Poll of the Day > So who here has a good understanding of DC motors...

(+) Yeah! (+)

This is the hardest challenge I could come up with for myself.

I have a pipe dream of creating an animatronic rodent that looks and moves like a real animal to the point of being completely indistinguishable. I've gone through many different kinds of mechanisms, such as hydraulics and electroactive polymers, and have come to the understanding that DC motors are king.

I am very experienced with electronics/programming and will design the circuitry and motor drivers however they need to be, down to the machine learning model. I have found a quirk in DC motors that may allow me to unconventionally lower the torque of the motor on purpose through custom drivers to allow for more natural movement.

This is my issue, I need the most efficient motor to insert into the joint, of, say, the leg of a rat, and match it 1:1. I need it to be as small as possible. I have limited experience with engineering, and have found the smallest motors have very high RPM but low torque. Why is this? I need the RPM to match my target, but it seems they make these motors for drones with as high RPM as possible. Gearboxes are inefficient, noisy, and consume space. Is it the only way?

I should probably take this to r/robotics.

I think you might be off base... many animals have rather robotic-jerky motions in the first place, particularly birds.

I kinda wanted to make a gearbox bird, but I have no industrial know-how

Lokarin posted...

I think you might be off base... many animals have rather robotic-jerky motions in the first place, particularly birds.

I kinda wanted to make a gearbox bird, but I have no industrial know-how

Hence lowering the voltage per motor via custom motor drivers, reducing torque and jerky movement.

The issue is that "4 limb" animals have around 600 muscles, and that doesn't matter what animal it is. Say we cut that number in half since a single motor can be used in reverse. This means that a proper replication will have 300 motors. Say I magically come up with driver circuitry that can handle that many motors without going overweight, I currently have no mechanism to tie that to requirements for the RPM/torque needed per joint, no knowledge of what kind of motors I even need.

So maybe I need to focus strictly on intel. Perhaps I should build this out in a 3D engine first, and utilize ML to optimize the requirements of each motor. I ask it to learn to walk around with a perfectly accurate skeleton and refine the minimum requirements.

Alternatively, I learn the anatomy of the rodent so well that I can calculate the requirements of each joint solely based on muscle weight and cell type ratio. I still need that data in dataset form.

Currently working on building out a quadroped ragdoll in Godot... I did this in Unity previously, but the truth is guys, Godot was better than Unity just enough for devs to make the jump. The truth is that Unity would not have died if users weren't sick of their shit up until this point anyway.

Unity's C# support is still veeery bad and it limited what I could do with the language, just because the engine was poorly designed. I couldn't get machine learning, I couldn't get QUIC sockets. Also, yes, engineers actually use game engines to help design robots.

I'm going to use an evolutionary algorithm in tandem with machine learning where appropriate to approximate the necessary requirements for each motor. This will also serve to create a ML model to walk around. In the future I will use this framework to help design the circuitry, using the same algorithms, which will require many motor drivers.

Sup, EE here.
I'm a bit lost with what your problem is and what your proposed solution would be. What exact motor are you using and what's your exact Voltage and Amperage? In the datasheet you normally have a chart you can refer to for the RPM and torque.

Additionally, that's not a quirk that you're using. That's just how DC motors work.

Oh, the reason why robotics tends to use stepper motors instead of DC motors is they don't have any internal feedback, so you're way more likely to burn out your motor when, say, the motor is caught. It's fine if you're in a completely controlled system, but if you're dealing with real machines out in real life you're going to cause issues.

You could always create a feedback loop, but you should look up control loop programs and the math associated with it if you're serious about it.

shadowsword87 posted...

Sup, EE here.
I'm a bit lost with what your problem is and what your proposed solution would be. What exact motor are you using and what's your exact Voltage and Amperage? In the datasheet you normally have a chart you can refer to for the RPM and torque.

Hey, this is me just using PotD as my rubber duck. I have come up with some solution, but it's from the perspective of a software engineer. Not your perfect engineer math. Mine is just to use evolutionary algos. Very complex and arduous.

My goal is that I want to replicate a rat. I don't have the physics background to match, say, the joint of a leg of a rat, to a motor that can match the torque and speed 1:1.

If I could come across some way to do the math, that would make this process faster. But that means I need someone to have actually documented how powerful a rat's leg is, so I guess this crosses over into biology.

Oh, ok so this is just theoretical then. You don't have a breadboard setup or a raspberry pi or ESP32 or something physical.

