PRECISION AUDIOFROM MIC TOLOUDSPEAKER
Full signal-chain audio engineering for embedded and connected products.
Great audio in an embedded product is disciplined signal-chain engineering from microphone placement through codec configuration, DSP algorithm tuning, and acoustic chamber validation. Ankh covers every layer: analog front-end design, I2S/TDM/PDM bring-up, beamforming and AEC on ARM Cortex-M, A2DP and LE Audio with LC3, USB Audio Class 2.0, and free-field measurement against IEC 60268 — delivering −95 dBFS noise floors and sub-0.1% THD+N in production.
What We Engineer
Audio Codec Selection & Integration
Codec selection and configuration (ES8388, TLV320AIC3x, ADAU1761, WM8960) for target SNR, power, and interface requirements — bring-up covers I2S/TDM master-slave configuration, MCLK PLL programming, and register-map validation against SNR and THD+N test vectors.
Microphone Front-End Design
MEMS microphone selection (Knowles SPH0645, TDK ICS-43434, InvenSense ICS-52000) with acoustic porting geometry to hit frequency response and polar pattern — PDM decimation filter configuration, analog LNA design for electret capsules, and array layout for beam-steering.
Speaker Amplifier Design
Class D amplifier design in BTL and SE configurations covering IC selection (MAX98357, TAS2770, TPA3110), LC output filter optimization for EMI, and speaker impedance matching — validated via swept-sine THD measurements from 20 Hz–20 kHz.
DSP Algorithm Development
AEC, ANC, and two/four-mic beamforming on ARM Cortex-M4/M7 using CMSIS-DSP — cycle budget profiling, FFT block size optimization, and double-talk detection validated with pre-recorded test corpora.
Bluetooth Audio Integration
A2DP SBC/AAC and LE Audio with LC3 codec integration covering BT stack bring-up (Zephyr, ESP-IDF, custom), isochronous channel configuration for LE Audio, and latency characterization for TWS earbud applications — with BT SIG QDID certification.
USB Audio Class 2.0
UAC2 descriptors, isochronous endpoint firmware, and host-side feedback endpoint logic for sample-accurate synchronization at 24-bit/96 kHz — clock recovery using SOF feedback preventing buffer underrun across Windows, macOS, and Linux without custom drivers.
Audio Path Signal Chain Design
SNR budget allocation from microphone self-noise through ADC quantization noise, op-amp voltage noise, and PCB parasitics — signal flow diagrams, LTspice noise simulations, and empirical measurements to verify design margin.
Acoustic Testing & Validation
Per IEC 60268 and IEC 61672 — free-field frequency response, polar pattern, noise floor at −95 dBFS, swept-sine THD+N, and impedance vs. frequency curves in formal test reports.
EMI Filtering for Audio Circuits
Ferrite-bead networks, common-mode chokes, and ground-plane partitioning designed and validated to suppress GSM 217 Hz buzz, switching regulator ripple coupling, and radiated susceptibility from co-located RF modules.
How We Work
Requirements & Signal Chain Spec
We produce a signal chain specification defining every gain stage, dynamic range allocation, and interface standard before schematic work begins. Microphone directivity requirements and acoustic use case drive codec, ADC resolution, and DSP algorithm selection—giving the firmware team a clear interface contract.
Component Selection & Schematic Design
Components are selected against datasheet SNR data and validated reference designs, with schematics capturing the full analog path including AC-coupling corner frequencies and ESD protection. A preliminary noise budget spreadsheet aggregates component-level contributions and predicts system SNR before layout begins.
PCB Layout & Audio Ground Strategy
Star-ground topology prevents digital return currents from crossing the analog path; MEMS mics are placed for acoustic porting; I2S/PDM clock traces are length-matched with ground return shielding. Switching loops stay remote from mic inputs, with ferrite footprints on every analog supply rail.
