<text-capital>A<text-capital>fter designing a company’s network, Meter tests the network for adequate coverage, reliability, and speed. We also hold ISPs accountable to the SLAs they give our customers. Through our work, we’ve noticed companies are typically given a theoretical expectation of WiFi performance that, in practice, is more consistent with wired connections.
We’ll try to give a synopsis of what's behind WiFi's inherent limitations, from physical and channel interference to how bandwidth is shared between devices. This post will also explain how to think about the speeds you actually need.
How WiFi speed is measured
When purchasing a business internet connection or network hardware (e.g., access points, security appliance), the price will include a specific data rate quoted in Mbps or Gbps. This refers to the amount of data that can be transmitted on your network per second.
When an ISP quotes a 500 Mbps symmetrical connection, they’re telling you that you’ll be able to download and upload 500 megabits of data per second; if the connection is asymmetrical, download and upload speeds will differ with download typically being quicker. Similarly, an access point that advertises 1.6 Gbps is claiming that it’s theoretically possible for the network to enable 1.6 gigabits of data transfer per second . For reference, watching a Youtube video at 1080p HD uses ~5 Mbps.
But, given the limitations of WiFi compared to wired internet connections, in both cases your actual data rate will be much lower than what you’re quoted.
Why WiFi is slower than wired
WiFi allows you to connect to a network wirelessly using radio waves. A client device (e.g., smartphone, laptop, etc.) sends data via radio waves to an access point, which connects to the broader internet and responds to your device via radio waves.
Wired connections send data via electrical pulses over a physical cable. You can give any client device a direct internet connection by connecting it to a router via an ethernet cable.
By transmitting data “through the air”, WiFi is more susceptible to failure. For example, the distance your device is from the nearest access point, the line of sight your device has to an access point, overlapping WiFi channels, dynamics of WiFi protocols, and device limitations all contribute to WiFi performance issues.
As you roam farther from the access point your device is connected to, the signal strength of the radio waves used by WiFi degrades. Even if you place multiple access points around your space to always ensure a short distance to an access point, client devices prioritize maintaining a WiFi connection, so they tend to stick to the access point they’ve been communicating with rather than shift to the new nearest access point.
Client devices need a clear line of sight to an access point to receive adequate WiFi signal. This can be interrupted by physical interferences like glass, concrete, metal, and brick. Also, radio waves emitted by devices and access points can bounce off objects in your space, causing the signal to be received more than once and in a difficult to decipher format.
WiFi must operate in the 2.4 GHz and 5 GHz frequency bands . The 2.4 GHz band has greater coverage but slower speeds, while the 5 GHz band has faster speeds but less coverage. It’s common for low-power devices like smart thermostats to connect to the 2.4 GHz band, while high-power devices like phones and laptops connect to the 5 GHz band.
Each band is divided into smaller bands called channels. The 2.4 GHz band is 100 MHz wide and has 11 channels that are each 20 MHz wide. If we only concentrate on channels in the 5 GHz band that WiFi doesn’t need permission to use (highlighted in green in the below image), we have eight channels that are each 20 MHz wide .
An access point sits on a channel with a width. For instance, an access point sitting on channel 36 with a width of 40 MHz will span across channels 36 and 40 .
Before transmitting data, a device checks if the channel is clear. If it’s clear, it sends a preamble that describes how the data will be sent. If two devices that communicate with the same access point notice a channel is clear and can’t hear the other device’s preamble, they could accidentally send data at the same time on the same channel. This is how channel interference happens.
You can try to mitigate this by having adjacent access points use different channels.
Competition from other devices
Since WiFi bands are a shared resource, only one device can communicate on a channel at any given time. Client devices must wait for the channel to be clear to send data, otherwise their data will collide. As a result, each client device only gets a portion of the bandwidth; for example, a 100 Mbps connection shared between five online devices means each device gets 20% of the bandwidth.
Overhead when transmitting data
When a client device or an access point transmits data, it must first wait until the channel is clear, then wait a little bit longer so that others waiting to transmit don’t all start at exactly the same time. Next, it’ll send a slow preamble . All of this work required to allow data to be transmitted uses up airtime .
Beyond all of this back and forth, there’s extra data wrapped in each transmission. WiFi’s radio waves are turned into units of bytes that a computer can understand. These units (i.e., frames) are of three types:
- Management: how an access point can start exchanging data with a device.
- Control: how a device acknowledges data can be sent and confirms data has been received.
- Data: contains the actual data to send and headers that specify where it’s going.
As a result, only some percentage of the airtime is used to transmit the actual data, resulting in slower speeds.
