Absolute Encoder Quasi Direct Drive Motor
Estimate QDD joint feedback resolution and compare encoder architectures for robotics.
* Note: The calculated output step assumes a perfectly rigid transmission. In practice, a quasi direct drive motor's true position accuracy depends on bearing clearance, gear backlash, and system stiffness.
Published by QDD Motors · Published and reviewed
Method and worked example
For a single-turn motor encoder with b bits and a reduction of N:1, the ideal output increment is 360° ÷ (2b × N). The result assumes an ideal transmission and evenly spaced counts. It describes quantization only; it is not an accuracy, repeatability, or startup-pose guarantee.
| Input or comparison | Value | How to interpret it |
|---|---|---|
| Motor encoder resolution | 16 bit = 65,536 counts/turn | Vendor-reported inner-loop resolution for this model. |
| Reduction ratio | 9:1 | Nine motor turns per output turn. |
| Ideal output quantization | 589,824 increments/turn; 0.000610° or 2.20 arcsec/increment | Computed from the motor encoder and reduction ratio; not measured output accuracy. |
| Published gearbox backlash | 15 arcmin = 900 arcsec | A separate vendor-listed mechanical figure, about 410 ideal increments; it is not a complete joint-error budget. |
The AK80-9 figures are for the V3.0 KV100 listing reviewed on . The listing does not identify its encoder as single-turn or multi-turn absolute. Confirm the exact revision, encoder range, accuracy, backlash measurement conditions, and drive startup behavior before using it as a sourcing specification [cubemars_ak80].
Startup check: a 9:1 single-turn ambiguity
With an ideal 9:1 reduction, one motor turn moves the joint 40°. If the motor encoder reports only its single-turn angle and the joint can travel through a full revolution, that same motor reading can match nine joint positions spaced 40° apart. More bits make each reading finer; they do not identify which output turn occurred while power was off.
| Feedback configuration | Question before motion | Minimum next check |
|---|---|---|
| Single-turn motor encoder | Can allowed joint poses share this motor reading? | Prove a unique mapping over travel or establish a reference. |
| Multi-turn motor encoder | Are turns retained and unpowered movements captured? | Test the documented retention method through power-loss cases. |
| Output absolute encoder | Does its range cover joint travel, and is its zero calibrated? | Verify mounting, offset, wraparound, and startup validity. |
These are kinematic and integration checks, not a measured fault rate or a robot safety function. Verify the encoder range, retained turn state, drive behavior, and allowed joint travel before enabling motion.
Executive Summary
- Absolute position is range-limited: An absolute encoder can provide position without a homing move inside its measuring range. Single-turn versus multi-turn coverage, startup calibration, drive support, and motion while unpowered determine whether the robot can safely recover a joint pose [broadcom_absolute].
- Resolution is not accuracy: A motor encoder and reduction ratio yield ideal output quantization. Backlash, compliance, sensor error, load, and calibration remain in the actual joint error budget.
- Torque sensing is a separate decision: Current-based estimates can support proprioceptive control, but they rely on motor calibration and drivetrain models; use direct torque sensing when the application requires it [wensing].
- Check the complete actuator: For example, CubeMars lists one 16-bit magnetic inner encoder and 15 arcmin backlash on the AK80-9 V3.0, but its product page does not specify single-turn or multi-turn absolute behavior [cubemars_ak80].
Feedback Architectures: Single vs. Dual Loop
A motor-side encoder supports rotor feedback. Adding a joint encoder measures the load after the transmission, which can improve load-position feedback when backlash or compliance matters. Whether either sensor is absolute depends on the required measuring range and the drive.
| Feature | Motor-side encoder only | Motor + output encoder |
|---|---|---|
| Primary Use Case | Lower sensor count and simpler control when output position can be inferred adequately. | Load-position feedback where output error through the transmission matters. |
| What is measured | Rotor angle; output position is inferred from ratio and a transmission model. | Rotor angle and load-side joint position directly. |
| Power-up pose | Depends on encoder range, retained turn count, calibration, and drive startup behavior. | Depends on output encoder range and reference retention; motor commutation still needs suitable rotor feedback. |
| Backlash Handling | Cannot directly observe motion or deflection downstream of the motor encoder. | Measures load position through dead zones; feedback cannot physically remove clearance. |
| Cost and validation | Lower sensor, cabling, and integration cost; validate inference error under load. | Adds sensor, mounting, wiring, synchronization, drive support, and loop-tuning work. |
Need to select a feedback layout for your joint? Request a joint feedback review with your travel, accuracy target, and power-off motion conditions.
Risks, Mitigation & Evidence
| Risk | Consequence | Mitigation / alternative |
|---|---|---|
| Overspecifying motor encoder resolution | Paying for finer quantization that does not reduce transmission, mounting, or sensor error. | Compare all error terms in compatible units. On the cited AK80-9 V3.0 listing, 15 arcmin backlash is about 410 times the ideal 16-bit, 9:1 increment; confirm measurement conditions with the vendor. |
| Assuming dual-loop eliminates backlash | Position error or unstable response if the control loop, update rate, or transmission is unsuitable. | Use load feedback to observe output behavior, then validate controller stability and bandwidth. It cannot physically remove gear clearance. |
| Assuming "absolute" means full joint pose after power loss | A single-turn motor encoder may not know output turns or downstream motion while unpowered. | Specify single-turn or multi-turn range, retention method, output feedback, homing policy, and safe startup behavior for the actual system. |
Low-ratio actuator designs can reduce motor inertia reflected to the output, while current-based force estimates still depend on calibration and drivetrain losses [seok] [wensing]. Dual-loop feedback can measure load-side position, but actual performance depends on the drive and cannot remove physical gear clearance [synapticon_dualloop].
