Falcon 500 vs MiniCIM Drivetrain Simulation

My team has been debating going all-in and dropping $1.2k on some Falcons. It’d be much cheaper for us to stay on SRXs. I decided to do a quick simulation of a sprint:

  • 4 Falcons / Robot, 6 miniCIMs / Robot
  • Global current limit of 400A / Robot (ie 100A / Falcon, 66A / miniCIM)
  • Geared for the same speed, ~14 fps free (8:1 on the falcons)
  • Constant 12V applied
  • Voltage sag is ignored
  • Free current is ignored
  • Rotational Inertia of the wheels is ignored

Here’s the summary graph:

And here’s the python3 code:
sprint.py (3.0 KB)

These results seem a little underwhelming given the hype. If they’re correct, we’d likely go with miniCIMs.
Is there an implication I’m missing? Did I make a mistake or simplify too much in the simulation?

EDIT: I had made some mistakes with constants; it’s fixed now. Results are the same as before, save for a constant multiple.

8 Likes

I am not the best person to evaluate the math being done here, but even if the performance does align this closely in practice, thinking about the weight and space savings the Flacons offer is enormous. I would find that hard to give up. They definitely are expensive and unfortunately what is worth the money will vary from team to team, but the weight savings and the 2 PDP slots are a huge factor here.

5 Likes

Last year we ended up landing at a 180A global current limit to prevent brownouts. Might want to rerun your sim at that limit, but your results will be pretty similar.

At the 3miniCIM -> 2 Falcon drive decision point, power is pretty equivalent. Power to weight ratio, power to volume ratio, power to component count ratio, and power to PDP slot ratios are not equivalent.

If you are a cash-constrained team who would normally reuse your drivetrain over several years for cost purposes and invest the savings in your manipulators, it’s difficult to make a compelling case for early-adopter investment in Falcons if it isn’t already a “new drive components” year.

3 Likes

I’ve realized there’s a mistake in my code. Working on it right now.
It’s been fixed.

4.4lbs < 12.96lbs

25 Likes

I think this is going to end up being a significant oversight. The Mini CIM’s efficiency peaks at 57%; the Falcon 500’s peaks at just about 90%. Both motors are covering the same distance in about the same time, so they’re putting out about the same mechanical power. But that 70% decrease in inefficiency is going to mean the Falcon 500’s see significantly less voltage sag, so they can maintain a system voltage a lot closer nominal.

3 Likes

I haven’t looked at your code yet, but the curves look about the right shape anyway. AriMB is right though, the voltage drop at 400A is going to be huge (past brownout), so you should probably model that (you can assume a system internal resistance, including the battery, of say 0.03 ohms)

For a sanity check, I recommend you compare your mini Cim curve to other drive train simulations (search CD), such as @JesseK’s: https://github.com/flybotix/drivetrainsim

1 Like

I’d be interested to see a 4 (full size) CIM drivetrain added to that chart just for comparison, a lot of teams (including mine) have been using that basically forever.

1 Like

To add to the weight savings, you can also utilize cheaper/smaller gear boxes when only 2 motors per side are needed.

We never jumped to a 3 Mini-CIM drivetrain before, we weren’t ready to try other drivetrains beyond the KOP chassis. If we wanted to add more power to the drive we now have an interesting option. Previously we would have used all the Mini-CIMs and controllers we collected over time and would have to spend on new gearboxes. Now we could buy a set of NEOs or Falcon 500s and just feed them through a standard Toughbox which we have lots of. That’s an appealing option for teams looking to upgrade their kit chassis vs buying 3 motor transmissions. It doesn’t help that you can’t buy the Mini-CIM with the EVO pinion already pressed on, which might be difficult for a number of kit bot teams to do on their own.

3 Likes

The data shows that there isn’t an earth shattering performance difference between the motor options.

However, if you are a new team or a team that hasn’t made an investment in an electrical ecosystem, I think there’s a good argument that you can’t beat the Falcon. When you compare the prices of a NEO with a high resolution encoder and speed controller, a Minicim with a good speed controller (that can do closed loop / motion control) and an encoder, and a Falcon, the Falcon’s high points stand out. Simplified electrical connections, less space, better performance, and a good encoder.

