Inside a QDD Motor: Taking Apart the RobStride 04 Joint Actuator
イチケン / ICHIKEN
Quasi-direct-drive (QDD) motors have become a standard component in the joints of four-legged and humanoid robots. In this video, the presenter calls them one of the core technologies of physical AI. They bought a RobStride 04, the highest-torque model in RobStride's lineup, ran it, and then took it apart. The goal was to explain why this kind of motor suits robot joints and what is inside one.
What a QDD Motor Is
A QDD motor pairs a motor with a gear of low reduction ratio. "QDD" stands for quasi-direct drive, which in Japanese is called a "semi-direct-drive motor." The presenter says the design spread quickly around the time four-legged robots such as Unitree's appeared, and that it is now widely used in humanoid robot joints.
The unit in the video is the RobStride 04. Its rated peak torque is 120 Nm and its reduction ratio is 9:1. The presenter bought it on AliExpress for 39,000 yen and said that getting a motor like this for about 40,000 yen is a very good deal.
Why Robot Joints Are a Hard Problem
The presenter starts from the Chinese humanoid robots many viewers have seen in videos. These robots jump and land, shrug off a push from a person, and in some movements move more nimbly than humans. The joint motors are a key part of what makes this possible.
A robot moves its joints much as a person does, and moving them while carrying its own body weight takes a lot of torque. A motor alone produces surprisingly little torque, the presenter explains. More precisely, a motor small enough to fit inside a joint has limited torque. There are two ways to get more: make the motor bigger, or use gears to trade speed for torque.
High-Gear-Ratio Motors and Their Weaknesses
Industrial robots have long used high-gear-ratio motors. These combine a high-speed motor with a large reduction gear, for example around 100:1. Their main strength is that they produce a lot of torque for their size.
In robots that face external forces, the presenter says, they have three weaknesses:
- No backdrivability. Because the reduction ratio is so high, pushing on the joint from outside barely moves it. A sudden impact, such as a landing, has nowhere to go, and the gear teeth can crack.
- Poor torque estimation. Friction in the gearbox and other parts makes it hard to estimate output torque from motor current. A separate torque sensor is sometimes needed.
- Unsuited to disturbances. Industrial arms mostly do positioning, so these weaknesses don't matter much there. Quadrupeds and humanoids are hit by disturbances all the time, so they need caution.
Direct-Drive Motors: Ideal for Control, Too Heavy
At the other extreme is direct drive, where the motor's rotation goes straight to the output with no gears. To show how this feels, the presenter used a motor taken from a robotic arm. They noted it isn't strictly a direct-drive motor, but it shows the behavior clearly. With no gears there is no friction or play, and it turns smoothly under a finger, like a joint. Turning freely like this means the motor is backdrivable. The motor current also tells you the torque almost exactly. The presenter calls it the ideal motor from a control standpoint.
The catch is that without gears, the motor must produce all the torque itself. That makes it bigger and heavier, which is a problem for a humanoid that has to hold up its own weight. The presenter added that attaching a gearbox to the shaft of this same arm motor would turn it into a high-gear motor.
Three Requirements, and Why QDD Fits
The presenter lists three things a quadruped or humanoid joint needs at the same time:
- High torque relative to weight. Low weight matters most, and torque is essential for supporting and moving the body.
- Impact resistance. This comes down to backdrivability: whether the joint can absorb the shock of landing after a jump.
- Torque awareness. The robot estimates the force on the ground from joint torque and adjusts that force.
High-gear motors fall short on impact resistance because they aren't backdrivable. Direct-drive motors struggle with torque per unit mass. QDD is the common way to meet all three.
The presenter describes typical QDD traits, while noting that they vary by product. The motor is flat and wide to raise torque. The gear ratio is kept low, around 10:1. The shaft can usually just barely be turned by hand, so it can absorb impacts, and torque can still largely be estimated from current. The approach became widely known through MIT's Cheetah robot research. It is used mainly in quadrupeds but is spreading to humanoids.
