`Lte ..
`
`----
`
`The UMTS Long Tern1 Evolution
`FROM THEORY TO PRACTICE
`
`Edited by: Stefania Sesia • Issam Toufik • Matthew Baker
`
`SECOND EDITION
`
`Including Release 10 for LTE-Advanced
`
`j
`
`;.i
`I •f
`. j
`
`I
`I
`
`IPR2022-00457
`Apple EX1008 Page 1
`
`
`
`LTE - The UMTS
`Long Term Evolution
`
`From Theory to Practice
`
`Second Edition
`
`Stefania Sesia
`ST-Ericsson, France
`
`Issam Toufik
`ETSI~ Fronce
`
`.Matthew Baker
`Alcatel-Lucent, UK
`
`@WILEY
`
`A John Wiley & Sons, Ltd., Publication
`
`IPR2022-00457
`Apple EX1008 Page 2
`
`
`
`This edition first published 201 I
`© 20 I I John Wiley & Sons Ltd.
`
`Registered office
`John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ,
`United Kingdom
`
`For details of our global editorial offices, for customer services and for information about how to apply
`for permission to reuse the copyright material in this book please see our website at www.wiley.com.
`
`The rights of the authors to be identified as the authors of this work have been asserted in accordance
`with the Copyright, Designs and Patents Act 1988.
`
`AH rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
`transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or
`otherwise, except as permitted by the UK Copyright, Designs and Patents Act 1988, without the prior
`permission of the publisher.
`
`Photograph on cover courtesy of Alcatel-Lucent. from the ngConnect LIB-equipped car.
`3GPP website reproduced by permission of© 3GPPT.M.
`
`\Vitey also publishes its books in a variety of electronic formats. Some content that appears in print
`may not be available in electronic books.
`
`Designations used by companies to distinguish their products are often claimed as trademarks. AH
`brand names and product names used in this book are trade names, service marks, trademarks or
`registered trademarks of their respective owners. The publisher is not associated with any product or
`vendor mentioned in this book. This publication is designed to provide accurate and authoritative
`information in regard to the subject matter covered. It is sold on the understanding that the publisher is
`not engaged in rendering professional services. If professional advice or other expert assistance is
`required, the services of a competent professional should be sought.
`
`Library of Congress Cataloging-in-Publication Data
`
`Sesia, Stefania.
`LTE--the UMTS long term evolution: from theory to practice/ Stefania Sesia, Issam Toufik,
`Matthew Baker. - 2nd ed.
`p.cm.
`Includes bibliographica1 references and index.
`ISBN 978-0-470-66025-6 (hardback)
`1. Universal Mobile Telecommunications System. 2. Long-Term Evolution (Telecommunications)
`I. Toufik, lssam. JI. Baker, Matthew (Matthew P.J.) Ill. Title.
`TK5103.4883.S47 20 I I
`621.3845' 6-dc22
`
`2010039466
`
`A catalogue record for this book is available from the British Library.
`
`Print ISBN: 9780470660256 (H/B)
`ePDF ISBN: 9780470978511
`oBook ISBN: 9780470978504
`epub ISBN: 9780470978641
`
`Printed in Great Britain by CPI Antony Rowe, Chippenham, Wiltshire.
`
`IPR2022-00457
`Apple EX1008 Page 3
`
`
`
`10
`
`Link Adaptation and Channel
`Coding
`
`Brian Classon, Ajit Nimbalker, Stefania Sesia and
`lssam Toufik
`
`10.1
`
`Introduction
`
`The principle of link adaptation is fundamental to the design of a radio interface which
`is efficient for packet-switched data traffic. Unlike the early versions of UMTS (Universal
`Mobile Telecommunication System), which used fast closed-loop power control to support
`circuit-switched services with a roughly constant data rate, link adaptation in HSPA (High
`Speed Packet Access) and LTE adjusts the transmitted information data rate (modulation
`scheme and channe1 coding rate) dynamically to match the prevailing radio channel capacity
`for each user. Link adaptation is therefore very closely related to the design of the channel
`coding scheme used for forward error correction.
`For the downlink data transmissions in LTE, the eNodeB typically selects the modulation
`scheme and code rate depending on a prediction of the downlink channel conditions.
`An important input to this selection process is the Channel Quality Indicator (CQI)
`feedback transmitted by the User Equipment (UE) in the uplink. CQI feedback is an
`indication of the data rate which can be supported by the channel, taking into account the
`Signal-to-Interference-plus-Noise Ratio (SINR) and the characteristics of the UE's receiver.
