Journal

Notes from the bench.

Design choices, calculations and lessons from robotics, embedded systems, industrial automation and software. Follow a worked example, investigate a failure or look inside one of my projects.

30 posts

Hardware-grade robotics workbench featuring a differential-drive rover chassis, a three-link aluminium planar arm and a matte-navy controller enclosure.

Slow down to speed up: the engineering case for an unhurried start

Rushing the first week of a project feels productive and usually costs the most time later. Why I spend longer on the problem before touching the solution, with NASA's cost data behind it.

Differential-drive mobile rover with a cylindrical spinning LiDAR turret on a calibration grid floor.

Why a rover's position gets worse the further it drives

A 0.1% difference in wheel diameter puts a differential-drive robot 0.17 m off after 10 m and 0.67 m off after 20 m. The maths behind odometry drift and the calibration that fixes most of it.

Local open-weight language model architecture connecting document ingestion, retrieval, memory and evaluation.

Testing a document-grounded AI one stage at a time

When a retrieval-augmented system gives a wrong answer, the cause can sit in the index, the retriever, the model's reading or the final answer. How I test each stage separately so the failure points to a fix.

Three-link planar kinematics simulation showing joint angle arcs inside a dashed circular reach envelope.

BRIDGE: designing the robot body and controller together

A September humanoid study puts mechanical design inside the learning loop. A look at its reported results, an arithmetic check, and what is actually available to reproduce.

Top-down technical corridor visualisation showing an autonomous mobile robot with concentric clearance rings and braking envelope contours.

Test the robot when the data stops

A practical ROS 2 fault-injection plan for stale observations, delayed commands and controlled recovery, with an illustrative stopping-distance calculation.

ESP32 instrumentation board on brass standoffs wired to a machined aluminium track with four Hall-effect sensor modules.

The second life of a product

What a product reveals about its design when someone needs to open it, find a part and put it back into service.

Hardware test rig holding modular HUB75 LED matrix panels connected via Ethernet patch leads to microcontroller receiver boards.

In praise of the volume knob

A small control raises a larger design question: how much attention should an everyday adjustment demand?

ESP32 instrumentation board on brass standoffs wired to a machined aluminium track with four Hall-effect sensor modules.

A smooth sensor trace can hide real motion

How an 80 Hz signal becomes indistinguishable from 20 Hz when sampled at 100 samples per second, and what to check before trusting a trend.

Isometric render of a modular stainless-steel food-processing and conveyor automation line with station status indicators.

99% accuracy can still miss every defect

A worked inspection example showing why defect recall, false rejects and review workload belong beside the headline accuracy score.

Engineering Mastery Lab learning dashboard showing five mechatronics courses and guided laboratories.

The experience trap: when hiring protects the past

Years in a role can tell us something. They cannot tell us everything. A case for assessing capability, supporting graduates and rewarding people who keep learning.

KiCad 3D render of the VW-MTC1 four-layer ESP32-S3 controller board showing USB-C, microSD and sensor connectors.

A noisy sensor needs a decision rule

How hysteresis, debounce and explicit states turn a fluctuating measurement into a useful event without hiding the timing cost.

Panelogram desktop application rendering a six-shelf retail bay to millimetre scale with capacity metrics.

A local app needs a way back

Why offline use, saved data, export and recovery need separate acceptance tests in a small personal or business app.

Interactive M/M/1 model

Little's Law Queue Simulator

Analytical M/M/1 illustration with 1.0 min mean service time (60 jobs/h). See how mean waiting grows nonlinearly as utilisation climbs from 50% to 95%.

0m5m10m15m20m50%65%80%95%
Mean wait (Wq)
4.0 min
Total flow time (W)
5.0 min
Arrival rate
48.0 jobs/h
Mean WIP (L)
4.0 jobs
Modelled WIP buffer (4.0 jobs)

Formula: Wq = (rho / (1 - rho)) x 1.0 min. Analytical M/M/1 illustration, not measured factory data.

Contact

Let's make the next system work.

Discuss an engineering role, a project or a research collaboration. Tell me what you are building, where it gets difficult and what a useful result would look like.

  1. Send the formIt is sent from this page, without publishing my email address.
  2. I reply to youTo the address you give, with questions or a next step.
  3. Project work starts with DefineA short, fixed-fee phase that sets the scope and the tests.

0 of 4000 characters

Sent from this page, so my email address is never published.