CompoundLearn

Multi-Link Operation (MLO) Explained: Wi-Fi 7 and Wi-Fi 8 Interview Prep

Multi-Link Operation (MLO) — the Wi-Fi 7 (802.11be) / Wi-Fi 8 (802.11bn) feature that lets one MLD use several bands at once, cutting latency.

Quick answer

Multi-Link Operation (MLO) is the architectural change that defines IEEE 802.11be (Wi-Fi 7) and that IEEE 802.11bn (Wi-Fi 8) builds on (see /topics/wifi-8).

Wireless interviews use MLO questions to test whether a candidate can navigate a multi-mechanism design space rather than recite a single feature (see /topics/wifi-standards-interview-guide).

Editorial review

Written by

CompoundLearn editorial team

Wireless / RF / hardware engineering

Reviewed by

CompoundLearn editorial team

Wireless / RF / hardware engineering

Last reviewed

Built from curated topic maps, editorial validation, and subject-matter review so the page stays aligned with the interview intent and the current content pipeline.

What it is

Multi-Link Operation (MLO) is the architectural change that defines IEEE 802.11be (Wi-Fi 7) and that IEEE 802.11bn (Wi-Fi 8) builds on (see /topics/wifi-8). Before MLO, a Wi-Fi client could associate to one band at a time — moving between 2.4 GHz, 5 GHz, and 6 GHz required a full disassociation and a fresh association. After MLO, a single Multi-Link Device (MLD) maintains affiliated links on multiple bands simultaneously and the access point scheduler steers each frame to the most appropriate link. MLO is the substrate for the throughput aggregation, link redundancy, and latency-steering features that distinguish Wi-Fi 7 and Wi-Fi 8 from earlier generations. The MLO architecture has three layers. First, the Multi-Link Device — both AP MLD and non-AP (client) MLD are defined. Each MLD aggregates multiple physical link endpoints, called affiliated STAs (or APs), under a single MLD MAC address. Upper layers see one device with one address; the link layer manages many radios with their own per-link MAC addresses. An 802.11be MLD can have up to 15 affiliated links. Second, the link operating modes. STR (Simultaneous Transmit and Receive) lets the device transmit on one link while receiving on another, but requires about 30 dB of combined frequency separation plus antenna isolation between the radio chains. Non-STR is the alternating mode used when isolation is insufficient — typically 2.4 GHz + 5 GHz on small devices. EMLSR (Enhanced Multi-Link Single Radio) is the power-efficient mode where one radio chain switches between links via a low-power frame detector that wakes the high-capability receiver. Third, the link selection policies. The scheduler steers traffic per-class: voice and video to the lowest-jitter link, bulk traffic to any available link, latency-critical frames optionally duplicated across links. IEEE 802.11bn (Wi-Fi 8) extends this with per-frame link selection (not just per-flow) and ties it to Restricted Target Wake Time and Non-Primary Channel Access (NPCA) for tighter latency control. Case study: a Wi-Fi 7 AR/VR headset wants 5-millisecond round-trip latency on motion-tracking frames. The MLD associates on 5 GHz and 6 GHz. The scheduler keeps the 6 GHz link reserved for motion frames (clean spectrum, lowest jitter) and uses 5 GHz for video texture streaming (bulk, high-throughput, can tolerate jitter). When the 6 GHz link briefly degrades from microwave interference, the scheduler steers the motion frames to 5 GHz for the affected microseconds and reverts. This level of per-frame steering is what MLO enables; no single-link Wi-Fi 6 deployment can deliver the same latency floor. MLO interacts directly with the OFDMA resource-unit scheduling surface (see /topics/ofdma-resource-units-explained) because every frame steered onto a link still has to fit into an RU on that link, and the broader Wi-Fi 8 reliability mechanisms are covered at /topics/wifi-8-80211bn-ultra-high-reliability.

