Building an ESC

[ziggyfish]

I am working on a project where I need some ESC (electronic speed controllers)s. The project is a quad copter with 4 three-phase brushless motors.

The brains of the thing will be https://www.olimex.com/Products/SOM/A20/ , the reason for picking this board is for a lot of reasons, but namely I am have a degree with an AI major and would like to play with the algorithms I learned (such as multi-agent systems, neual networks, etc). I also want to learn as much as I can about electronics with this project. So as such I would like to build my own board which will have 4 ESCs as well as sensors (MPU9250, MS5611-01BA03 and Venus838FLPx, all of which use SPI).

I have two options on controlling the speed via the CPU, I can either use a DAC, or PWM. After watching one of the best explanations of what PWM on actually is on YouTube and considering all of my components will be using SPI. I plan on using a DAC to control speed (and its relatively cheaper than going with a PWM, because I would need 4 microcontrollers to maintain the PWM signal just for the motors). This also makes my circuit considerably simpler and modular as well.

So the next thing I need to do is convert that analog DC to three-phase AC to turn the motors. And this is where I am a bit stuck, the ON SEMICONDUCTOR STK5F1U3E2D-E does what I want (I think anyway), however its $95.

So my question is this, how do I convert  DC into three-phase AC, which can support 50A without spending $95?

[ziggyfish]

And just to add a little more information, I am thinking about using http://www.farnell.com/datasheets/1487306.pdf as my DAC.

Thanks

Brendan

[ziggyfish]

Thanks for your reply,

The goal is to build my own ESCs (to understand how they work internally). I am not sure why you say I still need to use PWM at some point, so will look into that. From what I understand the reason most ESCs use PWM is to convert a digital RC signal from the receiver into an analog signal or am I missing something here? Because my main board will handle the communication (will be via WIFI or some other means), I assume using a DAC would be a better choice, as I would get a better response time (8?s vs 125?s based on a 8khz PWM).

If I wanted a quick an easy way to build a quad copter I would just go out an buy one pre made, but then what's the point of the project? The idea of building my own ESCs is to learn the electronics part of software development which I have done for 10+ years.

Thanks

Brendan

[David Hess]

The motors will have such slow response compared to the electronics that the different between using a DAC and PWM for the control signal to the brushless motor driver is insignificant.

As far as the brushless motor driver itself, 50 amps is not trivial and is going to require a discrete output stage.

How were you planning on commutating the motors?  Do they have hall effect sensor outputs built in?

I would either use one of the various dedicated 3 phase brushless motor controller ASICs designed to operate with discrete output stages (something like an Allegro A4915) or program my own using the right microcontroller to do the same thing.  With a microcontroller, control signals and feedback could be via SPI or some other serial link which has the advantage that noise from the motor controller can be isolated with optical coupling.

The motor winding are driven with a PWM signal to raise efficiency in the driver while producing a proportional output.  This is especially important in high power designs where the waste heat from a linear output stage can be considerable.

[ziggyfish]

As far as the brushless motor driver itself, 50 amps is not trivial and is going to require a discrete output stage.

How were you planning on commutating the motors?  Do they have hall effect sensor outputs built in?

I know 50 amps isn't trivial, that's the next step, after I workout getting the motors to work. I have a few ideas of how to do it.

I was hoping on using the DAC as a threshold voltage to a circuit that would produce the 3 waves needed to turn the motors.

I would either use one of the various dedicated 3 phase brushless motor controller ASICs designed to operate with discrete output stages (something like an Allegro A4915) or program my own using the right microcontroller to do the same thing.  With a microcontroller, control signals and feedback could be via SPI or some other serial link which has the advantage that noise from the motor controller can be isolated with optical coupling.

I will look into that A4915, however I see its a DC 3 phase brushless motor controller. The motors I want to use are AC, so there could be a problem with that. I was looking at a circuit called a 3 phase inverter (I noticed that the $95 IC uses a version of the same circuit), which looks like it can do what I want and it allows me to control it with my DAC. But I am not 100% sure on that yet. I've attached a diagram, so you can see how I see it working. The boxes named "Magically create sine waves" is what I am trying to solve.

The motor winding are driven with a PWM signal to raise efficiency in the driver while producing a proportional output.  This is especially important in high power designs where the waste heat from a linear output stage can be considerable.

From what I understand real ESCs use fairly big caps to smooth out the PWM signal, which means it ends up being the same signal as the one generated from my DAC without the need for the caps.

