5G NR Waveforms: CP-OFDM and DFT-s-OFDM, and When the Uplink Should Use Which
Why NR uses CP-OFDM everywhere but keeps DFT-s-OFDM as an uplink option, what PAPR really costs, and how the network chooses per UE.
Two waveforms, one difference
Both NR waveforms are OFDM with a cyclic prefix. The only difference is one optional block in the transmitter: a DFT applied to the modulation symbols before they are mapped to subcarriers. Without it you have CP-OFDM; with it you have DFT-s-OFDM (the spec calls the block transform precoding; LTE called the result SC-FDMA). That one block changes the shape of the transmitted signal in a way that matters enormously for a battery-powered transmitter.
The mental model: many voices vs one voice
CP-OFDM sends data on hundreds of subcarriers in parallel. Each subcarrier is a separate "voice" with its own random amplitude and phase; when they add up, occasionally they all peak together, producing a signal whose peaks are 10–12 dB above its average — a high peak-to-average power ratio (PAPR). Advantages: every subcarrier is independent, so per-subcarrier equalisation and MIMO are simple and spectral efficiency is maximal.
DFT-s-OFDM first spreads each modulation symbol across all the allocated subcarriers with a DFT. The result behaves like a single-carrier signal: at any instant you are essentially transmitting one modulation symbol, so the envelope varies only as much as the constellation does. PAPR drops by roughly 2–3 dB versus CP-OFDM (and much more with π/2-BPSK). The cost: the subcarriers are no longer independent, which complicates MIMO (single layer only in Rel-15) and makes frequency-selective scheduling less natural.
Why PAPR is the whole argument
A power amplifier is linear only up to a point. If the signal's peaks exceed that point they are clipped, causing distortion (EVM) and spectral regrowth (adjacent-channel interference). To keep the peaks inside the linear region, the transmitter must reduce its average power by roughly the PAPR — this is back-off. Every dB of back-off is a dB less transmit power, which is a dB less uplink coverage or a dB more battery drain.
- A gNB has a large, expensive, efficiently cooled PA; it can afford back-off and gains more from CP-OFDM's flexibility and MIMO. Hence downlink: CP-OFDM only.
- A UE at the cell edge is power-limited; 2–3 dB matters more than MIMO it cannot use anyway. Hence uplink: DFT-s-OFDM available.
- A UE near the gNB is not power-limited and wants throughput and MIMO. Hence uplink: CP-OFDM also available.
NR is the first 3GPP system to offer both in the uplink and let the network choose per UE.
The mechanics
In the transmitter chain (TS 38.211 §6.3.1):
bits → scrambling → modulation → layer mapping → [transform precoding] → precoding → RE mapping → IFFT + CP
Transform precoding, when enabled, takes M modulation symbols (M = 12 × number of allocated RBs) and applies an M-point DFT. The spec restricts M to values of the form 2^a·3^b·5^c so the DFT is efficient. The output is mapped to M contiguous subcarriers (contiguous allocation is required in Rel-15; Rel-16 adds a limited interleaved option for unlicensed spectrum).
Enabling it is a per-UE RRC choice, with one dynamic hook:
msg3-transformPrecoder(SIB1): whether Msg3 during random access uses DFT-s-OFDM — the UE has no dedicated configuration yet, so this is cell-wide.transformPrecoderin PUSCH-Config: enabled / disabled for scheduled PUSCH.- With DCI 0_0 the UE follows msg3-transformPrecoder; with DCI 0_1 it follows the dedicated setting.
Consequences of enabling it:
| CP-OFDM (UL) | DFT-s-OFDM (UL) | |
|---|---|---|
| Layers | 1–4 | 1 |
| Allocation | any (type 0 or 1) | contiguous, RB count 2^a 3^b 5^c |
| Modulation | QPSK–256QAM | π/2-BPSK, QPSK–256QAM |
| DM-RS sequence | Gold sequence | Zadoff-Chu-based (low PAPR too) |
| PAPR | high | low |
| Typical use | cell centre, MIMO | cell edge, coverage, small devices |
How a network decides
There is no standardised rule; typical implementations switch a UE to DFT-s-OFDM when its power headroom report shows it is at or near maximum power (or its uplink SINR falls below a threshold), and back when headroom returns. Some operators enable DFT-s-OFDM for all UEs in coverage-challenged bands; others leave CP-OFDM as default and use DFT-s-OFDM only for Msg3.
What changed from LTE
LTE: downlink OFDM only, uplink SC-FDMA only — a rigid choice made when UE PAs were weaker. NR: CP-OFDM as the common baseline for both directions (so uplink and downlink processing are symmetric, and UL MIMO works naturally), with DFT-s-OFDM retained as an option for exactly the case it was designed for. Same technology as LTE's SC-FDMA; different role.
