Enhanced ShockBurst (ESB) Protocol Overview

• Enhanced ShockBurst (ESB) is a connectionless 2.4 GHz protocol featuring lightweight packet transmission, explicit ACK/retry control, and high temporal determinism for on-demand sensing and synchronization.
• Studies demonstrate ESB’s ability to achieve sub-millisecond latency and nearly doubled forward throughput compared to BLE, with significantly reduced packet active time and energy consumption.
• ESB supports varied operating modes—such as acknowledged unicast and NOACK broadcast—for applications ranging from implantable sensing to sports synchronization, although it requires additional security measures due to the absence of native encryption.

Searching arXiv for papers on Enhanced ShockBurst to ground the article in current literature. Enhanced ShockBurst (ESB) is Nordic Semiconductor’s proprietary, lightweight, packet-based, connectionless protocol for the company’s 2.4 GHz radios and MCUs. In recent arXiv literature, it is treated not as a general replacement for Bluetooth Low Energy (BLE), but as a protocol family optimized for immediate transmit/receive, low stack overhead, explicit ACK/retry control, and high temporal determinism in workloads such as on-demand sensing, burst-mode telemetry, and low-latency multi-node synchronization (Zhou et al., 8 Apr 2026, Krull et al., 8 Sep 2025, Zhou et al., 20 May 2026).

1. Protocol definition and radio architecture

ESB operates in the 2.4 GHz ISM band on Nordic proprietary PHYs and is supported across Nordic SoCs, including nRF-series devices. The surveyed studies use nRF54L15 and nRF5340 development kits and characterize ESB as connectionless and master-initiated: a transmitter can send immediately after radio initialization, without advertising, discovery, GATT/L2CAP procedures, connection events, or service discovery (Zhou et al., 8 Apr 2026, Zhou et al., 20 May 2026).

At the packet level, ESB frames comprise a preamble, address, header or control information, payload, and CRC. Explicit ACK packets are standalone frames and may carry payload. ESB also exposes programmable addresses, CRC-based payload integrity checks, dynamic payload length, and multiple logical pipes for traffic selection. In the sports-synchronization study, a single ESB data pipe was used for uniform broadcast reception, and no PID use was reported for the tested configurations (Zhou et al., 8 Apr 2026, Krull et al., 8 Sep 2025).

The protocol’s bidirectional semantics differ materially from BLE. In standard acknowledged operation, reverse traffic is carried inside the ACK payload, so the reverse link depends on forward-link activity. The receiver cannot independently initiate a transmission in the same manner as a connected BLE peripheral or central. This dependency is central to later trade-offs in reverse throughput, reliability, and symmetry (Zhou et al., 8 Apr 2026, Zhou et al., 20 May 2026).

PHY support is a major differentiator. One comparative study uses ESB at 4M PHY against BLE at 2M PHY, identifying the higher bit rate as a key driver of lower packet time and energy; another synchronization study profiles ESB at 1 Mbps and 2 Mbps and focuses on 2 Mbps for deterministic command broadcast (Zhou et al., 8 Apr 2026, Krull et al., 8 Sep 2025). Nordic’s 2.4 GHz PHYs are described as GFSK-based at the radio level, although the on-demand sensing paper emphasizes timing and energy rather than modulation details (Zhou et al., 8 Apr 2026).

2. Transaction semantics, packet timing, and operating modes

For acknowledged point-to-point use, ESB is modeled as a forward payload transmission followed by explicit ACK exchange, software processing, and optional automatic retransmission. The on-demand sensing study decomposes packet latency as

$$L_{\text{packet}} = t_{\text{wake}} + t_{\text{sync}} + t_{\text{tx}} + t_{\text{ack}} + t_{\text{processing}},$$

with a useful payload-rate model

$t_{\text{tx}} = \frac{8(H+P)}{R},$

where $H$ is an effective header-plus-overhead term, $P$ is payload bytes, and $R$ is the PHY bit rate (Zhou et al., 8 Apr 2026).

That study further derives an “effective” overhead from measured durations for a 244-byte transaction. For ESB at 4 Mbit/s, the measured radio-active time is approximately $1.28$ ms, giving an effective transmitted size of about $640$ B and hence $H \approx 396$ B. For BLE at 2 Mbit/s, the measured radio-active time is approximately $2.60$ ms, giving an effective transmitted size of about $650$ B and hence $t_{\text{tx}} = \frac{8(H+P)}{R},$0 B. The paper explicitly states that this is not a literal header length; it subsumes radio ramp, ACK spacing, and control overhead (Zhou et al., 8 Apr 2026).

