Download - Motivation
C80216m-08_218r2
Project IEEE 802.16 Broadband Wireless Access Working Group <http://ieee802.org/16>
Title 16m Resource Block Indexing Schemes Evaluation
Date Submitted
2008-03-10
Source(s) Hujun Yin, Ping Wang Intel CorporationJuejun Liu, Zhigang Rong HuaweiShanpeng Xiao, Wenqi Liao China MobileXiaolu Dong, Ying Du CATRHongyun Qu, Huiying Fang ZTEShiqiang Suo CATTJianhua Liu Alcatel Shanghai-BellXin Su, Xiaofeng Zhong Tsinghua University
E-mail: [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected]
Re: IEEE 802.16m-08/005: Call for Contributions on Project 802.16m System Description Document (SDD).
Target topic: “Downlink Control Structures”.
Abstract Evaluation of the different resource block indexing schemes for 16m.
Purpose For discussion
Notice This document does not represent the agreed views of the IEEE 802.16 Working Group or any of its subgroups. It represents only the views of the participants listed in the “Source(s)” field above. It is offered as a basis for discussion. It is not binding on the contributor(s), who 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 802.16.
Patent Policy The contributor is familiar with the IEEE-SA Patent Policy and Procedures:<http://standards.ieee.org/guides/bylaws/sect6-7.html#6> and <http://standards.ieee.org/guides/opman/sect6.html#6.3>.Further information is located at <http://standards.ieee.org/board/pat/pat-material.html> and <http://standards.ieee.org
/board/pat>.
2
C80216m-08_218r2
Motivation
• Define the methodologies to compare the resource indexing schemes
• Compare various resource indexing schemes
3
C80216m-08_218r2
Resource Indexing Metrics• Worst case number of bits per-allocation
• Calculate the worst case number of bits required for an allocation• Commonly used metric• For worst case design• Ignores allocation pattern• Ignores the size of allocation
• Worst case bit cost per-resource block allocated• Calculate the worst case bit cost for one RB• For worst case design• Ignores allocation pattern• Takes the size of allocation into consideration
• Average number of bits per-allocation• Calculate the average number of bits required for an allocation assuming a particular
allocation pattern• For average performance design• Take allocation pattern into consideration• Ignores the size of allocation
• Average number of bits per-resource block allocated• Calculate the average bit cost to allocate one resource block• For average performance design• Take allocation pattern into consideration• Take the size of allocation into consideration
4
C80216m-08_218r2
Allocation Pattern Modeling
• Uniform distribution [RB_min, RB_max]• The number of resource block allocated in each
allocation is uniformly distributed over 1 RB (RB_min) and the maximum number of RBs (RB_max)
• Traffic dependent based on EVM traffic model• Select packet size distribution based on traffic model
in EVM (e.g. AMR 12.2 VoIP traffic model)
• Fixed MCS @ 1bits/sec/Hz or generate MCS distribution from SLS
• Calculate resource block allocation distribution from the packet size distribution and MCS distribution
5
C80216m-08_218r2
Resource Allocation Evaluation Methodologies• Method 1:
• Assume resource blocks are always contiguous in one allocation• Calculate the maximum number bits x(n) required of all possible allocations Hmax
• Method 2: • Assume resource block are always contiguous in one allocation• Calculate the worst case bit cost for one RB
• Method 3: • Assume resource block are always contiguous in one allocation• Assume allocated resource block size follow distribution P(n)• Assume required number of bits to index resource block size n is x(n)• Calculate the average number of bits required for one allocation
• Method 4: • Assume resource block are always contiguous in one allocation• Assume allocated resource block size follow distribution P(n)• Assume required number of bits to index resource block size n is x(n)• Calculate the average number of bits required for one RB allocation
max
1avg )()(
RB
n
nPnxH
nnPnxRRB
n
/)()(max
1avg
NnxH n ,....,1,)max(max
NnnxR n ,....,1,)/max(max
6
C80216m-08_218r2
3GPP LTE Resource Allocation Approaches: Downlink• A resource allocation field in each PDCCH includes two parts, a type field and information consisting of the actual
resource allocation. PDCCH with type 0 and type 1 resource allocation have the same format and are distinguished from each other via the type field. For system bandwidth less than or equal to 10 PRBs the resource allocation field in each PDCCH contains only information of the actual resource allocation. PDCCH with type 2 resource allocation have a different format from PDCCH with a type 0 or type 1 resource allocation. PDCCH with a type 2 resource allocation do not have a type field.
