WLAN evolution from HIPERLAN type 2 to MEDIAN
`
`S. Zeisberg, M. Bauling, A. Finger
`Communications Laboratory, Dresden University of Technology
`Institut fiir Nachrichtcntechnik, Mommsenstr. 13, 01069 Dresden
`Email: zei@ifn.et.tu-dresdcn.dc
`
`Abstract: HIPERLAN type 2 system being currently
`in the final standardisation process in ETSI BRAN
`is expected to operate in the 5.15 - 5.35 GHz or
`5.470 - 5. 725 GHz or 5.725 - 5.925 GHz frequency
`range while the JMEDIAN system to be standardised
`in the near fiiture is expected to work at least in
`fi‘equency bandfrom 61.0 to 61.5 GHz allocated to
`Industrial, Scientific and Medical applications
`(ISM. Both systems are based an orthogonal
`frequency division modulation (OFDM) physical
`layer scheme and will support quality of service
`requests for ATM as Well as for 11’ based networks.
`Cost eflicient
`terminal development and broad
`market acceptance can be supported by designing
`systems in a compatible and evolutionary way. A
`renewed MEDIAN system design is presented here,
`based on experience made in the ACTS project
`AC006 MDMN considering the evolutionary
`approach regarding recent HIPERLAN type 2 /
`IEEE 802.11a harmonised standardisation actions.
`
`1.
`
`INTRODUCTION
`
`2
`the HIPERLAI’\T type
`in
`changes
`Recent
`standardisation due to harmonisation with the IEEE
`802.11a
`show the
`need
`of world wide
`harmonisation in system design and development in
`order
`to ensure mass market production for
`component manufacturer. Decreasing production
`costs due to high volumes enable fast market
`introduction because of cost dependent higher
`customer acceptance.
`
`MEDIAN demonstrator has been designed and
`build and experience shows.
`that
`the technical
`approach of the MEDIAN project
`to transmit
`wireless gross data rates of more than 200 Mbitls at
`60 GHZ radio has been realised successfully. The
`demonstrator consists of one base station and two
`
`mobile stations each consisting of physical layer,
`media access control and interworking units. User
`data up to 150 Mb it!s can be transmitted. MEDIAN
`demonstrator has been designed and built using
`market place available technology. For this reason
`many choices in demonstrator system design had to
`be
`far
`from optimal. After
`finishing
`the
`demonstrator design a future MEDIAN system
`design has started. This design incorporates more
`
`trying to keep a certain
`optimal solutions but
`compatibility to the latest HIPERLAN type 2
`development. Performance of the HJPERLANl2
`and new MEDIAN system is determined by
`computer simulations incorporating measured local
`oscillator phase noise and radio channel data.
`
`2. HIPERLAN! 2 T0 MEDIAN DESIGN
`
`layer
`physical
`2
`Current HIPERLAN type
`parameters are 20 MHz channel spacing and sample
`rate, COFDM modulation
`scheme,
`64
`point
`complex FFT,
`48
`subcarrier
`used for
`data
`transmission, 800 ns guard time, BPSK, QPSK,
`l6 QAM and 64 QAM subcarrier modulation using
`coherent demodulation. Forward error correction is
`
`applied using standard convolutional code with
`Constraint length 7 and code rate 0.5. According to
`map one PDU onto integer numbers of OFDM
`symbols the mother code will be punctured to
`required code rates of gI].E and n. The DLC-PDU
`length is 54 bytes. The ideas presented here base on
`use as much as possible components of the
`HlPERLANtZ system for
`the future MEDIAN
`system. There are physical
`layer parts like FFT,
`subcarrier modulation, coding and digital filters, the
`PDU length of 54 bytes and upper layer protocols
`to be re-useable for MEDIAN. Two new MEDIAN
`
`design discussions shall consider what is possible
`and what is convenient.