`Section I 0.2 explains the principles of link adaptation as applied in LTE; it also shows how
`the eNodeB can select different CQI feedback modes to trade off the improved downlink link
`adaptation enabled by CQI against the uplink overhead caused by the CQI itself.
`The LTE specifications are designed to provide the signalling necessary for interoperabil(cid:173)
`ity between the eNodeB and the UEs so that the eNodeB can optimize the link adaptation,
`
`LTE - The UMTS Long Term Evolution: From Theory to Practice, Second Edition.
`Stefania Sesia, Issam Toufik and hfatthew Baker.
`© 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons. Ltd.
`
`IPR2022-00457
`Apple EX1008 Page 4
`
`
`
`216
`
`LTE - THE UMTS LONG TERM EVOLUTION
`
`but the exact methods used by the eNodeB to exploit the information that is available are left
`to the manufacturer's choice of implementation.
`In general, in response to the CQI fee<lback the eNodeB can select between QPSK,
`l 6QAM and 64QAM 1 schemes and a wide range of code rates. As discussed further
`in Section 10.2.1, the optimal switching points between the different combinations of
`modulation order and code rate depend on a number of factors, including the required quality
`of service and cell throughput.
`The channel coding scheme for forward error correction, on which the code rate adaptation
`is based, was the subject of extensive study during the standardization of LTE. The chapter
`therefore continues with a review of the key theoretical aspects of the types of channel
`coding studied for LTE: convolutional codes, turbo codes with iterative decoding, and a
`brief introduction of Low-Density Parity Check (LDPC) codes. The theory of channel
`coding has seen intense activity in recent decades. especially since the discovery of turbo
`codes offering near-Shannon limit performance. and the development of iterative processing
`techniques in general. 3GPP was an early adopter of these advanced channel coding
`techniques, with the turbo code being standardized in the first version of the UMTS as
`early as 1999. Later releases of UMTS (in HSPA) added more advanced channel coding
`features with the introduction of link adaptation, including Hybrid Automatic Repeat reQuest
`(HARQ), a combination of ARQ and channel coding which provides more robustness
`against fading; these schemes include incremental redundancy, whereby the code rate is
`progressively reduced by transmitting additional parity information with each retransmisslon.
`However, the underlying turbo code from the first version of UMTS remained untouched.
`Meanwhile, the academic and research communities were generating new insights into code
`design, iterative decoding and the implementation of decoders. Section 10.3.2 explains how
`these developments impacted the design of the channel coding for LTE, and in particular
`the decision to enhance the turbo code from UMTS by means of a new contention-free
`interleaver, rather than to adopt a new LDPC code.
`For the LTE uplink transmissions, the link adaptation process is similar to that for the
`downlink, with the selection of modulation and coding schemes also being under the control
`of the eNodeB. An identical channel coding structure is used for the uplink, while the
`modulation scheme may be selected between QPSK, l6QAM and, for the highest category
`of UE, also 64QAM. The main difference from the downlink is that instead of basing the
`link adaptation on CQI feedback, the eNodeB can directly make its own estimate of the
`supportable uplink data rate by channel sounding, for example using the Sounding Reference
`Signals (SRSs) which are described separately in Section 15.6.
`A final important aspect of link adaptation is its use in conjunction with multi-user
`scheduling in time and frequency, which enables the radio transmission resources to be
`shared efficiently between users as the channel capacity to individual users varies. The
`CQI can therefore be used not only to adapt the modulation and coding rate to the channel
`conditions. but also for the optimization of time/frequency selective scheduling and for inter(cid:173)
`cell interference management as discussed in Chapter 12.
`
`1 Quadrature Phase Shift Keying (QPSK) and Quadrature Amplitude Modulation (QAM).
`
`IPR2022-00457
`Apple EX1008 Page 5
`
`
`
`LINK ADAPTATION AND CHAN1''EL CODING
`10.2 Link Adaptation and CQI Feedback
`In cellular communication systems, the quality of the signal received by a UE depends on
`the channel quality from the serving cell, the level of interference from other cells, and the
`noise level. To optimize system capacity and coverage for a given transmission power~ the
`transmitter should try to match the information data rate for each user to the variations in
`received signal quality (see, for example, [!, 2] and references therein). This is commonly
`referred to as link adaptation and is typically based on Adaptive Modulation and Coding
`(AMC).