QDD has limits too, the presenter says. Its torque density is lower than a high-gear motor's, and holding a demanding posture for a long time makes the motor quite hot. Depending on the robot or joint, designers may raise the reduction ratio or switch to other approaches.
Specifications and First Run
The user manual lists 120 Nm peak torque, but the motor can't sustain it. The rated continuous torque is 40 Nm, and the manual's graph shows 120 Nm is available for only about 10 seconds. The presenter guesses that running at 120 Nm for long might burn out the motor driver.
The input voltage is 48 V, and control is over CAN. The whole motor controller is inside the unit, so a 48 V supply and a CAN bus are enough to run it. The presenter drove it with an ESP32 microcontroller and a CAN transceiver.
In the demo, the presenter found the rotation surprisingly smooth. Setting acceleration to around 300 made it spin quite fast. When stopped, it held its position firmly and the shaft wouldn't turn when pushed by hand, which they attributed to the 120 Nm rating. They called it very useful as a ready-to-run module, since you can start building a robot without designing the motor side.
Heat was already noticeable. After about 10 minutes of running with no load, the motor reached around 40°C. The presenter expects more heat once it is in a robot and expects cooling to be very important.
Opening It Up: Thermal Path and Outer-Rotor Layout
The first part removed was the outer plate. It carries a thermally conductive silicone mat that connects the plate to the windings inside. The presenter links this to the 40°C no-load result: heat seems to be a real challenge, so the design gives it a path out.
Turning the output shaft also turns the internal rotor, which spins faster than the output because of the reduction. The stator with the windings is fixed, and the rotor carries permanent magnets, so this is an outer-rotor, rotating-field motor. The presenter notes this type is common in drone motors. An outer rotor has more rotor inertia, but it gives more torque than an inner-rotor design, which the presenter thinks is why it suits a QDD motor. Inside there is also a bearing and a planetary gear. The presenter split the parts into electrical and mechanical groups.
Stator and Rotor
The motor is three-phase. The stator has a very high pole count and a laminated iron core. The presenter explains the trade-off:
- Benefits: a high pole count keeps the core from growing much, and it shortens the coil ends, which don't contribute to torque.
- Costs: each turn of the output shaft takes more electrical cycles. That raises iron losses in the core and requires a shorter control cycle.
Neodymium magnets are glued to the inside of the rotor, with plastic spacers to align them. Centrifugal force presses the magnets outward against the rotor, so the presenter doesn't think they risk flying off. They saw no sign that the rotor had been balanced. Their possible explanations are that it is precise enough not to need balancing, or that its weight and speed make balance a non-issue. They also suggested four holes in one part might be doing some balancing.
The Drive Board
The drive board holds the microcontroller, sensors, communication ICs, and power semiconductors. It originally had a moisture-proof coating. Because it sits inside the motor, it was covered in a thick layer of gear grease. Wiping it with ethanol seemed to remove the coating as well, which surprised the presenter.
The microcontroller is a GigaDevice GD32F303, a 32-bit Arm Cortex-M4 part with DSP instructions and an FPU. The presenter considers it well suited to motor control.
Under another silicone thermal mat is the three-phase inverter. It uses silicon MOSFETs from Oriental Semi, in pairs running in parallel. The presenter called them quite high-performance:
| Parameter | Value | |---|---| | VDS | 100 V | | On-resistance | 4.3 mΩ | | Pulsed drain current | 560 A | | Continuous drain current | 140 A |
The presenter suspects low on-resistance parts were chosen because heat dissipation is clearly a challenge. Heat flows from the MOSFETs into the board, through the silicone mat, and out to the aluminum housing, which the board presses tightly against. The coils also produce heat, so the presenter thinks long periods of heavy work will be hard without good cooling. Current sensing uses three sets of shunt resistors, one per phase. The presenter notes that precise motor control depends heavily on accurate current measurement.
A Mysterious Metal Block on the Board
A solid block of metal is mounted on the board. The presenter says such parts are usually used to lower the effective resistance of the copper traces. Here, they suspect it also stores heat by adding heat capacity.
