Sometimes headline figures can be misleading. With both cellular and Wi-Fi claiming to have broken the Gigabit barrier, you might be surprised to learn that wired Gigabit Ethernet can still outperform. We clarify the terminology to ensure you are making a fair comparison. This is important when dimensioning Ethernet backhaul for Wi-Fi and small cells.
Ethernet evolved from simplex wired to TDD wireless
When Ethernet was first invented, it shared the same wires used for transmit and receive, checking for other transmissions before sending out packets. This TDD mode is still found in some parts of a home or enterprise network, such as where passive hubs are used to share capacity.
In these cases the total data rate for both transmit and receive is 1 Gbps.
Wi-Fi took the principles of simplex Ethernet and applied them to a wireless link. Also using TDD (time division duplex), signals are sent and received on the same shared carrier frequency. Several channels can be aggregated to achieve high peak rates of 1Gbps or more. But these are simplex, with the peak data speed including both uplink and downlink flows.
Full Duplex Ethernet now widespread
More commonly now, we find switched wired Ethernet, which is full duplex using separate pairs of twisted wires and can transmit and receive at 1Gbps rates down and up simultaneously. Effectively this achieves peak datarates of 2Gbps per cable.
The total throughput of an Ethernet router can be many Gbps, although more often these aggregate traffic to a single broadband link into the internet or a data centre.
If today’s wired Ethernet were marketed like Wi-Fi, it would be headlined as 2Gbps technology.
Cellular data was duplex but now becoming TDD
Cellular data started off as full duplex (e.g. for GPRS, 3G UMTS and now FDD-LTE). We’ve seen real world data rates increase from the trickle of early GPRS (a few kbps) to LTE with over 100Mbps in a full sized 20MHz carrier.
Although cellular data uplinks have poorer performance than downlink, cellsites can send and receive simultaneously and achieve very high throughput rates.
TD-LTE is a more recent introduction and is simplex (ie sending or receiving at any one time), and so the peak data rates quoted for that are for combined uplink and downlink performance. Often the timeslots allocated for uplink and downlink are statically assigned, rather than dynamic (as with Wi-Fi) which limits flexibility.
Carrier aggregation across different bands can double peak rates to 300Mbps or more. 3GPP standards already allow for aggregation of up to four bands, with even more planned for a future release.
There are even combinations of FDD and TDD using carrier aggregation. This is where you have to start looking more carefully at the headline rate.
LAA is a good example. A likely combination would be one LTE licenced carrier (say 20MHz FDD) with thee TD-LTE 20MHz carriers at 5GHz. The peak download rate could be as much as 1Gbps (as demonstrated by AT&T at Mobile World Congress) but the uplink would be limited to around 150Mbps.
Wired Ethernet requires good quality cables of the correct specification. Cat 5e (rather than just Cat 5) is required for Gigabit speed. Cat 6 is best. Longer lines, poor termination and high external interference can all limit capacity. That may result in packet loss and retransmission or the circuit dropping back to 100Mbps mode of operation.
Generally speaking, the high quality of professional cabling installations for enterprise and commercial buildings ensures full performance throughout a property.
Some of that line transmission rate is taken up with overheads, such as packet framing. A typical 1500 byte IP packet size brings it down to about 974 Mbps. Jumbo Ethernet frames of 9000 bytes improve the efficiency somewhat but are rarely used for wide area Internet service.
Wireless performance degrades for many reasons including interference, environment, mobility and backward compatibility to handle older devices. In some cases, the ambitious QAM modulation level drops back in less than ideal signal/noise conditions. 256 QAM is quite ambitious but feasible in a well planned indoor environment.
The real killer to total speed is the number of concurrently connected devices. This is much more difficult to manage in an unco-ordinated best effort system such as Wi-Fi, resulting in a considerable drop-off in the total throughput achievable.
Another trick to hype the peak speed
Wi-Fi access point vendors are now using a format such as AC1300 or AC2400 to brand their equipment. The AC relates to 802.11ac specification while the number relates to the peak throughput in Mbps.
What’s in the small print is that the peak throughput may not be achievable for a single end user device. Sometimes the capacity of the 2.4GHz band (which ac doesn’t use) is added. The use of 4x4 MU-MIMO doubles performance and throughput of the access point, but that is shared between multiple clients.
So when will a Wi-Fi access point outperform the capacity of a wired Gigabit Ethernet cable? One vendor (Ubquiti) assesses this to be somewhere between two to three 4x4 MU-MIMO 80MHz radios depending on the ratio of the downlink/uplink traffic as shown in the following chart:
Some thoughts about small cell backhaul capacity
The duplex operation of a 3G or FDD-LTE small cell is more closely matched to wired Gigabit Ethernet than TDD technologies. It can support a full 1Gbps downlink at all times.
In practice, an enterprise Small Cell network is much less constrained over the air than by the external backhaul link to the core network. This is the usually first aspect to be addressed if there are Enterprise small cell capacity issues.
But if in future there was a need for faster wired speeds in-building, there are other technologies now available. N-BaseT provides up to 2.5Gbps over Cat5e and up to 5Gbps over Cat 6 as we’ve previously explained.
More capacity can also be added by increasing the number of small cells, each with their own dedicated Ethernet cable.