`
`3. First MEDIAN Design Discussion
`The first approach is to use the HIPERLAN/2
`design with a sampling rate of 100 Msamplesls
`instead of 20 Msamplesls. Thus a guard time of 160
`us would be achieved. Due to the less dispersive
`radio channel at 60 GHz a guard time of 160 ns is
`presumed to be sufficient [3]. Using HIPERLAN/2
`equipment
`in 16 QAM mode with 5 times the
`original sample rate can provide a physical layer
`fulfilling I‘vaDIAN data rate requirement. This
`provides a gross data rate of 240 Mbitls. Applying a
`forward error correction with code rate 0.625
`
`enables to carry one DLC-.PDU in four OFDM
`symbols (768 bit). Thus'the net bit rate is 150 Mbps
`for 16QAM. That subcarrier modulation for normal
`and QPSK for bad channels is
`foreseen. The
`forward error correction to be applied in the future
`
`c7803-5435-4199rsiooo o 1999 IEEE
`
`2656
`
`VTC ‘99
`
`Microsoft Corporation
`
`Exhibit 1015-00001
`
`Microsoft Corporation
`
`

`

`MEDIAN system needs to be more sophisticated
`than the standard scheme proposed for HIPERLAN
`type 2 due to the minimum code rate 0.625. Thus
`the TURBO coding scheme is foreseen, which is
`not compatible to the current HIPERLAN/Z. This
`point is under investigation in order to identify as
`much as possible decoder elements to be used in
`both systems.
`
`This very simple approach to implement a new
`MEDIAN system would save a lot of development
`time and costs but there are some disadvantages
`that should be considered. Thus the use of only 64
`subcarrier in a frequency band of 100 MHz seems
`to be a waste of this precious resource. Because of
`the slowly changing MEDIAN indoor radio channel
`the orthogonality of the subcarriers is not greatly
`disturbed by Doppler shifts and this allows a closer
`subcarrier spacing.
`
`is the 16QAM used to reach the
`Another point
`target net bit rate of [50 Mbps for wireless ATM.
`This modulation method is very sensitive in fading
`and frequency selective channel environment and
`needs a channel estimation before demodulation.
`
`To get sufficient results a relative high SNR is
`required. If differential encoded QPSK or 8PSK
`modulation is used there is no need for channel
`
`estimation. Additionally this modulation methods
`are more robust in radio channel environment and
`
`need a lower SNR to reach comparable error rates.
`Since the nonlinearities of the RF front end high
`power amplifier (I—[PA) are more critical in 60 GHz
`region the crest factor (CF) reduction requires more
`effort then in HIPERLAN/Q. Here a system with
`only 64 subcarriers shows a CF distribution that
`more often reaches higher values than the Rayleigh
`pdf valid for infinity number of subcarriers. A CF
`reduction coding is not applied due to the relatively
`high number of used subcarriers. In a system with
`even more subcarriers the amplitudes of the inphasc
`(I) and quadrature {0) signal components in time
`domain reaches nearly a Gaussian distribution. That
`results in a reduced probability of higher CF5.
`Because of the above mentioned problems with the
`simple transformation of the HIPERLAN/2 system
`into a MEDIAN system it is not favoured for a new
`MEDIAN design. But it
`is maybe an interesting
`alternative until a standard is adopted.
`
`4.
`
`SECOND MEDIAN DESIGN DISCUSSION
`
`to change
`A more sophisticated design needs
`number of subcarriers in one OFDM-symbol. Other
`parameters
`like PDU length and upper
`layer
`protocols are assumed still similar to HIPERLAN/Z
`
`system concept. Vv’ith a number of four times the
`number of subcarriers of the HIPERLAN/Z system
`the efficient radix 4 algorithm to get a 256 point
`FFT can be used. Due to the increased number of
`
`in time
`subcarriers the magnitude of the signal
`domain fits better a Rayleigh distribution. By this
`reason the average Crest Factor
`is
`lower
`in
`comparison to the system with the 64 point FFT.
`On the other hand the system gets more sensitive
`for phase noise and Doppler shifts. The influence of
`the noise introduced by the local oscillators is
`investigated by simulations with the given phase
`noise mask in Figure 3a.
`
`For subcarrier modulation differentially encoded
`PSK is chosen because of the robustness in fading
`environment and the possibility of simple non-
`coherent demodulation. The coding scheme is
`changed to a code with minimum code rate of 1/3.
`A detailed description of the new MEDIAN system
`is given in the following sections.
