Skip to content
GitHubYouTubeInstagram

From a button to a spinning motor.laptop → radio → SystemCore → CAN → motor

A button press on the driver station laptop becomes a spinning motor on the robot. This course follows it the whole way: across the network to the robot controller, over the CAN bus to a motor controller, and out to the motor on power from the battery.

5 chapters · about 32 minutes

  1. 01 · 6 minWhat is on the robotThe parts of the control system, and the four kinds of wire between them.
  2. 02 · 6 minMAC and IP addressesWho a device is, and where it is: 10.38.47.x on our robot.
  3. 03 · 5 minWhat a switch doesWatch a switch learn where every device is.
  4. 04 · 8 minThe CAN busTwo twisted wires, read as a difference, with a terminator at each end.
  5. 05 · 7 minPower and wire gaugeBattery to motor: what the wire and connectors cost in volts and heat.
Chapter 1 of 5 6 min

What is on the robot

The parts of the control system, and the four kinds of wire between them.

A driver presses a button on the laptop, and a motor on the robot spins. On the way, the command crosses Wi-Fi, Ethernet, a computer and a CAN bus, and the motor spins on power from the battery. Our 2026 offseason robot has 21 motor controllers, 4 encoders, a gyro and 3 cameras, and one computer that talks to every one of them. Nearly all of it hangs off four kinds of link: power from the battery, CAN for commands and sensor readings, Ethernet for the cameras and the laptop, and one short USB cable from the computer to a CAN adapter.

LabTour the control system

Tap a part, or pick one from the menu. Its wires light up. Then switch between the PDP 2.0 and the PDH and watch what happens to the CAN wiring.

PowerCANEthernetUSB

Wired by—

Simplified: a real robot has many more wires and parts, and the layout is ours. Facts are from Limelight's SystemCore spec sheet, Vivid-Hosting's radio docs, CTRE's and REV's own guides, R609 of the 2026 game manual, and our 2026 offseason robot's code for the device counts.

LabLook at the real parts

Drag to turn a part, scroll or pinch to zoom. Pick a part, then a link, and every connector for that link lights up. Try CAN on the PDP 2.0.

Scroll here to load the 3D model.
PowerCANEthernetUSB

Links on this part—

SystemCore, the PDP 2.0 and the PDH are their makers' own STEP files, colored to match photos of the real parts. The radio is the simpler box model from our robot CAD. Port details are from Limelight's SystemCore spec sheet, CTRE's PDP 2.0 guide, REV's PDH specs and Vivid-Hosting's port table.

The controller

SystemCore is the computer that runs the robot code on every FRC robot from 2027, replacing the roboRIO our 2026 robots ran. It has 5 CAN buses, a gigabit Ethernet port, 4 USB-A ports and 6 ports for sensors that can each be set up as digital, PWM or analog. It runs on 5 to 26 V, a range that covers a 12 V battery sagging under load. Limelight's spec sheet for it is still marked alpha, so these numbers may change.

Power distribution

Power goes from the battery through the 120 A main breaker to one power distribution device, which splits it into circuits with their own breakers. Two are common. CTRE's PDP 2.0, which our robots use, has 24 slots for 40 A breakers and takes the main power on bolts. It isn't on CAN, so it reports nothing, and each Talon FX reports its own current instead. REV's PDH has 20 channels for 40 A breakers, 3 for 15 A and one you can switch on and off from code, and it reports every channel's current over CAN.

The radio

The radio is the robot's link to the driver station. The VH-109 has four Ethernet ports: RIO for the controller, AUX1 and AUX2 for cameras and switches, and DS for a laptop on a cable. The next two chapters follow the button press through it.

The one thing to remember Follow the wires: power from the battery, CAN for commands and sensors, Ethernet for cameras and the laptop, and USB from SystemCore to a CANivore.

Next: MAC and IP addresses ↓

Chapter 2 of 5 6 min

MAC and IP addresses

Who a device is, and where it is: 10.38.47.x on our robot.

The button press starts on the driver station laptop. The laptop sends it to the robot as a network message, and on the field it has to reach one computer on our robot, not a camera on it or any of the other five robots. Two kinds of address make that happen, and they do different jobs.

