Chapter 9

LTE URLLC

Abstract

This chapter describes the evolution of the LTE technology toward Ultra-Reliable-Low-Latency-Communications. It starts by providing a background as to why such a substantial step in evolving the LTE technology was taken. The background is followed by a detailed description of the changes both to the physical layer design and idle and connected mode procedures. This involves everything from how the physical channels are in detail designed to the resource allocation used. The chapter is written assuming the reader has some basic LTE knowledge, and describes mainly the changes introduced on top of ‘legacy LTE’

Keywords

Third generation partnership project (3GPP); 3GPP RAN; LTE; 4G; 5G; mMTC; Critical MTC; cMTC; URLLC; sTTI

9.1. Background

In Release 15 of the 3GPP RAN specifications, the work on IoT expanded from being focused on the low-end part of the IoT spectra (ultra-low device cost, no stringent requirements on latency, etc) to also cater for more critical IoT applications. For more background on this development, see Chapters 1 and 2. Considering the global success of LTE, being the most advanced 3GPP technology (to that date), it was a natural choice to develop this technology toward more critical IoT applications. In parallel, work was also initiated in NR, see Chapter 2, targeting critical IoT applications, see Chapter 11.
To understand why a significant development of the LTE specification was needed to accommodate cMTC applications and services, let's first look at the abbreviation itself. URLLC stands for Ultra-Reliable and Low Latency Communication. From a RAN perspective, to make a radio link reliable, an improvement of the experienced radio signal quality (i.e. SINR) would quickly reduce the error rate of the transmission. However, if a too high SINR is required for the service to be considered ultra-reliable, the cellular network would contain many coverage holes where the reliability cannot be achieved. Furthermore, the use of time-based repetitions, as many of the techniques in this book utilizes to improve the reliability, will have a direct impact on the latency. Also, as described in Chapters 1 and 2, the extreme requirements on a low latency bound from the IMT-2020 requirements (upper bounded by 1   ms) would make the Release 14 LTE specification insufficient since the shortest possible transmission duration over-the-air already consumes the 1   ms latency budget (the IMT-2020 requirement also include e.g. alignment delay and processing time). It is however important to note that the aim of the LTE URLLC work was not only to cater for the extreme requirements of IMT-2020, but also to prepare LTE for other use cases with a less stringent latency bound and/or reliability requirement. The IMT-2020 requirements, however is the most stringent, and hence to a large extent what LTE is evolved toward fulfilling.
Knowing the limitations of the Release 14 LTE technology and the targeted URLLC performance, 3GPP Release 15 initiated work toward fulfilment of the IMT-2020 requirements for URLLC in two-stages.
It can be noted that improvements are limited to devices in connected mode and exclude any modification to PSS/Synchronization Signals (Primary and Secondary Synchronization Signals), Physical Broadcast Channel, Physical Control Format Indicator Channel and Physical Random Access Channel, the initial access procedures, SIB (System Information Block) signaling and Paging procedures. One exception is the work on reduced control plane latency, see Section 9.3.1.1, where processing times have been reduced to accommodate the IMT-2020 control plane latency requirements [3].
The following chapter will make no distinction from where a specific functionality originated, i.e. from which work item, but will describe the LTE URLLC functionality at the end of Release 15, including the work performed for both sTTI, URLLC work items and the work on improved control plane latency.
Throughout this chapter the terminology short transmission time interval (sTTI) will be used to refer to the functionality of shortening the air interface transmission duration to slot or subslot level.
The reader is assumed to have basic LTE knowledge and for a more complete coverage of LTE, the reader is referred to e.g. Ref. [4]. This chapter focuses on explaining the difference in the URLLC design compared to Release 14 LTE specifications. The latter is often referred to as ‘legacy LTE’.

9.2. Physical layer

This section describes the physical layer of LTE URLLC. First, the reader is provided with the design principles in evolving the LTE technology toward URLLC, followed by a detailed review of the final design chosen.
In addition to the main thinking behind the LTE re-design, the reader is also provided with a detailed review of the changes to the physical resources, the downlink and uplink physical channels, and the impact on processing timelines and timing advance from shortening the transmissions time.

9.2.1. Radio access design principles

To vastly improve the air interface latency and to improve reliability has a significant impact to the overall system design, as the following sections will show. Hence, some design principles are needed to ensure proper system operation:

9.2.2. Physical resources

To fulfill the IMT-2020 latency requirement of 1   ms from the radio protocol layer (see Chapter 1), the latency over the air interface need to be significantly less than 1   ms. Considering that the subframe in LTE is already 1   ms which is the transmission duration used by the data channels (PDSCH/PUSCH), this had to be shortened. To limit the additional network complexity from URLLC, but allow enough flexibility, two additional transmission durations of Physical Downlink Shared Channel (PDSCH) and Physical Uplink Shared Channel (PUSCH) were defined: slot transmission and subslot transmission. A slot was already defined in LTE as the split of a subframe in two equal parts (each of 0.5   ms duration), but typical data channel transmission was still restricted to subframe duration, which changed with the introduction of sTTI. With the introduction of URLLC, a slot is further split into three subslots, each of two or three-symbol duration. The new radio frame in LTE is shown in Fig. 9.1.
As can be seen from the figure, two different subslot patterns have been defined. Their use depends on if the transmission takes place in the uplink or downlink, and in case of downlink, in which symbols the data transmission starts (in the 3GPP specifications. This symbol index is referred to as l DataStart ) and is typically dependent on how many symbols are used for PDCCH, see Table 9.1.
It could be noted that the 3GPP specification does not mention the use of different ‘subslot patterns’, but this is effectively how the design can be seen.
Fig. 9.2 shows the subslots available for data transmission depending on the direction and, for the downlink, the symbol index where the data starts. It can be seen that the subslot number in downlink starts either with ‘0’ or ‘1’ depending on the starting symbol index for data. One can note that from subslot number 2 onwards, the subslot borders are aligned.
This design might seem complicated, having different subslot lengths in the subframe, and also, using different subslot patterns depending on certain given conditions. The main reason for this is to comply with the design principle of backwards compatibility mentioned in Section 9.2.1. In the downlink, the data transmission in LTE can start in different symbol indices and could be occupied by PDCCH in the, up to three, first symbols (four symbol PDCCH is not supported in sTTI operation), excluding them from possible data transmission. Since the subslot boundary aligns with the slot boundary it is impossible to split the seven symbols of a slot into an even number of subslots.
image
Fig. 9.1 Subslot patterns in the LTE frame structure.

Table 9.1

Subslot pattern used depending on data symbol start and transmission direction.
Direction Starting index of data symbol in the subframe (l DataStart ) Subslot pattern
Uplink 0 1
Downlink 1 1
2 2
3 1

image

image
Fig. 9.2 Subslots available for data depending on direction.
In legacy LTE, the size of the downlink control region is signaled via Physical Control Format Indicator Channel containing the CFI. Since the subslot pattern is dependent on the start of the downlink data (l DataStart ), see Fig. 9.2, which follows the CFI, and, since the PDCCH mapping changes with the size of the control region, there is a need for the device to acquire the CFI before processing the rest of the subframe. Considering the reliability improvements of including repetitions of the downlink data (see Section 9.2.3.3) and the use of multiple attempts of the dynamically changed per TTI (DCI) decoding (see Section 9.2.3.3) there is a risk that the CFI reception becomes the bottleneck, in terms of performance. To prevent this, the CFI can be configured by RRC for each device (which then determines the l DataStart ). Once configured with CFI, the device is no longer expected to decode CFI from PCFICH, removing the potential bottleneck.
Some restrictions have also been agreed to limit the complexity. The use of extended cyclic prefix is for example not supported in case of subslot or slot operation. This is motivated by the reduced link budget implied by sTTI operation (see Chapter 10), and that extended cyclic prefix primarily is used for large cell ranges.
image
Fig. 9.3 Downlink physical channels in LTE URLLC (impacted channels compared to legacy LTE is filled).

9.2.3. Downlink physical channels and signals

The set of physical channels used in the downlink in case of sTTI operation are shown in Fig. 9.3. For a reference to the transport channels and logical channels that the physical channels are associated with, please refer to Section 5.2.4.
With the introduction of LTE URLLC, the only impacted channels on the downlink are the Physical Downlink Shared Channel, PDSCH and the Short Physical Downlink Control Channel (SPDCCH), out of which SPDCCH is a completely new physical channel.
A PDSCH transmitted using slot or subslot duration are referred to as slot-PDSCH and subslot-PDSCH respectively. Similarly, for downlink control, slot-SPDCCH and subslot-SPDCCH is used when referring to a specific transmission duration.