• Resource Allocation Type 0
In resource allocations of type 0, a bitmap indicates the resource block groups that are allocated to the scheduled UE. The size of the group is a function of the system bandwidth that is shown:
DLRBN
RBG Size System Bandwidth
(P)
1 ≤10
2 11 - 26
3 27 - 64
4 64 - 110• Resource allocation type 0 is charaterized by the following:
>> Grouping of RBs (in frequency domain): Group size may depend on system BW
>> Bitmap indicates the RB groups to use
At most 28 bits for 110 RB system BW
At most 13 bits for 25 RB system BW
>> Setting the limit on control signaling overhead.
A A B B C C A A A A C C
RB
RB group
Frequency
Bitmap To A: 100110 To B: 010000 To C: 001001
7
C80216m-08_218r2
3GPP LTE Resource Allocation Approaches: Downlink
• Resource Allocation Type 1
In resource allocations of type 1, a bitmap indicates to a scheduled UE the resource blocks from the set of resource blocks from one of the P resource block group subsets where P is the resource block group size associated with the system bandwidth that is shown:
DLRBN
RBG Size
SystemBandwidth
(P)
1 ≤10
2 11 - 26
3 27 - 64
4 64 - 110
RB
Sub-group 1
Frequency
Sub-group 2
• Resource Allocation Type 1 is characterized by the following:
>> In this method, all resource blocks are divided into several sub-groups, wherein one example case dividing into two sub-groups is illustrated.
>> The scheduler uses 1 sub-group to allocate resource to one UE.
>> With this method, the number of signaling bits can be reduced because the scheduler uses one sub-group which consists of smaller number of RBs. The bitmap size can be expressed Nrb/m2, where m2 is the number of sub-groups.
8
C80216m-08_218r2
3GPP LTE Resource Allocation Approaches: Downlink
# contiguous RBs # contiguous RBs
A A A
RB Frequency
A A
1st island 2nd island
Start point Start point
• Resource Allocation Type 2
In resource allocations of type 2, the resource allocation information indicates to a scheduled UE a set of contiguously allocated physical or virtual resource blocks depending on the setting of a 1-bit flag carried on the associated PDCCH. PRB allocations vary from a single PRB up to a maximum number of PRBs spanning the system bandwidth. For VRB allocations, the resource allocation information consists of a starting VRB number and a number of consecutive VRBs where each VRB is mapped to multiple non-consecutive PRBs.
A type 2 resource allocation field consists of a resource indication value (RIV) corresponding to a starting resource block ( ) and a length in terms of contiguously allocated resource blocks ( ). The resource indication value is defined by:
startRB CRBsL
• Resource allocation type 2 is characterized by the following:
>> In this method, the scheduler allocates resource with the unit of the island which consists of several contiguous RBs. The allocation information indicates the start point and the number of contiguous RBs.
>> With this method, the number of signaling bits can be reduced because of compact expression thanks to contiguous RB allocation. However, when the number of islands is large, e.g., more than 2, the number of signaling bits becomes large. Also, it is a drawback that the number of signaling bits varies depending on the number of islands. It impacts on the receiver complexity in the UE.
>> The signaling size for this allocation can be expressed , where m3 is the number of islands.
3m2/)1Nrb(Nrblog 2
)1()1(
))1(2/)1(
startDLRBCRBs
DLRB
DLRB
startCRBsDLRB
DLRBCRBs
RBNLNNRIVelse
RBLNRIVthenNLif
9
C80216m-08_218r2
Tree Based Allocation
• Option 1: Radix-2 Binary tree
• Option 2: Annular tree
• Option 3: Binary tree/bitmap hybrid
10
C80216m-08_218r2
Radix-2 Binary Tree• Bit overhead
- Number of base nodes: N; - Number of Radix-2 based Nodes: - Number of Total Nodes based on the radix-2 tree: - Bit Overhead: ceil(log2(L))
))((log22 NceilM
• Bit Overhead of Radix-2-based Tree in case of N = 16.
