Options for Brushless Motor Control?

[Ether]

Originally Posted by r691175002

Hi, what kind of options are available for closed loop control of brushless motors in the 60A range.

The only controller I am aware of is by roboteq:
http://www.roboteq.com/brushless-dc-...er-hall-inputs

I have used their brushed controllers in the past with some success but their controllers are a bit large (form-factor wise) for my application. I've also tried looking into brushless servo controllers but there seems to be almost nothing compact.

In the past I have used PID control on standard hobby ESCs but that can hardly be called a solution and a lot of problems arise when trying to reverse or accurately control low speeds.

Questions:

- three phase brushless?

- 6-step or sinusoidal commutation?

- Hall, encoder, or resolver ?

- do you need the commutation done by the hardware or do you have enough processor power to do it in software?

- you said 60A but didn't mention the voltage. Are you talking 12V here?

[Tristan Lall]

Check out the speed controllers used for model aircraft motors. There are numerous manufacturers catering to that market. In particular, the F3A electric aircraft use very substantial speed controllers for their motors.

I've used the Schulze future 32.55K in a UAV application, but don't have much to compare it to in terms of making a recommendation. I wouldn't characterize the reliability as fantastic (one of three failed and had to be sent for repair), but the setup at 40 V and about 60 A (maximum power) did work well enough. It's pretty compact (about 3 in × 1.25 in × 0.5 in), but Schulze also makes many smaller speed controllers, for smaller loads.

The other catch is that the closed-loop control is implemented inside of the speed controller—you'll have no influence over it, except maybe to set the firmware mode via DIP switches. The motors are sensorless (no encoders), so the speed controller is constantly monitoring the three leads to determine the timing of the motor. (This seemed to cause rough starts, but once spinning, everything was smooth.)

[r691175002]

I hope to use hobby motors because they are cheap and powerful for their weight. This model in particular would be something I am considering:
http://www.modelmotors.cz/index.php?...e=14&line=GOLD

The motor would be three phase. I don't have a real need for either 6-step or sinusoidal but low speed control is a concern so sinusoidal would be nice. There would be an encoder on the motor.

I will be running the output through a gearbox and each wheel will be independent. There is a chance I will be using mecanum wheels so precision on the level of servo drivers would be ideal.

There will be enough processing power to do it in software although I'd like to avoid having to write a controller since it seems rather complicated.

The voltage is likely 25.2V but I'm not sure if we will be using lipo or life batteries.

There are a lot of great boards for hobby and rc planes, but none of them can really accomplish what I need.

[Ether]

Originally Posted by r691175002

I hope to use hobby motors because they are cheap and powerful for their weight. This model in particular would be something I am considering:
http://www.modelmotors.cz/index.php?...e=14&line=GOLD

The motor would be three phase. I don't have a real need for either 6-step or sinusoidal but low speed control is a concern so sinusoidal would be nice. There would be an encoder on the motor.

I see only 3 wires coming out of that motor. Where's the encoder?

I will be running the output through a gearbox and each wheel will be independent. There is a chance I will be using mecanum wheels so precision on the level of servo drivers would be ideal.

Sensorless motors are fine for propellers on hobby RC planes, but not so much for your app (a ground vehicle) which requires high start-up torque and acceleration.

[r691175002]

The plan was to add an incremental encoder to the motor shaft or at some point in the gearbox but upon further consideration a sensored motor would likely be more than accurate enough.

I would be likely adding my own sensor regardless of the type used.

[Ether]

Originally Posted by r691175002

The plan was to add an incremental encoder to the motor shaft or at some point in the gearbox but upon further consideration a sensored motor would likely be more than accurate enough.

The sensor in the motor is used for commutation. You can't commutate a motor with a sensor located on a geared-down gearbox shaft.

I would be likely adding my own sensor regardless of the type used.

If the motor has a sensor you don't need an external sensor. You can know vehicle position and velocity by counting motor commutations (unless it's all done in hardware and hidden from you).

[r691175002]

I was envisioning a controller that used a rotary encoder to keep track of the motor shaft - just like a brushless servo.
Regardless of the sensor or how it is mounted, as long as there is a way to get an absolute or relative (combined with some kind of start-up sequence) reading on the motor shaft commutation is possible. If I mounted the encoder in the gearbox backlash could be a problem but that's a separate issue.

There are a few products that can accomplish what I want but they are invariably giant expensive boxes intended for CNC or similar applications:
http://www.tolomatic.com/products/it...cfm?tree_id=92
The tolomatic box requires only an incremental encoder.