9.2.3.1. Downlink reference signals

Subframe DMRS - any
CRS - not in MBSFN-subframes
Subcarrier spacing 15   kHz
CRS bandwidth Full system bandwidth
DMRS bandwidth Same as associated SPDCCH/PDSCH
CRS frequency location According to Figure 5.9 in every PRB
DMRS frequency location According to Fig. 9.4 (subslot) and 9.5 (slot) in affected PRBs

The reference signals used for the PDSCH and SPDCCH can be configured to be either a Cell-specific Reference Signals (CRS) or a device-specific Demodulation Reference Signals (DMRS).
The definition and mapping onto physical resources of the CRS is not changed in LTE URLLC and is the same as for legacy LTE, as also described in Section 5.2.4.3.
The mapping of the DMRS however changes with the use of slot and subslot transmissions. The reason to re-define the DMRS is mainly to avoid the delay and buffering at the receiver of OFDM symbols arriving earlier than the DMRS and to have, an as accurate as possible, channel state information.
In case of DMRS-based downlink transmission, the device can assume the precoder of the DMRS and payload used by the network to be the same over a predefined set of PRBs. This is referred to as a precoder resource group (PRG) and is always set to two resource blocks in frequency for subslot/slot-PDSCH and for the associated DMRS-based SPDCCH. Using a PRG size larger than one reduces the complexity of the channel estimation for the device. A given PRG is defined relative to the physical resource grid (not relative to the resource allocation) and hence PRG k, contains physical resource blocks (PRB) 2k and 2k+1. The use of a PRG size of two opens up for a design of the DMRS pattern over two consecutive resource blocks in frequency. This helps to avoid channel estimation errors due to extensive extrapolation of the channel estimate for resource elements (REs) ‘outside’ the span of the DMRS resource in frequency, while at the same time keeping the DMRS density low (compared to having the DMRS pattern over a single resource block). For example, in Fig. 9.4, the REs that are outside the span of the DMRS is only one at high end of the frequency range for the baseline pattern and pattern v2.
The possible subslot-PDSCH DMRS patterns are shown in Fig. 9.4. Up to four antenna ports (where one port is used per layer in case of MIMO transmission) are defined. In case of 2-layer transmission, the two ports 7 and 8 are multiplexed using orthogonal cover codes (OCC), over two OFDM symbols. As the name implies, the codes applied to the repeated symbols are orthogonal, see Ref. [4] for details.
In case of 4-layer transmission, the same OCC in time is used over a second pair of ports 9 and 10 placed in other positions in frequency.
Which one of the four DMRS patterns to use is selected based on pre-defined rules. The rules have been decided to avoid too many restrictions in network configurations and to ensure backwards compatibility.
To select a pattern, the rules below are followed:
The baseline pattern is selected:
Pattern v0, v1 and v2 is selected:
image
Fig. 9.4 Subslot-PDSCH DMRS pattern.
image
Fig. 9.5 Slot-PDSCH DMRS pattern.
Pattern v0 is also selected:
The subslot-DMRS pattern is used for both subslot-PDSCH transmission and for DMRS-based SPDCCH associated with subslot-based or slot-based transmission.
The slot-PDSCH DMRS pattern is only spanning one resource block and shown in Fig. 9.5. In this case, the pattern is also following the v shift parameter, i.e. mod(v shift ,3)   =   x determines the shift to be used (v0, v1 or v2) in normal subframes. In MBSFN subframes, the baseline pattern is used.
The DMRS pair is mapped to symbol index three and four in the slot if the PDSCH transmission is in the first slot of the subframe, while it is mapped to symbol index two and three of the slot if the PDSCH transmission is in the second slot.

9.2.3.2. Slot/subslot-SPDCCH

Subframe Any
Duration 1 or 2 OFDM symbols (CRS)
2 or 3 OFDM symbols (DMRS)
Repetitions No
Subcarrier spacing 15   kHz
Bandwidth Any - granularity of one PRB for CRS
Any - granularity of two PRB for DMRS
Frequency location Any, with the restriction of the DMRS to align with the PRG grid

9.2.3.2.1. General
To ensure low latency a short periodicity of scheduling opportunities was introduced new scheduling information for each subslot or slot being scheduled, as illustrated by the lower figure in Fig. 9.6.
Adopting such a design, the downlink control channel had, to a large extent, be redesign for sTTI operation.
image
Fig. 9.6 Possible options for the downlink control channel occurrence (only the bottom approach was adopted).
In this clause, the new control channel, SPDCCH, is described.
It should however be noted that the control information associated with sTTI operation can be carried by both PDCCH and SPDCCH. Considering the mapping of PDCCH to the first symbols in the subframe, PDCCH is only used to schedule sTTI#0 (both in case of slot and subslot operation).
9.2.3.2.2. SPDCCH resource set
A SPDCCH is contained within a SPDCCH resource set, i.e. a set of possible resources that the SPDCCH can be mapped to.
The resource set is configured by RRC signaling with the following configuration options:
The details of the configurations will be further described in this section.
A device can be configured with up to two SPDCCH resource sets in each subframe type (MBSFN and non-MBSFN) in the device specific search space. In case of subslot operation, the subslot numbers applicable to each SPDCCH resource set are configured. In Fig. 9.7, SPDCCH resource set 0 has been configured to apply to subslot number 0, 2 and 5, and hence SPDCCH resource set 1 applies to subslot number 1, 3 and 4.
The main reason to allow different SPDCCH resource sets depending on subslot number is the potential variation in SPDCCH performance depending on the subslot it is mapped to. Since the SPDCCH is rate-matched around for example the CRS, DMRS and CSI-RS, the code rate for a given AL can vary over time, which is not desirable. To compensate for this behavior, two resource sets can be used where one could be configured with a higher AL and is configured to apply in subslots experiencing a larger overhead (a higher level of rate-matching).
image
Fig. 9.7 SPDCCH resource sets.
9.2.3.2.3. Mapping to physical resources
Provided that we have configured the physical resources for the SPDCCH resource set, there are multiple ways on how to map a given SPDCCH to the resource set. As already mentioned, the mapping can be either distributed or localized. When to use which mapping can depend on multiple factors, such as service type and reference signal type. However, for cMTC services for fulfilling the URLLC requirements a distributed mapping using CRS-based channel estimation will provide a good trade-off between overhead and performance. Using a distributed mapping will increase diversity and hence improve reliability.
To assist the mapping, the concept of short resource element groups (SREG) and short control channel elements (SCCE) are introduced (similar to REGs and CCEs used in legacy LTE).
An SREG is defined over one OFDM symbol mapped to one resource block in frequency. That is, there are 12 REs that constitute a SREG. However, as already mentioned, not all 12 REs might be used for the SREG if e.g. CRS is mapped to the same physical resources. For a given SREG, the SPDCCH resources will be mapped to at most 12 REs.
An SCCE can be seen as a minimum building block of an SPDCCH. Each SCCE consists of a set of SREGs. The number of SREGs per SCCE is dependent on the SPDCCH resource set configuration, and are given in Table 9.2.
Each SPDCCH candidate belongs to an AL. For different ALs, different number of SCCEs are aggregated. The aggregation is carried out to achieve different level of coverage in the cell, considering that different UEs will experience different radio conditions.
The ALs defined for the SPDCCH candidates are 1, 2, 4 and 8. The AL corresponds to the number of SCCEs constituting the SPDCCH candidate, e.g. an SPDCCH of AL 4 consists of 4 SCCEs.
To be able to describe where an SPDCCH candidate is mapped onto the physical resources, the SREGs are first logically numbered. In the examples below, it is assumed that the SPDCCH resource set consists of 20 RBs (in reality, the size of the SPDCCH resource set is only limited by the system bandwidth).
The SREGs are numbered in different ways depending on the reference signal type of the SPDCCH:
The next step in constructing the SPDCCH is the mapping of SREGs to SCCEs. There are four different configurations that have an impact on how the mapping is done:
image
Fig. 9.10 SREG to SCCE mapping, Localized CRS-based SPDCCHillustrated for a 2 symbol SPDCCH resource set.
For CRS-based SPDCCH using localized mapping, the SREG index corresponding to SCCE index n is given by:
n · N SREG SCCE + j
image (9.1)
where
n = 0 , … , N SCCE − 1
image
N S C C E image is the number of SCCE in the SPDCCH resource set,
j = 0 , … , N SREG SCCE − 1
image
N SREG SCCE image is the number of SREGs per SCCE.
The mapping is illustrated in Fig. 9.10 for a 2-symbol CRS-based SPDCCH.
For CRS-based SPDCCH using distributed mapping, the SREG index corresponding to SCCE index n is given by:
m o d ( n , ⌊ N R B N S R E G S C C E ⌋ ) + ⌊ n ⌊ N R B N S R E G S C C E ⌋ ⌋ · N R B + j · ⌊ N R B N S R E G S C C E ⌋
image (9.2)
where
j = 0 , ⋯ , N S R E G S C C E − 1
image
N R B image is the number of resource blocks in frequency in the SPDCCH resource set.
This results in the SREG to SCCE mapping as shown in Fig. 9.11 for a two symbol CRS-based SPDCCH.
image
Fig. 9.11 SREG to SCCE mapping, distributed CRS-based SPDCCH, illustrated for a 2 symbol SPDCCH resource set.
image
Fig. 9.12 SREG to SCCE mapping, DMRS-based SPDCCH, illustrated for a 3-symbol SPDCCH resource set.
For DMRS-based SPDCCH using either localized or distributed mapping, the SREG indexes corresponding to SCCE index n is given by the same equation as for CRS-based localized mapping (Eq. 9.1), but since the SREG numbering is performed in a time first manner, the SCCEs are formed more condense in frequency (compare Figs. 9.10 and 9.12), see Fig. 9.12.
It should be noted that, so far, the construction of SREGs, SCCEs and their mapping in the logical domain does not say anything about where a given SCCE index is mapped onto the physical resources. This is however easily derived for a given SPDCCH resource set. Consider in the examples above an SPDCCH resource set of 20 resource blocks. In this case, the mapping is simply done per resource block from low to high frequency. An example of mapping the DMRS based localized mapping to the physical resources in two different contiguous sub-bands is shown in Fig. 9.13.
Hence, it can be noted that although the mapping of SREGs to SCCEs is localized, the actual mapping to physical resources can still be distributed, depending on the SPDCCH resource set configuration.
One restriction that has been adopted related to the SPDCCH resource set configuration for DMRS-based SPDCCH is that the PRB need to be configured in no smaller entities than in contiguous pairs of resource blocks (i.e. the same PRG, see Section 9.2.3.1).
Let's now look at the final mapping stage of the SCCEs to a SPDCCH candidates. For an SPDCCH candidate of AL one, the mapping is already given (since a SPDCCH candidate of AL one consist of one SCCE, which we already have the mapping for), but for higher ALs, the CCE indices constituting a given SPDCCH need to be determined.
image
Fig. 9.13 Mapping of SCCEs to physical resources.
For CRS-based SPDCCH, the logical SCCEs indices corresponding to SPDCCH candidate m of the SPDCCH search space at AL L in a given slot/subslot are given by
m o d ( Y ( L ) + L ⋅ ( m o d ( m ⋅ N S C C E L ⋅ M ( L ) , [ N S C C E L ] ) + i ) , N S C C E )
image (9.3)
where.
Y ( L ) image is the RRC configured starting position for AL L
i = 0 , … , L − 1
image
N S C C E image is the total number of SCCEs in the SPDCCH resource set
m = 0 , … , M ( L ) − 1
image
M ( L ) image is the number of SPDCCH candidates to monitor at AL L.
The resulting SPDCCH candidates for different aggregation levels (and different number of candidates) are shown in Fig. 9.14.
The localized mapping is not illustrated here but is achieved using the same SCCE to SPDCCH relation.
For DMRS-based SPDCCH and localized mapping, the same equation is used as in the mapping of SCCEs to SPDCCH candidates for CRS-based SPDCCH (Eq. 9.3). For the DMRS-based SPDCCH and distributed mapping, the SCCE to SPDCCH candidate mapping is performed according to:
m o d ( Y ( L ) + ⌊ m · N S C C E L · M ( L ) ⌋ , N S C C E L ) + i · [ N S C C E L ]
image (9.4)
The mapping is illustrated in Fig. 9.15.
The mapping of the SPDCCH candidate to REs are done in a frequency-first-time-second manner. This implies that for CRS-based SPDCCH and localized mapping, the REs between a higher and a lower AL can be completely overlapping, see Fig. 9.16. Hence, there could be ambiguity for the device on what AL has been sent by the network.
To solve this, an interleaver on SREG level has been introduced. For each SPDCCH candidate, the SREGs are interleaved using a block interleaver with L (i.e. the aggregation level) rows and 4 (the number of SREGS per SCCE for CRS-based SPDCCH) columns.
image
Fig. 9.14 SPDCCH candidates for CRS-based SPDCCH and distributed mapping.
image
Fig. 9.15 SPDCCH candidates for DMRS-based SPDCCH and distributed mapping.
image
Fig. 9.16 Overlapping REs for different SPDCCH candidates for CRS-based SPDCCH and localized mapping without modified bit mapping.
9.2.3.2.4. Overview
To tie it all together, Fig. 9.17 shows an overview of the different steps in the creation of an SPDCCH candidate depending on the SPDCCH resource set configuration.
See Sections 9.2.3.2.2–9.2.3.2.3 for more details.

