Qualcomm Patent US12695557: The Subband Trick That Could Cut 6G Latency in Half

💡 Qualcomm was granted US patent 12,695,557 on July 29, 2026, for a method that lets 6G phones transmit and receive simultaneously in different parts of the same frequency band - a technique called Subband Full Duplex (SBFD). In Qualcomm's own tests, SBFD cut uplink latency by roughly 50% and lifted uplink throughput by up to 79% compared to standard 5G time-division duplex. If that holds at commercial scale, it would be one of the most consequential air-interface advances since OFDM.

UL throughput gain (ideal mitigation)+79%

UL latency reduction-50%

UL throughput gain (typical)+44%

Coverage gain at cell edge+6 dB

What patent US12695557 actually claims

Issued by the USPTO on July 29, 2026, patent US12695557 covers methods and apparatus that let a wireless user equipment (UE) - the technical term for a phone, tablet, or connected device - operate in a full-duplex mode by splitting its carrier bandwidth into distinct uplink and downlink subbands. In a standard 5G TDD network, the same frequency band is shared between uplink and downlink traffic but time-sliced: the base station alternates between sending and receiving, so neither can do both at once. This patent describes how a UE can instead receive in one subband while simultaneously transmitting in another, within the same physical channel - not switching back and forth, but genuinely doing both at once.

The specific contribution is on the device side. Earlier work addressed how base stations handle SBFD; this patent focuses on how the UE itself manages the simultaneous operation: how it reports its SBFD capability, how it receives scheduling signals, and how it handles the interference between its own transmit and receive chains. That last part - self-interference cancellation at the UE - is the hardest engineering problem in full-duplex wireless and the core of what this patent protects.

With the UE side now secured in IP, the full-duplex picture for 6G comes into focus. But to understand why this matters, you need to see the problem it solves.

The problem: 5G's timing dilemma

Every smartphone today uses one of two duplexing strategies to split traffic between the two directions. FDD (Frequency Division Duplex) uses two separate frequency bands - one up, one down - which never interfere with each other. TDD (Time Division Duplex) uses a single band but alternates rapidly between transmit and receive "slots." Most mid-band 5G runs on TDD because mid-band spectrum is scarcer and FDD wastes half of it.

TDD's problem is timing. When the network needs more uplink capacity - for a video upload, a live AI query, or a sensor stream from an autonomous vehicle - it must wait for an uplink slot. In 5G NR, the most common slot configuration gives roughly 20-25% of time to uplink and 75-80% to downlink. That asymmetry was tolerable when users mostly consumed content, but 6G's use cases flip the equation: always-on AI assistants generate constant two-way traffic; ISAC (Integrated Sensing and Communication) requires simultaneous transmit and sense; and XR experiences need sub-10ms roundtrips that TDD slot boundaries break.

The industry has known about this problem for years. What has lacked is a practical, patent-protected approach that works in commercial handsets. This is what Qualcomm now claims.

The subband solution: splitting the carrier

Subband Full Duplex does not attempt the hardest version of full-duplex (identical frequency, same device, simultaneously - that remains a research challenge). Instead, it divides a wide carrier into two regions: one carries downlink, the other carries uplink, and both run at the same instant. A typical Qualcomm prototype configuration uses a 400 MHz carrier split into 100 MHz uplink and 300 MHz downlink. The two subbands are active concurrently, not in sequence.

The gain from this arrangement is not merely additive. Because the uplink subband is always open, the UE can send immediately when it has data rather than queuing for the next uplink slot. This is what drives the roughly 50% latency reduction: latency in TDD includes not just propagation time but also the wait for the next uplink opportunity. Remove that wait, and every time-sensitive application - robotics, autonomous driving, AI inference in the cloud - benefits directly.

Coverage also improves. Because uplink power is no longer competing with downlink within the same time slot, the uplink subband can be allocated power continuously, pushing the maximum coupling loss by about 6 dB - roughly doubling the edge-of-cell reach for uplink traffic.

Doubling reach is impressive. The interference challenge is what almost ruins it.

The interference problem and Qualcomm's fix

The fundamental challenge in any full-duplex radio is that your transmitter and your receiver are physically close to each other. A phone's transmit power can be 100 milliwatts; the received signal from a distant base station might be a billionth of that. The transmitter's signal leaks into the receiver's front end and overwhelms the incoming data - this is called self-interference.

At the base station, the fix uses separate, physically isolated antenna arrays for transmit and receive - a luxury a phone does not have. At the UE, Qualcomm's approach relies on a combination of careful subband placement (the gap between UL and DL subbands provides some natural isolation), analog cancellation circuits in the RF front-end, and digital cancellation in the baseband. The patent covers the signaling framework that supports this: how the network tells the UE which subbands are allocated, how the UE acknowledges its cancellation capability, and how scheduling adapts dynamically.

Qualcomm's test data shows the penalty for imperfect cancellation is real but manageable: throughput gain falls from 79% to 44% when interference mitigation is not ideal. The 35-percentage-point gap between "ideal" and "typical" is the engineering headroom that remains - and the reason UE-side patents in this space matter so much for future 6G handset design.

Understanding who else is racing in this direction makes the stakes clearer.