Why interviewers ask

Wireless interviews use MLO questions to test whether a candidate can navigate a multi-mechanism design space rather than recite a single feature (see /topics/wifi-standards-interview-guide). A strong answer separates the four MLO axes — MLD architecture, operating mode (STR / Non-STR / EMLSR), link selection policy, and the 802.11bn extensions (NPCA, per-frame selection, enhanced MLO) — and explains how a design choice on one axis constrains the others. Three signals matter most to interviewers. First, RF reasoning: can the candidate explain that STR requires physical antenna isolation and not a digital baseband trick? Many candidates fabricate a "MAC-layer self-interference cancellation" that does not exist. The correct mental model is that STR is gated by RF isolation, roughly 30 dB of combined frequency separation plus antenna isolation. Second, address-hierarchy clarity: can the candidate distinguish the MLD MAC address (upper-layer identifier) from the per-link MAC (link-layer frame exchange)? Confusing the two is a fast tell that the candidate has memorized terms without understanding the architecture. Third, power-vs-concurrency trade-off: can the candidate explain why EMLSR exists alongside STR, and when each is the right choice? EMLSR for battery-constrained devices that prefer power over concurrency; STR for AP-class devices that can afford separate radios. Roles where these questions matter most include wireless systems engineers, modem firmware engineers working on MLO scheduling, RF systems engineers designing antenna layouts for Wi-Fi 7+ products, and product managers scoping multi-radio bills of materials. MLO is the most commonly-asked Wi-Fi 7 interview topic and the deepest Wi-Fi 8 interview surface after Co-BF (see /topics/beamforming).

Common mistakes

The most common mistake is treating MLO as "Wi-Fi with more antennas." MLO is not just hardware — it is an architectural change with an MLD address hierarchy, an explicit set of operating modes, and a scheduler-level link selection policy. A candidate who skips to "MLO uses multiple radios" misses every interesting interview content. A second common mistake is the STR self-interference mechanism. Candidates often invent a "digital baseband cancellation algorithm" or a "MAC-layer cancellation" that removes transmit leakage. No such mechanism exists in 802.11be or 802.11bn. STR depends on physical RF isolation — about 30 dB of combined frequency separation plus antenna isolation. If the device cannot achieve that, it operates Non-STR. A third gap is the address hierarchy. Some candidates treat the MLD MAC address and per-link MAC address as the same thing; some assume only the MLD MAC matters at the link layer. The correct picture is two-layered: MLD MAC for upper-layer addressing and inter-MLD communication, per-link MAC for link-layer frame exchange and PHY operations on each band. A fourth mistake is the EMLSR power story. Candidates sometimes describe EMLSR as "doubling throughput by running both radios at full power continuously." That is the opposite of what EMLSR does. EMLSR runs one radio at a time to save power; the low-power detector wakes the high-capability receiver only when needed. A fifth gap is forgetting backward compatibility. MLO is a Wi-Fi 7 feature — but the access points and clients still serve Wi-Fi 6 and Wi-Fi 5 stations on the same channels. Mixed-generation fleets are the norm. A deployment plan that assumes every client is MLO-capable on day one will produce wrong cost and coverage estimates. Other gaps: assuming Wi-Fi 8 invented MLO (it did not — Wi-Fi 7 introduced it; Wi-Fi 8 extends per-frame link selection on top), assuming MLO doubles or triples throughput in all conditions (it depends on which links are available and free of interference), and assuming MLO eliminates the need for proper antenna design (it amplifies the need — multi-radio devices have stricter antenna isolation requirements than single-radio designs).