9.2.3.3. Slot/subslot-PDSCH

Subframe Any
Basic TTI Subslot, slot
Repetitions Up to 6 transmissions
Subcarrier spacing 15   kHz
Bandwidth See Section 9.3.2.9.1
Frequency location See Section 9.3.2.9.1

image
Fig. 9.17 Overview of the SPDCCH procedure.
The PDSCH, is used to transmit unicast and broadcast data. For sTTI the focus is on unicast transmission. The data packet from higher layers is segmented into one or more transport blocks (TB), and PDSCH transmits one TB at a time. The PDSCH for sTTI transmission is inherited from LTE and hence in this section, the differences introduced with sTTI will be the focus. For more details on the PDSCH design for LTE, an interested reader is referred to e.g. Ref. [4].
When shortening the transmission duration using subslot and slot operation, the available resources, for a given bandwidth, compared to subframe transmission is heavily reduced. To achieve roughly the same code rate for different Modulation and Coding Schemes (MCS), the information bits mapped to the subslot/slot need to be reduced, compared to subframe operation. To limit the impact to the specification, this was achieved by using the existing TBS tables in LTE, but scaling the TBS value by a fixed factor, α, depending on if the transmission duration is subslot or slot:
After scaling, the resulting TBS value is rounded off to the closest valid TBS value in the existing table. It can be noted that this approach will result in a varying code rate over the different transmission opportunities, e.g. due to the overhead and subslot duration, but was chosen due to its simplicity in implementation.

Table 9.3

Transmission modes for slot/subslot-PDSCH.
Transmission mode Description Note
TM1 Single antenna transmissions FDD, TDD
TM2 Transmit diversity FDD, TDD
TM3 Open-loop codebook-based precoding/transmit diversity FDD, TDD
TM4 Closed-loop codebook-based precoding FDD, TDD
TM6 1-layer closed-loop codebook-based precoding FDD, TDD
TM8 Non-codebook-based precoding, up to 2 layers TDD
TM9 Non-codebook-based precoding, up to 4 layers FDD; TDD
TM10 Non-codebook-based precoding, up to 4 layers FDD, TDD
As for legacy LTE, a variety of transmission modes are supported using multiple antenna transmission and/or reception. Multiple-input multiple output, MIMO, transmission is supported with a maximum of four MIMO layers for both CRS- and DMRS-based transmission. Irrespective of the number of MIMO layers, a single codeword is used. Hence, only a single HARQ bit need to be reported from the device for each received downlink TTI, which improves Short Physical Uplink Control Channel (SPUCCH) performance.
The supported transmission modes are listed in Table 9.3. The downlink transmission modes for sTTI are configured separately from subframe operation.
For the 5G URLLC requirements (see Chapter 10), there is no high demand for high throughput but rather for a high reliability. Increasing diversity will generally increase reliability and hence transmit diversity is the expected mode of operation when reliability is of primary concern.
The mapping of the PDSCH to the physical resources follows the slot boundaries for slot operation and the subslot layout for subslot operation, see Fig. 9.2.
In case of slot operation in TDD, PDSCH transmission is supported in DwPTS for the first slot except in special subframe configuration 0 and 5, and in the second slot in case of special subframe configuration 3, 4 and 8, see Fig. 9.18. Each transmission in the two slots of DwPTS and UpPTS are scheduled independently.
For DMRS based PDSCH, when shortening the transmission duration, a natural consequence is that the DMRS overhead is increased since at least one DMRS symbol should be associated with each subslot/slot-based data transmission. However, when reducing the transmission duration to two or three symbols, as in the case of subslot operation, the overhead can become significant. To solve this, a sharing mechanism has been introduced to allow multiple subslots sharing the same DMRS. A DMRS can be shared between two consecutive subslots. Whether DMRS is included in a given subslot is indicated to the device dynamically via DCI.
9.2.3.3.1. Blind repetitions
image
Fig. 9.18 Slot- PDSCH transmission in different special subframe configurations.
To realize the blind repetitions, a scheme has been adopted where the DCI indicates the number of repetitions used from that DCI onwards. The repetition value is decremented for each additional repetition. Hence, the device need only detect and decode one of these DCIs in order to understand the number of repetitions remaining. However, if the first DCI(s) are missed, the associated blind repetitions will also be missed. Hence, it is still of importance for the network to ensure good downlink control reception to increase the probability of the DCI being received in the initial transmission (ensuring reception of all blind repetitions). After decoding the DCI, the device assumed that the rate-matching and the resource allocation is the same for the remaining repetitions as in the TTI where the DCI was received.
The DCI can indicate number of transmissions 1,2,3, and, 4 or 6. Which one of {4,6} that is used is configured by RRC signaling.
An example is shown in Fig. 9.19 where four total transmissions are configured. The DCI decrements the number of blind repetitions with increasing TTI. On the left side of the figure, the behavior from the network is shown while to the right, the understanding of the device. It is assumed here that the device misses the DCI of the initial transmission but receives it in the first blind repetition. The DCI indicates that two repetitions remain. Based on this and the knowledge that the resource allocation and the rate-matching for the remaining repetitions follow the decoded DCI, the remaining repetitions can be received and combined to achieve the processing gain required to improve the reliability.
image
Fig. 9.19 Blind PDSCH repetitions.
Sharing of the DMRS between subslots to reduce overhead, cannot be used in case of blind repetitions.
In case of blind repetitions, the processing timeline (see Section 9.2.5) for the HARQ-ACK feedback to be reported by the device is based on the last repetition in the sequence.