Their reasoning is that a joint motor often takes sudden loads, which briefly heat the area around the inverter. The silicone pad to the housing gives low thermal resistance at steady state, but transient thermal resistance is unavoidably high. The metal block may cover the time lag before heat reaches the housing. The presenter offered this as their interpretation of the design, not as confirmed.
The board also has space for many MLCCs that were never fitted. An electrolytic capacitor is on the other side, but its ESR and ESL are higher than MLCCs, so the presenter says behavior under sudden loads would need checking. They then noted that the control system's minimum operating voltage is 24 V against the 48 V rating. On a 48 V supply, they don't expect a sudden load to crash the control system.
Gate Driver, CAN, and Magnetic Sensors
The other side of the board mostly holds communication and sensor ICs:
- Gate driver: Texas Instruments DRV8353, which drives the MOSFETs.
- CAN transceiver: NXP TJA1042, paired with the CAN controller built into the GigaDevice MCU.
- Magnetic sensors: two ICs, one from Infineon and one from AMS. The presenter says AMS's non-optical analog mixed-signal sensor division was acquired by Infineon in 2026, so in effect both are Infineon sensors.
The middle sensor is an AMS magnetic rotary position sensor. It sits directly over the rotor shaft, where a magnet is attached. Holding a magnetic viewer up to it shows the poles clearly. This lets the sensor read which way the rotor is facing.
The second sensor is off the rotor axis. When the rotor is in the housing, a plastic gear on the rotor drives another gear, and this sensor reads that gear's angle. That gear's shaft is also a magnet. The two gears have different tooth counts, so their magnetic patterns drift apart as the motor turns. From that offset, the system can work out where the rotor is. The presenter identifies this as the vernier principle.
Knowing the starting angle of the output shaft and how many turns it has made lets the system know each joint's angle as soon as the program starts. The presenter thinks this is detected quite well.
As for why the design uses magnetic sensors instead of an optical absolute encoder, the presenter guesses three reasons. Optical sensors are expensive. They may struggle with grease spattering inside. There may also simply be no room for one in a QDD motor. A magnetic setup can detect rotation angle for a few hundred yen, so the presenter suspects cost and environment drove the choice.
Overall, the presenter observed that the semiconductors are mostly from Chinese makers, while some communication and sensor parts come from Western manufacturers.
The Mechanical Side
The reduction gear is a planetary gearset:
- Sun gear: built into the rotor.
- Planet gears: three.
- Carrier: serves as the output shaft.
- Ring gear: built into the stator.
The bearings are from C&U and S&F, mostly C&U, including two large-diameter bearings in the output shaft. The presenter notes that these mechanical parts are important, and that they were basically all sourced within China.
Conclusion
The teardown showed that the RobStride 04 combines an outer-rotor motor with a planetary gearset. That layout delivers high torque from a small package while staying backdrivable. The presenter was surprised a motor like this costs only a few tens of thousands of yen, and liked most that it can be plugged in and used right away. Their remaining concern is heat: the motor reached about 40°C with no load, and the design relies on silicone pads and possibly heat-storing metal to cope.
This video is brought to you by the Nikkan Kogyo Shimbun.
Hello. Since I got a QDD motor, I'd like to take it apart and see what's inside.
A QDD motor consists of a motor combined with a low-ratio gear. QDD stands for quasi-direct drive. In Japanese, it is called a "semi-direct-drive motor." Around the time four-legged robots like Unitree started appearing, it began spreading rapidly. Recently, it has been widely used in the joints of humanoid robots. It could be described as one of the core technologies of physical AI.
I purchased a QDD motor, model 04, from RobStride. I decided to buy the model with the highest torque in the lineup. The torque is 120 Nm, and the reduction ratio is 9:1. This time, I bought it on AliExpress for 39,000 yen. I think being able to get this motor for about 40,000 yen is a really great deal. I'll post a link in the description, so if you're interested, please check it out.