Evidence & Sources
Sources reviewed .
Wensing et al. (2017): Proprioceptive Actuator Design in the MIT Cheetah
Studies a proprioceptive actuator for impact mitigation and high-bandwidth physical interaction in dynamic legged robots. It supports discussing actuator transparency and force estimation; it is not a specification for absolute encoders.
Limit: The results concern the MIT Cheetah design and do not establish a universal commercial QDD ratio, encoder type, or torque-estimation accuracy.
Seok et al. (2012): Actuator Design for High-Force Proprioceptive Control in Fast Legged Locomotion
Presents actuator design for high-force proprioceptive control in fast legged locomotion and discusses the trade between motor, gearing, and reflected inertia.
Limit: Its specific actuator and locomotion goals do not define a universal 1:10 ratio limit or guarantee the performance of another transmission.
CubeMars: AK80-9 V3.0 KV100 product specifications
The vendor lists 9:1 reduction, one magnetic inner-loop encoder at 16-bit resolution, and 15 arcmin backlash for the AK80-9 V3.0 KV100.
Limit: The product page does not specify single-turn or multi-turn absolute behavior. Specifications are model- and revision-specific; confirm the current manual and datasheet with the supplier.
Synapticon: Dual-loop position control with two encoders
Explains the motor-side inner loop and load-side position loop and how load feedback can respond to backlash and mechanical compliance.
Limit: Implementation details and achievable performance depend on the selected drive, encoder interfaces, transmission, and controller tuning.
Broadcom: Absolute encoder single-turn and multi-turn overview
Distinguishes single-turn absolute position from multi-turn position tracking and describes absolute position reporting at power-up.
Limit: This is a product-family overview; verify measurement range, turn retention, interface, calibration, and startup behavior for the exact encoder and drive.
Texas Instruments: Motor-control startup and initial-position detection
Describes rotor-position initialization and alignment/IPD startup options before sensorless FOC enters closed-loop operation.
Limit: Covers TI motor-control implementations and does not define startup requirements for every encoder-equipped servo drive.
Frequently Asked Questions
What defines a quasi direct drive motor?
Quasi-direct-drive (QDD) describes a high-torque motor paired with a relatively low-ratio transmission to balance torque density, back-drivability, and output speed. There is no universal gear-ratio cutoff; judge the complete actuator by its reflected inertia, friction, backlash, continuous torque, and thermal limits.
Why use an absolute encoder instead of incremental for a quasi direct drive motor?
An absolute encoder reports position within its configured measuring range without first searching for an index or home mark. That can avoid a homing move, but it does not guarantee safe, motion-free torque at power-up: confirm single-turn versus multi-turn coverage, the encoder-to-electrical-angle calibration, drive support, data validity, and the joint pose after power-off motion.
Does the absolute encoder replace a torque sensor in QDD applications?
No. Motor current and a calibrated torque constant can estimate motor torque; a transmission and dynamics model can then estimate output torque or contact force. Friction, acceleration, temperature, and gear compliance add error. Use an output torque sensor when the application needs direct measurement, tighter force accuracy, or independent validation.
Should I mount the encoder on the motor shaft or the joint output?
The drive needs suitable rotor feedback for its commutation and speed loops; this can be absolute or incremental, depending on the drive and startup strategy. A second encoder at the joint measures load position after the transmission and can expose backlash or deflection that the motor encoder cannot see. It adds hardware, wiring, latency, and control-tuning work.
How does gear ratio affect the encoder resolution at the joint?
For an ideal reduction N:1, a motor encoder with 2^b counts per turn corresponds to 2^b × N theoretical motor-derived increments per output turn. This is quantization, not accuracy: backlash, compliance, eccentricity, sensor error, and control behavior remain. The motor rotor inertia reflected to the output also grows with N², so a larger ratio is not a free improvement.
What encoder does the AK80-9 V3.0 use?
CubeMars lists the AK80-9 V3.0 with a 9:1 reduction, one 16-bit magnetic inner-loop encoder, and 15 arcmin backlash. Its published page does not identify that encoder as single-turn or multi-turn absolute, so verify the exact revision and manual before treating it as a power-up absolute-position reference. These values describe one model, not a QDD-wide standard.
Does an absolute encoder always retain the robot joint position while power is off?
No. A single-turn encoder gives position within one shaft revolution. A multi-turn encoder adds a turn count, but retention depends on its design and whether the count is preserved. A motor-side encoder also cannot determine the joint output position across gearbox motion, backlash, or an unpowered external movement; check the whole feedback chain.
Does higher bit depth mean better position accuracy?
It means finer quantization for the specified measuring range. Accuracy also depends on the encoder error specification, mounting and alignment, gearbox backlash, stiffness, load, temperature, calibration, and the control loop. Compare those error terms in compatible units before paying for more bits.
When is a second encoder at the output worth the cost?
Consider one when load-side position matters more than motor-side inference, such as precision positioning through a compliant or backlash-prone transmission. It can measure the resulting load error, but it cannot physically remove gear play. Confirm that the drive supports dual-loop feedback and validate stability, update rate, wiring, and output-sensor protection.
Discuss Your Quasi Direct Drive Requirements
Are you integrating a quasi direct drive motor into an exoskeleton or quadruped? Send us your torque, speed, back-drivability requirements, and preferred absolute encoder interface.
Inquiry Email
Include target torque/speed, quantity, and delivery location.
Instant Chat
+86 188 5797 1991
Opens a direct WhatsApp thread with the RFQ message prefilled.