Can any brushless outrunner motor be used as a gimbal stabilization motor?

From looking at the construction of the gimbal motors, they do not appear to be much different than a standard brushless motor. However, they are able to be controlled like servo's, and I do not understand how. Does anyone have experience at these? I do not understand how a motor ment to spin fast with lowish torque, (high KV, as some may call that, or think of that) can be used to precisely control position, other than downwearing and feedback control.

Here is a video showing what I mean. Notice how slow that motor turns without gears? and I know they must have a lot of torque to be stabilizing a hefty camera.

sort by: active | newest | oldest
Wired_Mist2 years ago

21 replies, Lol

"I just do not know how they are used like servo's, to hold a certain position".

You would need to power one to two coils, and the magnetic draw will be what holds the motor still. If just one coil is energized then (Obviously) the motor will hold position over the activated coil. Energize two at once and it will Hold position directly in between the two coils.

For more Effective Positioning you can Energize 2 coils and balance the current delivered to each coil.

IE (Three coil brush-less motor, like a Computer Fan) 100% duty on coil A will center it on what i'm going to call point A,

50% duty on coil A and 50% duty on coil B will hold the motor between points A + B.

100% on coil A, 50% on coil B will hold the motor between point A / B but will be closer to point A

I believe that you need an H-Bridge, per coil, to drive these like how you described. but you may be able to devise a way to do it with some extra transistors and diodes. (Let me get back to you on that design; It's getting a bit Late for drawing up Schematics :P)

-max- (author)  Wired_Mist2 years ago

I do have a understanding of the basic functional working of a ESC, which is like 3 totem pole drivers on each phase, and the FETs are driven with the appropiate voltages to the gate from something like a microcontroller.

The basic problem is as you slowly turn you will need to supply power to 2/3 coils and ground the third. As the motor turns and you need to energize the third coil, one of the other two (Depending on what direction you want to turn) will have to be grounded. So you can either use 3X H-Bridges and set each motor contact high or low as needed. OR You may be able to use 3X FET's to energize each coil as needed and a second NPN Fet on each contact that can ground the coil. ( Sorry I'm used to 8 Wire Universal Steppers, Much more user friendly !)

You will also need to tell the position of the Motor. I would use a Rotary encoder but you may be able to use around 3 hall-effect sensors to tell what the position of the motor is. 3X hall-effect vs the single sensor on a CPU fan because you need a more Accurate reading then if the motor is turning or not.

Out or curiosity...

Why Can't you use Servo's? Seems like a much simpler solution for something like a Go-Pro. I assume you are using this on the quadra copter you mentioned on a diff post. The Cam should be light enough that servors could work Nicely :)

There may even be a Kit Available somewhere !

Servos don't make fluid enough movements as gimbal motors and are jittery.

Wow, someone's Cruising through the Archives ! lmao

Yeah, I think you're probably right. Think some gearing might help?

Would still make a nice test rig to learn the basics I suppose :P

Hahahaha the post are pretty old.

Lower gearing would help, but it is gears in general that don't allow you to switch directions fast and further cause the jitters. If you were making a head tracker for a plane you'd probably be just fine to use servos as you are not switching directions much like with stabilization.

-max- (author)  Wired_Mist2 years ago

They have dead band space, or something, and are jittery. I am using a mobius camera, which is smaller/lighter, and have tried servo's. Currently, the servos connect directly to the flight FC, and when driven by it, (I think with 10 bit software resolution, from atmega328p) you can actually see the 'steps' the servo's make. Also, they rotate a bit slow, and corrections occur only at 50Hz. (I am able to sometimes drive them at 160Hz, and get smoother operation from them, but if there is any mechanical stress or shock load to them, they will start to parasitically oscillate and lose all the performance and holding ability, as they just buzz and carry on.) Getting nicer, digital servo's may improve it, but probably only marginally.

Here is a video demonstrating a severe version of the issue. Luckily I have not had that much jitter, but you get the idea

They are also not that fast for quick corrections, as I realised. That may just be because they are not good quality servo's, I got them for like less than $5 each.

Currently I have bigger issues than getting a nice stable picture, as my quad needs LOTS of PID tuning to dampen out oscillations, and gain control over it better.

At $5 a piece you could but better ones.

Btw... That Vid looks nice as is! unless you plan to do some real filming from it I'd almost stick with what you have. Some new servo's may be all you need :)

-max- (author)  Wired_Mist2 years ago

I suppose so. I certainly am not going to be investing in a gopro anytime soon, and nearly all the $60-$160 gimbals are designed for that specifically, usually for the phantom. Does anyone know of some good digital servo's?

