Lakhasly

Not all devices on 5G NR have to support all bandwidths, which is a change from LTE. Furthermore, 5G NR supports adaptive bandwidth, letting devices move to a low-bandwidth, low-power configuration when appropriate, and gearing up to higher bandwidths only when necessary. This creates the opportunity for very low average power devices that can still deliver high performance -- IoT networks, for example, which normally only need small amounts of data for telemetry, but nevertheless need to be able to update their firmware for security and feature patches. The 5G NR specification refers to these different configurations as 'bandwidth parts', and in theory a device can support multiple bandwidth parts simultaneously on the same channel, although the first 5G NR release limits devices to one bandwidth part at a time.

Within a subchannel, data is divided up into frames of ten milliseconds each, further subdivided into ten 1ms subframes. Those subframes are themselves divided into slots of 14 OFDM symbols apiece. Thus, wider bandwidth subchannels have more OFDM symbols per second and each slot thus gets shorter, but the basic frame structure stays the same. At the lowest subcarrier spacing, 15kHz, the frames are identical to LTE, simplifying compatibility.

LTE and similar systems allocate bandwidth to different devices by slot, but 5G NR has a mechanism for a transmission to start within a slot, effectively creating what are called 'mini-slots'. This is especially useful for the high bands, which can have very large OFDM symbols and thus the ability to use just a few to carry a relatively short message improves both channel reuse and latency. Another potential advantage is if, or when, 5G expands to unlicensed spectrum, which normally comes with a 'listen before use' rule to prevent interference. If a channel appears quiet, the ability to start a transmission without having to wait for a slot boundary reduces the chance of another device grabbing the channel.

Other low-latency adaptations in 5G NR are tight requirements for data transmissions to start after a channel is granted, and restrictions on processing delay for data streams. This is achieved in the higher network layers by changing header structures so that processing can begin without the full packet information being known, and at the physical layer by having the radio receive essential information from reference and downlink control signals instead of deriving it from the symbol stream.