Resource Grid: RE, RB & Bandwidth in LTE 4G

Introduction

The resource grid is how LTE turns "spectrum" into something a scheduler can hand out. OFDM (downlink) and SC-FDMA (uplink) both slice the channel into a two-dimensional array of subcarriers in frequency and OFDM symbols in time, and TS 36.211 gives every cell in that array a precise name and size. Everything the physical layer does — carrying user data, placing reference signals, positioning the control channels — is ultimately a statement about which coordinates on this grid are used for what.

A UE lives on this grid from the moment it is scheduled. A downlink assignment tells it exactly which resource blocks in which subframe carry its PDSCH; the control region at the front of every subframe is where it hunts for that assignment; the reference-signal positions are the fixed pilots it uses to estimate the channel before demodulating anything. Capacity, coverage and overhead are all read off the same sheet.

Understanding it matters because throughput is, at bottom, an accounting problem on this grid: total resource elements in an allocation, minus the ones spent on reference and control signals, times the bits each remaining element carries. Get the geometry of the grid right — the atom, the block, the scheduling unit, the bandwidth ladder — and that whole calculation, and most capacity questions with it, becomes routine.

Why LTE needs a resource grid

Concretely, the grid solves the multiplexing problem. Many UEs, plus reference signals, plus control channels, all have to share one channel without colliding — and they have to do it dynamically, because who needs resources changes every millisecond. By defining a fixed set of named units (the resource element, the resource block, the PRB pair) on a shared time-frequency array, LTE lets the scheduler allocate chunks of the sheet to different users each subframe, while reference signals and control regions occupy known, reserved coordinates every UE can predict. The result is that a UE knows exactly where to find its own data, where the pilots are for channel estimation, and where the control channel that pointed it there lives — all from the same coordinate system.

The Time-Frequency Grid

LTE transmits with OFDM in the downlink and SC-FDMA in the uplink, so the physical layer is naturally two-dimensional: a set of closely spaced subcarriers stacked in frequency, each carrying a stream of OFDM symbols in time. TS 36.211 defines every unit on this grid, and everything else — scheduling, reference signals, control channels — is just a rule about which coordinates get used for what.

Formally, the grid for one antenna port in one slot has NRB × 12 subcarriers along the frequency axis and 7 (or 6) OFDM symbols along the time axis. Each point (k, l) in that array — subcarrier index k, symbol index l — addresses exactly one Resource Element. Larger channel bandwidths do not change the shape of a single block; they simply stack more blocks up the frequency axis. This is the single most important structural idea in LTE: the grid scales in count, never in the size of its cells.

Resource Element: the Atom of the Grid

The smallest thing you can point at in LTE is a single square on the sheet. That square is the Resource Element, and nothing in the physical layer is smaller.

A Resource Element (RE) is exactly one subcarrier by one OFDM symbol. It is the atom of the physical layer: nothing finer is addressable. Each RE carries one modulation symbol — one QPSK, 16QAM, 64QAM or (in later releases) 256QAM constellation point — so the number of bits an RE holds depends only on the modulation currently in use: 2 bits for QPSK, 4 for 16QAM, 6 for 64QAM, 8 for 256QAM. An RE is identified by the index pair (k, l), where k is the subcarrier and l is the OFDM symbol within the slot.

Not every RE carries user data. Some are reserved for the cell-specific reference signals (CRS) the UE needs for channel estimation, some for the primary and secondary synchronization signals (PSS/SSS), some for the broadcast channel (PBCH), and a block at the start of each subframe for the control channels. But whatever a square holds, it is still one RE — the unit never changes, only its assigned purpose does. This is why physical-layer capacity is ultimately an accounting exercise: count the total REs in an allocation, subtract the ones spent on reference and control signals, and what remains is what carries the transport block.

Resource Block: 12 Subcarriers Over a Slot

Individual REs are too fine-grained to schedule one at a time, so LTE groups them into a standard block — and here LTE differs from NR in a way worth pinning down precisely.

A Resource Block (RB) is 12 consecutive subcarriers in frequency, spanning one 0.5 ms slot in time. With normal cyclic prefix a slot holds 7 OFDM symbols, so one RB contains 12 × 7 = 84 Resource Elements. This is the crucial point for a fresher: in LTE the RB is a two-dimensional tile — it has a time dimension baked into its definition. In NR the RB is defined in frequency only (12 subcarriers, no time span). Same name, different definition; do not carry the NR meaning across.

