build projects
Home Projects FPGA Neural DSP Accelerator
Major Capstone Major Project • Verilog HDL
VLSI & Digital Systems VLSI • ECE • Embedded Systems Final Year Capstone (7th/8th Sem)

FPGA Neural DSP Accelerator

A dedicated digital hardware coprocessor designed on a Xilinx FPGA. Implements pipelined Multiply-Accumulate (MAC) DSP blocks for real-time edge audio feature extraction and 1D convolution with zero soft-processor overhead.

Bench-tested hardware preview and circuit schematics for FPGA Neural DSP Accelerator: Hardware-accelerated fixed-point CNN convolution engine implemented in synthesizable Verilog HDL.
Hardware Bench Preview

Component Bill of Materials (BOM)

Total: ₹4,280
1. Xilinx Spartan-6 / Spartan-7 FPGA Development Board
Qty: 1 ₹2,850
2. Platform USB JTAG Cable & Programmer Interface
Qty: 1 ₹650
3. I2S Master Audio Codec (PCM5102A DAC / INMP441)
Qty: 1 ₹320
4. 1.3" SPI TFT Display for On-Chip Register Telemetry
Qty: 1 ₹280
5. High-Speed Logic Analyzer Test Header Pinout
Qty: 1 ₹180

Academic Deliverables Included

  • University Major Project Dissertation with RTL register schematics and FSM state charts
  • Clean synthesizable Verilog source code (.v) and comprehensive testbench suite (.tb)
  • Vivado / ISE synthesis reports, timing constraint files (.xdc), and power dissipation tables
  • Examiner Viva defense presentation with DSP math breakdown and RTL walkthrough

Bench Testing Benchmarks

  • Synthesizable Verilog HDL with static timing closure achieved at 100MHz clock
  • 8-stage parallel MAC pipeline reducing inference latency by 14x vs ARM MCU
  • RTL simulation testbench verified in ModelSim / Vivado with bit-exact vectors
  • On-chip BRAM utilization under 65% leaving room for custom student modifications