Huawei Proposal Clarifications

Document Sample
Huawei Proposal Clarifications Powered By Docstoc
					       March 2006                                                                                                              doc.: IEEE 802.22-06/0050r0

                                        Huawei Proposal Clarifications
                               IEEE P802.22 Wireless RANs                                                                   Date: 2006-03-20
 Authors:
            Name                              Company                               Address                          Phone                                   email

 Linjun Lu                             Huawei Technologies                     Shenzhen, China            0086-755-28973119                 lvlinjun@huawei.com
 Soo-Young Chang                       Huawei Technologies                     Davis, CA, U.S.            1-916 278 6568                    sychang@ecs.csus.edu
 Jianwei Zhang                         Huawei Technologies                     Shanghai, China            86-21-68644808                    zhangjianwei@huawei.com
 Lai Qian                              Huawei Technologies                     Shenzhen, China            86-755-28973118                   qlai@huawei.com
 Jianhuan Wen                          Huawei Technologies                     Shenzhen, China            86-755-28973121                   wenjh@huawei.com
 Vincent K. N. Lau                             HKUST                         Hong Kong, China             852-2358-7066                     eeknlau@ee.ust.hk
 Roger S. Cheng                                HKUST                         Hong Kong, China             852-2358-7072                     eecheng@ee.ust.hk
 Ross D. Murch                                 HKUST                         Hong Kong, China             852-2358-7044                     eermurch@ee.ust.hk
 Wai Ho Mow                                    HKUST                         Hong Kong, China             852-2358-7070                     eewhmow@ee.ust.hk
 Khaled Ben Letaief                            HKUST                         Hong Kong, China             852-2358-7064                     eekhaled@ee.ust.hk


Notice: This document has been prepared to assist IEEE 802.22. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this
document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein.

Release: The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an IEEE Standards
publication; to copyright in the IEEE’s name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEE’s sole discretion to permit others to
reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this contribution may be made public by IEEE 802.22.

Patent Policy and Procedures: The contributor is familiar with the IEEE 802 Patent Policy and Procedures http://standards.ieee.org/guides/bylaws/sb-bylaws.pdf including the statement
"IEEE standards may include the known use of patent(s), including patent applications, provided the IEEE receives assurance from the patent holder or applicant with respect to patents essential
for compliance with both mandatory and optional portions of the standard." Early disclosure to the Working Group of patent information that might be relevant to the standard is essential to reduce
the possibility for delays in the development process and increase the likelihood that the draft publication will be approved for publication. Please notify the Chair
Carl R. Stevenson as early as possible, in written or electronic form, if patented technology (or technology under patent application) might be incorporated into a draft standard being developed
within the IEEE 802.22 Working Group. If you have questions, contact the IEEE Patent Committee Administrator at patcom@iee.org.

   Submission                                                                        Slide 1                                                               Jianwei Zhang, Huawei
   March 2006                                                         doc.: IEEE 802.22-06/0050r0

Co-Authors:
Name               Company                Address            Phone             email

Edward K. S. Au          HKUST            Hong Kong, China   852-2358-7086     eeedward@ee.ust.hk
Peter W. C. Chan         HKUST            Hong Kong, China   852-2358-7086     peter@ee.ust.hk
Ernest S. Lo             HKUST            Hong Kong, China   852-2358-7086     eeern@ee.ust.hk
Lingfan Weng             HKUST            Hong Kong, China   852-2358-7086     lingfan@ee.ust.hk
Zhou Wu             Huawei Technologies   Shenzhen, China    86-755-28979499   wuzhou@huawei.com
Jun Rong            Huawei Technologies   Shenzhen, China    86-755-28979499   rongjun@huawei.com
Jian Jiao           Huawei Technologies    Beijing, China    86-10-82882751    jiao_jian@huawei.com
Meiwei Jie          Huawei Technologies   Shenzhen, China    86-755-28972660   jiemingwei@hauwei.com




Submission                                   Slide 2                                   Jianwei Zhang, Huawei
  March 2006                  doc.: IEEE 802.22-06/0050r0




  Part 1: Channel Sensing




Submission          Slide 3             Jianwei Zhang, Huawei
  March 2006                         doc.: IEEE 802.22-06/0050r0




Part 1: Channel Sensing
 Guard intervals for extra quiet period in TDD WRAN system




Submission              Slide 4                Jianwei Zhang, Huawei
    March 2006                                     doc.: IEEE 802.22-06/0050r0

                                Background

• Synchronous Quiet Period
         a period in which all WRAN devices stop transmission in all channels
          available in the system
         used for sensing the signals in all channels of the system without
          interfering the system itself
         useful to enhance awareness to the surrounding radio environment


• Can the sensing accuracy be further enhanced?




  Submission                       Slide 5                   Jianwei Zhang, Huawei
   March 2006                                                                                    doc.: IEEE 802.22-06/0050r0

                             Fully Utilize the Guard Intervals
• Guard Intervals
    –         When using OFDMA at the physical layer, guard intervals should be inserted at
              the switching points of transmission
               OFDM symbols of different users can be synchronized at BS.


               CPE1 (d=0)         1    2     3    4                  1       2       3       4     1    2     3     4        …



                                  Downlink sub-frame        GI       Uplink sub-frame              Downlink sub-frame




               CPE2 (d=R)              1     2    3    4         1       2       3       4              1     2     3    4       …


               [R: cell radius]
                                       Downlink sub-frame        Uplink sub-frame                 GI    Downlink sub-frame
   –          We can use these guard intervals as extra quiet periods for sensing!

 Submission                                            Slide 6                                                    Jianwei Zhang, Huawei
       March 2006                                                                      doc.: IEEE 802.22-06/0050r0

                              Related Work by I2R (Singapore)
•      All CPE should have a mandatory quiet period with fixed length at the
       switching point from downlink (DL) to uplink (UL).
                                                                              TTG2


                                                    TTG1

BS          DL Subframe                     Sense                                     UL 1                              UL 2

                                      TRS     Tss




CPE1                    DL Subframe                   Sense                   UL 1

                  DL1                         DS1         Tss      SSRTG                DL1




CPE2                                        DL Subframe                       Sense           UL 2




                                                    DL2                 DS2     Tss   SSRTG           DL2


     Submission                                               Slide 7                                Jianwei Zhang, Huawei
    March 2006                                                                               doc.: IEEE 802.22-06/0050r0

            Related Work by I2R (Singapore)
• Disadvantages
   –      Guard intervals from uplink to downlink have not been utilized.
   –      Since a quiet period of fixed length is inserted to all CPEs (regardless
          of their distances to base station), for the CPEs at the edge of the cell
          in which guard intervals are usually not required, the uplink
          transmission of these CPEs will be deferred  not effective

               CPE1 (d=0)         1    2     3    4                  1       2       3       4    1    2     3      4        …



                                  Downlink sub-frame        GI       Uplink sub-frame             Downlink sub-frame




               CPE2 (d=R)              1     2    3     4        1       2       3       4             1     2      3    4       …


               [R: cell radius]
                                       Downlink sub-frame        Uplink sub-frame                GI    Downlink sub-frame


  Submission                                           Slide 8                                                   Jianwei Zhang, Huawei
   March 2006                                     doc.: IEEE 802.22-06/0050r0

                           Our Proposed Design

• Assumptions
  –       TDD (time division duplex) deployment
  –       OFDMA (orthogonal frequency domain multiplexing access) is used in
          both uplink and downlink

• Main Features
  –       Adaptive Guard Interval Control
  –       Asynchronous Quiet Period




 Submission                        Slide 9                  Jianwei Zhang, Huawei
  March 2006                                                                               doc.: IEEE 802.22-06/0050r0

                                  Our Proposed Design
   • Feature: Adaptive Guard Interval Control
    Conventionally, CPE1 should wait for CPE2 during the uplink
     transmission such that their first uplink symbols are synchronized at BS.
    We relax the above constraint:


       CPE1 (d=0)      1    2    3     4                  1       2       3       4    1    2    3     4         …



                       Downlink sub-frame        GI       Uplink sub-frame             Downlink sub-frame




       CPE2 (d=R)           1    2     3    4         1       2       3       4             1    2     3     4       …



                            Downlink sub-frame        Uplink sub-frame                GI    Downlink sub-frame



             * CPE2’s first UL symbol is synchronized with CPE1’s second UL symbol

Submission                                            Slide 10                                              Jianwei Zhang, Huawei
  March 2006                                                                    doc.: IEEE 802.22-06/0050r0

                     Advantages of Adaptive GI Control
     For those CPEs being close to BS: they can start transmission in advance
      (1)     Length of guard intervals from DL to UL can be shortened
      (2)     More OFDM symbols can be transmitted
     For those CPEs being far away from BS
      (1)     Uplink transmission will no longer be deferred
      (2)     Number of transmitted OFDM symbols remains unchanged
     If considering some practical limitations such as the hardware limitation or the delay
      spread of the multi-path channel, a gap should be guaranteed between the DL and
      UL sub-frame when operating the adaptive GI control.

                             1    2    3     4               1    2    3    4     1     2   3     4    …


                             Downlink sub-frame              Uplink sub-frame     Downlink sub-frame
             CPE3 (0<d<R)                     Without adaptive guard interval control

                             1    2    3     4       1       2    3    4    5     1     2   3     4    …


                             Downlink sub-frame          Uplink sub-frame         Downlink sub-frame

                                             With adaptive guard interval control
Submission                                        Slide 11                                       Jianwei Zhang, Huawei
  March 2006                                                        doc.: IEEE 802.22-06/0050r0

                           Our Proposed Design

 •     Feature: Asynchronous Quiet Period
  Guard intervals from UL to DL can also be used as extra quiet period for
   channel sensing.
  Depending on the demand for sensing accuracy, some OFDM symbols
    can be replaced by the sensing period
    – Flexibility is ensured
    – BS notifies the assignment of such sensing periods to the CPEs by
      using the proposed Sensing Period Assignment (SPA) message.


                 1    2     3    4      1    2     3     4     5        1    2     3      4      …
  CPE3 (0<d<R)
                 Downlink sub-frame         Uplink sub-frame            Downlink sub-frame

                                 With adaptive guard interval control



Submission                              Slide 12                                       Jianwei Zhang, Huawei
     March 2006                                                   doc.: IEEE 802.22-06/0050r0

                              SPA Message Format

                 Syntax                                         Notes

SPA_Message_Format() {

    Management Message Type   Indicates the type of SPA message

    Connection ID             Indicates the user to whom the message is sent

    Start Time                Indicates the start time of the sensing period, in unit of OFDM symbols

    Duration                  Indicates the duration of the sensing period, in unit of OFDM symbols

}




Submission                             Slide 13                                Jianwei Zhang, Huawei
   March 2006                                      doc.: IEEE 802.22-06/0050r0

                                 Conclusion
• Adaptive Guard Interval Control
         For CPEs being close to BS, more OFDM symbols can be transmitted
         Guard intervals from DL to UL can be shortened
         For CPEs being far away from BS, their uplink transmission will no
          longer be deferred
         Performance Gain: Assume cell size is 33km and frame length is 5ms,
          the round-trip delay is 0.22ms  4.4% of bandwidth can be used!
• Asynchronous Quiet Period
         Guard intervals from UL to DL can also be used for channel sensing
         Flexibility: some OFDM symbols can be replaced by sensing period
         Sensing Period Assignment (SPA) message: one kind of MAC
          management message

 Submission                        Slide 14                   Jianwei Zhang, Huawei
  March 2006                       doc.: IEEE 802.22-06/0050r0




Part 1: Channel Sensing
Region-based Bayesian method for RF sensing in WRAN system




Submission             Slide 15              Jianwei Zhang, Huawei
   March 2006                             doc.: IEEE 802.22-06/0050r0

                             Background

• The WRAN system needs to detect the presence of
  incumbent systems and avoid the interference to the
  incumbent system

• Detection of Incumbents
   – Presence of the incumbents
   – Locations of the incumbents




 Submission                    Slide 16             Jianwei Zhang, Huawei
   March 2006                                      doc.: IEEE 802.22-06/0050r0

                       Detection of Incumbents

• Detect the presence of the incumbents
   – The subband needs to be vacated in the whole cell/sector
   – Lower spatial efficiency


• Detect the locations of the incumbents
   – When the operation range of incumbent is small, the subband may be
     used without interfering to the incumbent.
   – Higher spatial efficiency
   – Complexity grows exponentially with the number of targets
   – Many previous work requires: knowledge of number of targets,
     knowledge of signatures, and detection of time of arrivals, etc.