For the record, engineering doesn't do perfect math. We have different models for different levels of complexity with different levels of abstraction. A wire heating up from electricity flowing through it can be modeled a half dozen different ways depending on what you're doing.

shadowsword87 posted...

Oh, ok so this is just theoretical then. You don't have a breadboard setup or a raspberry pi or ESP32 or something physical.

No, I imagine my final setup will be 3 custom ordered PCBs with many custom motor controllers (I got the CM4 specifically for this). I got into CAD to throw together a joint, but realized I have to do a lot of info gathering before I can even start.

shadowsword87 posted...

Oh, the reason why robotics tends to use stepper motors instead of DC motors is they don't have any internal feedback, so you're way more likely to burn out your motor when, say, the motor is caught. It's fine if you're in a completely controlled system, but if you're dealing with real machines out in real life you're going to cause issues.

You could always create a feedback loop, but you should look up control loop programs and the math associated with it if you're serious about it.

I would prefer to come up with a solution that intentionally isn't 100% efficient so that it matches the organic movement of an animal. I think steppers and servos are rigid and noisy, on top of being done to death, and it wouldn't be very interesting to me. I generally find most robotics projects kind of unimpressive for those reasons.

When designing my PCBs, I will make sure I take into account burning out my motors, as well as reading feedback, so that I can gauge how much pressure is being put on my motors and feed that back to my CPU. That's much appreciated.

Yellow posted...

No, I imagine my final setup will be 3 custom ordered PCBs with many custom motor controllers (I got the CM4 specifically for this). I got into CAD to throw together a joint, but realized I have to do a lot of info gathering before I can even start.

Uh, this thing?
https://www.raspberrypi.com/products/compute-module-4/?variant=raspberry-pi-cm4001000
Best of luck to you I suppose.

Yellow posted...

When designing my PCBs, I will make sure I take into account burning out my motors, as well as reading feedback, so that I can gauge how much pressure is being put on my motors and feed that back to my CPU. That's much appreciated.

Also remember that without an absolute positioning on the motor, you're going to need to be dead reckoning the exact position. With slippage especially you're going to just drift what your micro thinks the position is, and what's physically there. Little errors that build up over time until your saftey measures take over.
That's the reason why people use steppe motors, going with what people have already used lets you grow from them. You'll basically have to reinvent the wheel.

shadowsword87 posted...

Also remember that without an absolute positioning on the motor, you're going to need to be dead reckoning the exact position. With slippage especially you're going to just drift what your micro thinks the position is, and what's physically there. Little errors that build up over time until your saftey measures take over.
That's the reason why people use steppe motors, going with what people have already used lets you grow from them. You'll basically have to reinvent the wheel.

Oh no... I don't like hearing that.

I'll think it over a bit before responding. But while I am, what if I just took a servo motor, removed to gearbox, kept the measuring mechanism, and just created a custom driver?

I've heard someone jumping from one field to another while disrespecting the field they're jumping to is a common thing.

Maybe that could work, I just work with motors I don't know the exact details of them. That seems real hard though.

Yellow posted...

I've heard someone jumping from one field to another while disrespecting the field they're jumping to is a common thing.

Yeah, it's very easy to look over and say something isn't hard, it's just following the datasheet. Until they actually sit down and build boards and see things working badly. Making something bad is easy, making it good is hard.

this probably already exists for like $5 on temu

your second paragraph should have been your topic title wtf

So my biggest hangup here is how small DC motors (apparently?) trade off torque for speed at smaller sizes.

I look at small motors with higher torque and they all have gearboxes. Gearboxes have some drawbacks that I consider dealbreakers. (noisy, jerky, bigger)

I (naively) assume there is no such thing as a high torque DC motor smaller than a centimeter without a gearbox, and I assume this is just a limitation of physics. That, or there's no market for such motors.

I also maintain that I do not need the convenience of the internal controller board of a stepper/servo motor, and am capable of making my own. I don't really care about what's easy, I have no time limit to learn or build.

I say all this looking at my tiny 30 mm micromotor that can spin a fan fast enough to fly, but will stop moving if I touch the blade. I'm a software engineer, not an engineer.

That's a physical requirement, yes.
Machines and electronics aren't magic, as nice as it would be. We're still limited by physics, or by price.

If you start using rare earth magnets and gold wire filaments I'm sure you could squeeze a few extra footpounds out of it, but you would skyrocket the price.