9.2.4. Uplink physical channels and signals

9.2.4.1. Uplink reference signals

Subframe Any
Subcarrier spacing 15   kHz
DMRS bandwidth Same as associated PUSCH/PUCCH
DMRS frequency location Same as associated PUSCH/PUCCH

The reference signals in the uplink span the bandwidth of the associated allocation, as for legacy LTE operation, whether it being uplink control or uplink data.
The Demodulation Reference Signals (DMRS) associated with the physical uplink control channel are re-used from legacy LTE.
The reference signals associated with uplink data transmissions can use one out of two repetition factors in time domain, within the OFDM symbol duration. Using no repetition factor, the DMRS occupies all REs in frequency, while if a time-domain repetition is used, every other REs in frequency will be occupied. This stems from the properties of the Fourier transform. The repetitions in time domain will correspond to using interleaved frequency domain multiple access (IFDMA) modulation in frequency domain (where every Nth subcarrier will be occupied, transforming to N repetitions within the ODFDM symbol in time domain).
When using multi-layer transmission from a single user (SU-MIMO), the DMRSs of the different layers are multiplexed using different cyclic shifts. The DMRSs are transmitted on the same physical resources but with sequences that are orthogonal, as long as the delay spread of the propagation channel does not exceed the cyclic shift applied in case IFDMA is not configured. In case IFDMA is configured, the layers are multiplexed using different cyclic shifts and subcarriers offsets in the IFDMA configuration (also commonly referred to as different ‘combs’).
The use of IFDMA or not, and the cyclic shift to apply are each indicated by a 1-bit field in the DCI.
One of the main benefits of using IFDMA is that two devices can share the same OFDM symbol for reference signal transmission but can be allocated different resources in frequency. A sharing of DMRS is possible also if no IFDMA modulation is used (by the use of different cyclic shifts) but in this case, the resource allocation of the two users need to be fully overlapping in order to keep the orthogonality between the DMRSs.
IFDMA transmission of the DMRSs for two devices using different resource allocations is illustrated in Fig. 9.20.
image
Fig. 9.20 Uplink IFDMA DMRS transmission.
As with subslot-PDSCH transmissions (see Section 9.2.3.3), to minimize DMRS overhead, DMRSs can be shared across subslots also in the uplink. As in the downlink, the sharing of DMRS is dynamically indicated to the device. However, instead of an indication of the DMRS present or not, two bits are used to provide flexibility of using different subslot allocations, providing different level of DMRS overhead and scheduling options. The, up to four, combinations per subslot in the subframe is shown in Table 9.4. The DMRS can be shared amongst at most the three consecutive subslots of a slot.
An illustration of the signaling is shown in Fig. 9.21.
In case of slot operation, the DMRS overhead does not vary and the DMRS is self-contained in each slot with the DMRS symbols positioned as for subframe operation in legacy LTE, see Fig. 9.22.

Table 9.4

PUSCH DMRS pattern for dynamic scheduling through uplink grant.
Bit value Subslot#0 Subslot#1 Subslot#2 Subslot#3 Subslot#4 Subslot#5
00 R D D R D R D R D R D R D D
01 D D R D R D D D R D R
10 D D D D |R D D
11 D D |R D D |R

image

‘R denotes a DMRS symbol, ‘D’ denotes a data symbol, ‘|’ denotes the subslot boundary.

image
Fig. 9.21 Example of subslot PUSCH DMRS sharing including the DMRS pattern signaling.
image
Fig. 9.22 DMRS positions in case of slot-operation.

9.2.4.2. Slot/subslot-SPUCCH

Subframe Any
Basic TTI Slot/subslot
Repetitions No
Subcarrier spacing 15   kHz
Bandwidth 1 PRB (SPUCCH format 1/1a/1b/3)
1,2,3,4,5,6,8 PRB (SPUCCH format 4)
Frequency location Any PRB
Frequency hopping If applicable, between 2 PRB locations

9.2.4.2.1. General
The SPUCCH, is used to carry the following types of Uplink Control Information (UCI):
In contrast to other PUCCH designs in LTE, SPUCCH does not carry downlink Channel State Information (CSI), which is only carried by subslot-PUSCH and slot-PUSCH.
Different SPUCCH formats exist. The difference between the different formats is mainly the number of HARQ-bits it can carry and the multiplexing rate of the physical channel (i.e. how many UEs can be simultaneously transmitting on the same physical channel).
The SPUCCH format selection is based on the number of HARQ-ACK bits to be transmitted. The size of the HARQ-ACK bit field depends on multiple factors:
The device will determine the number of HARQ-ACK bits to be carried by a slot-SPUCCH/subslot-PUCCH considering the factors mentioned above and select the SPUCCH format accordingly.
An overview of the different SPUCCH formats is given in Table 9.5.
For slot operation, in case both SPUCCH format 3 and 4 are configured, SPUCCH format 4 is used from 12 bits onwards.
The channel coding used for SPUCCH is determined based on the payload size, irrespective of format 3 or 4 being used. A Reed-Muller block code is used for payload sizes of 3 bits up to 11 bits. For payload sizes between 12 and 22 bits, a dual Reed-Muller code is used (the same block code applied twice), and in case of payload size above 22 bits, a tail biting convolutional code with an 8-bit CRC is used.

Table 9.5

SPUCCH formats.
SPUCCH format Applicability HARQ Ack size [bits] # Resource blocks Frequency hopping
format 1/1a/1b
Subslot
Slot
 1 or 2 1
Yes
Yes or No
format 3 Slot 3–11 1 No
format 4
Subslot
Slot
 ≥3 1,2,3,4,5,6 or 8
No
Yes

image

For all PUCCH formats, up to four time-frequency resources can be configured by RRC signaling. The resource to use is indicated to the device in a 2-bit field in the DCI. The same functionality exists in legacy LTE for PUCCH format 3.
9.2.4.2.2. SPUCCH format 1/1a/1b
9.2.4.2.2.1. Slot
Slot-SPUCCH format 1/1a/1b is similar to PUCCH format 1/1a/1b in that it is based on coherent demodulation of modulated reference signal sequences. In the modulation of the sequences, either BPSK (SPUCCH format 1a) or QPSK (SPUCCH format 1b) is used.
However, in contrast to PUCCH format 1/1a/1b, the configuration of the time-frequency resources to be used is done through RRC signaling (as already mentioned in Section 9.2.4.2.1) and dynamically indicated in DCI which resources to use (see Section 9.3.2.7.1). This can be compared to PUCCH format 1a/1b, where the resources are implicitly given by the CCE index associated with the downlink data transmission.
Slot-SPUCCH format 1/1a/1b can be configured by RRC to either apply frequency hopping or not.
In case frequency hopping is not enabled, the format is identical to a single slot of PUCCH format 1/1a/1b (similar as the design choice of slot-SPUCCH format 3, see Section 9.2.4.2.3), including the use of cyclic shifts (CSs) and OCC to multiplex users.
In case of frequency hopping, the hop takes place after either the third symbol (for the first slot in a subframe) or after the fourth symbol (for the second slot in a subframe). The reason to have different hopping patterns is to align with the uplink subslot pattern, see Figs. 9.23 and 9.2.
How the cyclic shift and OCC-code applied varies over time is different depending on if frequency hopping is used or not. In case frequency hopping is not enabled, the cyclic shift and OCC is generated in the same way as for PUCCH format 1/1a/1b in legacy LTE. This allows users of slot-SPUCCH and PUCCH to be multiplexed as in the legacy case. In case frequency hopping is enabled, how the cyclic shift varies over time is instead taken from PUCCH format 2/2a/2b, as for the case of subslot-SPUCCH (see Section 9.2.4.2.2.2).
9.2.4.2.2.2. Subslot
image
Fig. 9.23 Slot-SPUCCH format 1/1a/1b
The difference lies mainly in how the resources are configured (aligned with the configuration for slot-SPUCCH format 1, see Section 9.2.4.2.2.1) and how the SPUCCH format is detected at the receiver.
As for the detection, the format is based on sequence selection rather than DMRS based coherent demodulation. This means that there is no DMRS symbol used for phase reference/demodulation, but each symbol can instead be independently and non-coherently detected by the receiver.
The same sequence is used as for PUCCH format 1/1a/1b using cyclic shifts (CSs) to separate the sequences. Up to eight CSs can be allocated. The detected CS provides the HARQ feedback   +   (potential) SR information. The reason to use a sequence-based design was mainly to enable frequency hopping between the symbols in a subslot since using a design with coherent demodulation requires one reference symbol sequence and one modulated sequence transmitted transmitted on the same frequency resources to enable demodulation.
As seen from Fig. 9.24, the frequency hop in a 2 OFDM symbol (2 os) SPUCCH format 1/1a/1b design occurs naturally between the two symbols. For a 3 os SPUCCH format 1/1a/1b, the hopping occurs after the first symbol, which allows the 3 os subslot at the end of the subframe (see Fig. 9.2) to still provide frequency diversity, in the case of SRS transmission.
As with legacy PUCCH format 1, there is a cyclic shift variation over time to randomize interference. There is however no OCC applied since in each PRB being transmitted there is only one time-domain symbol, and hence OCC is not possible to apply (the orthogonality is lost in the frequency hop). Since the legacy randomization of cyclic shifts is more complex for PUCCH format 1/1a/1b, the randomization of cyclic shifts follows instead the one used for PUCCH format 2/2a/2b. The consequence of this is that subslot-SPUCCH format 1/1a/1b cannot be easily multiplexed with legacy PUCCH format 1/1a/1b, but instead with PUCCH format 2/2a/2b. As for all (S)PUCCH formats, the format itself can be multiplexed to accommodate different users on the same physical resources by the use of different cyclic shifts.
image
Fig. 9.24 Subslot-SPUCCH format 1/1a/1b.
9.2.4.2.3. SPUCCH format 3
Why is not frequency hopping used in this format? As with other formats, this would allow for improved performance. This is true. However, using frequency hopping would essentially convert slot-SPUCCH format 3 to slot-SPUCCH format 4 (see Section 9.2.4.2.4.2) since OCC can no longer be used, due to the frequency hop. Hence, slot-SPUCCH format 3 is an alternative to slot-SPUCCH format 4 (over the payload ranges where they overlap), where a higher user multiplexing rate is achieved instead of an improved link performance.
image
Fig. 9.25 Slot-SPUCCH format 3.
By using the same design as PUCCH format 3, multiplexing is also achieved with existing PUCCH format 3 users.
9.2.4.2.4. SPUCCH format 4
9.2.4.2.4.1. General
As with PUCCH format 4 in legacy LTE, the slot/subslot-SPUCCH format 4 is similar to slot/subslot-PUSCH. It is based on coherent DMRS-based demodulation with all available bits used to provide an efficient channel coding rather than a high user multiplexing rate (only one user is intended for a given resource). The modulation used is QPSK and the DMRS sequence is generated in the same way as for PUCCH format 4.
The bandwidth allocated to SPUCCH format 4 can also vary depending on RRC configuration and dynamic signaling in the DCI, with allowed PRB allocations of 1,2,3,4,5,6 or 8.
9.2.4.2.4.2. Slot
Slot-SPUCCH format 4 is illustrated in Fig. 9.26. As for slot-SPUCCH format 1/1a/1b, the frequency hopping pattern is dependent on the slot that the SPUCCH is transmitted in. One DMRS symbol is used in each sequence of OFDM symbols sharing the same frequency allocation, to enable coherent demodulation.
image
Fig. 9.26 Slot-SPUCCH format 4.
9.2.4.2.4.3. Subslot
Subslot-SPUCCH format 4 is illustrated in Fig. 9.27. As can be seen, no frequency hopping is used due to the coherent demodulation. A front-loaded DMRS design (the DMRS is placed in the first symbol) is adopted which helps the receiver in performing a channel estimate early in the reception to allow for quick demodulation.
image
Fig. 9.27 Subslot-SPUCCH format 4.