 Submission                      Slide 17                     Jianwei Zhang, Huawei
    March 2006                                   doc.: IEEE 802.22-06/0050r0

                 Region-based RF Sensing Algorithm

• Partition the cell/sector into a number of disjoint regions

• For each region, decide whether some incumbents exist
    Higher spatial efficiency
    The number of targets need not be known a priori
    Complexity does not exponentially grow with the number of targets

• Control overhead
    Each CPE feedbacks incumbent types and the corresponding subband




  Submission                     Slide 18                   Jianwei Zhang, Huawei
    March 2006                                                   doc.: IEEE 802.22-06/0050r0

                             The Bayesian Method (1)

Define d = The vector {di} of all sensors.
and    Cij = Cost of deciding Hi(), given Hj() is true

The decision rule of the Bayesian method is
                                                           1
                 i ( )  arg min Ci ( )  arg min  P(d, H j ( ))Cij
                 *
                             i0,1            i0,1
                                                          j 0

where  is a subset of PIT region iff i*() = 1.

Cost matrix example:
                               C00            C01  0 10 
                             C                     1 1 
                                C10           C11        
  Submission                             Slide 19                          Jianwei Zhang, Huawei
     March 2006                                                                      doc.: IEEE 802.22-06/0050r0

                                      The Bayesian Method (2)
For simplicity, assume conditional independence,

                P(d, H 0 ( ))  P( H 0 ( ))  P(d i  0 | H 0 ( ))  P(d i  1 | H 0 ( ))
                                                        iS , 0                         iS ,1

                P(d, H1 ( ))  P( H1 ( ))  P(d i  0 | H1 ( ))  P(d i  1 | H1 ( ))
                                                       iS , 0                         iS ,1


P(di  0 | H 0 ( ))  P(di  0, H 0 ( Ai  ) | H 0 ( ))  P(di  0, H1 ( Ai  ) | H 0 ( ))

                                                                          
                       1  PF ,i e  ( Ai  )  1  PD,i ( Ai  )  1  e  ( Ai  )   
P(di  1 | H 0 ( ))  P(di  1, H 0 ( Ai  ) | H 0 ( ))  P(di  1, H1 ( Ai  ) | H 0 ( ))
                                                              
                       PF ,i e  ( Ai  )  PD ,i ( Ai  ) 1  e  ( Ai  )   
P(d i  0 | H1 ( ))  1  PD ,i ( )
P(d i  1 | H1 ( ))  PD ,i ( )

   Submission                                           Slide 20                                   Jianwei Zhang, Huawei
  March 2006                                       doc.: IEEE 802.22-06/0050r0


                         The Bayesian Method (3)

  We assume the detection process of a sensor is
  modeled by Bernoulli trials.
   Each IT within its detection region is an i.i.d. trial.
   The probability of detecting a particular IT is independent
    of its position.
            Flag di = 1, iff at least one IT is detected.




Submission                         Slide 21                  Jianwei Zhang, Huawei
       March 2006                                                                                                                  doc.: IEEE 802.22-06/0050r0

                                                             The Bayesian Method (4)

PD1i)  P(di  1 | Exact one incumbent transmitte r is within Ai )
  (
    ,


PD,i ( )  P(di  1 | H1 ( )),   Ai

P(d i  1, H1 ( ))
 P(Some IT in  is detected)  P(Some IT exists in  but none is detected) P(Some IT is detected in Ai  )
   
 1 e
            ( ) PD1,i)
                    (
                             e     ( ) PD1,i)
                                              (
                                                                   
                                                       e  ( ) 1  e
                                                                                  ( Ai  ) PD1,i)
                                                                                              (
                                                                                                      
                                        ( Ai  ) PD1,i)
                                                    (
                                                                   ( Ai ) PD1,i)
                                                                             (
 1  e  ( )  e  ( ) e                               e


                                                                                                 ( Ai ) PD1,i)
                                                                                                           (
                                                                                                                          ( )   ( Ai  ) PD ,i
                                                                                                                                               (1)
                                                                                           e                       e         e
                                                        PD ,i ( )  1 
                                                                                                                   1  e  ( )




  Submission                                                                                Slide 22                                                  Jianwei Zhang, Huawei
  March 2006                                               doc.: IEEE 802.22-06/0050r0

                               Algorithm Flow




       System Initialization      Compute PIT (potential               Compute
                               incumbent transmitter) region       Protection Region
Submission                         Slide 23                           Jianwei Zhang, Huawei
                                March 2006                                                                                                         doc.: IEEE 802.22-06/0050r0

                                                                                  PIT Region Distribution
• 400 CPEs, 4 ITs (incumbent transmitters)
• Detection radius = 10 grids (Grid space = 50m)
• 5km by 5km square region
                                                    CPE Number = 400, IT Number = 4                                                                          CPE Number = 400, IT Number = 4
                                100                                                                                                     100

                                90                                                                                                      90

                                80                                                                                                      80

                                70                                                                                                      70
y (grid point)




                                                                                                             y (grid point)
                                60                                                                                                      60
                 Grid Index Y




                                                                                                                         Grid Index Y
                                50                                                                                                      50

                                40                                                                                                      40

                                30                                                                                                      30

                                20                                                                                                      20

                                10                                                                                                      10

                                 0                                                                                                       0
                                      0   10   20    30    40        50      60    70   80   90     100                                       0   10   20     30    40        50      60   70   80   90   100
                                                                Grid Index X                                                                                             Grid Index X
                                                          x (grid point)                                                                                             x (grid point)

                                                      Union Algorithm                                                                                       Region-based Algorithm

                     Submission                                                                   Slide 24                                                                 Jianwei Zhang, Huawei
                                  March 2006                                                                                               doc.: IEEE 802.22-06/0050r0

                                                            Region-based vs. Union Algorithms

• PF,i = 0.01 (per CPE per subband)
                                0.4                                                                                          0.7

                                              Union Alg                                                                                                                      Union Alg
Ratio of areas of PIT region




                               0.35           Region Based Alg                                                               0.6                                             Region Based Alg




                                                                                                      Probabilitymiss miss
                                0.3                                                                                          0.5
     Ratio of PIT Region




                                                                                                                      of
                               0.25                                                                                          0.4




                                                                                                       Probability of
                                0.2                                                                                          0.3



                               0.15                                                                                          0.2



                                0.1                                                                                          0.1



                               0.05                                                                                           0
                                      0   2         4       6       8       10       12   14     16                                0   2      4     6       8       10        12      14        16
                                                                                 2                                                                                       2
                                                          CPE density (#CPE/km )                                                                  CPE density (#CPE/km )
                                                   CPE density (#CPE/km2)                                                                     CPE density (#CPE/km2)




                               Submission                                                 Slide 25                                                       Jianwei Zhang, Huawei
                                         March 2006                                                                                          doc.: IEEE 802.22-06/0050r0

                                                               Region-based vs. Union Algorithms

• PF,i = 0.1 (per CPE per subband)
                                   0.8                                                                                         0.7
                                                 Union Alg                                                                                                                     Union Alg
                                                 Region Based Alg                                                                                                              Region Based Alg
Ratio Ratio of PIT Region region




                                   0.7                                                                                         0.6

                                   0.6




                                                                                                        Probability missmiss
                                                                                                                               0.5
      of areas of PIT




                                   0.5




                                                                                                          Probability of of
                                                                                                                               0.4
                                   0.4
                                                                                                                               0.3
                                   0.3

                                                                                                                               0.2
                                   0.2


                                   0.1                                                                                         0.1


                                    0                                                                                           0
                                         0   2         4       6        8       10     12   14     16                                0   2     4       6        8       10     12      14         16
                                                             CPE density (#CPE/km 2)                                                                 CPE density (#CPE/km 2)
                                                       CPE density (#CPE/km2)                                                                      CPE density (#CPE/km2)




                                   Submission                                               Slide 26                                                          Jianwei Zhang, Huawei
                                    March 2006                                                                                                       doc.: IEEE 802.22-06/0050r0

                                                    Sensitivity to Estimate of CPE density (1)

• PF,i = 0.1 (per CPE per subband)
• Actual average number of IT per km2 = 0.16 IT/km2
                                           Ratio of PIT region -- lambda (Actual lambda = 0.16)                                                      Probability of miss -- lambda (Actual lambda = 0.16)
                              0.7                                                                                                           0.35
                                                                                      Union Algorithm                                                Union Algorithm
Ratio of areasRegion region




                                                                                      Region Based Alg                                               Region Based Alg
                              0.6                                                                                                            0.3




                                                                                                                     Probabilitymiss miss
     Ratio of PIT of PIT




                              0.5                                                                                                           0.25




                                                                                                                      Probability of of
                              0.4                                                                                                            0.2


                              0.3                                                                                                           0.15


                              0.2                                                                                                            0.1



                              0.1                                                                                                           0.05



                               0                                                                                                              0
                                 -4         -3           -2          -1           0               1              2                              -4     -3          -2          -1           0               1    2
                               10        10           10           10           10          10              10                                10     10         10           10           10          10        10
                                                                                           2                                                          lambda used in the algorithm (#estimated IT/km 2)
                                           lambda used
                                           Expected in the algorithm (#estimated IT/km
                                                       number of IT per km2                  )
                                                                                                                                                          Expected number of IT per km2


                              Submission                                                              Slide 27                                                              Jianwei Zhang, Huawei
            March 2006                                                                                                               doc.: IEEE 802.22-06/0050r0

                                      Sensitivity to Estimate of CPE density (2)
20




                                        Ln( P(H0()) / P(H1()))                                   Ln( P(di=0|H0()) / P(di=0|H1()))
10




                                                                                                                                               Black curve
  0
                                                                                                                                               = Sum of three curves
                                                                                                                                                         P(di  0 | H 0 ) i P(di  1 | H 0 ) 
                                                                                                                                               P( H ) iS                   S ,1
                                                                                                                                                                                                
                                                                                                                                           ln      0       ,0
                                                                                                                                                                                                
                                                                                                                                               P( H1 )  P(d i  0 | H1 )  P(d i  1 | H1 ) 
-10
                                                                           Decision Curve: <0 corresponds to   PIT region
                                                                                                                                                                                               
                                                                                                                                                       iS , 0            iS ,1             
         Ln( P(di=1|H0()) / P(di=1|H1()))
-20




-30




-40
    -4              -3                   -2                   -1                   0                    1                      2           3
  10             10                    10                   10                   10                   10                      10          10
                                                                   Estimated 
                                                            Estimate of 
         Submission                                                                      Slide 28                                                    Jianwei Zhang, Huawei
   March 2006                                              doc.: IEEE 802.22-06/0050r0
                   Region-based Algorithm: Transceivable Region


• Downlink System
   – Ideal antenna with 120-degree beam-width and front-to-back ratio GFB of 13dB.
   – Uniform gain within main beam and constant attenuation of 13dB outside.
   – Cell radius is 33km; path loss exponent in a cell, pl = 3.
   – 10 circular clusters of CPEs, with radius of 3km, center uniformly distributed
   – For every cluster, 100 CPEs are uniformly distributed within it.
   – Pth      = Maximum WRAN signal power allowed in the protection region
   – PRmin = Minimum required receiving power of a CPE: Pth + 3dB
   – Drr      = The radius of the receivable region
   – Dpro     = The minimum distance between BS and protection region, Dpro.

                                                      1
                                            Pth     pl

                                            P  D pro
                                     Drr         
                                            R min 

 Submission                            Slide 29                         Jianwei Zhang, Huawei
  March 2006                                                                                   doc.: IEEE 802.22-06/0050r0
                             Region-based Algorithm: Transceivable Region
                                                            Region based Algorithm
                     40




                            Sector of a WRAN cell
                     35


                                                                             A cluster


                     30                                                                   Cyan point: CPE that reports d=0




                                                                                                    Red point: CPE that reports d=1

                     25




                     20
       y (km)
                km




                     15




                                                              Yellow region: PIT region
                     10
                                                                     Blue star: IT



                     5




                     0


                                                          CPE Transceivable Region


                     -5
                      -20             -10           0                10                   20                          30                40
                                                                     km
                                                                           x (km)
Submission                                              Slide 30                                                             Jianwei Zhang, Huawei
   March 2006                               doc.: IEEE 802.22-06/0050r0

                Comparison of Usable Subbands
• Average number of subbands available to the
  system:
FDD
    – Region-based Algorithm:          32 subbands / base station
    – Union Algorithm:                 12 subbands / base station


TDD
    – Region-based Algorithm:          31 subbands / base station
    – Union Algorithm:                 12 subbands / base station



 Submission                 Slide 31                  Jianwei Zhang, Huawei
                                                                                                    March 2006                                                doc.: IEEE 802.22-06/0050r0

                                                                                                                             Complexity Comparison
                                                                                                  14
Complexity Ratio of Region-based to Union Algorithm
                                                  Complexity of Region Based Alg over Union Alg




                                                                                                  12
                                                                                                                                                                    At a reasonable CPE density,
                                                                                                                                                                    the complexity of the region-
                                                                                                  10
                                                                                                                                                                    based algorithm is about 10
                                                                                                                                                                    times of the union algorithm
                                                                                                   8                                                                and     its   complexity    will
                                                                                                                                                                    converge to less than 14 times
                                                                                                   6                                                                of the union algorithm.