15 16 17 18 19 20 21 22 23 24 25 26 3027 28 29
7 8 9 10 11 12 13 14
3 4 5 6
1 2
0
12 ML
RBs Allocated 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
GranularitySupported
1 2 X 4 X X X 8 X X X X X X X 16
Bit Overheadx(n) [Bits]
5 5 2x5 5 2x5 2x5 3x5 5 2x5 2x5 3x5 2x5 3x5 3x5 4x5 5
11
C80216m-08_218r2
Annular Tree• Annular Channel Tree Structure
>> The base nodes are located on the outmost circle. The channel tree maintains a triangular structure for the outmost three levels and then switches to a power of two structure for the remaining levels. Such a structure results in a balance between the overhead associated with making time-frequency resource assignments and flexibility in the assignments. The levels of annular channel tree can be extended with the number of base nodes increasing. For the systems supporting the number of base nodes in the range of 2^n and 2^(n-1), n>2, the structures of annular channel tree are similar, except the number of null base nodes( null base node does not map to physical time-frequency resources).
>> Bit overhead - Number of base nodes: N; - Number of Radix-2 based Nodes: - Number of Total Nodes based on the annular tree:
- Bit Overhead ceil(log2(L))
))((log22 NceilM 14 ML
Number of Nodes (N = 16)
Levels of Nodes [ceil(log2(N))]
Number of Continuous or Spaced Resource Blocks to Be Allocated [for
example, N = 16]
2 ceil(log2(N)) = 16 1 1
2 ceil(log2(N)) = 16 2 2
2 ceil(log2(N)) = 16 3 3
2 ceil(log2(N))-1 = 8 4 4
2 ceil(log2(N))-2 = 4 5 6
2 ceil(log2(N))-3 = 2 6 10
2 ceil(log2(N))-4 = 1 7 16
12
C80216m-08_218r2
Annular Tree Structure>> Take example - Number of base nodes: N = 32; - Number of Radix-2 based Nodes: M=25 = 32; - Number of Total Nodes based on the annular tree: L = 4M-1 = 4*32-1=127; - Bit Overhead for Option 1:log2(L) = log2(127) = 7bits
01 2 3
4
5
6
7
8
9
10
11
12
13
14
15
16171819
20
21
22
23
24
25
26
27
28
29
30
31
13
C80216m-08_218r2
Annular Tree Indexing
Number of Nodes (N)
Levels of Nodes [ceil(log2(N))]
Granularity of Continuous or Spaced Resource Blocks
to Be Allocated
Relationships
2 ceil(log2(N)) 1 1 G0
2 ceil(log2(N)) 2 2 G1
2 ceil(log2(N)) 3 3 G2
2 ceil(log2(N))-1 4 4 G3
2 ceil(log2(N))-2 5 6 G4 = G3+2
2 ceil(log2(N))-3 6 10 G5 = G4+4
…… …… ……
1 L = ceil(log2(N))
+3
N GL =GL-1 +2(L-4) (L>=4)
• Granularity of Continuous or Spaced Resource Blocks to be Allocated for Annular Channel Tree Structure
14
C80216m-08_218r2
Annular Tree Indexing (cont’d)• Granularity of Continuous or Spaced Resource Blocks to be Allocated for Annular Channel
Tree Structure is illustrated in case of N = 16.
0 1 2 3 4 5 6 7 8 9 10 11 1512 13 14
16 17 18 19 20 21 22 23
Granularity:1
24 25 26 27 28 29 30 31
0
32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
16
48 49 50 51 52 53 54 55
56 57 58 59
60 61
62
Granularity:2
Granularity:3
Granularity:4
Granularity:6
Granularity:10
Granularity:16
Tria
ng
ula
r S
tru
ctu
re f
or
Ou
tm
ost T
hree L
evels
A
Pow
er o
f Tw
o
Stru
ctu
re
15
C80216m-08_218r2
Annular Tree Average Overhead
RBs Allocated 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Granularity upported
1 2 3 4 X 6 X X X 10 X X X X X 16
Bit Overhead [Bits] 6 6 6 6 2x6 6 2x6 2x6 2x6 6 2x6 2x6 2x6 2x6 3x6 6
• Average Bit Overhead of Continuous or Spaced Resource Blocks to be Allocated for Annular Channel Tree Structure is illustrated in case of N = 16.
