5G NR Modulation: From π/2-BPSK to 1024QAM, and Why Each One Exists
The modulation schemes NR supports on each link, how the MCS index ties modulation to code rate, and the design reasoning behind π/2-BPSK, 256QAM and 1024QAM.
The mental model: how many bits per resource element
A modulation scheme decides how many bits ride on one resource element. QPSK carries 2, 16QAM 4, 64QAM 6, 256QAM 8, 1024QAM 10. Each step up doubles-plus the constellation density and needs roughly 5–6 dB more SNR to keep the same error rate. So modulation is a ladder the scheduler climbs as the channel gets better and descends as it gets worse — but it never climbs alone: the code rate climbs with it, and the pair is chosen together via the MCS index.
What NR supports, and where
| Scheme | Bits per RE (Q_m) | Downlink | Uplink CP-OFDM | Uplink DFT-s-OFDM | Notes |
|---|---|---|---|---|---|
| π/2-BPSK | 1 | — | — | ✔ | Lowest PAPR; extreme coverage |
| QPSK | 2 | ✔ | ✔ | ✔ | Cell edge, all control channels |
| 16QAM | 4 | ✔ | ✔ | ✔ | |
| 64QAM | 6 | ✔ | ✔ | ✔ | Default top of Table 1 |
| 256QAM | 8 | ✔ | ✔ (UE capability) | ✔ (UE capability) | Table 2; needs high SNR and low EVM |
| 1024QAM | 10 | ✔ (Rel-17, FR1) | — | — | Fixed-wireless / near-cell-centre |
Control channels are fixed: PBCH, PDCCH and PUCCH formats 2/3/4 use QPSK (PUCCH 3/4 may use π/2-BPSK); PUCCH 0/1 use sequences and BPSK/QPSK. Only the shared channels (PDSCH, PUSCH) walk the ladder.
Modulation never travels alone: the MCS index
A 5-bit MCS index selects a row from one of the MCS tables in TS 38.214 (§5.1.3.1 for PDSCH, §6.1.4.1 for PUSCH). The row gives modulation order Q_m and target code rate R. Their product Q_m × R is the spectral efficiency in bits per RE.
Three PDSCH tables exist:
- Table 1 — up to 64QAM, indices 0–28 usable.
- Table 2 — up to 256QAM, indices 0–27 usable; activated by
mcs-Table = qam256for DCI 1_1 with C-RNTI. - Table 3 — up to 64QAM but with lower code rates ("low spectral efficiency"), for URLLC reliability; selected via MCS-C-RNTI.
Rel-17 adds a 1024QAM table (Table 5.1.3.1-4) with a matching CQI table. The uplink has parallel tables, plus transform-precoding variants whose lowest rows are π/2-BPSK.
The tables deliberately overlap at each modulation boundary: a high-rate QPSK row and a low-rate 16QAM row have nearly the same efficiency, so the scheduler can pick whichever survives the channel better.
Why each scheme is there
QPSK — the robust floor. Every control channel uses it because control must decode at the cell edge; the constellation points are far apart and the amplitude is constant, so it tolerates poor SNR and phase error.
16QAM / 64QAM — the LTE-era workhorses. Nothing new in NR; the LDPC code behind them is what changed.
256QAM — was optional in LTE, standard in NR (and in the uplink too). It pays off only when SNR is high and the transmitter's EVM is low; the spec tightens the EVM requirement to 3.5 % for 256QAM (versus 8 % for 64QAM). Field results consistently show a clear gain when the UE is static and near the cell, marginal gain and higher BLER when moving. It is a cell-centre tool.
1024QAM (DL, Rel-17) — 10 bits per RE, EVM ≤ 2.5 %. Targets fixed wireless access and very short links where SNR > 30 dB is realistic.
π/2-BPSK (UL, DFT-s-OFDM only) — the interesting one. Plain BPSK with DFT-s-OFDM already has low PAPR; rotating every other symbol by 90° removes the 180° phase transitions that cause the envelope to pass through zero. The result is a nearly constant-envelope signal, so the UE's power amplifier can run right at its rated power with almost no back-off — a gain of a few dB of effective transmit power for coverage-limited UEs. Combined with optional frequency-domain spectral shaping, it is the lowest-PAPR waveform NR offers. It exists purely for coverage, at 1 bit per RE.
