A device can sit next to a router and report 866 Mbps, while real application throughput is only a few hundred megabits or less. BLE has the same trap: the PHY may say 2M, but GATT notifications do not deliver 2 Mbps of application data. The gap is not always marketing abuse. Bandwidth, rate, and throughput are different layers.
The common mistake is treating a larger advertised rate as a promise that application data will arrive at that speed.
channel bandwidth -> PHY rate -> MAC usable capacity -> transport throughput -> application payload
Each step loses capacity to protocol overhead, waiting time, acknowledgements, retries, contention, encryption, fragmentation, and application processing.
Bandwidth Is Spectrum Width
Bandwidth describes how much spectrum a signal occupies. It is not the same as application data speed.
In Wi-Fi, 20 MHz / 40 MHz / 80 MHz / 160 MHz usually refer to channel bandwidth. A wider channel can carry more subcarriers and enable a higher PHY rate, but it also has practical costs:
- More exposure to neighboring networks and non-Wi-Fi interference
- Fewer available channels and harder planning
- Less chance of sustaining high-order modulation at the coverage edge
- Shared airtime across devices instead of private capacity for one client
BLE is different. Classic BLE data channels are narrowband by design. The goal is low power, short packets, frequency hopping, and low implementation cost. NB-IoT is even more explicit: it trades raw speed for coverage, power, and network capacity.
Bandwidth is part of the capacity ceiling, not the final throughput.
PHY Rate Is Air-Interface Bit Rate
PHY rate describes how many bits the physical layer can carry under a specific modulation, coding, spatial-stream, and channel-width condition.
In Wi-Fi, the selected rate depends on:
- Channel bandwidth
- Modulation order
- Coding rate
- Number of spatial streams
- Guard interval
- Current link quality and rate adaptation
In BLE, 1M PHY, 2M PHY, and LE Coded PHY are not application-throughput promises. 2M PHY raises the air-interface bit rate, but packet spacing, connection events, ATT/GATT overhead, controller buffers, and host scheduling still exist. LE Coded PHY sacrifices speed for coverage and robustness.
Higher PHY rates usually require better signal quality. When the field rate drops, the protocol may simply be choosing a safer modulation and coding scheme.
Throughput Is What the Application Receives
Throughput measures how much useful data is delivered upward over time.
Between the PHY and the application, many things consume capacity:
- Preamble, headers, checksums, and inter-frame spacing
- MAC acknowledgements and backoff
- Encryption, aggregation, fragmentation, and reassembly
- Link-layer retransmission
- IP, TCP, TLS, MQTT, CoAP, and other upper-layer overhead
- Application acknowledgement, serialization, and storage work
For Wi-Fi, even a single-client speed test includes AP CPU, driver behavior, NAT, TCP windows, server distance, and same-channel contention. For BLE, throughput is often constrained by connection interval, packets per connection event, MTU, DLE, notification behavior, and system scheduling.
The difference between advertised rate and application throughput is normal. The useful question is whether the gap matches the link condition and protocol configuration.
Retries Consume Both Capacity and Latency
Wireless is not a stable cable. When packets are lost, many protocols retry at the link layer or above.
Retries have two costs:
- The same application data consumes more airtime
- Later data waits longer, increasing latency and jitter
That is why “still connected” does not mean “still usable” near the coverage edge. The link may not be down, but much of the airtime is spent on lower rates and retries.
In Wi-Fi, acceptable RSSI with poor throughput often points to high noise, heavy same-channel contention, hidden nodes, AP load, or unstable rate adaptation. In BLE, a connection can remain alive while notifications are unstable because connection events spend too much time on retransmission or the connection parameters do not fit the traffic burst.
Design for Sustained Usable Capacity
Wireless design should care more about sustained usable capacity than lab peak rate.
Useful judgments:
- For low latency, inspect latency distribution and retries, not just average throughput
- For battery life, higher throughput is not always better; wake time and retry cost matter
- For many devices, single-client peak rate matters less than shared airtime capacity
- For long coverage, a slower but stable link may be better than a high advertised rate
Bandwidth, rate, and throughput live at different layers. Bandwidth is spectrum resource, PHY rate is air-interface speed under a modulation and coding condition, and throughput is the useful data actually delivered. Separating them turns “why is it below the advertised rate?” into an analyzable wireless problem.
Continue Reading
- What PHY, Modulation, and Coding Actually Solve: why PHY rate changes with link quality
- Why Wireless Communication Cannot Avoid Channels, Interference, and Retries: why contention and retries consume real throughput