16
C80216m-08_218r2
Tree/Bitmap Hybrid• Hybrid of a Power of Two Channel Tree and Bitmap Based Assignments
>> The parent node is signaled using the power of two channel tree, while the child nodes are signaled using a bitmap, where each bit corresponds to one channel tree node. The delineation between the power of two signaling and the bitmap signaling is a system parameter, thereby making the assignment overhead flexible. >> Take example that the number of base nodes is 16 for allocations with bitmap indication. - Example 1: Parent Node 6 + bitmap ‘1000’ = base node 27 - Example 2: Parent Node 4 + bitmap ‘1110’ = base nodes 19, 20, 21 - Example 3: Parent Node 0 + bitmap ‘1100’ = channel tree nodes 3, 4
>> Bit Overhead [Which level of the child nodes is the bitmap indicated in relation to the parent nodes?] - Number of Base Nodes: N; - Number of Radix-2 Nodes: M = 2ceil(log2(N)); - System parameter delineating between the power of two signaling and bitmap signaling: ; - Bit Overhead for Option 2:
),......,4,2,1(_ MLinD LinD
LinD
M_)
_
*2(log2
17
C80216m-08_218r2
Tree/Bitmap Hybrid Indexing
• Granularity of Continuous or Spaced Resource Blocks to be Allocated for Hybrid of a Power of Two Channel Tree and Bitmap Based Assignments [in case of N = 16]
Number of Nodes[ceil(log2(N))-log2(D_Lin) - K]
Bitmap Size[log2(D_Lin) =2]
Levels of Parent Nodes [K]
Granularity of Resource Blocks
Relationships
4 log2(D_Lin) 0 1,2,3,4 G0
2 log2(D_Lin) 1 2,4,6,8 G1 = 2K*G0
1 log2(D_Lin) 2 4,8,12,16 G2 = 2*G1 = 2K*G0
Note: The granularity depends on the number of bits for the bitmap (D_Lin). In the above example, the granularity of resource blocks to be allocated is 1,2,3,4,6,8,12, and 16.
18
C80216m-08_218r2
Tree/Bitmap Hybrid Indexing (cont’d)
15 16 17 18 19 20 21 22 23 24 25 26 3027 28 29
7 8 9 10 11 12 13 14
3 4 5 6
1 2
0
Parent Nodes (3 bits)
Bitmap Nodes (4 bits)
Granularity:1 (Bitmap: 1000)
Granularity:2 (Bitmap:1100)
Granularity:3 (Bitmap:1110)
Granularity:4 (Bitmap:1111)
0#
Pra
ten
t N
od
es
(3
,4,5
,6)
Granularity:2 (Bitmap:1000)
Granularity:4 (Bitmap: 1100)
Granularity:6 (Bitmap:1110)
Granularity:8 (Bitmap:1111)1
# P
rate
nt
No
de
s
(1,2
)
2#
Pra
ten
t N
od
es
(0
)
Granularity:4 (Bitmap:1000)
Granularity:8 (Bitmap: 1100)
Granularity:12 (Bitmap:1110)
Granularity:16 (Bitmap:1111)
• Granularity of Continuous or Spaced Resource Blocks to be Allocated for Hybrid of a Power of Two Channel Tree and Bitmap Based Assignments is illustrated [N = 16].
19
C80216m-08_218r2
Tree/Bitmap Hybrid Average Overhead
• Average Bit Overhead of Continuous or Spaced Resource Blocks to be Allocated for Hybrid of a Power of Two Channel Tree and Bitmap Based Assignments is illustrated in case of N = 16.