MAC: the address on the device

A MAC address is 48 bits, written as six pairs of hex digits, like 02:1A:2B:3C:4D:5E. The first three pairs name the company that made the network chip, and the maker gives every chip its own last three. The examples in this course start with 02, which marks an address as made up rather than a maker's. It stays with the device: plug a Limelight into a different robot and its MAC is the same.

IP: the address on the network

An IP address is four numbers from 0 to 255. It can change, because it says where the device is plugged in, not what it is. Part of the address names the network, and a subnet mask says which part. With 255.255.255.0, two devices talk directly only if their first three numbers match.

FRC uses the same plan on every robot, so any team's laptop knows where to look. Your team number splits into two: all but the last two digits (TE) and the last two (AM). Team 3847 is 38 and 47, so our robot network is 10.38.47.x. Team 1002 would be 10.10.2.x.

AddressWho has it
10.38.47.1The radio
10.38.47.2SystemCore
10.38.47.3Kept for the field
10.38.47.4The gateway: the field's access point, or the radio off the field
10.38.47.5A driver station laptop, if set by hand
10.38.47.6 to .19Extra devices on the robot: cameras, a Jetson
10.38.47.20 to .199Handed out to laptops on the field
10.38.47.200 to .219Handed out by the robot's radio

SystemCore doesn't have 10.38.47.2 built in. It asks for an address with DHCP, and the radio gives it that one. You can also reach it by name, as robot.local, thanks to mDNS. With no radio to give it an address, SystemCore picks a 169.254.x.x one itself, which a device on 10.38.47.x can't reach. Plugged into a laptop by USB-C, it's at 172.26.0.1 on Windows and 172.27.0.1 on a Mac or Linux.

LabGive the Limelight an address

Start with the good address, then load each of the others in the menu and read why it fails. Then try to find every way to break it yourself.

Can talk to the radioCan't

  • On the radio's network—A 10.38.47.x address with 255.255.255.0, so the radio and the Limelight agree.
  • Address is free—Nothing else on the robot has it.
  • In the block for extra devices—10.38.47.6 to .19.
IP address (you set)—
Subnet mask—
MAC address (never changes)—

Simplified: only the address rules matter here, and the radio is assumed to sit at 10.38.47.1 with no other devices. The blocks are WPILib's address plan. The MAC address is made up for illustration; real ones are set by the maker.

Move the sliders and toggles and the MAC never changes, because it belongs to the device. Everything else on the screen belongs to the network. A device on the wrong network fails the same way whether the address is wrong by one digit or by a lot.

The one thing to remember The MAC address belongs to the device and never changes. The IP address belongs to the network, and on our robot every device is 10.38.47.something.

Next: What a switch does ↓

Chapter 3 of 5 5 min

What a switch does

Watch a switch learn where every device is.

Our 2026 offseason robot has three Limelight 4s and a robot controller on Ethernet. That's four devices, and the radio has three ports for the robot: RIO, AUX1 and AUX2 (its fourth, DS, is for a laptop on a cable). The usual fix is a small box that adds ports. What that box does with each message decides whether the network stays fast or fills with traffic nobody asked for.

A hub shouts, a switch learns

The simplest box with several ports is a hub. Whatever comes in one port, the hub repeats out of every other port, so every device hears every message and ignores the ones that aren't its own. That works with two devices. With five, each device is sitting through four other conversations.

A switch keeps a table of MAC addresses, one per port. Every message carries the sender's MAC, so each time one arrives the switch writes down which port that MAC is on. When a message comes in for a MAC it has in the table, it sends it out of that one port. If the MAC isn't in the table yet, it does what a hub does and sends the message out of every other port. The device that answers teaches the switch where it is.

LabWatch a switch learn

Send a message between two devices in Switch mode, then send the same one again. Then fill the table, send between two devices it knows, and switch to Hub for the same message.

The message, and who received itSender

  • Teach it all four devices—Send a message from each device in Switch mode.
  • Send a message only one device hears—Send between two devices the switch has already learned.
Devices that got the message—
MACs the switch knows—
PortDeviceMAC in the table

Simplified: one message at a time, no delays, and every device answers only when you send. Real switches also forget a MAC after a while. The MAC addresses are made up for illustration.

The first message from the laptop reaches every device, because the switch doesn't know where SystemCore is. The second time the laptop and SystemCore talk, only they hear it. In Hub mode nothing is ever learned, so every message reaches every device, and the table stays empty.