                                                                                                   4



                                                                                                   2
                                                                                                       0       2   4     6        8       10        12   14    16
                                                                                                                                               2
                                                                                                                       CPE density (#CPE/km2
                                                                                                                       CPE density (#CPE/km ) )

                                                                                                  Submission                             Slide 32                          Jianwei Zhang, Huawei
   March 2006                                     doc.: IEEE 802.22-06/0050r0

                                Conclusion
• More efficient use of space for frequency reuse
   – Smaller PIT and protection regions  larger transceivable region
   – CPEs inside the receivable region can use the channel
   – Noticeable gain in the number of usable channels per cell

• Compared with union algorithm:
   – Large reduction in PIT region
   – Small increase in probability of miss
   – The tradeoff can be controlled by the cost matrix

• Moderate Computation Complexity
• Low overhead for sensing report

 Submission                      Slide 33                    Jianwei Zhang, Huawei
  March 2006                      doc.: IEEE 802.22-06/0050r0




  Part 1: Channel Sensing
 The MAC management message for channel sensing




Submission            Slide 34              Jianwei Zhang, Huawei
   March 2006                            doc.: IEEE 802.22-06/0050r0

                              Proposal

• Design of MAC Management Messages for channel
  sensing of the CPE’s
• Our proposed RF sensing algorithm suggests the
  following information is sufficient for satisfactory
  performance in sensing report of CPE’s
   – Incumbent type
   – Channel occupied by the incumbent

• Design Criteria: Reduce control overhead



 Submission                   Slide 35             Jianwei Zhang, Huawei
  March 2006                                                             doc.: IEEE 802.22-06/0050r0
                      Measurement Request (MS-REQ)
                        Syntax                Size                           Notes
             Management Message Type         8 bits    The type of MAC management message
             Transaction ID                  16 bits   The transaction ID of the corresponding
                                                       channel measurement request for which a
                                                       report is required by the BS.
             System Type                     8 bits    The   type   of    incumbent    system     to   be
                                                       measured. See Table 2. If this field is 0, the
                                                       CPE should sense all incumbent systems.
             Start frame                     8 bits    Indicate the frame from which the channel
                                                       measurement starts.
             Duration                        8 bits    The actual duration of the measurement (units
                                                       in frame).
             Full_Report                      1 bit    0: Does not request full report.
                                                       1: Request full report.
             CINR_Flag                        1 bit    0: Does not request CPE to report CINR
                                                       (Carrier Interference-to-Noise Ratio).
                                                       1: Request CPE to report CINR for each
                                                       channel of each incumbent type.

             Table 1. Measurement Request (to be cont’d)                                  Jianwei Zhang, Huawei
Submission                                   Slide 36
  March 2006                                                               doc.: IEEE 802.22-06/0050r0

                           Measurement Request (MS-REQ)


             Channel_List_Format                1 bit      0: Channel-basis
                                                           1: Interval-basis
             If (Channel_List_Format=0){
                 Channel-basis Channel List    Variable    See Table 3.
             }
             else{
                 Interval-basis Channel List   Variable    See Table 4.
             }

         Table 1. (Cont’d) Measurement Request




Submission                                      Slide 37                             Jianwei Zhang, Huawei
  March 2006                                         doc.: IEEE 802.22-06/0050r0

                          System Types
                 System Type           Description
                      0                 All types
                      1                  802.22
                      2                  ATSC
                      3                  NTSC
                      4                  Part 74
                      5                   DVB
                    6-255               Reserved

               Table 2. System Types




Submission                  Slide 38                           Jianwei Zhang, Huawei
  March 2006                                                              doc.: IEEE 802.22-06/0050r0

                   Channel-/Interval- basis Channel Lists
                 Channel-basis Channel List     Size                         Notes
             Number of Channels, c             8 bits     Total number of channels the CPE should
                                                          measure.
             for i = 1:c {
                  Channel Index                8 bits
             }

             Table 3. Channel-basis Channel List


                 Interval-basis Channel List    Size                         Notes
             Number of Intervals, d            8 bits     Total number of channel intervals the CPE
                                                          should measure.
             for i = 1:d {
                  Starting Channel Index       8 bits     The index of the starting channel
                  Number of Channels           8 bits     The number of channels in the current
                                                          interval
             }

             Table 4. Interval-basis Channel List
Submission                                     Slide 39                                       Jianwei Zhang, Huawei
  March 2006                                                         doc.: IEEE 802.22-06/0050r0

                             Measurement Report (MS-REP)
                         Syntax              Size                       Notes
             Management Message Type        8 bits
             Transaction ID                16 bits
             Report_Detail                  1 bit     0: Incremental Measurement Report
                                                      1: Full Measurement Report
             If(Report_Detail=0){
                 Incremental Measurement   Variable   See Table 6.
             Report
             }
             else{
                 Full Measurement Report   Variable   See Table 7.
             }

         Table 5. Measurement Report



Submission                                 Slide 40                                Jianwei Zhang, Huawei
  March 2006                                                           doc.: IEEE 802.22-06/0050r0
                             Incremental Measurement Report
                       Syntax            Size                          Notes
         Number of System Types, n0      8 bits    Number of system types that has been
                                                   detected.
         for i = 1:n0{
             System Types                8 bits    See Table 2.
             Number of Channels, mi      8 bits    Number of channels occupied by current
                                                   incumbent system.
             for j = 1:mi{
                 Start frame             8 bits    Indicate the frame from which the channel
                                                   measurement starts.
                 Duration                8 bits    The actual duration of the measurement
                                                   (units in frame).
                 Channel Index           8 bits
                 Leave or Arrive         1 bit     1: the channel is occupied by the incumbent
                                                   system.
                                                   0: the channel is released from the
                                                   incumbent system.
                 if(CINR_Flag=1){
                     CINR                8 bits
                 }
             }
         }

        Table 6. Incremental Measurement Report
Submission                                  Slide 41                                     Jianwei Zhang, Huawei
  March 2006                                                            doc.: IEEE 802.22-06/0050r0
                                     Full Measurement Report
                           Syntax               Size                      Notes
             Number of System Types, n1         8 bits   Number of system types that has been
                                                         detected.
             for i = 1:n1{
                 System Types                   8 bits   See Table 2.
                 Number of Channels, mi         8 bits   Number of channels occupied by
                                                         current incumbent system.
                 for j = 1:mi{
                     Start Frame                8 bits   Indicate the frame from which the
                                                         channel measurement starts.
                     Duration                   8 bits   The actual duration of the
                                                         measurement (units in frame).
                     Channel Index              8 bits
                     if(CINR_Flag=1){
                         CINR                   8 bits
                     }
                 }
             }

             Table 7. Full Measurement Report

Submission                                  Slide 42                                   Jianwei Zhang, Huawei
  March 2006                doc.: IEEE 802.22-06/0050r0

               Example




Submission       Slide 43             Jianwei Zhang, Huawei
   March 2006                                  doc.: IEEE 802.22-06/0050r0

                               Conclusion

• Compatible with IEEE802.16 MAC Management Messages
   – Flexible
   – Support various schemes of control channel assignment

• Design Criteria: Reduce control overhead
   – Interval-basis Channel List
   – Incremental Measurement Report




 Submission                        Slide 44              Jianwei Zhang, Huawei
  March 2006                           doc.: IEEE 802.22-06/0050r0




  Part 1: Channel Sensing
     Pilot design for channel estimation
     and interference detection in WRAN system




Submission                Slide 45               Jianwei Zhang, Huawei
   March 2006                                               doc.: IEEE 802.22-06/0050r0

                   Proposed Downlink Pilot Design
    ●    ●    ●    ●    ●    ●    ●    ●    ●   … Block 0
    ●    ●    ●    ●    ●    ●    ●    ●    ●   … Block 1
    ●    ●    ●    ●    ●    ●    ●    ●    ●   … Block 2
    ●    ●    ●    ●    ●    ●    ●    ●    ●   … Block 3
                                       … Block 4
                                       … Block 5
                                       … Block 6
                                       … Block 7
                                       … Block 8
                        ……………………………………….
                           ●: Pilot subcarrier : Data subcarrier

Major consideration
 The channel is slow varying
 The subcarrier spacing is about several KHz
 To facilitate the interference detection

 Submission                          Slide 46                         Jianwei Zhang, Huawei
   March 2006                                            doc.: IEEE 802.22-06/0050r0

                   Proposed Uplink Pilot Design

                      ●●●●                   Block 0
                                         Block 1
                                         Block 2
                           ………………
                       ●: Pilot subcarrier : Data subcarrier


Major consideration
 The channel is slow varying
 The subcarrier spacing is about several KHz
 Subband-based OFDMA




 Submission                       Slide 47                         Jianwei Zhang, Huawei
      March 2006                                           doc.: IEEE 802.22-06/0050r0
                        Interference Detection

                                  Channel




                                  Channel + Interference




•    Left graphs stands for the constellation of pilots on the same subcarriers of
     different OFDM blocks
•    Right graphs stands for the constellation of corresponding received signals
•    Interference  symmetric structure of the constellation will be destroyed
•    No matter the interference varies or not
•    No matter what constellation size used                          Jianwei Zhang, Huawei
    Submission                        Slide 48
   March 2006                                                    doc.: IEEE 802.22-06/0050r0

                    Interference Detection Algorithm
• System model: Y = P*H + n  one subcarrier
• Pk,i: Pilot on the k-th subcarrier of i-th OFDM block.
• Pk,i = - Pk,i+1
• Hypothesis test:
              H0:       |Yk,i + Yk,i+1|2 = |Pk,i*Hk + nk,i + Pk,i+1*Hk + nk,i+1|2
                                         = |nk,i + nk,i+1|2
              H1:       |Yk,i + Yk,i+1|2 = |Pk,i*Hk + Ik,i + nk,i + Pk,i+1*Hk + Ik,i+1 + nk,i+1|2
                                         = |Ik,i + Ik,i+1 + nk,i + nk,i+1|2
              P(|Yk,i+Yk,i+1|2 > threshold | H0) = Palarm

• |Yk,i + Yk,i+1|2 given H0  χ2 distribution.

 Submission                                 Slide 49                           Jianwei Zhang, Huawei
   March 2006                                                                                                                    doc.: IEEE 802.22-06/0050r0

                                                                     Simulation Model and Parameters
                                                             Noise
 signal
                                                             +
                                                                                        Remove
                                                                                          CP
                                                                                                                   FFT       • Interference generated in
                                                             Interference
                                                                                                                               time domain  more
                                                                                                                               close to the real situation
AWGN                                                   Filter
                                                                                                                             • Interference on one
                                            0
                                                                                                                               subcarrier of different
                                           -10

                                           -20
                                                                                                                               OFDM blocks varies
                                                             Frequency response
    Power profile of filter, 20 log |H|




                                                                                                                             • False alarm probability is
                                   10




                                           -30
                                                             of the filter.
                                           -40

                                           -50
                                                                                                                               set to 0.01
                                           -60

                                           -70
                                                                                                                             • Noise power is known a
                                           -80

                                           -90                                                                                 prior
                                          -100
                                                 100   200   300   400   500      600    700   800   900   1000
                                                                     k th subcarrier



 Submission                                                                                                       Slide 50                 Jianwei Zhang, Huawei
  March 2006                                                                                                                                                                          doc.: IEEE 802.22-06/0050r0
                                                                    Simulation Results --- 2 Pilots
                            1                                                                                                                                                1

                           0.9                                                                                                                                              0.9

                           0.8                                                                                                                                              0.8
   Detection Probability




                                                                                                                                                    Detection Probability
                           0.7                                                                                                                                              0.7
                                                         INR = 0dB                                                                                                                                         INR = 10dB
                           0.6                                                                                                                                              0.6

                           0.5                                                                                                                                              0.5

                           0.4                                                                                                                                              0.4

                           0.3                                                                                                                                              0.3

                           0.2                                                                                                                                              0.2

                           0.1                                                                                                                                              0.1

                            0                                                                                                                                                0
                                 0   100   200   300    400   500   600   700                       800      900   1000                                                           0   100   200    300    400   500   600   700   800   900   1000
                                                       Subcarrier Index                                                                                                                                  Subcarrier Index