`
`Physical Layer Description:
`
`The basic functions of the new MEDIAN physical
`layer are scrambling / descrambling, ending /
`decoding, interleaving / deinterleaving, subcarrier -
`modulation
`/
`demodulation with
`differential
`
`encoding, OFDM modulation / demodulation and
`time and frequency synchronisation.
`
`Scrambling / Descrambling and Interleaving /’
`Deinterleaving:
`is
`used as proposed for
`the
`HIPERLAN/Z system [5]
`
`Coding / Decoding: In this section a coder with the
`coding rate R=l/3 is foreseen. All needed code
`rates in the different physical
`layer modes are
`obtained by puncturing the mother code. For
`decoding soft decision decoding is foreseen with
`the technical progress in mind providing the ability
`for iterative decoding.
`Subcarrier Modulation
`
`/ Demodulation:
`
`The
`
`OFDM subcarriers are modulated by differential
`encoded QPSK (DQPSK) or DSPSK between
`adjacent subcarriers in the same OFDM symbol.
`For this reason two phase reference subcarriers are
`introduced per OF DM symbol.
`Due tot differential modulation with noncoherent
`demodulation no channel
`estimation is used.
`
`Coherent demodulation will be investigated further
`for its advantage in SNR requirement. Decoding
`with channel state information (CSI) is foreseen.
`The subcarrier modulation dependent parameters of
`the potential
`future MEDIAN basic modes are
`listed in the following table.
`
`0-7303-5435-4/991'51000 c 1999 IEEE
`
`2657
`
`VTC ‘99
`
`Microsoft Corporation
`
`Exhibit 1015-00002
`
`Microsoft Corporation
`
`

`

`
`
`
`
`
`
`
`
`OFDM symbols
`I Data bits per
`Net bit rate
`Modulation Coding rate R Coded bits per
`per PDU
`OFDM symbol
`OFDM symbol
`
`
`
`
`DQPSK
`26
`2l6x2= 432
`
`D8PSK
`
`
`216x3— 64s __ SOMbps
`Table 1: Physical layer modes ofpotentialfuture iUEDMN system.
`
`V23
`
`
`
`/ Demodulation:
`
`OFDM Modulation
`
`After
`subcarrier modulation the complex symbols are
`mapped to an OFDM symbol. The general structure
`is shown in Figure 1. For the data there are 216
`subcarriers
`and
`for
`reference
`symbols
`two
`subcarriers in each OFDM symbol. The DC carrier
`and the band edge carriers are not used and set to
`zero. These carriers are used to cope with adjacent
`
`and out of band
`(AC1)
`interference
`channel
`emissions. In the time domain. a cyclic prefix with
`32 samples is added to the actual OFDM samples to
`compensate for the intersymbol
`interference (ISI)
`introduced by the multipath channel. The length of
`the guard interval is equal
`to 320 us. To reach a
`slightly higher throughput
`rate a shorter guard
`interval is possible.
`
`R2 D109
`
`D216
`
`R1 D1
`
`D108
`
`—‘l
`
`Dar 7 complex data symbols
`
`Rx - phase reference symbols
`DC - unused DC-subcarrier
`
`Ex - unused band edge subcarn‘er
`
`Figure 1: Subcarrierfrequency allocation
`
`The band of 500 MHz in the range 61.0 — 61.5 GHz
`which is currently foreseen for MEDIAN could be
`divided in five channels with 100 MHZ bandwidth.
`Then the distance between the outmost carriers of
`
`amplifier output backoff requirements, the absolute
`transmission power and decisions about spurious
`power emission limits outside the range 61.0 —
`61.5 GHz this value can be increased.
`
`two adjacent channels is 30.469 MHz. The distance
`to the edges of the MEDIAN frequency band is
`only a half this value. Depending on power
`
`The values for mean OFDM parameters of the
`created new MEDIAN system are given in the
`following table.