ESB supports several distinct operating modes that materially alter latency and reliability behavior. In acknowledged unicast, explicit ACK is enabled and automatic retransmission uses a fixed microsecond-level delay, reported as $t_{\text{tx}} = \frac{8(H+P)}{R},$1s by default in the nRF54L15 studies (Zhou et al., 8 Apr 2026, Zhou et al., 20 May 2026). In contrast, the sports synchronization work disables ACK entirely in “NOACK” broadcast mode to avoid ACK storms and preserve one-to-many command delivery, while retaining automatic retransmission with two retries and a minimum inter-retry spacing of $t_{\text{tx}} = \frac{8(H+P)}{R},$2s (Krull et al., 8 Sep 2025).

The sports paper also profiles dynamic versus static payload modes, CRC16 versus CRC8 versus CRC off, and automatic versus manual versus manual-start TX modes. Dynamic payload mode yields lower latency than static payload mode, and manual start yields the lowest mean latency and standard deviation among the tested TX modes. To minimize critical-path overhead, the best reported configuration uses a pre-constructed 1-byte payload on the NET-core, eliminating inter-core IPC and runtime formatting from the immediate transmission path (Krull et al., 8 Sep 2025).

These measurements show that “ESB” is not one monolithic operating point. It encompasses acknowledged unicast for energy-efficient, low-overhead data transfer, and NOACK multi-retransmit broadcast for deterministic sub-millisecond synchronization. A plausible implication is that protocol selection within ESB itself is workload-specific rather than universal.

3. Quantitative performance relative to BLE

On the same nRF54L15 platform, ESB and BLE have been compared head-to-head for packet latency, transmission energy, throughput, and duty-cycled wake-up behavior. In the on-demand sensing study, ESB latency scales with payload and is highly deterministic: at 244 B, the reported packet latency is approximately $t_{\text{tx}} = \frac{8(H+P)}{R},$3 ms, and at 2 B it is approximately $t_{\text{tx}} = \frac{8(H+P)}{R},$4 ms. BLE remains around $t_{\text{tx}} = \frac{8(H+P)}{R},$5 ms on average across payloads, with standard deviation greater than $t_{\text{tx}} = \frac{8(H+P)}{R},$6 ms, reflecting connection-event timing jitter (Zhou et al., 8 Apr 2026).

For a single 244-byte packet in that study, ESB has $t_{\text{tx}} = \frac{8(H+P)}{R},$7 ms radio-active time and $t_{\text{tx}} = \frac{8(H+P)}{R},$8J energy, whereas BLE has $t_{\text{tx}} = \frac{8(H+P)}{R},$9 ms radio-active time and $H$0J. The same paper reports maximum effective throughput of approximately $H$1 kbps for ESB and $H$2 kbps for BLE, a power-throughput slope of approximately $H$3 mW/kbps for ESB versus $H$4 mW/kbps for BLE, and standby power of $H$5 mW for ESB versus $H$6 mW for BLE (Zhou et al., 8 Apr 2026).

The hybrid-framework study reports a closely aligned but not numerically identical benchmark on nRF54L15. There, ESB at 4M PHY transmits a 244-byte packet in approximately $H$7s with about $H$8J packet energy, corresponding to about $H$9 nJ/bit; BLE at 2M PHY requires approximately $P$0s and about $P$1J, or about $P$2 nJ/bit. It also reports standby power of $P$3 mW for ESB versus $P$4 mW for BLE, and the same approximate doubling of maximum forward throughput: about $P$5 kbps for ESB and $P$6 kbps for BLE (Zhou et al., 20 May 2026).

The most pronounced differences arise under intermittent, duty-cycled operation. In the on-demand sensing paper, a 10 s sleep-wake cycle with one packet yields BLE warm-up time of $P$7 ms and energy of $P$8J, versus ESB warm-up time of $P$9 ms and energy of $R$0J. The hybrid-framework study reports an even smaller ESB warm-up cost of $R$1 ms and $R$2J, while retaining BLE at $R$3 ms and $R$4J. Although the absolute ESB values differ across the two studies, both report order-of-magnitude reductions in wake-up overhead for ESB relative to BLE (Zhou et al., 8 Apr 2026, Zhou et al., 20 May 2026).

For ultra-low-latency broadcast control, the sports synchronization paper reports even shorter timings. Using 2 Mbps, dynamic payload, CRC off, manual TX, a 1-byte pre-built payload, NOACK broadcast, and two retransmissions, it achieves mean APP-core to APP-core device-to-device latency of $R$5s and mean NET-core to NET-core latency of $R$6s, with an ESB-library pre/post-air interval of $R$7s (Krull et al., 8 Sep 2025).