MLO operating modes — STR vs Non-STR vs EMLSR

ModeConcurrencyPower profileRF requirementTypical use
STR (Simultaneous Transmit and Receive)Full — transmit on one link while receiving on anotherHigh — both radios active concurrently~30 dB combined frequency separation + antenna isolationAP-class devices; well-isolated 5+6 GHz client radios
Non-STRAlternating — only one link active at a timeMedium — radios alternateInsufficient isolation between bandsSmall client devices with adjacent-band radios (2.4+5 GHz)
EMLSR (Enhanced Multi-Link Single Radio)Switch — one radio chain serves multiple links via detectorLow — single-radio operation with wake-up detectorSingle radio chain plus low-power frame detectorBattery-constrained clients (phones, wearables, IoT)

Sample interview questions

  1. A Wi-Fi 7 phone is configured as a Multi-Link Device with 5 GHz and 2.4 GHz affiliated links. The vendor lists MLO support but the device operates in Non-STR mode and never enters STR. What is the most likely physical reason?
    • A. The 5 GHz and 2.4 GHz radio chains share antenna real estate and lack 30 dB of combined frequency separation plus antenna isolation, so transmitting on one band desensitizes the receiver on the other. ✓
    • B. The MLD MAC address is configured incorrectly; STR requires a separate primary MAC per link.
    • C. STR is forbidden by IEEE 802.11be for any pair of bands; only 802.11bn supports STR.
    • D. The phone is running 4096-QAM, which disables STR.

    Option A is correct. STR mode requires the two radio links to be sufficiently RF-isolated — transmit leakage from one link must be far enough below the noise floor of the receiver on the other link that no desensitization occurs. The rule of thumb is around 30 dB of combined frequency separation plus antenna isolation. A small device like a phone with 5 GHz and 2.4 GHz antennas placed close together typically does not meet that bar, so the MLD falls back to Non-STR alternating operation. 5 GHz + 6 GHz pairs typically do meet the bar because the radios can use separate antennas and the bands are further apart. Option B is wrong because MLD address configuration is independent of STR capability — the address hierarchy is a MAC-layer property; STR is a PHY/RF property. Option C is wrong because STR is defined in 802.11be Wi-Fi 7 itself, not just in 802.11bn; 802.11bn extends MLO but did not introduce STR. Option D is wrong because STR mode selection is decoupled from modulation order; a station running 1024-QAM with adequate RF isolation still operates STR if isolation is sufficient.

  2. An MLD with EMLSR enabled receives a 100-microsecond frame on a monitored link. The high-capability receiver needs roughly 50 microseconds to wake up. What design consideration follows?
    • A. The incoming frame must carry a signal duration long enough to cover the receiver wake-up latency, or the high-capability receiver will not be ready to decode the data portion. A short frame (below the wake-up budget) can be missed entirely. ✓
    • B. The high-capability receiver is unnecessary for EMLSR; the low-power detector decodes all frames itself.
    • C. EMLSR cannot receive any frame below 200 microseconds in total duration regardless of signal duration.
    • D. The wake-up latency is a hardware bug; correctly implemented EMLSR has zero wake-up time.

    Option A is correct. EMLSR depends on a two-stage receive path: the low-power frame detector identifies that a frame has started and triggers the high-capability receiver to wake up. The signal duration of the incoming frame — the time spent in the preamble and signal field before the data portion — must be long enough to cover the receiver wake-up latency. If a frame is too short, the high-capability receiver is not ready in time and the frame is dropped. This is why EMLSR signaling protocols typically support multiple signal durations (e.g., increasing the preamble length when wake-up is needed) instead of one fixed value. Option B is wrong because the low-power detector identifies frame presence; it does not have the demodulation capability of the high-capability receiver. Option C is wrong because EMLSR can receive short frames if the signal duration is configured correctly relative to wake-up latency; the constraint is signal duration, not total frame duration. Option D is wrong because wake-up latency is a physical property of the high-capability receiver hardware; it is the design constraint EMLSR is built around, not a bug.