`The degrees of freedom for the AMC consist of the modulation and coding schemes:
`
`217
`
`• Modulation Scheme. Low-order modulation (i.e. few data bits per modulated symbol,
`e.g. QPSK) is more robust and can tolerate higher levels of interference but provides
`a lower transmission bit rate. High-order modulation (i.e. more bits per modulated
`symbol, e.g. 64QAM) offers a higher bit rate but is more prone to errors due to its
`higher sensitivity to interference, noise and channel estimation effors; it is therefore
`useful only when the SINR is sufficiently high.
`
`• Code rate. For a given modulation, the code rate can be chosen depending on the radio
`link conditions: a lower code rate can be used in poor channel conditions and a higher
`code rate in the case of high SINR. The adaptation of the code rate is achieved by
`applying puncturing (to increase the code rate) or repetition (to reduce the code rate)
`to the output of a mother code, as explained in Section 10.3.2.4.
`
`A key issue in the design of the AMC scheme for LTE was whether all Resource Blocks
`(RBs) allocated to one user in a subframe should use the same Modulation and Coding
`Scheme (MCS) (see, for example, (3-6]) or whether the MCS should be frequency-dependent
`within each subframe. It was shown that, in general~ only a smaH throughput improvement
`arises from a frequency-dependent MCS compared to an RB-common MCS in the absence
`of transmission power control, and therefore the additional control signalling overhead
`associated with frequency-dependent MCS is not justified. Therefore in LTE the modulation
`and channei coding rates are constant over the allocated frequency resources for a given
`user, and time-domain channel-dependent scheduling and AMC are supported instead. In
`addition. when multiple transport blocks are transmitted to one user in a given subframe
`using multistream Multiple-Input Multiple-Output (MIMO) (as discussed in Chapter 11),
`each transport block can use an independent MCS.
`In LTE, the UE can be configured to report CQis to assist the eNodeB in selecting an
`appropriate MCS to use for the downlink transmissions. The CQI reports are derived from the
`downlink received signal quality, typically based on measurements of the downlink reference
`signals (see Section 8.2). It is important lo note that, like HSDPA, the reported CQI is not a
`direct indication of SINR. Instead, the UE reports the highest MCS that it can decode with a
`BLER (BLock Error Rate, computed on the transport blocks) probability not exceeding 10%.
`Thus the information received by the eNodeB takes into account the characteristics of the
`UE's receiver, and not just the prevailing radio channel quality. Hence a UE that is designed
`with advanced signal processing algorithms (for example, using interference cancellation
`techniques) can report a higher channel quality and, depending on the characteristics of the
`eNodeB's scheduler, can receive a higher data rate.
`
`IPR2022-00457
`Apple EX1008 Page 6
`
`
`
`218
`
`LTE - THE UMTS LONG TERM EVOLUTION
`
`A simple method by which a UE can choose an appropriate CQI value could be based on
`a set of BLER thresholds, as illustrated in Figure 10.1. The UE would report the CQI value
`corresponding to the MCS that ensures BLER,; 10-1 based on the measured received signal
`quality.
`
`<i>
`
`··y
`l
`
`t
`Iii
`l .. ).
`
`10-1
`
`0::
`w
`-'
`"'
`
`10-2
`
`... '
`
`.
`
`j
`
`f
`
`f
`
`r · - t
`
`0
`
`5
`
`. El.
`I
`·c
`
`-e - QPSK, r-113
`'I' - QPSK, r-1/2
`-
`- + - QPSK, r-213
`- B - QPSK, r=4/5
`-e- 16QAM, r=1/3
`_,,,_ 16QAM, r-112
`l
`--1- 16QAM, r=2/3
`. . -e- 16QAM, r-4/5
`I
`. -~- . ·-v- 64QAM, r=112
`·-0- 64QAM, r=1/3
`·-+- 64QAM, r=2/3
`·r- ' --ti]- 64QAM, r=4/5
`I
`... $
`
`15
`
`20
`
`25
`
`,T ······+··
`I
`. ·i- -.
`
`I
`I
`
`I
`I
`
`I
`
`. .