Taken together, the three studies consistently place ESB in a regime of sub-millisecond to low-millisecond latency, near-halved packet time and energy relative to BLE, approximately doubled forward throughput, and markedly lower wake-up overhead in burst-mode or on-demand operation.

4. Bidirectional behavior, broadcast scaling, and reliability

ESB’s strongest quantitative advantages are on the forward link, but its reverse-link behavior is structurally constrained. In acknowledged mode, reverse data is carried in ACK payloads, so reverse throughput is directly determined by ACK size and by the presence of forward traffic. Because ACK packets are not themselves acknowledged, reverse payloads may be lost (Zhou et al., 8 Apr 2026, Zhou et al., 20 May 2026).

This dependence is visible in measured throughput trade-offs. In the on-demand sensing study, ESB with a 2 B ACK payload reaches forward throughput up to approximately $R$8 kbps while reverse throughput is negligible, with slope $R$9. With a 132 B ACK payload, aggregate throughput is approximately $1.28$0 kbps and partial symmetry appears, with $1.28$1. With a 252 B ACK payload, forward and reverse become nearly symmetric at approximately $1.28$2 kbps and $1.28$3 kbps, aggregate throughput rises to approximately $1.28$4 kbps, and $1.28$5 (Zhou et al., 8 Apr 2026).

The hybrid-framework paper formalizes this trade-off with fitted models:

$1.28$6

where $1.28$7 is ACK payload in bytes. The paper states that these fits reproduce measured saturation points within at most $1.28$8 for $1.28$9 and at most $640$0 for $640$1 (Zhou et al., 20 May 2026).

BLE exhibits a different structure. With a 7.5 ms connection interval, forward and reverse throughput share a fixed capacity with slope approximately $640$2, and aggregate throughput is about $640$3 kbps. In the on-demand sensing benchmark, the measured slope is $640$4 with aggregate throughput around $640$5 kbps; in the hybrid-framework paper, a 100 ms connection interval yields about $640$6 kbps forward and reverse maxima with slope approximately $640$7 (Zhou et al., 8 Apr 2026, Zhou et al., 20 May 2026).

Reliability results are similarly mode-dependent. In broadcast NOACK operation for sports synchronization, CRC16 gives mean D0-D7 latency of $640$8s with $640$9 valid packets, CRC8 gives $H \approx 396$0s with $H \approx 396$1 valid packets, and CRC off gives $H \approx 396$2s but only $H \approx 396$3 valid packets, corresponding to a success rate of $H \approx 396$4 and packet error rate of approximately $H \approx 396$5. In the CRC-off case, one duplicate and fifteen corrupted payloads are observed among received packets (Krull et al., 8 Sep 2025).

For reverse ACK payloads in the hybrid framework, loss is strongly RSSI-dependent: approximately $H \approx 396$6 at about $H \approx 396$7 dBm, approximately $H \approx 396$8 at about $H \approx 396$9 dBm, approximately $2.60$0 at about $2.60$1 dBm, and approximately $2.60$2 above about $2.60$3 dBm. The same paper states that ESB’s 4M PHY is more sensitive to low RSSI and interference than BLE 2M, and throughput degrades faster as RSSI declines from $2.60$4 to $2.60$5 dBm (Zhou et al., 20 May 2026).

For multi-node systems, the sports study uses a shared ESB data pipe and NOACK broadcast, which scales naturally to star or broadcast topologies because one transmitter can push commands to multiple listening nodes without waiting for responses. However, collisions were not evaluated, the study assumes a single transmitter, and it explicitly notes that scheduling or channelization would be required in multi-transmitter scenarios (Krull et al., 8 Sep 2025).

5. Application domains and hybrid system integration

A central application domain in the literature is on-demand sensing, where a device spends most of its time in low-power states and transmits only when an event trigger occurs. The on-demand sensing study implements an implantable loop recorder prototype using a MAX30003 ECG analog front end sampled at 128 Hz, with FIFO interrupts triggering transmission. BLE is kept in continuous connection mode to avoid discovery and connection delays that risk FIFO overflow, whereas ESB is evaluated in both standby and on/off modes (Zhou et al., 8 Apr 2026).

In that prototype, ESB on/off mode wakes on FIFO threshold 31 and returns to system-off sleep after ACK. The reported MCU communication power is approximately $2.60$6 mW at threshold 31, while the ECG sensor AFE averages approximately $2.60$7 mW regardless of FIFO configuration due to fixed sampling rate. Relative to BLE connection mode, ESB on/off yields approximately $2.60$8 total system power reduction while maintaining continuous ECG delivery without overflow (Zhou et al., 8 Apr 2026).

