PAPR (Peak-to-Average Power Ratio) in OFDM and OFDMA Wireless Systems Interview Prep

PAPR reduction techniques — trade-offs

Sample interview questions

1. Why does an OFDM signal have higher PAPR than a single-carrier modulation like GSM's GMSK?
   • A. OFDM has higher PAPR because it uses more transmit power than GSM does at the same data rate.
   • B. OFDM has higher PAPR because the cyclic prefix doubles the energy of each symbol relative to its average.
   • C. An OFDM symbol is a sum of many independently-modulated subcarriers; by the central limit theorem the time-domain amplitude approaches a complex-Gaussian distribution and the magnitude follows a Rayleigh-like distribution with long tails — so instantaneous peaks can be 8-12 dB above the average. GMSK is constant-envelope (or nearly so) and has PAPR close to 0 dB. ✓
   • D. OFDM and GMSK have identical PAPR; the difference is purely in spectral efficiency.

   Option C is correct. The defining property of OFDM is that each symbol is a sum of many independently-modulated complex subcarriers. By the central limit theorem, as the number of subcarriers grows, the time-domain amplitude distribution approaches complex Gaussian, and the magnitude follows a Rayleigh-like distribution with long tails. Peaks at 8-12 dB above the average are typical. GMSK, by contrast, is a constant-envelope (or near-constant-envelope) modulation by design — the modulator shapes the phase trajectory so the amplitude stays nearly flat, yielding PAPR close to 0 dB. Option A is wrong because PAPR is a property of the waveform statistics, not the average power level. Option B is wrong because the cyclic prefix does not change the PAPR statistics; it adds redundancy in the time domain but does not alter the subcarrier sum. Option D is wrong because the PAPR difference between OFDM and GMSK is substantial — typically 8-12 dB versus near 0 dB.

2. Why does 5G NR commonly configure the uplink with DFT-spread OFDM (SC-FDMA / transform precoding) for coverage-limited (cell-edge) UEs instead of pure OFDM?
   • A. DFT-s-OFDM is mandated by 3GPP for all uplink transmissions regardless of channel conditions; cell-edge framing is incidental.
   • B. DFT-s-OFDM applies a DFT precoder before the OFDM IFFT, producing a waveform that behaves more like single-carrier with 2-4 dB lower PAPR than pure OFDM. The lower PAPR lets cell-edge mobile UEs run their power amplifiers closer to saturation, improving link budget and battery life — both binding constraints at cell edge. ✓
   • C. DFT-s-OFDM has higher PAPR than pure OFDM but better spectral efficiency, so it is preferred at cell edge where bandwidth dominates.
   • D. DFT-s-OFDM is used only for control channels; the data channel always uses pure OFDM regardless of UE location.

   Option B is correct. DFT-spread OFDM (sometimes called SC-FDMA) inserts a DFT precoder before the IFFT, which collapses the multi-subcarrier sum into a near-single-carrier waveform. The resulting PAPR is 2-4 dB lower than pure OFDM. At cell edge, mobile UEs are PA-power-limited and battery-limited; lower PAPR means the UE can transmit closer to its PA saturation point, recovering link budget and saving battery. 5G NR lets the gNB enable or disable transform precoding (DFT-s-OFDM vs pure OFDM) for PUSCH per UE by configuration; it commonly configures transform-precoded DFT-s-OFDM for coverage-limited cell-edge UEs precisely for this PAPR/link-budget reason. Option A is wrong because 5G NR uplink supports both DFT-s-OFDM and pure OFDM and the gNB picks based on channel state; it is not unconditionally mandated. Option C is wrong because DFT-s-OFDM has LOWER PAPR than pure OFDM — the direction is reversed. Option D is wrong because DFT-s-OFDM is a data-channel option on the uplink, not a control-only mode.

