Under steady Wi-Fi interference, a standard Bluetooth Low Energy stack delivered at most 83% of its packets, against about 99% on a clean channel (KU Leuven, EMC Europe 2022). Without adaptive frequency hopping, a separate testbed saw link reliability fall to 79.9% (Poirot and Landsiedel, eAFH, arXiv 2021), a person standing between two radios added a median 17 dB of loss (IEEE 802.15 contribution 15-08-0434, 2008), and a USB 3.0 hard drive raised the 2.4 GHz noise floor by nearly 20 dB (Intel, USB 3.0 RFI white paper 2012). Dropouts are rarely a single failure: they are a link budget eaten from several sides at once, then papered over with buffering. This roundup sits next to our Bluetooth audio market statistics, which covers shipments; here the focus is the radio itself. We aggregated data from the Bluetooth SIG, IEEE 802.15, KU Leuven, Lund University, IEEE Transactions on Consumer Electronics, PLOS ONE, Parks Associates, and peer-reviewed and preprint measurement studies.
TL;DR
- Bluetooth Classic hops across 79 channels of 1 MHz; Bluetooth Low Energy uses 40 channels of 2 MHz (Bluetooth SIG, Bluetooth Technology Overview).
- A single Wi-Fi channel 1 transmission covers BLE data channels 0 to 9 (KU Leuven, EMC Europe 2022).
- No adaptive hopping under constant Wi-Fi: 79.9% link reliability; with jammed channels excluded: 99.8% (Poirot and Landsiedel, eAFH 2021).
- Uncoded LE 1M and LE 2M links hit a packet error rate of 1 under 20 MHz WLAN interference (Lodro et al., arXiv 2021).
- A person 0.3 m in front of the receiver on a 3 m link: median 17 dB extra loss (IEEE 802.15 contribution, 2008).
- Ear-to-ear link loss at 2.45 GHz: 48 dB with antennas in the outer ear, 92 dB in the ear canal (Lund University, 2011).
- Swapping the passengers in a pair of bus seats moved Bluetooth attenuation by about 10 dB (Leith and Farrell, PLOS ONE 2021).
- Classic A2DP music streaming typically carries 100 to 200 ms of latency (Nick Hunn, Introducing Bluetooth LE Audio 2022).
- An external USB 3.0 hard drive raised the 2.4 GHz noise floor by nearly 20 dB (Intel white paper, 2012).
- US internet households averaged 17.3 connected devices in 2024, up from 8.4 in 2015 (Parks Associates, 2025).
1. The Shared 2.4 GHz Band: Channels, Hopping and the Wi-Fi Overlap
Bluetooth’s defense against interference is mobility, not power. A Classic connection can hop up to 1,600 times per second across 79 channels (Rohde and Schwarz, Application Note 1C108, 2016), so a narrowband interferer only hits a fraction of the packets. The weakness is the shape of Wi-Fi: one 20 MHz Wi-Fi channel blankets a whole block of Bluetooth channels at once, which is why a busy router hurts more than a single noisy gadget.
Adaptive frequency hopping (AFH: the radio marks channels it finds noisy as “bad” and stops using them) is the standard fix, but it has a floor. A Classic link may not shrink its hop set below 20 channels, so in a band crowded by several Wi-Fi networks there may simply be no clean spectrum left to retreat into. The Rohde and Schwarz AFH application note walks through how the hop set shrinks and recovers.