                                                                                                    1

                                                                                                   0.9

                                                                                                   0.8
                                                                           Detection Probability




                                                                                                   0.7
                                                                                                                                  INR = 20dB
                                                                                                   0.6

                                                                                                   0.5

                                                                                                   0.4

                                                                                                   0.3

                                                                                                   0.2

                                                                                                   0.1

                                                                                                    0
                                                                                                         0   100    200   300    400   500   600   700                       800      900   1000
                                                                                                                                Subcarrier Index                                                                Jianwei Zhang, Huawei
Submission                                                                                                            Slide 51
  March 2006                                                                                                                                          doc.: IEEE 802.22-06/0050r0

                                                                      Simulation Results --- 5 Pilots
                              1                                                                                                           1

                             0.9                                                                                                         0.9

                             0.8                                                                                                         0.8
     Detection Probability




                                                                                                                 Detection Probability
                             0.7                                                                                                         0.7
                                                           INR = 0dB                                                                                                    INR = 10dB
                             0.6                                                                                                         0.6

                             0.5                                                                                                         0.5

                             0.4                                                                                                         0.4

                             0.3                                                                                                         0.3

                             0.2                                                                                                         0.2

                             0.1                                                                                                         0.1

                              0                                                                                                           0
                                   0   100   200   300    400   500   600    700   800   900   1000                                            0    100   200   300    400     500   600    700   800    900   1000
                                                         Subcarrier Index                                                                                             Subcarrier Index


                              1



                                                                                                                  •
                             0.9

                             0.8
                                                                                                                                                   Threshold does not change
     Detection Probability




                                                                                                                  •
                             0.7

                             0.6
                                                           INR = 20dB
                                                                                                                                                   But use
                                                                                                                                                                         N 1

                                                                                                                                                                         Y
                             0.5
                                                                                                                                                                                                         2
                             0.4                                                                                                                                                     k ,i    Yk ,i 1
                             0.3                                                                                                                                             i 1
                             0.2                                                                                                                                                      N 1
                             0.1

                              0
                                   0   100   200   300    400   500    600   700   800   900   1000
                                                                                                                                                   to do the hypothesis test
                                                         Subcarrier Index

Submission                                                                                            Slide 52                                                                       Jianwei Zhang, Huawei
   March 2006                                       doc.: IEEE 802.22-06/0050r0

                                Conclusion

• Pilot design for both downlink and uplink
• Interference detection
   – Do not require extra overhead
   – No matter the interference is varying or not
   – No matter the constellation size used
   – Performance only depends on interference to noise ratio




 Submission                      Slide 53                     Jianwei Zhang, Huawei
   March 2006                                      doc.: IEEE 802.22-06/0050r0

              Joint Interference Detection & Decoding
• Existence of Narrowband Interference in WRAN
• Avoids Transmission in Interference Jammed Subcarriers
    – Transmitter may not know the existence of interference due to bursty
      nature of interference

• Receiver Detect Interference
    – Pilot based approaches
    – Data based approaches
          • Based on estimated data
          • Based on correlation of channel fading in frequency and time domain

• Existing Decoders Require Interference Knowledge
    – Performance determined by the accuracy of the interference detection
 Submission                       Slide 54                   Jianwei Zhang, Huawei
  March 2006                                           doc.: IEEE 802.22-06/0050r0

                                System Model
                                  Parallel Channel
                                 Par al l el Channel




    Enc      IL        Mod                                   Dem       DI L       Dec




                                               For Each Codeword


       Enc        IL     Mod                               Dem       DIL        Dec


                               Fading    AWGN Interference


Submission                         Slide 55                        Jianwei Zhang, Huawei
  March 2006                                      doc.: IEEE 802.22-06/0050r0

                     Existing Decoding Schemes
• Optimal Maximum Likelihood Decoding
     – Decoding metric is optimized differently for noise and interference
     – Require noise and interference statistics (position and power)


• Conventional Decoding
     – Only require interference position; not require interference power
     – Ignore (erase) interference jammed symbols
     – Decoding metric is Euclidean distance (Optimal metric for AWGN)
     – Undetected interference corrupts decoder because of metric mismatch

     all require interference detector

Submission                       Slide 56                    Jianwei Zhang, Huawei
   March 2006                                             doc.: IEEE 802.22-06/0050r0

                        Joint Erasure Decoding
• Given the number of erasures, search all possible
  codewords x with all possible erasure positions e

                                   
                           min min x  y
                              e           x
                                                  2

                                                  e   
• Determine the number of erasures
    – Apply sufficiency criteria

• Achievable performance
    – Maximum Likelihood decoding with the exact knowledge of the noise and
      interference statistics



 Submission                            Slide 57                     Jianwei Zhang, Huawei
    March 2006                                                         doc.: IEEE 802.22-06/0050r0

                       Implementation – Product Trellis

• Erasure Indicator Trellis
                                 ε1       ε2              ei  1i
                             0
                                                              ei  0
                             1
                                                          ei  1

• Bit Trellis
                                      a                   a
                                                 c
                                          d          e

                                                 f
                                      b                   b

• Product Trellis
                                                                                              [a,0]
               [a,0]
                                                                                              [a,1]




  Submission                                   Slide 58                          Jianwei Zhang, Huawei
   March 2006                                          doc.: IEEE 802.22-06/0050r0

                              Sufficiency Criteria

• Error Checking Code Based
   – Output the first candidate codeword that passes error checking and
     terminate decoding

• Path Metric Difference Based
   – Calculate path metric difference of consecutive candidate codewords
          • Metric difference is decreasing
          • Metric difference is small after all interference are erased
   – If the metric difference is less than a threshold Dec , then output the
     candidate codeword & terminate decoding




 Submission                          Slide 59                     Jianwei Zhang, Huawei
    March 2006                                    doc.: IEEE 802.22-06/0050r0

                              Complexity Reduction
• Find the most likely path sequentially
• Demodulator marks symbol erasures
    – Erase the symbol if any of the corresponding bit is marked as an erasure
      by decoder
    – Erase the symbol based on the channel output

                                              Undetectable left to decoder




                 detectable


  Submission                       Slide 60                  Jianwei Zhang, Huawei
  March 2006                           doc.: IEEE 802.22-06/0050r0

               Simulations – Fixed SIR or Jams
• Rate-½ 64-state convolutional code
• 16QAM with Gray mapping
• 864 subcarriers
• Profile A multipath fading channel; constant over each
  packet
• Fixed SIR or number of jammed subcarriers
• Sufficiency criterion: path metric difference based
• Demodulator does not mark erasure based on channel
  output

Submission                 Slide 61              Jianwei Zhang, Huawei
                    March 2006                                                                  doc.: IEEE 802.22-06/0050r0
                                            Simulations – Fixed SIR or Jams
                                             SIR=0dB, SNR=20dB                            5 Jams, SNR=20dB
           -2                                                                             -1
      10                                                                             10
                       Conventional                                                                                          Conventional
                       Proposed                                                                                              Proposed
                       Maximum Likelihood                                                                                    Maximum Likelihood


                                                                                          -2
                                                                                     10


           -3
BER




                                                                               BER
      10


                                                                                          -3
                                                                                     10




           -4
      10                                                                                  -4
                0        2         4         6         8       10       12           10
                                                                                          -10          -5                0                        5
                                Number of Jammed Subchannels
                                                                                                             SIR(dB)
           The proposed decoder
           (1) almost achieves the performance of the optimal decoder
           (2) reduces sensitivity to the number of jammed subcarriers
           (3) is insensitive to interference power
           Submission                                               Slide 62                                Jianwei Zhang, Huawei
                   March 2006                                                                       doc.: IEEE 802.22-06/0050r0
                                    Simulations – Fixed SIR or Jams
                                   SIR=0dB, SNR=20dB                                     5 Jams, SNR=20dB
                  -4                                                                        -4
           x 10                                                                      x 10
                                                                               7.4
                                                      11                                         SIR=-10dB
      11                                                                                         SIR=-5dB
                                                                               7.2
                                                                                                 SIR=0dB
      10                                                                                         SIR=5dB
                                                      9                         7
       9
                                                      7                        6.8
       8
BER




                                                                         BER
       7                                                                       6.6
                                                          5
       6
                                                                               6.4
       5                                                  3
                                                                               6.2
       4
                                                          1
       3                                                                        6

       2
           0           10   20     30           40   50        60              5.8
                                            2                                        0            10         20   30       40      50     60
                                 Dec /                                                                               2
                                                                    Dec /
      Optimal threshold of the path metric difference based sufficiency criterion is
      (1) almost independent of number of jammed subcarriers
      (2) almost independent of interference power
       Threshold can be determined offline
               Submission                                     Slide 63                                            Jianwei Zhang, Huawei
  March 2006                                     doc.: IEEE 802.22-06/0050r0

       Simulation – W/O Interference Detector
• 864 subcarriers
• 20 OFDM symbols per packet
• One codeword per OFDM symbol
• 2 OFDM pilot symbols for
      – Interference detection
      – Frequency domain LS channel estimation

• 32 jammed subcarriers (wireless microphone)
• SIR uniformly distributed in [-20dB,10dB]
• Sufficiency criterion: path metric difference based

Submission                       Slide 64                  Jianwei Zhang, Huawei
                    March 2006                                                                          doc.: IEEE 802.22-06/0050r0

           0
                                         Simulation – W/O Interference Detector
      10
                                                                                           0
                                                                                      10

           -1
      10




                                                                  Packet Error Rate
           -2
BER




      10


                                                                                           -1
                                                                                      10
           -3
      10
                     Conventional                                                                   Conventional
                     Proposed                                                                       Proposed
                     Conventional (ID)                                                              Conventional (ID)
           -4        Proposed (ID)                                                                  Proposed (ID)
      10
               16      17          18      19      20   21   22                                16     17          18      19      20   21   22
                                         SNR(dB)                                                                        SNR(dB)
                (1) Without interference detector (red)
                      Great gain over conventional decoder for BER and PER
                      Complexity increase by 1.5 times for PER=0.1 relative to conventional
                (2) With interference detector (blue)
                      Smaller gain for BER but significant gain for PER
                      Complexity increase by 15% for PER=0.1 relative to conventional
                                                         with or without interference Jianwei Zhang, Huawei
                (3) Proposed decoder performs similarly 65
                Submission                         Slide
                                                                                       detector
    March 2006                                  doc.: IEEE 802.22-06/0050r0

                 Simulation – W/O Channel Estimation Error
• Random interference for each carrier with probability 0.04
• SIR uniformly distributed in [-20dB,10dB]
• 2 OFDM pilot symbols for frequency domain LS channel
  estimation
• Each codeword is transmitted through 200 carriers and 10
  OFDM symbols
• Each convolutional codeword is encoded by CRC
• Demodulator marks erasures
• Sufficiency criterion: CRC and path metric difference based
    – CRC generator polynomial is 435(octal )
  Submission                    Slide 66                  Jianwei Zhang, Huawei
                    March 2006                                                                                            doc.: IEEE 802.22-06/0050r0
           0
                                      Simulation – W/O Channel Estimation Error
      10
                                                                                                            0
                                                                                                       10
           -1
      10

           -2
      10




                                                                                     Word Error Rate
                                                                                                            -1
                                                                                                       10
           -3
      10
BER




                                                  Gain of joint over separate
           -4
      10
                                                                                                            -2
                      Conventional                                                                     10            Conventional
      10
           -5
                      Proposed1 (separate)                                                                           Proposed1 (separate)
                                                                                                                     Proposed2     (joint)
                      Proposed2 (joint)
                      Maximum Likelihood                                                                             Maximum Likelihood
           -6
                      Conventional (CE error)                                                                        Conventional (CE error)
      10                                                                                                             Proposed1 (CE error)
                      Proposed1 (CE error)
                      Proposed2 (CE error)                                                                  -3       Proposed2 (CE error)
                                                                                                       10
               10     11      12       13     14      15      16      17        18                              10   11     12       13      14        15    16        17   18
                                            SNR(dB)                                                                                        SNR(dB)
      (1) Without channel estimation error (solid)
                 Joint erasure marking and decoding Performs closely to optimal decoder
                 Complexity increase by 50% for WER=0.01 relative to conventional decoder
      (2) With channel estimation error (dashed)
                 Joint erasure marking and decoding is less sensitive to channel estimation error than separate
                  erasure marking and decoding using demodulator only
                 Complexity increases by twice for WER=0.01 relative to conventional decoder
               Submission                                              Slide 67                                                                Jianwei Zhang, Huawei
   March 2006                            doc.: IEEE 802.22-06/0050r0