9.2.4.3. Slot/subslot-PUSCH

Subframe Any
Basic TTI Slot/subslot
Repetitions 2, 3, 4 or 6 (SPS)
Subcarrier spacing 15   kHz
Bandwidth See Section 9.3.2.9.2
Frequency location See Section 9.3.2.9.2

The PUSCH is primarily used to transmit unicast data. The data packet from higher layers is segmented into one or more TB, and PUSCH transmits one TB at a time.
PUSCH is also used for transmission of UCI when aperiodic CSI transmission is triggered by setting the CSI request bit in DCI (see Section 9.3.2.7.1 and 9.2.3.10), or, to carry HARQ-ACK bits from PUCCH, in case of collision between PUSCH and PUCCH (and the device does not support, and is not configured with, simultaneous transmission of PUSCH and PUCCH).
As with the PDSCH, there is no new physical shared channel defined for the uplink, instead the transmission duration of the PUSCH is shortened when transmitting subslot-PUSCH or slot-PUSCH. The slot-PUSCH and subslot-PUSCH is inherited from LTE and hence in this section, the differences introduced with sTTI will be the focus. For more details on the PUSCH design for LTE an interested reader is referred to e.g. Ref. [4].
As with PDSCH, the TBS is scaled depending on subslot or slot transmission (see Section 9.2.3.3), using the same scaling factors of α   =   1/2 for slot, and using either α   =   1/12 or α   =   1/6 for subslot depending on if the subslot contains one or two data symbols respectively.
As for subframe-based PUSCH, slot-PUSCH and subslot-PUSCH can be configured with transmission mode 1 and 2, see Table 9.6. The downlink transmission modes for sTTI are configured separately from subframe operation. MIMO is also supported on the uplink with up to four MIMO layers configured. As for PDSCH (see Section 9.2.3.3), a single codeword is used, irrespective of the number of MIMO layers.
The mapping of the PUSCH to physical resources follows the slot boundaries for slot operation and the subslot layout for subslot operation, see Fig. 9.2. PUSCH can however not be mapped to symbols where the DMRS is mapped since the DMRS is transmitted over the full resource allocation, see Section 9.2.4.1.
In case of slot operation in TDD, PUSCH is only supported in UpPTS in special subframe configuration 10, see Fig. 9.18.

Table 9.6

slot/subslot-PUSCH transmission modes.
Transmission mode Description
TM1 Single antenna
TM2 Transmit diversity

9.2.5. Timing advance and processing time

We have already seen how the time to transmit a packet over the air is improved by reducing the actual transmission time (to a subslot or slot operation). Another contributing factor to the overall latency is the processing time at the eNB and device. In case of eNB, the processing time is largely up to implementation, especially with the introduction of asynchronous operation with sTTI, see Section 9.3.2.6. On the device side however, a specified processing timeline needs to be strictly followed. This relates for example to the HARQ-ACK response to a downlink assignment and an uplink data transmission in response to an uplink grant. That is, for the network to properly schedule the uplink resources, and know when and where to expect what in the uplink, a processing timeline is required.
For cMTC services, the processing timeline of the device is an important factor since it determines the number of blind transmissions or number of HARQ retransmissions possible to perform within a given latency bound (see more details in Chapter 10), and is thus an important factor in reaching the 5G URLLC requirements.
The processing timeline is usually expressed in TTIs. For pre-release 15 devices, a processing timeline of n+4 is used. For example, if the downlink assignment is sent in subframe n, the associated HARQ-ACK response will be sent by the device in subframe n+4.
With shortened transmissions, the total amount of data to be processed, for a given bandwidth allocation, roughly scales with the transmission duration. However, other aspects, such as channel estimation will not. One of the more important processing related parameters that is independent of transmission duration is the timing advance, TA.
The timing advance is the advancement in transmission timing performed by the device on the uplink. The advancement is performed for the transmission to be received at the eNB aligned with the downlink frame timing. The amount of timing advance used by the device is set to correspond to twice the propagation time of the signal. This is since the downlink signal that the device synchronizes to, will be delayed by the same amount as the uplink signal received at the eNB. This is illustrated in Fig. 9.28.
The figure illustrates the case of a n+4 timing for two TTI durations (τT), with one being roughly 1/3 of the other one. As can be seen, since the propagation delay (τP) is not dependent on the transmission duration, the time remaining for processing (τProc) by the device is greatly reduced, from roughly 7τP to 1τP.
Hence, reducing the transmission time puts a high requirement on the device with regards to processing time, especially if the timing advance needs to be large. At the same time, allowing a too small timing advance will limit the propagation delay supported, and hence the cell size (including any fiber delay from the digital unit to the radio unit).
The above problem becomes most pronounced in the case of subslot transmission. In this case, the specification solves it by supporting multiple maximum timing advance values, each associated with a different processing timeline. The device can indicate its capability out of two possible processing timeline sets (Set 1 or Set 2). In case of Set 1, the timing advance values are associated with either n+4 or n+6 timing, while for Set 2, the timing advance values are associated with n+6 and n+8 timing, see Fig. 9.29.
image
Fig. 9.28 Timing advance with different transmission duration but the same propagation delay (top figure: long transmission duration; bottom figure: short transmission duration).
image
Fig. 9.29 Timing advance and cell size.
In case of slot operation, the processing timeline is not configurable and always equals n+4 slots.
The cell size supported for different configured transmission times and associated processing timelines is shown in Fig. 9.29.
In case of subslot operation, the processing timeline is configured by RRC. The TA set(s) supported by the device is indicated by capability signaling separately for:
Only a single set, using an associated processing timeline is used by a device in a PUCCH group (a group of carriers using the same (S)PUCCH). Since Set 1 has a more strict timing than Set 2, it is assumed that if Set 1 is supported, the device implicitly also supports Set 2 for that configuration.
In case subslot is configured on the downlink and slot on the uplink (see Section 9.3.2.1) for FDD operation, the lowest maximum timing advance (for slot or subslot) is assumed.
Furthermore, in case subslot is configured on the downlink and slot on the uplink, the processing timeline follows the data direction (with some constraints on the fixed subslot and slot structures), e.g. if data is transmitted in the downlink, the processing timeline is based on the subslot processing timeline of the device. This is shown in Fig. 9.30 for downlink assignment to HARQ-ACK transmission, and in Fig. 9.31 for uplink grant to PUSCH transmission. Since the timing follows the direction of the data, there are three possible timelines for PDSCH, each associated with a different processing timeline, while only a single timeline for the uplink (since slot-transmission only supports n+4 timing).
The minimum processing timelines shown above for subslot and slot operation are for FDD. In case of TDD, the minimum delay of n+4 for slot operation cannot always be fulfilled depending on the restriction in the TDD configuration. This details of this are however not outlined in this book. Interested readers are referred to Ref. [7].
image
Fig. 9.30 Processing timeline for downlink subslot PDSCH to uplink slot HARQ ACK.
image
Fig. 9.31 Processing timeline for downlink subslot and uplink slot and PUSCH grant to PUSCH.