RBs Allocated 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Granularity Supported 1 2 3 4 X 6 X 8 X 10 X 12 X X X 16
Overhead for Bitmap 4 bits
Overhead for Parent Nodes 3 bits
Bit Overhead [bits] 7 7 7 7 2x7 7 2x7 7 2x7 7 2x7 7 2x7 2x7 2x7 7
20
C80216m-08_218r2
Uniformly Distributed Resource Allocation• Uniformly distributed resource allocation within [RB_min, RB_max]• Contiguous resource block for each allocation • Number of resource blocks=128
Approaches Annular Channel
Tree
Hybrid Channel Tree
LTE Resource Allocation Radix-2-Based Tree
Bitmap:4bits
Bitmap:8 bits
Type 0* Type 1* Type 2
Bit Overhead ceil(log2(4*128-1)) = 9 bits
4+6 = 10 bits
8+5 = 13 bits
32 bits 32 bits ceil(log2(128*(128+1)))
= 14 bits
ceil(log2(2*128-1)) = 8 bits
Worst case overhead per allocation
45 bits 40 bits 39 bits 32 bits 32 bits 14 bits 48 bits
Worst case bit cost per RB
9 10 13 8 32 14 8
Average overhead per allocation
27 bits 26 bits 26bits 32 bits 32 bits 14 bits 28 bits
Average bit cost per RB
0.7296 0.7314 0.7840 1.0146 4.0585 0.5943 0.7621
Note: assume that the number of nodes is 128, which is corresponding to 20MHz bandwidth.
• *Note that LTE resource allocation type 0 and typ1 is not suitable for small packet size (e.g. VoIP) traffic due to the fact that the granularity of allocation for type 0 is 4RBs and that the granularity of allocation for type 1 is 1RB in case of the 20MHz (N = 128).
21
C80216m-08_218r2
VoIP with MCS Distribution
• VoIP packet size: 44bytes (AMR 12.2)• Resource block size: 12x6• Number of resource blocks: 128• MCS distribution generated from WiMAX SLS• Contiguous resource allocation
22
C80216m-08_218r2
Index MCS SE PMF
3 QPSK (1/2) Rep 2 0.5 0.318
4 QPSK (1/2) STBC 1 0.209
5 QPSK (3/4) STBC 1.5 0.058
6 16 QAM(1/2) STBC 2 0.207
7 16 QAM(3/4) STBC 3 0.037
8 64 QAM(1/2) STBC 3 0.089
9 64 QAM(2/3) STBC 4 0.031
10 64 QAM(3/4) STBC 4.5 0.024
11 64 QAM(5/6) STBC 5 0.016
12 QPSK(1/2) SM 2 0.001
13 QPSK (3/4) SM 3 0.000
14 16QAM(1/2) SM 4 0.000
15 16QAM(3/4) SM 6 0.007
16 64QAM(1/2) SM 6 0.004
17 64QAM(2/3) SM 8 0.000
18 64QAM(3/4) SM 9 0.000
19 64QAM(5/6) SM 10 0.000
MCS Distribution
Sample MCS distribution from WiMAX SLS with mixed mobility(60% PB3, 30% VA30, 10% VA 120)
23
C80216m-08_218r2
Overhead for VoIP Allocations Worst case bit
overhead per allocation
Worst case bit cost per RB
Average bit overhead per
allocation
Average bit cost per RB
Annular Tree 18 bits 4.5 12 bits 2.5221
Hybrid Tree (Bitmap: 4bits)
20 bits 5 14 bits 2.8024
Hybrid Tree (Bitmap: 8bits)
26 bits 6.5 18 bits 3.6431
LTE Resource Allocation
Type 0* 32 bits 4 32 bits 2.0107
Type 1* 32 bits 16 32 bits 8.0428
Type 2 14 bits 7 14 bits 3.5187
Radix-2 Based Tree
16 bits 5.33 14 bits 3.0728
• *Note that LTE resource allocation type 0 and typ1 is not suitable for small packet size (e.g. VoIP) traffic due to the fact that the granularity of allocation is 4RBs and that the granularity of allocation for type 1 is 1RB in case of the 20MHz (N = 128).
24
C80216m-08_218r2
Conclusions
• Worst case bit cost per allocation is not a good design criterion– Average bit cost per allocation provides better measure of bit
overhead for realistic traffic patterns – Average/worst case bit cost per RB provides better measure of
normalize bit overhead for each RB allocated
• Uniformly distributed allocation pattern is not a good pattern to evaluate the resource indexing overhead– More realistic traffic model should be used– Traffic model that is sensitive to resource indexing overhead
(e.g. VoIP) should be used – More accurate MCS level distribution model should provide
better resource indexing overhead calculation