Which bench numbers we calibrate, and how often

Filed from the vice · Bay View, Milwaukee

I maintain the calibration schedule here, which is a slightly grand way of describing a laminated card on the wall next to the loft machine and a set of certified reference masses in a drawer. It is the least interesting part of running a build bench and it is the part that makes every other number on this website mean anything.

The gram scale, weekly

The digital scale reads to a tenth of a gram and it is checked every Monday against a set of calibration masses at 10, 100 and 500 grams. It almost never drifts. When it does, it is because something got spilled on the platform or because it was moved and is no longer level.

Weekly is more often than the manufacturer asks for. It takes ninety seconds and it means that if a drift ever does appear, the maximum amount of work built on a wrong number is one week's worth. That ratio is the whole reasoning behind every interval on the card.

The swing balance scale, monthly

Balance scales drift more than digital gram scales because they are mechanical and because they get knocked. Ours goes through a three point check on the first of the month using a purpose-built reference club that never leaves the shop and whose true reading is known.

This is the gauge I trust least between checks, and it is also the one whose readings we publish the tightest tolerance on. Half a point across a set is not a lot of room. If this instrument is out by even a third of a point, a set that measures matched is not matched.

The loft and lie machine, quarterly and after every incident

Quarterly on the calendar, and immediately any time the machine gets used hard on a stubborn cast head or anything gets dropped near it. The check uses two reference heads of known loft and lie.

Worth being honest about a limitation here: a loft and lie machine measures the head as clamped, and how a head sits in the clamp introduces a small amount of operator variation. That is why the same person measures a set before and after wherever possible, and why the sheet records who took the measurement. It is not a certification, it is traceability.

Every number on a build sheet has a person's name attached to it for the same reason every weld has a stamp. Not to assign blame. To make it possible to find the cause.

The frequency analyser, twice yearly

Checked in March before the season and again in September, using three reference shafts kept specifically for the purpose. Frequency analysers are fairly stable instruments, but clamp wear changes readings gradually and the clamp is the part that does all the work.

The bigger source of error with frequency measurement is not the instrument at all, it is technique: clamp pressure, tip mass, and how the shaft is plucked. Which is why one person does all the frequency work here rather than sharing it around.

The launch monitor, at every session

Not calibration in the strict sense, but the equivalent discipline. The unit is levelled and its alignment confirmed at the start of every session day, and the ambient temperature gets logged. A radar unit that has been nudged half a degree out of alignment will produce beautifully precise numbers that are quietly wrong.

This is also why every fitting starts with the golfer's own club. If the baseline and the test are taken on the same unit on the same day under the same conditions, then even a systematically offset unit still gives you a valid comparison. Absolute numbers are fragile. Differences are robust.

Why any of this is your business

Because a tolerance window is a claim, and a claim you cannot audit is a marketing line. Our published build window says half a point on swing balance and plus or minus a thirty-second of an inch on playing length. Those figures are only meaningful if the instruments producing them are checked on a schedule that somebody actually keeps.

The card is on the wall by the loft machine. It is not hidden and you are welcome to read it when you are in for a session. If a shop cannot show you something equivalent, treat their tolerance claims as aspirations rather than specifications.

Back to the bench log