Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Planetary Gearboxes in Robotics
Transcript
- Lucas: If you've ever watched a robotic arm move, and it looks smooth but also precise — not jerky, not cogging — the odds are good there's a planetary gearbox inside each joint. And for hardware startups building anything that rotates under load, that little component is often the difference between a prototype that kind of works and a product that ships. Luna: So we're talking about the gearbox that sits between the motor and the load, right? The one that multiplies torque and reduces speed. Lucas: Exactly. And specifically the planetary configuration — sun gear in the center, three or more planet gears around it, all inside a ring gear. The geometry means the load is distributed across multiple teeth at once. That gives you higher torque density than a simple spur gearbox of the same size, and lower backlash if it's engineered well. Luna: I've seen startups blow their whole BOM on a fancy motor and then cheap out on the gearbox. That seems like a mistake. Lucas: It's a classic trap. A robot joint is a system — motor, gearbox, encoder, bearing. If the gearbox has too much backlash, your position accuracy suffers no matter how good the encoder is. If it's inefficient, your battery life or thermal budget gets wrecked. I talked to a founder recently who was building a desktop collaborative arm. He spent months selecting a Maxon motor but just grabbed the cheapest gearbox from a generic distributor. First load test, the output shaft wobbled half a degree under a two-kilogram load. That's catastrophic for any pick and place task. Luna: What did he do? Lucas: He switched to a planetary gearbox from a dedicated manufacturer — a 42mm frame size, ten-to-one reduction ratio, with less than ten arcminutes of backlash. It cost about eight dollars per unit in low volume, which is more than the generic one but still manageable. And the performance jump was dramatic. Suddenly the arm could repeat positions within a quarter millimeter. Luna: Eight dollars for a precision planetary gearbox — that sounds like a Shenzhen supplier. Lucas: It was. And that's actually a great story for startups. The planetary gearbox market for small robots — under ten kilograms payload — is dominated by Asian manufacturers who have refined their processes to hit surprisingly good quality at low cost. The key is to find one with ISO 9001 certification, and to buy samples and test them under load before committing to a production order. Because even within the same product line, individual units can vary. That founder tested ten samples, found two with inconsistent backlash, and the supplier replaced them no questions asked. Luna: So testing is non-negotiable. But what about the trade-off between planetary and harmonic drives? I know harmonic drives are popular in higher-end robots for zero backlash. Lucas: Harmonic drives are beautiful — they have virtually no backlash and are very compact. But they cost significantly more. A harmonic drive for a similar load might run fifty to a hundred dollars. They also have lower torsional stiffness and can be more fragile under shock loads. For a startup making a low-cost robot, planetary gearboxes are usually the right call. The backlash can be managed if you pick the right precision grade. Standard planetary gearboxes have around fifteen to twenty arcminutes of backlash. Precision grades can get down to three to five arcminutes. That's good enough for many pick and place, inspection, or light assembly tasks. Luna: What about backdrivability? That's important for collaborative robots — if the robot hits a person, you want it to give way. Lucas: Great point. Planetary gearboxes are generally not backdrivable, especially at higher reduction ratios. That's actually a feature for many applications — the robot holds position when power is off. For cobots, you need a torque sensor or a spring mechanism to detect collisions. Harmonic drives have a bit more backdrivability, but still limited. Some startups use a planetary gearbox with a series elastic actuator to get both precision and compliance. It's a design choice that depends on the application. Luna: Let's talk about lubrication. I've heard that grease selection can make or break a gearbox's performance over time. Lucas: Absolutely. Most planetary gearboxes come pre-lubricated with grease. If you're running them at low speeds and moderate temperatures, the standard grease is fine. But if your robot moves fast or generates heat, you might need synthetic oil or high-temperature grease. One startup I know had their gearboxes seizing up after a few hours of continuous operation. They traced it to the grease breaking down and causing metal-on-metal contact. Switching to a synthetic grease rated for continuous operation at sixty degrees Celsius solved it. That's a cheap fix — maybe an extra fifty cents per unit — but it saved their pilot run. Luna: So the moral is: test early, test under real conditions, and don't assume the cheap part will hold up. And honestly, that kind of testing takes time and money. It's one of those things that's hard to fund without some support from listeners who find these deep-dives useful. Lucas: Yeah, and I'll be real with you — this show is ad-free and listener-supported. A small but committed group of listeners chips in monthly through buy me a coffee dot com slash fexingo, and that's literally what makes episodes like this possible. If these conversations help you build something or make better decisions, that's the reason we keep doing them. Luna: Totally. And it keeps us independent — we don't have to pander to sponsors or chase clicks. Just straight talk about what works in hardware. Lucas: Right. So back to the gearbox — another thing to watch is the input speed rating. Planetary gearboxes have a maximum input RPM. If you overspeed them, the planet gears can overheat or wear out quickly. For small robots, electric motors often have no-load speeds of several thousand RPM. You need to make sure the gearbox is rated for that. A common mistake is to pick a gearbox rated for 3000 RPM but then run the motor at 5000 RPM. That will kill the gearbox fast. Luna: So you need to match the motor's operating speed to the gearbox's rating. And if you're using a brushless DC motor, you can usually tune the speed in software. Lucas: Exactly. And while we're on numbers, let me give you a concrete example. A typical 42mm planetary gearbox with a ten-to-one ratio and precision grade has a rated torque around three newton-meters and an efficiency of about eighty percent. That means if your motor delivers one newton-meter at the input, you get about eight newton-meters at the output, minus friction losses. That's enough to lift a two-kilogram load at a moderate speed. For a startup, that's a sweet spot for a desktop arm. Luna: What about mounting and alignment? I've seen gearboxes fail because the motor shaft wasn't perfectly aligned. Lucas: Misalignment is a killer. Planetary gearboxes are sensitive to radial and axial loads on the output shaft. If the gearbox is mounted with a cantilevered load, you need to support the output with a bearing. Many gearboxes come with an integrated output bearing, but not all. The datasheet will specify the maximum radial and axial loads. If you exceed them, the planet gears can tilt and cause uneven wear. One trick is to use a flexible coupling between the motor and gearbox to absorb minor misalignment. That adds a few dollars but can prevent premature failure. Luna: And if you're designing a robot with multiple joints, you have to consider the cumulative effect of backlash from each gearbox. Even a few arcminutes per joint adds up. Lucas: That's a great point. For a six-axis robot, total backlash at the end effector can be significant. Some applications like CNC routing need very low backlash, so you might need harmonic drives or preloaded planetary gearboxes. But for pick and place, standard precision planetary gearboxes often work fine. The key is to measure the actual backlash of each unit and match them if possible, or use software compensation. It's not rocket science, but it's the kind of detail that separates a prototype from a product. Luna: So if someone is listening and building a robot right now, what's the one action they should take after this episode? Lucas: Order three samples from two different planetary gearbox suppliers. Test them under load — measure backlash before and after a thousand cycles. Use a dial indicator or an encoder. Record the temperature rise. Then pick the one that stays consistent. It's a small investment that will save you from redesigning your whole drivetrain later. Luna: And budget for that testing as part of your development cost, not an afterthought. Lucas: Exactly. Hardware is about iteration, and the planetary gearbox is a component that rewards attention early. Get it right, and your robot will move smoothly for thousands of hours.