ALPHA VISION · ICE HOCKEY
Concept and pilot proposal · not a product

The sport sells the hardest shot.
Almost nobody measures it.

Shot speed is the number this sport puts on the screen, in the programme and in the highlight. In most places it is quoted, it was inferred from video: two positions, a few milliseconds apart, and a division. A Doppler sensor does not infer it. It measures the velocity itself, with nothing inside the puck and nothing on the player. That much is straightforward physics. What it does inside a steel and glass arena is the open question, and this page is about how we would answer it.

Behind goal
Sensor position
0
Hardest shot, true, km/h
0
What it would read, km/h
132 m
Detection range, puck estimate
0 %
Worst shot, low by
01 · The principle

Inference and measurement are not the same number

A camera does not see speed. It sees where the puck was in one frame and where it was in the next, and divides by the time between them. At sixty frames a second a puck travelling at forty five metres per second moves three quarters of a metre between frames, so the answer depends entirely on how precisely those two positions were resolved, and every error in them is amplified by the division. A Doppler sensor reads the frequency shift of the returned signal, and that shift is the velocity. Nothing is differenced and nothing is assumed.

What optics does

Infers speed from two resolved positions. Excellent at identity and at tracking many slow things at once, which is why it owns player tracking and should keep it.

What radar does

Measures the velocity of one fast object directly, in any light, with nothing inside the puck and no chip to charge, replace or certify.

Why it matters here

The hardest shot is a headline, a broadcast graphic and increasingly a data product. It is the one number in this sport where the difference between measured and inferred is commercially visible.

Where we stop

Radar sees motion, not identity. It will not tell you who took the shot, and it will not track twelve skaters. Those are optical problems and we would fuse with whatever already solves them.

02 · Mounting

Range is free here. The mounting position is the decision.

A puck reflects far more strongly than a golf ball, so the whole rink sits inside the detection envelope and range never becomes the constraint. What does matter is angle, because a Doppler sensor measures only the part of the velocity pointing along its own line of sight. The good news is that a shot on goal is travelling almost straight at a sensor placed behind that goal, which makes it close to the ideal geometry. Lift the sensor into the roof and you give a few per cent back. Put it at the side boards and shots on goal become the worst case while cross ice passes become the best. Pick a position and see what each shot would read.

ONE RINK · PLAN VIEW · 60 × 30 M · behind the goalsimulated physics
reads within 2 % reads 2 to 8 % low reads more than 8 % low component the sensor cannot see

03 · What we would have to prove

The honest list, before anyone asks for it

We have never run this sensor in an ice rink. The physics of measuring a puck is not in doubt, and we are not going to pretend the rest of it is settled because of that. These are the questions a pilot exists to answer, and we would rather publish them than be asked for them.

OPEN QUESTION 01 · MULTIPATH · WHAT THE SENSOR ALSO HEARSsimulated physics
Board and glass reflectivity 45 %
the puck, direct path ghost, arrives via a reflection too weak to detect

Every flat surface is a second route to the puck

A rink is a box of boards, glass and steel. The signal reaches the puck directly, and it also reaches it after bouncing off a board, which takes longer and comes back at a different apparent range and a different apparent speed. As far as the sensor is concerned each of those is a target. The map on the left shows where they appear to be.

Why this is honest rather than alarming. A tracker rejects most of these, because a puck obeys physics that a reflection does not: it cannot be in two places, it cannot accelerate at the rate a ghost appears to, and it persists between frames while a ghost flickers. We have built exactly that rejection layer for a faster and harder problem. What we cannot do from a desk is tell you how many survive in a real building with a real crowd. That is the number the pilot produces, and it is the reason there is no accuracy figure anywhere on this page.

Open question 01

The building

The widget above is a model. A real arena adds a crowd, a roof structure, advertising surfaces and an ice sheet that reflects differently when it is freshly resurfaced. No amount of calculation settles it. One sensor in one rink for one evening tells us more than a month of modelling.

Open question 02

The mounting

Where the sensor goes decides the geometry above. Behind the goal, in the roof, or both. This is a question for the people who run the building as much as for us, because it has to survive a season and an ice resurfacer.

