The choice that decides whether an audio product works is not a logo on the box — it is a budget. Latency, synchronisation, channel count, fidelity and range: each wireless standard spends that budget differently, and each hits a wall somewhere. Here is where, and how to match the link to the product.
Nearly every complaint about wireless audio comes back to one thing: the sound does not arrive when it should. On headphones it is a game that reacts a beat late; on a television it is lips that move before the words land. Both are the same defect — latency — and ear and eye are far less forgiving of it than most specifications assume. Audio lagging video by more than roughly 45 milliseconds reads as out of sync, and a plain Bluetooth link routinely spends four to five times that budget before the first sample reaches the speaker.
So the disciplined way to choose a link is to write down what the product actually has to deliver. Six parameters do almost all the work: latency — does audio have to track a picture or an action? Synchronisation — must several speakers play the same instant to microsecond accuracy? Channel count — stereo, or discrete 5.1/7.1? Fidelity — is lossy fine, or do you need high‑res and lossless? Range and robustness — on‑body, or across a room through walls and a crowded 2.4 GHz band? And power — a coin cell, or mains? Write those six down and the field narrows to one or two links before you open a single datasheet.
The shift that unlocks the rest: bandwidth is almost never the constraint. Determinism and timing are.
Classic Bluetooth (A2DP) is the universal default and the source of most lip‑sync pain: compressed stereo over SBC or AAC, with an encode–transmit–buffer–decode pipeline that routinely runs 150–250 ms, and no real multichannel or high‑res story.
Bluetooth LE Audio (LC3) is the genuine step forward, built on the Isochronous Channels added in the Bluetooth 5.2 feature set. It cuts latency to the 20–30 ms range, carries a defined presentation‑delay timestamp so multiple sinks render in step, and adds multi‑stream and broadcast (Auracast). For earbuds, headphones and simple multi‑speaker setups it closes most of the old gap. Its ceiling is bandwidth and range.
Buffered Wi‑Fi streaming (Sonos, AirPlay, Chromecast) is the inverse profile: enormous bandwidth, excellent fidelity, rock‑solid multi‑room sync and whole‑home range — but a deliberate buffer that adds hundreds of milliseconds. Superb for music; wrong for anything tied to a picture, unless the video source itself is doing the streaming and can hold the picture back.
Real‑time (tweaked) Wi‑Fi — the WiSA‑class approach — is a Wi‑Fi‑band radio run in a proprietary low‑latency mode: scheduled airtime instead of best‑effort contention, tight cross‑device time sync, a lean real‑time transport and uncompressed multichannel PCM. Latency in the single‑digit milliseconds, speakers locked to each other and to the screen. It is the only one of the four that satisfies the full real‑time, high‑fidelity, room‑scale requirement at once.
The bandwidth gap is easy to quantify. Uncompressed stereo is about 1.4 Mbps at CD quality and 4.6 Mbps at 24‑bit/96 kHz; a 5.1 mix at 24/96 is roughly 14 Mbps and 7.1 around 18 Mbps. Bluetooth LE carries about 2 Mbps on its radio, and even an efficient codec like LC3 sits far below that, so high‑resolution multichannel is a Wi‑Fi job by arithmetic alone. Which is the same point from the other side: once you are on Wi‑Fi, bandwidth stops being the constraint and timing becomes the whole job.
Put those profiles next to real products and a clean gradient appears. The closer the audio sits to a single body, the more Bluetooth and LE Audio win on power and simplicity. The more it has to fill a room — more channels, more distance, tighter sync to a screen — the more the decision crosses over to Wi‑Fi. Note the gaming headset sitting off the main colour: a personal device whose latency demand still pushes it onto a dedicated low‑latency link rather than ordinary Bluetooth. And the one corner where only a real‑time Wi‑Fi link will do is discrete surround tied to a picture.
When a vendor says they have built a low‑latency Wi‑Fi audio solution, it is worth being precise, because the instinct to picture a custom radio is almost always wrong. The physical layer is not the problem — a standard 802.11 radio moves an audio stream’s worth of bits in a fraction of the budget. What a general‑purpose Wi‑Fi part and its stock stack are not built to deliver is determinism: stock Wi‑Fi is best‑effort and contention‑based, it buffers for throughput, and it carries no shared clock across devices. That jitter is what forces big safety buffers, and the buffers are what break sync.
So the work sits above the PHY. It is a standards‑compliant radio run in a proprietary mode: a scheduled, TDMA‑like discipline in the MAC in place of free‑for‑all contention; hardware timestamping disciplined by a PTP‑style protocol so every speaker shares a microsecond‑accurate clock; a lean real‑time transport instead of stock TCP/IP; and an end‑to‑end delay budget held so tightly the system can delay the picture to meet the sound. Owning the silicon helps — it is how the timestamping, the scheduling hooks and an integrated audio path get baked in — but the differentiator is the timing and transport layer, not a new kind of radio. In short: the same Wi‑Fi, driven very differently.
The standards are moving this way on their own. Wi‑Fi 6 added OFDMA and trigger‑based scheduling so an access point can hand out deterministic slots, and Target Wake Time helps timing and power; Wi‑Fi 7 adds Restricted TWT, protected airtime for latency‑sensitive traffic. Each generation narrows how much a vendor must tweak — but on a network you do not control, guaranteed real‑time behaviour still means running your own scheduled, synchronised layer on top. And increasingly the same SoC is asked to run edge AI — on‑device noise suppression, beamforming, room correction, wake‑word — so the audio is improved before it ever leaves the chip.
This is the seam we work in — where RF, Wi‑Fi silicon, real‑time firmware and on‑device AI are co‑designed rather than bolted together.
Because the link adds latency. The audio is encoded, transmitted, buffered and decoded — on classic Bluetooth about 150–250 ms — while the picture arrives with almost none. Past roughly 45 ms of audio behind video the ear and eye stop agreeing, and you see lip‑sync error.
Not on its own. Bluetooth 5.0 was mostly a Low Energy range‑and‑speed upgrade and left Classic audio unchanged. The real improvement is LE Audio with the LC3 codec, built on the Bluetooth 5.2 Isochronous Channels, which cuts latency to about 20–30 ms and adds a shared presentation‑delay timestamp for sync.
Near mirror images. LE Audio nails low latency and sync and sips power, but is limited on bandwidth and range. Buffered Wi‑Fi nails fidelity, channel count and whole‑home range but adds hundreds of milliseconds. Real‑time Wi‑Fi is the one that does both.
For buffered music streaming, yes. For lip‑synced, multi‑speaker audio, not out of the box: stock Wi‑Fi is contention‑based and un‑synchronised, so you either exploit Wi‑Fi 6/7 scheduling on a network you control, or run your own scheduled, time‑synchronised layer.
Keeping the standard Wi‑Fi PHY and RF and changing how it is driven: scheduled MAC access instead of contention, hardware timestamping with PTP‑style sync, and a lean real‑time transport. The same Wi‑Fi operated in a proprietary low‑latency mode, not a new radio.
Start from the six parameters. Personal, on‑body audio points to Bluetooth or LE Audio; multiroom music to buffered Wi‑Fi; discrete 5.1/7.1 tied to a screen to real‑time Wi‑Fi.
Written for engineers. Share it with one.
The RF front ends behind the link.
AIOn-device audio intelligence.
PRODUCTOur wireless silicon and IP.
KNOWLEDGEThe same call, for asset tracking.
KNOWLEDGEWhat Bluetooth is adding next.
SERVICELinux, BSPs, bootloaders, drivers.
Bring us the six parameters — latency, sync, channels, fidelity, range, power — and we will help you choose the link, and build the silicon and firmware to make it real.