  3. A network engineer is troubleshooting why an MLD client cannot establish a multi-link association. Packet captures show the AP MLD advertising one MLD MAC address and three affiliated AP link MAC addresses. The client uses the MLD MAC address to authenticate but transmits data frames using the per-link MAC. Is this normal?
    • A. Yes — this is the standard 802.11be MLO address hierarchy. The MLD MAC address is the primary identifier for the device and for inter-MLD communication, while per-link MAC addresses isolate PHY and link-layer frame exchange per band. ✓
    • B. No — both the MLD MAC and the per-link MAC must be the same address for MLO to work.
    • C. No — data frames should always use the MLD MAC address; per-link MACs are only used during association.
    • D. No — only one of the three affiliated links should have a per-link MAC; the others should reuse the MLD MAC.

    Option A is correct. The 802.11be MLD address hierarchy has two layers by design. The MLD MAC address serves as the upper-layer identifier — the address that TCP/IP sees, the address that authentication and key management use, and the address other MLDs use for inter-MLD communication. Each affiliated link, however, has its own per-link MAC address that is used for link-layer frame exchange and PHY operations on that specific band. This separation is what allows the MLD to look like a single device to upper layers while operating as multiple radios at the link layer. The capture description matches normal MLO behavior. Option B is wrong because the MLD MAC and per-link MAC are explicitly different addresses in the 802.11be design; sharing them would defeat the address-isolation goal. Option C is wrong because data frames at the link layer are exchanged using the per-link MAC of the link they traverse; upper-layer addressing remains the MLD MAC. Option D is wrong because every affiliated link gets its own per-link MAC; the count of affiliated MACs equals the count of affiliated STAs / APs, not a partial subset.

