See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/220761985

Codeword to layer mapping for multiple layers MIMO systems

Conference Paper · January 2010

DOI: 10.1145/1815396.1815585 · Source: DBLP

CITATIONS READS
2 2,244

4 authors, including:

Jian Wang Hua Zhou

University of Electronic Science and Technology of China Ofinno
277 PUBLICATIONS   3,827 CITATIONS    10 PUBLICATIONS   20 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Chinse Ebiz User Group Survey View project

Distributed AI View project

All content following this page was uploaded by Hua Zhou on 06 April 2017.

The user has requested enhancement of the downloaded file.

 Codeword to Layer Mapping for Multiple Layers MIMO
Systems
Jie Zhang Jian Wang
Fujitsu Research & Development Fujitsu Research & Development
Center Co., Ltd Center Co., Ltd
Beijing, P.R. China 100025 Beijing, P.R. China 100025
Jie.zhang@cn.fujitsu.com jianwang@cn.fujitsu.com
ABSTRACT
Currently multiple-input multiple-output (M IM O) technology
has been widely studied in both industrial and academic areas as it
can be utilized to transmit multiple spatially parallel streams. The
number of data streams (or layer), supported over M IM O channel,
is the maximum number of Tx/Rx antennas, where one encoded
block is conveyed over one layer or aggregated several layers. In
some cases, the transmission quality on different layers are not the
same, which can be utilized to optimize the encoded stream to
layer mapping in order to improve the link level performance. In
this paper, relying on non-equal importance of information and
parity bits in decoding, a new method to map the information bits
onto high quality layer, whereas parity bits onto low quality
layers is proposed. The performance gain obtained from the
optimized mapping scheme is verified via simulation results.
Categories and Subject Descriptors
C.2.1 [Computer-Communication Networks]: Network
Architecture and Design – wireless communication.
General Terms
Algorithms, Standardization, Performance.
Keywords
M IM O; Precoding; Layer mapping; 3GPP LTE.
1. INTRODUCTION
In modern high rate wireless communication systems, multiple
antennas are deployed at both transmit and receive side to support
multiple parallel spatial transmission streams. This provides a
mechanism to increase the reliability of signal reception (diversity),
Permission to make digital or hard copies of all or part of this work
for personal or classroom use is granted without fee provided that
copies are not made or distributed for profit or commercial
advantage and that copies bear this notice and the full citation on
the first page. T o copy otherwise, or republish, to post on servers
or to redistribute to lists, requires prior specific permission and/or a
fee.
"IWCMC’10, June 28– July 2, 2010, Caen, France. Copyright ©
2010 ACM 978-1-4503-0062- 9/10/06/...$10.00"
824
Hua Zhou and Jianming Wu
Fujitsu Research & Development
Center Co., Ltd
Beijing, P.R. China 100025
zhouhua@cn.fujitsu.com
or rate of information transfer (multiplexing), or a combination of
these aspects [4]. To facilitate multiple spatial stream
transmission in parallel, M IM O channel is split into several layers,
each layer maybe pre-coded with a given pre-coding vector
according to maximum post-SNR criterion. In this scenario, these
layers can be used to send one or more symbol streams of encoded
block. Commonly, the norm of the precoder for each layer is
normalized to unit so that the transmit power on each layer is
balanced. When total transmit power constraint is applied, the
norm of each precoder is one of number of layers.
Precoded M IM O technology has already been adopted in realistic
systems, for example, 3GPP LTE Rel-9 included dual-layer
beamforming in work items [1], and LTE Rel-10 is considering an
enhanced M IM O transmission with support of multiple layers
beamforming and/or multiple codeword streams [3]. M oreover, in
IEEE 802.16m standardization, an advanced M IM O technology is
also adopted in consideration of multiple layers and encoded
streams to boost system performance by means of both transmit
diversity and spatial multiplexing.
In current progress of LTE Rel-10 standardization, special
considerations are taken on power amplifier requirement in uplink
when applying codebook based precoding. Thus, it is desirable to
achieve low PAPR at transmitter in uplink. It is observed that
codebook could affect PAPR directly. Therefore codebook should
be designed to make the precoded M IM O with low PAPR.
Recently, cubic metric preserving (CM P) codebook [3] is
proposed to be used for M IM O precoding due to its unique
codebook property with low PAPR compared to cubic metric
friendly (CM F) codebook. However, the codeword in CM P
codebook is power imbalanced between different layers according
to inherent structure of CM P codebook, i.e., some layers
experience high post-processing SNR, and some layers experience
low post-processing SNR.
In general, a Turbo coded codeword stream is composed of two
parts; the first one belongs to information symbols constructed
from parity bits for error correcting. Because information bits are
what we convey from transmit to receive sides, it should be more
important and more protected than parity bits [2]. Current
codeword stream to M IM O layer mapping scheme is to map one
or more streams onto M IM O layers sequentially [5]. In other
words, the mapping rule doesn’t take into account differential
importance of information and parity symbols.
In this paper, we consider a general scenario behaving power
imbalance between different M IM O layers, i.e., the transmission
quality between different layers is not equal at all. In this case, if
we map the codeword stream(s) sequentially (some information
symbols are mapped onto low quality layers, and some onto high
quality layers), the decoding cannot be beneficial from differential
quality of M IM O layers for information symbols. In
consideration of such a drawback above, thus, we propose a
stream to layer mapping scheme that is able to take advantage of
transmission quality difference between different layers. In the
proposed mapping scheme, information symbols, requiring high
reliability transmission, are mapped onto high SNR layers,
whereas parity symbols, requiring low reliability transmission, are
mapped onto relatively low SNR layers. To investigate the
performance improvement from the optimized stream to layer
mapping mechanism, in this paper we exemplify a common case
with layer power imbalance caused by rank-3 CM P codebook
which is discussed in 3GPP LTE-Advanced standardization.
The mapping rules in Fig. 1 can be formulated as,
xLayer j (i)  sLi j (1)
2. CODEWORD TO LAYER MAPPING
In M IM O transmission, the original information stream is
assumed encoded by one or more symmetric coder to generate two
parts of encoded bits, i.e., information bits and parity bits. Each of
them is interleaved separately, constellation mapped, and then to
form one or more codeword streams. It is assumed that the
information and parity symbols are arranged such that the first M
symbols contain all the information bits and the remained symbols
contain parity bits. The codeword stream(s) is mapped onto
different layers, and the symbols on each layer are precoded to be
transmitted via all transmit antennas.
2.1 Conventional Mapping Rules
Here, we exemplify a mapping scheme with one or two codeword
streams, linking to 2, 3 or 4 M IM O layers, alternatively. The
conventional mapping rule for one codeword to 2, 3 and 4 layers is
shown in Fig. 1(a), (b) and (c), respectively, where si (i = 0, 1, …,
K) is output of encoder and modulator to be transmitted over
layers. According to these mapping rules, the encoded stream is
transmitted through 2, 3 or 4 layers via multiple antennas.
which means sLi  j is mapped onto xLayer j (i) , the i-th symbol
of j-th layer of L total layers.