3. A Wi-Fi 7 access point is struggling to sustain 4096-QAM in production despite indoor short-range deployment. What is the most likely PAPR-related root cause?
   • A. 4096-QAM has higher PAPR than 256-QAM because each symbol carries more bits, and the AP cannot meet the higher PAPR budget.
   • B. 4096-QAM requires LDPC decoding and Wi-Fi 7 access points use BCC, so the AP cannot decode 4096-QAM frames.
   • C. 4096-QAM is restricted to 320 MHz channels and most APs operate at 160 MHz.
   • D. 4096-QAM has very small per-symbol Euclidean distance and requires EVM around 1.3 percent rms (about -38 dB) or better. The OFDM PAPR forces PA back-off; if the back-off is not deep enough, residual PA nonlinearity drives the EVM above the 4096-QAM threshold and the modulation falls back to 1024-QAM or 256-QAM. Hitting 4096-QAM in production typically requires deeper back-off, envelope-tracking PA, or digital predistortion. ✓

   Option D is correct. Higher-order QAM shrinks the spacing between constellation points, so the EVM requirement tightens accordingly. The IEEE 802.11be EHT transmitter EVM limit for 4096-QAM (MCS12/13) is about -38 dB (roughly 1.3 percent rms or better) — about 3 dB tighter than the 1024-QAM limit (about -35 dB). The OFDM waveform PAPR forces PA back-off; if the back-off is insufficient, residual AM-AM and AM-PM distortion at the PA pushes EVM above the threshold and rate adaptation falls back. Production 4096-QAM in Wi-Fi 7 typically requires envelope-tracking PA architectures, digital predistortion, and careful baseband EVM budgeting — all working together. Option A is wrong because PAPR is a waveform property of OFDM and is independent of modulation order; the per-symbol bit count does not change the time-domain amplitude statistics. Option B is wrong because Wi-Fi 7 access points support LDPC for high-order MCS (LDPC has been in Wi-Fi since 802.11n); the modulation cannot be decoded with BCC at 4096-QAM but the chipset does support LDPC. Option C is wrong because 4096-QAM is supported across the channel widths defined in IEEE 802.11be, not restricted to 320 MHz.

Frequently asked questions

What is PAPR and why does it matter for OFDM-based wireless systems?

PAPR (Peak-to-Average Power Ratio) is the ratio of the instantaneous peak power of a waveform to its average power, typically expressed in dB. OFDM signals are sums of many independently-modulated subcarriers, so their time-domain amplitudes follow approximately a complex-Gaussian distribution and the instantaneous peak can be 8-12 dB above the average — a much higher PAPR than constant-envelope waveforms like GSM or single-carrier modulations. High PAPR is expensive because the power amplifier (PA) must be backed off from its saturation point to keep the peaks linear, which wastes DC power and reduces average TX power. PAPR is a defining engineering constraint in every OFDM/OFDMA system: Wi-Fi 4 through Wi-Fi 8, LTE, 5G NR, DVB-T, and DOCSIS.

How does PA back-off translate the PAPR cost into real power and efficiency loss?

A power amplifier delivers maximum efficiency near its 1 dB compression point — the input level where output power starts to compress nonlinearly. To transmit a signal with 10 dB PAPR without clipping the peaks (which causes spectral regrowth and ACPR violations), the PA must be backed off by 8-10 dB from saturation. PA efficiency typically falls as a steep function of back-off — a Class AB PA might run at 50 percent efficiency at saturation but only 15-20 percent backed off 8 dB. The wasted power becomes heat, drives larger heat-sinks, and reduces battery life on a phone or sensor. The mobile and access-point industries spend enormous engineering effort reducing PAPR or recovering efficiency at high back-off through Doherty and envelope-tracking PA architectures.

What PAPR-reduction techniques are used in real systems?