Also, can I get more resolution out of a arduino? I would think anything less that 12 bits is a bit low. I may end up having ditch the flight controller's software 10 bit PWM to the servo's, and if I use a separate ATmega chip entirely, then is it possible to get 12 bit PWM out of it, hardware, software, or otherwise?

I can't help you source servo's any better then google, but yes you can get 12 bit with a TLC5940.


12bit Res and 16 outputs! Even has a nice Library for Arduino :)

-max- (author)  Wired_Mist2 years ago

I do have a ton of these puppies: Burr Brown DAC70BH DACs

They are driven in the most fundamental way: literally 16 input pins, and one output, and the rest of the pins are supporting circuitry! It is SOOO 80's


-max- (author)  Wired_Mist2 years ago

I suppose so. I certainly am not going to be investing in a gopro anytime soon, and nearly all the $60-$160 gimbals are designed for that specifically, usually for the phantom. Does anyone know of some good digital servo's?

Also, can I get more resolution out of a arduino? I would think anything less that 12 bits is a bit low. I may end up having ditch the flight controller's software 10 bit PWM to the servo's, and if I use a separate ATmega chip entirely, then is it possible to get 12 bit PWM out of it, hardware, software, or otherwise?

-max- (author)  Wired_Mist2 years ago

"You would need to power one to two coils, and the magnetic draw will be what holds the motor still. If just one coil is energized then (Obviously) the motor will hold position over the activated coil. Energize two at once and it will Hold position directly in between the two coils."

That makes sense, but the reason I think there is more to it is that it needs to be able to hold it's position at any angle, or amount of rotation, and somehow overcome 'magnetic cogging' or whatever those steps are called. Holding using just this method without feedback would probably not work too well unless current was really high. I suppose it would work.

[the rest of that comment]

OK then, so the motor can literally be driven with really slow, as low as 0Hz 3 phase power to hold precise position. I did not want to think that at first because it seems to be really wastefull method, when the freq. is not high enough to allow the effects of inductive reactance limit current, and instead, rely on parasitic wasteful resistance do the current limiting.

If they are actually using a brush-less motors that's the only way I can see it happening. Otherwise they are using High Res Stepper motors or maybe servos.

I have a few from photocopiers here that have a 1.8* step. using the half step method (that I believe Steve mentioned somewhere) means the ones I have would need 400 steps to complete a rotation. Evenn without gearing that fairly smoothe as is.

rickharris2 years ago

AFAIK the brushless motor is effectively a stepper motor. Pulsing the coil creates a rotating magnetic field which drags the magnets in the rotor around. In general use this happens very quickly to give the high speed you want.

If you make those pulses very slow then you get precision stepping. In this case the acellerometer output is being use to step he motor 1 step at a time.

Not entirly true I'm afraid.

A stepper motor uses fixed positions and you can only use full or half steps for the positioning.

They also need quite a high impulse current to work properly.

A brushless motor like used for this project can be set freely to any position by adjusting the magnetic field between two coil pairs.

The great advantage over a stepper motor is the smooth movement.

Where a stepper motor at very low speeds tends to "jump" a brushless motor will still make a smooth movement.

Downside is the more complicated driver for the brushless motor and that the holding power on a fixed postion is far less.

A stepper motor contains stationary coils, often 2 but 4 or more may be used.

The rotor of the stepper motor is a permanent magnet.

All this is in a brushless motor.

However I will agree that the brushless motor has many more coils and magnets because of the nature of the use. They are all 3 phase motors the coils are linked into 3 groups 1 for each phase.

This doesn't detract from the essential operating mechanism which is the same as a stepper motor a rotating magnetic field is used to move a permanent magnet attached to the output shaft.

Because of the higher frequency of the magnets in the brushless motor it can make much smaller steps. I see little difference practically between the stepper motor and the brushless motor.

The 2 images below show a Brtushless motor and a stepper motor.


See a skewed rotor but the stator may be also skewed to avoid cogging !


Induction motor. Brushless motors are not skewed.

Please trust me, about an early 3ph brushless regenerative machine we made at Lear Motors using a skewed deep bar rotor in 1980.

The brushless motors the OP is talking about are made for model air craft use and are a somewhat different construction.

Yes but while I no longer get email from ibles, anything is open for commenting !

Lmao !

Yeah I have like 7 diff windows open in an effort to follow Quetions >.<

And asking to follow questions is kaput also, while Kiteman is getting regular e-mail all this time.

Slowly moving to http://www.jigidi.com/ for simple fun !