Precoding
s0 s2 s4 ... sK/2-1
Codeword to
s0 s1 s2 ... sK layer mapping
s1 s3 s5 ... sK/2

(a) One codeword to two layers

s0 s3 s6 ... sK/3-2

Precoding
Codeword to
s0 s1 s2 ... sK layer mapping
s1 s4 s7 ... sK/3-1

s2 s5 s8 ... sK/3-1

(b) One codeword to three layers

s0 s4 s8 ... sK/4-4
Precoding

s1 s5 s9 ... sK/4-2
Codeword to
s0 s1 s2 ... sK layer mapping
s2 s6 s10 ... sK/4-2

s3 s7 s11 ... sK/4-1

(c) One codeword to four layers

Fig. 1. Codeword stream to layer mapping rules for one stream and 2/3/4 layers pre -coded MIMO systems.

825
In LTE Rel-8/9, maximum two codeword streams shall be M oreover, the stream to layer mapping with more than 2
supported for maximum 4 layers pre-coded M IM O configuration. codeword streams and up to 8 layers is also being discussed with
Fig. 2(a) ~ (c) show the mapping rules for the cases of 2 codeword the similar manner as in Fig. 2. As the focus of this paper is to
j propose the optimized mapping scheme, which is independent of
streams and 2 ~ 4 layers M IM O where si (i = 0, 1, …, K; j = 0, 1)
is the i-th constellation of j-th codeword stream. In Fig. 2(a), two the specific number of codeword streams and/or layers, the
codeword streams are mapped onto two layers respectively. Fig. mapping rules with more than 2 streams and up to 8 layer are
2(b) maps the first stream to the first layer, and maps the second omitted for space limitation.
stream on the second and third layers alternately. When there are
2.2 Power Imbalance between Different Layers
two codeword streams and four layers, the first stream is mapped
Note in Fig. 1 and Fig. 2 that the length of codeword and number
onto the first two layers and the second stream onto the last two
of resource elements are chosen appropriately to illustrate the
layers.
mapping rules in the above two figures. It can be observed that the
The mapping rules for the case of 2, 3 and 4 layer in Fig. 2 can be stream is mapped onto different layers alternately no matter how
expressed as (2.a), (2.b) and (2.c) respectively. many codewords and/or how many layers exist. This is optimal
xLayer j (i)  si
j
, (2.a) when the transmission quality on different layers is exactly the
same. However, in the case that there exists power imbalance
xLayer 0 (i )  si
0
between different layers, (i.e., transmission quality on different
xLayer j (i )  s2i  j 1 ( j  1,2)
1
layers is different) the mapping rules above may suffer from
, (2.b) performance degradation. In this case, more efficient mapping
xLayer j (i )  s2i  j ( j  0,1)
0
. (2.c) method to incorporate such power imbalance should be designed.
xLayer j (i )  s ( j  1,2)
1
2 i  j 2