The radio does this job itself for its four ports. When you run out, plug a small switch into an AUX port, which Vivid-Hosting's port table lists for cameras and switches. From SystemCore on, the command leaves Ethernet behind: the robot code turns it into a CAN message, and that's the next chapter.

The one thing to remember A hub repeats every message to every port. A switch learns which MAC address is on which port, and after that sends each message only where it belongs.

Next: The CAN bus ↓

Chapter 4 of 5 8 min

The CAN bus

Two twisted wires, read as a difference, with a terminator at each end.

The robot code on the controller decides the intake should run, and now it has to tell a motor controller. Our 2026 offseason robot has 26 devices to tell things to and hear from: 21 Talon FX motor controllers, 4 CANcoders and a Pigeon 2 gyro. Running a separate wire to each would be a bundle as thick as your wrist, so they share a pair of wires instead. That shared pair is the CAN bus.

Every device hears every message and picks out the ones meant for it, by its ID. A CTRE device's ID can be 0 to 62, and two devices of the same kind on one bus can't share one. A robot can have more than one bus: ours puts 24 of the 26 on a CANivore, and the two intake rollers on the roboRIO's own bus.

Two wires, one signal

At rest, both wires sit at 2.5 V, which is a 1. To send a 0, a device pulls CAN-H up to 3.5 V and CAN-L down to 1.5 V, a 2 V gap. The receiver only looks at the gap. That's a differential signal. A motor next to the wire makes electrical noise, and if the noise lifts CAN-H by 1 V it usually lifts CAN-L by about 1 V too, and the gap doesn't change.

That only works if the noise reaches both wires equally, and that's what the twist is for. FRC CAN wire is 22 AWG yellow (CAN-H) and green (CAN-L) twisted together, so each wire spends half its length on each side of any noise source.

LabNoise on the bus: twisted or not

Turn the noise up with the wires twisted, then switch to untwisted and watch the red bits. Try turning it up until reading CAN-H alone fails.

CAN-HCAN-LA bit read wrong

  • Read every bit right—Reading the gap between the wires, with the noise turned up.
Wrong, reading the gap—
Wrong, reading CAN-H alone—

Simplified: twisted wire is modeled as the same noise on both wires, and untwisted as noise on CAN-H only; real twisted pairs cancel most of it, not all. Voltages are Texas Instruments' (2.5 V rest, 3.5 V and 1.5 V dominant). The receivers decide halfway between the two levels, which is simpler than a real CAN transceiver's threshold. The noise is made up and the same every time, so runs compare.

The ends of the bus

A signal that reaches the end of a wire can bounce back, like a wave off a wall, and garble the bits behind it. The cure is a 120 Ω resistor across CAN-H and CAN-L at each end, which matches the cable's 120 Ω impedance so the signal has nowhere to bounce. There are two, one at each end. SystemCore has 120 Ω termination built in, and a CANivore has a terminator that is on unless you turn it off in Tuner X. REV's PDH and CTRE's original PDP have one you can switch. The PDP 2.0 has none, because it isn't on CAN at all.

You can check the bus with a multimeter, with the robot off. Measure between CAN-H and CAN-L. The two 120 Ω resistors sit in parallel, so the meter should read about 60 Ω, according to CTRE's troubleshooting page.

Terminators on the busThe meter readsWhat it means
0openNeither end is terminated, or the bus is broken between you and them.
1120 ΩOne end is missing its resistor. Check the other end.
260 ΩCorrect.
340 ΩA third terminator was added somewhere in the middle.

Resistors in parallel divide, so n terminators read 120 Ω divided by n. A very low reading means a short between the wires.

How full is the bus?

Every message takes time on the wire, so a bus has a capacity. The share of it in use is bus utilization, and CTRE says to keep it under 90%. Our CANivore ran at 66 to 81% on 2026-09-04, with every motor sending its status signals 250 times a second. We cut that in two steps, first the signals nobody reads fast, then position and velocity to 100 Hz on the motors that lead and 20 Hz on the ones that follow. In qual 45 it sat at 44 to 61%.

CAN 2.0 and CAN FD

The other way to make room is a faster bus. Classic CAN 2.0 sends every bit at 1 Mbit/s and fits 8 bytes of data in a frame. CAN FD keeps the same two wires and the same terminators, but fits 64 bytes in a frame and sends the data part faster.