-max- (author)  rickharris2 years ago

I do understand the similarities of the brushless motors to stepper motors, however, there is a reason why stepper motors are not used for camera stabilization. Because the movements they make are 'stepped.'

Instead, it needs to be a small, slow rotation, with strongish holding torue to the camera. You can see in that video link above what I mean. See how slowly, yet smoothly it can turn? Then add a heavy camera and mount to the shaft of it for it to turn in all directions, and you can see what I mean.

The brushless motor has many magnets which give it a higher resolution than a standard stepper. The slow - smooth motion can be created by applying as Steve says sinusoidal waveforms to the coils.

-max- (author)  rickharris2 years ago

Even though all those 'coils' are wired as a 3 pole system, either as a delta or wye config?

OK then, so what I said way down on this question VVV really is the case! You can apply literally any frequency (including 0Hz) to the 3 phases and position the motor exactly where you want it? Because at those really low frequencies, for slow, steady movement, doesn't the coils internally saturate and then all the power used by the motor is eventually due to just the resistance? I do know it can be very difficult to turn the mototr shaft if the the 3 phases are shorted togther.


Not entirly true I'm afraid.

A stepper motor uses fixed positions and you can only use full or half steps for the positioning.

No, because you can microstep, giving a sinusoidal and co-sinusoidal excitation to the motor.

-max- (author)  rickharris2 years ago

I originally considered that, however, from my experience, such a 'step' would be really large and not precise, especially for a motor with such few poles. each 'step' is like 1/8th of a turn at least. That of course does not matter when rotating 1000's of RPM, but for small, really finely controlled things like camera stabilization, that would be horrifically bad.

As you can see in the video, the motor does not just jump from one position to another, it very slowly and elegantly rotates from point A to B. Something I would have imagined only possible after appropriate gearing.

rickharris2 years ago
iceng2 years ago

Keep in mind that motor manufacturers incorporate slot skewing up to one slot pitch to reduce shaft cogging torque in both PM and brushless machines.

-max- (author)  iceng2 years ago

Is that why the shaft on some larger induction motors looks skewed? Like all the short-circuit coils are at a weird angle? If not, then what do you mean by slot skewing?


-max- (author)  iceng2 years ago

OK I just found your pictures a bit higher up. I took apart one of the cheap motors I have, and it did not seem to be built like that, where I would assume the outrunner magnets are skewed. There are all parallel to the shaft.

Fundamentally, DC motors are TORQUE machines, not POSITION machines, whether they are brushed or brushless. They not rely for their action on slip induced currents, as in an induction motor, but in classic Fleming law reaction between a current and a field.

-max- (author)  steveastrouk2 years ago

"Fundamentally, DC motors are TORQUE machines, not POSITION machines"

I do know that, and that is why I asked the question to begin with: How can they be directly used in these 3 axis gimbals [link below]? Apparently the torque they produce strong enough to keep a hefty camera stable w/o and down-gearing, and presume that there is some sort of feedback control that takes place, probably a control loop from a gyro/accelerometer sensor mounted on the same bit as is the camera, streaming tilt/pan/roll data to a microcontroller, which controls the motors indirectly though some motor drivers. What I find weird, is that there is no gears, as you would find as in the most basic simple servo motor.


(my previous comment was really addressed to the other posters here, who aren't seeing the distinction between BLDC and stepper action)

If the system were balanced, you would only be moving the inertial mass, not the static mass, and for modest excursions, modest motors would work.

The nice thing about the motors you've shown is they are well cooled, if not very efficient magnetically.

The torque from the motor is proportional to its diameter^2 (actually, IIRC, its BND^2/L , or BNdersquerdl as my electric machines lecturer used as his mnemonic 30 off years ago. B is the flux in the gap, N is the number of turns on the coils, D is diameter and L is the length of the machine. The motors we;re looking at look fairly wide, so you'd get quite a bit of torque out of them. The key has to be for how long you are demanding high torques.

-max- (author)  steveastrouk2 years ago

That equation make a lot of sense, considering the equation for a cylinder volume. (see the similarities?)

That formula does not seem to take into account the current through the coils, and electrical-side of things. Is it a formula that was derived experimentally for the exact motors you speak of?

Its a highly generic proportionality for electric machines in general. Clearly you take it and multiply it by the current, but its only a proportionality, to give you an idea of scaling.

-max- (author)  steveastrouk2 years ago

Ahh, OK. So there is a multiplied K constant in there somewhere. It does not seem to take the look of a function though, what's on the other side of the '∝' or '≈'? Is it just torque? Such that T∝(BND²)/L

Its dodgy as hell, but the basic proportionalites are all in there,

T∝(BND²)/L only remembering a bit more its

T∝BND²L - since a longer motor isn't going to REDUCE torque.

