Radio capacity is easy to sell and easy to measure. Transport capacity is neither, which is why it is routinely the constraint that limits what a 5G network actually delivers. The arithmetic is not complicated, but it has to be done before the radio order is placed rather than after the first site goes live.
Start from what the radio can produce
Take a macro configuration with 320W and 16-layer MIMO with three-carrier aggregation: up to 10 Gbps downlink and 1.3 Gbps uplink per centralised unit, serving three to six cells and up to 3,000 user equipments. Even a mid-tier 8×40W eight-layer configuration produces around 5 Gbps downlink and 450 Mbps uplink per CU.
No operator dimensions transport to peak — that would be absurd. But the ratio matters: the 4G site being replaced was typically backhauled at 1 Gbps and frequently less. The replacement can generate five to ten times that. A cell-site router at 10 Gbps is not generous; it is the correct order of magnitude.
Three numbers to work from
Busy-hour throughput, not peak. Dimension against measured busy-hour traffic with a growth factor, and revisit it annually. Most operators in this region are seeing traffic growth that outruns their transport refresh cycle.
Uplink separately. TDD is asymmetric by configuration, and the applications that justify 5G commercially — fixed-wireless access with real usage, video contribution, industrial telemetry, surveillance — are uplink-hungry in ways consumer mobile was not. Dimensioning uplink as a fixed fraction of downlink is how fixed-wireless propositions disappoint.
Aggregation ratios. A cell site at 10 Gbps behind a metro ring that aggregates thirty of them needs 100 Gbps class aggregation, with a statistical multiplexing assumption you can defend. This is what the 100G site router class exists for — the access layer of a metro network carrying 5G application traffic, aggregating mixed 3G, 4G and 5G mobile, fixed access, enterprise and transport traffic from multiple vendors onto one path.
Synchronisation is not a transport detail
In TDD — which is what n41 and n78 5G is — every cell has to agree on when it transmits and when it listens. If synchronisation drifts, cells transmit into each other’s receive windows. The symptom is not an outage; it is degraded throughput and rising interference across a cluster, which is far harder to diagnose and can persist for months.
GPS and IEEE 1588v2 have to be designed into the transport layer, with holdover behaviour specified for GPS loss. Treat sync as a first-class design requirement with its own acceptance test, not as a checkbox on the router datasheet.
The multi-vendor reality
The transport layer is where vendor boundaries physically collide — radio from one supplier, transmission from another, an existing microwave estate from a third, and an enterprise access network that predates all of them. A site router’s job is to sit at that point of transformation between mobile radio, fixed access, enterprise and transport networks and make one coherent path out of it, prioritising and processing voice, data and signalling for reliable transport across any available backhaul.
That is an integration problem before it is a product problem. It is also why transport is the layer where an operator most often ends up arbitrating between its own suppliers.
A practical checklist
- Dimension against measured busy hour with an explicit growth factor, not against peak or against last year.
- Dimension uplink independently, against the fixed-wireless and enterprise plan.
- Specify cell-site routing at 10G where the radio configuration justifies it — which is most macro sites.
- Size metro aggregation at 100G class with a defensible multiplexing ratio.
- Design GPS and 1588v2 synchronisation with holdover, and test it as a named acceptance item.
- Confirm the transport plan can carry 3G and 4G traffic from the existing estate throughout the migration.
Do this before the radio purchase order. Transport discovered late is the most expensive kind of late.