Talk to usTalk to usEngineering deep-dives, architecture decisions and the reasoning behind them.
The site volunteered, the integration was unbilled, the benefit was measured by its advocates, and nobody owned it afterwards. Six structural reasons, and the reframing that fixes them.
Consequence, likelihood and lead time — and it is lead time that inverts the intuitions built on price. Why a €50 sensor can outrank a €50,000 drive, and the three metrics that tell you if you got it right.
They are not competitors: a register protocol, an information model and a transport pattern. What each is genuinely good at, the two failure modes that recur, and why the choice is per link rather than per company.
The screen is drawn on a quiet afternoon and read during an upset. Why colour should be reserved for abnormal, why a bar beats a number, and the one test that tells you whether the design works.
A FAT moves risk before the machine moves. Most of them demonstrate the happy path instead — here is the list of failures to force, and why a test that passes everything first time tested nothing.
Protective earth and signal reference are different jobs sharing a word, and confusing them is where most noise starts. Where to bond a shield, why the frequency changes the answer, and what to measure first.
The arithmetic is trivial and the definitions are where it breaks: the denominator, the ideal cycle time, the stop threshold and where quality is counted. Four choices that move the number thirty points.
The pilot of ten was configured by hand by the people who built it. At a thousand, the manual steps stop being slow and start being the defect. Identity, claiming, verification and the paths nobody plans for.
Segmentation is not a firewall purchase, it is a list of what talks to what. Why VLANs alone are not a boundary, how to write a conduit down, and the exception that quietly eats the design.
Nyquist is about the highest frequency present, not the highest one you care about — and everything above half your rate folds back in permanently. What rate each job actually needs, and the six things to write down.
Four clocks touch every reading and nobody decided which one counts. What that costs, how much accuracy each use actually needs, and the short list of decisions that fixes it for good.
The link will fail. Whether that costs you an inconvenience or a hole in the record comes down to how the gateway buffers, what it drops first, and how politely it drains when the connection returns.
Tell us what you're building and we'll help you scope the first deployment.