and TACK denote transmission times
`of a data frame and an ACK, respectively. Let
`LDATA denote the payload length in bytes.
`
`antennae other than frequency diversity via bit
`interleaved coded modulation, and claimed that
`OFDM is very well suited for use with multiple
`antennae, for example, as an optional mode in
`IEEE 802.16, with the cost of an additional
`antenna and a radio frequency (RF) front-end.
`In [5] the authors proposed a combined scheme
`of BL and trellis-coded modulation (TCM). BL
`is the process of modulating a different number
`of bits on each carrier based on the signal-to-
`noise ratio (SNR) of the carriers. IEEE 802.11a
`adopts an equal number of bits per carrier,
`which is the simplest form of BL. BL is better
`suited to a multipath channel. However, it
`requires feedback from the receiver to the trans-
`mitter, and the communication from the receiver
`to the transmitter must happen faster than chan-
`nel changes. On the other hand, TCM combines
`
`TUL
`MT (54 Mb/s)
`MT (216 Mb/s)
`
`TUL =
`
`8LDATA
`2Tp + 2TPHY + TDIFS + TSIFS + (CWmin – 1)Tslot/2
`
`MT =
`
`8LDATA
`
`TDATA + TACK + TDIFS + TSIFS + (CWmin – 1)Tslot/2
`
`500
`Payload size (bytes)
`(a)
`
`1000
`
`1500
`
`50
`100
`150
`(ii) Data rate (Mb/s)
`(Payload = 1500 bytes)
`
`200
`
`500
`1000
`(iv) Payload (bytes)
`(Data rate = 216 Mb/s)
`
`1500
`
`0.7
`
`0.6
`0.5
`0.4
`
`0.3
`0.2
`
`Normalized overhead
`
`1
`
`0.95
`
`0.9
`
`0.85
`
`0.8
`
`0.75
`0
`
`Normalized overhead
`
`(b)
`
`50
`100
`150
`(i) Data rate (Mb/s)
`(Payload = 100 bytes)
`
`200
`
`500
`1000
`(iii) Payload (bytes)
`(Data rate = 6 Mb/s)
`
`1500
`
`80
`
`70
`
`60
`
`50
`
`40
`
`30
`
`20
`
`10
`
`Maximum throughput (Mb/s)
`
`0
`
`0
`
`ormalized overhead
`
`0.95
`0.9
`0.85
`0.8
`0.75
`0.7
`0.65N
`
`1
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0
`
`Normalized overhead
`
`n Figure 1. a) The MT and TUL for IEEE 802.11a; b) normalized overhead vs.
`data rate and payload.
`
`84
`
`IEEE Wireless Communications • December 2005
`
`Exhibit 1033
`Panasonic v. UNM
`IPR2024-00364
`Page 4 of 11
`
`

`

`XIAO LAYOUT 12/1/05 12:13 PM Page 85
`
`The wireless LAN
`standards, ETSI
`BRAN HIPERLAN/2
`and IEEE
`802.11a/11g,
`offer transmission
`rates up to 54 Mb/s
`in the 5 GHz or
`2.4GHz band. The
`two standards differ
`primarily on the MAC
`layer. The actual
`throughputs achieved
`are also highly
`dependent on MAC
`protocols.
`
`2 ms
`
`MAC frame
`
`MAC frame
`
`MAC frame
`
`MAC frame
`
`BCH
`
`FCH
`
`ACH
`
`DL phase
`
`DiL phase
`(optional)
`
`UL phase
`
`RCHs
`
`Train
`
`SCH
`
`SCH
`
`LCH
`
`LCH
`
`LCH
`
`Type
`
`INFO
`
`CRC
`
`Type
`
`SN
`
`CL
`
`DATA
`
`CRC
`
`4 bits
`
`52 bits
`
`2 bytes
`
`2 bits
`
`10 bits 12 bits
`
`48 bytes
`
`3 bytes
`
`9 bytes
`
`54 bytes
`
`(a)
`
`TO = TBCH+Pre + 26NDown + 32NUP + TACH
`+ TRCH+Pre + (NDown + 2NUP)TSCH + TTurn
`
`ThroughputHIPERLAN/2 =
`
`[(2000 – TO)/TLCH] · 48 · 8
`
`2000
`
`(b)
`
`(b/s)
`
`n Figure 2. a) HIPERLAN/2 MAC frame; b) equations for overhead and throughput.
`
`According to our previous contributions in [2],
`the MT and TUL are shown in Fig. 1a.