A second application domain is body-worn multi-node synchronization in sports. There the objective is not sustained payload throughput but deterministic command broadcasting suitable for biosignals at 500 Hz to 1000 Hz. The paper argues that NET-to-NET latency of approximately $2.60$9 ms is about $650$0 of a 2 ms period and $650$1 of a 1 ms period, while broadcast delivery to multiple receivers bounds relative skew by the reported jitter of roughly $650$2s standard deviation. On that basis, ESB is positioned as suitable for millisecond- and sub-millisecond alignment of sampling or state changes across devices (Krull et al., 8 Sep 2025).

A third direction is explicit hybridization with BLE. The Enhanced-BLE framework uses Nordic’s Multiprotocol Service Layer (MPSL) to time-division multiplex BLE and ESB on a single transceiver, with BLE connection events given priority and ESB granted timeslots only in idle windows. If a BLE event arrives while ESB is transmitting, MPSL cancels ESB to maintain BLE connection stability (Zhou et al., 20 May 2026).

In the measured hybrid system, ESB transmits immediately after wake while BLE advertising, connection, and service discovery proceed in the background. With a 100 ms advertising interval, the first ESB packet occurs about $650$3 ms after wake, BLE connects at about $650$4 ms, service discovery completes at about $650$5 ms, and then traffic hands over to BLE. Reported transition latencies are approximately $650$6 ms for BLE-to-ESB and $650$7 ms for ESB-to-BLE restoration from standby mode; from shutdown, ESB-to-BLE recovery rises to about $650$8 ms because BLE stack reinitialization dominates (Zhou et al., 20 May 2026).

These case studies show three distinct system roles for ESB: a low-overhead forward telemetry channel for energy-constrained sensing, a deterministic broadcast control plane for synchronized wearables, and a high-efficiency forward subchannel inside a BLE-centered multiprotocol architecture.

6. Trade-offs, limitations, and protocol selection

The most persistent misconception in the recent literature is that ESB is simply a “faster BLE.” The measurements do not support that simplification. ESB’s advantages are strongest when the system values immediate transmission, deterministic timing, fast wake-up, and high forward-link efficiency. BLE retains clear advantages in standardized interoperability, built-in security, and robust symmetric bidirectional communication (Zhou et al., 8 Apr 2026, Zhou et al., 20 May 2026).

Security is the most explicit limitation. ESB has no built-in encryption, authentication, or replay protection in the cited studies; payloads are plaintext, and address matching is the only link-level gating mechanism. BLE, by contrast, integrates AES-128, ECDH Secure Connections, pairing, and key management. Where ESB is deployed in security-sensitive contexts, application-level or system-level protection must be added (Zhou et al., 8 Apr 2026, Zhou et al., 20 May 2026).

Interoperability is similarly constrained. BLE is an open standard with native support on phones and PCs, whereas ESB is proprietary and restricted to Nordic radios with preconfigured endpoints. The literature therefore places ESB primarily in closed, purpose-built links rather than heterogeneous consumer ecosystems (Zhou et al., 8 Apr 2026).

Robustness also depends strongly on link conditions and operating mode. At low RSSI, ESB throughput degrades faster than BLE because 4M PHY is less tolerant of weak SNR. Reverse ACK payloads are especially vulnerable because they are not themselves acknowledged or retransmitted. In NOACK broadcast, disabling CRC lowers latency but measurably increases corruption risk and duplicates, which requires receiver-side de-duplication and idempotent command handling (Krull et al., 8 Sep 2025, Zhou et al., 20 May 2026).

Energy trade-offs are not one-directional across all ESB use cases. For event-driven sensing, ESB’s wake-up advantage dominates and yields very large savings. For always-listening synchronization networks, however, the sports paper notes that keeping the receiver continuously active to achieve deterministic latency trades timing determinism for lower energy efficiency than BLE’s duty-cycled connection events (Krull et al., 8 Sep 2025).

Protocol choice in the cited work is therefore conditional. ESB is favored for sub-millisecond event bursts, fast wake-up, deterministic broadcast, and high payload efficiency with minimal stack overhead. BLE is favored when the system requires smartphone or PC interoperability, standardized services and profiles, integrated security, symmetric bidirectional transfers, or complex connection and state management. The hybrid Enhanced-BLE framework treats these not as mutually exclusive choices but as complementary radio roles on the same hardware (Zhou et al., 8 Apr 2026, Zhou et al., 20 May 2026).