Precoding
s00 s10 s20 ... sK0 s0 0 s1 0 s2 0 ... sK 0
Codeword to
layer mapping
s01 s11 s21 ... sK1 s0 1 s1 1 s2 1 ... sK 1

(a) T wo codewords to two layers

s0 0 s1 0 s2 0 ... sK 0

Precoding
s0 0 s1 0 s2 0 ... sK 0
Codeword to
layer mapping s0 1 s2 1 s4 1 ... sK/2-11
s0 1 s1 1 s2 1 ... sK 1

s1 1 s3 1 s41 ... sK/21

(b) T wo codewords to three layers

s00 s2 0 s4 0 ... sK/2-10

Precoding

s00 s10 s20 ... sK0 s10 s3 0 s4 0 ... sK/20

Codeword to
layer mapping
s01 s11 s21 ... sK1 s01 s2 1 s4 1 ... sK/2-11

s11 s3 1 s4 1 ... sK/21

(c) T wo codewords to four layers

Fig. 2. Codeword stream to layer mapping rules for 2 streams and 2/3/4 layers pre -coded MIMO systems.

826
Authors in [2] proposed adaptive mapping scheme to map the
codewords to layers to take into account the instantaneous link
quality difference on different layers. Receiver estimates the
channel state information (CSI), determines the SNR level of all
M IM O layers, then feds back this information to transmitter so as
to adapt codeword stream to layer mapping to real-time M IM O
channel. Obviously, this mechanism of closed loop M IM O
requires additional overhead of feedback and suffers from feedback
delay.
In some cases, the power imbalance is fixed and independent of
instantaneous channel realization, thus the mapping scheme can be
optimized and kept unchangeable. For example, in UL pre-coded
M IM O with rank 3, to reduce the PAPR at transmit side, CM P
codebook is proposed by many references [3] where transmit
power is imbalanced between the first and other two layers. Thus
the link quality between them is different. This motivates us to
design optimized stream to layer mapping scheme to incorporate
such discrimination.
For channel coding scheme adopted in LTE, coded block is
comprised of two parts of constellation, one is from information
bits with high priority and the other is from parity bits with
relatively low priority. Compared with mapping the coded block
symbol alternately, better performance can be achieved when
information symbols are mapped onto high quality layer and
parity symbol onto low quality layer.

High SNR layer

s0(I) s1(I) ... sK(I)

Precoding
Codeword to
s0(I) ... sM(I) s0(P) ... sN(I) layer mapping
... sM(I) s0(P) s1(P)

s2(P) S3(P) ... sN(P)

Information symbols Parity symbols Low SNR layers

(a) Optimized mapping from one codeword to three layers

High SNR layer

s0(I)0 s0(I)1 ... sQ(I)0 sQ(I)1

Precoding
s0(I) 0 ... sM(I) 0 s0(P) 0 ... sN(I) 0

Codeword to
layer mapping ... sM(I)0 ... ... sK(I)1
s0(I)1 ... sK(I)1 s0(P)1 ... sL(I)1
... ... ... sN(I)0 sL(I)1
Low SNR layer

(b) Optimized mapping from two codeword to three layers

Fig. 3. Optimized mapping rules from one or two codeword stream to 3 layers.
2.3 Optimized Mapping Scheme
Since transmission quality is different on layers and importance is
also different for information and parity symbols, we propose a
scheme with optimized stream to layer mapping rules as follows.
Here the link quality difference between layers is assumed to be
introduced by SNR imbalance on different layers of codebook.
In Fig. 3, the modified stream to layer mapping scheme in
consideration of the link quality difference between layers is
illustrated where si(I)j , si(P )j (i = 0, 1, …; j = 0, 1) denote the i-th
information and parity symbols of j-th codeword stream
respectively. Here, layer one possesses high transmission SNR
and layer 2 and 3 possess relatively low SNRs, therefore the
important symbols should be mapped on the first layer and parity
symbols on other two layers.
Note that the length of two code block may be different and so
does the ratio of information to parity symbols for two streams,
the essence of the modified mapping scheme is to alternately map
the information symbols of two code blocks onto the high SNR
layer as more as possible, and then map the remained symbols,
maybe including parity symbols, onto other layers.
Compared with adaptive mapping scheme proposed in [2], the
newly designed mapping method depends on the predetermined
power imbalance between layers, thus doesn’t need additional
feedback overhead.
3. SIMULATION RESULTS
In this section, the performance improvement of the modified
mapping scheme is evaluated via link level simulations. Layer
quality difference is assumed to be introduced by codebook with
imbalanced power allocation between layers, where power on the