9.3. Idle and connected mode procedures

In this section, the idle and connected mode procedures are described. Since the work on LTE URLLC was mainly concerning user plane latency and reliability in connected mode, only a brief review on the reduced control plane latency to achieve the IMT-2020 requirements is provided. Related to connected mode procedures, the allowed configurations, multiplexing options between control and data, and the handling of collisions between physical channels are some of the aspects handled by this section.

9.3.1. Idle mode procedures

9.3.1.1. Control plane latency

As mentioned in Section 9.1, the main work in the LTE URLLC design is on the connected mode procedures. There is however one exception related to fulfilling the IMT-2020 requirements on the control plane latency, see Section 2.3, where a maximum of 20   ms shall be fulfilled.
Control plane latency is defined by ITU-R [3], as “the transition time from a most “battery efficient” state (e.g. Idle state) to the start of continuous data transfer (e.g. Active state)”. Hence, what needs to be analyzed is the initial access procedure where the device goes from a battery efficient idle mode state to an active state transferring data.
To achieve minimum specification impact, lowering the control plane latency was achieved by changes to device processing times and assumptions on processing times in the network.
A high-level signal diagram of the messages involved in case of the RRC Resume procedure is shown in Fig. 9.32.
As can be seen, the multi-step procedure involves multiple processing steps after receiving the messages transmitted, both at the device and the eNB.
Device processing delay after receiving the Random Access Response has been decreased by 1   ms, from 5   ms to 4   ms.
Also, the processing delay from the reception of the RRC connection resume, and reception of the uplink grant to the transmission of the associated connection resume complete has been reduced, assuming a typical 15   ms–7   ms. It is reasonable to assume that devices evolve with time and hence that some latency reduction can be expected compared to when the procedure was first introduced in Release 13. However, the significant reduction from 15   ms to 7   ms that has been specified comes at a cost, in order for the device to support it. The network is restricted in that the RRC message shall only include MAC and physical layer (re-)configurations and shall not include (re-)configurations of DRX, SPS, MIMO operation or to any secondary cell. Furthermore, the uplink grant of the connection resume complete must be sent on PDCCH DCI format 0 using the Common Search Space.
image
Fig. 9.32 Control plane latency.
The support of a reduced control plane latency is an optional capability that the device can indicate. However, the network will not know that the device is supporting it when the device triggers the RACH preamble transmission. Hence, after detecting the Physical Random Access Channel and giving grant to the RRC connection resume request, the network will have to assume that the device either has 4   ms or 5   ms processing delay and hence allocate resource for both reaction times. In order for the network to control if a shorter processing is allowed to be used by the device or not, the activation of the reduced control plane latency is done via the broadcasted system information.
The resulting control plane latency is evaluated in 10.3.2 where the latency component of each step is outlined.

9.3.2. Connected mode procedures

9.3.2.1. Configurations

The allowed (per cell) configurations for slot and subslot operation are shown in Table 9.7.
The device will be configured to operate an sTTI configuration (i.e. subslot or slot transmission and reception) on a per PUCCH group basis. Hence, it is not possible to be configured with different sTTI lengths on different carriers within the same PUCCH group.
For sTTI (subslot or slot) operation, the network can, on a per-subframe basis, schedule the device interchangeably with sTTI and subframe transmissions. A typical implementation would however not be expected to switch frequently between the two, but e.g. use subframe operation for RRC reconfiguration, and when leaving a TCP slow-start (in which case a fast reaction time, using sTTI is primarily needed to ramp up the TCP throughput). The sTTI operation however, comes with a penalty in overhead both from the reference signals and the control channel, and is not advisable to be used, unless providing a performance benefit compared to subframe operation.

9.3.2.2. Multiplexing of PDSCH and SPDCCH

9.3.2.2.1. General
In legacy LTE, the downlink data (PDSCH) and the associated control (PDCCH) is not multiplexed in the same time domain OFDM symbols but are instead time-domain multiplexed. This changes with the introduction of sTTI. Not allowing any multiplexing of control and data in the same OFDM symbol in this case would severely restrict the data capacity since the control is sent in most of the symbols in the subslot. For some configurations, data transmission would even be prevented (e.g. for DMRS-based control signaling).
The multiplexing of control and data has been implemented on two levels, either by RRC configuration (semi-statically changed) or by signaling in the DCI.

Table 9.7

Allowed combinations of slot and subslot (sTTI) operation.
Downlink Uplink Frame structure type1
Slot Slot FDD, TDD
Subslot Slot FDD
Subslot Subslot FDD

FDD, TDD are in the 3GPP specifications also referred to as Frame Structure Type 1 and 2 respectively.

Irrespective of the configured mode, the device will always rate-match around its own DCI scheduling the downlink data.
9.3.2.2.2. RRC-based multiplexing
Four modes are defined for RRC-based multiplexing of data and control. The different modes are configured per SPDCCH resource set and relate to the device behavior on how the rate-matching for PDSCH is performed in association to the SPDCCH resource set:
It could be expected that different modes are used depending on the traffic situation in the cell. For example, in a light traffic load, when a single device is scheduled in a given TTI, there is no reason to have it rate-match around the whole SPDCCH resource set, but instead to rate-match around its own DCI (RRC mode 1). In a heavier traffic load, where multiple devices are expected to be scheduled, there is no functionality to communicate to a given device which DCIs resources are used by the other devices. In this case, rate-matching around the full set is recommended.
The different rate-matching behaviors are illustrated in Fig. 9.33, where it is assumed that both SPDCCH resource sets are configured with the same rate-matching RRC mode.
As can be seen, the different modes allow for a good level of flexibility. However, the rate-matching behavior cannot be changed frequently, due to the use of RRC signaling.
Hence, it is motivated to introduce a more flexible rate-matching behavior that can follow the momentary traffic behavior. This is solved by including information in the DCI in each sTTI about the rate-matching behavior of the device.
9.3.2.2.3. DCI-based multiplexing
DCI-based multiplexing of data and control can also be used instead of RRC-configured rate-matching.
DCI-based multiplexing is configured by RRC, and the device behavior is associated with three different possible modes.
As with the RRC-based rate-matching, different DCI modes can be used depending on the traffic situation and depending on if one or two SPDCCH resource sets are configured. For example, in case of high traffic load and large variations in the resource allocation over time between the two SPDCCH resource sets, DCI mode 0 could be advisable.
If only one SPDCCH resource set is configured, the use of DCI mode 1 clearly provides most flexibility (top part of Fig. 9.35).
The different modes are illustrated in Figs 9.34–9.36.
image
Fig. 9.34 DCI-based rate-matching mode 0 for sTTI.

9.3.2.3. Scheduling request

As for normal LTE operation, if the device wants to indicate the need for dynamic uplink scheduling from the network, it can send a Scheduling Request (SR). An SR-only transmission is carried by SPUCCH format 1. In case of HARQ and SR multiplexing in the same SPUCCH, the format can be adopted to carry both SR and HARQ bits. The multiplexing capacity of subslot-SPUCCH is reduced compared to PUCCH since no OCC can be used, and hence up to 12 cyclic shifts can be used for multiplexing.
To allow for as low latency as possible, the SR periodicities allowed to be configured to a given device is 1 sTTI (subslot or slot). The allowed range of SR periodicities for subslot operation and slot operation are listed in Tables 9.8 and 9.9 respectively.

9.3.2.4. UCI on PUSCH
image
Fig. 9.35 DCI-based rate-matching mode 1 for sTTI.
In case of slot operation, the mapping of UCI onto PUSCH follows the same principles as for n+4 operation as shown in Fig. 9.37. The PMI/CQI bits related to the precoder and CSI are rate-matched with the data, and so is the rank indication (RI). The HARQ-ACK bits punctures the PUSCH data (instead of using rate-matching) to avoid the case that the downlink assignment associated with the HARQ-ACK is missed. If using rate-matching, the eNB and the device rate-matching would in this case be different (the device assuming HARQ-ACK bits not present). The HARQ-ACK bits are placed close to the DMRS to ensure a good channel estimate, and hence better performance.
image
Fig. 9.36 DCI-based rate-matching mode 2 for sTTI.
In case of subslot operation, the mapping is somewhat different.
For a subslot containing two data symbols, the bits for:
For a subslot containing one data symbols, the bits for:
 

Table 9.8

Scheduling request periodicities – subslot.
Subslot SR periodicity
1 subslot
2 subslots
3 subslots
4 subslots
5 subslots
6 subslots (1   ms)
2   ms
5   ms
10   ms
 

Table 9.9

Scheduling request periodicities – slot.
Slot SR periodicity
1 slot
2 slots (1   ms)
2   ms
5   ms
10   ms
Fig. 9.38 illustrates the different mapping options for subslot-operation.
Furthermore, in case of subslot operation, the beta-offset (a scaling of the baseline code rate for the UCI) is allowed to take two values (configured by RRC) for HARQ-ACK and RI. This is to compensate for the potential degradation that can be experienced due to switching transients at the device. The switching transients stems from RF related imperfections at for example power ramp-up/ramp-down. In the switching transient period, the signal structure is not defined during (in the general case), up to 10   μs, and can thus assume to carry no information. A transient can occur (depending on scheduling) at the start of the first symbol and at the end of the last symbol. In case of a possible transient occurring (known to the network), a higher value of beta-offset can be indicated in the DCI to ensure proper HARQ-ACK reception.
The beta-offset values are independently configured for subframe-, slot- and subslot operation.