`Figure 1a shows the MT and TUL for IEEE
`802.11a. As illustrated in the figure, the TUL
`upper bounds the MT at a 54 Mb/s data rate and
`the MT at a 216 Mb/s data rate. When the pay-
`load size is 1500 bytes, the TUL is about 75.24
`Mb/s. The existence of the TUL shows that by
`simply increasing the data rate without reducing
`overhead, the enhanced throughput is bounded
`even when the data rate goes infinitely high. In
`other words, reducing overhead is necessary for
`IEEE 802.11 standards to achieve higher
`throughput.
`Figure 1b(i) and (ii) show normalized over-
`head vs. data rate. The normalized overhead
`increases as the data rate increases. The normal-
`ized throughput almost reaches 1 after 180 Mb/s
`when the payload size is 100 bytes. The normal-
`ized throughput reaches 70 percent after 180
`Mb/s when the payload size is 1500 bytes. Figure
`1b(iii) and (iv) show normalized overhead vs.
`payload size. The normalized overhead decreas-
`es as the payload size increases. The normalized
`throughput almost reaches 1 when the payload
`size is very small.
`In summary, the normalized overhead is
`extremely large when either the data rate is high
`or the frame is small.
`
`HIPERLAN/2
`The wireless LAN standards, European
`Telecommunications Standards Institute (ETSI)
`Broadband Radio Access Network (BRAN)
`HIPERLAN/2 [10] and IEEE 802.11a/11g, offer
`transmission rates up to 54 Mb/s in the 5 GHz
`or 2.4 GHz band. In this section we compare
`IEEE 802.11a and HIPERLAN/2 in terms of
`throughput upper limit. The two standards differ
`primarily in the MAC layer. The actual through-
`puts achieved are also highly dependent on
`MAC protocols. HIPERLAN/2 employs central-
`ized control, where a scheduler at an AP allo-
`cates resources in a MAC frame. Another
`difference in MAC protocols is the packet length
`adopted: HIPERLAN/2 adopted fixed length
`packets, and IEEE 802.11 adopted variable
`length packets.
`In HIPERLAN/2 a MAC frame is transmit-
`ted in a period of 2 ms (Fig. 2a). Each MAC
`frame comprises time slots for broadcast control
`(BCH), frame control (FCH), access feedback
`control (ACH), data transmission in downlink
`(DL), direct link (DiL), and uplink (UL) phases,
`and random channels (RCHs). DL and UL are
`used when data has to be transmitted. DL, UL,
`and DiL phases consistent of two types of proto-
`col data units (PDUs): the short transport chan-
`
`IEEE Wireless Communications • December 2005
`
`85
`
`Exhibit 1033
`Panasonic v. UNM
`IPR2024-00364
`Page 5 of 11
`
`

`

`XIAO LAYOUT 12/1/05 12:13 PM Page 86
`
`To overcome
`overhead of the IEEE
`802.11 MAC, we
`propose several
`efficient MAC
`enhancements, in
`which we adopt
`frame aggregation
`concept. The idea of
`frame aggregation is
`to aggregate
`multiple MAC/PHY
`frames/payloads
`into a single (or
`approximately
`single) transmission.
`
`nel (SCH) and long transport channel (LCH).
`SCHs are for control data and have a size of 9
`bytes. LCHs are for normal data and have a size
`of 54 bytes, including 6 bytes overhead and 48
`bytes payload. A train of SCH and LCH packets
`is transmitted in DL and/or UL. The duration of
`BCH is fixed; others are not. When transmitted,
`there is a physical preamble for the whole MAC
`frame, DL, DiL, UL, and RCH, respectively. We
`do not consider DiL since it is optional and not
`important for conclusions.
`Let TBCH+Pre, TFCH, TACH, TRCH+Pre, TSCH,
`TLCH, TDPre, and TUPre denote the transmission
`times for a BCH with the preamble, FCH, ACH,
`RCH with preamble, SCH, LCH, preamble in
`DLs, and preamble in ULs, respectively. Let TCD
`and TCU denote the transmission times for con-
`trol signal in DL (an SCH for ACK per session)
`and control signal in UL (the SCH for ACK and
`resource reservation per session), respectively.
`Let TGU, TTurn, and TO denote guard time in
`UL, radio turnaround time, and total overhead
`time, respectively. Data rates are 6, 9, 12, 18, 27,
`36, and 54 Mb/s, and the corresponding TLCH
`are 72, 48, 36, 24, 16, 12, and 8 µs, respectively.