| Metric | Value | Source |
|---|---|---|
| Frequency band Bluetooth uses (2.4 GHz ISM band) | 2400 to 2483.5 MHz | Bluetooth SIG, Understanding Bluetooth Range |
| Bluetooth Classic (BR/EDR) channels | 79 channels, 1 MHz spacing | Bluetooth SIG, Bluetooth Technology Overview |
| Bluetooth Low Energy channels | 40 channels, 2 MHz spacing (3 advertising, 37 data) | Bluetooth SIG, Bluetooth Technology Overview |
| Maximum hop rate of a Bluetooth Classic connection | 1,600 hops per second (1600 Hz) | Rohde and Schwarz, Application Note 1C108, 2016 |
| Minimum number of channels a Classic link must keep after AFH removes noisy ones | 20 of 79 | Rohde and Schwarz, Application Note 1C108, 2016 |
| Bluetooth Low Energy data channels covered by Wi-Fi channel 1 alone | channels 0 to 9 | KU Leuven (Pang et al.), EMC Europe 2022 |
Context: the Rohde and Schwarz document is a test-equipment application note summarizing the Bluetooth Core Specification, not a product claim. Most recent available data for the hop-rate and channel-floor figures: Rohde and Schwarz, 2016; the underlying specification values have not changed since.
2. Wi-Fi Coexistence: Measured Packet Delivery and Error Rates
The lab numbers show how much of the “Bluetooth is flaky” reputation is really a scheduling problem. Without adaptive hopping, constant Wi-Fi interference left a BLE link at 79.9% reliability; simply excluding the jammed channels restored 99.8% (Poirot and Landsiedel, eAFH, arXiv 2021). One lost packet in five is an audible stutter for streamed audio; one in five hundred can be hidden.
The catch is that the standard algorithm is not always that good. In a peer-reviewed test using the BLE channel selection algorithm #2, a stock stack topped out at 83% packet delivery ratio (PDR: share of packets that arrive) with static Wi-Fi on channel 1, and 82% when the Wi-Fi hopped between channels 1, 6 and 11 (Pang et al., KU Leuven, EMC Europe 2022). Radio mode matters as much as hopping: the coded long-range PHYs (physical-layer modes that add error-correction coding) held up where the faster uncoded modes collapsed (Lodro et al., arXiv 2021).
| Metric | Value | Source |
|---|---|---|
| Link-layer reliability of a BLE link without AFH under constant Wi-Fi interference | 79.9% | Poirot and Landsiedel, eAFH, arXiv 2021 |
| Same link with channels excluded by measured packet delivery ratio | 99.8% | Poirot and Landsiedel, eAFH, arXiv 2021 |
| Reliability under continuous jamming: no AFH vs the eAFH scheme | 62% vs 99.4% | Poirot and Landsiedel, eAFH, arXiv 2021 |
| Maximum packet delivery ratio of a standard BLE stack with static Wi-Fi interference on channel 1 | 83% | KU Leuven (Pang et al.), EMC Europe 2022 |
| Share of packets lost on the same BLE connection with no interference | around 1% | KU Leuven (Pang et al.), EMC Europe 2022 |
| Packet delivery ratio with Wi-Fi switching between channels 1, 6 and 11: standard stack vs modified stack | 82% vs 93% | KU Leuven (Pang et al.), EMC Europe 2022 |
| Packet error rate of uncoded LE 1M and LE 2M modes under 20 MHz WLAN interference, line of sight | 1 (every packet in error) or no packets detected | Lodro et al., arXiv 2021 |
| Worst packet error rate of the coded LE 500K mode under the same interference | 0.333 | Lodro et al., arXiv 2021 |
Context: a packet error rate (PER) of 1 means every packet failed; 0.333 means one in three. The eAFH and Lodro papers are arXiv preprints; the third study was presented at the 2022 International Symposium on Electromagnetic Compatibility (EMC Europe). Classic audio links and BLE data links behave differently, so treat these as measurements of the mechanism, not of any specific headset.
3. Body Blocking, Head Shadowing and the Link Budget
A body in the path is one of the largest single losses a 2.4 GHz link sees indoors. A person standing 0.3 m in front of the receiver on a 3 m link added a median 17 dB of loss, and 34 dB in the rare 5-sigma case (IEEE 802.15 contribution 15-08-0434, Samsung and Seoul National University, 2008). Every 10 dB is a tenfold drop in received power, so a single body can erase most of a modest radio’s safety margin.
