Why your prototype passed the bench and failed the dyno

There's a moment every hardware team enjoys: the first prototype motor arrives, you wire it up, give it power, and it spins. Smooth, quiet, fast. Somebody films it on a phone. It feels like the hard part is over.

It isn't. A brushless motor spinning with nothing attached to the shaft is doing the one thing it will almost never do in your product. You've confirmed it isn't dead. You've confirmed almost nothing else.

No-load tells you the easy 10%

Spinning free, a motor draws a small current to overcome its own friction and windage. The number you read is dominated by bearing drag and iron losses, not by the work you actually care about. Two motors that look identical at no load can behave completely differently the moment you ask them to push against something.

What the free spin does tell you is narrow but real: the windings are connected correctly, the rotor is balanced enough not to shake itself apart, the bearings aren't notchy, and the commutation is roughly sane. Useful as a smoke test. Useless as validation.

The whole job of a motor is to convert electrical power into mechanical work under load. If you never load it, you never tested the thing it's for.

What the dyno actually measures

A dynamometer puts a controllable, measurable load on the shaft so you can walk the motor across its real operating range and record what happens. The output isn't a single number — it's a set of curves, and the shape of those curves is where the truth lives.

• Torque vs speed. The defining curve. It tells you whether the motor can actually hold your load at the speed you need, with margin, or whether it's living right at the edge where a hot day or a low battery tips it over.
• Efficiency across the range. Peak efficiency on a spec sheet is a single dot. Your product lives in a region, and a motor that's 90% efficient at one point can be 70% where you actually run it — that missing 20% comes out as heat.
• Temperature under sustained load. This is the one that ends programs. A motor can hit its torque target for thirty seconds and then thermally run away over ten minutes of duty. The bench never shows you this. The dyno, with a thermocouple and patience, shows you exactly when and where.
• Back-EMF and Kv under real current. The constants you were handed shift under load and temperature. Measuring them loaded is how you find out whether your controller assumptions still hold when the magnets are warm.

The failures that only appear under load

Here are the ones we see most often — every one of them invisible on a free-spinning bench:

Thermal runaway in the windings

As copper heats, its resistance climbs, which means more loss, which means more heat. Under a demanding duty cycle a marginal design finds a new, much higher equilibrium temperature — or never finds one at all. You discover this at minute eight of a loaded run, not in the first ten seconds.

Demagnetisation at the corner

Push high current through a hot motor and you can partially demagnetise the rotor magnets. The motor survives, keeps spinning, but its torque constant has quietly dropped. Everything still "works." It's just permanently weaker, and nothing on the bench would have told you.

Cogging and ripple you can feel but not see

Torque ripple that's irrelevant on a free shaft becomes vibration, noise, and control instability when coupled to a real load — especially in robotics and anything doing precise positioning. You need the load in place to characterise it.

The controller-motor mismatch

A motor and its drive are one system. Timing, current limits and commutation that look fine unloaded can collapse into stalls or audible whine the moment torque demand rises. Testing the motor without testing it through its intended controller, under load, is testing half the product.

Why this is the cheapest insurance you'll buy

A dyno session before you commit to tooling costs a small fraction of a single production run. Finding a thermal problem after you've ordered 5,000 stators costs you the tooling, the parts, the schedule, and the customer's patience. The asymmetry is enormous, and it always points the same way: load it early.

This is exactly why we won't call a prototype "validated" on a spin test. A prototype is validated when there's a curve next to the spec and the curve clears it with margin — measured, not assumed. Everything before that is a motor that happens to turn.

If you've got a prototype that spins and you're not yet sure what it does under load, that gap is precisely the work worth doing before anything goes near production.