Common confusions
- "DFT-s-OFDM is used in the downlink." Never in NR.
- "DFT-s-OFDM is a single-carrier system, not OFDM." It is OFDM with a DFT spread; it keeps the CP, the subcarrier grid and the FFT receiver.
- "Low PAPR means higher capacity." Spectral efficiency is essentially equal; DFT-s-OFDM wins on coverage/power, loses on MIMO and scheduling flexibility.
- "π/2-BPSK works with CP-OFDM." It is defined only with transform precoding.
- "The UE chooses the waveform." The network configures it; the UE reports capability.
Spec pointers
- TS 38.211 §5.3 — OFDM baseband signal generation; §6.3.1.4 — transform precoding; §6.4.1.1 — PUSCH DM-RS with/without transform precoding
- TS 38.214 §6.1.3 — UE procedure with transform precoding
- TS 38.331 — transformPrecoder, msg3-transformPrecoder
- TS 38.101-1 §6.2 — UE output power and maximum power reduction (MPR) per waveform/modulation
Recap
One optional DFT before the IFFT turns CP-OFDM into DFT-s-OFDM. CP-OFDM: maximal flexibility and MIMO, high PAPR — used for all downlink and for capable uplinks. DFT-s-OFDM: single-layer, contiguous, low PAPR — uplink only, for coverage-limited UEs and Msg3, and the home of π/2-BPSK. The network picks per UE, typically on power headroom.
Interview questions
The questions below are drawn from this topic and phrased the way they tend to come up. Try each one out loud before revealing the answer.
18 questions
- CP-OFDM in the downlink always. In the uplink, either CP-OFDM or DFT-s-OFDM, configured by the network.
- DFT-s-OFDM applies a DFT — transform precoding — to the modulation symbols before mapping them to subcarriers. Everything after that stage is identical.
- It spreads each modulation symbol across all allocated subcarriers, so the time-domain signal resembles a single-carrier waveform with a substantially lower peak-to-average power ratio.
- The UE's power amplifier must back off from saturation by roughly the PAPR to stay linear. Lower PAPR means less back-off, more mean transmit power, and therefore more coverage.
- On the order of 2 to 3 dB, which translates directly into uplink link budget at cell edge.
- Only contiguous allocations are possible, MIMO is limited to a single layer, and frequency-selective scheduling within the allocation is lost because each symbol is spread across the whole allocation.
- Near the cell centre, where power is not the constraint and the extra spectral efficiency, multi-layer MIMO and flexible allocation are worth more than link budget.
- At cell edge, for coverage-limited devices, and for RedCap or IoT devices where transmit power is the binding constraint.
- It is configured semi-statically by RRC per bandwidth part, not switched per slot, because it changes the whole transmit chain.
- LTE used SC-FDMA — essentially DFT-s-OFDM — for the uplink with no alternative. NR made CP-OFDM the default and DFT-s-OFDM the option, reversing the emphasis.
- Uplink MIMO and flexible scheduling matter more in NR, and modern power amplifiers plus smaller cells make the PAPR penalty less decisive than it was.
- π/2-BPSK, which lowers PAPR further still and is the deepest coverage option in the uplink.
- Yes. The gNB must apply an inverse DFT after equalisation to recover the modulation symbols, which is a modest additional processing step.
- Spreading symbols across subcarriers destroys the per-subcarrier independence that spatial multiplexing relies on for separating layers.
- Not at all structurally — both waveforms add a cyclic prefix in the same way. The DFT sits before subcarrier mapping, well ahead of that stage.
- The base station is not power-limited in the same way and needs multi-layer MIMO and frequency-selective scheduling. The PAPR advantage is not worth those losses.
- The 3GPP name for the DFT stage. Configuring transform precoding as enabled is what selects DFT-s-OFDM.
- CP-OFDM gives flexibility and MIMO, DFT-s-OFDM gives 2 to 3 dB of uplink coverage, and the network picks per bandwidth part according to whether the link is power-limited.
Related Articles
5G NR Transmitter and Receiver Chain: Following the Bits From MAC to Antenna and Back
A block-by-block walk through the NR physical-layer transmit chain and its mirror-image receiver, explaining what each stage does and why it sits where it does.
5G NR Synchronisation Signals (PSS and SSS): How a Phone Finds a Cell From Nothing
PSS and SSS explained from the UE's point of view: what each signal gives the receiver, how the PCI is built from them, and why NR has 1008 cell identities.
5G NR SS/PBCH Blocks and Bursts: Beams, Timing and the 5 ms Window
How PSS, SSS and PBCH are packaged into one SSB, how multiple SSBs form a burst set for beam sweeping, and the Case A–E timing patterns.