True wireless earbuds face the hardest version of the problem: the signal between the left and right bud has to travel around the head. Simulations at 2.45 GHz put that ear-to-ear loss at 48 dB with antennas in the outer ear and 92 dB inside the ear canal, and a realistic anatomical head model lost 30 dB more than the standard SAM phantom (a simplified plastic head shape used for antenna testing) (Chandra and Johansson, Lund University, 2011). Their later IEEE Transactions on Antennas and Propagation model confirmed that leaving out the pinna (the outer ear) underestimates the loss.
| Metric | Value | Source |
|---|---|---|
| Median extra loss from a person standing 0.3 m in front of the receiver, 3 m link at 2.4 GHz | 17 dB | IEEE 802.15 contribution 15-08-0434, 2008 |
| Same geometry, 2-sigma case (exceeded about 4.56% of the time) and 5-sigma case | 24 dB and 34 dB | IEEE 802.15 contribution 15-08-0434, 2008 |
| Median extra loss from a person standing side-on halfway along a 5 m link | 5 dB | IEEE 802.15 contribution 15-08-0434, 2008 |
| Ear-to-ear link loss at 2.45 GHz with antennas in the outer ear (simulation) | 48 dB | Chandra and Johansson, Lund University, 2011 |
| Ear-to-ear link loss with antennas placed in the ear canal (simulation) | 92 dB | Chandra and Johansson, Lund University, 2011 |
| Additional ear-to-ear loss in a realistic anatomical head model compared with the SAM phantom | 30 dB higher | Chandra and Johansson, Lund University, 2011 |
| Minimum receiver sensitivity required by the Bluetooth specification | -70 dBm to -82 dBm, depending on PHY | Bluetooth SIG, Understanding Bluetooth Range |
| Average receiver sensitivity of LE Coded (125K) implementations | -103 dBm | Bluetooth SIG, Understanding Bluetooth Range |
Context: the body-loss figures come from an anechoic chamber (a room lined to absorb reflections) with patch antennas, so real rooms with reflections usually soften the shadow. Most recent available data: IEEE 802.15 contribution, 2008, and Lund University, 2011. The Bluetooth SIG range guide shows why LE Coded matters against losses of this size: an average sensitivity of -103 dBm is 21 dB better than the -82 dBm minimum the specification sets for the slowest coded mode (-103 minus -82).
4. Crowds, Trains, Buses and Walls: Real-World Measurements
The best public field data on Bluetooth propagation in crowded spaces came, unexpectedly, from pandemic contact-tracing research. In a Dublin commuter bus, changing the people sitting in a pair of seats shifted Bluetooth attenuation by about 10 dB (Leith and Farrell, PLOS ONE 2021). For audio, that means the same seat and the same headset can work on one trip and stutter on the next.
Metal vehicles turn distance upside down. In a light-rail tram, handsets under 2 m apart and up to 5 m apart showed similar signal ranges, which the authors attribute to reflections from metal walls, floor and ceiling (Leith and Farrell, PLOS ONE 2020). Where the phone sits on the body matters as much: phones in trouser pockets read around -80 to -90 dB against around -65 dB on a table in the same authors’ earlier domestic and outdoor tests, a gap of 15 to 25 dB (-65 minus -80, and -65 minus -90). Stage and broadcast crews fight the same physics on a larger scale, as our touring wireless microphone spectrum statistics show.
