Low-Power Wake-Up Signal (LP-WUS)

• Low-Power Wake-Up Signal (LP-WUS) is a control primitive that uses a dedicated low-power receiver to monitor for wake-up triggers while the main radio remains off.
• It employs techniques such as OOK modulation, Manchester decoding, and OFDM enhancements to achieve low-complexity detection and significant energy savings.
• Standardized in 3GPP and IEEE frameworks, LP-WUS demonstrates practical gains including up to 99% delay reduction and 80% power savings in optimized network scenarios.

Searching arXiv for recent LP-WUS and wake-up receiver papers to ground the article in the current literature. arXiv query: "Low-Power Wake-Up Signal 3GPP wake-up receiver IEEE 802.11ba wake-up radio" Low-Power Wake-Up Signal (LP-WUS) denotes a low-power wake-up signal, wake-up frame, wake-up call, or lightweight downlink indication that is continuously monitored by a dedicated low-power receiver while the main communication radio remains asleep. Across wake-up radio (WUR) IoT networks, UWB-based on-demand localization, IEEE 802.11ba wake-up operation, and 3GPP NR Release 18 and 5G-Advanced Release 19, LP-WUS has the same architectural purpose: it moves channel monitoring and wake-up triggering onto an ultra-low-power path and reserves the high-power radio for the short intervals in which data, paging, or ranging actually occur. Reported outcomes span up to 90–99% delay reduction in sparse massive-access IoT polling, anchor average power of approximately $15.53\,\mu\mathrm{W}$ with $5.01$ years of operation in an on-demand RTLS scenario, and up to $80\%$ power savings for NR UEs in $\mathrm{RRC\_IDLE}$ and $\mathrm{RRC\_INACTIVE}$ (Deshpande et al., 2023, Cortesi et al., 29 Apr 2025, Wagner, 14 Jul 2025).

1. Concept and architectural placement

LP-WUS appears under different names but with a closely related function. In WUR-enabled IoT polling, it is the wake-up frame sent by a gateway to solicit uplink data from battery-powered sensors; in WakeLoc, LP-WUS corresponds to the wake-up call (WuC) emitted by an active tag to wake anchors and passive tags; in 5G-Advanced Release 19, LP-WUS is a low-power OOK-modulated downlink signal detected by a low-power receiver (LR), which then activates the main radio (MR) for paging reception (Deshpande et al., 2023, Cortesi et al., 29 Apr 2025, Wagner, 14 Jul 2025).

Context	LP-WUS form	Wake-up target
WUR-enabled IoT polling	Wake-up frame over WUR	PCR and main MCU of sensors
WakeLoc RTLS	Wake-up call (WuC)	Anchor MCU/UWB transceiver and passive-tag RX
NR Rel-19	OOK-modulated downlink LP-WUS	UE main radio from IDLE/INACTIVE

In all three cases, the always-on path is deliberately simpler than the main data plane. WUR-enabled IoT assumes that sensors keep the high-power Primary Communication Radio (PCR) and main MCU asleep while the WUR continuously listens. WakeLoc places the wake-up functionality in a nano-/microwatt UWB WuR ASIC model compatible with IEEE 802.15.4-2011 wake-up protocol and supports addressing. NR Rel-19 similarly assigns continuous or duty-cycled monitoring to an LR that performs energy detection or simple coherent detection rather than full I/Q processing. This common partitioning is the defining systems property of LP-WUS (Deshpande et al., 2023, Cortesi et al., 29 Apr 2025, Wagner, 14 Jul 2025).

A recurrent design consequence is that LP-WUS often carries very limited information but strongly constrains the timing of the subsequent high-power procedure. In NR Rel-19, LP-WUS carries at most $B_{\max}=5$ bits to wake UE subgroups or “all UEs,” while LP-SS provides coarse synchronization and measurements. In WakeLoc, the WuC carries only the wake-up signaling and the DW3000 UWB data path carries ranging and localization packets. In multicast WUR IoT, the LP-WUS may wake a unicast target or an entire multicast group, after which data contention or scheduled retransmission resolution occurs on the PCR (Wagner, 14 Jul 2025, Cortesi et al., 29 Apr 2025, Deshpande et al., 2023).