`Let NDown and NUp denote the number of ses-
`sions in UL and DL, respectively. From [10] we
`have TFCH = [(NDown + NUp)/2] 36 µs, TCD =
`NDownTSCH, TCU = 2NUpTSCH, TDPre = NDown 8
`µs, TUPre = NUp 12 µs, and TGU = NUp 2 µs. We
`have overhead and throughput shown in Fig. 2b.
`From [10] we have TBCH+Pre = 36 µs, TACH = 12
`µs, TRCH+Pre = 28 µs, and TTurn = 12 µs. We
`obtain overhead and throughput shown in Fig.
`2b. When both the data rate and control rate go
`infinitely high, we have TSCH = 0 and TLCH = 0.
`Furthermore, when the data rate goes infinite,
`the total overhead TO will be a fixed number,
`and ThroughputHIPERLAN/2 will be infinitely large
`(∞). In other words, the throughput upper limit
`of HIPERLAN/2 is ∞.
`IEEE 802.11 employs a protocol similar to the
`stop-and-wait protocol so that every packet is
`acknowledged. In each transmission cycle there is
`a fixed overhead time, including spacing, ACK,
`and other overhead, independent of the data rate.
`For a given payload size, no matter how high the
`data rate is, the transmission time is larger than
`the fixed overhead time. Therefore, IEEE 802.11a
`has a throughput limit, which is 75.24 Mb/s when
`the payload size is 1500 bytes. On the other hand,
`HIPERLAN/2 employs a MAC frame (transmit-
`ted in 2 ms) in which a train of LCHs can be
`transmitted within 2 ms. As the data rate goes
`infinitely high, the number of LCHs that can be
`potentially transmitted is infinite. Therefore,
`HIPERLAN/2’s throughput upper limit is ∞. In
`other words, HIPERLAN/2 does not have a
`throughput upper limit. The conclusion is that
`HIPERLAN/2 is much more scalable than IEEE
`802.11 for much higher data rates.
`IEEE 802.11 MAC ENHANCEMENTS
`To overcome the overhead of the IEEE 802.11
`MAC, we propose several efficient MAC
`enhancements in which we adopt the frame
`aggregation concept. The idea of frame aggrega-
`tion is to aggregate multiple MAC/PHY frames/
`payloads into a single (or approximately single)
`
`transmission. We classify frame aggregation
`mechanisms into many different and orthogonal
`aspects:
`• Distributed vs. centrally controlled: Frame
`aggregation can be used under both the con-
`tention-based DCF and the contention-free
`PCF. The former is a distributed mechanism,
`and the latter is centrally controlled.
`• Ad hoc vs. infrastructure: There are two types
`of 802.11 networks: infrastructure (BSS) in
`which an AP is present, and ad hoc (IBSS) in
`which an AP is not present. Frame aggrega-
`tion can be used in both ad hoc and infra-
`structure networks.
`• Uplink vs. downlink: In infrastructure net-
`works, frame aggregation can be used by both
`UL and DL transmissions. UL transmissions
`are those from stations to the AP, and DL
`transmissions are those from the AP to sta-
`tions.
`• Single-destination vs. multi-destination:
`Under some frame aggregation mechanisms,
`frames can be aggregated only if they have the
`same destination (MAC/PHY) addresses.
`Under other proposed frame aggregation
`mechanisms, frames can be aggregated even if
`they have different destination (MAC/PHY)
`addresses.
`• PHY-level vs. MAC-level: Frames can be aggre-
`gated at both the PHY and MAC levels. At
`the PHY level, contents of PHY frames retain
`integrity. At the MAC level, PHY frames are
`changed, and MAC frames may or may not be
`changed. However, payloads of MAC frames
`cannot be changed.
`• Single-rate vs. multirate: Aggregated frames
`can use the same transmission rate, and can
`use different transmission rates for multi-des-
`tination frame aggregation.
`• Immediate ACK vs. delayed ACK: Aggregated
`frames can be acknowledged via separate
`ACK frames immediately following a SIFS
`time, or a single delayed group ACK frame
`can be adopted after aggregated frame trans-
`missions.
`• No spacing vs. SIFS spacing: Under delayed
`ACK frame aggregation, two consecutive
`aggregated PHY frames may be separated by
`nothing or a SIFS time.
`Frame aggregation mechanisms have many
`benefits. First of all, since transmitting longer
`frames may lead to better throughput than trans-
`mitting shorter frames, by adopting these mecha-
`nisms the system can achieve the throughput of
`transmitting longer frames. The second and most
`important benefit is that these mechanisms can
`reduce overhead. Without these mechanisms, each
`frame transmission needs a separate set of over-
`head (headers, IFSs, backoff time, and/or ACKs).