                          Conclusion
• The proposed decoding scheme almost achieves the optimal
  decoder performance without knowing the interference statistics
• Threshold of sufficiency criterion does not depend on
  interference characteristics and can be determined offline
• Complexity increase is reasonably small especially for high
  SNR or with an interference detector
• Performance loss due to channel estimation error is much
  smaller than that of conventional decoding scheme
• Therefore, it is robust and effective to combat unknown
  interference in practical situations


 Submission                Slide 68                Jianwei Zhang, Huawei
  March 2006                            doc.: IEEE 802.22-06/0050r0




 Part 2: Radio Resource Allocation
  Effective and flexible structure for CPE CSIT collection
  at base station for TDD/FDD OFDMA architecture




Submission                Slide 69                Jianwei Zhang, Huawei
  March 2006                          doc.: IEEE 802.22-06/0050r0


                         Motivation

   Radio resource is very scarce

               Design good resource allocation
               algorithm to fully utilize the resource


               CSIT is a crucial input



Submission                Slide 70              Jianwei Zhang, Huawei
   March 2006                                      doc.: IEEE 802.22-06/0050r0

                              CSIT Acquisition
• Using the reciprocity of the uplink and downlink channel
         CSIT of the excited subchannels of those currently uplink-active CPEs
          of TDD system

• Using feedback
         CSIT of the un-excited subchannels of those currently uplink-active
          CPEs of a TDD system
         CSIT of the currently uplink-inactive CPEs of TDD system
         CSIT of all the CPEs of FDD system



               Very important to design a good
                 CSIT collection mechanism
 Submission                        Slide 71                   Jianwei Zhang, Huawei
   March 2006                                     doc.: IEEE 802.22-06/0050r0


                Features of Downlink WRAN System

• BS knows the QoS requirements and queueing states of
  all the CPEs
   BS can determine which CPEs have higher priority and are more urgent

• Maximum Doppler frequency is very small
   The CSIT can be updated rather infrequently

• Variation of Doppler frequency among CPEs is limited
   The CSIT update frequencies of CPEs are similar


              Polling-based CSIT feedback mechanism
 Submission                     Slide 72                    Jianwei Zhang, Huawei
   March 2006                                      doc.: IEEE 802.22-06/0050r0

              Main Features of Our Proposed Structure

• Centralized polling at the BS
         BS decides which CPEs to poll based on QoS requirements, queueing
          states, etc.
         BS decides for each selected CPE which subband to estimate based
          on power mask, history, etc.
         BS decides for each selected CPE through which subchannels to
          convey CSIT

• Placement of the polling information
         For currently active CPEs, the polling information is contained in the
          UL-MAP
         For currently inactive CPEs, the polling information is contained in
          some broadcast channel

 Submission                        Slide 73                   Jianwei Zhang, Huawei
  March 2006                                                                    doc.: IEEE 802.22-06/0050r0
                          CSIT Collection Request Message (1)
                           Syntax                    Size (bits)                            Remarks
 CSIT_Collection_Request() {
                                                                   N_DL_RCID is the number of selected downlink-active-
 N_DL_RCID                                               8         only CPEs and both-downlink-and-uplink-active CPEs that
                                                                   are in this subband
 for i = 1: N_DL_RCID {
               Downlink RCID                             8
               Feedback_Control( )                    variable
       }
                                                                   0: no selected CPE is uplink-active-only
 UL_RCID_flag                                            1
                                                                   1: there are selected CPEs that are uplink-active-only
 If {UL_RCID_flag == 1}{
                                                                   N_UL_RCID is the number of selected uplink-active-only
  N_UL_RCID                                              8
                                                                   CPEs that are in this subband
  for i = 1: N_UL_RCID {
                Uplink RCID                              8
                Feedback_Control( )                   variable
           }
   }
                               CSIT_Collection_Request for active CPEs (to be cont’d)
Submission                                         Slide 74                                       Jianwei Zhang, Huawei
  March 2006                                                               doc.: IEEE 802.22-06/0050r0

                          CSIT Collection Request Message (2)
                                                              0: no CID is used
 CID_flag                                             1
                                                              1: CID is used

 If {CID_flag == 1}{

                                                              N_CID is the number of selected CPEs that are switched
     N_CID                                            8
                                                              to this subband

     for i = 1: N_CID {

              CID                                    16

              Feedback_Control( )                  variable

          }

      }

 }


                                    CSIT_Collection_Request for active CPEs (Cont’d)



Submission                                       Slide 75                                   Jianwei Zhang, Huawei
  March 2006                                                                   doc.: IEEE 802.22-06/0050r0
                                   Feedback Control Message (1)
                          Syntax                    Size (bits)                           Remarks

  Feedback_Control() {
                                                                  0: estimate the downlink CSI of this subband
       Subband_change_flag                               1        1: in the next frame estimate the downlink CSI of the
                                                                  subband specified by Subband Index
       If{Subband_ change_flag==1}{
             Subband Index                               8        At most 256 6MHz subband
         }
       Else{
                                                                  0: use default quantization level, L=a
         Quantization_level_flag                         1
                                                                  1: use specified quantization level
         If{ Quantization_level_flag ==1}{
                                                                  Assume there are at most 4 additional quantization
                   Quantization level, L=b               2
                                                                  precision levels
         }
                                                                  0: use default number of subchannels, N=c
         Feedback_ch_constraint_flag                     1
                                                                  1: use specified number of subchannels
         If{ Feedback_ch_constraint_flag==1}{
                   Number of subchannels, N=d            6        Assume 64 subchannels in a subband
               }
                                        Feedback_Control Message (to be cont’d)                  Jianwei Zhang, Huawei
Submission                                         Slide 76
  March 2006                                                               doc.: IEEE 802.22-06/0050r0

                             Feedback Control Message (2)
                                                              0: use default number of OFDM symbols, M=e
           Feedback_symb_constraint_flag                1
                                                              1: use specified number of OFDM symbols
           If{Feedback_symb_constraint_flag==1}{
                                                              Assume at most 4 OFDM symbols can be used to do
                 Number of OFDM symbols, M=f            2
                                                              feedback
             }
            for j=1:N{
                  Subchannel Index                      6
                     }
       }
   }


                                           Feedback_Control Message (cont’d)




Submission                                         Slide 77                                Jianwei Zhang, Huawei
  March 2006                                                                  doc.: IEEE 802.22-06/0050r0
                       CSIT Collection Request Message (1)
                           Syntax                          Size (bits)                       Remarks
 CSIT_Collection_Request() {
 N_CID                                                         8         N_CID is the number of selected inactive CPEs
 for i = 1:N_CID{
         CID                                                   16
         Subband Index                                         8         At most 256 6MHz subband
                                                                         0: use default quantization level, L=a
       Quantization_level_flag                                 1
                                                                         1: use specified quantization level
       If{ Quantization_level_flag ==1}{
                                                                         Assume there are at most 4 additional
              Quantization level, L=b                          2
                                                                         quantization precision levels
          }
                                                                         0: use default number of subchannels, N=c
       Feedback_ch_constraint_flag                             1
                                                                         1: use specified number of sub-channels
       If{ Feedback_ch_constraint_flag==1}{
               Number of subchannels, N=d                      6         Assume 64 subchannels in a subband
          }

                               CSIT_Collection_Request for inactive CPEs (to be cont’d)
Submission                                      Slide 78                                         Jianwei Zhang, Huawei
  March 2006                                                         doc.: IEEE 802.22-06/0050r0

                      CSIT Collection Request Message (2)

                                                                0: use default number of OFDM symbols, M=e
         Feedback_symb_constraint_flag                      1
                                                                1: use specified number of OFDM symbols

         If{Feedback_symb_constraint_flag==1}{

                                                                Assume at most 4 OFDM symbols can be used to
              Number of OFDM symbols, M=f                   2
                                                                do feedback

          }

         for j=1:N{

              Subchannel Index                              6

                 }

     }

 }


                                 CSIT_Collection_Request for inactive CPEs (Cont’d)

Submission                                       Slide 79                            Jianwei Zhang, Huawei
    March 2006                                       doc.: IEEE 802.22-06/0050r0

               Main Features of Our Proposed Structure
• Overhead reduction
          For currently active CPEs, 8-bit RCID is used instead of the 16-bit CID
           to identify CPEs
• Flexibility
          Default constraint on the number of subchannels and the number of
           OFDM symbols that a CPE should use to do feedback is known to both
           the BS and the CPEs
          BS has the option to allocate more or less subchannels and/or OFDM
           symbols for each CPE to do feedback, depend on the QoS requirement
           or the urgency of the downlink traffic
          Default CSIT quantization level is known to both BS and CPEs
          BS has the option to increase or decrease the quantization level to
           adjust the precision of the feedback
  Submission                        Slide 80                    Jianwei Zhang, Huawei
   March 2006                                                             doc.: IEEE 802.22-06/0050r0
         Main Features of Our Proposed Structure
• CPEs decide which subchannel CSIT to feedback based
  on the channel condition
             Using predefined modulation and coding scheme, given the number of
              subchannels, OFDM symbols that are used to convey CSIT, and
              the CSIT quantization level, each CPE knows it can feedback the CSIT
              of say c number of subchannels
             For FDD system, the CPE should choose c number of subchannels
              with the largest gains
             For TDD system, the CPE should choose c number of un-excited
              subchannels with the largest gains
                                                  Size (bits)                      Remarks
               CSIT_Feedback_Format() {                         If Q-bit feedback is allowed, then c  Q (6 x ) 
               for i = 1:c {
                         Subchannel Index             6
                         Subchannel Gain              x
                       }
               }
 Submission                                     Slide 81                                      Jianwei Zhang, Huawei
                                            CSIT_Feedback_Format
  March 2006                           doc.: IEEE 802.22-06/0050r0




   Part 2: Radio Resource Allocation
    Downlink multiuser resource allocation algorithm
    for OFDMA-based QoS-enabled WRAN system




Submission                Slide 82               Jianwei Zhang, Huawei
   March 2006                                       doc.: IEEE 802.22-06/0050r0

                                 Background

• What’s challenging for 802.22?
  (1)     Interference Avoidance to Incumbent Users (IU)
  –       No cooperation possible between incumbent & WRAN systems
           Preventive measures should be chosen at the WRAN transmitter
  –       Unknown BS-IU channels & incompatible system structure
           Isotropic transmission reduces the effective cell coverage
           Transmit-side interference pre-cancellation is impossible

  (2)     Broad available spectrum for each cell: (~180MHz, 30 TV channels)
  –       covered by multiple OFDM symbols instead of one
  –       max. one subband per each CPE
           Simultaneous multi-band channel estimation is not possible

 Submission                        Slide 83                    Jianwei Zhang, Huawei
      March 2006                                      doc.: IEEE 802.22-06/0050r0

                                Key Related Work
     Paper
•    [Wong99] C. Y. Wong, R. S. Cheng, K. B. Letaief, and R. Murch, “Multiuser
              OFDM with adaptive subcarrier, bit and power allocation,” IEEE
              Journal on Selected Areas of Communications, vol. 17, no. 10, pp.
              1747-1758, Oct. 1999.

     US Patent
•    [Li05]        X. Li, H. Liu, K. Li, and W. Zhang, “OFDMA with Adaptive
                   Subcarrier-Cluster Configuration and Selective Loading,” US
                   Patent, US6947748 B2, Sep-20 2005.

     US Patent Application
•    [Cho05]       Y.-O. Cho, et al, “Method for Allocating Subchannels in an OFDMA
                   Mobile Communication System,” US Patent Application,
                   US2005/0180354 A1, Aug-18, 2005.
    Submission                       Slide 84                   Jianwei Zhang, Huawei
  March 2006                                         doc.: IEEE 802.22-06/0050r0

                         Definitions




               Concept of band, subband, subchannel and subcarriers
Submission                  Slide 85                                  Jianwei Zhang, Huawei
   March 2006                                        doc.: IEEE 802.22-06/0050r0

                             Proposed Algorithm
• Features
  for (1) Interference Avoidance to Incumbent Users (IU)
  –       Peak power constraint, namely power mask, for every subband.
  –       Sectored antenna adopted for reducing the performance sensitivity to
          any nearby incumbent users (from a cell to only a sector).
  for (2) Broad available spectrum for each cell: (~180MHz, 30 TV channels)
  –       Two-layer resource allocation algorithm:
          Layer-1: subband allocation
          – distribute users over subbands exploiting knowledge of power mask.
               avoid over-congestion of subbands.
          Layer-2: in-subband subchannel and power allocation
          – maximize subband throughput: QoS-enabled, priorities allowed.