Only the data part. Every frame starts with the sender's ID, and that part is how the bus decides who talks: if two devices start at once, both drive the wires together, bit by bit, until one sees a bit that isn't its own and stops. That only works if every device can see the whole bus within one bit, and the wire sets that limit, so the ID goes at 1 Mbit/s on both kinds of CAN. Once one device has won, it's the only one talking, and it can switch to a faster rate for the data.

CAN 2.0CAN FD
Data in one frame8 bytesup to 64 bytes
The ID, which decides who talks1 Mbit/s1 Mbit/s
The data1 Mbit/s10 Mbit/s on a CANivore; SystemCore is rated for up to 8 Mbit/s
One Talon FX at CTRE's default rates4.1% of the bus1.8% of the bus
Where you find itthe roboRIO's own busa CANivore, and SystemCore's five buses

A CAN FD bus needs every device on it to speak FD, because a CAN 2.0 device reads an FD frame as an error. Phoenix 6 devices like the Talon FX, CANcoder and Pigeon 2 do. How fast SystemCore will run its buses in 2027 isn't settled yet: its spec sheet says up to 8 Mbit/s, and it may run lower.

LabRead the bus: three real logs

Start on practice match 8, where a wire was cut, then compare it with qual 45 and with the restart. Find what changed in utilization and in the error counter.

Bus utilizationTransmit error counter

Typical utilization—
Highest error counter—

The traces are the CANivore's logged bus utilization and transmit error counter from three of our Chezy Champs 2026 logs, with utilization sampled once a second. The error counter is logged only when it changes, so it shows as spikes, drawn against 255, where a device gives up and goes bus-off.

A bus that goes quiet looks the same in the log whether the wire is cut, a terminator is missing or a device is dead: utilization falls and the error counter spikes. The multimeter check separates them. With the robot off, a cut that splits the bus leaves each half with one terminator, so you read 120 Ω from either end.

The one thing to remember CAN is two twisted wires read as a difference, with one 120 Ω terminator at each end. With the power off, the meter should read 60 Ω.

Next: Power and wire gauge ↓

Chapter 5 of 5 7 min

Power and wire gauge

Battery to motor: what the wire and connectors cost in volts and heat.

The command has reached the Talon FX. The motor still needs power to spin, and it gets it through two wires, one out from the power distribution device and one back to it. A wire has a little resistance, so it uses up some voltage as current passes through it. The volts it uses are gone before the motor sees them, and the same power turns into heat in the wire.

Two things set a wire's resistance

AWG is the size of the copper, and a smaller number is a thicker wire. Copper is about 1.0 mΩ per foot in 10 AWG, 1.6 mΩ in 12 AWG, 0.4 mΩ in 6 AWG and 0.16 mΩ in 2 AWG. Every 3 gauge numbers halve the resistance. The current goes out and comes back, so a run that is 6 ft one way is 12 ft of wire.

FRC also has rules on what wire you may use. R622 in the 2026 manual sets the minimum for each breaker: 12 AWG on a 31 to 40 A breaker, 14 AWG on 21 to 30 A, 18 AWG on 6 to 20 A. The breaker is there to keep the wire from overheating, so that rule is about safety, not performance. A wire that passes inspection can still lose enough volts to hurt the robot.

LabWire the turret: 10 AWG or 12 AWG

Switch between 10 and 12 AWG and drag the length. Find the longest run where 12 AWG passes both targets, then see how much longer 10 AWG can go.

Volts lost in the wireThis lab's 0.75 V target

  • Volts lost at the peak—At most 0.75 V, so the motor still gets nearly what the PD sends.
  • Heat at the peak—At most 40 W in the wire.
Peak current—
Lost at the peak—
Heat at the peak—

Simplified: solid copper at 20 °C, and the wire doesn't heat up. Current is the turret's logged supply current from Chezy Champs qual 45 on our 2026 offseason robot: 50 A at the peak of this window, 65 A over the whole match. We haven't measured the run, so its length is yours to set. The 0.75 V and 40 W targets are this lab's, not a rule.

At 50 A, 10 AWG loses 0.60 V over a 6 ft run and 12 AWG loses 0.95 V. That's 30 W against 48 W of heat, in a wire that isn't cooled. 12 AWG passes the lab's targets up to 4 ft, and 10 AWG up to 7 ft. The Kraken X60 ships with 18 in, 10 AWG power leads. A Chief Delphi thread reports about 1 V lost at peak, and about 60 W in the wires, on an 8 ft run of 12 AWG to a shooter motor, the same size of loss the lab shows.