Before we actually start taking it apart, let me introduce today's sponsor. The annual 2026 edition of the Kosen Special Feature from the Nikkan Kogyo Shimbun is now available. Kosen—that is, the National Institutes of Technology—are educational institutions dedicated to nurturing the talent that will support the industries of the future. A special feature highlighting their appeal is now available.
Here's what you might be curious about: Look at all the fields where graduates of technical colleges are making their mark! What kinds of activities do technical colleges actually offer? We feature plenty of things like these. In this installment, we'll look at the life story of a senior who graduated from a National College of Technology and his journey to finding a new job. It also includes a list of national universities that accept transfers from technical colleges. It's full of tips to help you think about your future path.
As you can see, there have also been special features in the newspaper. It has been published in booklet form and is now being distributed to technical colleges across the country. Even if you don't subscribe to the Nikkan Kogyo Shimbun, the web version is now available for free. We are also currently accepting entries for our reader giveaway, with prizes such as digital multimeters and digital calipers. Please check it out right away by clicking the link in the description.
First of all, what exactly is a QDD motor? I think a lot of people have seen videos of Chinese humanoid robots. When a robot jumps and lands, or when a person pushes it, it absorbs that shock. When it comes to certain movements, they can even move more nimbly than humans. One of the components that enables this movement is the motor built into the joint.
When a humanoid robot is moving, it moves its joints just like humans. Moving your joints while supporting your own body weight requires a great deal of torque. However, the torque of the motor alone is surprisingly low. Rather than saying it is low, if you try to build a motor that fits inside a joint, the torque will decrease. To increase torque, you either make the motor itself larger or use gears to reduce speed and increase torque.
A long-standing staple in industrial robots and other applications is a high-gear-ratio motor that significantly reduces speed via a gear system to amplify torque. On the other hand, some people avoid using gears because they dislike their weaknesses. Direct-drive motors have also been the subject of research for many years. That said, for four-legged robots that move around vigorously and humanoid robots, both types have their pros and cons. That's why QDD motors have recently come into widespread use. First, let's take a look at the features of high-gear-ratio motors and direct-drive motors.
First, a high-gear-ratio motor consists of a high-speed motor with a large reduction gear, with a ratio of, say, 100:1. One of its key features is that it can generate quite a lot of torque for its size. However, it does have its weaknesses. When used in robots, even when I press on the joint from the outside to try to move it, it barely moves. The reason is that the reduction ratio is too high, resulting in poor reverse-drive capability, or backdrivability.
For example, when a sudden impact occurs, such as when landing, with nowhere for that impact to be absorbed, the gear teeth can end up cracking. Furthermore, because of the high friction in components such as the gearbox, it is difficult to estimate how much torque is currently being applied based on the current flowing through the motor. In some cases, a separate sensor may be required to measure torque.
Since industrial robot arms are primarily used for positioning, there's no problem with using a high-gear-ratio motor. But caution is required when using this with four-legged robots that are constantly subjected to external disturbances, or with humanoid robots.
Next is the direct-drive motor. This allows the motor to rotate without using gears, directly transmitting the rotation to the joint. This is a motor that was removed from a robotic arm. It's not a direct-drive motor, but I'm going to show you this because the movements are easy to understand.
Since there are no gears, there is no friction or play, and if you turn it with your finger, it moves smoothly and easily. This part actually moves like a joint. When a joint rotates smoothly, it means it has backdrivability. If you look at the current flowing through the motor, you can determine the torque being produced with near-perfect accuracy. From a control perspective, it's the ideal motor.
However, since torque cannot be amplified by the gears, the motor alone must generate all the required torque. If designed that way, the motor itself would become larger and heavier. If used in a humanoid robot designed to support its own weight, that is a little difficult. By the way, I said this motor was something I had removed from the robotic arm. You can attach a gearbox to the end of this shaft. That would make it a high-gear-ratio motor.
