Month 3 of 6
Mechanical design, CAD and manufacturing your own parts
Mechanical design, CAD and manufacturing your own parts
This is the month that separates people who assemble kits from people who build robots
Every robot you build after this will contain parts that exist only because you designed them, and being able to go from an idea to a physical bracket in an afternoon changes what projects are possible for you
What to learn
1CAD
Pick one tool and go deep rather than sampling all of them
Decision framework:
- Onshape Free if you are happy designing in public. It runs in a browser, works on any machine including a Chromebook, and its assembly and mate system behaves the way robot joints actually behave. The catch is real: every document is public on the free tier
- Fusion Personal if you want CAM and 3D-print integration later. Free on a renewable 3-year term for non-commercial use under $1,000 a year of revenue, but with limited import and export file types
- FreeCAD if you are somewhere cloud CAD or card payment is a problem, or you object to a licence that can be revoked. Version 1.1 landed in March 2026 and it is genuinely usable now
- SOLIDWORKS for Makers at $48/year if you want the industry-standard tool, with the caveat that native files are watermarked and will not open in commercial SOLIDWORKS
Resources
Onshape Learning Center, Fundamentals: CAD (free with account)
The only free structured CAD curriculum that ends in a credential you can put on a CV, and it ships a dedicated robotics-competition track
Product Design Online, Learn Autodesk Fusion in 30 Days (free)
Thirty modelled objects in thirty days, which is the fastest route from never having opened CAD to confident parametric sketching
MangoJelly Solutions FreeCAD tutorials (free)
The best FreeCAD teacher for makers, organised as short targeted lessons rather than one monolithic course
Protolabs Network, Design for 3D printing (free)
The design-for-manufacture half of CAD: wall thickness, orientation, tolerances, supports, snap-fits, and when to use STL versus 3MF versus STEP
What to focus on
Practice task
model a bracket that holds the exact servo you bought, with correctly sized screw holes and a shaft clearance, from the manufacturer's datasheet drawing rather than by eye. Then print it and see if it fits, which it probably will not the first time, and that failure is the lesson
23D printing
Verified printer prices, September 2026:
- Creality Ender-3 V3 SE, $199
- Bambu Lab A1 mini, $219.99
- Bambu Lab A1, $299.99, with the 256mm bed you will want for larger brackets
- Creality K1C, $369, enclosed and hardened for carbon-fibre filaments
- Bambu Lab P1S, $799, enclosed CoreXY for ABS and ASA
Filament, and when to use each
- PLA and PLA+ for prototype brackets, jigs, and the SO-101 arm itself, which specifies PLA+ at 15% infill and 0.2mm layers. Stiffest of the easy materials, and it creeps under sustained load and softens around 55 to 60°C
- PETG for the default real robot part: chassis plates, gearbox housings, servo mounts, anything that must survive a drop. Tough with far better layer adhesion than PLA, and stringy
- ABS and ASA for parts near hot motors and for outdoor rovers, and they warp badly without an enclosure
- Nylon for gears and cable guides, and it is easier to order than to print
- Carbon-fibre filled for stiff structural links, and it needs a hardened nozzle because it is abrasive
- TPU for feet, bumpers and compliant gripper fingers
Resources
OrcaSlicer Calibration wiki (free)
Temperature, flow, pressure advance, retraction and tolerance calibration in a recommended running order, and it is the most useful slicing document for anyone who wants parts that fit
Teaching Tech 3D Printer Calibration (free, interactive)
A printer-agnostic interactive walkthrough that takes you through every calibration in sequence
CNC Kitchen (free)
Instrumented, repeatable strength tests on infill, walls, threaded inserts and orientation, which is where print orientation stops being folklore and starts being data
Clearance and Tolerance 3D Printer Gauge (free STL)
Print this once and you know your machine's real clearance for press fits and sliding fits, which every bracket and bearing seat you design afterwards depends on
If you cannot buy a printer:
- Fab Labs worldwide, roughly 2,875 of them, searchable by country
- Public library makerspaces, free or near-free in much of the US
- Craftcloud compares quotes across a network spanning 95 countries and routes to a manufacturer near you, which is the right choice outside the US and EU
- JLC3DP starts at $1.00 per part for MJF nylon and FDM, with 3-day builds
The honest economics: an SO-101 arm needs roughly 1kg of PLA+, which is about $20 to $25 of filament on your own machine against $30.99 for a ready-printed set. A printer does not pay for itself on one build. It pays for itself on iteration, because the tenth revision of a gripper finger costs 40 cents and 25 minutes at home, against $8 and a week from a service
Practice task
print the tolerance gauge, write down your machine's actual clearance numbers, then design and print a two-part snap-fit enclosure for your ESP32 that closes without glue. Iterate until it clicks properly. This is the loop that all mechanical design is made of
3Actuators, transmissions and why robots are hard
Understanding gear reduction, backlash and torque density is what separates a robot that works in a video from a robot that works repeatedly
What to focus on
Practice task
design and print a simple planetary or cycloidal reducer for a NEMA17 stepper or a hobby motor, and accept that the first one will be bad. Measure the backlash by holding the output and rocking it, then redesign to reduce it. Reference builds are on Instructables and Hackaday if you want a starting geometry
4Build a real robot arm
This is the capstone of the month, and it is the single best hardware purchase in this entire roadmap
The SO-101 is an open-source 5-DOF arm plus gripper from TheRobotStudio and Hugging Face, designed to be built as a leader and follower pair so you can hand-guide one and have the other mirror it
That teleoperation setup is what lets you record demonstrations, which is what month six is built on
Official bill of materials, verified in the repo: $229.88 US for a leader and follower pair, $121.94 for a single follower arm, excluding 3D printing
github.comWhere to buy, verified prices
- Seeed Studio SO-ARM101 Pro servo kit, $277.99, motors and control boards without printed parts
- Seeed Studio printed parts set, $30.99, if you have no printer
- Robonine SO-ARM101 complete kit, $349.00, shipping from Delaware
- WowRobo via OpenELAB, $325.99 printed parts plus servos, $419.99 unassembled full kit, $489.99 fully assembled
Cheaper alternatives at every tier
- $0: everything in the LeRobot stack runs in MuJoCo simulation before hardware exists, which is the answer if you cannot import anything
- $50 to $80: the EEZYbotARM MK2, free STLs, built from MG996R hobby servos and printed parts. It teaches linkage kinematics rather than servo-bus protocols Link: www.thingiverse.com/thing:1454048
- $122: a single SO-101 follower arm if you print the parts yourself. You lose teleoperation and keep the entire software path
- $199.99: Hiwonder xArm 1S, the cheapest arm with intelligent bus servos that report position and voltage Link: www.hiwonder.com/products/xarm-1s
Practice task
build the SO-101, calibrate every servo, and teleoperate the follower with the leader. Then design and print your own gripper fingers to replace the stock ones, in TPU, and test them on three objects of different shapes. The arm is the platform for months five and six, and the custom fingers are the part that proves you can design as well as assemble
Month 3 Milestone
By the end of this month you should be able to:
Check yourself
2–3 hours a day is the design load.