`With these mechanisms, instead of several sets of
`overhead for different frames, only one set of over-
`head will be used. Finally, these mechanisms can
`reduce the average delay. Without these mecha-
`nisms, the second or a later frame is transmitted at
`a much later time. With these mechanisms, it is
`transmitted at almost the same or earlier.
`One issue is how long the total length of
`aggregated frames should be. One possible solu-
`tion is that the number of aggregated frames
`should not be larger than a threshold, and the
`
`86
`
`IEEE Wireless Communications • December 2005
`
`Exhibit 1033
`Panasonic v. UNM
`IPR2024-00364
`Page 6 of 11
`
`

`

`XIAO LAYOUT 12/1/05 12:13 PM Page 87
`
`Frame aggregation
`mechanisms have
`many benefits.
`Without these
`mechanisms, each
`frame transmission
`needs a separate set
`of overhead. With
`these mechanisms,
`instead of several
`sets of overhead for
`different frames,
`only one set of
`overhead will
`be used.
`
`Time
`
`Time
`
`Time
`
`Time
`
`Indicate the beginning of frame
`aggregation in subtype field
`
`Indicate the end of frame
`aggregation in subtype field
`
`Busy
`
`DIFS Random
`backoff
`
`PHY
`Frame 1
`
`PHY
`Frame 2
`
`PHY
`Frame k SIFS ACK
`
`(a) PHY level version 1: distributed, no spacing,
`delayed ACK, single-destination, single-rate
`
`Busy
`
`DIFS Random
`backoff
`
`PHY
`Frame 1
`
`PHY
`Frame 2
`
`PHY
`Frame k SIFS
`
`ACK
`request
`
`SIFS ACK
`
`(b) PHY level version 2: distributed, no spacing,
`delayed ACK, single-destination, single-rate
`
`PHY
`Frame 2
`
`PHY
`Frame 2
`
`PHY
`Frame k SIFS ACK
`
`Header
`
`DIFS Random
`backoff
`
`Busy
`
`(c) PHY level version 3: distributed, no spacing,
`delayed ACK, single-destination, single-rate
`
`Busy
`
`DIFS Random
`backoff
`
`PHY
`Frame 1
`
`SIFS
`
`PHY
`Frame 1
`
`SIFS
`
`PHY
`Frame k SIFS ACK
`
`(d) PHY-level version 4: distributed, SIFS-spacing,
`delayed-ACK, single-destination, single-rate
`
`Busy
`
`DIFS Random
`backoff
`
`PHY
`Frame 1
`
`SIFS
`
`PHY
`Frame 1
`
`SIFS
`
`PHY
`Frame k SIFS
`
`ACK
`request
`
`SIFS ACK
`
`(e) PHY level version 5: distributed, SIFS spacing,
`delayed ACK, single-destination, single-rate
`
`Time
`
`Busy
`
`DIFS Random
`backoff
`
`PHY
`Frame 1
`
`SIFS ACK SIFS
`
`PHY
`Frame 2
`
`SIFS ACK SIFS
`
`Time
`
`Time
`
`Time
`
`Time
`
`(f) PHY level version 6: distributed, immediate ACK,
`single-destination, single-rate
`
`header
`MAC
`header
`
`Random
`backoff
`
`PHY
`
`SIFS ACK
`
`tail
`PHY
`
`MAC
`Frame k
`
`Lk
`
`MAC
`Frame 2
`
`L2
`
`MAC
`Frame 1
`
`L1
`
`(g) MAC level version 1: distributed, single-destination,
`single-rate
`
`SIFS
`
`ACK
`
`tail
`PHY
`
`MAC
`Frame k
`
`MAC
`Frame 1
`
`MAC
`Frame 2
`
`header
`
`Random
`backoff
`
`PHY
`
`(h) MAC level version 2: distributed, single-destination,
`single-rate
`
`SIFS ACK
`
`header
`
`PHY
`FCS
`
`Payload
`
`k
`
`Lk
`
`Payload
`
`2
`
`L2
`
`Payload
`
`1
`
`L1
`
`MAC
`header
`
`header
`
`Random
`backoff
`
`PHY
`
`Busy
`
`DIFS
`
`Busy
`
`DIFS
`
`Busy
`
`DIFS
`
`(i) MAC level version 3: distributed, single-destination,
`single-rate
`
`A single PHY frame
`
`n Figure 3. Distributed, single-destination, and single-rate frame aggregation.
