MEMS Inertial Sensors for ISRO

M2D2 Lab, IISc · with ISRO · Guided by Prof. G K Ananthasuresh · Jun 2022 – Jul 2024

Role: Design lead
Team: IISc M2D2 Lab with ISRO
Tools: COMSOL Multiphysics · CleWin · KLayout · MATLAB · SolidWorks

What I did

Designed a dual-axis capacitive accelerometer and an accordion-like disk resonator gyroscope (ADRG) through parametric sweeps of beam, ring and electrode geometry
Ran COMSOL eigenfrequency and electromechanical studies across five gyroscope topologies, with <5% error versus fabricated devices
Prepared CleWin/KLayout fabrication layouts and co-authored the ICMNS 2024 paper

SEM micrographs of fabricated ADRG showing accordion-like ring structure at multiple magnifications (28X, 200X, 450X)

Over a two-year collaboration between IISc and ISRO, this project aimed to develop a complete MEMS Inertial Measurement Unit for space navigation applications. The IMU comprised two core sensors: a dual-axis capacitive accelerometer and an Accordion-like Disk Resonator Gyroscope (ADRG). Starting from an existing ring-and-spoke gyroscope concept, extensive parametric optimization was performed on both devices using COMSOL multiphysics simulations. Both devices were fabricated at CeNSE, with the gyroscope validated via laser Doppler vibrometry achieving <5% eigenfrequency error.

12.1 µm/rad/s Mechanical sensitivity: 3–4× prior designs

6.77 kHz Drive-mode eigenfrequency: matched within <5% in test

181.6k Q factor Highest of the five topologies tested

46.4% ΔC₀/C₀ Largest relative sense-capacitance change

Problem Statement

Need: ISRO required high-precision inertial sensors for space navigation applications
Prior art: Existing DRG topologies (standard DRG, Honeycomb DRG (HDRG), Concentric DRG (CDRG), and Gear-tooth DRG (GDRG)) each advanced sensitivity incrementally but always traded electrode placement against mechanical performance

We set out to find a ring geometry that would raise mechanical sensitivity and quality factor without hurting noise performance or making the device harder to fabricate.

Methodology

The project followed a design-simulate-fabricate-test workflow spanning both accelerometer and gyroscope development.

Design & Simulation

Accelerometer Design

The MEMS accelerometer employs a proof mass suspended by folded-beam flexures with interdigitated comb-drive electrodes for differential capacitive sensing.

Optimization Accelerometer optimization parameters

A parametric study varied beam width (5, 6, 7, 8 µm) and out-of-plane thickness (10, 20, 30, 60 µm) to optimize:

Eigenfrequency (~3 kHz target)
Displacement sensitivity
Capacitance sensitivity (ΔC/g)
Cross-axis sensitivity
Quality factor
Brownian noise equivalent acceleration

Gyroscope Topology Comparison

Five DRG topologies were evaluated using COMSOL Multiphysics. The novel Accordion-like DRG (ADRG) introduces accordion-shaped spoke connections that increase mechanical compliance while maintaining structural integrity.

Held constant across all five topologies for a fair comparison.

COMSOL eigenfrequency mode shapes of ADRG analyzed by FEM — Mode shapes of ADRG analyzed by FEM showing eigenfrequencies from 2344 Hz to 8068 Hz

Note Why accordion spokes?

Accordion-shaped spokes increase radial compliance while maintaining tangential stiffness. This geometry allows for larger modal displacements under Coriolis forces, directly improving mechanical sensitivity without compromising structural integrity or introducing additional thermoelastic damping paths. Eigenfrequency studies identified drive and sense mode frequencies, with mode matching critical for gyroscope sensitivity.

Mechanical Sensitivity (×10⁻⁶ µm/rad/s):

DRG

3.83

HDRG

4.27

CDRG

3.61

GDRG

2.76

ADRG

12.1

Data Complete topology comparison table

Disk Resonator Gyroscope Topology Comparison (9 rings, 8mm OD, 10µm ring width, 60µm thickness)

DRG Type	Freq (Hz)	Q (k)	ΔC₀/C₀ (%)
DRG	12,720	103.62	14.37
HDRG	11,306	105.42	16.39
CDRG	12,611	99.99	26.36
GDRG	14,761	91.58	29.40
ADRG	6,771	181.55	46.44

Fabrication

Mask Layout Generation

Once mechanical and electrode parameters were finalized through simulation, designs were converted into semiconductor fabrication mask layouts.