Because the spacing is fixed at 15 kHz, an RB always occupies 12 × 15 kHz = 180 kHz of bandwidth, in every LTE cell, at every carrier frequency, in every release. That constancy is why the RB is the natural currency for talking about LTE bandwidth: count the RBs and you know the spectrum. The table below shows the two cyclic-prefix cases side by side — the frequency dimension is identical, only the symbol count per slot differs.

The extended CP inserts a longer guard interval between symbols to absorb multipath in large or hilly cells; the price is one fewer symbol per slot, so throughput drops slightly. Notice that even the exotic 7.5 kHz MBSFN case still fills the same 180 kHz RB with 72 REs — the grid geometry rearranges but the block bandwidth is invariant.

The Scheduling Unit: a Resource Block Pair

Here is a subtlety that trips people up. The RB is defined over one 0.5 ms slot, but the scheduler does not actually allocate a single slot — it allocates a full 1 ms subframe at a time.

The unit the eNodeB scheduler works in is the Physical Resource Block (PRB) pair: two RBs stacked in time — the two consecutive slots that make up one 1 ms subframe. So while the RB is the definitional building block, the PRB pair (2 slots = 1 subframe) is what a UE is actually granted in a downlink assignment or uplink grant. A PRB pair therefore contains 2 × 84 = 168 REs under normal CP, before subtracting whatever is taken by reference signals and the control region.

How the scheduler names those RBs to the UE is itself standardised: DCI uses one of three resource-allocation types. Type 0 grants whole resource-block groups (RBGs) via a bitmap — the RBG size is 1, 2, 3 or 4 RBs depending on channel bandwidth; Type 1 grants RBs within a subset for finer, scattered placement; and Type 2 (used by the compact DCI format 1A) grants a single contiguous run of RBs described by a start and length. The allocation still resolves to PRB pairs on the grid — the types only differ in how compactly the grant is encoded.

It helps to hold three time units in your head at once. The OFDM symbol (roughly 71.4 μs with normal CP) is the finest time step; seven of them make a slot (0.5 ms); two slots make a subframe (1 ms), which is the TTI; and ten subframes make a radio frame (10 ms). The RB lives at the slot level, but scheduling, HARQ and the whole grant machinery live at the subframe level — which is exactly why the PRB pair, not the bare RB, is the currency of a real allocation.

Subcarrier Spacing: Fixed at 15 kHz

Everything above rests on one number that LTE never negotiates: the subcarrier spacing.

The normal LTE subcarrier spacing is Δf = 15 kHz. It was chosen so that the OFDM symbol duration (1⁄15 kHz ≈ 66.7 μs, plus the cyclic prefix) is short enough to track the fast fading of high-speed mobiles yet long enough that the cyclic-prefix overhead stays modest. From 15 kHz everything else is derived: the RB is 12 × 15 kHz = 180 kHz, and the sampling and FFT sizes of the whole system flow from it. Crucially, this value does not change with channel bandwidth, carrier frequency, or CP length — a 1.4 MHz cell and a 20 MHz cell use identical 15 kHz tones; the wide cell just has more of them.

There is exactly one exception. For MBSFN (Multicast-Broadcast Single-Frequency Network) transmissions, LTE defines a reduced spacing of Δf = 7.5 kHz. Halving the spacing doubles the symbol length, which lengthens the effective cyclic prefix so that signals from many transmitters spread over a large single-frequency area still combine constructively at the receiver instead of causing inter-symbol interference. At 7.5 kHz an RB is 24 subcarriers over the same 180 kHz, with 3 long symbols per slot — still 72 REs, still 180 kHz. The 7.5 kHz mode is confined to MBSFN subframes; ordinary unicast traffic always uses 15 kHz.

Bandwidth: Counting Resource Blocks

Since every RB is a fixed 180 kHz, LTE's supported channel bandwidths are simply different RB counts. This is one of the most-asked interview tables in all of LTE, so it is worth committing to memory rather than deriving under pressure.

The RB ladder 6 / 15 / 25 / 50 / 75 / 100 maps to 1.4 / 3 / 5 / 10 / 15 / 20 MHz. Multiply the RB count by 12 to get the occupied subcarriers, and by 180 kHz to get the occupied bandwidth. From this one row you can reconstruct almost any LTE capacity figure on the spot.

Two refinements make the table honest. First, notice the occupied (transmission) bandwidth is always less than the nominal channel bandwidth. A 20 MHz carrier actually uses 18 MHz of subcarriers; the remaining 2 MHz is split into guard bands at the two edges so the signal decays before it reaches the neighbour's spectrum and so the transmit spectrum mask can be met. For the wider channels the occupied fraction is 90%; for the narrow 1.4 MHz channel it is about 77% (1.08 of 1.4 MHz), because a minimum practical guard is proportionally larger on a small carrier. The edge RBs therefore do double duty as the outermost usable resource and the buffer against adjacent-channel leakage.