`
`IEEE Wireless Communications • December 2005
`
`87
`
`Exhibit 1033
`Panasonic v. UNM
`IPR2024-00364
`Page 7 of 11
`
`

`

`XIAO LAYOUT 12/1/05 12:13 PM Page 88
`
`In single-destination
`approaches, frames
`can be aggregated if
`they are available,
`and have the same
`source and
`destination
`addresses. The total
`length of aggregated
`frames should be
`smaller than a
`threshold, which is
`called aggregation
`threshold.
`
`total length of aggregated frames should be
`smaller than another threshold, which is smaller
`than or equal to the fragmentation threshold.
`The purpose of these mechanisms is not to build
`a huge frame, but a reasonably sized frame since
`huge frames may have a bad effect on fairness
`and/or efficiency. Furthermore, frame aggrega-
`tion is not a reversed mechanism of fragmenta-
`tion. In fact, the proposed aggregated
`mechanisms require that the total length of the
`aggregated frames be smaller than the fragmen-
`tation threshold. Therefore, there will be no
`aggregated frame that was originally generated
`by a previous fragmentation mechanism. On the
`other hand, an aggregated frame will not be
`fragmented since the total length is smaller than
`the fragmentation threshold. Next, we discuss
`some frame aggregation mechanisms.
`DISTRIBUTED, SINGLE-DESTINATION, AND
`SINGLE-RATE FRAME AGGREGATION
`Distributed, single-destination, and single-rate
`frame aggregation mechanisms are shown in Fig. 3,
`where Figs. 3a, 3b, 3c, 3d, 3e, and 3f are PHY-level
`frame aggregation methods, and Figs. 3g, 3h, and
`3i are MAC-level frame aggregation methods. At
`the PHY level contents of PHY frames remain
`unchanged, and at the MAC level PHY frames are
`changed; MAC frames may or may not be changed.
`Note that the payloads of MAC frames cannot be
`changed. In single-destination approaches, frames
`can be aggregated if they are available, and have
`the same source and destination addresses. The
`total length of aggregated frames should be smaller
`than a threshold, called the aggregation threshold.
`Figure 3a shows that k PHY frames are sent
`one by one with no spacing between PHY frames.
`The first aggregated frame (PHY frame 1) should
`indicate the beginning of the frame aggregation in
`one of the subtype fields, and the last aggregated
`frame (PHY frame k) should indicate the end of
`frame aggregation, followed by a SIFS time and
`an ACK from the destination station. The desti-
`nation can identify the PHY boundaries via PHY
`preambles. The ACK frame should have a bitmap
`field, in which each bit can be used to acknowl-
`edge one frame: a bit value of 1 stands for suc-
`cessfully receiving a frame; it is 0 otherwise.
`Figure 3b has minor differences from Fig. 3a
`as follows. Instead of indicating the end of frame
`aggregation in the last aggregated frame (PHY
`frame k), an ACK request frame should send
`directly from the source, and the destination
`respond with an ACK frame after SIFS time. This
`scheme adds additional overhead, but the system
`becomes more robust since if the last aggregated
`frame (PHY frame k) in the scheme in Fig. 3a is
`lost, the destination does not obtain an explicit
`indication that an ACK needs to be sent back.
`Figure 3c shows another variant of Fig. 3a:
`instead of using the first aggregated frame to indi-
`cate the beginning of the frame aggregation, anoth-
`er header frame is transmitted earlier in which
`more information is included such as the number
`(1 byte) of aggregated frames to follow and the
`total length/duration (2 bytes). The drawback is
`that an additional header PHY frame is needed.
`Figure 3d is another variant of Fig. 3a, and
`Fig. 3e is a variant of Fig. 3b. In both Figs. 3d
`
`and 3e, a SIFS time is used to separate PHY
`frames. A drawback is that additional overhead
`of SIFS time is needed. An advantage is that the
`schemes make PHY frames’ boundaries clearer.
`Figure 3a, 3b, 3c, 3d, and 3e all adopt delayed
`ACK, whereas Fig. 3f adopts immediate ACK.
`The drawback of Fig. 3f is that it has more over-
`head. One advantage is that it becomes more
`robust under a noisy channel.
`Roughly speaking, comparing schemes in Figs.
`3a–3f, we observe that if a scheme has more over-
`head, it becomes less efficient, but is more robust
`in terms of error control, explained as follows.
`The scheme