Let's take a look at what's needed for the joints of four-legged robots moving around here and humanoid robots. First of all, they need to generate a lot of torque relative to their weight. Being lightweight is the most important thing, and torque is also extremely important for supporting the body and moving.
Next is impact resistance. This leads to the question of whether it has backdrivability. For example, when you jump and land, it all comes down to whether the joints can absorb the impact. Then there's understanding torque. This is necessary to estimate the force applied to the ground based on the torque at the joint and adjust that force. All three of these elements are necessary at the same time.
Since high-gear-ratio motors do not have backdrivability, from the perspective of impact resistance, they're a bit of a mixed bag. With direct-drive motors, it becomes more difficult to meet the torque-per-unit-mass requirements. A common approach for meeting all three of these requirements at the same time is a QDD motor.
It depends on the type of QDD motor, but the motor is designed with a flat shape to increase torque. The gear ratio is kept low, at about 10:1. For the most part, the shaft has enough backdrivability that it can just barely be turned by hand. In other words, it can absorb the impact when it's struck. The characteristic of being able to determine torque from the current flowing through the circuit is also largely preserved.
QDD motors became widely known through MIT's research on a robot called Cheetah. Nowadays, they are used primarily in four-legged robots, and they are also becoming widely used in humanoid robots. However, QDD isn't a one-size-fits-all solution either. It doesn't have the same torque density as a high-gear-ratio motor, and if you hold a strenuous posture for a long time, the motor itself generates quite a bit of heat. Therefore, depending on the robot or joints, it may be necessary to increase the reduction ratio. We also switch between this method and others as needed.
I'll try using it before taking it apart. The specifications are also listed in the user manual. The peak torque is 120 Nm, but it can't run at that all the time. The rated torque is 40 Nm. Here is the graph. In fact, it can operate at 120 Nm for only about 10 seconds. The continuous-duty rating is 40 Nm. If you operate it for a long time at a torque of 120 Nm, I guess the motor driver might burn out or something.
The input voltage is 48 V. Control is handled via CAN communication. All of the controllers for motor control are contained within this unit. So, if we can incorporate a 48 V power supply and a CAN bus for control, it's designed to be ready to use right away. In this project, I'm using an ESP32 microcontroller and a CAN transceiver.
Now let's actually put it into action. When I pressed the "Start" button, it began to turn. I think this is actually pretty good, don't you? It's currently in the upright position, but if it flips over, that's nice. There are definitely comebacks, too. The rotation is actually surprisingly smooth.
For example, should we set the acceleration to around 300? When you do that, it spins pretty fast. I think it's pretty good. By the way, it also holds the position securely when stationary. Even if I try to move it like this, the axis won't turn. As you'd expect from 120 Nm, the torque is quite powerful.
As you can see, it's ready to use pretty quickly. Since it's a module that can be put into action right away, it's very useful when building robots. I think it's great that you can start building robots without giving the motor part much thought.
I was just playing around, running it for about 10 minutes without a load, but it's getting pretty lukewarm. The temperature is around 40 degrees. Since this is how it performs under no load, once it's actually installed in the robot, I think it will generate even more heat. Cooling seems like it will be very important.
From here on, we'll take it apart further. First, remove this plate. I got it. This plate comes with a heat-conductive silicone mat. This is used to thermally connect the outer plate to the inner winding. I did a little experiment just a moment ago, and even when running under no load, the temperature rose to about 40°C. I guess it must be pretty tough in terms of heat, after all. That's why it's designed this way so that heat can escape to the outside.
When you turn the output shaft, it looks like this: the internal rotor is also spinning. Because of the reduction gearing, the rotor is spinning faster than the output shaft. The armature, which consists of windings, is fixed. The rotor is equipped with permanent magnets. It appears to be an outer-rotor type motor. This is a rotating-field motor. This outer-rotor type is commonly used in drone motors and the like.
Here's what it looks like when you spin it while it's disassembled. It's a bit choppy because it's currently being disassembled, but it's spinning smoothly. With an outer-rotor type, although the rotor's inertia increases, it has higher torque compared to the inner-rotor type. In a QDD motor like this one, I suppose the outer-rotor type has an advantage.