| Metric | Value | Source |
|---|---|---|
| Bluetooth attenuation in the tram at 1.5 m vs at 4 m | around 52 dB vs around 60 dB | Leith and Farrell, PLOS ONE 2020 |
| Bluetooth attenuation on a commuter bus at 1.5 m and at 3.5 m | around 75 dB at both distances | Leith and Farrell, PLOS ONE 2021 |
| Change in attenuation from swapping the people seated in a pair of bus seats | about 10 dB | Leith and Farrell, PLOS ONE 2021 |
| Signal loss through a blockwork (concrete block) wall | around 20 dB | Leith and Farrell, arXiv 2020 |
| Received signal with phones in trouser pockets vs phones on a table | -80 to -90 dB vs around -65 dB | Leith and Farrell, arXiv 2020 |
| Maximum packet error rate without interference, metal sheet blocking the path vs removed | 0.34 vs 0.05 | Lodro et al., arXiv 2021 |
Context: the tram, bus, wall and pocket values are received signal strength figures reported by Android phones (attenuation in dB, as the papers print them), not audio dropout counts. Most recent available data: Leith and Farrell, 2020 to 2021. The metal-sheet result was measured in a reverberation chamber (a metal room designed to create many reflections). No stadium-specific Bluetooth packet-loss study with public figures turned up in our search.
5. Latency and Buffering: How Bluetooth Hides Dropouts
Bluetooth audio links are built to retransmit lost packets, and every retransmission opportunity has to fit inside the delay budget before the audio is played. Robustness and latency are therefore the same dial. Classic A2DP music streaming from a phone or computer typically carries 100 to 200 ms of latency (Nick Hunn, Introducing Bluetooth LE Audio, 2022), fine for a playlist but a long time for live voice or games.
LE Audio rebuilt the trade-off around a fixed presentation delay (a set waiting time before playback, so every earbud plays the same sample at the same moment). Profiles for hearing aids and telephony can demand 20 ms or less, which leaves far fewer chances to retransmit through the interference measured in section 2. Hearing aid users set the tightest bound, because their devices mix streamed sound with sound arriving naturally through the ear; the battery cost of that streaming is covered in our hearing aid Bluetooth streaming statistics, and the gaming side in our rhythm game audio latency calibration data.
| Metric | Value | Source |
|---|---|---|
| Typical latency of Classic A2DP music streaming from a phone or computer | 100 to 200 ms | Nick Hunn, Introducing Bluetooth LE Audio, 2022 |
| Typical overall delay of mSBC-based HFP voice calls | around 30 ms | Nick Hunn, Introducing Bluetooth LE Audio, 2022 |
| Delay between streamed and natural sound at which hearing aid users begin to hear echo | much more than 30 to 40 ms | Nick Hunn, Introducing Bluetooth LE Audio, 2022 |
| Presentation delay that hearing aid (HAP) or telephony and media (TMAP) profiles may require | not more than 20 ms | Nick Hunn, Introducing Bluetooth LE Audio, 2022 |
| Broadcast presentation delay every receiver must support; higher values should be avoided | 40 ms | Nick Hunn, Introducing Bluetooth LE Audio, 2022 |
| Left and right earbud playback alignment needed to keep the stereo image stable | within 25 microseconds | Nick Hunn, Introducing Bluetooth LE Audio, 2022 |
| Length of one Bluetooth Classic time slot | 625 microseconds | Rohde and Schwarz, Application Note 1C108, 2016 |
Context: these figures are design targets and typical values from an introductory book on the LE Audio specifications, not a measured sample of products. Real-world headset latency depends on the phone, the chipset and the codec in use.
6. The PC and the Home: USB 3.0 Noise, Microwave Ovens and Device Density
Some of the worst interference sits a few centimeters from the Bluetooth antenna. An external USB 3.0 hard drive raised the 2.4 GHz noise floor by nearly 20 dB, and a notebook’s own USB 3.0 connector by about 25 dB (Intel, USB 3.0 Radio Frequency Interference Impact on 2.4 GHz Wireless Devices, 2012). The noise floor is the background radio noise a receiver has to hear above, so raising it shrinks usable range much like moving the transmitter farther away. USB 3.0 signals at 5 Gbit/s and its data spectrum runs from DC to 5 GHz, which puts energy straight into the 2.4 to 2.5 GHz range that Bluetooth uses.