Unused data streams
Many access points will advertise having four transmission antennas (i.e., 4x4 MIMO), theoretically allowing the access point to transmit data at four times the max transmission rate of each radio on the access point (e.g., if each radio has a max transmission rate of 400 Mbps, the access point is advertised as 1.6 Gbps). Beyond the max transmission rate not being possible for reasons outlined above, nearly all client devices have one or two data streams, which leaves the remaining streams unused. Several technologies have been proposed to allow different devices to connect to different data streams concurrently . However, most client devices can’t take advantage of them, resulting in only one client device being connected to an access point at a given time.
Reality of speed tests
After understanding the reality of WiFi speeds, it’s important to note:
- Network reliability must be taken into account along with WiFi speeds, and
- You likely don’t need WiFi speeds anywhere close to what you’re quoted.
Taking a speed test doesn’t provide the full picture. Speed tests are usually taken with a single device after a network has been built. While this can identify issues of distance and physical interference, it doesn’t reveal other issues that crop up when an entire workplace is using a network. For instance, an access power may have its power set too high causing competition between too many devices, or adjacent access points may be configured to use the same channel causing devices on both access points to overlap. It’s important to consider the reliability of the network in conjunction with speeds sampled.
Speeds you realistically need
Meter supports many fast growing technology companies with high data usage from both sales and engineering teams. Amidst all of the Zoom calls and data set uploads, we find companies use a much lower data rate than they’re quoted from their ISP or network hardware manufacturer. For instance, below is a network usage graph of a technology company with ~200 employees (re-plotted to preserve anonymity). Notice that during the time of the day with the highest usage, they only hit 210 Mbps.
This is not to say you should buy the internet connection or network hardware with the lowest quoted data rate. As mentioned, ISPs can’t always meet their SLAs and network hardware is grossly overoptimistic. We typically recommend internet connections in the range of 250 Mbps to 1 Gbps and use network hardware that we design and manufacture. If you need guaranteed internet speed and reliability at desks, we recommend installing wired ethernet drops.
In short, WiFi has inherent limitations that lower its speed and reliability such as distance to an access point, physical and channel interference, device competition, requirements of WiFi protocols, and device flaws. Fortunately, we need speeds much lower than what we’re sold.
Special thanks to
Derrick Pallas and Lachy Groom
- 1.6 Gbps means 1.6 billion 0s and 1s are being sent per second between an access point and your device.
- The frequency of a radio wave is simply the number of times it oscillates per second. By operating in the 2.4 GHz and 5 GHz frequency bands, its signal must oscillate 2,400,000,000 and 5,000,000,000 times per second, respectively. Use of frequencies is regulated by the FCC to prevent frequencies from interfering with one another.
- While the FCC generally requires licenses to use airwaves, it leaves some frequency bands open. WiFi operates in unlicensed bands (2.4GHz and 5GHz), meaning WiFi devices don’t need permission to use the bands. Some WiFi devices will try using other bands to get more bandwidth but must boot themselves off when the primary user (e.g., radar) needs to use the band.
- When a client device transmits data to the access point, it’ll be transmitting over the channels that the access point sits on (this assumes the device supports the same channel width as the access point; if an access point sat on channel 36 and supported a width of 40MHz, but the client only supported 20MHz, the client would only transmit over channel 36, not channels 36 and 40).
- The preamble must be sent very slowly so that every client, no matter how old or far away, understands that the transmission is taking place, even if they can’t demodulate the actual data being sent (demodulation takes power, so if the message is not for you it makes sense to only focus on knowing that the channel is being used). This data can then be transmitted at a much faster speed because everyone knows a transmission is taking place.
Given all the work required to begin transmitting, individual packets for a particular destination can be aggregated into a single large transmission. The receiver will then acknowledge which pieces of the aggregate were received without error so that the transmitter can retransmit anything that wasn’t received properly. To make sure that this exchange can take place uninterrupted, clients will often first make a Request To Send (RTS) and the access point will respond with a Clear To Send (CTS) that includes a duration during which other devices should not transmit, even if the channel seems clear. When access points transmit, they often first transmit a CTS-to-self that has the same effect. If other devices are behaving properly, this Virtual Carrier Sense enables advanced features to work in environments with older devices without the older devices needing to understand or even hear the transmission.
- Note that this uses airtime, not bandwidth. Airtime is how long a transmitter uses spectrum (i.e., radio waves in the frequency range that WiFi is allowed to operate in). Bandwidth is the amount of spectrum the signal is using.
- MU-MIMO (stream sharing) in 802.11ac or HEW (sub-channel sharing) in 802.11ax.