The more I search the more it looks like the best option is building it ourselves but every time I hear terms like flux vector I feel like it is a job better left to experts.

Of course, given the limited options I seem to have I'm willing to go with whatever I can get.

[Ether]

Backlash, torsional windup, and sensor resolution make a gearbox-mounted sensor a very poor choice for commutation. You want the commutation sensor mounted internally to sense the shaft position near the rotor for best results.

Trapezoidal commutation is not that hard to code if you are using motors with Hall commutation sensors. Interrupt on change of Hall states and flip the appropriate FETs in the 3-half-bridge driver circuit. Just make sure the rest of your code is written properly so that you have very low software latency.

[AdamHeard]

If you want cheap cheap, I reccommend okhobby and hobbyking.

Reliability varies product to product, but I have a lot from both running in a lot of my vehicles.

Originally Posted by Ether

I see only 3 wires coming out of that motor. Where's the encoder?



Sensorless motors are fine for propellers on hobby RC planes, but not so much for your app (a ground vehicle) which requires high start-up torque and acceleration.

Nearly all hobby grade ground vehicles use sensorless brushless motors versus sensored.

[Ether]

The OP hasn't given much detail about his application, but he did say 60 amps (post#1) at 25 volts (post#4) (times 4 for mecanum). That's 6000 watts total at stall.

I was assuming (perhaps erroneously) that he choose those specs because he needed high startup torque and acceleration at low speed. Sensorless brushless motors are poor performers under those conditions.

[r691175002]

Ether is correct that I will need some kind of feedback in this application. Most of the motion will be at very slow speeds or completely stopped and there will be a lot of reversals. The motors will also be fighting a lot of torque. Unfortunately most hobby controllers won't even let you reverse, so accurate positioning is out of the question.

There are a few reasons that I want to avoid off-the shelf sensored motors:
Because of their applications, they tend to be more expensive, heavier and harder to find than hobby motors. They are also mostly inrunners and I really want to avoid having to gear down 60k rpm four times.

I have no problem attaching an encoder to the back of a sensorless motor to get feedback.

I still find it weird that it is so hard to find something that will work. Closed loop brushed controllers practically grow on trees - brushless motors aren't that unusual...

[BornaE]

How about we go at this the right way.

Define your functional requirements,
Come up with ideas that meet those requirements,
Rank and select the best one.

My main question is why brushless when brushed motors are so easy to obtain with a very reasonable price?

-Borna

Edit:

-Requirements-

Robot pushing force:
Maximum robot Speed:
Voltage:

[Ether]

Originally Posted by BornaE

How about we go at this the right way.

Define your functional requirements,
Come up with ideas that meet those requirements,
Rank and select the best one.

You're assuming he hasn't already done this. Perhaps you're right... perhaps you're wrong. It's always good to ask though.

[r691175002]

Robot must carry ~15LB between two waypoints several kilometers apart without GPS. Terrain is generic grass/rocks/dirt/sand whatever is in the way. The lighter the robot is the better.

I don't consider brushed motors an option. For example, this 3000W motor from NPC:
http://www.npcrobotics.com/products/...at=20&mode=gfx
weighs 15 pounds. A brushless motor of the same power would weigh around a pound.


We still don't know exactly what the robot will look like and not much planning has been done yet, I'm just looking at some options. I'd like to avoid ackerman steering because some of the manoeuvres will need to be precise but it depends on what the software guys want. Crab drive/meccanum wheels and skid steering are also all on the table.

Given how light brushless motors can be it is my opinion that it is easier and lighter to have a separate motor for each powered wheel since there will be some form of suspension.

Hoping for a weight of ~40KG and a max speed of 4m/s.

[Ether]

Originally Posted by r691175002

Robot must carry ~15LB between two waypoints several kilometers apart without GPS. Terrain is generic grass/rocks/dirt/sand whatever is in the way.

Are you saying you want the robot to travel several kilometers over rocks and sand autonomously, with nothing but dead reckoning to keep it on track?

I don't consider brushed motors an option. For example, this 3000W motor from NPC:
http://www.npcrobotics.com/products/...at=20&mode=gfx
weighs 15 pounds. A brushless motor of the same power would weigh around a pound.

With a similar torque vs speed curve? Can you post a link please...

I'd like to avoid ackerman steering because some of the manoeuvres will need to be precise but it depends on what the software guys want. Crab drive/meccanum wheels and skid steering are also all on the table.

You're not going to get precise maneuvers on rocks and sand with mecanum wheels.