`I
`
`. r,
`
`I I
`
`10
`
`SNR
`
`Figure 10, I: Typical BLER versus Signal-to-Noise Ratio (SNR) for different MCSs. From
`left to right, the curves in this example correspond to QPSK, I 6QAM and 64Q~l\.1, rates
`1/3, 1/2, 2/3 and 4/5.
`
`The list of modulation schemes and code rates which can be signalled by means of a CQI
`value is shown in Table 10.1.
`AMC can exploit the UE feedback by assuming that the channel fading is sufficiently
`slow. This requires the channel coherence time to be at least as long as the time between
`the UE's measurement of the downlink reference signals and the subframe containing the
`correspondingly adapted downlink transmission on the Physical Downlink Shared CHannel
`(PDSCH). This time is typically 7-8 ms (equivalent to a UE speed of ~16 km/h at 1.9 GHz).
`However, a trade-off exists between the amount of CQI information reported by the
`UEs and the accuracy with which the AMC can match the prevailing conditions. Frequent
`reporting of the CQI in the time domain allows better matching to the channel and inter(cid:173)
`ference variations, while fine resolution in the frequency domain allows better exploitation
`of frequency-domain scheduling. However, both lead to increased feedback overhead in
`the uplink. Therefore, the eNodeB can configure both the time-domain update rate and the
`frequency-domain resolution of the CQI, as discussed in the following section.
`
`10.2.1 CQI Feedback in LTE
`The periodicity and frequency resolution to be used by a UE to report CQI are both controlled
`by the eNodeB. In the time domain, both periodic and aperiodic CQI repo1ting are supported.
`
`IPR2022-00457
`Apple EX1008 Page 7
`
`
`
`LINK ADAPTATION AND CHANNEL CODING
`
`219
`
`Table 10. l: CQI values. Reproduced by permission of© 3GPP.
`
`CQI index
`
`Modulation
`
`Approximate code rate
`
`Efficiency
`(information bits per symbol)
`
`0
`
`2
`3
`4
`5
`6
`7
`8
`9
`IO
`11
`12
`13
`14
`15
`
`'Out-of-range'
`QPSK
`QPSK
`QPSK
`QPSK
`QPSK
`QPSK
`16QAM
`!6QAM
`16QAM
`64QAM
`64QAM
`64QAM
`64QAM
`64QAM
`64QAM
`
`0.076
`0.12
`0.19
`0.3
`0.44
`0.59
`0.37
`0.48
`0.6
`0.45
`0.55
`0.65
`0.75
`0.85
`0.93
`
`0.1523
`0.2344
`0.3770
`0.6016
`0.8770
`1.1758
`1.4766
`1.9141
`2.4063
`2.7305
`3.3223
`3.9023
`4.5234
`5.1152
`5.5547
`
`The Physical Uplink Control CHannel (PUCCH, see Section 16.3.1) is used for periodic CQI
`reporting only while the Physical Uplink Shared CHannel (PUSCH, see Section 16.2) is
`used for aperiodic reporting of the CQI, whereby the eNodeB specifically instructs the UE to
`send an individual CQI report embedded into a resource which is scheduled for uplink data
`transmission.
`The frequency granularity of the CQI reporting is determined by defining a number of sub(cid:173)
`bands (N), each comprised of k contiguous Physical Resource Blocks (PRBs). 2 The value of
`k depends on the type of CQI report considered and is a function of the system bandwidth. In
`each case, the number of sub-bands spans the whole system bandwidth and is given by N =
`r N~5 I kl, where N~5 is the number of RBs across the system bandwidth. The CQI reporting
`modes can be Wideband CQI, eNodeB-configured sub-band feedback, or DE-selected sub(cid:173)
`band feedback. These are explained in detail in the following sections. In addition, in the
`case of multiple transmit antennas at the eNodeB, CQI value(s) may be reported for a second
`codeword.
`For some downlink transmission modes, additional feedback signalling consisting of
`Precoding Matrix Indicators (PM!) and Rank Indications (RI) is also transmitted by the UE,
`as explained in Section 11.2.2.4.
`
`10.2.1.1 Aperiodic CQI Reporting
`
`Aperiodic CQI reporting on the PUSCH is scheduled by the eNodeB by setting a CQI request
`bit in an uplink resource grant sent on the Physical Downlink Control CHannel (PDCCH).
`
`2Notc that the last sub-band may have less thank contiguous PRBs.
`
`IPR2022-00457
`Apple EX1008 Page 8
`
`
`
`220
`
`LTE - THE UMTS LONG TERM EVOLUTION
`
`The type of CQI report is configured by the eNodeB by RRC signalling. Table 10.2
`summarizes the relationship between the configured downlink transmission mode (see
`Section 9.2.2.1) and the possible CQI reporting type.
`
`Table 10.2: Aperiodic CQI feedback types on PUSCH for each POSCH transmission mode.
`
`PDSCH
`transmission mode
`(See Section 9.2.2. l)
`
`PDSCH transmission scheme
`assumed by UE for deriving CQI
`
`Wideband
`only
`
`eNodeB-configured UE-selectec
`sub-bands
`sub-bands
`
`CQI type
`
`Mode 1
`
`Mode2
`
`Mode 3
`
`Mode4
`
`Mode5
`
`Mode6
`
`Mode7
`
`Mode gb with
`PMI/RI configured
`
`.tv1ode gb without
`Pivll/RI configured
`
`Mode9"' with
`PMI/RI configured
`and > l CSI-RS port'
`
`Mode 9t· otherwise
`
`Single antenna port
`
`Transmit diversity
`
`Transmit diversity if Ri=l,
`otherwise large-delay CDDa
`
`Closed-loop spatial multiplexing
`
`Multi-user MIMO
`
`Closed-loop spatial multiplexing
`using a single transmission layer
`
`Single anlenna port
`if one PBCH antenna port,
`otherwise transmit diversity
`
`X
`
`X
`
`Closed-loop spatial multiplexing
`
`X
`
`Single antenna port
`if one PECH antenna port,
`otherwise transmit diversity
`
`Closed-loop spatial multiplexing
`
`X
`
`Single antenna port
`if one PBCH antenna port,
`otherwise transmit diversity
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`n Cyclic Delay Dlversity.
`
`b Introduced in Release 9.
`
`c Introduced in Release l 0.
`
`d See Section 29. i .2.
`
`The CQI reporting types are as follows:
`
`• Wideband feedback. The UE reports one wideband CQI value for the whole system
`bandwidth.
`
`IPR2022-00457
`Apple EX1008 Page 9
`
`
`
`CHANNEL CODING
`
`221
`
`• eNodeB-configured sub-band feedback. The UE reports a wideband CQI value for
`the whole system bandwidth. In addition, the UE repmts a CQI value for each sub(cid:173)
`band. calculated assuming transmission only in the relevant sub-band. Sub-band CQI
`reports are encoded differentially with respect to the wideband CQI using 2-bits as
`follows:
`
`Sub-band differential CQI offset= Sub-band CQI index - Wideband CQI index
`
`Possible sub-band differential CQI offsets are (:5 -1, 0, + l,;:,, +2}. The sub-band size k
`is a function of system bandwidth as summarized in Table I0.3.
`
`Table I0.3: Sub-band size (k) versus system bandwidth for eNodeB-configured aperiodic
`CQI reports. Reproduced by permission of© 3GPP.
`
`System bandwidth
`(RBs)
`
`Sub-band size
`(k RBs)
`
`6-7
`8-10
`11-26
`27-63
`64-110
`
`(Wideband CQI only)
`4
`4
`6
`8
`
`• UE-selected sub-band feedback. The UE selects a set of M preferred sub-bands of
`size k (where k and Mare given in Table I0.4 for each system bandwidth range) within
`the whole system bandwidth. The UE reports one wideband CQI value and one CQI
`value reflecting the average quality of the M selected sub-bands. The UE also reports
`the positions of the M selected sub-bands using a combinatorial index r defined as
`M-1
`
`r= L {:-_s:)
`
`j,:,,Q
`
`where the set lsi}f!o 1
`
`, l ~ S; ~ N, S; < Sj+1 contains the lvt sorted sub-band indices and
`
`r}={(;)
`
`0
`
`y
`
`ifx;:,,y
`
`if X <y
`
`is the extended binomial coefficient, resulting in a unique label r E (0, ... , (;)- I}.
`Some possible algorithms for deriving the combinatorial index r in the UE and
`extracting the infonnation from it in the eNodeB can be found in [7, 8].
`The CQI value for the M selected sub-bands for each codeword is encoded differen(cid:173)
`tially using two bits relative to its respective wideband CQI as defined by
`
`Differential CQI
`= Index for average of M preferred sub-bands - Wideband CQI index
`Possible differential CQI values are {5 + 1, +2, +3,;:,, +4}.
`
`IPR2022-00457
`Apple EX1008 Page 10
`
`
`
`222
`
`LTE - THE UMTS LONG TERM EVOLUTION
`
`Table 10.4: Sub-band size k and number of preferred sub-bands (M) versus downlink
`system bandwidth for aperiodic CQI reports for DE-selected sub-band feedback.
`Reproduced by permission of© 3GPP.
`
`System bandwidth
`(RBs)
`
`6-7
`8-lO
`11-26
`27-63
`64-110
`
`Sub-band size
`(k RBs)
`(Wideband CQI only)
`2
`2
`3
`4
`
`Number of preferred sub-bands
`(M)
`
`(Wideband CQI only)
`I
`3
`5
`6
`
`10.2.1.2 Periodic CQI Reporting
`
`If the eNodeB wishes to receive periodic reporting of the CQI, the UE will transmit the
`reports using the PUCCH.3
`Only wideband and OE-selected sub-band feedback is possible for periodic CQI reporting,
`for all downlink (PDSCH) transmission modes. As with aperiodic CQI reporting, the type of
`periodic reporting is configured by the eNodeB hy RRC signalling. For wideband periodic
`CQI reporting, the period can be configured to4
`):
`
`{2, 5, 10, 16, 20, 32, 40, 64, 80, 128, 160} ms, or Off.
`
`While the wideband feedback mode is similar to that sent via the PUSCH, the 'UE(cid:173)
`selected sub-band' CQI using PUCCH is different. In this case, the total number of sub-bands
`N is divided into J fractions called bandwidth parts. The value of J depends on the system
`bandwidth as summarized in Table 10.5. In this case, one CQI value is computed and reported
`for a single selected sub-band from each bandwidth part. along with the corresponding sub(cid:173)
`band index..
`
`Table 10.5: Periodic CQI reporting with lJE-selected sub-bands:
`sub-band size (k) and bandwidth parts (J) versus downlink system bandwidth.
`Reproduced by permission of© 3GPP.
`
`System bandwidth
`(RBs)
`
`Sub-band size
`(k RBs)
`
`Number of bandwidth parts
`(J)
`
`6--7
`8-10
`11-26
`27-63
`64-110
`
`(Wideband CQI only)
`4
`4
`6
`8
`
`I
`
`2
`3
`4
`
`3 Lf PUSCH transmission resources are scheduled for the UE ln one of the periodic sub.frames. the periodic CQI
`report is sent on the PUSCH instead, as explained in Section 16.4 and Figure 16.15.
`4These values apply to FDD operation; for TDD, see [9, Section 7.2.2]
`
`IPR2022-00457
`Apple EX1008 Page 11
`
`
`
`LINK ADAPTATION AND CHANNEL CODING
`
`223
`
`10.2.1.3 CQI Reporting for Spatial Multiplexing
`
`If the UE is configured in POSCH transmission modes 3, 4, 8 or 9,5 the eNodeB may use
`spatial multiplexing to transmit two codewords simultaneously to the UE with independently
`selected MCSs. To support this, the following behaviour is defined for the UE's CQI reports
`in these modes:
`
`• If the UE is not configured to send RI feedback, or if the reported RI is equal to
`I, or in any case in transmission mode 3,6 the UE feeds back only one CQI report
`corresponding to a single codeword.
`• If RI feedback is configured and the reported RI is greater than 1 in transmission modes
`4or 8:
`
`For aperiodic CQI reporting, each CQI report (whether wideband or sub-band)
`comprises two independent CQI reports for the two codewords.
`For periodic CQI reporting, one CQI report is fed back for the first codeword,
`and a second three-bit differential CQI report is fed back for the second codeword
`(for both wideband and sub-band reporting). The differential CQI report for the
`second codeword can take the following values relative to the CQI report for the
`first codeword: :<;-4, -3, -2, -1, 0, +!, +2,;:: +3.
`
`10.3
`
`Channel Coding
`
`Channel coding, and in particular the channel decoder, has retained its reputation for being
`the dominant source of complexity in the implementation of wireless communications~ in
`spite of the relatively recent prevalence of advanced antenna techniques with their associated
`complexity.
`Section 10.3.1 introduces the theory behind the families of channel codes of relevance
`to LTE. This is followed in Sections 10.3.2 and 10.3.3 by an explanation of the practical
`design and implementation of the channel codes used in LTE for data and control signalling
`respectively.
`
`10.3.1 Theoretical Aspects of Channel Coding
`This section first explains convolutional codes, as not only do they remain relevant for small
`data blocks but also an understanding of them is a prerequisite for understanding turbo codes.
`The turbo-coding principle and the Soft-Input Soft-Output (SISO) decoding algmithms are
`then discussed. The section concludes with a brief introduction to LDPC codes.
`
`10.3.1.1 From Convolutional Codes to Turbo Codes
`
`A convolutional encoder C(k, n, m) is composed of a shift register with m stages. At each
`time instant, k infonnation bits enter the shift register and k bits in the ]ast position of
`
`5Mode 8 from Release 9 onwards~ mode 9 from Release 10 onwards.
`6If RI feedback is configured in transmission mode 3 and the RI fed back is greater than l, although only one
`CQI corresponding to the first codeword is fed back, its value is adapted on the assumption that a second codeword
`will also be transmitted.
`
`IPR2022-00457
`Apple EX1008 Page 12
`
`
`
`224
`
`LTE - THE UMTS LONG TERM EVOLUTION
`
`the shift register are dropped. The set of n output bits is a linear combination of the
`content of the shift register. The rate of the code is defined as R,. = k/n. Figure 10.2 shows
`the convolutional encoder used in LTE [10] with m = 6, n = 3, k = I and rate Re= 1/3.
`The linear combinations are defined via n generator sequences G =[go •... , gn-il where
`g, = [ge.o, g,,1, ... , g,,.]. The generator sequences used in Figure 10.2 arc
`
`go= [101 101 1] = [133](oct),
`
`g 1 =[II I 100 I]= [17l](oct),
`
`g2 = [I I IO 1 0 I]= [165](oct).
`
`dfi G2 = 165 (octal)
`')-----+<cB--------->ffi------>ft+.')-(cid:157)
`
`Figure 10.2: Rate 1/3 convolutional encoder used in LTE with m = 6, n = 3, k = l [10].
`Reproduced hy permission of© 3GPP.
`
`A convolutional encoder can be described by a trellis diagram [ 11 ], which is a representa(cid:173)
`tion of a finite state machine including the time dimension.
`Consider an input block with L bits encoded with a rate l/11 (i.e. k = I) convolutional
`encoder, resulting in a codeword of length (L + 111) x n bits, including m trellis termination
`bits (or tail bits) which are inserted at the end of the information block to drive the shift
`register contents back to all zeros at the end of the encoding process. Note that using tail bits
`is just one possible way of te1minating an input sequence. Other trellis termination methods
`include simple truncation (i.e. no tail bits appended) and so-caUed tail-biting [12]. In the
`tail-biting approach, the initial and final states of the convolutional encoder are required
`to be identical. Usually, tail-biting for feed-forward convolutional encoders is achieved by
`initializing the shift register contents with the last m information bits in the input block. Tail(cid:173)
`biting encoding facilitates uniform protection of the information bits and suffers no rate-loss
`owing to the tail bits. Tail-biting convolutional codes can be decoded using, for example, the
`Circular Viterbi Algorithm (CVA) [13, 14].
`Let the received sequence y be expressed as
`
`y= ~x+n
`
`(JO.I)
`
`where n = [110, 111, .•• , n,, ... , "<L+mJx(n-IJl and ne ~ N(O, No) is the Additive White Gaus(cid:173)
`sian Noise (AWGN) and Eb is the energy per bit. The transmitted codeword is x =
`[xo, xi •... , Xe, . ..• XL+m-1] where Xe is the convolutional code output sequence at time
`
`IPR2022-00457
`Apple EX1008 Page 13
`
`
`
`225
`LINK ADAPTATION AND CHANNEL CODING
`instant t for the input information bit ie, given by Xt = [xc.o, ... , xe,n-1J- Equivalently,
`Y = [Yo, Y1, ... , y,, ... , YL+m-il where Yr= [yf.o, ... , YE,n-d is the noisy received version
`of x. (L + m) is the total trellis length.
`
`10.3.1.2 Soft-Input Soft-Output (SISO) Decoders
`
`In order to define the performance of a communication system~ the codeword error probability
`or bit e1TOr probabHity can be considered. The minimization of the bit error probability is in
`general more complicated and requires the maximization of the a posteriori bit probability
`(MAP symbol-by-symbol). The minimization of the codeword/sequence error probability is
`in general easier and is equivalent to the maximization of A Posteriori Probability (APP) for
`each codeword; this is expresse<l by the MAP sequence detection rule, whereby the estimate
`X of the transmitted codeword is given by
`x = argmax P(x I y)
`
`(10.2)
`
`X
`
`When all codewords are equiprobable, the MAP criterion is equivalent to the Maximum
`Likelihood (ML) criterion which selects the codeword that maximizes the probability of the
`received sequence y conditioned on the estimated transmitted sequence x. i.e.
`
`x = argmax P(y Ix)
`'
`Maximizing Equation ( 10.3) is equivalent to maximizing the logarithm of P(y Ix), as log(·) is
`a monotonically increasing function. This leads to slmplified processing.7 The log-likelihood
`function for a memoryless channel can be written as
`
`(10.3)
`
`log P(y Ix)= I; I; log P(y;,j I x;,j)
`
`L+m-i n-1
`
`(I 0.4)
`
`_FO
`For an AWGN channel, the conditional probability in Equation (10.4) is P(y;,j I x;,j) ~
`N( -../E,,x;,j, No), hence
`
`i=O
`
`log P(y Ix) ex: IIY, -
`
`(10.5)
`
`-,,/E,,x,il 2
`The maximization of the metric in Equation (10.5) yields a codeword that is closest to
`the received sequence in terms of the Euclidean distance [15]. This maximization can be
`performed in an efficient manner by operating on the trellis.
`As an example, Figure 10.3 shows a simple convolutiona1 code with generator polynomi(cid:173)
`als g0 =[IO I) and g 1 = [l l 1) and Figure 10.4 represents the corresponding trellis diagram.
`Each edge in the trellis corresponds to a transition from a state s to a state s', which can be
`obtained for a particular input information bit. In Figure I 0.4, the edges are parametrized with
`the notation if/ x,.o x,. 1 , i.e. the input/output of the convolutional encoder. The shift registers
`of the convolutional code are initialized to the all-zero state.
`In Figure 10.4, M(y; Ix;)= :E1:J log P(y;,; I x;,1) denotes the branch metric at the ith trellis
`
`step (i.e. the cost of choosing a branch at trellis step i), given by Equation (10.5). The Viterbi
`Algorithm (VA) selects the best path through the trellis by computing at each step the best
`
`7The processing is simplified because the multiplication operation can be transformed to the simpler addition
`operation in the logarithmic domain.
`
`IPR2022-00457
`Apple EX1008 Page 14
`
`
`
`226
`
`LTE - THE UMTS LONG TERM EVOLUTION
`
`g0=[!0l], 5(oct)
`
`ie
`
`Xf,0
`
`xe,1
`
`g 1=[ll l], 7(oct)
`Figure 10.3: Rate½ convolutional encoder with m = 2, m = 2, k = 1, corresponding to
`generator polynomials g 0 = [ 1 0 1] and g 1 = [ 1 1 1 ].
`
`0/00
`001k---'=-,,-----,-,+---7,,
`OIL!
`•
`
`Ol
`
`· .
`ll
`I/00,,,
`-· ·_/
`10 0101-:·~
`•·. All!O
`)/IO
`..
`l l .. ····l/ll·-··.
`
`Ml I 1(y;x}j
`£=]
`
`£=L+m-1
`
`Time
`
`Figure 10.4: Trellis corresponding to convolutional code with generator polynomials
`go=[10l]andg1 =[111).
`
`partial metric (the accumulated cost of following a specific path until the f'h transition) and
`selecting at the end the best total metric [16, 17]. It then traces back the selected path in the
`trellis to provide the estimated input sequence.
`Although the original VA outputs a hard-decision estimate of the input sequence, the VA
`can be modified to output soft information8 along with the hard-decision estimate of the
`input sequence [IS]. The reliability of coded or information bits can be obtained via the
`computation of the a posteriori probability.
`Let it be the information bit which enters the shift register of the convolutional code at
`time e. Assuming BPSK (Binary Phase Shift Keying) modulation (0-> + I, and I -> -!), the
`Log Likelihood Ratio (LLR) of an information symbol ( or bit) ie is
`
`8 A soft decision gives additional information about the reliability of the decision [15].
`
`IPR2022-00457
`Apple EX1008 Page 15
`
`
`
`LINK ADAPTATION AND CHA