Jadavpur University DSP Validation

Jadavpur University — DSP Validation Project Proposal

Source: newdocs/Jadavpur_University_Proposal_Sections_3_to_7.docx Date: 2026-03-02

Objectives

Design, configure, validate, and freeze a robust Digital Signal Processing (DSP) architecture for industrial vibration monitoring using IEPE accelerometers and/or advanced MEMS vibration modules with embedded ADC and DSP capabilities.

Specific Objectives

• Establish a mathematically correct and industry-compliant signal chain from sensor to feature output
• Define and freeze sampling rate, anti-alias filtering, FFT scaling, windowing, and feature extraction parameters
• Validate RMS, Peak, Crest Factor, Envelope, Kurtosis and FFT outputs against known mechanical references
• Develop a repeatable validation framework for lab-scale test rigs
• Prepare validated data outputs suitable for integration with edge and cloud systems

Scope of Work

1. Sensor Interface & Configuration

• IEPE signal conditioning validation
• MEMS SPI interface validation (if applicable)
• Verification of sampling rate and decimation configuration

2. Signal Processing Architecture

• Digital high-pass filtering (>= 1 Hz)
• Band-limited RMS computation (e.g., 1-1000 Hz)
• FFT implementation with correct amplitude scaling
• Envelope processing using bandpass + rectification + low-pass method
• Spike energy and kurtosis computation

3. Validation & Testing

• Test rig validation under controlled mechanical excitation
• Repeatability testing (+/- 5% RMS tolerance)
• Frequency accuracy validation (+/- 1% at 1X RPM)

4. Documentation & Freeze

• Mathematical definitions of all computed parameters
• DSP configuration freeze document
• Final validation report

Methodology

Phase	Description
Phase 1 — System Review	Review sensor physics, ADC configuration, and signal chain
Phase 2 — Algorithm Validation	Mathematical verification of RMS, FFT scaling, window correction, and envelope detection
Phase 3 — Experimental Validation	Data acquisition from controlled test rig, cross-validation of computed parameters with reference mechanical values
Phase 4 — Optimization & Freeze	Finalization of DSP configuration and performance documentation
Phase 5 — Reporting & Handover	Submission of validated configuration parameters and recommendations

Deliverables

• Validated and operational DSP architecture document
• Mathematical formulation document
• Sampling and filtering configuration report with details
• FFT scaling validation note with sample outputs
• Envelope processing definition and validation results with sample outputs
• Experimental validation dataset and operational results
• Final technical report suitable for accurate and operational industrial deployment

RAPID AI Relevance

This project directly supports Phase 3 (DSP Dual-Path Integration) of RAPID AI v2.0:

• The validated DSP outputs (RMS, Peak, Crest Factor, Envelope, Kurtosis, FFT) map directly to the DSPMetrics model in schemas_v2.py
• The tolerance thresholds (+/- 5% RMS, +/- 1% frequency) inform the DSP_TOLERANCE config values in config_v2.py
• The freeze document will define the exact contract for what edge DSP devices send to RAPID AI
• Module A v2’s _cross_validate_dsp() function compares these DSP outputs against internally recomputed values
• The validation framework ensures DSP metrics sent to RAPID AI are trustworthy (supporting the “physics authority” principle)

Gap Map (DSP Deliverables vs RAPID AI)

DSP Deliverable	RAPID AI Status	Integration Point
RMS (band-limited)	Module A recomputes	DSPMetrics.overall_rms, cross-validated
Peak	Module A recomputes	DSPMetrics.peak, cross-validated
Crest Factor	Module A recomputes	DSPMetrics.crest_factor, cross-validated
Kurtosis	Module A recomputes	DSPMetrics.kurtosis, cross-validated
Envelope RMS	Not currently computed	DSPMetrics.envelope_rms (new capability)
Spike Energy	Not currently computed	DSPMetrics.spike_energy (new capability)
HF Kurtosis	Not currently computed	DSPMetrics.hf_kurtosis (new capability)
FFT Spectrum	Partial (Module A)	Future: spectral cross-validation
Sampling/Filter config	Not tracked	Future: DSP metadata in MeasurementPoint