I have a generic 5010 outrunnig drone motor. It was sold with zero documentation. Other then “360 Kv” which is printed on the side of the motor. I don’t know if this is accurat yet. And I can count 14 magnets for 7 pole pairs.
Simple question: What limits to use to prevent buring this up. But I’d also like to extract every bit of perfeormance I can from this motor
I’ll be using this in a robot with a 9:1 belt reduction so it will be holding torque at low RPM and will not have the benifit the free air-blast cooling a drone motor would usually get. BTW, it is working on open loop but not yet in closed loop FOC.
I took some measurements Lead Pair Resistance Inductance
red — black 0.24 Ohm 0.14 mH
black-Yellow 0.24 0.10
yellow — red 0.24 0.16 Average 0.24 0.133
If you’d got any documentation coming with the motor it would be useless anyway.
The best way to find out it’s limits is to add a thermistor between the windings and read the temp.
Those painted stators and the neodym magnets will be the weak spots here.
I wouldn’t go higher than 80°C.
If you feel adventurous you can do a torture test until the motor fails, while reading the temp.
If your MCU supports it you can also write a temp based torque-control loop
Motor’s have typical operating points, and you can specify the voltage and current they are expected to consume at that point.
However, if you are tampring with the drive electronics, you should think in terms of maximal waste heat it can dissipate. The amount of mechanical power it produces/it’s efficiecy has a large impact on this. This will change greatly depending on the motor drive style and operating conditions like rpm and torque regime etc.
One thing I’ve long wanted to try is to try to ascertain the resistance of the windings in order to ascertain their temperature. That would be pretty useful. You’d have to be able to measure the current through the motor, which is a common feature of motor drive, but also the back emf which is a less common feature in motor drives although it is a thing. From that and the drive voltage you could calculate winding resistance, which will go up with temperature, but it’s not by that much, for copper.
You can assume zero efficiency and just think of it like a stepper motor, if you want to do that I would just hook it up to an adjustable power supply, and use an infrared thermometer or FLIR camera if you can or something and just put it on low, then extrapolate from there to ascertain the power dissipation capability.
There must be some indication from other people/documentation, however the airflow will be very different and have a large impact too.
Oh, one idea is you could take a pico and have a separate system to measure the resistance of the windings using the ADC of the pico. That would be useful. Like, galvanically isolated from the motor driver.
Those are more like a motor kit than a finished product. Everything is well made except they only put about half as much copper on as you can fit if you rewind it by hand. RCTimer makes similar form factor with somewhat better winding fill, and has some listed stats (but as o_lampe says, they’re fairly useless).
To find the current limit, I just put it in torque mode, prevent the rotor from turning, and keep turning up the current until it gets uncomfortably hot. Then back off a bit. Voltage is basically as much as you want. Probably 12-24V.
So, no one is willing to guess at a starting point?
With 1/4 Ohm per phase, I think 2 volts means 8 amps and 16W.
I think I’m going to have to epoxy a thermistor to the stator and monitor it with software.
Yes, the motor could be better but these sell for $20 each and the project goal is to see what is the lowest cost per joint for a small quadruped robot. I think this is the cheapest motor that could possibly work.
It has similar winding resistance but is likely to have better thermals.
Cyberdog can do 6A continuous for 10mins, so perhaps your motor can do 3A continuous?
I thought the same but I guess they traded max. copperfill with best air ventilation
For a robot, I would add small fans to the motor enclosures, like the Arctos has.
I used these motors for my V1 quadruped, you can dig around in my github to find the parameters I used. Also, you can check out my blog to read more about my V1 quadruped build.
The recommendation for everyone trying closed loop for the first time is to use torque/ voltage mode, you will quickly learn the voltage limit (voltage where motor speed plateaus). Set this voltage as the limit and then try velocity mode.
You did the same thing I’m trying to do now. Basically a re-design of the ODRI “Solo” robot. The magor difference between yours and mine (other then you project being a year ahead of mine) is that I am using the native propellor mount on the motor to connect the first timing belt pully. This means I have the motor flipped around 180 degrees. I also have the motor included inside slightly bulkier leg segments. I think you might get better cooling with exposed motors but (1) I wanted to used the propeller mounts and (2) I want the robot to walk outdoors and was concerned about contamination from mud and water.
Also, I have already started looking at other motors with more poles and I’m rethinking gears vs pulleys.
I am going to study your web site and github in detail. I will likey build something slightly different. I am going to run the robot in simulation so I will have a good idea of the required torque and speed needed and then test one leg until I have about a 50% design margin for torque and speed.
I already have a quad powered by hoby servos. It walks but not well. I am re-thinking the motion of walking. My theory is that animals and people use constant (or zero) jerk motion planning and something like a time horizon cost based planning
I think the cost functions associated with the motion planning of a biological entity are quite complex and multi-faceted. You see this for example when you have a non-perfect specimen, e.g. injury or deformation. Then the motions take into account things like the pain associated with the motion, and the resulting trajectory can be quite different (non-optimal from the point of view of the perfect specimen).
Old individuals and young individuals, even when in perfect condition, may have quite different ideas of the optimal trajectory, as younger specimens are generally far more generous with their energy budget.
And certainly there are differences associated with the type of body pattern - for bipeds, balancing with the help of the upper body will be more important, while quadrupeds can shift their centre of gravity in ways that bipeds cannot. Horses can’t assume all the poses that cats can, etc…
So I think motion in biological specimens is a result of reinforcement learning neural network type “algorithms” and not as pre-designed as the way we move our motor based robots. I think this is most obvious as we observe baby organisms learning to move.
I’d like to see an AI-optimized quadruped robot who suddendly decides to jump on his rear-legs like a kangaroo
The bots from Boston Dynamics have already learned to trott, which does not only look more natural, but also reduces wear and energy consumption.
Way to go IMHO.