CleWin: initial layout design and layer definition
KLayout: advanced editing and DRC (Design Rule Check) verification

SOI Microfabrication

Devices were fabricated at the Centre for Nano Science and Engineering (CeNSE) at IISc using SOI-based surface micromachining.

Batch of fabricated MEMS gyroscope dies from CeNSE fabrication run

Wire Bonding

Wire bonding connected the MEMS die pads to the PCB carrier for electrical interfacing.

Wire-bonded GDRG die showing M2D2 lab markings — Wire-bonded DRG die mounted showing M2D2 lab markings and concentric ring structure

Process SOI microfabrication steps

Start with SOI wafer (60 µm device layer, 2 µm BOX)
Photolithography for device patterning
Deep reactive ion etching (DRIE) through device layer
HF vapor release of buried oxide
Critical point drying to prevent stiction
Metal deposition for electrodes and bond pads

Info Pinout details

Custom pinout schematics were designed specifying electrode-to-pin mappings: the gyroscope required 16 individually addressed outer electrodes plus 4 common connections. This allows independent drive/sense control of each electrode pair for differential operation.

Test Setup

The fabricated ADRG was characterized using a Polytec MSA-500 Micro System Analyzer (Laser Doppler Vibrometer) at atmospheric pressure and room temperature.

Method How the measurement worked

The device was electrostatically driven with AC+DC excitation applied to opposing drive electrodes (0° and 180° phase). Frequency sweeps measured the mechanical frequency response, with the resonance peak identifying the eigenfrequency. The LDV measured out-of-plane velocity at multiple points across the resonator to verify mode shapes matched FEM predictions.

Gyroscope wire bonding pinout with 16 outer electrodes — Gyroscope pinout: 16 outer electrodes + 4 common

Dual-axis accelerometer pinout with X/Y sense and feedback electrodes — Accelerometer pinout: X/Y sense and feedback electrodes

Equipment Full test equipment list

Polytec MSA-500 Micro System Analyzer
Keysight 33500B Waveform Generator
Stanford Research SR830 Lock-in Amplifier
Keithley 2400 SourceMeter (DC bias)
Custom vacuum chamber (for Q-factor measurements)

Results

LDV frequency response showing measured resonance at 6.8 kHz — LDV frequency response validating 6.8 kHz resonance prediction

6.8 kHzMeasured resonance

6.771 kHzFEM prediction

~0.4%Error

The target was under 5%, and the simulations held up once the devices were made.

            Key Finding
            ADRG performance advantages
          
            ADRG achieved highest quality factor (181.55k) among five DRG topologies tested
ADRG demonstrated best relative sense capacitance (46.44% ΔC₀/C₀)
Mechanical sensitivity of 12.1×10⁻⁶ µm/rad/s: 3-4× higher than prior designs
LDV validation showed <5% eigenfrequency error between simulation and measurement

          

            Publication
            Conference paper details
          
            Sai Pranav Avva, Suman Acharya, Kandula Eswara Sai Kumar, Sudhanshu Shekar, Karthik Raveendranath, and G K Ananthasuresh. "A Novel Accordion-like Disk Resonator Gyroscope."
            International Conference on Micro, Nano, and Smart Systems (ICMNS), July 11, 2024.
            
            The paper presents a comparison of five DRG topologies (DRG, HDRG, CDRG, GDRG, ADRG) across mechanical sensitivity, Q factor, capacitance change, and Brownian noise floor metrics.

Where Simulation and Reality Disagreed

The hardest part was the gap between the capacitance I simulated and the capacitance we measured on real devices. The resonant frequencies came out within 5% of the COMSOL predictions, but capacitance varied a lot more, so I couldn't take the simulated sensing numbers at face value. Working out how far to trust the model, and designing with that margin in mind, was the biggest lesson of the project.

Full Report

Conference presentation slides.

Skills Technical skills demonstrated

Simulation: COMSOL Multiphysics, FEM eigenfrequency/thermoelastic analysis
CAD/Layout: CleWin, KLayout, mask design, DRC
Fabrication: SOI micromachining, DRIE, HF vapor release
Characterization: LDV, impedance analysis, wire bonding
Analysis: MATLAB, Python data processing

Team Contributions & credits

I led parametric design optimization for both sensors, performed COMSOL multiphysics simulations achieving <5% eigenfrequency error, generated CleWin/KLayout fabrication masks, designed wire bonding pinouts, and first-authored the ICMNS 2024 conference paper.

Guided by Prof. G K Ananthasuresh and Karthik Raveendranath (ISRO), with simulation support from Sai Kumar and Suman Acharya, and fabrication by Sudhanshu Shekhar.

Funded by ISRO; fabricated and characterized at CeNSE.