Firmware Bring-Up & DSP Integration
Codec bring-up validates PLL lock, sample rate, and volume control before DSP algorithms are integrated; AEC and beamforming are tuned against near-end/far-end test corpora at defined SNR levels. FFT sizes, overlap, and fixed-point scaling are profiled to meet the real-time budget without clipping artifacts.
Acoustic Characterization & Tuning
Free-field measurements using calibrated reference microphones capture frequency response, polar patterns, THD+N, noise floor, and maximum SPL before onset of limiting. EQ biquad coefficients are derived from measurement data, and hardware/firmware are iterated together until results satisfy the original specification.
Production Test & Acceptance Criteria
An Audio Precision-based automated test station validates every unit against SNR, THD+N, frequency response, and channel balance limits in under 15 seconds. Test specification, Python or LabVIEW scripts, acoustic coupler fixtures, and golden-unit reference recordings are delivered to the CM.
The Engineering Behind Every Decibel
Noise Budget Engineering
Reaching −95 dBFS co-located with a switching charger demands PSRR analysis, often an LDO post-regulator on the analog rail, and a noise budget aggregating mic self-noise, ADC input-referred noise, and resistor Johnson noise. Models predict A-weighted noise floor within ±2 dB of measurement, documented so the budget survives component substitutions.
Electronics DesignReal-Time DSP on Constrained MCUs
AEC convergence is tuned against the acoustic path impulse response in the actual form factor — cabinet resonance directly affects NLMS tap count — with every function profiled using DWT cycle counters to verify the real-time budget. Battery products duty-cycle DSP during silence while a low-power VAD preserves wakeword detection.
Firmware DevelopmentBluetooth Audio: A2DP to LE Audio
LE Audio with LC3 over isochronous channels enables synchronized multi-stream audio and hearing-aid profiles that A2DP cannot support. CIG/CIS parameters are configured for sub-5 ms inter-earbud synchronization; LC3 integration covers bitrate selection (32–320 kbps), PLC under RF packet loss, and ISOAL segmentation in the BT HAL.
RF & Wireless Engineering
Full signal-chain audio engineering for embedded and connected products.
Audio Engineering Across Ankh's Product Range
Audio touches products far beyond consumer speakers — medical devices need clean acoustic alarms, industrial monitors need reliable voice command in high-noise environments, and wearables need ultra-low-power always-on microphones.
Wearable Electronics
Always-on wakeword detection with PDM mic, sub-100 µA voice activity detector, in-ear ANC
View ProductMedical Devices
IEC 60601-1-8 alarm audio, stethoscope acoustic amplification, patient voice command interface
View ProductPOS & Kiosk Hardware
Far-field microphone array for kiosk voice ordering, 4-mic beamforming, AEC against speaker playback
View ProductAI Vision Systems
Audio event detection synchronized with video, USB Audio Class 2 capture for ML training data collection
View ProductIndustrial Equipment Monitoring
Acoustic anomaly detection microphone front-end, high-SPL capsule selection for machinery environments
View ProductWireless Communication
Full-duplex VoIP with AEC, narrowband/wideband codec selection for constrained RF links
View ProductSmart Home Automation
Multi-room LE Audio synchronization, low-latency speaker feedback for alarm and chime applications
View ProductRuggedized Devices
IP67-sealed acoustic port design, noise-canceling mic for high-ambient-noise push-to-talk radios
View ProductEnd-to-End Signal Chain Ownership
Codec register errors causing pop artifacts, insufficient PSRR causing 217 Hz GSM buzz, DMA buffer glitches at wrap — these boundary failures vanish when both sides are owned by the same team. Noise budgets, analog ground planes, I2S DMA drivers, and NLMS AEC parameters are co-designed so cross-domain failures are caught before tapeout.
Algorithm Expertise on Resource-Constrained Targets
AEC residual echo 12 dB over spec from a cabinet-induced 4 ms late reflection requires extending tap count or adding a nonlinear post-filter — CPU cost of each must be known before the decision is made. Fluency in NLMS, MVDR beamforming, and spectral subtraction means products pass first certification rather than cycling through DSP tuning.