 Submission                         Slide 86                    Jianwei Zhang, Huawei
  March 2006                                                      doc.: IEEE 802.22-06/0050r0

                                2-Layer Algorithm

                                        Dynamic Frequency Selection Block


                                                   Layer-1 Allocation
               Knowledge of transmit              Subband Assignment
               power mask on every
               subband in every
               sector


                                                   Layer-2 Allocation
                                                       In-subband
               Knowledge of                        Subchannel, Power
               channel gain of the                 and Rate Allocation
               assigned subband




                  The two-layer structure of the multiuser resource allocation
                                           algorithm

Submission                             Slide 87                              Jianwei Zhang, Huawei
     March 2006                                                                          doc.: IEEE 802.22-06/0050r0

                            Layer-1: Subband Assignment

Intuition:      Subband with smaller allowed maximum transmit power should handle less
                CPEs
Step 1:         For each sector, eliminate those unserviceable subbands, defined as
                those subbands with the power mask value smaller than a threshold.
Step 2:         Define Pmm,b,c as the average power mask per subchannel of subband b, i.e.
                the peak possible transmit power per subchannel, in sector c. Let Kc be the
                total number of users in sector c. For each sector c, the number of users
                allocated to subband b, represented by Kb,c, is done according to the following
                equation:


                   K b ,c  K c 
                                                                                        
                                       f subband Pm m,b ,c | Pm m,1,c ,, Pm m, Nb ,c  f sectorPm m,b ,c | Pm m,b ,c ,, Pm m,b , L 

                                                    P                                       
                                                                | Pm m,1,c ,, Pm m, Nb ,c  f sectorPm m,b ,c | Pm m,b ,c ,, Pm m,b , L 
                                    Nb

                                    f
                                    b 1
                                           subband   m m,b ,c




   Submission                                            Slide 88                                             Jianwei Zhang, Huawei
        March 2006                                                                      doc.: IEEE 802.22-06/0050r0

                             Layer-1: Subband Assignment

Step 2: (cont’d)         where Nb is the number of serviceable subbands and L is the
                                 number of sectors.
                         Both f sectorPmm,b,c | Pmm,b,c ,, Pmm,b, L  and f subbandPmm,b,c | Pmm,1,c ,, Pmm, N ,c    b


                         should be non-decreasing functions of Pm m,b,c .
Example functions:
If the objective is to maximize the minimum average user data rate, we can use:
(i)                                                                  
                     f subband Pmm,b,c | Pmm,1,c ,, Pmm, Nb ,c  log 1   b Pmm,b,c 
where b can be set to the average channel power gain to noise ratio.
                     f sectorPm m,b ,c | Pm m,b ,c ,, Pm m,b , L   Pm m,b ,c /  Pm m,b ,c
                                                                                  L
(ii)
                                                                                 c 1

- (i) approximates the rate of each subchannel in subband b of sector c.
- (ii) reflects the relative number of possible subchannel allocation across
  different sectors for that subband b.

      Submission                                       Slide 89                                   Jianwei Zhang, Huawei
     March 2006                                                                       doc.: IEEE 802.22-06/0050r0
                             Layer-1: Subband Assignment
Step 3:         Randomly select Kb,c users for subband b in sector c.
Remarks:        Step 3 is indeed up to the vendors. e.g. Assignment can be done based on
                user classes so that users of higher class may be distributed to a subband with
                larger power mask.

Example:        Advantages of exploiting one-dimensional (within sector) and two-dimensional
                (across sector & subband) power mask against equal user allocation.
Objective:
maximize the minimum average user data rate.
System Settings:
3 sectors, 2 subbands, 40 subchannels per subband, 60 users per sector.
(i) 1-D (Single-sector) allocation
                    f sectorPmm,b,c | Pmm,b,c ,, Pmm,b, L   1

(ii) 2-D (Multi-sector) allocation

                   f sectorPm m,b ,c | Pm m,b ,c ,, Pm m,b , L   Pm m,b ,c /  Pm m,b ,c
                                                                                L


                                                                               c 1
   Submission                                         Slide 90                                  Jianwei Zhang, Huawei
  March 2006                                   doc.: IEEE 802.22-06/0050r0

                       Layer-1: Performance

 Subchannel power masks in the example for the Layer-1 algorithm:


                        Subband 1                Subband 2

                    Subchannel Power         Subchannel Power
                         Mask                     Mask

    Sector 1                20                       20

    Sector 2                40                       20

    Sector 3                40                        0




Submission                       Slide 91                 Jianwei Zhang, Huawei
   March 2006                                                       doc.: IEEE 802.22-06/0050r0

                                  Layer-1: Performance

  Subchannel allocation and subchannel data rate for the Layer-1 algorithm
  example with 40 subchannels per subband:

                   Equal Allocation              1-D (single-sector)              2-D (multi-sector)
                                                     Allocation                      Allocation
              Subband 1      Subband 2      Subband 1          Subband 2     Subband 1          Subband 2
Sector
          #Sub- Data #Sub- Data #Sub- Data #Sub- Data #Sub- Data #Sub- Data
            ch.  Rate   ch.   Rate  ch.   Rate  ch.   Rate  ch.   Rate  ch.   Rate
          Alloc- per   Alloc- per  Alloc- per  Alloc- per  Alloc- per  Alloc- per
           ated subch. ated subch. ated subch. ated subch. ated subch. ated subch.

  1           8     4.3923   20   4.3923     8        4.3923   20   4.3923   8      4.3923      20        4.3923


  2           16    5.3576   20   4.3923    16        5.3576   20   4.3923   16     5.3576      20        4.3923


  3           16    5.3576   0        0     16        5.3576   0       0     16     5.3576       0          0


 Submission                                Slide 92                               Jianwei Zhang, Huawei
  March 2006                                                                  doc.: IEEE 802.22-06/0050r0

                                     Layer-1: Performance
 Effect of different user allocation algorithms on the subband data rate per
 user with 60 users per sector: (Differences are highlighted)

                   Equal Allocation                   1-D (single-sector)                     2-D (multi-sector)
                                                          Allocation                             Allocation
              Subband 1         Subband 2          Subband 1          Subband 2          Subband 1          Subband 2
Sector
             No. of   Bits     No. of   Bits     No. of     Bits     No. of    Bits     No. of    Bits    No. of      Bits
             users    per      users    per      users      per      users     per      users     per     users       per
             Alloc-   User     Alloc-   User     Alloc-     User     Alloc-    User     Alloc-    User    Alloc-      User
             ated              ated              ated                ated               ated              ated

   1          30      1.1713    30      2.9282     30       1.1713    30       2.9282    17      2.0670     43        2.0429


   2          30      2.8574    30      2.9282     33       2.5976    27       3.2536    30      2.8574     30        2.9282


   3          30      2.8574    30        0        60       1.4287     0         --      60      1.4287      0          --


Submission                                       Slide 93                                     Jianwei Zhang, Huawei
    March 2006                                       doc.: IEEE 802.22-06/0050r0

                          Layer-1: Performance

Advantage of 1-D allocation over Equal Allocation:
- realized in Sector 3:
 min. average rate per user increases from 0 to 1.4287.


Advantage of 2-D allocation over its 1-D counterpart (also Equal Allocation):
- realized in Sector 1:
 min. average rate per user increases from 1.1713 to 2.0429.




  Submission                       Slide 94                     Jianwei Zhang, Huawei
    March 2006                                        doc.: IEEE 802.22-06/0050r0

                     Layer-2: In-subband Allocation

Objective:
– Maximize subband throughput by subchannel (a group of pre-selected
   subcarriers) and power allocation.
– Support differentiated-QoS service.
– Allow flexible tradeoff between max. throughput and fairness among users.
Problem Formulation:
– divided into two cases:
   (i)   individual subcarrier power gain is known.
   (ii) average channel power gain is known,
– The proposed algorithm is optimal for case (i), and almost optimal for case (ii)
   if every subchannel is within the coherence bandwidth.

  Submission                        Slide 95                    Jianwei Zhang, Huawei
    March 2006                                                                                                  doc.: IEEE 802.22-06/0050r0

                             Layer-2: In-subband Allocation

Problem Formulation:

                             Subchannel Sharing Factor                                                 Power allocated to user k on subchannel i
                                      K                      M
                     max            w 
                   k ,i {0 ,1} , 
                                                       k                 k ,i   f k ,i ( Pk ,i )
                  P 0              k 1                   i 1
                  k ,i             
                                                                                                                                lQoS _ Class ( k )
                     M     K                 Ni
                                                                                                             e.g.   wk   k
subject to             P
                     i 1 k 1
                                   k ,i
                                          n ( i ) 1
                                                           k ,n ( i )    PTotal

                                                                                                         rate control (0 k 1)
                       K
                         k ,i  1  i  {1,2,, M }                                                                         priority control
                      k 1                                                                                                   (lQoS_Class(k)  0)


                    0  Pk ,n (i )  Pkmnask  PTotal
                                       , (i )                                                      i   ,2,, M , k   ,2,, K , n(i)   ,2,, N i 
                                                                                                         1                1                   1


               Power mask                                               #subchannels                                 #users                  #subcarriers in
                                                                                                                                              subchannel i
  Submission                                                                    Slide 96                                         Jianwei Zhang, Huawei
    March 2006                                                                                             doc.: IEEE 802.22-06/0050r0

                                 Layer-2: In-subband Allocation

Define the rate function f k ,i ( Pk ,i ) as:
                                                             Ni
                                      f k ,i ( Pk ,i )     f
                                                           n ( i ) 1
                                                                        k ,n ( i )   ( Pk ,n ( i ) )

                                                    h            2
                                                                                                       
                                                                    Pk ,n ( i )                       
                                          Ni                                                                               Ni
                                       log 2 1                                                                                  Pk ,n (i )  Pk ,i
                                                       k ,n ( i )
                                                                                                               with
                                       n ( i ) 1          n    2
                                                                                                                        n ( i )1
                                                                                                      
where  n
        2
                            is the noise power,

          hk ,n ( i )
                        2     is the average channel power gain of subcarrier n(i) in
                              subchannel i, and
                         1.5
                                    is a factor bridging the gap between ideal minimum
                     ln( 5BER)
                                      power required (using mutual information) and actual
                                      required transmission power (using practical modulation
                                      schemes for a given rate)

  Submission                                                 Slide 97                                                  Jianwei Zhang, Huawei
      March 2006                                                                doc.: IEEE 802.22-06/0050r0

                             Layer-2: In-subband Allocation

Proposed algorithm
-    by relaxing  k ,i  {0,1} to  k ,i  [ 0,1] , the problem becomes convex and
     method of Lagrangian can be applied to obtain the optimal solutions.
Algorithm Details:
Step 1: Initialization.
        Initialize Ω  0 .
Step 2: Select the optimal CPE for each subcarrier for a given value of Ω
        CPE k is selected ( k ,i  1 ) for subcarrier i according to the following criterion:
                             *



                    1
                                   if Gk ,i (Ω )  maxGk ,i (Ω ) 
                  
                   *                                    k
                                                                                  i
                    0
                   k ,i
                                    otherwise
                   
     where                                                        ~        ~                        ~
                                  Ni                           Ω        Ω     Ni
                                                                                                 Ω     
                 Gk ,i (Ω )  wk   f k ,n (i ) ( f ' 1n (i ) 
                                                                w
                                                                      ) 
                                                                       w        1 f ' 1,n(i )  w
                                                                                                 
                                                                                                        
                                                                                                        
                                  n (i ) 1           k,                               k
                                                                k        k   n (i )           k    
    Submission                                     Slide 98                                     Jianwei Zhang, Huawei
    March 2006                                                                       doc.: IEEE 802.22-06/0050r0

                                  Layer-2: In-subband Allocation
     ~
with Ω defined as

                                                            1  Ω      
                                    
                                 m ask
               wk f' k,n(i ) Pk,n(i )                if f' k,n(i ) 
                                                                     w    Pk,nask
                                                                         
                                                                               m
                                                                                 (i )
                                                                     k 
           ~                                                        1  Ω 
           Ω  Ω                                      if 0  f' k,n(i )    Pk,nask
                                                                         w 
                                                                                      m
                                                                                        (i )
                                                                         k
                                                            1  Ω 
               wk f' k,n(i ) 0                      if f' k,n(i )    0
                                                                     w 
               
                                                                     k
Step 3: Compute the optimal allocated power for each CPE for a given value of Ω.
The optimal average power for user k on subchannel i is:

                                                 Ω
                                                   ~
                                                                                                 
                                                             *  wk      n 2
                                                                                                   
             *
           c k .n ( i )     k .i f ' 1n ( i ) 
                              *
                                       k,        w
                                                          k .i ~ 
                                                                Ω                                
                                                                                                  
                                                                                               2
                                                  k                 hk ,n (i )                 
  Submission                                             Slide 99                                      Jianwei Zhang, Huawei
       March 2006                                                         doc.: IEEE 802.22-06/0050r0

                                                Layer-2: In-subband Allocation

Step 4: Coarse adjustment of Ω.
If ( Ω  0 ),
            K     M       Ni
    If (     c
           k 1 i 1 n ( i ) 1
                                    k ,n (i )    PTotal ),


           Set Ω  0 for some small Ω . Go back to Step 2.
    Else
           Optimal solutions obtained; algorithm terminated.
    End
Else
            K    M       Ni
    If (     c
           k 1 i 1 n ( i ) 1
                                  k ,n (i )    PTotal ),


           Ωlower  Ω; Ω  2Ω. Go to Step 2.