827
 first layer is twice as on the second and three layers [3]. The and 64QAM 1/2. M oreover, the performance gain increases when
simulation parameters are listed in T ABLE I. M CS level tends low. For 64QAM 1/2, SNR gain is about 0.4 dB,
In the case of one codeword three imbalanced layers, most however, it becomes 0.9dB for M CS of QPSK 1/2.
information symbols are conveyed on high SNR layer (coding rate Including the performance gain over conventional method, the
is about 1/3) when optimized mapping scheme is applied, whereas proposed mapping scheme doesn’t require additional overhead for
only 1/3 information symbols are transmitted over high SNR layer feedback if the power imbalance between layers comes from
when conventional mapping scheme is adopted. The performance codebook structure.
improvement is obvious because additional 2/3 important
Table 1. S imulation parameter settings
information symbols are well protected.
Parameter Value
0
10
Bandwidth 10M Hz
Prop. QPSK 1/2
Conv. QPSK 1/2 Antenna configuration 4x4
Prop. 16QAM 1/2
Conv. 16QAM 1/2 Channel model SCM -E
Prop. 64QAM 1/2
Conv. 64QAM 1/2 UE speeds of interest 3km/h
Number of RB
1 RB
allocation
Turbo Coding,
BLER

-1
10
0 2 4 6 8 10 12 14 16
SNR (dB)
Fig. 4. Illustration of performance improvement in BLER
with proposed stream-to-layer mapping scheme.
Now, we focus on the case of two codeword streams with the
same modulation and coding format (M CF) and code length. In
conventional mapping scheme (see Fig. 2(b)), all the information
symbols of two codeword are conveyed over high and low SNR
layer, respectively. According to Fig. 3(b), 2/3 information
symbols of both codewords are conveyed over high SNR layer
when optimized mapping method is used. The proposed mapping
method degrades the BLER performance of the first codeword, but
such a degradation is negligible because 2/3 important symbols are
already well protected. On the other hand, the performance
improvement of the second codeword is exaggerated because
additional 2/3 important symbols are protected, which is very
important for decoding. Fig. 4 compares the BLER performance of
the second codeword with LTE Rel-8 mapping and the proposed
mapping schemes.
In Fig. 4, BLER performance obtained from conventional and the
proposed mapping rules (see Fig. 2(b) and Fig. 3(b)) are plotted
for comparison. It shows obvious performance improvement in
BLER for three cases of M CS levels, say QPSK 1/2, 16QAM 1/2
Channel coding
QPSK/16QAM /64QAM 1/2
Number of CWs and
2 CWs to 3 layers
layer numbers
M IM O detector M M SE detector
Channel estimation Ideal channel estimation
18 20
4. CONCLUSIONS
In this paper, an optimized stream-to-layer mapping rule for
multiple layer transmission to improve the link level performance
has been introduced. The system performance is improved by
differential mapping for information and parity bits. Another
important property of the proposed scheme is that the mapping
rule doesn’t require additional overhead for feedback.
5. REFERENCES
[1] 3GPP RP-090359, Work Item Description for "Enhanced DL
transmission for LT E”, CMCC, T SG-RAN Meeting #43,
Biarritz, France, 3 rd -6th March 2009.
[2] 3GPP R1-094318 “Issues on Layer Shifting for Dual-Layer
Beamforming”, IT RI, T SG-RAN Meeting #58bis, Miyazaki,
Japan, 12th – 16th October 2009.
[3] 3GPP R1-094255, “Rank3 Precoding Design for LT E-A UL-
MIMO”, Fujitsu, T SG-RAN Meeting #58bis, Miyazaki, Japan,
12th – 16th October 2009.
[4] A. Paulraj, R. Nabr, and D. Gore, Introduction to Space-Time
Wireless Communications, Cambridge University Press, NY,
2003.
[5] 3GPP T S 36.211. “3rd Generation Partnership Project:
T echnical Specification Group Radio Access Network; Evolved
Universal T errestrial Radio Access (E-UT RA); Physical
Channels and Modulation (Release 8)”.

828

View publication stats