Production-Ready Acoustic Test Infrastructure
Consistent acoustic quality across 50,000 units requires production test infrastructure, not just a working prototype. Production test fixtures, IEC 60268-grounded procedures, and Audio Precision automated scripts are standard outputs of every audio engagement — tolerance limits derived from the noise budget, golden-unit files calibrated, and the CM operates the station without audio expertise.
Deliverables
Audio Signal Chain Specification
Written signal chain specification defining every gain stage, dynamic range allocation, noise budget, interface standard, and performance target — the contract between hardware and firmware teams and reference for production test limits.
Annotated Schematics & Layout Guidelines
Full Altium or KiCad audio subsystem schematics with design decision annotation notes, plus PCB layout guidelines for ground partitioning, keepout zones, and decoupling strategy.
Codec & Peripheral Driver Source Code
Production-quality C drivers for the selected codec, PDM/I2S DMA engine, and peripheral ICs — with register-level bring-up sequences, error handling, and unit test hooks.
DSP Algorithm Implementation
AEC, ANC, beamforming, or noise suppression algorithm source code in C with CMSIS-DSP optimization, tuning parameter documentation, test corpus methodology, and performance characterization on target hardware.
Acoustic Measurement Report
Free-field frequency response, polar pattern, THD+N, noise floor, and maximum SPL measurements in a formal test report with pass/fail assessment against the signal chain specification.
Production Test Specification & Scripts
Automated production test sequences covering frequency response, SNR, THD+N, and channel balance — Python or LabVIEW scripts compatible with Audio Precision hardware, with pass/fail limit files.
Bluetooth Audio Profile Validation Report
Bluetooth audio profile validation report covering A2DP/AVRCP or LE Audio compliance, latency measurements, and codec quality assessment under simulated packet loss conditions.
BOM & Supplier Recommendations
Audio-specific BOM with preferred and alternate sources, lead time notes, safety stock recommendations, and PSRR/SNR comparison table for viable second-source codecs.
All audio deliverables are provided in source-editable formats with documentation sufficient for your team or a subsequent contractor to maintain and extend.
Often Paired With Audio Engineering
Electronics Design
Audio subsystem schematics and PCB layout guidance integrated into the full product electronics design.
ExploreFirmware Development
Codec drivers, I2S DMA engines, and DSP integration delivered as part of the full firmware stack.
ExploreRF & Wireless Engineering
Bluetooth audio profile qualification and coexistence testing with Wi-Fi and other 2.4 GHz radios.
ExploreEmbedded UI/UX
Haptic feedback, LED audio indicators, and voice prompt playback coordinated with the user interface design.
ExploreWork That Demonstrates the Standard
Far-Field Voice Interface for Industrial Kiosk
Ankh designed a four-mic circular PDM array feeding a Cortex-M7 running delay-and-sum beamforming and NLMS AEC, with EMI filtering eliminating 217 Hz GSM buzz from the co-located cellular modem. Free-field testing confirmed −94 dBFS A-weighted noise floor with 35 dB ERLE under double-talk, yielding >97% word recognition accuracy after cloud ASR integration.
Read the Case StudyLE Audio TWS Earbuds with ANC
Ankh implemented LC3 over Bluetooth isochronous channels for a TWS earbud, achieving 4.2 ms inter-earbud synchronization while running a hybrid feedforward/feedback ANC suite on Cortex-M33 within a 6 ms frame budget. THD+N at rated SPL measured 0.07%, and the product completed Bluetooth SIG qualification via QDID inheritance.
Read the Case StudyReady to engineer audio that actually sounds right?From mic selection to production test — we own the full signal chain.
Tell us your SNR requirement and acoustic use case and Ankh will scope the hardware, firmware, and DSP engagement to hit your noise floor and THD targets on the first spin.