Several approaches, each with different trade-offs. Clipping and filtering: hard-limit the time-domain peaks, then filter to suppress the spectral regrowth — simple but introduces in-band distortion (raises EVM). Tone reservation: dedicate a set of subcarriers as cancellation tones whose values are chosen to cancel peaks — adds overhead but no in-band distortion. Selective Mapping (SLM) and Partial Transmit Sequences (PTS): generate multiple candidate waveforms with different phase rotations and pick the lowest-PAPR one — requires side-information signaling. DFT-spread OFDM (SC-FDMA): used on the LTE / 5G NR uplink, applies a DFT precoder before the OFDM IFFT so the resulting waveform behaves more like single-carrier, with 2-4 dB lower PAPR than pure OFDM. The 5G NR uplink uses DFT-s-OFDM as the default at cell edge for exactly this reason; the downlink stays as pure OFDM because the base station PA can afford the back-off.

Why does IEEE 802.11bn introduce Distributed Resource Units (DRU) and what does that do to PAPR?

The Distributed Resource Unit (DRU) in IEEE 802.11bn (Wi-Fi 8) spreads the data subcarriers of an OFDMA Resource Unit across the entire transmission bandwidth instead of grouping them contiguously. The motivation is power-boost coverage gain for far stations — spreading energy across more frequency improves the link budget at the edge of the cell. The PAPR implication is subtle: because the DRU still uses an OFDM waveform with many simultaneously-active subcarriers, the underlying PAPR distribution remains Gaussian-like and high. The DRU does not fundamentally change PAPR; it changes where the energy is in the frequency domain. For the broader OFDMA RU framework, see /topics/ofdma-resource-units-explained. For the 802.11bn mechanism context, see /topics/wifi-8-80211bn-ultra-high-reliability.

How does PAPR interact with PA architecture choices like Doherty and envelope tracking?

Conventional Class AB PAs have a fixed efficiency curve and lose efficiency steeply at back-off. Doherty PAs use a two-amplifier structure (carrier + peaking) that maintains higher efficiency over a wider back-off range — typically 6 dB of efficient back-off versus 0 dB for Class AB. Envelope-tracking (ET) PAs dynamically vary the supply voltage to follow the envelope of the modulated signal, keeping the PA near its efficient operating point even as the instantaneous amplitude swings up and down. ET adds complexity (the envelope-tracking power supply is its own DC-DC converter) but recovers 5-15 percentage points of efficiency at high PAPR. Cellular base-station PAs combine Doherty + digital predistortion (DPD); mobile handset PAs increasingly use ET. The PAPR-to-PA story is one of the dominant cost-and-power drivers in any wireless transmitter design.

What PAPR interview questions catch candidates who do not understand the PA-efficiency story?

The fastest filter is asking why OFDM has higher PAPR than single-carrier modulations. Candidates who say "because there are more subcarriers" are partially right but miss the underlying explanation — the sum of many independent complex sinusoids approaches a complex-Gaussian distribution whose magnitude has a long-tailed Rayleigh distribution, hence the high peak-to-average. A second filter is asking what PA back-off costs in efficiency: strong candidates know that Class AB efficiency falls steeply with back-off (50 percent at saturation, 15-20 percent at 8 dB back-off) and can name Doherty + ET as the mitigations. A third filter is the DFT-s-OFDM choice in 5G NR uplink — candidates who can explain why LTE/NR uplink uses DFT-spread to reduce PAPR for cell-edge mobile UEs are showing system-level reasoning that connects PHY waveform design to PA cost.

How does PAPR interact with EVM and the per-symbol Euclidean distance for high-order MCS?

Tightly. The PA back-off chosen for a given PAPR target determines the residual nonlinearity (AM-AM and AM-PM distortion) the signal experiences; this nonlinearity manifests as EVM at the receiver constellation. Higher-order modulations like 4096-QAM in IEEE 802.11be have very small per-symbol Euclidean distance, so they need very low EVM — typically around 1.3 percent rms or better, i.e., better than -38 dB. Operating a 10 dB PAPR signal close to PA saturation breaks that EVM target. Hitting 4096-QAM in production requires either deeper PA back-off (lower efficiency), more aggressive PAPR reduction, ET-based PA architectures, or digital predistortion — usually all four together. The PAPR-EVM trade-off is the central reason 4096-QAM rates are advertised peaks that real links rarely sustain.

Related topics