A Bandwidth Part (BWP) is a contiguous set of physical resource blocks (PRBs) on a given carrier. These RBs are selected from a contiguous subset of the common resource blocks for a given numerology (u). It is denoted by BWP. Each BWP defined for a numerology can have following three different parameters.
- Subcarrier spacing
- Symbol duration
- Cyclic prefix (CP) length
BWP Configuration Properties
- UE can be configured with maximum 4 BWP for Downlink and Uplink but at a given point of time only one BWP is active for downlink and one for uplink.
- BWP concept enable UEs to operate in narrow bandwidth and when user demands more data (bursty traffic) it can inform gNB to enable wider bandwidth.
- When gNB configures a BWP , it includes parameters: BWP Numerology (u) BWP bandwidth size Frequency location (NR-ARFCN), CORESET (Control Resource Set)
- With respect to Downlink ,UE is not expected to receive PDSCH, PDCCH, CSI-RS, or TRS outside an active bandwidth part
- Each DL BWP include at least one CORESET with UE Specific Search Space (USS) while Primary carrier at least one of the configured DL BWPs includes one CORESET with common search space (CSS)
- With respect to uplink, UE shall not transmit PUSCH or PUCCH outside an active bandwidth part
- UEs are expected to receive and transmit only within the frequency range configured for the active BWPs with the associated numerologies. However, there are exceptions; a UE may perform Radio Resource Management (RRM) measurement or transmit sounding reference signal (SRS) outside of its active BWP via measurement gap
BWP Activation/Deactivation and Switching
According to 38.321-5.15 Bandwidth Part (BWP) operation, BWP selection (or BWP switching) can be done by several different ways as listed below.
- Dedicated RRC Signaling
- Over PDCCH channel Downlink control information (DCI)- DCI 0_1 (UL Grant) and DCI 1_0 (DL Scheduling)
- By bwp-inactivityTimer – ServingCellConfig.bwp-InactivityTimer
- By MAC CE (Control Element)
DCI based mechanism, although more prompt than the one based on MAC CE, requires additional consideration for error case handling, i.e. the case when a UE fails to decode the DCI containing the BWP activation/deactivation command. To help to recover from such a DCI lost scanarios, the activation/deactivation of DL BWP (or DL/UL BWP pair for the case of unpaired spectrum) by means of timer (bwp-inactivityTimer) is also introduced. With this mechanism, if a UE is not scheduled for a certain amount of time, i.e. expiration of timer, the UE switches its active DL BWP (or DL/UL BWP pair) to the default one.
There is an initial active BWP for a UE during the initial access until the UE is explicitly configured with BWPs during or after RRC connection establishment. The initial active BWP is the default BWP, unless configured otherwise.
As per 3GPP Release 15, for a UE, there is at most one active DL BWP and at most one active UL BWP. The HARQ retransmission across different BWPs is supported when a UE’s active BWP is switched.
Why BWP is Required?
A wider Bandwidth has direct impact on the peak and user experienced data rates, however users are not always demanding high data rate. The use of wide BW may imply higher idling power consumption both from RF and baseband signal processing perspectives. In regards to this , new concept of BWP has been introduced for 5G-NR provides a means of operating UEs with smaller BW than the configured CBW, which makes NR an energy efficient solution despite the support of wideband operation.
Alternatively, one may consider to schedule a UE such that it only transmits or receives within a certain frequency range. Compared to this approach, the difference with BWP is that the UE is not required to transmit or receive outside of the configured frequency range of the active BWP, which attributes power saving from the following aspects:
- BWP concept reduce bandwidth processing requirement to transmit or receive narrow bandwidth
- BWP enable RF-Baseband interface operation with a lower sampling rates
- UE RF bandwidth adaptation can provide UE power saving at least if carrier bandwidth before adaptation is large.
BWP Allocation Types
Figure below represents the different BWPs types available for a UE. Considering typical use cases, Idle Mode BWP is smaller than Connected Mode BWPs.
Three types of BWP are available:
- Initial BWP
- Active BWP (UE Specific)
- Default BWP (UE Specific)
Initial BWP is used to performs Initial Access Process. It includes Parameters like RMSI (Requested Minimum System Information), CORESET* and RMSI Frequency location/bandwidth/SCS. It can be 24~96 PRBs with different settings and relaxed to wider BWP after RMSI decoding.
Active BWP is defined as UE specfic can also be used to BWP performs Initial Access Process. It is the first BWP where UE starts data transfer after RRC configuration/reconfiguration. The very first Active BWP should be different from the default BWP.
Default BWP is again UE specific BWP and configured during RRC reconfiguration, if it not configured then it can be assumed that Intial BWP is the default BWP. Every UE would switch back to default BWP when BWP timer expires.
