Inferensys

Glossary

Bandwidth Part (BWP)

A contiguous subset of the carrier resource blocks configured in 5G NR to enable bandwidth adaptation and power saving for user equipment with varying capability and throughput requirements.
Stylish WeWork-like workspace with hot desks and document wall, professional searching through enterprise knowledge base on a mounted ultrawide display, warm industrial pendants overhead.
5G NR RESOURCE ADAPTATION

What is Bandwidth Part (BWP)?

A Bandwidth Part (BWP) is a contiguous subset of carrier resource blocks configured in 5G NR to enable bandwidth adaptation and power saving for user equipment with varying capability and throughput requirements.

A Bandwidth Part (BWP) is a contiguous set of physical resource blocks (PRBs) on a 5G NR carrier, defined by a specific subcarrier spacing and cyclic prefix. It allows a User Equipment (UE) to operate on a narrower bandwidth than the full carrier, reducing power consumption and accommodating devices with limited baseband processing capability.

Multiple BWPs can be configured per serving cell, but only one active downlink and one active uplink BWP operate at a time. Dynamic switching is controlled via downlink control information (DCI), radio resource control (RRC) signaling, or inactivity timers, enabling seamless adaptation between high-throughput wideband operation and energy-efficient narrowband monitoring.

5G NR ADAPTATION MECHANISM

Key Characteristics of Bandwidth Parts

Bandwidth Parts (BWPs) are a fundamental 5G New Radio feature enabling dynamic adaptation of the UE's operating bandwidth. They allow a single carrier to serve devices with vastly different capabilities and throughput needs while optimizing power consumption.

01

Contiguous Resource Block Subset

A BWP is defined as a contiguous set of common resource blocks (CRBs) within a carrier's full channel bandwidth. Each BWP is configured with its own numerology (subcarrier spacing and cyclic prefix), allowing a single UE to switch between different OFDM parameter sets. A UE can be configured with up to four BWPs per serving cell in the downlink and uplink, but only one BWP is active at any given time per transmission direction.

02

Bandwidth Adaptation for Power Saving

The primary motivation for BWPs is UE power efficiency. A device can operate on a narrow BWP during low-activity periods for control channel monitoring, then dynamically switch to a wide BWP for high-throughput data bursts. This reduces the RF front-end bandwidth and baseband processing requirements, directly lowering power consumption. The adaptation is controlled via DCI-based switching, inactivity timers, or RRC reconfiguration.

03

Numerology Flexibility Per BWP

Each BWP is independently configured with a specific subcarrier spacing (SCS) and cyclic prefix type. This enables mixed-numerology operation on the same carrier:

  • Low-band BWP: 15 kHz SCS for wide-area coverage
  • Mid-band BWP: 30 kHz SCS for balanced throughput
  • High-band BWP: 60 kHz or 120 kHz SCS for low-latency URLLC This flexibility is critical for supporting diverse 5G use cases simultaneously on a single carrier.
04

BWP Switching Mechanisms

5G NR defines three methods for BWP transition:

  • DCI-based switching: The gNB signals a BWP ID in the scheduling DCI, triggering an immediate switch for data reception
  • Inactivity timer: The UE automatically falls back to a default BWP after a configured period without data activity
  • RRC reconfiguration: Higher-layer signaling changes the active BWP for long-term adaptation The switching delay is bounded by slot-level timing requirements specified in 3GPP TS 38.133.
05

Initial and Default BWP Configuration

Every UE is assigned an initial BWP for the random access procedure and initial connection setup. Additionally, a default BWP can be configured as a fallback narrowband resource. If the inactivity timer expires, the UE returns to the default BWP, ensuring that devices do not remain on wide bandwidths unnecessarily. The initial BWP is broadcast in SIB1, while dedicated BWPs are configured via UE-specific RRC signaling.

06

BWP and CORESET Relationship

Each BWP contains its own set of Control Resource Sets (CORESETs) and associated search spaces for PDCCH monitoring. This means the control channel configuration is BWP-specific:

  • A narrow BWP may have a single CORESET for minimal monitoring
  • A wide BWP can configure multiple CORESETs for multi-beam operation The UE's PDCCH blind decoding budget is managed per BWP, aligning control overhead with the active bandwidth.
BANDWIDTH PART (BWP) IN 5G NR

Frequently Asked Questions

Clarifying the mechanics, configuration, and operational advantages of Bandwidth Parts in 5G New Radio for adaptive spectrum utilization.

A Bandwidth Part (BWP) is a contiguous subset of the carrier resource blocks configured for a User Equipment (UE) on a 5G NR component carrier. It defines the operating bandwidth and numerology for the UE's active downlink and uplink transmissions. Unlike LTE, where a UE must support the entire carrier bandwidth, BWP enables bandwidth adaptation, allowing a device with limited capability to operate on a wideband carrier by tuning to a narrower frequency segment. Each BWP has its own subcarrier spacing (SCS) and cyclic prefix configuration, defined by a specific numerology index. A UE can be configured with up to four dedicated BWPs in the downlink and four in the uplink, but only one BWP per direction is active at any given time. The active BWP is switched dynamically via Downlink Control Information (DCI) or timers, enabling rapid transitions between wideband operation for high throughput and narrowband operation for power saving.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.