Microwave ovens look worse than they are. A Bluetooth piconet (a phone and its connected devices) within 1 m of a residential oven kept throughput above half its maximum, and at about 10 m saw very little degradation (Rondeau, D’Souza and Sweeney, IEEE Transactions on Consumer Electronics, 2004); a later Indiana State University study found the oven’s power setting drove BLE packet loss more than distance. The bigger structural trend is density: the number of devices competing for the same air has doubled in a decade (Parks Associates, 2025).
| Metric | Value | Source |
|---|---|---|
| Rise in 2.4 GHz noise floor with an external USB 3.0 hard drive connected | nearly 20 dB | Intel white paper, 2012 |
| Rise in 2.4 GHz noise floor near a notebook’s USB 3.0 connector | about 25 dB | Intel white paper, 2012 |
| Noise reduction from fully shielding the USB 3.0 hard drive | about 12 dB | Intel white paper, 2012 |
| Wireless mouse range with improved connector shielding and a fully shielded drive | no degradation out to approximately 8 feet | Intel white paper, 2012 |
| Bluetooth throughput with both devices within 1 m of a running microwave oven | more than half of maximum | Rondeau et al., IEEE Transactions on Consumer Electronics, 2004 |
| Average connected devices per US internet household, 2015 vs 2024 | 8.4 vs 17.3 | Parks Associates, 2025 |
| US internet households owning a smart home device, Q2 2014 vs Q4 2024 | 10% vs 45% | Parks Associates, 2025 |
Context: the USB 3.0 tests used a 2.4 GHz wireless mouse, not a Bluetooth headset, but the mechanism (a raised noise floor in the shared band) is the same. Most recent available data: Intel, 2012, and Rondeau et al., 2004. The white paper’s mitigations are still current advice: move 2.4 GHz receivers away from USB 3.0 ports, or onto a USB 2.0 port or extension cable.
Summary: Bluetooth Audio Dropout and 2.4 GHz Interference by the Numbers
| Metric | Value | Source |
|---|---|---|
| Bluetooth Classic channels | 79 channels, 1 MHz spacing | Bluetooth SIG, Bluetooth Technology Overview |
| Bluetooth Low Energy channels | 40 channels, 2 MHz spacing | Bluetooth SIG, Bluetooth Technology Overview |
| Minimum channels a Classic link must keep after AFH | 20 of 79 | Rohde and Schwarz, Application Note 1C108, 2016 |
| BLE data channels covered by Wi-Fi channel 1 | channels 0 to 9 | KU Leuven (Pang et al.), EMC Europe 2022 |
| BLE link reliability without AFH under constant Wi-Fi interference | 79.9% | Poirot and Landsiedel, eAFH, arXiv 2021 |
| Same link with jammed channels excluded | 99.8% | Poirot and Landsiedel, eAFH, arXiv 2021 |
| Maximum packet delivery ratio of a standard BLE stack under static Wi-Fi interference | 83% | KU Leuven (Pang et al.), EMC Europe 2022 |
| Packet error rate of uncoded LE 1M and LE 2M under 20 MHz WLAN interference | 1 (every packet in error) or no packets detected | Lodro et al., arXiv 2021 |
| Median extra loss from a person 0.3 m in front of the receiver, 3 m link | 17 dB | IEEE 802.15 contribution 15-08-0434, 2008 |
| Same geometry, 5-sigma case | 34 dB | IEEE 802.15 contribution 15-08-0434, 2008 |
| Ear-to-ear link loss at 2.45 GHz, antennas in the outer ear | 48 dB | Chandra and Johansson, Lund University, 2011 |
| Change from swapping the people seated in a pair of bus seats | about 10 dB | Leith and Farrell, PLOS ONE 2021 |
| Typical latency of Classic A2DP music streaming | 100 to 200 ms | Nick Hunn, Introducing Bluetooth LE Audio, 2022 |
| Delay at which hearing aid users begin to hear echo | much more than 30 to 40 ms | Nick Hunn, Introducing Bluetooth LE Audio, 2022 |
| Rise in 2.4 GHz noise floor with an external USB 3.0 hard drive | nearly 20 dB | Intel white paper, 2012 |
| Wireless mouse range after shielding fixes | no degradation out to approximately 8 feet | Intel white paper, 2012 |
| Average connected devices per US internet household, 2024 | 17.3 | Parks Associates, 2025 |
Methodology and Sources
- Bluetooth SIG, Bluetooth Technology Overview and Understanding Bluetooth Range: channel plan, frequency band and receiver sensitivity.
