Series: FRC Technical Foundations Unit: 2 — Making It Move Session length: ~3.5 hours (this is a long lesson; consider splitting across two sessions) Prerequisites: Lessons 1–2
A brushed or brushless DC motor has a beautifully simple behaviour: torque falls linearly as speed rises. Two numbers define the line.
Draw a straight line between those two points and you have the whole torque–speed curve. Two consequences fall out immediately.
Power peaks in the middle. Power is torque times speed. At free speed, torque is zero, so power is zero. At stall, speed is zero, so power is also zero. Maximum mechanical power occurs at roughly half free speed and half stall torque. If your mechanism spends its working life near either end of the curve, you are wasting most of the motor.
Stall is a thermal event, not a mechanical one. A Kraken X60 or NEO at stall pulls well over 100 A and turns essentially all of it into heat inside the windings. A motor held at stall for a few seconds gets hot. Held for twenty, it can be permanently damaged. This is why current limiting exists, and why “the arm holds position by stalling the motor against gravity” is a design error, not a design.
Rule 3.1 — Design mechanisms to operate near peak power, and never to rest at stall. If a mechanism must hold a load statically, use a brake mode, a hard stop, a counterbalance, or a ratchet — not motor current.
The modern FRC landscape is mostly brushless. You will encounter the NEO and NEO Vortex (REV, driven by SPARK MAX or SPARK Flex), the Kraken X60 and Kraken X44 (integrated TalonFX controller from CTRE), and the smaller NEO 550 for low-inertia jobs like swerve azimuth. Older CIM and mini-CIM motors still appear on training rigs and remain excellent teaching hardware.
What matters pedagogically is not the part number but that every one of them has a published curve, and you are expected to use it.
A gear reduction trades speed for torque. For a reduction ratio N (input revolutions per output revolution):
Output speed = Input speed / N
Output torque = Input torque × N × η
where η is efficiency. Ratios multiply through stages — a 4:1 followed by a 3:1 is 12:1 — and so do efficiency losses.
That efficiency term is where students get burned. Rough working figures:
Stage type | Efficiency per stage |
|---|---|
Spur gear pair | ~95% |
Planetary stage | ~90% |
Chain (#25 or #35) | ~95% |
Timing belt (HTD/GT2) | ~97–98% |
Worm gear | 40–70% |
Three planetary stages at 90% each is 0.9³ ≈ 73%. You planned for 100 N·m of output torque and you have 73. On a lift, that is the difference between working and not.
Rule 3.2 — Compute the ideal number, then multiply by realistic efficiency, then add margin. A design that only works at 100% efficiency does not work.
This is the skill. The process is always the same:
Do not do this arithmetic by hand every time. Use ReCalc or the JVN Mechanical Design Calculator — but do it by hand once, so you know what the calculator is doing.
Chain is tolerant, cheap, repairable at competition, and tolerates slight misalignment. Belt is quieter, lighter, more efficient, and requires accurate centre distances because you cannot easily take up slack.
For both: wrap matters. A sprocket with less than about 120° of chain wrap will skip under load. Tension matters too — a chain run slack will derail, a chain run tight will eat bearings and cost you efficiency. The classic guidance is about 6 mm of deflection at the midpoint of the span.
Two sets of wheels, each side driven together, turning by driving the sides at different speeds. Usually six or eight wheels, with the centre wheels dropped a few millimetres below the outer ones so the robot pivots about its middle rather than skidding the full length of its wheelbase.
Strengths: mechanically simple, extremely robust, cheap, easy to program, excellent at defence and at being defended against, forgiving of alignment errors.
Weaknesses: cannot translate sideways. To move 30 cm left, you must turn, drive, and turn back. Every alignment to a scoring position costs time.
Scrub is the mechanism that makes tank work and also what limits it: turning requires wheels to slide sideways across carpet. Longer wheelbases scrub more, which is why centre drop exists.
Four independently steered and independently driven modules. The robot becomes holonomic: it can translate in any direction while simultaneously rotating, with three independent degrees of freedom in the plane.
The competitive advantage is real and large. Alignment to a scoring location becomes a single motion rather than a three-part manoeuvre, and cycle times drop accordingly. Swerve is the default at the top of the field.
What it actually costs:
Should a first- or second-year team run swerve?
The honest answer is: usually not in year one, and it depends on your team’s bottleneck rather than your ambition.
If your team’s limiting factor is programming capacity — one or two students learning Java — swerve will consume the entire software budget and your manipulators will go unprogrammed. If your limiting factor is build hours, a COTS tank drivetrain can be built in a weekend, and swerve cannot.
If you have a stable programming subteam, a mentor who has run swerve, and the budget, then modern COTS modules have made swerve far more accessible than it was five years ago. The commonly recommended families are the SDS MK4/MK4i (deepest community documentation), REV MAXSwerve (compact, integrates naturally with a REV control system), WCP Swerve X2, and Thrifty Swerve (cost). They are not interchangeable — the choice fixes your motor purchases, your gear ratios, and your floor clearance.
The community guidance is consistent on one point: pick the module with the strongest support network near you, not the one with the best spec sheet.
Rule 3.3 — A reliable tank drive beats an unreliable swerve, every single time. Mecanum, for what it is worth, is the option that is usually worse than both: it has swerve’s alignment complexity with tank’s speed and neither one’s robustness. If you cannot run swerve, run tank.
[ANECDOTE SLOT] — Your own view on the swerve decision, ideally from having watched a team make it one way or the other. This is a section where a real opinion carries more weight than a balanced summary.
Time: 90 minutes. Equipment: Motor/gearbox bench with interchangeable ratios, tachometer or high-frame-rate video, a multimeter or current clamp, a known mass, a stopwatch.
For three different gear ratios on the same motor:
Evidence of learning: A three-row prediction-vs-measurement table, a calculated real efficiency for each configuration, and a written explanation of where the losses went. The gap between prediction and reality is the entire point of this exercise — do not let students treat it as an error to be hidden.
Time: 40 minutes. Equipment: ReCalc or JVN calculator, motor data.
Give each group a different written requirement, for example:
Each group selects a motor count, ratio, and wheel or drum size; states peak current; and confirms it fits under the available breaker.
Evidence of learning: A one-page design sheet per group, defended out loud to another group who must try to break the assumptions.
Time: 60 minutes. Equipment: A drivable chassis, field tiles or a marked floor area, a stopwatch.
Time three events on the same chassis: a 10 m sprint, a slalom, and an alignment task where the robot must reach a marked position and heading from a start point 1.5 m away and 1 m to the side.
Run it twice with different drivers. If you have both a tank and a swerve chassis available, run both and compare — particularly on the alignment task, which is where the difference actually shows.
Evidence of learning: A timing table across drivers and chassis, plus a written argument for which drivetrain suits your team’s game strategy, citing your own numbers.
A drivetrain that goes everywhere and does nothing scores zero. Lesson 4 — Manipulators: Intakes, Arms and Elevators is about the mechanisms that touch game pieces, and why compliance beats precision.