`
`Parameter
`
`
` Value
` 100 MHz
`Sampling rate fSZUT
`256
`
`FFT complex points
`
`
`256-1" = 2.56 ps
`
`Symbol part duration Tu
`
`Guard interval duration Tu
`32-T = 0.32 as
`
`
`
`
`OFDM symbol interval T5
`288T —*— 2.88 as
`
`216
`Number of data subcarriers
`
`
`Number ol‘ reference symbol snbcarrier
`
`_Subcarrier spacing
`
`
` 85.15625 MHz !
`Spacing between the two outmost subcarriers
`
`
`390.625 kHz
`
`Table 2: Second MEDIAN OFDMparameter set.
`
`o7m3-5435—41991510mc) 1999 WEE
`
`2658
`
`VTC ‘99
`
`Microsoft Corporation
`
`Exhibit 1015-00003
`
`Microsoft Corporation
`
`

`

`5. PERFORMANCE SIMULATIONS
`
`At this point of investigation several HIPERIAN/Z
`simulations and some MEDIAN simulations with
`
`the second MEDIAN design draft have been carried
`out. The PDU (54 byte) error rate for (D)QPSK and
`16QAM is depicted in Figure 2. The channel model
`used for both simulations
`(HIPERLAN/2 and
`MEDIAN) is Model A according to document [2].
`In this time and frequency selective channel model
`the doppler shifts are introduced according to a
`velocity of
`3 mfs and the carrier
`frequencies
`5.2 GHZ or 61 GHz.
`In Figure 2a HIPERLANJ'.
`and MEDIAN results for (D)QPSK are shown in
`one
`diagram.
`16QAM is
`only
`foreseen
`in
`I-HPERLAN/2 and is depicted in the diagram of
`Figure 2b. The different code rates are the result of
`mapping one PDU in an integer number of OFDM
`symbols.
`
`For QPSK channel model A is the worst case in
`performed HIPERLAN/Z simulations. That is due
`to the bad frequency selective behaviour of this
`model. There are large and deep fades in the
`frequency domain that can hardly corrected with
`the implemented code. The influence of the channel
`state information (CSI)
`is clearly visible. Soft
`
`1.00E+ii0a....
`
`Decision with C81 gives the best results but the
`technical burden should be taken in consideration.
`
`The DQPSK curve of the MEDIAN simulations is
`degraded in comparison to coherent QPSK, but in
`addition
`to
`the
`noncoherent
`differential
`
`demodulation the time selective fading is strongly
`increased. That is due to the fact of the increased
`
`subcarrier number. At a PDU error rate equal to 10'
`3 the difference between the sofi decision curves is
`
`is a rather good result. The better
`|.3 dB. That
`frequency selective behaviour in the case of more
`used subcarriers is one point that should mentioned
`here.
`In the case of a channel model especially
`designed for the 60 GHz channel and even with a
`strong direct path the PDU error
`rate can be
`improved further. Only such a channel model
`allows
`expressive
`statements
`about
`real
`transmission error rates. With the given Model A
`only the HlPERLANJZ and the new MEDIAN draft
`can be compared.
`
`For 16QAM the increased Ebi’No ratio should be
`pointed
`out.
`That
`is
`combined with
`the
`disadvantage of a needed amplitude correction of
`the received signal. By this remains that modulation
`method is not foreseen in the new MEDIAN design.
`
`PDU error tale, 16QAM, R=3l4
`
`i
`
`LUCIEHJU
`
`PDU error rate. (DlnPSK. R=112
`
`..
`
`
`#- sm Cac MECIAN [00%
`SaflDec HPERM
`LSoflDscCSlHlFERLANIQ
`
`
`
`
`
`tout-m
`
`-- ~
`
`
`
`1.00E-01
`
`
` a = ..
`LUBE-051
`i- —v—r
`l
`-i
`l ,\ ‘v—
`r—E'——I
`
`0
`
`2
`
`4
`
`6
`
`5
`
`l0
`
`i2 H 16
`Eb/NOlI'IdB
`
`19
`
`20
`
`2'2
`
`24
`
`25
`
`
`
`“s—i
`
`28
`
`30
`
`tour-oz ;.. 5
`
`
`
`
`
`
`::
`
`1.00502 -
`
`-‘ . a;
`p
`+HaMDac
`+ihrdDe¢ GSI
`SollDac
`noASollDacCS
`tnut-na *i—I'r
`024681012141513202224262330
`attends
`
`a) HIPERLAN/Z and new MEDIAN results
`
`b) HIPERLAN/2 results
`
`Figure 2: PERfor a transmission over HIPERLAN/Z channel A [2]
`
`the local
`Using the _6t| GHz channel modeIS,
`oscillator model, and the high power amplifier
`
`model derived in the MEDIAN AC006 project error
`rate performance simulations are performed.
