Backhaul is the transport segment that connects cell sites and other access nodes to the core network, carrying user traffic and signaling between the radio access network and the operator’s switching, routing and internet gateway functions. In telecommunications it is the hidden middle link: the radio interface serves the phone, but every byte must then travel over backhaul before it reaches the internet or another subscriber.
Backhaul capacity, latency and timing quality set a hard ceiling on what a base station can deliver. A site with excellent spectrum and modern antennas still performs poorly if its backhaul is congested, which is why operators plan backhaul together with radio capacity rather than as an afterthought.
How backhaul works in a mobile network
In a mobile network built on telecom fundamentals, traffic follows a chain: the device talks to a base station over the air, the base station sends packets over backhaul to an aggregation point, and aggregation routers forward them to the core network. In LTE the backhaul carries the S1 interface toward the core and often the X2 interface between neighboring sites; in 5G the equivalents are N2 and N3 toward the 5G core and Xn between gNodeBs.
Modern backhaul is almost always IP or Ethernet based, typically using MPLS or segment routing in the aggregation layer. Traffic is encrypted with IPsec where the transport is not trusted, and quality-of-service markings keep voice and signaling ahead of bulk data when links are busy.
Backhaul also distributes timing and synchronization. Time-division duplexing and many advanced radio features require base stations to share phase alignment within about plus or minus 1.5 microseconds, delivered through GNSS receivers at each site or through Precision Time Protocol (IEEE 1588) and Synchronous Ethernet carried over the backhaul itself.
Backhaul, midhaul and fronthaul
With centralized and virtualized radio access networks, the single backhaul link has split into three transport segments, each with very different requirements. Understanding the split explains why some sites need dark fiber while others work well with microwave.
| Segment | Connects | Typical latency budget | Capacity demand |
|---|---|---|---|
| Fronthaul | Radio unit to distributed unit | About 100 to 250 microseconds | Very high, scales with antennas and bandwidth |
| Midhaul | Distributed unit to central unit | A few milliseconds | Moderate, close to user traffic |
| Backhaul | Central unit or base station to core | Several to tens of milliseconds | Aggregated user traffic plus overhead |
Fronthaul protocols such as CPRI and eCPRI carry radio samples or lightly processed data, so their bandwidth grows with the number of antennas rather than with actual user demand. Backhaul, by contrast, carries ordinary packet traffic, which allows statistical multiplexing and makes it far more tolerant of distance and shared infrastructure.
Backhaul technologies and their trade-offs
Operators choose backhaul media site by site, balancing capacity, cost, deployment time and terrain. No single technology wins everywhere, so most national networks run a mix.
| Technology | Typical capacity per link | Typical reach | Main strength | Main weakness |
|---|---|---|---|---|
| Fiber | 10 to 100 Gbit/s and more | Tens of kilometers | Practically unlimited upgrade path | Civil works are slow and costly |
| Traditional microwave (6 to 42 GHz) | Hundreds of Mbit/s to a few Gbit/s | Several to tens of kilometers | Fast deployment without trenching | Needs line of sight and spectrum licenses |
| E-band millimeter wave (70/80 GHz) | Up to about 10 to 20 Gbit/s | Usually a few kilometers | Fiber-like capacity over short hops | Rain fade limits link length |
| Satellite | Tens to hundreds of Mbit/s | Global coverage | Reaches remote sites | Higher latency and recurring cost |
| Copper or leased lines | Tens of Mbit/s to about 1 Gbit/s | Short distances | Reuses existing assets | Insufficient for modern 5G sites |
Microwave links are engineered for availability targets such as 99.99 or 99.999 percent, and adaptive modulation lowers the data rate during heavy rain instead of dropping the link entirely. Satellite backhaul through low Earth orbit constellations has reduced latency to tens of milliseconds, making it a practical option for rural cells that once relied on geostationary links with roughly 600 milliseconds of round-trip delay.
Dimensioning backhaul capacity
Backhaul dimensioning estimates how much capacity a site needs at its busy hour, not its theoretical peak. Planners consider the peak rate of each sector, the expected average load, the number of sectors and carriers, and transport overhead from IPsec, tunneling and Ethernet framing, which commonly adds about 10 to 20 percent.
| Input | Effect on required capacity | Typical planning practice |
|---|---|---|
| Peak sector throughput | Sets the minimum burst capacity | Provision for at least one sector at peak |
| Busy-hour average load | Sets sustained capacity | Sum averages across all sectors |
| Carrier aggregation and bandwidth | Raises both peak and average | Recalculate when spectrum is added |
| Transport overhead | Adds a fixed percentage | Add margin for encryption and tunnels |
| Growth horizon | Determines upgrade timing | Plan for several years of traffic growth |
A common rule is to provision the larger of two values: the peak of a single sector, or the sum of busy-hour averages across all sectors. Sites in a daisy chain or ring must also carry traffic from downstream sites, so hub sites near fiber points of presence often need many times the capacity of an individual tail site.
Common backhaul problems
Backhaul faults often look like radio problems from the user’s perspective, which makes diagnosis tricky. Slow speeds with full signal bars are a classic sign that the bottleneck sits behind the base station rather than on the air interface.
- Congestion at the busy hour appears as falling throughput and rising latency while radio metrics remain healthy.
- Microwave fading during heavy rain reduces capacity on high-frequency hops, especially in E-band.
- Synchronization loss after a GNSS or PTP failure can force TDD sites to stop transmitting to avoid interference.
- Single points of failure in chained topologies take down several sites when one hub link fails.
- Underestimated growth after adding new spectrum leaves the radio side upgraded while backhaul stays fixed.
Operators mitigate these risks with ring or mesh topologies, monitoring of link utilization at short intervals, and upgrade triggers set well below full load, often around 70 to 80 percent busy-hour utilization.
In plain words
When you use your phone, the signal travels through the air only to the nearest cell tower. From there it still has a long way to go, and backhaul is the road that takes it from the tower to the rest of the network and the internet.
Think of a tower as a busy bus stop. The stop can be modern and large, but if the road leading away from it is narrow, everybody waits. A fast tower with weak backhaul feels slow, which is why operators spend a great deal on fiber cables and microwave dishes that most people never notice.
A real-world example
A regional operator upgrades a rural cell site from LTE to 5G by adding 100 MHz of mid-band spectrum. The radio upgrade raises the expected busy-hour demand from about 300 Mbit/s to roughly 1.5 Gbit/s, but the site still uses a 6 GHz microwave link limited to about 500 Mbit/s.
After launch, customers report full signal bars but disappointing speeds every evening. Monitoring confirms the microwave hop is saturated during the busy hour. Because trenching fiber would take many months, engineers install an E-band link of about 10 Gbit/s to a nearby hub site that already has fiber, keeping the old microwave link as a backup path.
Evening speeds rise to match the new radio capacity, and the backup link keeps the site online during heavy rain. The outcome shows that radio upgrades only pay off when backhaul is upgraded with them.