Bandwidth Parts Operations during Initial Access
The BWP parameters are used to configure the mobile operator between the UE and the cell. According to 3GPP TS
38.331 for each serving cell the network configures at least an initial bandwidth part, comprising of downlink
bandwidth part and one (if the serving cell is configured with an uplink) or two (if using supplementary uplink
– SUL) uplink bandwidth parts. Furthermore, the network may configure additional uplink and downlink
bandwidth parts.
The bandwidth part configuration is split into uplink and downlink parameters as well as into common and
dedicated parameters. Common parameters (in BWP-UplinkCommon and BWP-DownlinkCommon) are “cell
specific” and the network ensures the necessary alignment with corresponding parameters of other UEs. The
common parameters of the initial bandwidth part of the PCell are also provided via system information. For all
other serving cells, the network provides the common parameters via dedicated signaling.
| Step | Stage | DL BWP | UL BWP | Processing |
|---|---|---|---|---|
| 0 | PSS and SSS Decode | DL Synchronization | ||
| 1 | MIB decode | UE decode MIB and get CORESET #0 configuration | ||
| 2 | RMSI decode | CORESET #0 | Get Initial DL-BWP and Initial UL-BWP setting for RMSI decoding | |
| 3 | Msg-1-UE >——> gNB | Initial UL-BWP | Random Access Request to gNB | |
| 4 | Msg-2-UE <—–< gNB | CORESET #0 | Random Access Response (RAR) gNB | |
| 5 | Msg-3-UE >——> gNB | Initial UL-BWP | RRC connection request | |
| 6 | Msg-4-UE <—–< gNB | CORESET #0 | RRC connection setup Configure UE specific BWP (default/1st active/ other) BWP If not configured, still use initial BWP | |
| 7 | Msg-5-UE >——> gNB | 1st Active BWP | 1st Active BWP | RRC set-up completed Initial BWP is the 1st Active BWP if no additional configuration carried in Msg4 |
BWP Activation/Deactivation and Switching
The traffic patterns within one active data session can change frequently as the data rate may increase or decrease based on the type of service or the user behavior (accessing the internet and answering a phone call for example). It becomes very important to quick switch between different bandwidth parts to manage different power consumption for different data rates.
According to TS 38.321 BWP selection and switching can be done with different mechanisms as listed below:
- RRC-Based Adaptation: It is more suitable for semi-static cases since the processing of RRC messages requires extra time, letting the latency reach ~10 msec. Due to longer switching latency and signaling overhead, a RRC-based method can be used for configuring a BWP set at any stage of the call, or for slow adaptation type services (e.g., voice) where the resource allocation is not changing rapidly within the same data session.
- MAC CE (Control element):It is used upon initiation of Random Access procedure
- DCI-Based Adaptation: It is based on PDCCH channel where a specific BWP can be activated by BWP indicator in DCI Format 0_1 (UL Grant) and Format 1-1 (DL scheduling). This method better fits on-thefly BWP switching as using this method the latency is as low as 2 msec. However, this method requires additional considerations for error handling as UE may fail to decode the DCI with BWP activation/deactivation command.
- Timer-Based implicit fallback to default BWP is a mechanism designed to mitigate possible DCI errors. If the UE is not explicitly scheduled with a BWP after the timer expires, it will automatically switch to the default BWP.
In 4G/LTE, UEs support the maximum possible bandwidth of 20MHz. In 5G, transmission can go up to 400MHz per carrier. It's impractical to expect every UE to support such a high bandwidth. Therefore by design, it's possible for a 5G UE to communicate on a bandwidth smaller than the cell's channel bandwidth. This smaller portion is what's called Bandwidth Part (BWP).
Via RRC signalling, a UE is configured with multiple BWPs, in downlink and uplink. At PHY layer, the network dynamically activates a BWP for transmission or reception. Through such dynamic adaptation, BWPs allow a 5G system to use radio resources optimally to suit current needs.
In 4G and 5G, carrier aggregation is possible, which allows UEs to aggregate bandwidth across carriers. This article is concerned with only bandwidth parts configured on a single carrier.
Discussion
What are the benefits of having bandwidth parts in 5G NR?
What are some allocation scenarios of BWPs? 
A BWP is a contiguous set of Resource Blocks (RBs). It starts at a common RB and spans a specified number of RBs. Numerology, which determines sub-carrier spacing and cyclic prefix, is also an essential BWP configuration.
Some allocation scenarios are illustrated in the figure. In the simplest case of (a), a reduced BWP is configured for a UE of a lower bandwidth capability. Scenario (b) is useful for a UE having bursty traffic. When more data is to be sent, BWP2 is used. Note that even if BWP1 and BWP2 overlap, only one of them active at a time.
Scenario (c) shows two BWPs with different numerology, each meeting different service requirements. While Physical Resource Blocks (PRBs) of a BWP are all contiguous, there's no requirement that two BWPs have to be contiguous. This is apparent in scenario (d) where other services can be introduced between BWP1 and BWP2, although this option is not part of Release 15.
What are some technical details about BWP?
A UE is configured with multiple BWPs but only one is active at a time. Source: Swamy 2019.