9.3.2.5. Subframe and subslot/slot collisions

As described in Section 9.3.2.1, the device can be scheduled with subframe and subslot/slot transmissions interchangeably in different subframes. Furthermore, since the processing timeline between subframe and subslot/slot differ (see Section 9.2.5), the network might schedule uplink transmission of different lengths on the same carrier to be simultaneously transmitted by the device.
image
Fig. 9.37 UCI mapping for slot-PUSCH.
image
Fig. 9.38 UCI mapping for subslot-PUSCH.
image
Fig. 9.39 Overlapping transmissions of different length in the same power amplifier.
However, transmitting two physical channels of different lengths on the same carrier using the same power amplifier will cause imperfections to the signal, including phase discontinuities, that prevents coherent demodulation at the eNB, see Fig. 9.39. Hence, a handling of collisions of physical channel with different duration on the same carrier is needed.
Some general principles are followed:
The different collision cases are described in Table 9.10.
Piggy-backing HARQ-ACK on the shorter channel implies a larger impact on the performance (less energy transmitted but the same number of bits carried). To minimize the negative impact on performance, bundling of the HARQ-ACK bits is supported. Bundling the bits implies that if one or more of the bits being bundled indicates a ‘NACK’, the bundled bit will indicate a ‘NACK’. Bundling is performed over the spatial domain (i.e. the layers transmitted on the same carrier) and is always applied for subslot operation, while being configurable for slot operation.
All above collision cases are concerned with the collision on a given carrier. However, there are cases when different lengths can be transmitted on different carriers. For example, when falling back to subframe operation, when not configuring sTTI on all carriers, or when one PUCCH group is configured with subslot operation and the other PUCCH group with slot operation in the uplink. ‘
In this case, the device can either be capable, or not capable, of transmitting different transmission lengths on different carriers.

9.3.2.6. HARQ

In legacy LTE, synchronous HARQ operation is supported on the uplink with a fixed timing between the initial transmission of PUSCH and its retransmission. In the downlink, asynchronous HARQ is used, i.e. the network does not have to follow a fixed timeline between initial transmission and retransmission.
For sTTI, asynchronous HARQ is adopted both in downlink and in uplink. This also implies that there is no longer a need to involve the PHICH, see Ref. [4].
The total number of HARQ processes have also been increased with sTTI operation, due to the shorter transmission duration, to be able to provide full throughput. A maximum of 16 processes regardless of slot or subslot operation is used.
To allow a smooth transition between sTTI and subframe operation, the HARQ processes are shared. That means that an initial transmission carried out in one transmission duration can be retransmitted using another transmission duration. The support is conditional of that the maximum TBS size for the respective transmission duration is respected.

9.3.2.7. Scheduling

In this section, we describe how scheduling for uplink and downlink transmissions works.
When the network needs to schedule a device, it sends a DCI addressed to the device in one of the PDCCH or SPDCCH candidates (see Section 9.2.3.2 for more details) in the search space that the device monitors. The Cell RNTI, masking the DCI CRC, is used to identify the device. The DCI includes resource allocation (in both time and frequency domains), modulation and coding scheme, and information needed for supporting the HARQ operation.
In case of slot#0 or subslot#0, the device monitors the search space in the PDCCH region, while for slot#1 and subslot#1–5, the SPDCCH search space is used.
As with PDCCH decoding, the more SPDCCH candidates the device need to search for, the more complex the decoding procedure becomes and the higher risk for a false SPDCCH decoding. Some limitations are therefore assumed.
For SPDCCH, the following applies:
For PDCCH, the following applies:
The DCI format for both downlink and uplink have been designed with the same number of payload bits. If the number of used bits for a given format in a given direction does not align with the format in the other direction, padding of bits is applied. Using an aligned DCI format will allow the use of a downlink/uplink flag in the DCI content and reduce the number of blind decodes required (same decoding for downlink and uplink DCI).
9.3.2.7.1. Dynamic downlink scheduling
The base station schedules downlink transmission on PDSCH dynamically using DCI Format 7-1A to 7-1F.
A list of the DCI fields for the baseline format DCI format 7-1A is provided in Table 9.11.
For downlink DCI formats other than DCI format 7-1A, the fields of DCI format 7-1A are included as baseline. Additional fields are included according to Table 9.12.
The other formats include additional information related to MIMO information, which is similar to legacy LTE operation. Interested readers are referred to Refs. [6,7].

Table 9.11

DCI format 7-1A.
Information Size [bits] Possible settings
Flag for DL/UL differentiation 1
0 - format 7-0A/B depending on the configured uplink transmission mode
1 - format 7-1X, where X is the DCI format transmitted
Resource allocation variable Different bit spaces depending on resource allocation type 0 or 2, see Section 9.3.2.9.1
Modulation and coding scheme 5 Modulation used together with an indicative code rate, see Ref. [7]
HARQ process number 4 See Section 9.3.2.6
New data indicator 1 The bit is toggled when indicating to the device that the soft-buffer is to be flushed and that new data is transmitted for the signaled HARQ process. See also [7]
Redundancy version 2 Indicates the set of bits, from the encoded set of bits, selected for transmission. See also [7]
TPC command 2 See Section 9.3.2.8
Downlink Assignment Index (DAI) 2 or 4 Assists the device in understanding the number of downlink assignments being transmitted by the eNB, even though only a subset is detected. Used to align the device's and eNB's understanding in the number of HARQ-ACK bits reported. See also [7]
Used/Unused SPDCCH resource indication 2 See Section 9.3.2.2
SPUCCH resource indication 2 See Section 9.2.4.2
Repetition number 2 See Section 9.2.3.3

Table 9.12

DCI format 7-1B to 7-1G.
Information Size [bits] DCI format 7-1 Possible settings
B C D E F G
Precoding information 1 or 2 •
1 bit for transmission with 2 antenna ports
2 bits for transmission with 4 antenna ports
Precoding information 4 or 6 •
4 bits for 2 antenna ports
6 bits for 4 antenna ports
Precoding information 3 or 5 •
3 bits for 2 antenna ports
5 bits for 4 antenna ports
SRS request 0 or 1 • • • For TDD operation, if the device has indicated the capability and is configured with SRS request
Scrambling identity 1 • See Ref. [6]
Precoding information 2 • One or two layers (with or without transmit diversity)
DMRS position indicator 1 • • See Section 9.2.3.1
Antenna port(s), scrambling identity and number of layers 1 or 3 • • See Ref. [6]
PDSCH RE Mapping and Quasi-Co-Location Indicator 2 • See Ref. [7]

image

9.3.2.7.2. Dynamic uplink scheduling
The base station schedules uplink transmission on PUSCH dynamically using DCI Format 7-0A and 7-0B. DCI format 7-0B is used in case of MIMO transmission.
A list of the DCI fields is provided in Table 9.13.
9.3.2.7.3. Semi-persistent Scheduling
In addition to the dynamic scheduling in uplink and downlink, support has also been added for Semi-Persistent Scheduling (SPS) using slot/subslot transmission to reduce latency. A blind repetition-based scheme, similar to PDSCH (see Section 9.2.3.3), to support improved reliability has also been defined. SPS pre-allocates periodically (re)occurring resources using a configurable interval for the device to monitor and decode (on the downlink) and to transmit in (on the uplink). Using such pre-allocation of resources will for example eliminate the need for sending a SR to trigger scheduling in the uplink (reducing latency) and will also minimize the control overhead for the data transmission.
The use of SPS is also a key enabler for fulfilling the 5G requirements for URLLC as elaborated more upon in Chapter 10.
The SPS operation for sTTI is very similar to subframe-based operation in legacy LTE, see Ref. [4].