Open question 03

Occlusion

Twenty skaters and two goaltenders stand between the sensor and the puck for much of a game. How much of the play survives that, and whether the shot itself does, is measured rather than argued.

What we would not claim, whatever the pilot shows

Plainly

No puck spin, because we have never measured spin on any object. No player tracking, because video does it better and already does it. No officiating decisions, which rest with the governing body. And no accuracy figure at all until one has been measured in a building rather than calculated on a page.

What would make us stop

A bad night

If multipath in a real arena degrades the measurement past what the sport needs, we will say so and we will not build the product. That is what a validation pilot is for, and it is the reason this page is not a brochure.

04 · The pilot

One rink, one sensor, one half of a season

Deliberately small. The point is not to install a system, it is to answer the three questions above with evidence, in a building that has an ice resurfacer and a crowd and all the reflective surfaces we are worried about.

01
What we bring

The sensor, the mounting, the engineer and the analysis. There is no invoice attached to this and no obligation on the other side beyond letting us in and telling us what you actually need out of it.

02
What we measure

Shot speed on every shot the sensor sees, with the raw returns kept so the result can be checked rather than taken on trust. Scored head to head against whatever the building already uses.

03
What comes out

A written result, including the parts that went badly. If the building beats the sensor, that goes in the report and this page changes accordingly.

04
Who it is for

A league, a federation or a venue operator who already knows that the speed number on their screen is an estimate, and who would rather it was not.

05
Where it could go

If it works: broadcast graphics on measured numbers, a hardest shot record that means something, and a data feed that integrity sensitive markets can rely on because it was measured. All of that is downstream of a pilot that has not happened yet.

Fusion, not replacement

Wherever a venue already has camera tracking, the sensor sits underneath it rather than against it. Video knows who is on the ice and where everybody stands. It does not measure how fast the puck left the stick. Those are two different jobs and the second one is ours.

05 · Technology

The numbers, each with its class

The same sensor and the same fusion layer as everything else we build. What changes between sports is the object, and therefore how far away it can be seen.

Detection range

132 m estimate

To a puck at 24 GHz under fixed power, scaled from our measured link budget by reflectivity. A rink is 60 m long, so the whole surface sits well inside it. This is the one number in this sport that is comfortable.

Puck reflectivity

−25 dBsm estimate

Modelled, not measured. Only three sports balls have a published measurement at this frequency and a puck is not one of them. Measuring it is on the pilot list.

Clutter gate

2 m/s

Anything moving slower than this along the sensor line disappears into the returns from the building. A puck at rest or rolling gently is invisible. Shots are not, which is why the shot is the product.

Range resolution

1.5 m design

Set by the chirp bandwidth. How finely the sensor separates two objects in depth, and the reason nobody honest quotes centimetres at this bandwidth.

The heritage

±0.05 % measured

Certified velocity measurement on objects from 30 to 10,000 m/s. The affiliate instrumentation heritage this team comes out of, not a specification of a hockey product that does not exist yet.

What is missing

Everything measured

No accuracy figure from a rink. No multipath characterisation. No puck reflectivity measurement. No product. This page describes physics we are confident in and an experiment we have not run.

06 · Where this sits

Behind golf, and honest about it

Now

Physics and signal chain

The measurement chain runs on development hardware for our first sport. The same chain would measure a puck without modification, which is the entire reason this page exists.

Next

Puck reflectivity, measured

A short measurement campaign to replace the estimate above with a number. Cheap, quick, and it either confirms the range figure or moves it.

When a rink says yes

The validation pilot

One building, one sensor, the three open questions answered in writing. This is the gate. Nothing downstream of it is promised.

Only if the pilot passes

A product conversation

Broadcast graphics, a measured record, a data feed. We are not going to describe these in any more detail than that until there is evidence behind them.

Talk to us

We need one rink and one evening.

If you run a building, a league or a federation and the speed number on your screen has always bothered you, that is the conversation. Bring your scepticism about the arena, because it is our scepticism too.

Talk about a pilot

Ole Kjærgaard Schrøter · Alpha Vision · oks@weibelventures.com · +45 20 97 56 20