QuantalRF AG is sampling a monolithic linear front-end module that targets Wi-Fi 7 and Wi-Fi 8 WLAN designs and shifts error-vector-magnitude performance from the host SoC's digital-predistortion engine into the RF silicon itself. The Swiss company built the QWX27104 on GlobalFoundries' 8SW RF-SOI process and schedules volume production for Q1 2027.

The device addresses a structural problem in modern handset RF chains. Mobile systems built on nonlinear gallium-arsenide FEMs lean heavily on DPD to correct power-amplifier distortion. GaAs PAs are fast, but they lack linearity, so the correction burden lands on the baseband. That approach consumes computational resources, raises power consumption, and erodes design margin — and it degrades further as channel bandwidths widen to 160 and 320 MHz, which Wi-Fi 7 requires and Wi-Fi 8 will extend.

The QWX27104 inverts the tradeoff. QuantalRF engineers designed linearity into the FEM's patent-pending PA architecture rather than correcting for it downstream. Because the amplifier is linear across all power modes, the device reaches its target EVM with a lower-order, less complex DPD engine on the host SoC.

The claimed payoff is concrete. The company cites up to 3 dB more mask-compliant output power under the most demanding modulations at 160/320 MHz, greater tolerance to system, temperature, and process variations, and channel-to-channel EVM variation held within 0.5 dB. Available competitive GaAs parts, by QuantalRF's comparison, swing more than 4 dB across channels.

The module integrates three RF building blocks on one die: the linear PA, a low-loss single-pole, triple-throw switch, and a low-noise amplifier. The PA maintains EVM performance under varying VSWR conditions, and all RF ports match on-chip to 50 Ω, which cuts external components and PCB footprint — a direct concern for the target applications: smartphones, portable battery-operated devices, AI wearables, and AR/VR headsets.

The RF numbers define the device's positioning. Transmit output power with DPD applied runs 22 dBm at HE20/MCS0 with 3 dB of spectral-emission-mask margin, 20 dBm at VHT80 with −42 dB DEVM, 18 dBm at HE160 with −45 dB DEVM, and 16 dBm at EHT320 with −43 dB DEVM. The wideband 320 MHz figure matters most: holding −43 dB DEVM at that modulation is where GaAs-plus-heavy-DPD designs struggle.

Supporting specifications include 28 dB transmit gain in high-gain mode, current consumption of 252 mA at 21 dBm output, a 2 dB noise figure, and 15 dB receive gain. The device operates from 5,150 to 7,125 MHz — covering the full 5 GHz and 6 GHz bands — and runs from a 3.8 V supply for the PA and 1.8 V for the LNA. Packaging options cover both integration models: a 2 × 2 mm, 16-pin LGA and a flip-chip die smaller than 2 mm² for board-level and die-level integration respectively.

The competitive context is the slow migration of mobile WLAN front ends from GaAs toward silicon-based processes. RF-SOI gives QuantalRF a monolithic die where PA, switch, and LNA share one substrate, and the foundry relationship with GlobalFoundries provides a production path at scale. QuantalRF's claims rest on its own comparisons with unspecified competitive GaAs parts; independent benchmarking will come once customers evaluate the sampling silicon against incumbent FEMs from the established mobile RF suppliers.

With volume parts due in Q1 2027, the QWX27104's commercial test will be whether its engineered linearity translates into measurable SoC power savings and design-margin gains for handset and wearable OEMs as Wi-Fi 8 design cycles begin.