Merits of Star Topology for CAN Bus

CTRE did not make up the design requirements for CANbus. ISO takes care of the standard and all the transceiver manufacturers design their equipment using the standard. CTRE follows the standards just like every other OEM. I understand the attraction of the star topology but that is just not how CANbus was designed. My professional advice is to respect the engineers who created CANbus and use it in the manner standardized by the ISO folks.

2 Likes

"The Standard recommends a maximum un-terminated stub length of 0.3 m (12 inches) with a 1 Mbps signaling rate, but with slew rate control, reduced signaling rate, and careful design, longer stub lengths are easily obtained."

2 Likes

“The High-Speed ISO 11898 Standard specifications are given for a maximum signaling rate of 1 Mbps with a bus length of 40 m with a maximum of 30 nodes. It also recommends a maximum unterminated stub length of 0.3 m.”

So 11.8" stubs are allowed. The CTRE wires coming out of the Talons are about that long so adding wire to achieve a star topology is not a good idea. Also having stubs that are long in relation to the “main” run is a bad idea. #physics

1 Like

Unless you plan on hacking the Linux CANbus driver, we are running at 1Mbps. The careful design they are talking about is transceiver design, not wiring topology.

These type connectors are very cool and have many great purposes on FRC robots. But CANbus is not one of them. There is no freaking way these look anything like 120 ohms at 1Mhz.

We should be teaching students to look up the standards and the data sheets or accurately summarize for them. We should not be teaching them to search for whatever is convenient and “seems” to work.

FRC robots are complex mechatronics with some pretty fancy software. The proper engineering processes involve using all the subsystems we buy in the manner the design engineers intended. They invented and tested these things - trust them! Then test the things we design to make sure they perform as intended. If one skirts these basic tenants of good design the statistics will catch up with you. Why create unnecessary problems? Aren’t these robots hard enough to design and keep running?

3 Likes

I’m into electronics, but analog signal characteristics is not my cup of tea.
In your opinion, would the can connectors sold by CTRE not fall into the same bucket as the wagon lever nuts?

Would a geometry like this be more suitable for this application (I’m not endorsing these or this seller, never used them myself)?

1 Like

The folks at CTRE are pretty thorough. My guess is that the CTRE connectors are OK. The UOHGDPY connector might be OK. It is hard to tell from the outside. If the specs are not readily available look for keywords like the connector is for “signaling” or for “communications”.

I guess I could be misinterpreting them, but dont the wiring rules specifically require Daisy chained CAN connections:

It doesn’t explicitly say you can’t use star topology, but my reading of “directly or Daisy-chained” excludes star topology…

Maybe the rules were changed slightly over the years, but pretty sure in 2017, 3719 was forced to re-wire their star topology CAN bus at worlds to pass inspection.

They’re completely within the spec. They work just fine. Source: Seven failure free competitions.

5 Likes

Note that the spec for 12" stubs comes with a cost of needing a certain length on the main line. You can’t just run a 12" star topology, you need a certain amount of meat on the main on bus as well.

1072 has tried:

  • Automotive spade connectors (2017, bad idea)
  • Lever nuts (2018, worked great all season, although we had one fall off a few months later)
  • Soldering, which became 50% lever nuts after “temporary fixes” were applied.

Soldering was fine, but definitely keep some lever nuts on hand in case things go south and you need to make a quick swap. I’m pushing for lever nuts next year, however, as their clear casing makes it easy to inspect connections.

The CTRE CAN connectors are good if you plan on mounting them with the mounting holes. Otherwise, they need to be secured down to prevent them from whipping around and putting strain on the connections.

1 Like

You seem very confident that the thing that worked will not work.

4 Likes

You’re welcome to disregard specs. There’s plenty of ways that the rules can be broken before things really start to break. But the two worst problems that can pop up in the electrical system at competition are Ethernet issues and CAN issues, both because they can be extremely hard to debug. It’s not always just a simple “swap and re-flash firmware”.

CAN issues that stem from something as simple as stub length would be somewhere between “difficult” and “impossible” to debug at competition. I would be tearing my hair out swapping the entire control system before I suspected the layout, and even if I did suspect the layout first, I would have to rewire significant portions of the robot before I could narrow down the issue.

I’ve heard of teams run CAN through brushed slip rings before. It’s not impossible, and CAN is a very robust system, but I wouldn’t do it on a comp bot.

1 Like

If you’re concerned about distance between stubs feel free to add 12" of extra cable between each tee connector, and feel free to cut the stub wire shorter too, but I can tell you - it’s a non issue.

The IDC tee connector is the solution to CAN bus failures, because connector failures don’t interrupt the bus! We have had zero CAN failures this year.

3 Likes

FRC is simply not a competition of strictly following engineering standards and data sheets. Take a look at the the datasheet for the series of motors the CIM is specced from. Listed typical applications do not include a robot drive train, however they do include “Massager” and “Vibrator”. I hope your not using them in the manner the design engineers intended!

Likewise Blue Nitrile and Colson wheels were never intended to be robot drive wheels, one was made for conveyor belting and the other was a high quality non marking caster wheel. Yet they’ve become some of the most popular and reliable choices for FRC drivetrains.

Like real engineers we teach our students to develop new things by testing them thoroughly before putting them into practice. That’s what we did in the off season and it paid off, we managed to reduce our CAN issues this year to zero.

8 Likes

They are designed to connect power to LED strings. They work at DC, as in zero hertz. Your source is an insignificant sample compared to the efforts of the designers of CANbus. Were you monitoring the CANbus with a proper analyzer? Were you even recording problems ? Looking at the driver station logs?

The CIM motors were intended for general purpose use, basically any application that fits the torque curve without overheating. Your attempt at vulgar humor is disappointing in a forum where students read your diatribe.

Mentors should be teaching students proper engineering practices and processes. I do it every day - for 35 years now. If a young engineer at work made your arguments about CANbus I’d show them the door. It is ridiculous. It is mal-practice.

Are FRC users violating any recommendations or data from the OEM? No. Yes, these wheels were created for a specific purpose. But I hope you can understand that a wheel can have generic applications.

I doubt you are equipped with the equipment, expertise or the time to test CANBus like the inventors and the designers of the transceivers. I have the expertise and the equipment but not the time. I’m not telling you how to use CANbus. I’m telling you to respect the work of the engineers who designed this stuff and follow their design guidelines. This is what we should be teaching students.

You should reduce your CANbus problems to zero by properly crimping suitable connectors used in a topology described by the standards (or by CTRE if you don’t want to look it up). You can maintain your harness by testing the bus and connectors before each and every match.

Young engineer: Hey, this thing has worked flawlessly for quite some time now maybe its feasible.
You: The data says that’s bad, Clean off your desk.

Seems like a bit of a superiority complex to me.

If 5406 isn’t having any issues with it, and they want to present on a public form about their method, what is wrong with that?

15 Likes

:man_shrugging:

7 Likes

It is an anecdote. Anyone who will not follow simple guidelines (like CANBus) is unsuitable, not a serious person.

No complex, just the facts. I am an expert offering seasoned advice so students will not follow unprofessional practices.

… because some mentor or student might duplicate the mistake

1 Like

What mistake?

1 Like