I will also look into linear output stages.

[macboy]

The motor winding are driven with a PWM signal to raise efficiency in the driver while producing a proportional output.  This is especially important in high power designs where the waste heat from a linear output stage can be considerable.

From what I understand real ESCs use fairly big caps to smooth out the PWM signal, which means it ends up being the same signal as the one generated from my DAC without the need for the caps.

I will also look into linear output stages.

I'm afraid that you don't understand at all.

With brushed motors, ESCs use PWM to control the speed. The motor speed is controlled by the PWM duty cycle. More on-time = faster. Why PWM and not some analog DAC? Because PWM results in driving the motor alternately full-on and full-off, which results in very low losses in the components (MOSFETs) controlling the current to the motor. If you drive the motor with an analog signal (such as buffered/amplified output from a DAC) then you will have very high losses and a lot of heat in generated. It is not practical with these big motors.

Brushless motors are completely different. Controlling them isn't so simple as applying power, as it is with brushed motors. I can almost guarantee that you will not be successful in designing your own high power brushless speed controller.

Your best bet is to use four commercially available brushless ESCs. Then to control them, you will need to send the ESC a standard RC servo control signal, which itself is just a simple PWM signal. You need to generate a pulse with an on-time of 1.0 ms (full 'left') to 2.0 ms (full 'right'), repeating at roughly 50 to 100 Hz (the rate doesn't matter, just pulse width). It is relatively simple to generate that type of signal from any microcontroller.

[Richard Crowley]

Yes, listen to macboy, he speaks the truth. You did not mention what the application is, but all those PWM and brushless, high-current motor controllers have been developed for the quad-copter application where weight and efficiency are extraordinarily important.  Using analog techniques, especially at high current is no longer practical no matter what the application is.  This is the 21st century and the digital age didn't happen by accident.

[nowlan]

I know little about ESC in general, but freescale has some 3phase chips, and app notes to look at.
edit: Motorhome

However, im not sure building esc is what you wanted. Just control, in which case it offload the trouble to commercial/hobby esc, and build brain box like other drones.

[ziggyfish]

nowlan, thanks for the link and I will read through it. The reason I want to build own ESC is to learn and understand about electronics. I could easily just go out and buy 4 ESCs and connect them up and get the whole thing running within hours, but I won't learn anything from it. The programming side of this project for me will be very easy. However to chalange myself I want to build an ESC.

macboy, you are right, I thought maybe I would loose torque which is why I was trying to get a sine wave (i.e going positive and negative), but on second thought it may not matter.

With brushed motors, ESCs use PWM to control the speed. The motor speed is controlled by the PWM duty cycle. More on-time = faster. Why PWM and not some analog DAC? Because PWM results in driving the motor alternately full-on and full-off, which results in very low losses in the components (MOSFETs) controlling the current to the motor. If you drive the motor with an analog signal (such as buffered/amplified output from a DAC) then you will have very high losses and a lot of heat in generated. It is not practical with these big motors.

But what I don't understand is this if I increase the duty cycle to 100%, then there is no 3 phases created. All 3 inputs into the motor will be on all the time, which won't turn the motor. So can you maybe explain how that works?

[KIKi]

It doesn't work. Actually you can't use 100% PWM because there won't be any "electronic" commutation necessary for proper work of a brushless DC motor. Sorry if I'm wrong

[ziggyfish]

It doesn't work. Actually you can't use 100% PWM because there won't be any "electronic" commutation necessary for proper work of a brushless DC motor. Sorry if I'm wrong

BTW its a brushless AC motor, not DC. Something like, http://www.hobbyking.com/hobbyking/store/uh_viewItem.asp?idProduct=16234 (its not the one I will end up using, but it will be the same type of motor).

Thanks

Brendan

[Precipice]

That's a sensorless motor, and quite a high power one.
Looking at this from the outside, you need to read a lot of app notes and reference designs - I think Microchip had a bunch of good ones. Search for sensorless BLDC. At the moment, you're asking questions that mean you're a long way from starting, let alone finishing, this.
I'll agree with the people suggesting you try something easier first - perhaps a driver for a smaller BLDC motor with hall effect sensors. There's still _plenty_ of meat on that project, and you'll be blowing up cheaper transistors, and probably less often. Most of the lessons you learn from doing a with-sensor motor will translate straight into doing the rather harder sensorless one.

Also - "The programming side of this project for me will be very easy." - you may well look back on this with amusement, when you're finished.