2. Waveform and detection principles

A central LP-WUS design objective is compatibility with very low-complexity detection. Release 19 therefore adopts On-Off Keying (OOK) applied as Multi-Carrier OOK within NR OFDM. The LR operates on received OOK symbols $y_i$, forms symbol energies

$e_i = \|y_i\|^2 = y_i y_i^H,$

and performs Manchester decoding via pairwise energy comparison: $$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$ The same design includes channel coding for $B \le 5$, Manchester line coding, and Zadoff–Chu-based ON-sequences over $5.01$0 subcarriers. Exact false-alarm and detection probabilities are expressed through central and noncentral chi-square CDFs, and Gaussian approximations are also provided for large $5.01$1 (Wagner, 14 Jul 2025).

Release 18 studied a broader waveform set. The study compared OOK-based and OFDM-based LP-WUS, recommended LP-WUS bandwidth $5.01$2 MHz for Idle/Inactive, and evaluated payload sizes from $5.01$3 to $5.01$4 bits. Under the reported link-budget assumptions, OFDM WUS coverage is at least $5.01$5 dB better than OOK WUS, while OOK can require much larger resource overhead to achieve comparable coverage. In the idle-mode scenario described in the study, OFDM WUR overhead is approximately $5.01$6–$5.01$7, whereas OOK WUR overhead ranges from approximately $5.01$8 to $5.01$9, partly because OOK may require an additional periodic LP-SS and longer WUS durations (Hoglund et al., 2024).

IEEE 802.11ba addresses a related but distinct problem: a $80\%$0 MHz OOK WUS, optionally multiplexed across multiple $80\%$1 MHz channels, must remain compatible with low-power envelope detection while controlling instantaneous power swings. Complementary-sequence and Golay complementary pair constructions were proposed for frequency-division multiplexed wake-up signals, reducing worst-case multi-channel PAPR by more than $80\%$2 dB, and by up to $80\%$3 dB relative to draft examples, while improving error-rate robustness under severe PA distortion (Sahin et al., 2019). Sequence-based OOK construction over contiguous OFDM subcarriers was also used to maintain strict orthogonality with OFDM data, improve WURx performance in fading channels, and remove the interference floor at the OFDM receiver (Sahin et al., 2018).

Recent 5G LP-WUS signal design work based on DFT-s-OFDM reaches a different conclusion about waveform shaping. Rectangular-like OOK waveforms are described as aesthetically pleasing but not optimal for 5G LP-WUS scenarios because of limited robustness to channel frequency-selectivity and timing offset. The reported alternative is to shape the OOK spectrum and concentrate OOK symbol energy; under the NR-like setup studied in that work, frequency repetition gives strong BER gains up to about three repetitions, and “concentrated” OOK largely recovers the roughly $80\%$4 dB loss at $80\%$5 BER caused by a $80\%$6 timing offset (Pitaval et al., 2024). This suggests that LP-WUS design is not reducible to “simplest possible OOK,” but depends on whether the limiting impairment is false alarms, fading diversity, timing drift, PA nonlinearity, or receiver power.

3. LP-WUS as a control-plane primitive in WUR-enabled IoT

In WUR-enabled IoT, LP-WUS serves not merely as a wake-up impulse but as the downlink control primitive that determines how massive access is organized. A unicast TDMA-like policy uses one node per LP-WUS slot and is collision-free, but it wastes time polling idle nodes when traffic is sparse. The multicast alternative sends one LP-WUS to a group $80\%$7, reducing idle polling at the cost of possible collisions. With per-node arrival rates $80\%$8 and elapsed times $80\%$9, the group outcome probabilities are

$\mathrm{RRC\_IDLE}$0

$\mathrm{RRC\_IDLE}$1

$\mathrm{RRC\_IDLE}$2

The gateway sorts nodes by activation probability $\mathrm{RRC\_IDLE}$3, initializes a group with the highest-$\mathrm{RRC\_IDLE}$4 node, and adds low-$\mathrm{RRC\_IDLE}$5 nodes until the collision threshold $\mathrm{RRC\_IDLE}$6 is reached (Deshpande et al., 2023).

