Transmission Modes (TM1–TM10) in LTE 4G

Introduction

A Transmission Mode is the single per-UE setting that tells the physical layer how the downlink shared channel (PDSCH) is mapped onto the eNB's antennas for one particular UE. LTE defines ten of them, TM1 through TM10, spanning the whole history of the standard: TM1–TM6 arrived with Release 8/9 and are built on the always-on cell-specific reference signals; TM7–TM10 arrived from Release 9 through Release 11 and are built on UE-specific reference signals that unlock free-form beamforming and high-order MIMO.

The mode is chosen when the UE is set up on a cell and can be changed as its radio conditions evolve — a cell-edge UE might sit in robust transmit diversity, then be moved into high-throughput spatial multiplexing as it walks toward the cell centre. That change is an RRC event, not a per-subframe one, which is why the fallback mechanism described later in this page matters so much during the moments a mode is being reconfigured.

It matters because the receiver cannot decode PDSCH until it knows how the data was placed on the antennas, which reference signal to demodulate against, and which DCI format to look for. Bundling all of that into one numbered mode keeps the control channel compact and lets the network match the antenna technique to each UE's antenna count, channel quality, and mobility.

Why transmission modes are needed

Concretely, three problems force the design. First, the receiver must know the antenna scheme in advance — whether it is seeing one stream sent redundantly, several independent layers, or one steered beam — because each demands a different equaliser and a different assumption about the reference signals. Second, signalling must stay compact: negotiating the antenna scheme, pilot pattern, and feedback set message-by-message would bloat the PDCCH, so LTE freezes a coherent bundle as a numbered mode configured once by RRC. Third, no single technique is best everywhere — diversity suits the cell edge, multiplexing suits a strong channel, beamforming suits a coverage hole — so the network needs a clean way to match the technique to the UE. The transmission mode is that knob: set it, and the antenna scheme, reference signal, DCI format, and feedback report all follow.

What a Transmission Mode Is

The downlink of LTE is a MIMO system, and MIMO can be used in very different ways: send the same data over several antennas for robustness, send different data streams in parallel for throughput, or focus energy toward one UE with beamforming. Each of these needs a different receiver assumption, a different reference-signal pattern, and different feedback. Rather than negotiate all that on the fly, LTE bundles a coherent set of choices into a numbered Transmission Mode. Pick the mode, and everything else — antenna scheme, pilots, control format, feedback report — follows from it.

It is worth being precise about scope. A TM configures the downlink shared channel (PDSCH) only — it does not change how the UE decodes PDCCH, and it does not govern the uplink. The mode is strictly per serving cell and per UE: two UEs on the same cell can sit in different modes in the same subframe, and in carrier aggregation each component carrier carries its own transmissionMode. Moving a UE between modes — say from robust transmit diversity to high-throughput spatial multiplexing as it walks toward the cell centre — is done with an RRCConnectionReconfiguration, not per-subframe. Switching is therefore a deliberate, RRC-timescale decision, which is exactly why the fallback below matters during the reconfiguration gap.

The Key Split: CRS-based vs DMRS-based

If you remember only one thing about transmission modes, make it this division. It cleanly separates the ten modes into two eras of LTE, and it explains almost every difference in the big table further down.

TM1 through TM6 are CRS-based. The UE demodulates PDSCH against the Cell-specific Reference Signals (CRS) — the always-on, cell-wide pilots tied to physical antenna ports 0–3. Because the CRS pattern is fixed by the cell and shared by everyone, any precoding the eNB applies must be layered on top of the CRS and then signalled to the UE (as a codebook index, the PMI), so the UE can work out how the CRS it measures relates to the precoded layers it received. This ties these modes to a fixed precoding codebook and to the number of physical antenna ports the cell has. The upside is that CRS is always present, so channel estimation is cheap and robust; the downside is that the beam must come from a small standardised codebook rather than being freely shaped.

TM7 through TM10 are DMRS-based. The UE demodulates against UE-specific Demodulation Reference Signals (DMRS, on antenna port 5, or ports 7/8, or ports 7–14) that are precoded with the same weights as the data. The UE never needs to know the precoder — it simply estimates the effective channel (real channel times precoder) directly from the DMRS and demodulates. This is what makes flexible, non-codebook beamforming and higher-order MIMO possible: the eNB can apply an arbitrary beam, computed however it likes (from uplink sounding in TDD, from rich feedback in FDD), and the UE is none the wiser. It is also the direction 5G NR took wholesale — NR dropped CRS entirely and made every downlink transmission DMRS-based.

There is a subtlety worth naming: in the DMRS era, LTE split the two jobs that CRS used to do alone. CRS was simultaneously the demodulation pilot and the measurement pilot (the UE measured CRS to compute CQI/PMI/RI). From TM9 onward those jobs are separated: DMRS handles demodulation only, and a new sparser pilot, CSI-RS (Channel State Information Reference Signal), handles measurement across up to 8 (later more) antenna ports. That separation is the single design move that made 8-layer MIMO and, later, CoMP practical, because you can measure a large array with a light-weight periodic pilot while spending demodulation pilots only on the layers actually transmitted.

Three Ways to Use the Antennas

Every transmission mode is a flavour of one of three underlying ideas. Seeing them side by side makes the whole table click: diversity trades away rate for reliability, multiplexing trades away reliability for rate, and beamforming spends antennas on signal strength instead of either.

A fourth idea sits quietly behind the diagram: MU-MIMO, where the eNB reuses the same time–frequency resource to serve several UEs at once, each with its own precoder — spatial multiplexing across users rather than across layers of one user. It is the engine behind TM5 (codebook) and the multi-user side of TM9/TM10 (DMRS). Keep the three-plus-one taxonomy in mind — diversity, multiplexing, beamforming, multi-user — and every mode below lands in one of those buckets.

The CRS Modes: TM1–TM6

These are the classic Release-8/9 modes, all built on cell-specific reference signals (CRS) and the fixed precoding codebook. Because they share the CRS foundation, they differ only in how they combine the physical ports and what feedback they need. For each mode below, watch the four things that define it: antenna configuration, reference signal, DCI format(s), and use case.

• TM1 — single antenna port. Antenna config: one logical port, port 0. RS: CRS. DCI: DCI format 1 (or 1A). Plain SISO — no diversity, no multiplexing. Rank is fixed at 1, so feedback is CQI only (no RI, no PMI). Use case: a cell or UE with a single usable antenna, small cells, and as the simplest baseline against which everything else is compared.
• TM2 — transmit diversity. Antenna config: 2 or 4 ports. RS: CRS. DCI: DCI format 1 (or 1A). Scheme: Space-Frequency Block Coding (SFBC, an Alamouti-style scheme; SFBC combined with Frequency-Switched Transmit Diversity, FSTD, for 4 ports). The same data is sent redundantly across the ports, so it decodes reliably even in a poor channel — no rate gain (still one codeword, rank 1), pure robustness. Feedback is CQI only. Use case: the default/most reliable mode, cell-edge UEs, and any situation where keeping the link alive beats squeezing out throughput. This is also the technique the DCI 1A fallback uses when the cell has more than one port.
• TM3 — open-loop spatial multiplexing. Antenna config: 2 or 4 ports. RS: CRS. DCI: DCI format 2A (+ 1A). Scheme: multiple layers using large-delay Cyclic Delay Diversity (large-delay CDD) with a fixed cycling precoder. Rank up to 2 (2 ports) or up to 4 (4 ports). "Open-loop" means no PMI feedback — the UE reports only rank (RI) and CQI. Because the precoder cycles deterministically rather than tracking the channel, TM3 is robust to higher mobility where PMI would be stale by the time it is used. Use case: fast-moving UEs that still have good enough SINR to support more than one layer. When the channel only supports rank 1, TM3 gracefully falls back to transmit diversity.
• TM4 — closed-loop spatial multiplexing. Antenna config: 2 or 4 ports. RS: CRS. DCI: DCI format 2 (+ 1A). Scheme: multiple layers (rank up to 2 or 4) with precoding chosen from the codebook using the UE's reported PMI. Full RI + PMI + CQI feedback. This is the workhorse high-throughput mode for low-to-medium mobility: the eNB tracks the channel via PMI and matches the precoder to it, extracting maximum rate. Use case: the typical good-SINR, slow-moving connected UE that you want to run at peak throughput.
• TM5 — MU-MIMO. Antenna config: 2 or 4 ports. RS: CRS. DCI: DCI format 1D (+ 1A). Scheme: Multi-User MIMO — the eNB pairs several UEs on the same time/frequency resource with different codebook precoders, one rank-1 stream per UE. Feedback is PMI + CQI. DCI 1D adds a downlink power-offset field so a UE knows how power was split with its co-scheduled partner. Use case: capacity-limited cells with many well-separated UEs, squeezing more users into the same resources. In practice TM5 saw limited deployment; codebook MU-MIMO matured later in the DMRS modes.
• TM6 — closed-loop rank-1 precoding. Antenna config: 2 or 4 ports. RS: CRS. DCI: DCI format 1B (+ 1A). Scheme: closed-loop precoding restricted to a single layer (rank 1) with a PMI-selected beam from the codebook. It boosts SINR without multiplexing — one steered stream. Feedback is PMI + CQI. DCI 1B carries the precoding vector index and a confirmation bit. Use case: coverage-limited but low-mobility UEs where a single well-aimed beam does more good than trying to run two shaky layers.

The DMRS Modes: TM7–TM10

From Release 9 onward, LTE moved to UE-specific reference signals so the eNB could beamform freely, without a codebook and without telling the UE the precoder. These are the DMRS-based modes, and each one relaxes a constraint of the era before it.

• TM7 — single-layer beamforming. Antenna config: one beamformed layer on antenna port 5 (a virtual, UE-specific port formed from many physical antennas). RS: UE-specific DMRS on port 5. DCI: DCI format 1 (or 1A). The eNB applies any beam — often derived from uplink channel estimates via TDD reciprocity — and the UE demodulates against port 5's pilots, needing no PMI (feedback is CQI only). Use case: coverage extension and TDD reciprocity-based beamforming with large antenna arrays, in Release 9's first step beyond the codebook.
• TM8 — dual-layer beamforming. Antenna config: up to two layers on antenna ports 7 and 8, each with its own DMRS. RS: UE-specific DMRS (ports 7/8). DCI: DCI format 2B (+ 1A). This adds a second beamformed stream to TM7 — throughput and steering — and can be used for single-user rank-2 or for pairing two UEs (one layer each) as MU-MIMO. Feedback is RI + PMI + CQI when configured. Use case: Release-9 beamforming where the channel supports two layers, doubling the beamformed rate.
• TM9 — up to 8-layer SU/MU-MIMO (Rel-10). Antenna config: up to 8 spatial layers on antenna ports 7–14. RS: DMRS for demodulation and, separately, CSI-RS for channel measurement. DCI: DCI format 2C (+ 1A). This is the big one: splitting the two reference-signal roles is exactly what enables high-order SU-MIMO (up to 8 layers to one UE) and flexible MU-MIMO (layers split across UEs) over the largest arrays of pre-CoMP LTE. Feedback is full RI + PMI + CQI, measured on CSI-RS. Use case: LTE-Advanced peak throughput and dense multi-user scheduling with 8-antenna eNBs.
• TM10 — enhanced multi-point / CoMP (Rel-11). Antenna config: as TM9 (ports 7–14) but generalised across multiple transmission points. RS: DMRS + CSI-RS. DCI: DCI format 2D (+ 1A). It extends TM9 for Coordinated Multi-Point (CoMP): several transmission points cooperate to serve a UE (joint transmission, dynamic point selection, coordinated scheduling/beamforming). To make that work it adds multiple CSI processes (so the UE reports feedback for several point hypotheses) and explicit quasi-co-location (QCL) signalling — carried in DCI 2D's PDSCH RE-mapping and QCL indicator field — so the UE knows which reference signals share channel properties like delay spread and Doppler. Use case: interference-limited cell edges where coordinating neighbouring points beats fighting them.

The Full Table: Every Mode Side by Side

Here is all ten modes at once, with the six things you actually need to recall for each: antenna configuration, reference signal, associated DCI format(s), the transmission scheme, and the use case. Note the pattern in the DCI column — each mode-specific format matches the sophistication of its mode, and DCI format 1A is always available as the fallback.

One reading tip: each mode's feedback follows from its scheme. Codebook-precoded closed-loop modes (TM4, TM5, TM6, TM8) report PMI; open-loop or single-antenna modes (TM1, TM2, TM3, TM7) do not. Multi-layer modes (TM3, TM4, TM8, TM9, TM10) also report rank (RI). Every mode reports CQI, since the eNB always needs to know what modulation and coding rate the channel can bear.

The DCI Format 1A Fallback

The single most important thing that every one of the ten modes shares is the fallback. Whatever mode-specific DCI a UE is configured to expect, it always also monitors DCI format 1A in parallel. This is not an optional extra — it is designed into every TM, and it is the safety net that keeps the UE reachable.

There is a neat consistency here. The fallback scheme is not a random simple mode — it is TM2's transmit diversity (or TM1's single antenna in a one-port cell). So every UE, whatever its configured mode, can always be served with the most robust technique the cell physically supports, via a DCI guaranteed to be decodable. When the eNB doubts the UE's feedback — a fresh rank change, mid-handover, or a rank it does not trust — it simply schedules with DCI 1A and transmit diversity. This is also why DCI 1A serves procedures like Random Access Response scheduling and paging, where the network must reach UEs with no reliable channel state at all.

Summary

A Transmission Mode is one RRC setting (transmissionMode in PDSCH-Config) that fixes, for one UE, how PDSCH uses the antennas, which reference signal demodulates it, which DCI format schedules it, and what feedback the UE reports. The cleanest way to hold all ten in your head is the CRS-vs-DMRS split: TM1–TM6 demodulate against the always-on cell-specific pilots and are therefore stuck with the fixed precoding codebook and PMI signalling, while TM7–TM10 demodulate against precoded UE-specific pilots and so escape the codebook entirely, enabling free-form beamforming and, from TM9, high-order MIMO on the back of the DMRS/CSI-RS role split.

Within each half, map every mode to one of four antenna ideas: diversity (TM2), spatial multiplexing (TM3 open-loop, TM4 closed-loop), single-stream beamforming for SINR (TM6, TM7), and multi-user sharing (TM5, and the MU side of TM8–TM10). Remember that the multi-layer modes carry two codewords via the DCI-2 family, that feedback (RI/PMI/CQI) follows directly from the scheme, and that every mode keeps DCI format 1A monitored as an always-decodable transmit-diversity escape hatch. Get those three axes — reference signal, antenna idea, and fallback — and any mode's row in the big table becomes something you can reconstruct rather than memorise.

Interview Quickfire