If we take it apart further, it looks like this. There's a bearing here—so this is a planetary gear, right? The rotor and stator components are installed like this, but I'm going to remove them. It is mainly divided into electrical and mechanical components. First, I'll start by introducing the electrical components.
First, a three-phase motor is used. The stator is a multipole motor with a very high number of poles. The iron core has a laminated structure. If the number of poles is high, the iron core itself won't get that much bigger, and the coil ends, which do not contribute to torque generation, can also be made shorter. However, for every one revolution of the output shaft, this increases the number of electrical cycles needed. If that happens, iron losses in the core will increase, and we also need to shorten the control cycle.
Next is the rotor section. Neodymium magnets are attached to the inside of the rotor. Adhesive was used to secure them in place. Since they're designed to be pressed against the rotor by centrifugal force, it doesn't look like the magnets are going to fly off somewhere. Plastic spacers were used to align the magnets.
There are no signs that the rotor has been balanced. It might be made with such precision that there's no need to balance it in the first place. Or, given this weight and rotational speed, you might not even need to worry about balance. It might be something like that. Since there are four holes in this part, it's also possible that it's balanced by adjusting their depth.
Next, we'll take a look at the circuit board for the motor drive. It contains the microcontroller, various sensors, communication ICs, power semiconductors, and other components. This is the side with the microcontroller. Here's what the back looks like.
This circuit board originally had a moisture-proof coating applied to it. Since this circuit board is inside the motor, there was quite a thick layer of grease—the kind used to lubricate gears—stuck to it. I wiped it down with ethanol to remove it, but it looks like the moisture-proof coating came off as well. I heard that ethanol can strip off the moisture-proof coating, but I don't really have many experiences like that.
Let's take a closer look at each component. First, take a look at this. The microcontroller used was the GigaDevice GD32F303. This is a 32-bit microcontroller that uses the Arm Cortex-M4. It contains DSP instructions and has an FPU, so I think it's quite easy to use for motor control.
Next, we'll take a look at the power circuit section. I have a heat-conductive silicone mat here. I'll remove this one, too. This section shows the circuit for a three-phase inverter. It uses silicon MOSFETs from Oriental Semi. They're set up in pairs, running in parallel. This MOSFET is actually quite high-performance, with a VDS of 100 V, an on-resistance of 4.3 mΩ, and a pulse drain current of 560 A. The continuous drain current rating is 140 A, so it handles quite a high current. Since heat dissipation from this motor itself seems to be quite a challenge, it is possible that they are selecting MOSFETs with the lowest possible on-resistance.
The heat dissipation path for this power semiconductor involves heat transfer to the circuit board and, through this heat-conducting silicone mat, heat dissipation from the aluminum housing. As shown here, it fits snugly against the outer aluminum casing. Heat generation isn't limited to the inverter alone. There is also heat generated by the motor's coils. Therefore, during prolonged periods of heavy-duty operation, it's essential to ensure proper heat dissipation; otherwise, it looks like it's going to be tough.
For current detection in the inverter, a shunt resistor is used for each phase. There are three sets. If you try to control the motor precisely, the accuracy of current measurements is also quite important.
Something that looks like a chunk of metal is mounted on this circuit board. One of the goals is to lower the effective resistance of the copper foil on a printed circuit board. In this case, I'm wondering if it's being used to store heat by increasing the heat capacity. In particular, since this motor is installed in the robot's joints, there are many situations where a sudden load is applied. If used in this way, the area around the inverter will heat up momentarily.
Since heat is dissipated to the enclosure using a silicone heat-dissipating pad, although the steady-state thermal resistance is low, the transient thermal resistance will inevitably be high. By mounting such metal pieces, it compensates for the time lag in heat dissipation. I think that's how it's designed.
