Multi-IMU Precision Motion Sensor Board

2025-11-13

Overview

For Advanced Robotics at the University of Washington, I was tasked to create a small CAN-enabled IMU board to measure the motion of the our robot. We wanted to use to up to 4 IMUs to improve measurement precision.

Simulation

To understand how multiple IMUs would the resulting measurement, I ran a simulation on python with 8 IMUs, each with a different amount of noise based on the Farrenkompf noise model. This will most accurately simulate what real IMUs data may look like.

The first image shows all the different IMUs and their respective signal (in this case it is a sine wave) and second image shows the average value of all 8 IMUs compared against a regular sine wave.

System Architecture

Sensor Array: 8× LSM6DSVTR IMUs positioned maximally distributed across the board
Data Fusion: Averaging algorithm across all IMUs using high sampling rates (sine wave simulation with noise model)
Communication: CAN bus interface (PA11/PA12) via MCP2562T-E/MF transceiver
Thermal Management: Individual shunt resistors (RNCF0805DTE2K50) beneath each sensor, relay-controlled (VO1400AEFTR) for temperature stabilization

Hardware Design

MCU: STM32F042F4P6 (TSSOP20)

High-speed SPI multiplexing with 3:8 decoder (MC74HC138ADR2G)
Software NSS chip selection (PA1-PA4) for sequential IMU polling
Optimized for minimal switching latency across all sensors

Power System:

Input: 24V → 5V buck converter (AP63201QWU-7, 1A)
5V → 3.3V LDO (MIC5504-3.3YM5-TR, ≤300mA) for MCU
5V → 1.8V LDO (MIC5365-1.8YC5-TR, ~8mA) for IMU array
24V direct feed for heating elements

Physical Specs:

Form factor: 40mm × 40mm (Board Type C compliant)
Mounting: M3 holes at 23mm × 33mm spacing
Operating range: -40°C to 85°C

Technical Challenges