   Submission                                                 Slide 100             Jianwei Zhang, Huawei
    March 2006                                                                        doc.: IEEE 802.22-06/0050r0

                                      Layer-2: In-subband Allocation

Step 4 (Cont’d):
                K    M        Ni
   Elseif (               c          k ,n (i )    PTotal ),
               k 1 i 1 n ( i ) 1



        Go to Step 5.
   Else
        Optimal solutions obtained; algorithm terminated.
   End
End
Step 5: Fine adjustment of Ω.
                     K    M        Ni
   While (
                      c
                    k 1 i 1 n ( i ) 1
                                            k ,n (i )    PTotal     ) for some predefined tolerance level ,


  Submission                                                      Slide 101                      Jianwei Zhang, Huawei
  March 2006                                                                            doc.: IEEE 802.22-06/0050r0

                            Layer-2: In-subband Allocation

             Step 5 (Cont’d):
                    Repeat Step 2 and 3.
                            K    M       Ni
                    If (     c
                           k 1 i 1 n ( i ) 1
                                                  k ,n (i )    PTotal ),


                            Ωlower  Ω;
                                     K     M        Ni

                    Elseif (                     c         k ,n (i )    PTotal ),
                                    k 1 i 1 n ( i ) 1



                            Ωupper  Ω;
                    End

                    Ω  Ωlower  Ωupper / 2;
                    Repeat Step 2 and Step 3.
             End
Submission                                                      Slide 102                         Jianwei Zhang, Huawei
    March 2006                                                doc.: IEEE 802.22-06/0050r0

                             Layer-2: In-subband Allocation

In case of oscillations between two assignment profiles,
               K    M                            K   M
       Pu   c k ,i  PTotal   and     Pl   c k ,i  PTotal ,
               k 1 i 1                        k 1 i 1

a time-sharing ratio (  u :  l ) for these profiles can be calculated:

               PTotal  Pl
       u                       where      l  1  u
                Pu  Pl

so that on average the total power constraint is satisfied.




  Submission                             Slide 103                      Jianwei Zhang, Huawei
    March 2006                                       doc.: IEEE 802.22-06/0050r0

                     Layer-2: Channel Quantization

In our numerical results, the following simple channel quantization algorithm is
used:
Quantization lookup table construction:
1. Acquire the channel power gain distribution.
2. Identify the range of the channel power with a desirable probability of
  occurrence, say 90%.
3. Equally partition the corresponding range in the logarithm domain.
4. Set up the thresholds as the middle points of each interval in the logarithm
  domain.
5. Transform the thresholds into their corresponding thresholds in the original
  domain.

  Submission                       Slide 104                    Jianwei Zhang, Huawei
    March 2006                                      doc.: IEEE 802.22-06/0050r0

                       Layer-2: Rho Quantization

When time-sharing cannot be implemented, the following two algorithms can be
used:

Algorithm 1:
Step 1: Select the assignment profile closest to the Total Power Constraint.
Step 2: Perform optimal power allocation for that assignment set.

Algorithm 2: (shown good enough through numerical evaluation)
Select the assignment profile with the total power smaller than the Total Power
Constraint. In practice, perfect channel information feedback may not be possible
but limited number of bits is used instead.



  Submission                       Slide 105                   Jianwei Zhang, Huawei
    March 2006                                    doc.: IEEE 802.22-06/0050r0

                         Layer-2: Performance
Sum rate comparison of (i) optimal SPA, (ii) random SA & optimal PA and
(iii)random SA & equal PA with effects of channel quantization:

                                                        Legend:
                                                          - Perfect
                                                          - 3-bit quantization
                                                          - 1-bit quantization




  Submission                     Slide 106                   Jianwei Zhang, Huawei
  March 2006                                           doc.: IEEE 802.22-06/0050r0

                           Layer-2: Performance
Percentage loss of sum rate for the optimal subchannel and power allocation due to
channel quantization:




Submission                          Slide 107                      Jianwei Zhang, Huawei
    March 2006                                       doc.: IEEE 802.22-06/0050r0

                           Layer-2: Performance

Sum Rate Performance:

Optimal Subchannel and Power Allocation:

• 3-bit Channel Quantization is sufficiently good (~ 1% loss).

• 1-bit Channel Quantization is fairly good (~ 9% loss).


Random Subchannel Assignment with Optimal/Equal Power Allocation:

• Even 1-bit Channel Quantization gives apparently the same performance.




  Submission                       Slide 108                     Jianwei Zhang, Huawei
     March 2006                                           doc.: IEEE 802.22-06/0050r0
                             Layer-2: Performance
Number of iterations required for convergence with 3-bit channel quantization and power
constraint accuracy of 99.999998%:




   Submission                         Slide 109                       Jianwei Zhang, Huawei
     March 2006                                                 doc.: IEEE 802.22-06/0050r0
                                Layer-2: Performance
Complexity Issue:
Optimal Subchannel and Power Allocation
(i) Number of operations1  required:
    (Number of users)*(Number of subcarriers or subchannels2)
   *(Number of iterations3)
Random Subchannel Assignment with Optimal Power Allocation:
(i) Number of operations1  required:
   (Number of subcarriers or subchannels2)*(Number of iterations3)
Random Subchannel Assignment with Equal Power Allocation:
• Two steps: random subchannel assignment + peak power clipping according to the Power Mask
values.
Remarks:
1. includes mainly the calculation of power and rate.
2. when the same channel gain and power mask are used in a subchannel.
3. fairly independent of the total number of users, of order O(log(FFT Size)) assuming
   same #subchannels for all FFT sizes.
   Submission                             Slide 110                          Jianwei Zhang, Huawei
    March 2006                                      doc.: IEEE 802.22-06/0050r0
                          Layer-2: Performance
Percentage of the occurrence of subchannel sharing with the application of 3-bit
channel quantization.




  Submission                      Slide 111                    Jianwei Zhang, Huawei
    March 2006                                      doc.: IEEE 802.22-06/0050r0
                          Layer-2: Performance
Percentage loss of sum rate among the cases of subchannel sharing with
sharing factor quantization for the optimal subchannel and power allocation:




  Submission                      Slide 112                    Jianwei Zhang, Huawei
    March 2006                                         doc.: IEEE 802.22-06/0050r0

                            Layer-2: Performance

- Sharing rarely occurs (~2%).
- Actual loss due to rho-quantization in total data rate is negligible
(~0.01% loss with rho-quantization Algorithm 2 among scenarios
 with time-sharing).




  Submission                         Slide 113                     Jianwei Zhang, Huawei
   March 2006                                             doc.: IEEE 802.22-06/0050r0

                                       Conclusion
• Developed a two-layer resource allocation algorithm for
  the downlink IEEE 802.22 WRAN Systems, featuring
  –           interference avoidance to incumbent users
  –           user pre-distribution over subbands in a cell, avoiding over-congestion
                  of subbands in a way that subband with a larger power mask (max.
                  transmit power possible) should handle more CPEs
  –           efficient in-subband subchannel and power allocation for:
              (i)   maximizing subband throughput at affordable complexity,
              (ii) allowing QoS to be guaranteed,
              (iii) allowing prioritized transmission and flexible tradeoff between
                        maximum throughput and fairness among users.



 Submission                            Slide 114                     Jianwei Zhang, Huawei
  March 2006                           doc.: IEEE 802.22-06/0050r0




  Part 2: Radio Resource Allocation
   Joint dynamic frequency selection and power control
   with user specific transmit power mask constraints
   in uplink WRAN system using OFDMA scheme




Submission               Slide 115               Jianwei Zhang, Huawei
   March 2006                                        doc.: IEEE 802.22-06/0050r0

                                Background (1)
• Principles of WARN systems
         shares the VHF/UHF TV bands between 47MHz-910MHz which are being
          used by the licensed operators and other license-exempt (LE) devices.
         a main constraint is to avoid interference to incumbent services such as
          TV broadcasting (analog and digital) and Public Safety systems.
• Role of Dynamic Frequency Selection
         performs multiple-access control to provide QoS-guaranteed services
          required in the WRAN standard while not disturbing the service quality of
          the licensed users.
         involves user selection, rate adaptation as well as transmit power control
          (TPC).



 Submission                        Slide 116                   Jianwei Zhang, Huawei
   March 2006                                                                                    doc.: IEEE 802.22-06/0050r0
                                                Background (2)
Role of Dynamic Frequency Selection (Cont’)
         The spectrum occupation information, called geographical spectrum state
          information (GSSI), is obtained by data fusion and acts as the input
          information for dynamic frequency selection (DFS).
              – Usually full GSSI may not be easy to obtain.
          – Instead of full GSSI, one possible form of partial GSSI is transmit power masks
             imposed on all WRAN transmitters.


                                                                       CPE2
                                      CPE1
                                                                   ×
                                           ×{n , n }={1, 1} {n , n }={1, 6}
                                                 b   c
                                                                   b   c
                                                                                                 CPE3
                                                                                                   ×
                                                                                             {nb, nc}={2, 3}
                         {nb, nc}={5, 5}                     BS
                               ×j                                              ×CPE
                                                                                         k

                             CPE                                           {nb, nc}={3, 7}
                                                 {nb, nc}={5, 8}                                               nb: Index of subbands
                                                                                                               nc: Index of subchannels
                                                           × CPEn



 Submission                                              Slide 117                                              Jianwei Zhang, Huawei
    March 2006                                                       doc.: IEEE 802.22-06/0050r0
                                 Related Works
· In patent US2005180354 “Method for allocating subchannels in an OFDMA
  mobile communication system”, Cho et al. proposed resource allocation
  algorithms to maximum the transmission rates of all users by allocating
  subchannels and bits.
· The scheme introduced an adaptive modulation using linear programming
  into an existing scheme for a system including a single kind of users, thereby
  enabling simultaneous execution of the adaptive modulation for all users in a
  system including two kinds of users.
                                    max          z
                                 ck ,n , k ,n

                                         N
                       s.t. Rk   ck ,n   k ,n  z for all k
                                        n 1
                             K      N

                            
                             k 1 n 1
                                           f k (ck ,n )  k ,n /  k2,n  PT

                             K


  Submission
                            
                             k 1
                                      k ,n    1 for all n
                                         Slide 118                             Jianwei Zhang, Huawei
    March 2006                                                        doc.: IEEE 802.22-06/0050r0

                                  Related Works

· In paper “Multiuser OFDM with adaptive subcarrier, bit and power allocation,”
  Wong et al. considered a subcarrier, bit and power allocation problem in
  OFDM system.
· The objective is to the minimize the total transmitted power, given the
  minimum data rate requirement of each user.


                                                    K   N      k ,n
                             min
                            ck ,n ,  k ,n
                                                      2 f k (ck ,n )
                                                    k 1 n 1   k ,n
                                             N
                     s.t.   Rk   ck ,n   k ,n for all k
                                             n 1
                              K

                            
                            k 1
                                        k ,n      1 for all n


  Submission                                 Slide 119                          Jianwei Zhang, Huawei
      March 2006                                       doc.: IEEE 802.22-06/0050r0

                      Drawbacks of the Related Work

•    For the patent US2005180354, the problem considered here is actually a rate
     adaptive problem which maximizes a lower bound of all users’ throughput
     with respect to a transmit power budget.
•    Delay constraints and users’ priorities were not considered in this invention.
•    It cannot be applied in WRAN systems since it does not employ any
     technique to guarantee free interference to the incumbent users.
•    Subband allocation among multiple OFDM symbols was not investigated.




    Submission                       Slide 120                    Jianwei Zhang, Huawei
  March 2006                                                             doc.: IEEE 802.22-06/0050r0

                                   Our Proposed Algorithm
 • Two-Layers’ Design
                                     Dynamic Frequency Selection Block


                                              Layer-1 Allocation
             Knowledge of transmit           Subband Assignment
             power mask on every
             subband in every
             sector


                                              Layer-2 Allocation
                                                  In-subband
             Knowledge of                     Subchannel, Power
             channel gain of the              and Rate Allocation
             assigned subband




Submission                                  Slide 121                              Jianwei Zhang, Huawei
    March 2006                                                                          doc.: IEEE 802.22-06/0050r0
                              Layer 1 (Subband Allocation)
• Method 1: (Sum-Rate-Max Strategy)

Step 1: For each 6-MHz subband b, create a list of CPEs in descending order of
their transmit power mask values. CPEs with power mask values smaller than a
serviceable threshold predefined a priori are eliminated.