Who built it and why it matters now

Qualcomm Incorporated is the world's largest designer of mobile wireless chips by revenue and the single largest holder of standard-essential patents (SEPs) in 4G and 5G. Its 6G timeline is not abstract: the company has been prototyping 6G hardware since 2024, demonstrated a live SBFD prototype at Mobile World Congress Barcelona in February 2026, and is actively participating in 3GPP Release 19 and beyond - the standards process that will define 6G for commercial networks in the early 2030s.

The key rivals watching this space are Ericsson, Nokia, Samsung (whose RIS patent this site featured two weeks ago), Huawei, and MediaTek. Huawei leads in total 6G patent families but faces export restrictions limiting access to sub-10nm silicon. Ericsson and Nokia focus on the infrastructure side, not UE chipsets. That leaves Qualcomm and MediaTek as the two companies with both the chipset platform and the incentive to make SBFD work in actual phones - and Qualcomm just secured the UE-side IP.

A standard-essential patent in a future 6G air-interface standard means every phone maker that ships a 6G device must either license this technology or design around it. For a company that already earns roughly 20% of every smartphone's wholesale price through licensing, adding another layer of UE-side IP in 6G is a significant structural move.

But no single patent works alone.

The 6G ecosystem: where this patent fits

SBFD is one of three foundational air-interface technologies in Qualcomm's 6G Foundry program. The other two are Giga-MIMO (very large antenna arrays at the base station, operating in the 7-24 GHz upper mid-band spectrum) and probabilistic shaping (a coding technique that concentrates transmitted power in the signal states most likely to succeed). When combined, Qualcomm's tests show 2.3x-2.4x throughput improvement compared to 5G baselines - a multiplier that would make 6G a genuine generational leap.

SBFD also intersects with ISAC (Integrated Sensing and Communications). In an ISAC network, the base station emits a sensing waveform and listens for its echo to detect objects - people, vehicles, drones - in the vicinity. That sensing waveform is also a communications carrier. SBFD's ability to maintain continuous uplink enables connected devices to report sensing data in near-real time, without waiting for uplink slots, feeding the digital twins and AI inference pipelines that 6G is designed to support.

The interconnection runs deeper. SBFD needs AI-based interference cancellation (AI as an enabler), which in turn runs on dedicated AI inference chips (semiconductors), which must be manufactured in advanced packaging nodes built with 3D integration technologies. Multiple patents from multiple companies are already woven into the 6G stack - and US12695557 is now one of them.

Key facts: Qualcomm US12695557 at a glance

Detail	Value
Patent number	US12,695,557
Title	Full-Duplex Operations for User Equipments
Assignee	Qualcomm Incorporated
Jurisdiction	United States (USPTO)
Grant date	July 29, 2026
Core claim	UE-side subband full-duplex: simultaneous UL/DL in separate subbands of the same TDD carrier
Key metric	~50% UL latency cut; up to +79% UL throughput vs. standard TDD (Qualcomm tests, 2026)
6G relevance	Part of Qualcomm 6G Foundry: SBFD + Giga-MIMO + probabilistic shaping
Prior foundational work	US20210136696A1 (Qualcomm, filed Oct 2020) - sub-band full-duplex interference mitigation

So what does it mean for us?

Patent US12695557 is not a product launch - it is an IP stake in a standards race. 3GPP is expected to finalize 6G specifications around 2028-2029, with commercial deployments in South Korea, Japan, and the US following in the early 2030s. What Qualcomm is doing now is shaping the technical choices that will define the 6G architecture, so that when those deployments happen, their chipsets are indispensable.

For the innovation ecosystem, the SBFD patent matters because it resolves a long-standing ambiguity: the UE side of full-duplex was the missing piece. Base-station SBFD was already being prototyped by multiple vendors. Now the device side has an IP anchor, and the standards conversations at 3GPP can converge around an implementable approach.

For patent translation and technical IP work, this is exactly the kind of filing that creates multi-jurisdictional demand: a Qualcomm US patent of this strategic importance will be followed by family members in Europe (EPO), China (CNIPA), South Korea, and Japan - each requiring precise technical translation that holds up in patent office examination and, eventually, in licensing negotiations. The wireless standards world communicates in English, but IP rights are enforced jurisdiction by jurisdiction, in local languages. A concept as precise as "self-interference cancellation at the UE level" admits no ambiguity in any language.

Whether or not SBFD becomes the dominant 6G duplexing approach, the race it represents - for UE-side air-interface IP - will shape how 6G is built, licensed, and translated for a decade.

FAQ

Sources
USPTO Official Gazette Week 30 / 2026 - Qualcomm patents
RCR Wireless: Qualcomm 6G Foundry - SBFD performance data (2026)
Qualcomm ONQ: 6G Air Interface - Giga-MIMO and SBFD (2026)
Google Patents: US20210136696A1 - Qualcomm SBFD prior patent (filed 2020)

About the author

Dao Huy (Lucas) is a professional translator specializing in English, Chinese, and French into Vietnamese with over seven years of experience in patent translation, technical translation, and IP translation. His work includes localizing engineering documentation, patent specifications, and intellectual property filings for technology companies entering the Vietnamese market. Patents like Qualcomm's US12695557, which describe precise radio-frequency architectures and claim specific interference-cancellation signaling frameworks, represent exactly the kind of material where a missed technical nuance can invalidate protection across an entire jurisdiction worth billions in future licensing revenue.

If your organization needs patent translation, technical translation, or IP translation into Vietnamese - or if you are localizing a software product or technical platform for the Vietnamese market - Lucas offers precise, field-tested translation and localization services. Request a quote at daohuy.com.