Spectral efficiency at a glance
| Scheme | Max bits per RE at R ≈ 0.93 | Rough SNR needed (AWGN, 10 % BLER) |
|---|---|---|
| QPSK | ~1.9 | ~ 3 dB (at R≈0.5) to ~9 dB |
| 16QAM | ~3.7 | ~ 12–15 dB |
| 64QAM | ~5.6 | ~ 20–22 dB |
| 256QAM | ~7.4 | ~ 27–30 dB |
| 1024QAM | ~9.3 | > 33 dB |
These are order-of-magnitude figures to build intuition; real thresholds depend on channel, receiver and BLER target.
Common confusions
- "π/2-BPSK is a rotated QPSK." The constellation picture looks like QPSK, but each symbol carries one bit and uses alternating BPSK sets; that alternation is what suppresses the envelope dips. It is only defined with transform precoding (DFT-s-OFDM).
- "Higher modulation = higher throughput." Only if the SNR supports it. Pushing 256QAM on a moving UE mostly raises BLER and retransmissions.
- "The MCS index is the modulation order." It is a row selecting modulation and code rate.
- "256QAM is only for the downlink." NR supports it on PUSCH too, subject to UE capability.
- "Control channels use adaptive modulation." They do not; they adapt via aggregation level / repetition instead.
Spec pointers
- TS 38.211 §5.1 — modulation mappers (π/2-BPSK, QPSK, 16/64/256/1024QAM)
- TS 38.214 §5.1.3.1 — PDSCH MCS tables; §6.1.4.1 — PUSCH MCS tables
- TS 38.101-1 §6.4.2 — UE EVM requirements per modulation; TS 38.104 §6.5.2 — BS EVM
- TS 38.212 — channel coding (LDPC for data, Polar for control)
Recap
Modulation is bits per RE: 1 (π/2-BPSK, uplink DFT-s-OFDM only) to 10 (1024QAM, downlink, Rel-17). It is always chosen together with a code rate via the MCS index and one of three tables. QPSK is the control-channel and cell-edge floor; 256QAM and 1024QAM are cell-centre, low-EVM tools; π/2-BPSK trades bits for a constant envelope and a few dB of coverage.
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
- QPSK, 16QAM, 64QAM and 256QAM, with 1024QAM added in Release 17 for the downlink.
- π/2-BPSK, used with DFT-s-OFDM. Its low peak-to-average power ratio lets a power-limited UE transmit closer to its maximum output, which buys coverage.
- π/2-BPSK 1, QPSK 2, 16QAM 4, 64QAM 6, 256QAM 8, 1024QAM 10.
- A single index that jointly selects the modulation order and the target code rate, from which the transport block size is derived.
- They trade against each other for a given SNR. Combining them into one index keeps the DCI small and constrains the scheduler to combinations that are actually sensible.
- Rotating each successive symbol by π/2 avoids transitions through the origin, which is what reduces the peak-to-average power ratio.
- A higher SNR, because the constellation points sit closer together. Roughly 6 dB more per doubling of constellation size.
- It needs an SNR in the mid-thirties of dB together with very good phase noise and EVM performance. Those conditions occur mainly in fixed wireless or very close to the cell.
- Spectral efficiency is modulation order times code rate. Modulation sets the ceiling; the code rate determines how much of that ceiling is actually delivered.
- From CQI reported by the UE, adjusted by outer-loop link adaptation that tracks the observed HARQ acknowledgement rate against a target.
- Bias in the UE's CQI reporting and channel ageing between report and transmission. It nudges the effective SNR offset so the block error rate holds near target.
- Separate tables for 64QAM, 256QAM and low-spectral-efficiency operation. The low-SE table exists for URLLC, where very low code rates matter more than peak throughput.
- Its target block error rate is around 10^-5 rather than 10^-1, which requires code rates far lower than the standard tables provide.
- No. If the SNR cannot support it the block fails, HARQ retransmits, and effective throughput falls. The right choice is the highest order that the channel sustains.
- Error vector magnitude bounds how tightly constellation points can be distinguished. Transmitter EVM sets a practical ceiling on usable modulation order regardless of path loss.
- π/2-BPSK is only available with DFT-s-OFDM. Choosing CP-OFDM in the uplink forfeits that coverage option.
- Optimistic CQI reporting, channel ageing, or interference that appeared after the report. Outer-loop adaptation should correct it, so a persistent gap points at reporting configuration or mobility.
- Each step up the ladder doubles the constellation and adds roughly 6 dB to the SNR requirement, so modulation order is a direct exchange of robustness for bits per resource element.
Related Articles
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.
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.