Nice formatting BTW

-max- (author)  steveastrouk2 years ago

I was wondering why a a larger L would have been inversely proportional to torque, especially considering those small hobby motors used for micro quads are so small, but very long compared to the diameter. That makes a lot more sense.

I am sure that formula will not get me much further than an rough approximation, due to all the things that have not been considered, like efficiency, friction, timing of the brushes for brushed motors, and stuff like that. I figure more in-depth calculations will require high level physics and vector calculus and stuff like that. Totally not worth it compared to testing/experimentation which always yeilds accurate results!.

-max- (author)  steveastrouk2 years ago

That is 20 million too many variables that I don't know, considering I have not bought any of those generic 3 axis gimbals yet!

What you say about the load makes sense, I suppose the size (diameter) of these brushless outrunners allow higher torques than I give them credit for! (after all, it is quite hard to really slow down the crap A2212 motors from hong kong that I have when running full speed!)

(The reason for the original question was to see the feasibility of making the gimbal from scratch, and I needed a good understanding of the principles at work to make sure I get the right stuff. I thought that those motors may be somehow wired directly to the multiwii FCfor control, similar to servo's, and some sort of black magic occurred under the bells of the gimbal motors. however, like many other things, it will probably end up costing >= than just getting one ready-made, and of course there is risk of it not working anyway due to my lack of knowledge and building skills!)

Have look how the motors for these selv levelling ride on toys work, or to be precise how they are driven.

There are some Segway like instructables too.

These brushless motors are like three phase motors.

With the right addressing and good current / voltage control you can get the rotor in any position you like and with some good holding power too.

Maybe the Martinez controller is a good start for you?


Or have a look at Basecam:


Google is your friend, why not ask him more often? ;)

-max- (author)  Downunder35m2 years ago

If I am going to get 3 axis gimbal, I am definitely not designing it from scratch, as that proved difficult even for mounting a basic servo motor to control pitch for a mobius camera! I ended up using hot glue, which I do not like. I could not figure out how else to mount things together since the RC hobby stuff never has proper mounting hardware holes or anything. the screw holes on the servo do not line up nicely with the mobius case, and I do not have small enough screws to extend the size of the servo arm.

Making a 3 axis brushless gimbal without a 3D printer or other machining equipment (other than a metalworking lathe, saw/ basic woodworking stuff and dremel/hand tools) would be a near-impossible challenge which isn't worth it.

They are just your standard brush-less motor. It's all in the motion control board being used in conjunction with the motor controller. Based on the motion sensors position it's pulsing the motor forward and backwards to give you the slight movements. But it's by no means accurate and will easily overrun a desired possition.

-max- (author)  mpilchfamily2 years ago

What about the holding torque of the motor, even with feedback control? (even in servo's, there is a potentiometer which serves as feedback) How are they able to get away without using any gearing, to increase toruqe, and decrease RPM, and jitteriness from the motor, and the fact that is is not precise?

-max- (author) 2 years ago

After further research, I found that at least in terms of construction, they are in fact the same to 3 phase brushless motors, except that the wire is considerably thinner, and many more turns are used, which increase resistance by a lot, I guess so that the voltage from a really low frequency drive do not cause excessive current draw. (because obviously, the really low resistance of standard windings at really low speeds/frequencies, cannot rely on the inductive reactance of the coils to limit current, as would be the case of standard brushless outrunners.)

-max- (author) 2 years ago

I do understand that brushless motors are effectively 3 phase motors, I just do not know how they are used like servo's, to hold a certain position.

Are they literally being driven such that the coils are saturated with current from a really low frequency 3 phase power, whereas the 3 phases can slow down to a complete freeze for holding the motor stationary, and perhaps feedback control to prevent slip. (like if there is external torque applied to the motor, the ESC will begin to cycle the 3 phase power in the opposing direction to take into account the 'slip' and counteract that force?)

I understand how slip works in a induction 3 PH motor, but does the same principle apply to these permanent magnet motors? Rather than slipping, I would think that they 'skip' instead of 'slip,' as too much opposing force on the motor running on a fixed frequency drive from the ESC, and the magnets on the outrunner may eventually fail to keep up with the rotating magnetic field produced by the coils. However, thats just a guess, since the ESCs that I have are 'smart', and dynamically speed up or slow down the brushless motors based on back EMF, so that they effectively perform like brushed motors, with a similar linear(ish) torque/RPM curve. That is until they are completly stopped, then they just twitch and jerk, as the ESC keeps trying to get them going, then quits, and tries again periodicly.