Step 2: Create a list of CPEs in descending order of their maximum power
                                              ~
mask values across subbands. Define Pkmbask as the normalized power mask per
                                                ,

subchannel of user k on subband b.
   For k = Lmax (1) to Lmax ( K total )where k  Lmax ,
                                                                                                    
         (i)   b k  arg max 
                                   f           
                                             ~ m ask ~ m ask        ~ m ask
                                                                                
                                   subch Pk ,b | P1,b ,  , PK total ,b  f rate Pk ,b  
                                                                                         ~ m ask
                                                                                                    
                                                                                                     
                                                                                                     
                    b1,, N b 
                                  
                                   kL
                                               ~
                                                  ,     ~
                                                            b
                                                                      ~ ask
                                                                                          ~
                                                                                             
                                    b f subch Pkmbask | P1,m ask ,, PKmtotal ,b  f rate Pkmbask
                                                                                              ,      
                                                                                                     
                                                                                                     
                                                                                                    

         (ii) Remove CPE k from Lb for all b’s except bk.
    End                                                                                                  Jianwei Zhang, Huawei
  Submission                                            Slide 122
     March 2006                                                                  doc.: IEEE 802.22-06/0050r0

                            Layer 1 (Subband Allocation)
(Cont’)
The functions f subch Pkmask | P1,mask ,, PKmaskb  and f rate Pkmask  should be non-decreasing
                       ~        ~           ~                                ~
                         ,b        b             ,       total
                                                                    ,b
functions. For example,

                                          ~
                                             ,b                ~
                                                                 
                                    f rate Pkmask  log 1   b Pkmask
                                                                  ,b     
                                                                    
                                                                             b
                                ~ m ask ~ m ask    ~ m ask      ~ m ask L ~ m ask
                      f subch   Pk ,b | P ,b ,  , PKtotal ,b  Pk ,b /  Pk ,b
                                         1
                                                                         k 1


where b can be set to the average channel gain to noise ratio.

Step 3 (Optional): Perform subband re-assignment starting from the CPE with the
minimum f subchPkmask | P,mask,, PKmaskb  f rate Pkmask  .
                ~
                  ,b
                         ~
                          1b
                                   ~
                                        ,    total
                                                      ~
                                                        ,b




   Submission                                            Slide 123                         Jianwei Zhang, Huawei
    March 2006                                       doc.: IEEE 802.22-06/0050r0

                      Layer 1 (Subband Allocation)

• Method 2: (Round-Robin-Max Strategy)
Step 1: For each 6-MHz subband b, create a list of CPEs Lb in descending order
of the transmit power mask values. CPEs with power mask values smaller than a
serviceable threshold predefined a priori are eliminated.
Step 2: Sort the subbands in descending order of their maximum power mask.
Starting from index 1, i.e. the subband with the largest maximum power mask,
each subband takes turn to pick up one CPE with the maximum transmit power
mask. Any CPE selected in the previous subband will be subtracted from the list of
the latter subbands. Repeat Step 2 until the lists of all the subbands are empty.




  Submission                       Slide 124                    Jianwei Zhang, Huawei
  March 2006                                                   doc.: IEEE 802.22-06/0050r0

                An example for Layer-1 Algorithm (1)

    Subchannel Power Masks and the Approximated Subchannel Data Rate

                             Subband 1                                Subband 2
    CPE        Subchannel Power    Approx. Subchannel   Subchannel Power    Approx. Subchannel
                    Mask               Data Rate             Mask               Data Rate
    1                50                  5.6725                0                    0
    2                50                  5.6725                0                    0
    3                60                  5.9307               50                  5.6725
    4                30                  4.9542               20                  4.3923
    5                10                  3.4594               30                  4.9542
    6                40                  5.3576               20                  4.3923




Submission                               Slide 125                            Jianwei Zhang, Huawei
       March 2006                                                                        doc.: IEEE 802.22-06/0050r0

                            An example for Layer-1 Algorithm (2)

                       CPE Assignment and Corresponding Subchannel Data Rates


              Sum-Rate-Max                                     Round-Robin-Max                                          CPE-Max
       Subband 1                 Subband 2              Subband 1                Subband 2              Subband 1                     Subband 2
  CPE         Data          CPE         Data       CPE           Data       CPE         Data       CPE          Data          CPE            Data
selected    Rate per      selected    Rate per   selected      Rate per   selected    Rate per   selected     Rate per      selected       Rate per
             subch.                    subch.                   subch.                 subch.                  subch.                       subch.
   -               -         3         5.6725       3          5.9307        -               -      1         5.6725              -               -
   1         5.6725          -               -      -               -        5         4.9542       2         5.6725              -               -
   2         5.6725          -               -      1          5.6725        -               -      3         5.9307              -               -
   6         5.3576          -               -      -               -        4         4.3923       4         4.9542              -               -
   -               -         4         4.3923       2          5.6725        -               -      -               -             5         4.9542

   -               -         5         4.9542                                6         4.3923       6         5.3576              -               -




  Submission                                                Slide 126                                       Jianwei Zhang, Huawei
        March 2006                                                                   doc.: IEEE 802.22-06/0050r0
                            An example for Layer-1 Algorithm (3)
                     Subchannel Allocation and Corresponding Subchannel Data Rates
                    Sum-Rate-Max                             Round-Robin-Max                                    CPE-Max
            Subband 1              Subband 2          Subband 1                Subband 2          Subband 1               Subband 2
        #Subch.     Data     #Subch.      Data    #Subch.     Data        #Subch.     Data    #Subch.     Data      #Subch.      Data
        allocate    Rate     allocate     Rate    allocate    Rate        allocate    Rate    allocate    Rate      allocate     Rate
            d        per         d         per        d        per            d        per        d        per          d         per
                   subch.                subch.              subch.                  subch.              subch.                 subch.
           -         -            20     5.6725     15       5.9307            -       -      9 (8.7)    5.6725           -       -
           14                                                                17
                   5.6725         -        -         -            -                  4.9542   9 (8.7)    5.6725           -       -
         (14.3)                                                            (17.2)
           14                                        13                                          10
                   5.6725         -        -                 5.6725            -       -                 5.9307           -       -
         (14.3)                                    (12.5)                                      (10.4)
CPE
           12                                                                12
Order              5.3576         -        -         -            -                  4.3923   5 (5.2)    4.9542           -       -
         (11.4)                                                            (11.4)
                                                     12
           -         -            8      4.3923              5.6725            -       -         -          -            40     4.9542
                                                   (12.5)

                                                                             11
           -         -            12     4.9542                                      4.3923      7       5.3576           -       -
                                                                           (11.4)

Min.
CPE                      35.14                                        48.32                                      24.77
Rate
Sum
                         431.16                                       416.02                                   421.85
                                                                                                     Jianwei Zhang, Huawei
Rate Submission                                       Slide 127
   March 2006                                                                                     doc.: IEEE 802.22-06/0050r0

                                Layer-2 Allocation
• Objective
         maximize the weighted system capacity given the QoS requirements
          and power constraints
• Problem Formulation
                                       K                Nc

                             max
                              k ,nc
                                       w        k                     k , nc
                                                                                  f k , nc ( Pk , nc )          wk   k
                                                                                                                          lQoS _ Class ( k )

                                       k 1            nc 1
                             Pk ,n
                                  c


                subject to                    k ,n {0,1}  k , nc
                                                       c


                                                   K

                                              k 1
                                                                k , nc      1  nc

                                           0  Pk ,nc  Pkmask
                                                          , nc                              k , nc
                                           Nc

                                           
                                           nc 1
                                                       k , nc   Pk ,nc  Pkt              k



 Submission                             Slide 128                                                           Jianwei Zhang, Huawei
    March 2006                                                                              doc.: IEEE 802.22-06/0050r0

                                 Proposed Algorithm (1)
• Our proposed algorithm to solve Layer-2 problem is described a follows:
Step 1: Initialize all the Lagrangian multipliers u k to be zeros and set ck ,n  k ,n Pk ,n .                        c        c   c



Step 2: Selection of temporarily optimal CPE for each subchannel given the values of u k .
  For every subchannel and every CPE, compute
                                                         '1  uk   o
                                                                             uk  '1  uk
                                                                                  o            o
                                                                                                   
                              Gk ,nc (u )  wk  f k ,nc  f k ,nc 
                                       o
                                       k                                      f k ,nc 
                                                                                                  
  where                                        
                                                                  wk      wk          wk   
                                                                                                    

                                                                          u
                                              '
                                       wk f k ,nc (0)         f k',n1c ( k )  0
                                                                          wk
                                                                             u
                                 uk  uk
                                  o
                                                             0  f k',n1c ( k )  Pkmask
                                                                                     , nc
                                                                             wk
                                                                      ' 1 u
                                       wk f k ,nc ( Pk ,nc ) f k ,nc ( k )  Pk ,nc .
                                              '        mask                         mask

                                                                            wk


Then for each subchannel, we select the CPE k ' such that                                          k '  arg max Gk ,nc (uk )
                                                                                                                          o

                                                                                                           k

and accordingly set
                                              k ,n  1, k ,n  0 for all k  k '
                                               *
                                                 '
                                                          *
                                                                c
                                                     c




  Submission                                             Slide 129                                                 Jianwei Zhang, Huawei
    March 2006                                                                                doc.: IEEE 802.22-06/0050r0

                                            Proposed Algorithm (2)
Step 3: Compute the temporarily optimal power allocation.
         For each CPE in each subchannel, compute
                                                                                          o
                                                                                         uk
                                                   c*
                                                    k , nc      *
                                                                  k , nc   f    ' 1
                                                                               k , nc   ( )
                                                                                         wk
Step 4: Examine whether the total power limitation for each CPE is satisfied or not.
       Given the temporarily optimal values of ck ,nc and k ,n .
                                                *          *
                                                                                               c
               Nc
        If     c
               nc 1
                       *
                       k , nc    Pkt   has been satisfied for each CPE, stop.
        The optimal solutions have been obtained.
        Else, go to Step 5.


Step 5: Adjust the values of                   u k to satisfy the total power limitations.
        Denote               as the precision of the power allocation within a tolerance error.



  Submission                                                 Slide 130                                  Jianwei Zhang, Huawei
    March 2006                                                                                      doc.: IEEE 802.22-06/0050r0

                                                     Proposed Algorithm (3)
(Cont’)
                                  Nc

      While      Pkt     ck ,nc  Pkt
                              *
                                 haven’t been satisfied for all the CPE’s,
                       n 1       c

         Choose the CPE k that exceeds the most the total power limitation.
         Set that the lower bound uk(l ) to be the current value uk and the upper bound                                      (
                                                                                                                            uku )   to
         be wk f k',n (0) , where nc  arg max wk fk',n (0)
                             c                                                  c
                                                                  nc


                                       (ukl )  uku ) )
                                         (       (

            Set u
                     (m)
                     k
                                                          and repeat step 2 and step 3 using u ( m ) .
                                                2                                              k
                   Nc

            If    c
                  nc 1
                             *
                             k , nc
                                           Pkt
                                          ,
                                 set u ( l )  u ( m ) ;
                                      k         k
                        Nc

          Elseif  ck                          Pkt ,
                                  *
                                       , nc
                     nc 1
                                 set uk(u )  ukm).
                                               (



                                                          Nc
               Repeat until P     ck ,n  Pkt
                                       *
                                                k
                                                 t
                                                                                    is satisfied.
                                                                       c
                                                          nc 1
      End

  Submission                                                               Slide 131                          Jianwei Zhang, Huawei
    March 2006                                   doc.: IEEE 802.22-06/0050r0
                              Channel Quantization
· In practice, perfect channel information feedback may not be possible but
  limited number of bits is used instead.
· A simple channel quantization algorithm is provided where the index of a
  quantization table based on the estimated channel power gain is used as
  the channel feedback.

Quantization lookup table construction:
Step 1: Acquire the channel power gain distribution.
Step 2: Identify the range of the channel power with a desirable probability of
           occurrence, say 90%.
Step 3: Equally partition the corresponding range in the logarithm domain.
Step 4: Set up the thresholds as the middle points of each interval in the
           logarithm domain.
Step 5: Transform the thresholds into their corresponding thresholds in the
  Submission original   domain.    Slide 132               Jianwei Zhang, Huawei
  March 2006               doc.: IEEE 802.22-06/0050r0




                     End




Submission     Slide 133             Jianwei Zhang, Huawei