[KIKi]

It doesn't work. Actually you can't use 100% PWM because there won't be any "electronic" commutation necessary for proper work of a brushless DC motor. Sorry if I'm wrong

BTW its a brushless AC motor, not DC. Something like, http://www.hobbyking.com/hobbyking/store/uh_viewItem.asp?idProduct=16234 (its not the one I will end up using, but it will be the same type of motor).

Thanks

Brendan

There is no "pure" DC brushless  motor. In order to get any electric motor spinning it's necessary to have some kind of a variable electromagnetic field. In DC motors it's done with usage of brushes (commutator) that reverse current. Brushless motor in sequential order turns OFF and ON different windings (phases) on the stator and as a result we have variable field (that is why every DC brushless motor is AC). 

EDIT:

Examples from Atmel:

AVR430: ATAVRMC300 Hardware User Guide
AVR430: ATAVRMC300 Hardware User Guide - schematic

Hope it will help!

[ziggyfish]

That's a sensorless motor, and quite a high power one.
Looking at this from the outside, you need to read a lot of app notes and reference designs - I think Microchip had a bunch of good ones. Search for sensorless BLDC. At the moment, you're asking questions that mean you're a long way from starting, let alone finishing, this.
I'll agree with the people suggesting you try something easier first - perhaps a driver for a smaller BLDC motor with hall effect sensors. There's still _plenty_ of meat on that project, and you'll be blowing up cheaper transistors, and probably less often. Most of the lessons you learn from doing a with-sensor motor will translate straight into doing the rather harder sensorless one.

Yeah I agree, going with something simpler would help with understanding electronics, and would also learn about controlling DC motors as well. So good idea.

I estimate the project taking about 6-12 months to complete, so I know it won't be built tomorrow. The A20 comes tomorrow (FYI, I am really excited about). And there is a lot to do in that area, such as I need to rebuild the kernel so it uses the latest kernel, find or develop a few drivers for the other sensors, find a good solution to get camera to stream to something like an ipad or iphone (hoping to control it through an app I develop), so there is still a lot of stuff to do on this project before I get to the motors. But I am just trying to get enough information about the motors first, so that I can workout what components I need in order to get the whole thing working.

I did some searching just then, and I did find an 40A ESC schematic. It uses the ATMEGA8-16AU, as the micro-controller, and three IR2101 ( http://www.irf.com/product-info/datasheets/data/ir2101.pdf ), to control the 3 phase inverter circuit. Will see if I can find the link again, if someone wants it.

Also - "The programming side of this project for me will be very easy." - you may well look back on this with amusement, when you're finished.

I have previously built operating systems (from bootloader to GUI), Linux drivers, Windows drivers, and technical business applications. So yeah, the programming side will be fun.

[macboy]

nowlan, thanks for the link and I will read through it. The reason I want to build own ESC is to learn and understand about electronics. I could easily just go out and buy 4 ESCs and connect them up and get the whole thing running within hours, but I won't learn anything from it. The programming side of this project for me will be very easy. However to chalange myself I want to build an ESC.

macboy, you are right, I thought maybe I would loose torque which is why I was trying to get a sine wave (i.e going positive and negative), but on second thought it may not matter.

With brushed motors, ESCs use PWM to control the speed. The motor speed is controlled by the PWM duty cycle. More on-time = faster. Why PWM and not some analog DAC? Because PWM results in driving the motor alternately full-on and full-off, which results in very low losses in the components (MOSFETs) controlling the current to the motor. If you drive the motor with an analog signal (such as buffered/amplified output from a DAC) then you will have very high losses and a lot of heat in generated. It is not practical with these big motors.

But what I don't understand is this if I increase the duty cycle to 100%, then there is no 3 phases created. All 3 inputs into the motor will be on all the time, which won't turn the motor. So can you maybe explain how that works?

You need to read. A lot.
Right now you are at the point in the learning curve that you don't even realize how little you know. You don't know enough to even begin to ask the right questions. After a day or two of studying brushless motors, you will begin to understand how little you know today, and how much you still have to learn before you can come back here and start asking pointed questions.

[zapta]

Brushless motors are completely different. Controlling them isn't so simple as applying power, as it is with brushed motors. I can almost guarantee that you will not be successful in designing your own high power brushless speed controller.

Can you explain to the uninitiated how brushless ESC work? What kind of signal do they deliver to the motor, is it just a multi phase on/off that is phase sync'ed with the rotation? Do they sense the state of the motor in some way to sync the phase? How? 