`
`0-7803A5435-4/99!$10.UD © 1999 IEEE
`
`2659
`
`VTC ‘99
`
`Microsoft Corporation
`
`Exhibit 1015-00004
`
`Microsoft Corporation
`
`

`

`
`
`
`
`
`is
`.5-
`
`u
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`o
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`lNPlfI|fC[(13de 0-1th
`a 7'2“
`71m
`
`
`Favm{dam}Phlnalumina(dog
`
`
`410
`
`
`
`.
`
`
`430
`i
`' J
`
`
`it:a L—
`i
`|
`L
`‘
`l
`l
`-‘l‘D
`‘ _J
`1.
`‘ eons
`thus
`1 am
`I sour
`1.5.05
`1.Ev03
`1.5::
`
`5
`D
`5
`40
`an
`45
`an
`7:5
`
`
`' [HZ]
`Inptfl Fawel [dam]
`
`
`
`
`
`
`a) Phase noise mask of 56GHz local oscillator
`b) AM—AM and AM-PM conversion of 60GHz
`used in the MEDIAN demonstrator (approximated
`high power amplifier (HPA) used in the MEDIAN
`
`
`
`PSD curve based on measurements by Dassault
`demonstrator (model based on measurements by
`
`
`
`Electronique)
`Dassault Electronique)
`
`
`
`
`
`
`
`
`
`(1') Impulse response magnitude considered typical
`for the 60GH2 wireless indoor channel with non
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`c) Impulse response magnitude considered typical
`for the 60GHZ wireless indoor channel with line of
`
`
`(LOS). GZGHZ carrier
`sight
`measurement bandwidth
`
`frequency, 2GHz
`
`line of sight (NLOS). 62GI—lz carrier frequency,
`EGHZ measurement bandwidth
`
`Figure 3: IWOdQISfor 60 GHz Simulations
`
`6. CONCLUSIONS
`A future MEDIAN system design has been
`introduced. The simulation system models used for
`system error
`rate performance simulations have
`been
`provided.
`Compatibility with
`current
`IHPERLAN/Z standardisation
`process
`enables
`evolutionary introduction of the future MEDIAN
`system.
`7. ACKNOWLEDGEMENT
`
`The authors would like to thank all MEDIAN team
`
`fruitful discussions and friendly
`for
`members
`cooperation. The contribution of the Commission of
`the European Community enabling the MEDIAN
`project research work is highly acknowledged.
`
`8. REFERENCES
`[1]
`R. I. Kopmeiners, P. Wijk: ETSI BRAN
`30701E, Criteriafor Comparison
`
`[2}
`
`[3]
`
`[4]
`
`l5]
`
`J. Mcdbo, P. Schramm: ETSI BRAN
`3ERIOSSB,
`Channel
`Models
`for
`HIPERLAN/E
`in
`Diflerent
`Indoor
`Scenarios.
`K.Koora,
`S.Zeisberg,
`J'Hfibnerj
`J'Borowski, A'Fingcr:
`"Simple Channel
`Model for 60 GHz indoor wireless LAN
`design
`based
`on
`complex widebancl
`measurements.“ IEEE Proc. of V’l‘C'97,
`Phoenix, May 4-7, 1997, pp.lOO4-1008.
`
`J.Borowski, S. Zeisberg, P. Legg: ETSI
`BRAN WGSTD48, The MEDIAN physical
`layer.
`2
`HIPERLAN TYPE
`BEAN,
`ETSI
`Functional Specification Part 1 — Physical
`(PH?) layer
`
`0-7803-5435—43951030 to 1999135};
`
`2660
`
`VTC ‘99
`
`Microsoft Corporation
`
`Exhibit 1015-00005
`
`Microsoft Corporation
`
`