In my first post about the counting machine, I ended with a promise: the prototype proved the concept; the next version will prove the design. This is that next version.
The short version: I removed all the electronics. No Arduino, no stepper motors, no buttons. Version 2 is purely mechanical, and it’s in daily production at Replay Golf Supply, run by one person.
I haven’t had a chance to take proper photos or video yet, so this one is text only. Pictures will come later.
Why Drop the Electronics?
Version 1 worked, but the parts that failed were almost always the moving, powered ones:
- The count gate ran on two steppers over a plastic rail with plastic gears. Friction and wear meant it eventually stopped moving on its own, and I was positioning it by hand anyway.
- The gates bent under load, and the count gate sometimes let a ball or two slip through.
- Snap-fit rails were easy to remove but flexed out of alignment.
Looking at how the machine was actually used, the motors weren’t buying much. An operator runs the same count for hours: set the gate once for 50-ball bags, then do batch after batch. A gate you set by hand and lock in place does that job with nothing to burn out or recalibrate.
So the goal for v2 was reliability over automation.
How v2 Works
The core idea hasn’t changed: golf balls are all the same size, so the count is built into the geometry. Ten rails feed down toward the front of the table, and the distance between the count gate and the front gate decides how many balls are in a batch.
Setting the Count
The count gate is a bar that sits across the rails. Each side of the bar has two wires that hook over pins on the table, one set of pins per count. To set up, the operator moves the gate to the count they’ll be working with and drops the wires over that count’s pins. That’s the whole fixation mechanism, and it holds.
Tensioning the Gate
The gate is raised and lowered by a wire running over pulleys to a spool. The spool is what makes multiple counts possible: since the gate can sit at different distances along the table, the wire length changes, and the spool takes up the slack.
Once the gate is fixed in place, the operator uses the spool’s ratchet to tension the wire just enough that balls can roll underneath. The ratchet holds that setting, so it’s done once per count, not once per batch.
Running a Batch
- Fill the table with balls and attach the bags (or boxes) at the end of the table.
- Pull the lever. It closes the count gate, so only the balls between the two gates are free.
- Open the front gate by hand. The counted balls roll into the bags or box.
- Close the front gate and release the lever. The count gate lifts and the table refills for the next batch.
Compared to v1’s two-button sequence, it’s one more hand movement, but every part of it is something you can see and feel working.
What Else Changed
Rails
I kept the original rails, but they’re now fixed in place with screws instead of snap-fit mounts. That fixed the flexing and alignment problems from v1.
The Bag Clamp
Getting balls into bags fast turned out to matter as much as counting them. Each bag hangs under a tube at the end of the table and is held by a spring-loaded, 3D-printed clamp: lift the handle, slide the bag over the tube, let go.
The design took a while. A clamp that’s easy to open tends to be too weak to hold a bag full of golf balls, and one that holds well is a pain to open hundreds of times a day. It took several iterations to land on a spring and lever geometry that does both.
Results
All numbers below are my estimates from watching it in use, for counting 25,000 balls:
| Batch | Time | Notes |
|---|---|---|
| 300-ball boxes | ~2 hours | Depends on whether boxes are pre-made, need taping, or need a plastic liner |
| 50-ball bags | ~3 hours | Slower because of constant bag changing and tying |
| 72-ball bags | a bit under 3 hours | Fewer bag changes than 50s |
That’s one operator doing what used to take a team. For comparison, v1 took about 2.5 hours for 25,000 balls into 72-ball bags, while a team of six doing it by hand needed 1.5–2 hours. Per person, v2 is roughly 3–5× more output, and it almost never jams.
Most of the remaining time isn’t counting at all: it’s bag handling, tying and box prep. The machine is no longer the bottleneck.
What I Learned
The biggest lesson was that automation should earn its place. My first idea was a fully automated grader and counter. v1 trimmed that down to automated counting. v2 trimmed it to good geometry and simple mechanics, and that’s the version that survived production.
The other lesson was to watch where the time actually goes. Once counting was fast, the slow part became the bags, and the clamp ended up being one of the most worked-on parts of the machine.
Designed and built by Pavlo Solodrai, 2026.