Edit: apparently there are open source brushless ESC's   https://www.google.com/search?q=open+source+brushless+esc

[Kremmen]

BLDCs and their cousins PMSMs (Permanent Magnet Synchronous Motors) are both what the latter name says - synchronous AC motors.
In order to drive one successfully you must generate a rotating magnetic field in the stator which then reacts with the rotor field to induce torque in the the motor shaft. Once the shaft starts turning the rotor field orientation naturally starts to spin and the stator field must align with the rotation in such a manner as to maintain the desired torque.
In practice the stator field is generated in a 3 phase field coil array driven by 3 half bridges from a DC bus. To get the phasing right one either reads a set of (hall) sensors indicating rotor position or, in the sensorless case one measures the back emf induced into the momentarily undriven stator coil (at least 1 one the coils will always be undriven).
The minimum requirement in order to achieve commutation is to correctly identify the rotor field orientation to the nearest 60 degree sector. This is easiest accomplished with the sensor feedback doing just that. After that some kind of pwm scheme for coil current control will let you control torque, at least approximately.
A more advanced control scheme calls for properly orienting the stator field by what is called Field Oriented Control or vector control. The story is far too long to meaningfully explain here, but the idea is to compute from the 3 phase currents first the so called alpha and beta components by applying a Clarke transformation and then finally the stator direct and transverse field components using a Park transformation. Note that these are calculated in the rotating reference frame of the rotor making the calculation non-trivial.
Now to control the motor torque a suitable control scheme, such as classical PID, is applied to the d and q components. The controller has a setpoint value for d (this is 0 as a rule) and for q, the latter corresponding to the useful torque. The d and q controllers compute command variables for the direct and quadrature  current components which by inverse Park are transformed to alpha and beta commands and finally by inverse Clarke to the individual instantaneous phase currents.
The above calculations are done in hard real time as is obvious so there is a notable performance requirement. FPGAs are the tool of choice, or something like TI's TMS320 family of signal processors.
The above is the hard part. The actual speed control is business as usual and accomplished by nested control loops - outer speed loop and inner torque loop.

[ziggyfish]

You need to read. A lot.
Right now you are at the point in the learning curve that you don't even realize how little you know. You don't know enough to even begin to ask the right questions. After a day or two of studying brushless motors, you will begin to understand how little you know today, and how much you still have to learn before you can come back here and start asking pointed questions.

That's why I asked here, because I was stuck on something. Now I have some reading to do and I will figure the answer out myself (which is really all I want anyway), and will ask more questions after. That's how programmers work . It sounds like Kremmen knows a lot, so I will read up more on the terms he uses and ask a more directed question.

This week I have already learned about op amps, and 3 phase inverter circuits. So next will be what a classical PID, Clarke transformations, and a few other terms he uses.

[Kremmen]

OK, if you really want an in depth presentation into the art of sensorless motor control then i repeat the recommendation i have made earlier on this forum: read and understand James Mevey's master thesis on the subject; freely available at http://krex.k-state.edu/dspace/handle/2097/1507.
The ingress of the document reads as follows:
Abstract:
Working with the subject of sensorless motor control requires an understanding of several topical areas; this report presents an understanding that was gained during this research work. The fundamentals of electric motors (particularly brushless motors) are developed from first principles and the basic models are discussed. The theory of sinusoidal synchronous motors is reviewed (phasor analysis of the single phase equivalent circuit). The concept of a complex space vector is introduced and developed using a working knowledge of the sinusoidal synchronous motor. This leads to the presentation of the space vector model of the permanent magnet synchronous motor, in both the stationary and rotor reference frames. An overview of the operation of three-phase voltage source inverters is given, followed by an explanation of space vector modulation and its relationship to regular sinusoidal pulse width modulation. Torque control of the permanent magnet synchronous machine is reviewed in several reference frames and then rotor-flux-field-oriented-control is explained. Finally, some schemes for sensorless operation are discussed.

[ziggyfish]

So I have my 3 phase signal working the way I want it to. The next step is to scale up to 50A. The voltage output is at the correct.

The input signal is generated by the main board power supply (which is max 3A) throw a micro controller and other components. So my next question is how to I get the power from the battery (which doesn't have such limit) to power the motor.

Thanks

Brendan

[netdudeuk]

You could get the SimonK source code for some ideas.
 

[rob77]

So I have my 3 phase signal working the way I want it to. The next step is to scale up to 50A. The voltage output is at the correct.