Frequently asked questions

What is Multi-Link Operation (MLO) in IEEE 802.11be (Wi-Fi 7) and what problem does it solve?
Multi-Link Operation is a defining feature of IEEE 802.11be (Wi-Fi 7) that lets a single device — called a Multi-Link Device (MLD) — establish and use multiple radio links to an access point at the same time, typically across 2.4 GHz, 5 GHz, and 6 GHz bands. Before MLO, a Wi-Fi client could associate with only one band at a time and a roam between bands meant a full disassociation and reassociation. MLO solves three problems together: throughput aggregation by carrying data across multiple links, low-latency steering by sending latency-sensitive frames on the lowest-jitter link, and reliability by duplicating critical frames on more than one link. The architecture is multi-link per device, not multi-link per packet.
What is a Multi-Link Device (MLD) and how does its address hierarchy work?
A Multi-Link Device is the logical entity introduced in 802.11be that aggregates two or more physical link endpoints — called affiliated STAs (or APs) — under a single MLD MAC address. Both AP MLDs and non-AP (client) MLDs are defined. The address hierarchy has two layers: the MLD MAC address is the primary identifier used by upper layers and for inter-MLD communication, while each affiliated link has its own per-link MAC address used for link-layer frame exchange and PHY operations. This means an MLD looks like a single device to TCP/IP and like multiple radios at the link layer. An 802.11be MLD can have up to 15 affiliated links, with link ID assignment AP-initiated during association.
What are STR and Non-STR modes in MLO, and what physical condition decides which is used?
STR (Simultaneous Transmit and Receive) is the MLO mode where the device transmits on one link while simultaneously receiving on another. Non-STR is the alternating mode where one link is silent whenever the other is active. The deciding condition is radio isolation: STR requires the two radio links to be sufficiently decoupled — typically about 30 dB of combined frequency separation plus antenna isolation — so that transmit leakage from one link does not desensitize the receiver of the other. Pairs like 5 GHz + 6 GHz typically meet the isolation threshold; pairs like 2.4 GHz + 5 GHz often do not. STR delivers the full multi-link throughput benefit. Non-STR still benefits from link aggregation and steering but cannot run truly parallel TX/RX.
What is Enhanced Multi-Link Single Radio (EMLSR) and why is it a separate MLO mode?
EMLSR (Enhanced Multi-Link Single Radio) is a power-efficient MLO mode where the device uses one radio chain at a time and switches between links via a low-power frame detector. The detector monitors multiple links and wakes the high-capability receiver when traffic arrives. EMLSR appears as multi-link to the access point — the MLD advertises multiple affiliated STAs — but consumes far less power than running two radios continuously like STR. The trade-off is concurrency: EMLSR cannot transmit on one link and receive on another at the same instant, and the wake-up latency for the high-capability receiver imposes a minimum signal duration on the incoming frame. EMLSR is the mode of choice for battery-constrained MLDs like phones, wearables, and IoT clients.
How does MLO link selection work for latency-sensitive versus throughput-sensitive traffic?
In 802.11be, link selection is per-traffic-class. The MLD and AP MLD negotiate which links are available and the access point scheduler steers each frame onto the link that best matches its requirements. Voice and video frames go to the link with the lowest current congestion and jitter — typically the cleanest 6 GHz link. Bulk best-effort traffic can take any available link. Frames that need redundancy can be duplicated across links and the receiver discards the duplicate. IEEE 802.11bn (Wi-Fi 8) extends this with per-frame link selection that adjusts on a packet-by-packet basis instead of per-flow, and adds Restricted Target Wake Time scheduling so latency-sensitive frames pre-reserve a wake window on a specific link.
How does Non-Primary Channel Access (NPCA) in 802.11bn relate to MLO?
NPCA is an 802.11bn MAC mechanism that lets a device transmit on a non-primary 20 MHz subchannel when the primary channel is busy. While not strictly MLO, it interacts closely with MLO because both expand the access decision surface: NPCA expands within a single link by adding subchannel-level access, while MLO expands across links by adding band-level alternatives. A station with MLO and NPCA both available has more degrees of freedom for avoiding contention. NPCA is gated by an RSSI threshold to bound hidden-node risk, and must coordinate with Dynamic Sensitivity Control (DSC) — DSC adjusts the CCA threshold that NPCA depends on. The combined mechanism is what gives 802.11bn its reliability story.
What MLO interview questions catch candidates who only memorized the feature list?
The fast filter is the STR isolation question: ask what physical condition gates STR operation on an MLD. A candidate who says "Multi-Link Operation defines a digital baseband cancellation algorithm that removes self-interference at the MAC layer" is fabricating a mechanism that does not exist. The correct answer is RF isolation — about 30 dB of combined frequency separation plus antenna isolation. A second filter is the address hierarchy: a candidate who treats MLD MAC and per-link MAC as the same thing misses that upper layers see one address while the link layer manages many. A third filter is the EMLSR power story: a candidate who says EMLSR doubles throughput because it runs both radios at full power has it backwards; EMLSR runs one radio at a time to save power.

Related topics

Pillar

Siblings

Practice

Essential AI-Native Skills for Multi-Link Operation (MLO) Explained: Wi-Fi 7 and Wi-Fi 8

Modern engineering work increasingly uses AI tools for design and code review, debugging, documentation, test and testbench generation, and workflow automation. The goal is not to let AI replace engineering judgment — it is to move faster while keeping verification discipline.

  • Use AI to explain unfamiliar code, logs, waveforms, datasheets, or test failures.
  • Break large problems into small, reviewable steps you can verify independently.
  • Ask AI for hypotheses, then validate them against tests, measurements, simulations, or lab data.
  • Version-control your analysis scripts, testbenches, and configs — keep changes small and reviewable.
  • Document your assumptions, design tradeoffs, and debugging decisions.
  • Verify AI output before trusting it: run the checks that fit the domain — unit tests, linters, simulations, or bench/lab measurements.
  • Review AI output for correctness, edge cases, and real-world consequences.

Multi-Link Operation (MLO) Explained: Wi-Fi 7 and Wi-Fi 8 — coming to the question bank

The adaptive practice engine is already live for core wireless, RF, and ML systems. Multi-Link Operation (MLO) Explained: Wi-Fi 7 and Wi-Fi 8 isn't covered in the question bank yet — get notified when it's added.

One email when this topic launches. Nothing else. Unsubscribe in one click.