Second, there is one subcarrier you cannot use for data: the DC subcarrier at the exact centre of the downlink carrier. It is left unmodulated to avoid the DC-offset spike that direct-conversion (zero-IF) receivers produce at the centre frequency, and the local-oscillator leakage a transmitter puts there. This is why the occupied count is 12 × N_RB + 1 when you include the DC tone: the data-bearing subcarriers number 12 × N_RB, and the unused DC tone sits in the middle as one extra, silent subcarrier. So a 20 MHz downlink spans 1201 subcarrier positions — 1200 usable, 1 DC. (In the uplink, SC-FDMA does not reserve a DC subcarrier in the same way; the DC-tone rule is a downlink feature.)

Control Sub-Units and Per-Port Grids

User data is scheduled in RBs and PRB pairs, but the downlink control region — the first one to four OFDM symbols of each subframe, where the PDCCH lives — is measured in its own smaller units. It also matters that the grid is defined per antenna port, because that is where reference-signal overhead really comes from.

The control region tiles into two nested units. A Resource Element Group (REG) bundles 4 usable REs (contiguous in frequency within one OFDM symbol, skipping any REs occupied by reference signals). A Control Channel Element (CCE) bundles 9 REGs = 36 REs, and the CCE is the atomic currency of PDCCH allocation: every control transmission is measured in whole CCEs, aggregated at levels 1, 2, 4 or 8. These are the LTE control units; do not confuse them with the NR redefinition (where a REG is a full RB × 1 symbol = 12 REs and a CCE is 6 REGs). We introduce the names here and cover their aggregation and mapping on the PDCCH page.

How many symbols the control region spans is signalled per subframe by the PCFICH as a Control Format Indicator (CFI = 1, 2 or 3, and up to 4 on the 1.4 MHz / 6-RB bandwidth). A larger CFI trades data REs for control capacity when many UEs must be scheduled at once. Whatever the CFI, the control region always starts at symbol 0 of the subframe, so a UE reads the PCFICH first to learn where the control region ends and its PDSCH can begin.

Now the antenna-port point. In LTE, the resource grid is defined separately for each antenna port: port 0, port 1, and so on each have their own grid, and cell-specific reference signals (CRS) for one port occupy REs that are deliberately left empty (muted) on the other ports, so a UE can estimate each port's channel without interference. That is why a 2-port or 4-port cell spends more REs on CRS than a single-port cell — every extra port adds its own reference pattern and mutes the corresponding REs elsewhere. Figure 1 marked the port-0 CRS in symbols 0 and 4; port 1 uses the same symbols on the interleaved subcarriers, while ports 2 and 3 add sparser pilots in symbol 1. The upshot for capacity accounting: usable data REs per PRB pair fall as you add antenna ports, because reference-signal overhead climbs with port count.

Summary

The LTE resource grid is a two-dimensional map of subcarriers (frequency) against OFDM symbols (time), defined per antenna port per slot. Its atom is the Resource Element (one subcarrier × one symbol, carrying one modulation symbol of 2–8 bits). Twelve subcarriers over one 0.5 ms slot make a Resource Block — 84 REs with normal CP, 72 with extended — and, crucially, the LTE RB has a time dimension baked in, unlike NR's frequency-only RB. The scheduler actually grants a PRB pair (two RBs = one 1 ms subframe = the TTI, 168 REs before overhead), which also enables slot-level frequency hopping.

Because the subcarrier spacing is fixed at 15 kHz (bar the niche 7.5 kHz MBSFN case), each RB is always 180 kHz, and bandwidth is just a count of RBs: the ladder 6 / 15 / 25 / 50 / 75 / 100 maps to 1.4 / 3 / 5 / 10 / 15 / 20 MHz. Occupied subcarriers are 12 × N_RB (plus one unused downlink DC tone), occupied bandwidth is N_RB × 180 kHz, and the remainder is guard band. The control region at the front of each subframe is measured in its own units — REGs (4 usable REs) and CCEs (9 REGs = 36 REs) — and the grid being per-port is why CRS overhead grows with antenna count.

Hold the hierarchy — RE, RB, PRB pair, bandwidth ladder, control sub-units — and LTE capacity, control-channel placement and reference-signal overhead all become one accounting exercise on the same sheet of graph paper.

Quick Q&A