Table 9.13

DCI format 7-0A and 7-0B.
Information DCI format 7-0A DCI format 7-0B
Size [bits] Possible settings Size [bits] Possible settings
Flag for UL/DL differentiation 1
0 - format 7-0A
1 - format 7-1A/B/C/D/E/F/G depending on the configured downlink transmission mode
 1
0 - format 7-0B
1 - format 7-1A/B/C/D/E/F/G depending on the configured downlink transmission mode
Resource block assignment variable See Section 9.3.2.9 variable See Section 9.3.2.9
Modulation and coding scheme 5 See Table 9.11 and [7] 5 See Table 9.11 and [7]
HARQ process number 4 See Section 9.3.2.6 4 See Section 9.3.2.6
New data indicator 1 See Table 9.11 and [7] 1 See Table 9.11 and [7]
Redundancy version 2 See Table 9.11 and [7] 2 See Table 9.11 and [7]
TPC command for scheduled PUSCH 2 See Section 9.3.2.8 2 See Section 9.3.2.8
DMRS pattern 2 See Section 9.2.4.1 2 See Section 9.2.4.1
Cyclic shift for DMRS and IFDMA configuration 1 See Section 9.2.4.1 1 See Section 9.2.4.1
UL index 2 In case of TDD, and for certain configurations, the UL index is used to allow scheduling of multiple PUSCHs from the same DCI. See also [7] 2 In case of TDD, and for certain configurations, the UL index is used to allow scheduling of multiple PUSCHs from the same DCI. See also [7]
Downlink Assignment Index (DAI) 2 See Table 9.11 and [7] 2 See Table 9.11 and [7]
CSI request 1, 2 or 3 See Section 9.3.2.10 1, 2 or 3 See Section 9.3.2.10
SRS request 0 or 1 For TDD operation, if the device has indicated the capability and is configured with SRS request 2 For TDD operation, if the device has indicated the capability and is configured with SRS request
Beta offset indicator 1 See Section 9.3.2.4 1 See Section 9.3.2.4
Cyclic shift field mapping table for DMRS 1 See Section 9.2.4.1 1 See Section 9.2.4.1
Precoding information and number of layers – – 3 or 6 See Section 9.2.4.3. Depending on the number of antenna ports at the device (2 or 4)

image

Information that is not carried in the DCI is configured by RRC. The DCI is then dynamically used to activate and release the SPS operation, where for example the selected MCS can be changed dynamically. For the device to understand that the DCI is related to SPS, it is scrambled by SPS-C-RNTI. SPS for sTTI is supported both on the downlink and the uplink.
The device monitors all subslots/slots in the subframe for possible SPS activation and release.
For downlink SPS, some behavior related to DCI operation need to be changed. As described in Section 9.2.4.2, the SPUCCH resources to be used for the HARQ-ACK response is indicated by a 2-bit field, indicating one out of four resources. For SPS operation, the activation of SPS will indicate one of these resources to be used until the SPS operation is reconfigured.
The power control loop for SPS is handled by the same DCI format as in legacy LTE, that is, DCI format 3/3A, using a processing timeline (see Section 9.2.5) of n+4 (subframes). The TPC-index configured to a device is separate for sTTI and subframe operation. As for PUSCH and PUCCH power control (see Section 9.3.2.8), the power control loops are independent for sTTI and subframe operation.
In contrast to dynamic downlink scheduling, there is no support for DMRS sharing across subslots in the downlink (see Section 9.2.3.3 and 9.3.2.7.1) when SPS is configured.
However, on the uplink where the overhead from DMRS can be up to 50% (in case of two-symbol subslot with one DMRS symbol), there are two configurations supported, one with DMRS contained in each subslot and one where the associated DMRS can be placed outside of the subslot boundary. The two possible configurations are shown in Table 9.14 (which can be compared to Table 9.4 used in case of dynamic uplink scheduling).
To allow for ultra-low latency, the SPS periodicity can be configured as low as 1 subslot or 1 slot for subslot and slot operation, respectively. The possible configurations are shown in Table 9.15.
Another difference to the legacy LTE operation is that the SPS configuration can be associated with a DMRS using a configurable cyclic shift and potentially IFDMA modulation (see Section 9.2.4.1). This allows users to be simultaneously multiplexed on the same physical resources, using Multi User MIMO, which compensates for the increase in resources used in sTTI, where the frequency allocation is typically larger, and for cMTC services, the SPS periodicity is typically shorter. However, for cMTC services, it should be noted that increasing the user multiplexing rate on the same physical resources will have an implication in the reliability achieved.
To improve reliability in uplink SPS operation, there is a possibility of configuring a device to transmit blind repetitions for subframe, slot and subslot operation. The number of repetitions possible to configure for uplink SPS repetitions are 2, 3, 4 or 6. To ensure a low waiting time from the time the packet is delivered to lower layers until it can be transmitted over the air, multiple SPS configurations can be used, each with a different starting position of the repeated PUSCH sequence. Up to six configurations can be used. The total number of transmission (the initial transmission and all repetitions) cannot exceed the periodicity configured since that would result in overlapping SPS transmissions.

Table 9.14

Uplink SPS DMRS configurations.
Bit value Subslot#0 Subslot#1 Subslot#2 Subslot#3 Subslot#4 Subslot#5
0 R D D R D R D R D R D R D D
1 R D D D D |R R D D D |R R D R D D

image

‘R denotes a DMRS symbol, ‘D’ denotes a data symbol, ‘|’ denotes the subslot boundary.

Table 9.15

Downlink and uplink SPS periodicities.
sTTI SPS periodicities
sTTI [#] Subslot a operation [ms] Slot operation [ms]
1 0.2 0.5
2 0.4 1.0
3 0.5 1.5
4 0.7 2.0
6 1.0 3.0
8 1.4 4.0
12 2.0 6.0
16 2.7 8.0
20 3.4 10.0
40 6.7 20.0
60 10.0 30.0
80 13.0 40.0
120 20.0 60.0
240 40.0 120.0

image

An example is shown in Fig. 9.40, where four transmissions are used for each configuration (one initial transmission and three repetitions). Each configuration has a periodicity of four (s)TTIs and they each have different starting positions for the sequence of repetitions. In order to minimize the delay of the packet arriving, configuration#1 should be selected for the uplink transmission.

9.3.2.8. Uplink power control

The uplink power control for PUSCH and SPUCCH follow very similar principles as in legacy LTE operation.
9.3.2.8.1. PUSCH
image
Fig. 9.40 Multiple uplink SPS configurations.
P PUSCH = min { P CMAX 10 l o g 10 ( M PUSCH ) + P 0 + α · P L + Δ TF + f
image (9.5)
The closed loop parameter ( f ) image controlled by the TPC command in the DCI will only impact the power control loop related to subslot/slot transmission. That is, it will not have an impact to the power control loop used for subframe-based transmission, and hence they are independent.
The power headroom report (PHR), providing a rough indication to the network on the difference in power between the used transmit power and the maximum allowed power, is also similar to regular LTE operation.
A PHR can be transmitted by either subframe-based PUSCH or subslot/slot-PUSCH. The PHR is reported for all activated carriers. Both Type 1 and Type 2 PHR reporting is supported. As in legacy LTE, in case of.
Furthermore, the power headroom for a given carrier can be based on either a scheduled transmission or non-scheduled transmission. In case the transmission is not scheduled, it is assumed that the resource allocation is one resource block (since the allocation is not known) and that there is no MCS-specific (since not known) contribution in the power estimation.
In case the PHR is transmitted on subframe-based PUSCH, the PHR is subframe-based for all carriers, regardless if the carrier is configured with sTTI or not.
In case the PHR is transmitted on subslot/slot-PUSCH, carriers not configured with sTTI follow the subframe-based PHR reporting, while in case the carrier is configured with sTTI, the power headroom is calculated based on sTTI.
9.3.2.8.2. SPUCCH
P PUCCH F 1 / 3 = min { P CMAX P 0 + P L + Δ T x D + g + Δ F _ PUCCH + h
image (9.6)
P PUCCH F 4 = min { P CMAX P 0 + P L + Δ T x D + g + Δ F _ PUCCH + 10 log 10 ( M PUSCH ) + Δ TF
image (9.7)
However, to capture the difference in performance between PUCCH and SPUCCH, one of the parameters need to be modified, and this is done through the SPUCCH format dependent offset ( Δ F _ P U C C H image ). A specific offset can be configured (within a specified range) for each subslot-SPUCCH/slot-SPUCCH transmission.
As for PUSCH, the closed loop power control parameter ( g image ) updated through TPC in the DCI is independently applied for the PUCCH and the SPUCCH power control loops.

9.3.2.9. Resource allocation

9.3.2.9.1. Downlink
In the downlink, resource allocation type 0 and 1 is supported for subslot and slot transmissions. One of the two resource allocations is configured by RRC.
If the RBG is not a multiple of the system bandwidth, the size of the last RGB is increased to avoid unscheduled resources.
9.3.2.9.2. Uplink

9.3.2.10. CSI reporting

In legacy LTE operation, the Channel State Information (CSI) reporting from the device is done by performing measuring on the downlink resources, for the network to be able to apply channel dependent scheduling.
The CSI is either reported periodically or aperiodically, carried by PUCCH or PUSCH.
For sTTI operation, only aperiodic CSI reporting is supported, carried by PUSCH.
The triggering of CSI reporting from the DCI (see Section 9.3.2.7.1) follows the same processing timeline as subslot/slot-PUSCH and the downlink reference resources where the device is measuring is based on slot/subslot (depending on what is configured).
For more details on CSI reporting, the reader is referred to Ref. [4].

9.3.2.11. PDCP duplication