IN Brief:
- The SAF9800 combines AM and FM reception with a HiFi 5 DSP and neural-network processing.
- Hardware biquad acceleration supports filtering, noise reduction, classification, and acoustic-source separation.
- The device advances software-defined automotive audio while retaining an integrated analogue front end.
NXP Semiconductors has introduced an automotive radio and audio processor combining analogue AM and FM reception with digital signal processing, neural-network acceleration, and hardware filtering.
The SAF9800 integrates an automotive radio front end with a HiFi 5 DSP and machine-learning resources supporting noise reduction, acoustic classification, source separation, and software-defined audio functions. Hardware biquad accelerators increase filtering capacity without assigning every operation to the main DSP.
Processing performance exceeds that of NXP’s previous generation by more than tenfold. The additional capacity can serve conventional equalisation and tuning or newer functions including emergency-siren detection, mechanical-fault recognition, and the separation of speech or warning sounds from cabin noise.
Despite the growth of streamed audio and digital broadcasting, AM and FM reception remain within the architecture because they provide broad geographic coverage, emergency information, and compatibility across vehicle markets. Their analogue signal paths must operate beside switching converters, displays, processors, motors, communications links, and high-voltage electrified systems.
NXP’s EVAM-lite processing is designed to reduce amplitude-modulated interference in electric vehicles, where traction inverters, DC-DC converters, onboard chargers, and auxiliary power systems can introduce broadband and harmonic noise. RF layout, grounding, shielding, antenna design, and digital processing all contribute to the resulting reception quality.
A common hardware platform can support several vehicle variants, with software controlling which audio functions are enabled. Manufacturers can consequently reduce the number of separate electronic configurations that must be sourced, validated, and maintained across a vehicle range.
Software-defined audio retains analogue constraints
As more audio functions move into software, algorithms for noise suppression, sound enhancement, classification, and personalisation can change without replacing the full analogue signal chain. The quality of the source information, however, still depends on antenna matching, selectivity, grounding, conversion accuracy, and electromagnetic compatibility.
Interference or distortion introduced before digitisation cannot always be removed afterwards, even with a capable neural model. An integrated processor may improve coordination between the tuner and digital processing, but board layout and vehicle-level EMC remain essential to predictable reception.
Centralised electrical architectures are also changing the location of audio functions. Some processing can move into a domain or central computer, while time-sensitive reception, filtering, and conversion remain close to the radio hardware to avoid transporting raw or lightly processed signals across the vehicle network.
Keeping those functions locally integrated can reduce network traffic and simplify latency, although the radio processor then carries a larger software and cybersecurity burden. Feature updates must be authenticated, version controlled, and tested across vehicle variants so that an entertainment change cannot interfere with warnings, communications, or regulated functions.
Acoustic classification introduces uses extending beyond entertainment. Sirens, alarms, broken-glass signatures, unusual mechanical sounds, and spoken instructions can all be identified, provided that training data represents the noise, accents, cabin conditions, and operating environments encountered in service.
False detections create distraction, whereas a missed safety-related sound can carry more serious consequences. Thresholds, confidence measures, sensor placement, and fallback behaviour therefore require validation alongside the underlying model.
The SAF9800 sits within a broader processing portfolio that includes i.MX 95 application processors entering wider distribution. Those devices address higher-level industrial, automotive, and edge workloads, while the radio processor concentrates specialised analogue and audio functions closer to the signal source.
Automotive products also require supply and qualification periods extending well beyond normal consumer-audio cycles. The processor must withstand temperature extremes, voltage disturbances, vibration, and electromagnetic stress while remaining available throughout vehicle production and servicing.
Consolidating reception, DSP, acceleration, and filtering can reduce board area, external memory, power consumption, and component count. The corresponding concentration of functions increases the need for diagnostics, software isolation, fallback operation, and long-term support.
Analogue radio consequently remains part of the software-defined vehicle rather than becoming a separate legacy subsystem. Its integration with AI processing will be measured by reception quality, predictable acoustic classification, software maintainability, and reliable operation within a hostile electrical environment.
