5G Fundamentals : Functional Split Overview High Layer Split Low Layer Split (6 & 7)

7-2x (O-RAN)
3GPP
TR 38.801 1 2 3 4 5 6 7-3 7-2 7-1 8
Split Naming

(DL only)
Overview

eCPRI v1 A B C D ID IID E
Small Cell Forum PDCP-RLC RLC-MAC Split MAC MAC-PHY (nFAPI) II III IIIb
NGMN MAC-PHY I II III ~IV

OSI Layer 2 - Data Link OSI Layer 1 - Physical

Protocol
Stack

Logical Link Control (LLC) - OSI Layer 2(b) MAC - OSI Layer 2(a)

High Low MAC High PHY Low PHY RF

IP PDCP High RLC Low RLC
MAC

The lower the split option, the less Retrans-

RLC The higher the split option, the
Functional

mission Controller Scheduler

functions are in the CU. buffer
control less functions are in the RU.
View

Resource

Antenna N
Transport

Transport
Retrans- Header Add Trans­ DL-SCH Cyclic Analog

Symbols

Symbols
symbols

Antenna
symbols
Symbols

Layer N

carriers

carriers
Blocks

Blocks

Coded
Num­ Ciphering Segmen- Add RLC Multi­ CRC Coding + Rate Scram- Modula- Layer Pre- Element Beamf.
PDCP

Code-
words
Block

Code-
words
MAC

ports
SDU

SDU

SDU

Sub­

Sub­
RLC
Data mission com­ PDCP mission HARQ data iFFT Prefix Conver- Downlink

IQ

IQ
bering tation header plexing attach block seg matching bling tion mapper coding Mapper Port exp.
buffer pression header buffer transfer Insertion sion

Latency and Jitter

Req.
Lat.-

10ms Latency tolerant 1.5 - 10ms ~100µs HARQ Loop - Very tight latency requirements 250µs constrained
Scala­
bility

n.a. Scales with MIMO layers Scales with antenna port

(LTE / Gpbs)
Band-
width

UL DL UL DL UL DL UL DL UL DL UL DL UL DL UL DL UL DL UL DL UL DL
Based on 100MHz, 8 layers, 256 QAM Based on 100MHz, 32 Antennas
3 4 3 4 3 4 4.5 5.2 7.1 5.6 7.1 5.6 15.2 9.8 15.2 9.8 15.2 9.8 60.4 9.2 157.3 157.3

Option 8
Backhaul CU/DU Fronthaul RU CPRI

Low Layer Split

Option 7-2x
Backhaul CU/DU Fronthaul RU O-RAN

Option 7
Backhaul CU/DU Fronthaul RU eCPRI IID
Core Network

Option 6
Backhaul CU/DU Fronthaul RU nFAPI
CU and DU Mapping

High Layer Split

Backhaul CU Midhaul DU/RU Option 2

Double Split
Backhaul CU Midhaul DU Fronthaul RU Option 2 & 6

Backhaul CU Midhaul DU Fronthaul RU Option 2 & 7.2

High Layer Split Double Split Low Layer Split

+ Drastically reduced Bandwidth - CoMP extremely complex or even impossible + CU can easily be virtualized - High bandwidth and latency fronthaul + Ideal for CoMP = mobile - High Bandwidth
+ Ideal for non-mobile = FWA - Complex and expensive RRH (size, heat, cost) + Optimal for mobile and URLLC requirements + Cost effective RRH - Bandwidth scales with antenna ports (8, 7-1)
+ Latency Tolerant = long distances + Reduced cost - Very tight latency requirements
+ Processing in RRH = URLLC + Good scalability