Taken individually the features on the Falcon might not make it a ‘must have’ product for me. Taken as a whole this is what I want to use as my next motor/motor controller.

1 Like

Depends on where you are as a team, imo — I would much rather a team invest into infrastructural tooling such as a drill press or mini lathe or good hand tools or batteries before they bought into a slightly more powerful drivetrain than the kitbot.

Additionally, if a team chooses to buy new motors every year (which admittedly many don’t!), then the Falcons will be a slightly more expensive option over several years.

You do not need a high resolution encoder to run closed loop control with a NEO.

Agree with your other points.

3 Likes

The electrical plant math of this simulation doesn’t work in the real word.

The 400A collective motor draw is theoretically possible.

If you have a nearly perfect battery with 0.020 Ohms of resistance the battery terminal voltage drops to 4V at that load.

Said battery is now producing 3200W for the duration of that test. Between two and three corded hairdryers for comparison. All in a confined space. The Lead/Acid battery will get warm very quickly.

The RIO will reset when V <4.3V

Also, the main breaker trips at ~4 seconds in this condition, which is significantly faster than the individual 40A breakers will trip at 250% of rating.

2 Likes

I agree with this point to an extent. It should be said that the more complex your closed loop control becomes, the more you begin to need a higher resolution encoder. For example, I’ve heard that when running velocity control on a wheel ball shooter, in order to recover the velocity of your wheel quickly and accurately after it is slowed down from launching a ball, you need an encoder with a decently high CPR*.

I’m not disagreeing with the statement that you can run closed loop control with the Neo’s internal encoders (we used Neos with PID position control on our elevator this year and it worked just fine). But I am saying that there comes a point at which using a low resolution encoder in closed loop control does affect performance. It really depends on what you are trying to accomplish.

*At least if you want to shoot balls back to back to back to back like some teams did in 2017.

1 Like

It’s actually a bit worse than that, since the weight of the Falcons includes the built-in motor controller. To get the equivalent with the six miniCIMs, you would need six TalonSRXs at .26lbs each. So the actual ratio is 4.4lbs < 14.52lbs. And that doesn’t include any weight (small though it is) for encoders on the miniCIMs, even though that too is included in the Falcons’ weight.

1 Like

Our team has been using this tool to select motor / gearbox combinations for a couple of years. This has been updated to include the new Venom and Falcon options.
https://www.chiefdelphi.com/t/frc-1018-robot-speed-simulation-2019/354716/4

If you download, make sure you select the 191106 version.

Battery is simulated as a fixed Voltage with a series resistance - 12.5 Volts was chosen as a nominal case.

Velocity, Voltage, and Current traces are plotted from the simulation.
Motor and gearbox selections are pull-down menus - if you change gearboxes, make sure that both gearbox selections are changed.
If, while experimenting, a purple trace shows up in the velocity plot, the robot wheels are estimated to be slipping on the drive surface. Suggest reducing the current limit until the purple trace goes away.

1 Like

Last year we burned out 3 NEOs, and lost 2 rounds at champs due to CAN wires slipping out of the SPARK controller. These Falcons would solve the CAN wire slip issue, but not sure about burnout. Cooling vents probably too passive, but perhaps forced air at pits between matches might work. As other noted, $40 for a replacement NEO is way better than $140. Have to say, having only to worry about mounting the motor, versus both controller and motor is a big plus, since the wiring is direct and easier to shield uninterrupted from RIO to motor.

It’s worth noting that as compared to a NEO, the Falcon will have relatively significant reductions in input power pulled at matched torque output. The difference between input power and mechanical output power is turned into heat, thus the Falcons will run cooler (and likely with less failures) than the NEOs. This will be most significant for mechanisms that you are operating at stall (arms and elevators) but it will matter for drivetrains to some extent too.

If you want to know more, I did some calculations in this discussion on a different thread: