`"'
`
`0
`
`t:D ., �
`�
`(") �
`
`Q..
`
`(I)
`Q..
`El
`
`�
`�
`i I
`
`ACK
`
`@
`
`I
`1 +•r•I
`
`�
`
`QC
`
`8
`
`(a) LTEFDD; (b) LTE TDD
`
`Figure 12.10 HARQTiming:
`
`DL subframe)
`?? □
`
`as ordinary
`
`is treated
`
`subframe
`
`1 ·L���-
`
`1 +•r•� ,:
`
`(b) Conceptual example
`
`18
`
`I· 1
`
`8
`
`Data
`
`of TDD HARQ Timing (special
`
`I
`I
`1 -���
`I I I· 1
`I I· 1
`
`of FOO HARO Timing (propagation
`
`example
`
`(a) Conceptual
`
`I
`
`8
`
`I • I Jms I • L���
`o,mI
`
`I I
`
`delay and timing advance is ignored)
`
`I • I
`Jms I • I •c•I
`I
`
`I
`
`IPR2022-00468
`Apple EX1018 Page 441
`
`
`
`LTETDDMode
`
`379
`
`1 ms
`
`UL UL UL
`
`UL
`
`UL UL [Qi!
`
`-�--oca• ••�
`
`UL scheduling window
`
`--- -- -OCO•O--:il
`
`UL scheduling window
`
`
`
`
`
`
`
`Figure 12.11 Possible uplink multi-TT! scheduling
`
`
`
`12.3.4 HARQ Design for UL TT/ Bundling
`UL m bundling is used to improve the UL coverage by combining multiple UL Tris for one
`
`
`
`
`
`
`
`
`HARQ process transmission. In a system where it is expected that many retransmissions are
`
`
`
`
`
`
`needed to successfully transmit a packet in uplink, UL Tri bundling provides fast automatic
`
`
`
`
`
`
`retransmission so that the delay penalty is minor. Although many details of UL m bundling are
`
`
`
`the same for LTE FDD and TDD, specifically the HARQ timing is different due to the inherited
`
`
`
`special UL HARQ timing in LTE TDD. In terms of the 3GPP discussion, only TDD UUDL
`
`
`
`
`
`configurations #0, #1 and #6 have bundling fully defined in Release 8. To control the aspect of
`
`
`
`
`
`
`Tri bundling timing and HARQ process number after m bundling, the starting point of UL
`
`
`
`
`
`Tri bundling is limited and the number of bundled HARQ processes is fixed according to the
`
`
`
`bundle size. Thus the number of bundled HARQ processes is defined by:
`B dl d HARQ N, = [Original HARQ No x Bundled HARQ RITl
`
`un e -
`
`-0 Bundled_size x Original_HARQ_RIT
`(12. l)
`-J
`
`
`
`In Equation 1, Original_HARQ_No and Original_HARQ_RIT are fixed for each TDD
`
`
`
`
`
`
`configuration in the LTE uplink system, so if assuming for instance TDD configuration #1 and
`
`
`
`
`a bundle size of 4, the number of bundled HARQ processes is 2, as shown in Figure 12. 12.
`
`
`
`As to the principle of ACK/NACK timing, it is always tied to the last subframe in a bundle,
`
`
`which is exactly the same as the rule in FDD. The uplink m bundling for LTE FDD is described
`
`in Chapter IO.
`
`0
`PUSCH
`I
`No Bundling
`
`0
`PUSCH I
`With Bundling
`
`0
`
`z•dTX
`
`2 3
`
`0 0
`
`0
`
`..._____.__
`
`___, I · ..
`
`.______._____,
`
`I···
`
`TTis
`
`One HARQ pt·ocess to cover multiple
`
`
`
`
`
`
`#1 Figure 12.12 TTI bundling with TDD configuration
`
`IPR2022-00468
`Apple EX1018 Page 442
`
`
`
`380
`
`LTE for UMTS – OFDMA and SC-FDMA Based Radio Access
`
`12.3.5 UL HARQ-ACK/NACK Transmission
`In the same way as for FDD, the UL HARQ-ACK/NACK for LTE TDD is transmitted over the
`PHICH on PDCCH. The PHICH mapping and indexing is mostly the same for LTE FDD and
`TDD, i.e. the PDCCH in one DL subframe only contains the PHICH associated to a single UL
`subframe PUSCH. The only exception to this rule is the TDD UL/DL confi guration #0 where
`the PHICH associated has specifi c exception specifi ed.
`
`12.3.6 DL HARQ-ACK/NACK Transmission
`For both LTE FDD and TDD, the DL HARQ-ACK/NACK is transmitted on the PUCCH or
`the PUSCH depending on whether UL has simultaneous data transmission in the same UL
`subframe or not. In many cases the DL HARQ-ACK/NACK associated from more than one
`PDSCH, e.g. up to 9, will be mapped into a single UL subframe. This so-called multiple UL
`ACK/NACK transmission is, however, notably different from the FDD MIMO case in which
`the DL HARQ-ACK/NACK associated from a single PDSCH (e.g. two codewords) is mapped
`into a single UL subframe.
`The fact that multiple downlink transmissions may need to be acknowledged within a single
`uplink subframe, makes the design for good UL coverage for control channels in TDD even
`more challenging. A very special design arrangement has been created to accomplish this task
`while simultaneously respecting the single carrier property of the UL multiple access scheme
`when UE has to transmit multiple DL HARQ-ACK/NACKs. There are two DL HARQ ACK/
`NACK feedback modes supported in TDD operation of LTE which are confi gured by a higher
`layer on a per-UE basis:
`• ACK/NACK bundling feedback mode (the default mode), where a logical AND operation
`is performed per codeword’s HARQ ACK/NACK across multiple DL subframes PDSCH
`whose associated HARQ ACK/NACK is mapped into the same UL subframe.
`• ACK/NACK multiplexing feedback mode, where a logical AND operation is performed
`across spatial codewords within a DL HARQ ACK/NACK process. In Release 8 LTE TDD
`up to 4 bits DL HARQ ACK/NACK is supported per UL subframe, hence for UL/DL con-
`fi guration #5 this feedback mode is not supported.
`
`The ACK/NACK bundling feedback mode is the most aggressive mode to relieve the cover-
`age problem of multiple UL ACK/NACK transmission in TDD. The allocated DL resources
`have been decoupled from the required UL feedback channel capability; i.e. only a single
`DL HARQ-ACK/NACK is transmitted in a single UL subframe regardless of the number of
`associated DL subframes carrying PDSCH for the user. The single ACK/NACK is then cre-
`ated by performing a logical AND operation over all associated HARQ ACK/NACK per UL
`subframe. This way, TDD has the same number of HARQ ACK/NACK feedback bits and thus
`transmission formats on PUCCH as FDD per UL subframe. The ACK/NACK encoding and
`transmission format in PUSCH is the same as it is in FDD.
`Without proper compensation in the link adaptation and packet scheduling functions, the
`probability of DL HARQ NACK will increase causing more unnecessary DL retransmissions
`when using ACK/NACK bundling. Thus the control channel reliability is the key with ACK/
`NACK bundling. The second mode is more attractive for when the UE has suffi cient UL cov-
`
`IPR2022-00468
`Apple EX1018 Page 443
`
`
`
`LTE TDD Mode
`
`381
`
`erage to support multiple ACK/NACK bits on PUCCH. When in ACK/NACK multiplexing
`feedback mode, the status of each DL subframe HARQ-ACK/NACK, i.e. ACK, NACK or DTX
`(no data received in the DL subframe), one of the QPSK constellation points in certain derived
`PUCCH channels is selected for transmission on the UE side, and the eNodeB can decode the
`multi-bit HARQ ACK/NACK feedback by monitoring all constellation points from all associ-
`ated PUCCH channels. The exact mapping table can be found in [3].
`
`12.3.7 DL HARQ-ACK/NACK Transmission with SRI and/or CQI over PUCCH
`When in ACK/NACK bundling mode, if both bundled HARQ-ACK/NACK and SRI are to be
`transmitted in the same UL subframe, the UE transmits the bundled ACK/NACK on its derived/
`assigned PUCCH ACK/NACK resource for a negative SRI transmission, or transmits the bundled
`HARQ-ACK/NACK on its assigned SRI PUCCH resource for a positive SRI transmission. This
`operation is exactly the same as for FDD.
`When in ACK/NACK multiplexing mode, if both multiple HARQ-ACK/NACK and SRI are
`transmitted in the same UL subframe, the UE transmits the multiple ACK/NACK bits accord-
`ing to section 12.3.5 for a negative SRI transmission, and transmits 2-bit information mapped
`from multiple ACK/NACK input bits on its assigned SRI PUCCH resource for a positive SR
`transmission using PUCCH format 1b. The mapping between multiple ACK/NACK input bits
`and 2-bit information depends on the number of generated HARQ-ACK among the received
`DL subframe PDSCH within the associated DL subframes set K. The exact mapping table can
`be found in [3].
`
`12.4 Semi-persistent Scheduling
`Semi-persistent Scheduling (SPS) can be used with all the TDD UL/DL confi gurations. Many
`details of SPS are the same for LTE FDD and TDD, but in this section some TDD specifi c
`aspects related to SPS are detailed. The reader is referred to Chapter 10 for more informa-
`tion about SPS. To match the special frame structure of TDD, the SPS resource interval must
`be set to equal a multiple of the UL/DL allocation period (i.e. 10 ms) to avoid the confl ict of
`non-matching UL/DL subframes because the UL subframe and DL subframe do not exist
`simultaneously. Furthermore, LTE UL uses synchronous HARQ and there are some problems
`for most UL/DL confi gurations in TDD because the HARQ RTT is 10 ms. When UL SPS is
`used for VoIP traffi c (AMR codec periodicity is 20 ms), the second retransmission of a previ-
`ous packet will collide with the next SPS allocation, since the period of SPS resource is two
`times the RTT. The collision case is shown in Figure 12.13. In the fi gure the numbers 1, 2 and
`3 indicate different VoIP packets. For example, at the 20 ms point, the second retransmission
`of VoIP packet #1 collides with the initial VoIP packet #2. To solve this problem, two solutions
`are available to the network: dynamic scheduling and two-interval SPS patterns.
`Although confi gured for SPS, the UE will anyway listen to dynamic allocations on the
`PDCCH. Such allocations will always override an existing persistent allocation. By using
`dynamic scheduling at the collision point it is possible to mitigate the problem as shown in
`Figure 12.14: if the UE is asked for a retransmission, the UE will perform a retransmission
`unless it has an empty buffer. With these defi nitions, if there are other following idle subframes
`available, the eNodeB will next schedule the retransmission in the current subframe, and
`
`IPR2022-00468
`Apple EX1018 Page 444
`
`
`
`382
`
`LTE for UMTS -OFDMA and SC-FD MA Based Radio Access
`
`L
`Persistent resource
`I interval=20ms
`
`�
`
`�
`
`r-R��s+R��ns
`
`
`
`0 Persistent resource O Retransmissions
`
`t
`
`
`
`
`
`
`
`and a new transmission Figure 12.13 Collision between retransmissions
`
`Collision & Dynamic
`
`Scheduling
`�--------
`P1 P2"-,
`/
`\, D I) □P3
`---20 ms
`40 ms
`20 ms
`Oms
`transmission I
`D Delayed Initial
`
`Initial transmission 0 Retransmission
`
`,,.
`
`'-)-<. ______ , ,_,, 20 ms----
`
`
`
`reschedule the initial transmission that was supposed to take place on the SPS resources in
`
`
`
`
`
`
`
`following subframes at the collision point.
`
`
`
`
`The second solution is to use a two-interval SPS pattern. Here a two-interval SPS pattern
`
`
`
`
`means that two periods are used for semi-persistent scheduling while only one semi-persistent
`
`
`scheduling period is used and persistent allocation is carried out based on the pre-defined period
`
`
`
`
`
`
`
`
`in the conventional scheme. With a two-interval SPS pattern, a resource pattern with two dif
`
`
`
`
`
`ferent intervals (i.e. Tl, T2, Tl, T2 ... ) is used to avoid the main collision between the second
`
`
`
`
`
`retransmission of the previous packet and the SPS allocation. The procedure is given in Figure
`
`
`12.15, in which Tl is not equal to T2 and the combined set of Tl and T2 is always a multiple
`
`
`of lOms. The offset between Tl and T2 is several subframes, and a variable subframe offset is
`
`
`
`
`
`
`
`used to indicate the offset. The following formulas are used to calculate Tl and T2:
`
`Tl =SPS periodicity+subframe_offset(12.2)
`(12.3)
`T2= SPS periodicity
`
`
`
`
`
`Figure 12.14 Dynamic scheduling at collision point
`
`-subframe_offset
`
`IPR2022-00468
`Apple EX1018 Page 445
`
`
`
`LTETDDMode
`
`383
`
`t+--------40
`
`ms-------.i 0SPS resource ORetransmissions
`
`�
`
`
`
`Figure 12.15 Two-interval SPS patterns
`
`'SPS periodicity' will be signaled by RRC signaling;
`
`
`'subframe_offset' (positive value in
`
`
`
`
`
`
`
`Figure 12.15, but can also be negative) is implicitly defined according to the different TDD
`
`
`
`
`
`configurations and the starting point of a two-interval SPS pattern), and then Tl (the first time
`
`
`
`
`periodicity) and T2 (the second time periodicity) can be computed in terms of these equations.
`
`
`
`
`
`The allocation period always starts with the first time period T l . However, even configured with
`
`
`
`
`a two-interval SPS pattern, some residual collisions might still exist if the number of required
`
`
`
`
`
`retransmissions is large, i.e. 4. Any residual collisions can be avoided by means of dynamic
`
`
`scheduling as described earlier.
`
`
`
`12.5 MIMO and Dedicated Reference Signals
`
`LTE supports a number of different MIMO modes in DL, as already described in Chapter 5,
`
`
`
`
`
`
`
`
`covering both closed loop schemes with UE feedback information to the Node B. Together
`
`
`
`
`
`with the information of actually adapted downlink parameters from eNodeB this adds up to
`
`
`a significant amount of signaling to handle the DL closed loop MIMO. For TDD mode the
`
`
`
`
`
`earlier mentioned channel reciprocity can be exploited to mimic closed loop MIMO with a
`
`
`
`
`
`
`reduced signaling overhead. Under the assumption that the channel is identical for UL and DL
`
`
`
`we can estimate the channel in UL by using SRSs transmitted by the UE and then apply this
`
`
`
`
`
`channel lrnowledge for selecting the bestDL pre-coding matrix. In this way UE feedback can
`
`be reduced or even avoided.
`Going a step further, we can also eliminate the pre-coding matrix indication in the DL
`
`
`
`
`
`
`
`allocation message by using UE Spec ific Reference Signals (URS) to transfer this informa
`
`tion. Moreover, the use of URS decouples the physical transmit antenna from the UE detection
`
`
`
`
`
`
`complexity and system overhead resulting from having a cell spec ific reference signal for each
`
`
`
`transmit antenna. URS are spec ified for LTE Release 8, and can be used for both FDD and
`
`
`
`
`TDD modes. However, this transmission mode is especially attractive when considered in a
`
`
`
`TDD setting where channel reciprocity is available.
`URS are transmitted on antenna port 5 and they are generated with the same procedure as
`
`
`
`
`
`
`
`
`
`cell specific reference signals. The only difference is that the UE RNTI impacts the seed of the
`
`
`
`
`pseudo-random generator used to generate the code. The pattern for normal cyclic prefix and
`
`
`
`
`
`
`extended cyclic prefix can be seen in Figure 12.16(a) and Figure 12.16(b), respectively. There
`
`used for URS per PRB per l ms subframe so the additional overhead
`
`are 12 resource elements
`
`
`
`
`
`
`
`is rather large. With a normal cyclic prefix, cell specific reference signals on antenna port 0
`
`IPR2022-00468
`Apple EX1018 Page 446
`
`
`
`384
`
`(a)
`
`LTE for UMTS -OFDMA and SC-FD MA Based Radio Access
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`l=O
`
`1=6 l=O i
`
`l = 6
`
`even-numbered slots odd-numbered slots
`
`
`
`◄
`
`►◄
`
`►
`
`
`
`Antenna port 5
`
`(b)
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`
`Rs
`Rs
`l=(J l=5l=P l=5
`
`
`even-numbered slots odd-numbered slots
`◄
`►◄
`►
`
`
`
`Antenna port 5
`
`
`Figure 12.16 (a) URS pattern
`
`
`
`CP for normal CP; (b) URS pattern for extended
`
`IPR2022-00468
`Apple EX1018 Page 447
`
`
`
`LTE TDD Mode
`
`385
`
`and 1 and a control channel region of three symbols antenna, enabling URS will reduce the
`number of resources for data transmission by 10%. On the other hand this gives a very robust
`performance at the cell edge or for high UE velocity.
`One advantage of using URS is that pre-coding does not need to be quantifi ed. As data and
`reference signals are using the same pre-coding matrix, the combined channel can be directly
`estimated from the UE specifi c reference signals and then used for demodulating the signal.
`Due to the rather large overhead of URS, this mode is mainly expected to be used when
`Node B deploys more than four antennas. The Release 8 specifi cations do not allow for more
`than four different common reference signal ports, so in this case the only way to acheive UE
`specifi c pre-coding (beamforming) is to use URS.
`An issue which is vendor specifi c and where eNodeB implementations may differ is how
`antennas are deployed and pre-coding is determined when URS are used. Two different scenarios
`will be discussed here. In the typical macro cell with antennas mounted above the roof-top, the
`azimuth spread of the channel would be low and the optimal solution would be to have antennas
`with narrow spacing and use traditional angular beamforming to create a narrow beam which
`directs the power towards the UE while reducing the interference to other UEs. In this case,
`the system only relies on estimating the Direction of Arrival (DoA) from the UL transmission
`and this can be done without relying on channel reciprocity. Moreover, the standard allows
`for reuse of time and frequency resources by other users in the same sector. Suffi cient angular
`separation should be ensured to maintain performance.
`In a scenario with many scatterers around the base station, the azimuth spread would be larger
`and angular beamforming might not work very well. Another solution could then be to rely on
`pre-coding determined from the eigen-vectors of the complex channel matrix determined from
`UL sounding. When the azimuth spread is increased, the rank of the channel could become
`larger than one and UE could potentially benefi t from dual stream transmission. Although dual
`stream transmission is not supported in Release 8, it is a potential candidate for future releases.
`For more details on MIMO, beamforming and channel modeling see [4].
`
`12.6 LTE TDD Performance
`In this section the performance for LTE TDD is analyzed. For FDD mode, extensive analysis
`of performance has already been addressed in Chapter 9. As many of the observations provided
`there apply equally well to TDD mode, in the following we try to focus on the areas where
`TDD mode is different from FDD mode. First we look at the link performance, i.e. how well
`the receivers in eNodeB and UE can decode the physical channels. In general this is an area
`where there is little difference between FDD and TDD as reference signal patterns and channel
`coding are duplexing mode agnostic. After link performance, we discuss the link budget. Link
`budget for TDD is different from FDD because of the discontinuous transmission, so coverage
`for different bit rates in TDD will invariably be worse than for FDD. However, there are still a
`number of details to pay attention to in order to evaluate the TDD link budget properly.
`This section ends with a discussion on system performance. First we look at the best effort
`type of service where a fairly small number of users per sector are assumed to download large
`amounts of data. Assuming that we are designing a network to deliver a certain bit rate for a
`certain number of active users, then the most important difference between networks based on
`FDD and TDD technologies is that UL and DL in TDD mode would need double the system
`bandwidth compared to FDD. Data transmission would also need to be carried out with a larger
`
`IPR2022-00468
`Apple EX1018 Page 448
`
`
`
`386
`
`LTE for UMTS – OFDMA and SC-FDMA Based Radio Access
`
`bandwidth in the TDD system to achieve bit rates similar to FDD. Both system and transmis-
`sion bandwidths affect the operation of RRM algorithms and in this the impact to the system
`performance is analyzed.
`Performance assuming VoIP service is also evaluated and this gives a somewhat different
`perspective on the system performance as bit rates are low and users are many when we load
`the system with 100% VoIP traffi c. As VoIP is a symmetric service, TDD systems can have a
`higher VoIP capacity than FDD because the split between UL and DL resources can be adjusted.
`The increased HARQ round trip time for the TDD system is also shown to have some effect
`on the VoIP performance where many UEs are coverage limited.
`
`12.6.1 Link Performance
`The two most important factors impacting the link performance for data transmission are chan-
`nel estimation and channel coding. As reference signal design and channel coding are very
`similar for TDD and FDD, the link performance is also very similar. One source of difference
`is the discontinuous transmission in TDD. FDD receivers can use the reference signals from
`the previous subframe to estimate the channel. This is especially important for the DL link
`performance where the UE receiver should start decoding the control channel region as soon
`as it is received. In a TDD system, when a UE receiver decodes subframes transmitted right
`after UL→DL switching, it could not rely on reference signals from previous subframes and
`this could introduce some degradation to the channel estimation and thus coverage of control
`channels. The importance of such potential loss will depend on the UE implementation. If, for
`example, the UE could wait for the second column of reference signals then the performance
`degradation could be reduced.
`Another potential difference in link performance between FDD and TDD is related to the
`special subframe. As explained earlier the guard period is created by adjusting the number of
`symbols in DwPTS. When the DwPTS length is reduced we also eliminate some reference
`signals as the rule is not to move them to new locations. The potential loss is minor as refer-
`ence signals from a previous subframe could be taken into use to improve channel estimation.
`In the special case of UE specifi c reference signals we lose one column of reference signals
`even with full DwPTS length. In this case the UE cannot use reference signals from a previous
`subframe as these could have been transmitted with a different pre-coding.
`Short RACH performance is clearly worse compared to 1 ms RACH preamble and thus
`should be used only in environments where link budget is not foreseen to be an issue.
`
`12.6.2 Link Budget and Coverage for TDD System
`The link budget calculation aims at estimating the range of different bit rates. A detailed
`description of how to calculate link budgets for LTE is already given in Chapter 9. Here we
`focus on the differences between link budgets for TDD and FDD modes. The differences relate
`mainly to the limited maximum UE transmit power and in the following we therefore focus
`our attention on UL link budgets.
`The TDD UE cannot transmit continuously since the transmission must be switched off
`during the downlink reception, The UE will thus need to transmit with a larger bandwidth and
`a lower power density to achieve a similar bit rate to a FDD system. The lower power density
`is because the UE transmitter is limited on total maximum power, not on power per Hz.
`
`IPR2022-00468
`Apple EX1018 Page 449
`
`
`
`LTETDDMode
`
`387
`
`
`
`Suburban cell range
`
`LTE 2500 (TDD)
`
`LTE 2500 (FDD)
`
`LTE 2100 (FDD)
`
`LTE 900 (FDD)
`
`
`
`3.5
`0.0 0.5 1.0 1.5 2.0
`2.5 3.0
`km
`
`Figure 12.17 Uplink cell range for LTE FDD and TDD systems for 64 kbps
`
`As a simple example, if the downlink: uplink share is 3: 2, the UE transmission power den
`
`
`
`
`
`
`
`
`sity is reduced by IO x log I 0(2/5) = -4 dB as we need roughly 5/2 times the bandwidth for the
`
`
`
`
`TDD UL transmission. Another way of viewing this is that at a fixed distance from the base
`
`
`
`
`than the bit rate station the maximum achievable FDD bit rate will roughly be 2Y2 times larger
`
`
`
`achieved with maximum UE transmit power in a TDD system. Note that for DL, the power
`
`
`
`
`density can be assumed to be similar between FDD and TDD mode as the size of the power
`
`
`amplifier in eNodeB can be adapted to the system bandwidth.
`
`
`
`
`e over TDD do have an advantagFDD based systems So from a UL coverage perspective,
`
`
`
`
`
`systems due to the continuous transmission. Moreover, coverage with the TDD system can
`
`
`
`
`
`
`
`also be challenging because the TDD spectrum is typically situated at higher frequencies such
`
`
`
`
`as 2.3GHz or 2.5GHz. A cell range comparison for a suburban propagation environment is
`
`
`
`shown in Figure 12.17. The best coverage is obtained by using aFDD system at low frequency.
`
`
`
`The cell range for LTE900 FDD is four times larger (cell area 16 times larger) and LTE2500
`
`
`FDD is 80% larger than LTE2500 TDD. The assumed data rate is 64 kbps and the cell range
`
`
`
`
`is calculated with the Okumura-Hata propagation model with 18dB indoor penetration loss,
`
`
`
`
`
`50 m base station antenna height and -5 dB correction factor. For maximum LTE coverage,
`
`
`
`
`LTE TDD deployment at high frequency could be combined with LTE FDD deployment at a
`lower frequency.
`
`
`
`
`
`
`
`
`
`12.6.2.1 MCS Selection and UE Transmission Bandwidth for Coverage
`
`For a certain target bit rate different combinations of MCS and transmission bandwidth have dif
`
`
`
`
`
`
`
`
`
`
`
`
`
`ferent coverage and different spectral efficiency. From S hannon's information theory [5] we know
`
`
`
`
`that if we want to maximize coverage under a fixed total transmission power constraint we should
`
`
`
`
`
`
`increase MCS when physical layer spectral efficiency (SE) is< l bit/s/Hz and increase bandwidth
`
`
`
`when SE is > I b/s/Hz. As adjusting BW does not impact SE, but increasing MCS does, the link
`
`
`
`should be operated with an MCS which achieves a SE of I bit/sf/Hz (for LTE QPSK 2/3 would do
`
`
`
`
`
`
`the job). Then bandwidth can be adjusted to achieve the required bit rate with optimal coverage.
`
`IPR2022-00468
`Apple EX1018 Page 450
`
`
`
`388
`
`LTE for UMTS -OFDMA and SC-FD MA Based Radio Access
`
`
`
`UL Coverage for TDD and FOO, UE target bitrate 2 Mbps
`
`
`
`
`
`48
`
`43
`
`38
`
`33
`
`ID
`
`IL
`
`a:: 28
`'o
`'It 23
`
`18
`
`13
`
`8
`
`Data rates
`
`R:!sources
`
`2.50
`
`-UEBWFOO
`- -uE BWTOO
`-Bitrate FOO 2.00
`
`1.50 �
`
`J:J
`:!.
`ii
`
`1.00 :e
`
`ID
`
`0.50
`
`0.00
`
`--
`
`3
`0.00 0.10 0.20
`
`0.30 0.40 0.50
`
`Oista nee [km J
`
`Figure 12.18 Coverage and required number of physical resource blocks for a 2Mbps target bit rate
`
`
`
`
`
`
`
`
`for FDD and TDD UL. MCS in coverage limited region is QPSK ¾
`
`In Figure 12.18 an example of this process is given. AUE at the cell edge selects MCS
`
`
`
`
`
`
`
`
`
`QPSK ¾ and sets the transmission bandwidth according to the available transmit power and
`
`
`required SINR. Curves for both TDD and FDD UE are shown and assumptions are as given
`
`in [6]. From the figure we can see that when the UE moves towards the Node B, path loss
`
`
`
`
`is reduced and the link gain can be used to increase the UE transmission bandwidth. As the
`
`
`
`
`
`
`transmission bandwidth increases, the UE bit rate also increases. When the UE target bit rate
`
`
`
`
`
`is reached, and if the UE path loss is reduced further, we can start to increase MCS and reduce
`
`
`
`
`bandwidth to improve the SE while keeping the bit rate on target. At some point the maximum
`
`
`
`
`
`
`MCS is reached and only at this point can we start to reduce the UE total transmit power. While
`
`
`
`the TDD system supports 2Mbps in UL 300m from the eNodeB, FDD system increases the
`
`
`
`
`
`coverage for 2Mbps to 400 m. For more details on the interaction of UE transmit power, MCS
`
`
`
`and transmission bandwidth see Chapter 9.
`
`
`
`12.6.2.2 Coverage for Low Bit Rates
`
`
`
`
`
`When the data bit rate decreases, the relative overhead from header and CRC will increase. The
`
`
`
`
`
`
`
`effect of this is that even in FDD mode it does not make sense to schedule UEs in coverage
`
`
`
`problems with very narrow bandwidth and low MCS. An example for a data bit rate of 64 kbps
`
`
`
`is illustrated in Table 12.3. From this we can see that due to excessive overhead, the UE band
`
`
`
`width increase for UL in a TDD 3DU2UL configuration is only a factor 1.6, corresponding
`
`
`
`to a transmit power loss of 2 dB, not 4dB as in the example given above where overhead was
`not taken into account.
`
`IPR2022-00468
`Apple EX1018 Page 451
`
`
`
`LTE TDD Mode
`
`389
`
`Table 12.3 Required UE transmission bandwidth to support 64 kbps for TDD
`and FDD
`
`System
`
`TDD UL with 3DL/2UL FDD UL
`
`Service bit rate (kbps)
`MCS
`Data bits per TTI (bits)
`Header (3 byte)
`CRC (3 byte) (bits)
`Total bits per TTI (bits)
`Payload per PRB in physical layer
`(2 symbols for DM RS
`1 symbol for SRS)
`Required number of PRBs
`
`64
`QPSK 1/10
`160
` 48
`
`208
` 27 bits
`
`64
`QPSK 1/10
`64
` 48
`
`112
` 27 bits
`
` 8
`
` 5
`
`Table 12.4 Required SINR for VoIP in TDD and FDD with and without TTI
`bundling. UE transmission bandwidth assumed to be 360 kHz (2 PRB)
`
`System
`
`TDD (3DL/2UL or 2DL/3UL)
`
`FDD
`
`Bundling enabled
`Number of transmissions
`Required SINR
`
`No
` 5
`−3.3 dB
`
`Yes
` 8
`−5.3 dB
`
`Yes
`No
`12
` 7
`−4.7 dB −7.04 dB
`
`One way of reducing the overhead for low bit rates is to use TTI bundling, as described in
`section 12.3.3. When TTI bundling is enabled the maximum number of retransmissions within
`a certain time limit is increased. In that way we can have more aggressive MCS selection and
`thus lower the relative overhead from protocol header and CRC. Due to the reduced number of
`UL TTIs in TDD, the potential link budget improvement from TTI bundling is not as impor-
`tant as for FDD. As shown in Table 12.4, for a VoIP service in 3DL/2UL confi guration we can
`improve the number of transmissions of one VoIP packet within 50 ms from 5 to 8 TTIs, which
`is a 2 dB improvement to the link budget. We note that the link budget gain from TTI bundling
`in TDD mode is similar in both 2DL/3UL and 3DL/2ULconfi guration.
`Finally, when operating at very low UL bit rates far away from the base station, coverage
`on UL control channels could also become a limiting factor. The control channel for TDD
`has been designed so that if ACK/NACK bundling mode is selected then the required receiver
`sensitivity is similar between TDD and FDD.
`
`12.6.3 System Level Performance
`TDD mode is in many aspects similar to FDD and this is also valid when we analyze the systems
`from a performance point of view. In general, when we compare time and frequency duplexing
`there are some differences related to the different ways of using the spectrum which make it
`diffi cult to make a 100% fair comparison of spectral effi ciencies. Whereas a TDD mode system
`needs a guard period between UL and DL, a FDD system needs a large separation in frequency.
`Secondly if TDD systems have the same partition of UL and DL resources they can operate in
`adjacent bands; if not they need to be separated in a similar way to UL and DL for FDD.
`
`IPR2022-00468
`Apple EX1018 Page 452
`
`
`
`390
`
`LTE for UMTS – OFDMA and SC-FDMA Based Radio Access
`
`Another spectrum related issue is that for the TDD system to provide a similar capacity to a
`FDD system the DL and UL system bandwidth needs to be double that of a FDD system. This
`impacts the RRM and users will typically need to operate with larger transmission bandwidths.
`That can be challenging for UL data transmission due to the limited transmission power of
`the UE.
`One advantage for the TDD RRM solution is the possibility of exploiting channel reciproc-
`ity. In the current RRM framework (see Chapter 8) two parallel mechanisms are available for
`obtaining channel state information. For DL, the UE can be confi gured to feedback CQI, PMI
`and RI reports based on measurements of DL reference signals. In UL the UE can transmit SRSs
`so that Node B can measure the radio channel. For TDD mode, when channel reciprocity is
`present we ideally need only one of these mechanisms as the DL channel state can be inferred
`from the UL channel state or inversely. As mentioned earlier, there are challenges before this
`could work in a practical RRM solution, such as differences in UL/DL interference levels,
`different UL/DL antenna confi gurations and lack of UL/DL radio chain calibration, but on the
`other hand gains could be important. UL sounding, for example, can take up more than 10% of
`the UL system capacity. See section 12.2.4 for a further discussion of channel reciprocity.
`
`12.6.3.1 Round Trip Time for TDD Systems
`
`Feedback control loops are used for quite a few purposes in the LTE system. As TDD systems
`do not have continuous transmission and reception we might expect that the round trip time
`for such control loops would be increased for TDD systems, potentially degrading the system
`performance. However, due to the need for processing time, i.e. time for the UE or the Node B
`to decode and encode the control information, the typical round trip times between TDD and
`FDD are quite similar. Since the resulting delays are usually about 10 ms compared to 8 ms for
`the FDD RTT, the impact of the TDD frame structure is rather low and unlikely to impact on
`the performance of TCP/IP regardless of being run over LTE TDD or FDD.
`
`12.6.3.2 Scheduling
`
`One of the key LTE RRM features is channel aware scheduling, which is available for both
`UL and DL. Under the right conditions this feature can bring gains in spectral effi ciency of up
`to 50% with even more important improvements to the coverage. To maximize the scheduling
`gain it is important to have frequency selective channel knowledge and fl exibility in the control
`signaling to allocate UEs to the optimal frequency resources. Obtaining detailed frequency
`selective channel state information and enabling fl exible resource allocation in the frequency
`domain is very costly in terms of control signaling.
`For DL a number of different frequency resource allocation schemes are specifi ed in the
`standard. For channel aware scheduling, the most effective allocation scheme specifi ed gives a
`bit mask where each bit corresponds to