- Rohde and Schwarz, Application Note 1C108, Bluetooth BR/EDR Adaptive Frequency Hopping (November 2016): hop rate, time slot and AFH channel floor from the Core Specification.
- KU Leuven, Pang, Claeys, Hallez and Boydens, An Exponential Back-off Algorithm Based Interference Avoidance Strategy for Bluetooth Low Energy against Wideband Interference, EMC Europe 2022.
- Poirot and Landsiedel, eAFH: Informed Exploration for Adaptive Frequency Hopping in Bluetooth Low Energy, arXiv preprint, December 2021.
- Lodro, Gradoni, Vukovic, Thomas and Greedy, PER Measurement of BLE in RF Interference and Harsh Electromagnetic Environment, arXiv preprint, May 2021.
- Kang, Cho, Park, Won (Samsung Electronics) and Kim, Kang, Kim (Seoul National University), Path Loss due to Human Body Between Antennas, IEEE 802.15 contribution 15-08-0434, July 2008.
- Chandra and Johansson, Lund University, Antennas and Propagation for the Ear-to-Ear Propagation Channel for Binaural Hearing Aids, COST IC1004, 2011, and A Link Loss Model for the On-body Propagation Channel for Binaural Hearing Aids, IEEE Transactions on Antennas and Propagation, 2013.
- Leith and Farrell, Trinity College Dublin: light-rail tram study (PLOS ONE, September 30, 2020), commuter bus study (PLOS ONE, April 29, 2021) and Bluetooth received signal strength for proximity detection (arXiv, May 2020).
- Nick Hunn, Introducing Bluetooth LE Audio, first published January 2022.
- Intel, USB 3.0 Radio Frequency Interference Impact on 2.4 GHz Wireless Devices, white paper, April 2012.
- Rondeau, D’Souza and Sweeney, Residential Microwave Oven Interference on Bluetooth Data Performance, IEEE Transactions on Consumer Electronics, vol. 50, no. 3, 2004.
- Barge, Indiana State University, packet loss model for BLE wearables near residential microwave ovens, dissertation, 2015.
- Parks Associates, home networking research excerpt presented at WiFi NOW, April 2025.
- Data watch: several core measurements are old (the USB 3.0 white paper from 2012, the body-loss contribution from 2008, the ear-to-ear simulations from 2011, the microwave oven study from 2004) because the physics has not changed and few newer studies publish open figures; treat them as most recent available data. Household device counts conflict: Parks Associates reports 17.3 connected devices per US internet household in 2024, while NCTA (a US cable trade group crediting Nokia, Omdia and Charter collectively) says the typical household supports around 25; definitions differ, and we cite Parks. The KU Leuven abstract states residual packet loss of around 3% and 5%, while the body reports packet delivery ratios of 95% and 93%; we use the body figures. The eAFH preprint gives slightly inconsistent counts of excluded channels, so we use only its reliability results. Informal hobbyist latency tests of specific headsets were excluded for lack of a documented method. No public stadium or station packet-loss study with numeric results was found. Newer LE Audio and Auracast field measurements are expected as more broadcast deployments open.
Last updated: November 7, 2026. We update this roundup quarterly, and the next refresh is expected when Parks Associates releases its next home networking survey results.