The main path carries everything

The wire from the battery through the main breaker to the power distribution device carries the current of every motor at once. In qual 45 our offseason robot's total current peaked at 402 A, and the battery fell to 7.35 V at that moment. The roboRIO 2 and SystemCore both turn the motors off below 6.75 V by default, so the main path is the wire that matters most.

The path is short. On our robots each battery lead is about 8 in, and the run from the SB connector through the main breaker to the PDP 2.0 is about 1 to 1.5 ft on each of the red and black, so the loop is at most about 5 ft. Our robots run 2 AWG on it, with ring terminals on the PDP 2.0's bolts. CTRE's guide lists 4 AWG as the largest the PDP 2.0 takes, and we run 2 AWG anyway. REV's PDH takes up to 4 AWG in its tool-less connector, so teams that want 2 AWG on the battery side step down to 4 AWG at the PDH. R609 only asks for 6 AWG or larger.

The connector matters too. R609 requires one pair of Anderson SB connectors, and FIRST's batteries come with a red or pink SB50, so that's what most robots use. The SB50's contacts take 16 to 6 AWG and the SB120's take 10 to 1 AWG, so 2 AWG needs an SB120. Anderson rates the SB50 at 50 A (CSA) or 120 A (UL), and the SB120 at 130 A, or 240 A with 1 AWG. Our peak is 402 A, but for a fraction of a second.

LabThe main path: wire and connector

Start at 6 AWG with SB50, then switch to SB120, then move the wire to 4 and 2 AWG. Drag the run to see what a longer path costs.

As logged, 2 AWGWith the wire and connector you pickedBrownout

  • Lowest battery voltage—Stay above 6.75 V, the default brownout.
Lost at 402 A—
Of that, connectors—
Heat at 402 A—
More than our 2 AWG—

Simplified: the recording is from our 2026 offseason robot, which had a roboRIO, 2 AWG leads and SB120 connectors. For the setup you pick, the lab adds the difference in resistance times the current. Contact resistance is Anderson's datasheet figure with its test wire subtracted (about 0.16 mΩ for SB50, 0.06 mΩ for SB120), so it's an estimate. Total current is the robot code's sum of every device's supply current, and the battery voltage was logged about 9 times a second, so the true lows are a little lower. The 6.75 V line is the default brownout; our code lowers it to 6.0 V.

With a 1.25 ft run, 6 AWG and SB50 lose 0.74 V at 402 A, which is 0.44 V more than 2 AWG with SB120. That takes the log's lowest point from 7.35 V to 6.91 V, 0.16 V above the brownout. Most of the difference is the wire: swapping only the connector, from SB50 to SB120, is worth about 0.08 V at 402 A, and swapping only the wire is worth 0.37 V. The battery costs far more than either. The Motors course fits our battery's internal resistance at 14.4 mΩ, which is 5.8 V at 402 A. Better wire and connectors can't fix a weak battery, but on a robot this close to the line, 0.4 V is the difference between browning out and not.

The one thing to remember Every foot of wire costs volts in proportion to the current it carries, so put the thickest wire where the current is highest: the main path, which carries all of it.

QuizCheck yourself
  1. 1. Which link carries a motor command from the CANivore to a Talon FX?

  2. 2. A Limelight is set to 10.38.48.11 with mask 255.255.255.0. The radio is 10.38.47.1. Can they talk directly?

  3. 3. You change a Limelight's IP address. What stays the same?

  4. 4. A switch that has just been turned on gets a message for a device it has not heard from. What does it do?

  5. 5. Why does a differential receiver ignore noise picked up on a twisted pair?

  6. 6. With the robot off, a meter across CAN-H and CAN-L reads 120 Ω. What is the likely problem?

  7. 7. A CANivore runs CAN FD with a 10 Mbit/s data rate. Why does the ID at the start of every frame still go at 1 Mbit/s?

  8. 8. The turret draws 50 A through 6 ft of 12 AWG each way. About how many volts does the wire lose?

  9. 9. At 402 A on our main path, which saves more volts: 6 AWG to 2 AWG wire, or SB50 to SB120 connectors?

That's the Control Systems course.

Head back to all courses.