The baseline unicast model makes the polling overhead explicit. If $\mathrm{RRC\_IDLE}$7 packets are transmitted in one cycle over $\mathrm{RRC\_IDLE}$8 nodes, the cycle time is

$\mathrm{RRC\_IDLE}$9

and under Poisson arrivals the mean cycle time becomes

$\mathrm{RRC\_INACTIVE}$0

which requires $\mathrm{RRC\_INACTIVE}$1 for stability. The average delay for node $\mathrm{RRC\_INACTIVE}$2 is

$\mathrm{RRC\_INACTIVE}$3

These relations formalize why sparse traffic penalizes deterministic polling: the fixed term $\mathrm{RRC\_INACTIVE}$4 dominates (Deshpande et al., 2023).

Collision resolution after multicast LP-WUS is handled either by linear search (LS), in which the gateway polls each node individually, or by binary search (BS), in which the colliding group is recursively split into subgroups of similar aggregate activation probability. LS has $\mathrm{RRC\_INACTIVE}$5 worst-case behavior, whereas BS resolves sparse collisions in $\mathrm{RRC\_INACTIVE}$6 rounds. Monte Carlo evaluation with $\mathrm{RRC\_INACTIVE}$7, $\mathrm{RRC\_INACTIVE}$8 ms, $\mathrm{RRC\_INACTIVE}$9 ms, and $B_{\max}=5$0 iterations per scenario reported up to $B_{\max}=5$1 delay reduction for $B_{\max}=5$2 and up to $B_{\max}=5$3 for $B_{\max}=5$4 versus both ALOHA and unicast WUR in light traffic; more than $B_{\max}=5$5 of collisions were solved within $B_{\max}=5$6 rounds for both $B_{\max}=5$7 and $B_{\max}=5$8; and multicast energy was “never more than twice” TDMA in the tested regimes (Deshpande et al., 2023).

These results delimit the useful operating region of LP-WUS multicast. The strongest gains appear in sparse traffic and large-$B_{\max}=5$9 settings, especially for $y_i$0 packets/$y_i$1. As load increases, the grouping heuristic shrinks groups and effectively falls back toward unicast polling. A plausible implication is that LP-WUS is most valuable when the main inefficiency is not data transmission itself but the control-plane cost of discovering which nodes have data.

4. LP-WUS in NR Release 18 and 5G-Advanced Release 19

In NR, LP-WUS is embedded into a larger wake-up procedure tied to paging and DRX. Release 19 defines LP-WUS Occasions (LO), each associated with $y_i$2 paging occasions, and Monitoring Occasions (MO), with $y_i$3 MOs per LO. UEs in $y_i$4 and $y_i$5 monitor LP-SS and LP-WUS on the LR while the MR remains off; after subgroup-matched LP-WUS detection, the LR triggers MR startup, the MR acquires SSB and/or LP-SS for measurements, and paging is decoded at the associated PO. Every UE belongs to a subgroup, and LP-WUS carries up to $y_i$6 bits, allowing up to $y_i$7 subgroups plus “all” (Wagner, 14 Jul 2025).

LP-SS complements LP-WUS by supporting LR-based measurements and coarse synchronization. Its periodicity is configurable to $y_i$8 ms or $y_i$9 ms, and it spans $e_i = \|y_i\|^2 = y_i y_i^H,$0 OFDM symbols per configured beam. The network configures the LO-to-PO timing offset $e_i = \|y_i\|^2 = y_i y_i^H,$1 based on MR wake-up delay and SSB periodicity; the example given is a $e_i = \|y_i\|^2 = y_i y_i^H,$2 ms MR wake-up delay, for which $e_i = \|y_i\|^2 = y_i y_i^H,$3 ms guarantees at least three SSBs between LO and PO (Wagner, 14 Jul 2025).