The input signal is generated by the main board power supply (which is max 3A) throw a micro controller and other components. So my next question is how to I get the power from the battery (which doesn't have such limit) to power the motor.

Thanks

Brendan

simply buy a ready to go ESC for brushless motors. designing a 50Amp 3 phase bridge is far away from "having fun"... and actually it's nearly impossible for a newbie - you will burn more parts and equipment than the cost of the ESC

another point is the commutation of a BLDC motor - either you need hall sensors to know the rotor field's position or you need to go for a solution without hall sensors - monitoring the induced voltage on the "3rd" wire to get some feedback indicating the field's position. it's really not easy to drive a BLDC.

definitely nothing for newbies to design a BLDC ESC... trust me
another point is the cost - building a 50Amp ESC will be MUCH MORE EXPENSIVE than buying one

[ziggyfish]

So I have my 3 phase signal working the way I want it to. The next step is to scale up to 50A. The voltage output is at the correct.

The input signal is generated by the main board power supply (which is max 3A) throw a micro controller and other components. So my next question is how to I get the power from the battery (which doesn't have such limit) to power the motor.

Thanks

Brendan

simply buy a ready to go ESC for brushless motors. designing a 50Amp 3 phase bridge is far away from "having fun"... and actually it's nearly impossible for a newbie - you will burn more parts and equipment than the cost of the ESC

another point is the commutation of a BLDC motor - either you need hall sensors to know the rotor field's position or you need to go for a solution without hall sensors - monitoring the induced voltage on the "3rd" wire to get some feedback indicating the field's position. it's really not easy to drive a BLDC.

definitely nothing for newbies to design a BLDC ESC... trust me
another point is the cost - building a 50Amp ESC will be MUCH MORE EXPENSIVE than buying one

It depends on what you call fun. Over the past 2 months I have had so much fun learning about how electronics work, and how sub-circuits work. I also have been successful in building a circuit that can read the 3rd wire and power a small BLDC motor (I am a fast learner).

However I want to scale it up to the 50A. I know I can use a bunch of IRF7832s in parallel to support the 50A current however this can only be DC (from what I understand anyway). I wanted to know if there is another way which would allow true AC. An Op Amp does exactly what I am after, however the max amps is about 1A. Which means I would need a lot of them in parallel, which is not practical. Hence why I am asking this question.

[rob77]

However I want to scale it up to the 50A. I know I can use a bunch of IRF7832s in parallel to support the 50A current however this can only be DC (from what I understand anyway). I wanted to know if there is another way which would allow true AC. An Op Amp does exactly what I am after, however the max amps is about 1A. Which means I would need a lot of them in parallel, which is not practical. Hence why I am asking this question.

what you mean by true AC and how does an op-amp fit the picture ? what you need is a 3 phase bridge and feed it with 3 phase signal (you can PWM the 3 phases to simulate "true" AC)  generated by a micro-controller or FPGA or a dedicated BLDC controller chip.

[ziggyfish]

However I want to scale it up to the 50A. I know I can use a bunch of IRF7832s in parallel to support the 50A current however this can only be DC (from what I understand anyway). I wanted to know if there is another way which would allow true AC. An Op Amp does exactly what I am after, however the max amps is about 1A. Which means I would need a lot of them in parallel, which is not practical. Hence why I am asking this question.

what you mean by true AC and how does an op-amp fit the picture ? what you need is a 3 phase bridge and feed it with 3 phase signal (you can PWM the 3 phases to simulate "true" AC)  generated by a micro-controller or FPGA or a dedicated BLDC controller chip.

By true AC, I mean both positive and negative voltages. I have now been successful in creating a 3 phase inverter circuit and I can used IRF7832s in parallel to handle the 50 amps. However I would like to experiment with a circuit I created using op amps (it works in a simulator), which gives me perfect sine waves phase shifted 120 deg apart, but because op amps can't handle more than 1 A, it won't work without additional components (which BTW is what I am asking for).

how does op amps fit the picture? Because no current flows in or out of its inputs, there is a separation between the low current part of the circuit (powered by the A20) and the high (powered by the battery). Not only that op amps can also handle true AC which transistors can't.

But anyway, it looks like I need to ask on a different forum, because there seems to be a bad culture on this forum. A culture where you only get a responses like "actually it's nearly impossible for a newbie" and no actual answers which I can research further on to find the right solution. In programmers land, most forums never give you "your not good enough" answers (I actually don't know any that do).