Jastej
Sandhu
I build the parts of robots that nobody photographs. This includes the gearbox between the motor and the joint, the firmware between the encoder and the controller. All of it converges on one machine: the ASX-6. A 1.2m 6-axis robotic arm that I am designing and building from scratch.
dimensions in mm · angles ±0.5°
do not scale drawing · break sharp edges
- Interpret this portfolio per ASME Y14.100. All work shown was designed, built, tested, and — where noted — broken by the author.
- Pursuing a dual degree in Mechatronics and Business at Simon Fraser University, Burnaby BC.
- Primary focus: compact, low-backlash transmissions and actuator packages such as wave drives, cycloids and the test rigs that prove them out.
- Firmware between encoder and controller counts as part of the mechanism, with the same tolerance discipline as the hardware.
- The BBA half supplies cost, schedule, and supply-chain judgment for when the goal is a product, not a prototype.
- Material & finish: prototypes in PETG/other 3D printed materials depending on situational material requirements; steel where it matters.
later sheets inherit all general notes
queries — see SHT 05 approvals
Eccentric rolling-element wave-drive
A rare low-backlash transmission taken from obscure literature to a
machined/3D printed hybrid gearbox over five distinct iterations. This design mimics the principle that modern ball bearings operate on to provide high efficency
as a reducer; now at the heart of the ASX-6’s major joints.
This gearbox was designed for the base joint of ASX-6, and meant
to be used in conjunction with an input belt reduction. A through bore encoder shaft architecture has already been implemented
in the robotic arm's physical base joint.
The 2.3 arcminute number was measured using a laser deflection test on prototype 3. This prototype
had no machined parts, and was purely 3D printed. The final version of this gearbox is currently under testing with a 17-bit encoder, and graphical data will
be added to this project page soon. The final version is showing results in the ~1 arcminute range now, getting extremely close to the noise floor of the encoder
which sits at 15-bits as per datasheet and observed data (MT6828S). Graphical illustrations to be posted soon.
repeatability from N-step fwd/rev rig
magnetic encoder on output shaft
ASX-6 — six-axis robotic arm
A full-scale arm where every gearbox, geartrain, joint, and linkage is designed
from scratch. Dual encoders on every major joint (wrist roll & end-effector
roll excluded due to cost measures). Every project on this site feeds this build.
This build will use Eccentric Rolling-element Wave Drive gearboxes for every high load joint (Base, Shoulder, and Elbow joints)
and 3D printed strainwave gearboxes for low-load joints where a high ratio is required in a small package (Wrist pitch, wrist roll, and end effector roll).
Compliance is to be compensated in real-time using PID loops, and the gearbox output
encoders.
All gearbox output encoders are being radially aligned by 2X 10x15x4 bearings for maximum precision, with the encoders being MT6826S 17-bit
sensors. This project is being funded by a part-time sales job.
The motor scheme is a hybrid BLDC/stepper motor mix based on motor availability, and joint requirements. BLDC motors are being controlled using
FOC for the following joint: Shoulder, Elbow, and Wrist. Stepper motors are allocated and controlled using powerful stepper drivers. The design and build are both underprogress in a parallel scheme.
parts print continuously · components
staged ahead of each assembly phase
Force-balanced cycloidal drive
Two-phase cycloidal gearbox, disks phased 180° for vibration cancellation, and force balancing. Went head-to-head with the wave drive for the ASX-6 base joint, and lost the pick due to cost analysis, and manufacturing complexity comparison. 135mm envelope just like the wave-drive gearbox project on this website.
Field-oriented control · ODrive
Field Oriented Control (FOC) is found throughout the ASX-6 robotic arm project. Exploring affordable, yet powerful/precise FOC options brought me to ODrive. A clone ODrive pictured in the images above was purchased for ~$56 CAD as an ultra-cheap high performance FOC module. The cheap price tag caused me to explore the entire working system of an ODrive controller - from using DFU to flash firmware versions, changing ODrive tool versions, to encoder SPI communication variation research. Full start to high precision set-up with SPI encoders shown in this project. Universal across all open-source ODrive clones. This particular board is being swapped for a Flipsky Unit for high voltage reliability issues seen on the ultra-cheap option
AGV — path-teach platform
2×2 ft ESP-32 vehicle that learns a driven path once and repeats it on call, both in forward and reverse. Positioning is managed from step counts vs time recording. Omni-directional movement with the help of Mecanum wheels, paired with a custom 2:1 reduction on each wheel.
every build on this sheet feeds the ASX-6
open any detail for its full sheet
| Item | Description | Spec / note |
|---|---|---|
| Mechanical | ||
| 001 | SolidWorks · Fusion 360 | Parametric assemblies, DXF-driven profiles |
| 002 | Bearing & gear design | Preload, fits per manufacturer tables |
| 003 | GD&T · tolerance stacking | ASME Y14.5 · worst-case + RSS |
| 004 | DFM / DFA | Prototype → CNC handoff |
| Electrical · firmware | ||
| 005 | Field-oriented control | ODrive · SimpleFOC · loop tuning |
| 006 | ESP-32 · STM32 | C/C++ firmware, wireless control |
| 007 | Encoders & drivers | AS5047P · 17-bit absolute · TMC5160T |
| Software · math | ||
| 008 | Python · NumPy | Profile generators → DXF toolchain |
| 009 | MATLAB · Simulink | Kinematics, control models |
| 010 | Cycloid & wave-profile math | Rolling-contact + closure constraints |
| Fabrication · test | ||
| 011 | 3D printing | FDM perfected using python script for expansion modelling. |
| 012 | Repeatability rigs | Backlash measurement, encoder traces |
| 013 | CNC machining liaison | Drawings out, hardened steel back |
| 014 | Failure analysis | Wear tracking, honest post-mortems |
QTY: as required.
SUBSTITUTIONS: tools are interchangeable;
the discipline is not.
ask about any row in an interview
and get the full story
Let’s build
something
that moves.
Robotics, actuator development, drivetrain engineering, and adjacent motion-control roles. Happy to talk shop — wave drives, cycloids, FOC tuning, you name it.
| Function | Name / signature | Date |
|---|---|---|
| DRAWN | Jastej Sandhu | 2026-07-04 |
| CHECKED | Your signature here, recruiters welcomed | — |
| APPROVED | PENDING OFFER | — |
| jastejsandhu11@gmail.com | → | |
| /in/jastejsandhu | ↗ | |
| GITHUB | Coming soon | |
| RESUME | PDF · 1 page — issued on request |
Revision history — this website
| Rev | Date | Description |
|---|---|---|
| A | 2025-11 | Initial release — dark theme |
| B | 2026-05 | Space-glass revision |
| C | 2026-07 | Redrawn as an engineering drawing package |
| D | 2026-07 | Fusion — hazy-glass sheets over the starfield · ASX-6 added |
| E | 2026-07 | Nanowhite frost — static backdrop, frosted glass, performance pass |
This document contains no proprietary
information — share with your team