This section was designed to accommodate a large number of MLCCs, but they had not actually been installed. Although an electrolytic capacitor is mounted on the opposite side, compared to MLCCs, ESR and ESL are higher, so the behavior under sudden loads requires verification. However, the rated voltage of this motor itself is 48 V, but the minimum operating voltage, including the control system, is 24 V. As long as you're using 48 V, even if there's a sudden load, I don't think the control system will crash or anything like that.
Let's take a look at the other side, too. This section mainly features communication and sensor ICs. First, this is a gate driver for driving a MOSFET. It uses the DRV8353 from Texas Instruments.
Next up is this one. This is a CAN transceiver. The NXP TJA1042 is used. GigaDevice microcontrollers include a CAN controller, so it is designed to be used in combination with that.
There are two left. These ICs are magnetic sensors. Products from Infineon and ams are used. However, regarding ams, the non-optical Analog Mixed-Signal Sensor Division was acquired by Infineon in 2026. So, in effect, both are Infineon magnetic sensors.
The one in the middle is ams's magnetic rotary position sensor. The rotor is inserted like this, and the design is such that these sensors are positioned on top of that. The magnetic sensor is positioned exactly on the rotor shaft. This part of the rotor has a magnet attached to it. When you actually hold a magnet up to it, it looks something like this: the magnetic poles are clearly marked. In other words, using a magnetic sensor, it's designed so you can tell which direction the rotor is facing.
What is the magnetic sensor located off-center from the rotor's axis of rotation looking at? When the rotor is assembled with the housing, there is another gear attached to the tip of the plastic gear on the rotor, and it's looking at this angle. In terms of their relative positions, there are two magnetic sensors, but this will result in a shape that overlaps like this. In this part as well, the shaft is made of a magnet, and when you bring a magnet close to it, it spins smoothly. The magnetic sensor reads the direction of the magnet.
If you look closely, you'll see that the two gears have different numbers of teeth, so depending on the rotational speed, the magnetic pattern is shifting. Based on that misalignment, you'll be able to tell where the rotor is located right now. It's what's known as the Vernier principle. If the initial angle of the output shaft is known, also, since you can tell how many rotations it takes to break a joint, things like what angles the joints are at when the program is first launched, and so on—I suppose it's detecting that part quite well.
By the way, optical absolute rotary angle sensors are expensive, and in a dirty environment where grease and other substances tend to splatter inside, I wonder if they're hard to use. Or maybe there's just no room to fit one in a QDD motor, or something like that. I wonder if that might be related as well. With a magnetic type like this one, since you can set up rotation angle detection for just a few hundred yen, I suspect they've opted for a magnetic system due to cost and environmental considerations.
As such, when it comes to semiconductors, devices from Chinese manufacturers are generally in use, and regarding some communications and sensors, I found out that products from Western manufacturers are being used.
Next, we'll take a look at the mechanical aspects, mainly gearboxes. A planetary gear mechanism is used in the reduction gear. The sun gear is integrated with the rotor, and the carrier is the output shaft. It also comes with three planetary gears. And the ring gear is integrated with the stator.
Although numerous bearings are used in various parts of this mechanical assembly, the bearings used are from C&U and S&F. Basically, it's C&U. In particular, two large-diameter parts are used for the output shaft section. Mechanical parts like these are actually quite important, but it was basically made up of items available within China.
I took apart a QDD motor commonly used in four-legged robots and humanoid robots. As for the internal components, it consists of a combination of an outer-rotor motor and a planetary gear mechanism. Although small, it delivers high torque. It also has reverse drive capability. I hope you now understand how it's put together.
It's surprising that you can get a motor like this for just a few tens of thousands of yen, and above all, it's great that you can just plug it in and start using it right away. Anyone who is thinking of building a robot, please feel free to use this as a reference.
If you found this video helpful, please give it a thumbs up. If you have any questions or concerns, please feel free to leave a comment. Thank you for watching the video all the way to the end.
We are currently selling breadboard circuit boards made by Ichiken and electronics kits, such as pre-wired LED products. For more details, see the description section.
Article published