A UE can be configured with a maximum of four BWPs in downlink and another four in uplink. This is in addition to the initial BWPs configured via SIB1. Like UL, there's also Supplementary Uplink (SUL). UE can have four BWPs in SUL.
Even with multiple configured BWPs, only one is active at any one time; that is, UE transmits and receives within its active BWP and nowhere else. DL PDSCH/PDCCH/CSI-RS are received only within the active DL BWP but UE can use measurement gaps to perform measurements outside the active BWP. UL PUSCH/PUCCH are sent by UE only with the active UL BWP.
BWP switching means deactivating the currently active BWP and activating another configured BWP. In TDD, DL and UL BWPs differ only by the transmission bandwidth and numerology; and they're switched together.
There's also default BWP configured for DL and UL. If not configured, initial BWP is used as default. Default is used when there's not much to send/receive to/from the UE. It's activated when an inactivity timer expires.
How does a UE use BWP in RRC idle mode and RRC connected mode?
Illustrating BWP usage in RRC idle mode and connected mode. Source: Line t al. 2020, fig. 2.
A UE's access to the network starts with acquiring the Synchronization Signal Block (SSB) that consists of PSS, SSS and PBCH. This spans 4 OFDM symbols and 20 RBs. It contains the MIB.
MIB contains CORESET#0 configuration. This is used by UE to infer the initial DL BWP. UE receives and decodes the CORESET#0, which contains SIB1. SIB1 sets the initial BWP for both DL and UL. Initial BWP is named BWP#0. DL BWP#0 is configured such that it encompasses CORESET#0.
RACH access happens with UL BWP#0. Network responds with DL BWP#0 until RRC connection happens. Once RRC connection happens, UE can be configured with UE-specific BWPs.
The figure shows BWP#0 (24 RBs), BWP#1 (270 RBs) and default DL BWP#2 (52 RBs). This is an FDD example: DL switches to BWP#2 but UL stays at BWP#1. In TDD, BWP switching happens together for both DL and UL.
How does BWP adaptation or switching happen?
DCI-based bandwidth part switching. Source: Lin et al. 2020, table 1.
When a UE moves from idle mode to RRC connected mode, RRC signalling can configure UE-specific BWPs. RRC configuration or reconfiguration message may specify one of these to be activated. If so, the UE will do BWP switching. Due to RRC processing delay, this can be in the order of tens of milliseconds.
Once UE is configured with multiple BWPs, network can command UE to switch BWP using Downlink Control Information (DCI) in PDCCH. DCI format 1_1 for downlink assignment and format 0_1 for uplink grant are used. These formats contain the BWP indicator that can take 1 or 2 bits. If more than 2 BWPs are configured, 2-bit indicator is used.
The third way of switching is when BWP inactivity timer expires, which triggers a switch to default BWP. The timer ranges from 2-2560ms. It's maximum value relates to DRX inactivity timer.
What's CORESET and how is it relevant to BWP?
A CORESET lies within a BWP. Source: Saini 2019.
Control Resource Set (CORESET) is where the UE searches for downlink control signals. Like BWP, it's smaller than the carrier bandwidth. A CORESET can be anywhere but a UE is expected to process only CORESETs that are within its active BWPs. CORESETs are configured at cell level so that the configuration can be reused for any applicable BWP.
CORESET is where UE searches for PDCCH, though the network doesn't necessarily transmits PDCCH on every CORESET. Whereas in LTE, control region spans the entire carrier bandwidth, 5G NR optimizes this via CORESET.
LTE control region can vary and is specified by PCFICH. In 5G NR, CORESET size is configured via RRC signalling. CORESET spans up to three OFDM symbols. CORESET at the start of the slot facilitates scheduling decisions. CORESET at other places may be useful to reduce latency. In the frequency domain, CORESET is in multiple of six RBs.
A BWP can have up to three CORESETs. CORESETs are common or UE specific. Configured via MIB, CORESET#0 is used for SIB1 scheduling. After RRC connection, UE-specific CORESETs may be configured.
Compare 20MHz of LTE versus 100MHz (FR1) of 5G. Apart from the higher bandwidth, 5G's higher sub-carrier spacing translates to lower symbol duration, higher clock speeds and therefore higher power consumption. In FR2 mmWave spectrum, power consumption increases further due to antenna arrays and other RF components. At lower data rates, 5G at 100MHz has a lower power efficiency compared to 4G. The use of BWPs overcomes this.
Allocating a single bandwidth to a UE is also not the best use of radio resources. Bandwidth Part allows for dynamic adaptation. For example, consider three service requirements: eMBB/100Mbps/1ms, eMBB/15Mbps/0.5ms, URLLC/7Mbps/0.25ms. To meet these, a UE is configured with three BWPs, each with a different numerology, MIMO configuration, modulation, and so on. For example, B1 has more resource blocks and bandwidth to achieve 100Mbps; B3 has smaller bandwidth but gives lower latency due to a higher numerology.
