Ball Screw or Belt Drive? Real-World Selection for CNC Automation

Ball Screw or Belt Drive? Real-World Selection for CNC Automation

I’ve spent more than a decade bolting linear stages onto machines—some for micron-level cutting, some for slinging parts across a factory at 3 m/s. New engineers keep asking me the same question: “Which one is better?” There is no “better.” There is only “better for what you are actually doing.” Let me skip the textbook definitions and talk about what happens on the shop floor.

The fundamental difference that matters

A ball screw stage uses metal-on-metal rolling contact. That gives you stiffness and a direct relationship between motor rotation and table position. Load pushes back, the screw pushes harder, and deflection stays small. This is why a C3 ground screw can hold ±0.008 mm per 300 mm lead error, and with preloaded nut you get down to single-digit micron repeatability in a well-designed setup. The price you pay: rotating mass, critical speed limits, and cost climbing fast beyond 3–4 meters stroke.

A belt drive stage replaces the screw with a reinforced rubber belt and pulleys. The belt itself is a spring. Under acceleration or cutting load, it stretches. With a rotary encoder only, a belt stage might wander 0.1–0.2 mm when load changes. But you can run it at 3–5 m/s and accelerate at 2g without worrying about screw whip. Strokes of 6, 10, even 20 meters are practical. It is cheap per meter, and you replace a belt, not a precision-ground shaft.

Accuracy—when belts can work, and when they can’t

If your CNC application puts a tool against a workpiece, you want ball screw stiffness. I have seen shops try to run a light milling spindle on a belt stage, and surface finish went to hell because the whole carriage was bouncing under intermittent cutting forces. Adding a linear encoder and full closed-loop control helps with static positioning, but it does not fix dynamic stiffness. So my rule is: if you need better than ±0.02 mm under load, you stay with a properly sized ball screw, period.

Belts can do good repeatability—I’ve measured ±0.05 mm repeat on a well-tensioned belt stage with a linear encoder, moving a constant light payload. That is plenty for pick-and-place, packaging, or gluing. But you have to check belt tension every few months, and if the factory temperature swings 15 °C, the belt length changes enough to shift your home position. Ball screws also grow with temperature, but the effect is more predictable and you can map it or use a linear scale to compensate.

Speed and stroke—where belts win by a mile

A ball screw has a critical speed limit defined by its diameter, length, and end support. As a quick check, many shops use a DN value around 70,000. For a 25 mm diameter screw, that gives you about 2800 rpm max. With 20 mm lead, top speed is 56 m/min—under 1 m/s. You want 2 m/s and 3 meters stroke? You need a much thicker screw, and the cost skyrockets. Beyond 4 meters, you also deal with screw sag, requiring rotating nuts or extra supports that complicate the design.

Belt stages don’t care about length the same way. I’ve put together 8-meter belt gantries that run at 4 m/s, and the limiting factor was the linear rail and cable management, not the drive. If your cycle time is the money-maker and you are moving a 5 kg gripper instead of a 200 kg spindle, go belt.

Load and thrust—screw’s home turf

A 25 mm diameter ball screw with 5 mm lead can push over 1000 N continuously without breaking a sweat. That’s enough to drive the Z-axis on a mid-size VMC, pushing through material. Belt drives can handle loads too, but the limit is often the belt’s tensile strength and tooth shear. For a typical AT10 belt, you might get 500–800 N working load, but deflection under that load is significant. When you combine heavy payload with high acceleration, the belt stretches dynamically, and your positioning becomes mushy. If your machine does any real work—drilling, milling, pressing—the stiffness of a ball screw pays off immediately.

What I tell engineers before they spec

Don’t just look at “max speed” or “load rating” on a datasheet. Sketch out a duty cycle: mass, acceleration, external force, duty. A 50 kg load moving slowly is a completely different problem than 30 kg accelerating at 1.5g, 20 times a minute. With belts, calculate the dynamic stretch under worst-case deceleration. With screws, check the critical speed and column buckling limits at the stroke you need. If you ignore that, you will end up with either a system that shakes itself apart or one that’s oversized and wasting budget.

One more thing: if you need both high thrust and long stroke, you are in the “expensive compromise” zone. I’ve seen people try to couple two ball screws or go with rack and pinion, but sometimes a properly engineered belt stage with a heavy-duty linear rail and external scale is the more practical, serviceable solution—provided you can live with belt replacement every few years.

Quick decision cheat sheet

• Need ±0.02 mm or better under cutting load, rigid, sub-3-meter stroke: ball screw, no question.

• Stroke above 4–5 meters, speed 2 m/s or more, light-to-medium payload, duty cycle not hammering: belt drive with linear encoder.

• High-speed but also high-force (e.g., a pick-and-place lifting heavy parts fast): calculate carefully, and consider a screw if you can accept shorter stroke or a slower speed.

There’s no magic product that does everything. The right stage is the one that holds tolerance on your worst day of production, not on a lab test sheet. Choose based on your actual process forces and motion profile, and you’ll avoid the expensive rework I’ve had to fix more times than I’d like.