Engineering Template – Under-Car Sound Diagnostic System

Introduction

This project aims to develop a modular, attachable acoustic diagnostic device for vehicles. The system uses an array of microphones mounted on a flexible magnetic sheet, which adheres to the underside of a vehicle. During operation, it records a spatially aware “sound map” while driving. The data can then be analyzed to detect abnormal mechanical noises, localize their source, and compare them to a database of known vehicle sounds, enabling precise diagnostics and machine learning-based improvements over time.

Table of Contents

URS
Physics
Free Body Diagram
Design
BOM Bill of Materials
Electronics
Code
Modeling
Fabrication
Guides
Notes
Contact

URS

User Requirement Specifications

Constraints	Approach	Achieved
Size	Flexible sheet, ~36″ x 24″
Depth	< 1 inch
Height	Minimal to avoid dragging
Width	Scalable array layout
Weight	< 5 lbs total
Cost	<$200 for prototype
Power	Rechargeable LiPo or car battery
Decibels	Detect range: 10 dB–120 dB
Geometry	Grid or hex array
Case	Waterproof & dustproof enclosure
Assemble	Modular plug-and-play mic units
Temperature Range	-10°C to 60°C	car exhaust can reach up to 1200°F

Physics

• Sound Triangulation:
  t = d / v
  Where t is time delay between microphones, d is distance between microphones, and v is speed of sound (~343 m/s)
• Heatmap Generation:
  Use inverse square law + time delays + relative amplitudes for source localization
• Noise Signature Analysis:
  Fourier Transform (FFT) for frequency domain analysis
• Correlation Models:
  Use cross-correlation to match to known good models (i.e., r = sum(x_t * y_{t+lag}))

FBD

Design

The system consists of:

• Magnetic rubber sheet
• Grid of digital MEMS microphone
• Microcontroller, ESP32-S3 with USB-C native control
  -individual node SD card for sound wave recording.
• USB hub for all nodes. Ideally three USB hubs each with 10 ports.
• Raspberry pi is an open web server that links all of the nodes and communicates all commands.
• a book converter and wires will attach to the cars battery to power the node array.

Each microphone module plugs into a snap connector and can be replaced or expanded.

BOM

1. ESP32-S3 x 12
2. 10-Port Powered USB 3.0 Hub X 3
3. MEMs Microphone INMP441 x 24
4. SD Card x 24
5. Raspberry Pi x 1
6. Buck Converter: 12V → 5V 15A
7. Magnets
8. USB Cords

Electronics

Schematics

Basic schematic:
ESP32–> MEMs microphone
ESP32–> MEMs microphone

Wokwi

TBD: Simulate a quadrant of 4 mics and 1 ESP32 on Wokwi

Notes

• Use twisted pair wires for mic connections to reduce EMI
• Each mic can be timestamped for synchronization using ESP32 timers

Code

Sequence Diagram

1. System powers on
2. ESP32 initializes and syncs time
3. Mics start recording in buffer
4. Audio timestamped and saved or streamed
5. Data aggregated and processed into heatmap
6. Uploaded to cloud or displayed locally

Repo

TBD – Github link

Notes

• Consider using TensorFlow Lite for sound classification
• Use JSON metadata for recordings (make/model, mileage, etc.)

Modeling

Parametric Model

Model the case, mic holders, and flexible mount

Revisioning

Rev A – Fixed array
Rev B – Modular snap-in array
Rev C – Wireless mic test

Notes

• Create 2D drawing for laser cutting the magnetic sheet
• Waterproofing seals and vents for sound clarity

Fabrication

• [ ] 1. Design mic holder clips in CAD
• [ ] 2. Cut magnetic base to shape
• [ ] 3. Solder mic boards and ESPs
• [ ] 4. Mount to flexible sheet
• [ ] 5. Waterproof enclosure for battery and ESP
• [ ] 6. Power and test mic signals
• [ ] 7. Drive test with recording
• [ ] 8. Analyze sound and generate heatmap

Revisioning

Keep logs in GitHub issues for bugs and improvements

Notes

Hot-glue for temp tests; silicone sealant for final

Guides

Assembly Guide

TBD – Google Doc

Operating Guide

TBD – Instructions on charging, attaching, starting recording, and viewing results

Notes

Video demo idea: Show mic array going under car, drive around, isolate sound location, compare to database

Notes

• Investigate onboard AI chip (like ESP32-S3 or Edge TPU)
• Might be valuable for EVs where sound issues are more subtle
• Potential commercial applications: dealerships, inspection centers, race teams
• Build a sound profile library based on VIN scan