Release 18 framed the same problem as a design study and highlighted a sharper waveform trade-off. OFDM-capable WURs consume more power than OOK WURs, but OFDM WUS achieves markedly better coverage and dramatically lower system overhead. The study reports discontinuous WUR monitoring yields more than $e_i = \|y_i\|^2 = y_i y_i^H,$4 UE power saving relative to legacy Idle DRX, while continuous monitoring can consume more energy than the baseline because false alarms trigger expensive MR wake-ups. Under the reported assumptions, network energy overhead for OOK-based WUR can rise to approximately $e_i = \|y_i\|^2 = y_i y_i^H,$5 when LP-SS is frequent, or approximately $e_i = \|y_i\|^2 = y_i y_i^H,$6 when LP-SS is infrequent, whereas OFDM WUR can reuse PSS/SSS in SSB and incur no additional network energy beyond the baseline (Hoglund et al., 2024).

Release 18 waveform work also evolved beyond the initial study item. One comparative analysis of Rel-18 candidates found that full-band OOK, especially OOK-4, outperforms segmented-bandwidth approaches under frequency selectivity; OOK-4 improves over classical OOK-1 by approximately $e_i = \|y_i\|^2 = y_i y_i^H,$7 dB at $e_i = \|y_i\|^2 = y_i y_i^H,$8 BLER for $e_i = \|y_i\|^2 = y_i y_i^H,$9 in AWGN; and a repetition-overlay code can provide approximately $$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$0–$$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$1 dB BLER gain at fixed payload and resource, or substantially higher spectral efficiency by embedding $$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$2 bits across $$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$3 repetitions (Wagner et al., 2023). Standardized LP-WUS in NR is therefore best understood as a family of signaling and scheduling mechanisms, rather than a single waveform frozen at the study stage.

5. Representative systems beyond cellular

WakeLoc shows LP-WUS in an on-demand RTLS rather than a paging architecture. There, the LP-WUS is a UWB wake-up call of $$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$4 ms sent by an active tag. Anchors sleep with WuR on at $$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$5–$$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$6 nW idle and measured baseline power of approximately $$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$7; after WuC detection and a wake transition of approximately $$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$8, they respond at deterministic offsets

$$\hat{f}_k = 0 \text{ if } e_{2k} > e_{2k+1}; \text{ otherwise } \hat{f}_k = 1.$$9

enabling nested CC-SS-TWR with only $B \le 5$0 packets. In the scenario “5 tags, one on-demand localization per minute,” the reported average anchor power is approximately $B \le 5$1, corresponding to approximately $B \le 5$2 years on a $B \le 5$3 mWh coin cell; latency is approximately $B \le 5$4 ms for $B \le 5$5 anchors; and measured 2D errors are $B \le 5$6 cm for the active tag and $B \le 5$7 cm for the passive tag, both below the $B \le 5$8 cm FlexTDOA baseline (Cortesi et al., 29 Apr 2025).

A different LP-WUS use appears in heterogeneous LoRa networks with micro-watt wake-up receivers. Here the cluster head reconfigures an SX1276 to OOK and broadcasts a $B \le 5$9-byte wake-up beacon at $5.01$00 kbps. End devices listen continuously at $5.01$01, decode addresses at $5.01$02, and measure a $5.01$03 ms wake-up delay and $5.01$04 synchronization accuracy from the common beacon arrival instant. The subsequent schedule is deterministic: $5.01$05 with guard time $5.01$06 ms. Experimental results reported $5.01$07 packet delivery ratio, aggregate collection latency below a second across two hops, latency up to $5.01$08 lower than Listen-Before-Talk for nine end devices at the lowest data rate, and up to $5.01$09 years on a $5.01$10 mAh Lithium battery when polled every minute (Piyare et al., 2018).

Legacy Wi-Fi offers a third variant: LP-WUS synthesized from ordinary IEEE 802.11 transmissions rather than a dedicated wake-up PHY. In that design, wake-up information is encoded in frame durations, with $5.01$11, $5.01$12, and $5.01$13 frames generated at $5.01$14 Mb/s in 802.11b. A simple wake-up receiver consisting of an LNA, $5.01$15 MHz channel filter, envelope detector, RC low-pass filter, and $5.01$16-interval sampling estimates burst length with $5.01$17 tolerance. Under the tested conditions, the receiver correctly decoded frame duration with high probability down to at least $5.01$18 dBm and achieved approximately $5.01$19 dB better sensitivity than a CC2420-based 802.15.4 platform, implying about $5.01$20 range increase in a log-distance model with $5.01$21 (Yomo et al., 2012).

