I build systems where physical behavior meets computation.

02 / HOW I THINK

Every system has a limit.
The interesting work starts there.

01Constraints.

02Architecture.

03Timing.

04Computation.

05Physical behavior.

06Trade-offs.

07Implementation.

Not a list of disciplines. A method for deciding what the system needs—and what it can afford.

03 / SELECTED WORK

The architecture
is the argument.

One problem, traced from physical constraint to measured implementation.

PROJECT / 01

Digital hardware / real-time estimation · 2024

FPGA-Based Angular Position & Velocity Estimation

ESTIMATOR / SIGNAL PATHCLK 100 MHz

θ[n]ω[n]LATENCY / BOUNDED

01Problem

Observe motion without losing time.

Position arrives as an imperfect signal. Velocity must be inferred before latency destabilizes the loop.

02Architecture

Make the computation explicit.

Acquisition, conditioning, estimation, and output become a clocked signal path—not a software abstraction.

03Constraint

Resources are part of the equation.

Word length, multiplier count, and pipeline depth determine what the estimator can become.

04Decision

Build the datapath around the physics.

Parallel stages preserve throughput while fixed-point analysis controls error and silicon cost.

05Implementation

Model. Synthesize. Measure. Repeat.

MATLAB establishes the reference; Verilog carries the behavior into a Spartan-6 implementation.

06Result

Determinism, verified in hardware.

A bounded-latency estimator that resolves position and velocity at the cadence of the control system.

RESOLUTION

Hardware-validated, deterministic estimation architecture.

Constraint: Finite logic / Fixed-point precision / Cycle-level timing
Implementation: Verilog / Spartan-6 / MATLAB / Fixed-point DSP
Method: Translate the signal path into a purpose-built datapath, then validate behavior against a numerical reference.

Request full case study