Two Instrument Classes, One Golf Shot

Strip away the branding and there are two ways to find out what a golf ball is doing. You can bounce radio waves off it while it flies and read the frequency shift in what comes back, or you can photograph it several times in a few thousandths of a second and measure how far it moved between frames. Everything else is engineering around one of those two ideas.

Both work. Neither observes everything. And the parts each one does not observe get filled in by inference, which is a perfectly respectable thing to do as long as the sheet says which is which.

What the Radio Waves Actually See

A doppler unit sits behind the golfer and watches the ball recede. Because it is tracking the ball through a stretch of its flight rather than at a single instant, it is genuinely good at anything that reveals itself over time. Spin is the headline case: a spinning ball produces a small, characteristic modulation in the returned signal, and the unit reads the rate directly rather than deducing it.

The same property makes it strong on the later part of the flight, where it can watch the ball slow, and on club delivery, where the head is a large moving object with a distinctive signature.

Its weakness is the first few milliseconds. At the moment of separation the ball has barely moved, and there is very little flight for the unit to look at. Launch angle in particular is therefore derived from the early trajectory rather than seen at the face, and if the ball is stopped by a screen quickly, that derivation is working from a short sample.

What the Cameras Actually See

A photometric unit does the opposite. It fires a very short sequence of exposures in the interval around impact, and measures the ball's position and orientation in each one. That gives it the launch geometry directly: how fast the ball left, at what vertical angle, and in what direction, all read off images rather than inferred.

Spin is where it works harder. The unit needs something on the ball to track between frames, which is why marked balls exist and why dimple-pattern recognition is such a difficult engineering problem. When the marking is clean and the lighting is right, the result is excellent. When the ball is scuffed, the lighting is off, or the mark rotates out of view, the reading degrades, and the honest failure mode is to report nothing rather than a bad number.

Its other weakness is that it sees a tiny window. Everything after the ball leaves that window is outside its knowledge entirely.

Column Ownership

Here is how those two capability sets map onto the rows that appear on a fitting sheet. This table is the reason we run both units on every full-swing block rather than picking a favourite.

Notice the bottom three rows. Neither instrument observes carry indoors. Both compute it from the rows above plus assumptions about the air. A studio with the most expensive unit on the market and a studio with the cheapest one are, on those three rows, both showing you arithmetic.

Disagreement Is a Measurement in Its Own Right

Running two instruments is not about averaging them into a better number, although that is a pleasant side effect. It is about having a second opinion on every strike.

When the units agree within the published band, we have two independent routes to the same answer and the confidence in that row goes up accordingly. When they disagree beyond the band, something about that strike went wrong: the ball rotated its marking out of view, a low strike sprayed material across the lens, the radar picked up a neighbouring bay, or the contact was so far off centre that the early flight and the impact geometry genuinely tell different stories.

That strike then goes on the discard list with the reason attached. It is the single most common discard category we record, and the count itself is diagnostic:

• One or two disagreements in a block is normal and unremarkable.
• Six or more usually means the ball is wrong, worn, or badly marked.
• A cluster on one candidate club often means genuinely erratic contact, which is a finding.
• A cluster at one time of day has twice been a lighting problem in Bay 5.

A single instrument cannot tell you it is having a bad afternoon. It can only tell you what it thinks it saw, with the same confidence either way.

What to Ask a Studio That Runs One Unit

Most studios run one, and running one is entirely defensible. The questions worth asking are not accusations, they are just requests for the label.

Which of my columns did this instrument observe directly? Any technician who knows their equipment can answer in ten seconds, and the answer tells you which rows to lean on.

What is the repeatability on the columns it infers? If the answer is a number, excellent. If the answer is that nobody has checked, that is worth knowing before you make a component decision on that row.

What ball, and what marking? On a camera unit this determines whether spin is a strong reading or a struggling one. On a radar unit it matters much less.

Is carry observed or computed? Indoors it is computed, always, on every system ever built. A studio that says otherwise has misunderstood their own equipment, and that is worth knowing too.

Corrections: the angle-of-attack row originally read "observed" for both instrument classes. Amended 8 April after a reader pointed out that our radar derives it. He was right.