These systems demonstrate that LP-WUS is not tied to one radio technology or one service model. It can trigger paging, solicit uplink traffic, start ranging, establish slot timing, or encode compact addresses. The constant architectural pattern is that LP-WUS is a low-rate, low-complexity control signal whose value lies in what it allows the main radio to avoid doing.

6. Trade-offs, misconceptions, and open directions

A common misconception is that LP-WUS is simply “a very short OOK beacon.” The literature is more heterogeneous. Release 18 evaluates OOK-based, OFDM-based, and harmonized designs, and finds OFDM WUS coverage is at least $5.01$22 dB better than OOK WUS in the tested setting, while OOK can have roughly $5.01$23–$5.01$24 higher overhead for similar link budgets. Release 19 standardizes an OOK-based design precisely because it matches LR energy detection, but recent DFT-s-OFDM work argues that rectangular-like OOK is not optimal under frequency selectivity and timing offset (Hoglund et al., 2024, Wagner, 14 Jul 2025, Pitaval et al., 2024).

A second misconception is that lower-power monitoring automatically reduces total energy. The Rel-18 study reports that continuous WUR monitoring can consume more energy than baseline DRX because false alarms repeatedly wake the MR. The multicast IoT study likewise reports that multicast polling consumes more energy than TDMA because non-destined nodes listen to more LP-WUS frames and collision resolution adds LP-WUS transmissions, although energy is “never more than twice” TDMA in the tested regimes. WakeLoc shows the same design tension in another form: the $5.01$25 ms WuC is the dominant per-event cost for the active tag, even though anchor energy collapses because anchors remain asleep until triggered (Hoglund et al., 2024, Deshpande et al., 2023, Cortesi et al., 29 Apr 2025).

The practical importance of false alarms and structured decoding also appears in low-power baseband design. A 65 nm CMOS digital baseband for a structured wake-up beacon, operating with $5.01$26, $5.01$27, $5.01$28, and $5.01$29, consumed $5.01$30 at $5.01$31 kbps and $5.01$32 V, achieved approximately $5.01$33 wake-up beacon detection and $5.01$34 false alarm probability per listen interval, and was used to compensate about $5.01$35 dB of practical sensitivity loss in a $5.01$36 front-end (Mazloum et al., 2016). This suggests that LP-WUS performance is often dominated not only by RF waveform choice but by how much structure the wake-up receiver can exploit at sub-$5.01$37W cost.

Traffic adaptivity is another emerging theme. A forecasting wake-up signaling scheme based on a one-layer, $5.01$38-unit LSTM predicts inter-arrival times and snaps predicted sleep extensions to integer multiples of the base sleep interval $5.01$39. The reported prediction errors are below $5.01$40, false alarm and miss-detection probabilities are below $5.01$41 and $5.01$42, and energy consumption reduction reaches up to $5.01$43 relative to the best benchmark mechanism, while maintaining a $5.01$44 ms mean-delay target (Ruíz-Guirola et al., 2022). This line of work is consistent with the future directions proposed in multicast WUR polling—priority-based LP-WUS scheduling, predictive grouping via machine learning, adaptive collision thresholds, and hybrid LS/BS strategies (Deshpande et al., 2023).

Several limitations recur across the literature. WUR-enabled multicast assumes a pure collision channel, perfect LP-WUS reception, no capture effect, and fixed packet size. WakeLoc assumes ideal WuC detection behavior modeled from a noncommercial ASIC and emulates wake-up via GPIO interrupts in real-world tests. NR Release 19 explicitly does not discuss authentication or spoofing resistance for LP-WUS. These limitations do not negate the LP-WUS concept, but they delimit what current results mean: LP-WUS is already a unifying low-power control mechanism across wireless systems, yet its eventual operating envelope will depend on synchronization robustness, false-wake suppression, coexistence with legacy signals, and how much intelligence can be added to the low-power path without destroying the power budget.