Engineering notes
Notes from the bench
Short notes on problems we fix often. One problem each, why it happens, and what we check.
Industrial IoT connectivity
Adding connectivity to an existing product through its serial port
How to connect an instrument or machine that already works, without redesigning its main board: what to check, and where the approach stops working.
Reading Modbus energy meters: register maps, polling rates and byte order
Why two energy meters with the same Modbus registers can give different numbers, how fast you can really poll an RS485 bus, and how to handle counters.
Why an RS485 link works on the bench and dies in the field
Ground potential, surges and missing isolation: what kills RS485 links on site, and what we check on a schematic before the board is built.
PCB design
Battery life on paper versus in the field: measuring sleep current properly
How to measure microamp sleep current without a burden voltage error, why wake bursts dominate the average, and how to calculate real battery life.
BoM versus schematic checks before you order parts
The mismatches between a bill of materials and a schematic that waste a board spin, and the checks we run before any parts are ordered.
Designing a board so a jig can test it in two minutes
Test points, a programming header, a bed-of-nails fixture and a short test sequence: how a board is laid out so production testing stays fast.
Driving relays from a microcontroller: what to check before production
GPIO current limits, flyback diodes, boot-time states and contact ratings. The relay checks we run on every mains-switching board.
ESD on exposed connectors and buttons: where the zap goes
How IEC 61000-4-2 levels translate into TVS placement and return paths on a board, and what a TVS diode cannot fix on its own.
Wide-input power for Indian sites: surges, brown-outs and generator changeovers
How to protect and design a power input for mains that sags, spikes and switches to a generator, with specifics on MOVs, TVS parts and hold-up time.
Embedded firmware and MCU choice
ESP32 resets you only see on site: strapping pins and brown-out
Two hardware causes behind 'random' ESP32 and ESP32-S3 resets in the field, and the checks we run before a board goes to fabrication.
MQTT topic and payload design that still works at a thousand devices
Topic hierarchy, QoS choice, retained messages, payload versioning and last will: the decisions that are cheap on day one and painful to change later.
OTA updates that cannot brick a device
A/B slots, image verification, boot confirmation and rollback: how a firmware update can fail at any point, including power loss, and still leave a working device.
Watchdogs: the settings that turn a hang into a clean restart
Hardware and task watchdogs, where to feed them, how to pick a timeout, and how to log why the restart happened so the fault can be found.
What to log on a device so you can debug it from the office
Reset reasons, event counters and RAM ring buffers: the small set of data that lets you diagnose a field failure without driving to site.
Prototype to production
Cloud, apps and dashboards
BLE setup apps: why pairing is the most-reviewed screen in the app
Onboarding flow, Wi-Fi credential handover over BLE, error messages and Android and iOS permission differences that decide first-run success.
Dashboards a plant team actually opens every morning
Why most IoT dashboards are visited twice and then ignored, and the design choices that make a plant team check one before the first shift meeting.
Provisioning: how a device gets its identity at the factory or on site
Serial numbers, per-device credentials, certificates and claiming a device to a customer account: how to give every unit its own identity safely.
Storing time-series data from devices without a large cloud bill
Sampling versus reporting rate, downsampling, retention tiers and compression: how to keep device data useful for years at a small and predictable cost.
Micro-apps and AI agents
A morning energy report on WhatsApp: what to send and what to leave out
What a daily plant energy message should contain, how to keep it short, and what to know about WhatsApp Business API templates and opt-in.
AI agents with a person in charge: approval steps, logs and widening what an agent may do
How to give an AI agent a narrow job, put an approval step in front of anything that leaves the building, keep a log, and widen its permissions with evidence.
Asking your datasheets and test reports questions: what a document assistant gets right and wrong
A document assistant can find the right page in a datasheet fast. Here is how retrieval, tables, scanned PDFs and invented numbers decide whether to trust it.
Calibration tracking for instruments and gauges, before the audit
What a calibration register must hold: due dates, certificates, traceability and out-of-tolerance handling, so no gauge is found overdue at an audit.
Downtime logging that operators actually use
How to get honest downtime data from the shop floor: detect stops from machine data, ask for the reason with one tap, and keep the list short.
Micro-app or ERP module: how to decide where a shop-floor task belongs
Five questions we ask before choosing between an ERP module and a small purpose-built app for a shop-floor task, and the cases where the ERP should win.
Offline-first shop-floor apps: keeping work going when the network drops
How a shop-floor app keeps recording work when Wi-Fi fails: local storage, queued writes, conflict rules, and the records that must never be lost.
One-step device provisioning at the end of the line
How to write the serial number, credentials and cloud registration in a single operator action, and what to do when one of the three steps fails.
Quality checks with photos on a tablet: designing the checklist so it gets filled in
A tablet inspection checklist only helps if inspectors use it. Layout, photo rules, reject reasons and offline use that keep the records complete.
Shift handover summaries from machine events and operator notes
How to turn overnight machine events and short operator notes into one page for the morning meeting, and where a person must still check it.
Tracking returned units by serial number so repeat faults show up early
How a returns register tied to serial numbers and production batches shows repeat faults in weeks, and why symptom and root cause need separate fields.
Turning machine alarms into maintenance tickets with history attached
How to turn an alarm into one ticket with the machine's recent readings and past faults, so the technician starts with context and not a phone call.
Watching your BoM for end-of-life and shortage notices before they stop production
How to track lifecycle status, PCN and EOL notices and stock for every BoM line, so a last-time buy or second source is decided before the line stops.
What a production test station app should record for every unit
The fields a test station app should store for each serial number, from measured values and limits to firmware version and retest history, and why.
AI for IoT
Edge or cloud for machine learning on sensor data: how to decide
Five questions on latency, bandwidth, privacy, power and updates that decide whether a sensor model runs on the device or in the cloud.
Setting a machine's baseline before calling anything an anomaly
An anomaly alarm is only as good as the baseline behind it. How to collect one, handle operating modes, and set thresholds that do not cry wolf.
Vision on a production line: why lighting and camera position come before the model
Most failed inspection projects fail on light and geometry, not on the neural network. What to fix and fix first before you train anything.
Voice assistants for the shop floor: language, noise and what to log
What breaks when voice commands move from an office to a noisy plant: mixed languages, microphones, wake words, and what to record for improvement.
What an FFT on a microcontroller can and cannot tell you about a motor
Vibration monitoring at the edge: what a MEMS accelerometer and an on-chip FFT can detect on rotating machines, and where you need more.
Building a hardware product in India
BLE, Wi-Fi, LoRa or LTE Cat-1 for a product sold in India
How to pick the radio for a connected product in India: range, power, data, running cost, networks and approvals. A short decision guide.
Choosing a cellular module as 2G and 3G shrink in India
How to pick between LTE Cat-1, Cat-M and NB-IoT for a new device in India, and which availability and certification questions to settle first.
Sourcing components in India: authorised distributors, lead times and counterfeit parts
How to spot counterfeit risk, why authorised distributors reduce it, and which lifecycle checks to run before a part goes on your bill of materials.
What to ask a contract manufacturer before the first order
Questions on capability, testing, quality records, design feedback, minimum order quantity and traceability that separate a good build partner from a risky one.
WPC, TEC and BIS: which approvals a connected product may need in India
A general guide to what WPC, TEC and BIS each cover for connected hardware in India, and how to confirm what applies to your product.
Company news and talks are on our news page.
Tell us what you are building.
Book a discovery call