Summary of S4-260236: Device Microphone Array Characterization

Introduction and Motivation

This contribution addresses the characterization of UE microphone arrays, proposing an extension to the IMPro (Integrated Microphone Pressure frequency response) method to handle devices with non-identical microphone integrations. Traditional anechoic chamber measurements are time-consuming and require laboratory automation. The IMPro method combined with numerical acoustic modeling offers a more flexible alternative, but practical challenges arise when microphone integrations differ across the array.

The key technical challenge is that when microphones have different port lengths, cavity shapes, meshes, or ingress-protection structures, the relative inter-channel propagation delay (RIPD) between microphones can vary. This is particularly problematic when measurement devices are not accurately synchronized with the UE's audio clock, which can affect spatial audio algorithms that rely on relative timing information.

Factors Affecting Microphone Integration Variation

The document identifies multiple factors that can cause integrated microphone characteristics to vary within a single UE:

• Device shape and surface curvature around sound inlets
• Mixture of different microphone transducer types
• Device hardware mechanics and PCB layout
• Ingress protection meshes and gaskets
• Microphone inlet integration in earpiece or loudspeaker outlets
• Protection against user misuse

Experimental Validation

Test Setup

The source conducted experiments using:
- A simplified microphone integration jig with removable tubes of different lengths (5mm, 7mm, 8mm, 10mm)
- An array of commercial and custom-built probe microphones
- Analog transducers enabling accurate synchronization

Case Study: DoA Estimation Error

A case study was performed using DaCAS target Device B geometry with microphone positions #1-#4:
- Microphone port lengths: #1: 8.0mm, #2: 8.0mm, #3: 5.0mm, #4: 7.0mm
- DoA estimation performed using GCC-PHAT method
- Device aligned in landscape orientation in horizontal plane
- Direction estimation analyzed at 1-degree resolution across 360 degrees

Results: The largest DoA errors occurred towards both ends of the device in landscape orientation, demonstrating the impact of varying microphone port lengths on spatial audio algorithm performance.

Proposed Technical Solution

Extension to IMPro Method

The contribution proposes extending the IMPro method defined in the DaCAS-2 permanent document with the following technical framework:

2.1-2.2 IMPro Method Baseline

The integrated microphone pressure frequency response H_m(f) is calculated by:

H_m(f) = Y_m(f) / [X_p(f) · C_p(f)]

where:
- Y_m(f): measured integrated microphone output signal response
- X_p(f): probe signal output signal response at reference point at sound inlet
- C_p(f): calibration frequency response of probe microphone

2.3 Compensation Filter Target Response

Key definitions:
- M: number of device microphones
- p_m(t): pressure at m-th microphone sound inlet
- h_m(t): impulse response of integrated microphone using IMPro
- g_m(t): equalization filter for microphone signal compensation
- y_m(t): raw microphone signal output
- z_m(t): compensated microphone output

The compensated microphone output is defined as:
z_m(t) = g_m(t) * y_m(t)

The target equalization filter response compensates for integrated microphone response such that:
z_m(t) = p_m(t - τ_m)

In frequency domain:
G_m(f) = e^(-j2πfτ_m) / H_m(f)

within the range of target frequency response mask.

2.4 Proposed Integrated Microphone Calibration Method

The calibration procedure consists of four main steps:

Step 1: Perform IMPro measurements for all device microphones
- Prepare UE software for raw microphone recording
- Setup loudspeaker and device at 0.5-2m source distance
- Prepare measurement stimulus (sine sweep) at ~30dB above background noise
- Calibrate probe microphone(s)
- Perform IMPro measurement with probe at sound inlet while DUT records
- Time-align measured audio signals
- Calculate H_m(f) for each microphone

Step 2: Calculate compensation filter
- Measure H_m(f) from reference sample or average from multiple devices
- Design linear equalization filters according to equation (2)
- Ensure responses align within equalization filter frequency response mask and difference mask

Step 3: Implement equalization filter in UE software
- Process raw microphone signals y_m with filters g_m to produce compensated outputs z_m

Step 4: Verify compensated microphone output
- Check all microphone signals satisfy equalization filter frequency response mask (target timbre)
- Check all signals satisfy equalization filter frequency response difference mask (spatial information preservation)
- Verify z_m(t) is effectively a delayed version of pressure signal at sound inlet with constant delay τ_m

2.5 Applicability Condition for Synchronized IMPro Measurement

Key Technical Requirement: If the UE employs non-identical microphone integrations (different acoustic port lengths, meshes, gaskets, baffles, or cavity geometries) such that inter-microphone propagation delay differences may be non-negligible, the IMPro measurement must address the RIPD.

This requirement is fulfilled by either:

1. Synchronized IMPro acquisition: All probe signal(s) and all UE microphone signals acquired with shared, repeatable time base, OR

2. IMPro acquisition with probe-array: Simultaneous probe measurements at all relevant inlets during UE recording

For probe-array approach:
- Requires calibrated probe microphones calibrated simultaneously to common reference
- Preferably single multichannel recording with as many probes as UE microphones
- Enables inspection of probe positioning for all microphones before measurement
- Alternative: sequential approach with reference microphone for synchronization (requires ensuring no setup changes during microphone movement)

NOTE: This clause does not alter the IMPro calculation or compensation target defined in sections 2.2 and 2.3.

Conclusion

The source proposes extending the IMPro-based compensation method defined in the DaCAS-2 permanent document with the presented synchronized measurement approach. This enables accurate isotropic characterization of UE microphone arrays even when the measurement device does not guarantee accurate synchronization with the UE audio clock, particularly for devices with significantly different or unknown microphone port integrations.