Split 1 2 3 4 5 6 7-3 7-2 7-2x 7-1 8

+ Low bandwidth requirements. + Fundamentals for achieving a PD- + Very Low bandwidth requirements. + Low bandwidth requirements. + Low bandwidth requirements. + Bitrate scales with MIMO layers + Bitrate scales with MIMO layers + Bitrate scales with MIMO layers + Simplified interface + The required bitrate is more than half of + Small and cost effective RU.
+ Bitrate scales with MIMO layers. CP-RLC split have already been stan- + Low latency requirements. + Bitrate scales with MIMO layers. + Reduced latency requirements if HARQ + Significant bandwidth reduction com- + Reduced bandwidth requirements com- + Reduced bandwidth requirements com- + Open interface protocol specifically split option 8. + Easy to centralize CU/DU enabling co-
+ Separate User Plane and centralized dardized for LTE Dual Connectivity.* + More robust under non-ideal transport processing and cell-specific MAC func- pared to split option 7-3. pared to split option 7-1. pared to split option 7-1. designed to enable interoperability + Coordinated multi-point schemes are ordinated multi-point (CoMP) schemes.*
RRC/RRM.* + The 2-2 option enables centralization of conditions.* tionalities are performed in DU.* + Joint Transmission is possible.* + Coordinated multi-point schemes are + Coordinated multi-point schemes are between RUs and DUs from different possible if CU/DU are colocated.* + Majority of processing can be central-
the PDCP layer.* + vendors. + Transmit and receive joint processing is
Pros

+ It may in some circumstances provide + Possibility of reduced processing and Efficient interference management + Centralized scheduling is possible.* possible if CU/DU are colocated.* possible if CU/DU are colocated.* ized at a BBU hotel or CU-pool.*
benefits in handling some edge comput- + Option 2-2 allows a separate UP and a buffer requirements in DU.* across multiple cells and enhanced + Allows resource pooling for layers in- + Transmit and receive joint processing is + Transmit and receive joint processing is + Bitrate scales with MIMO layers possible.* + RUs can be used for different genera-
ing or low latency use cases where the centralized RRC/RRM.* + In option 3-2 Rx RLC is placed in the scheduling technologies such as CoMP, cluding and above MAC.* possible.* possible.* + Reduced bandwidth requirements com- tions of RAT (GSM, 3G, 4G)
user data needs to be located close to CU, there is no additional transmission CA, etc.* pared to split option 7-1.
the transmission point.* delay of PDCP/RLC reestablishment
procedures.*

- Very complex and expensive DU/RU. - Coordination of security configurations - Split 3-1 is more latency sensitive than - No benefits for LTE.* - Complex interface between CU and - May require subframe-level timing inter- - High bandwidth requirements. - High bandwidth requirements. - High bandwidth requirements. - Still relatively high bandwidth - Highest bandwidth requirements of all
December 2019

- It‘s not clear if this option can support between different PDCP instances for 3-2 due to the ARQ in CU and not DU.* DU.* actions between MAC layer in CU and - Relatively high latency requirements - Relatively high latency requirements - Relatively high latency requirements. requirement especially for the uplink. functional split options.
aggregation based on alternative 3C.* Option 2-2 required.* - Difficulty in defining scheduling opera- PHY layers in DUs.* - Complex timing for RU and CU/DU link.* - Complex timing for RU and CU/DU link.* - Bandwidth scales with number of RUs.* - Bandwidth scales with number of RUs.*
tions over CU and DU.* - Round trip fronthaul delay may affect - -
Cons

Very latency constrained. Very latency and jitter constrained.

- Limitations for some CoMP schemes.* HARQ timing and scheduling.* - Complex timing for RU and CU/DU link.* - Distance between RU and DU/CU limit-
ed to ~20km due to latency constraint.
- Interoperability between radio equip-
ment vendors not specified

> Best suited for low latency and/or edge > Suited for high layer split between CU > Low bitrate and latency insensitive > No specific advantage for use cases. > Ideal for scenarios where distances > Ideal for small cell deployments. > Suited for setup with limited fiber capac- > Current 5G eCPRI radios use this split > Ideally suited for virtualized RAN and > High fiber capacity available between > High fiber capacity available between
Rev. 3

computing scenarios. and DU. Very latency tolerant enabling midhaul connections between CU and greater than 20km between DU and CU ity in the fronthaul. option. virtual DU running on general purpose radio and centralized location. radio and centralised location.
Cases
Use

distances up to 40km. DU with non-ideal transport conditions.* need to be bridged. processing platforms. > Real time communication applications.
> Possible to integrate in Ethernet based
networks using Radio over Ethernet.

5G Fundamentals Functional Split Overview

* 3GPP TR 38.801 V14.0.0 (2017-03): „Study on new radio access technology: Radio access architecture and interfaces.”
www.cubeoptics.com