First Post on a AC Servo Drive Project + Current Loop Simulation

Introduction

Over the past year I have been working on a motor driver for an AC hobby motor (total time about a month due to school and work) - I decided to document my past developments and continue the project as I am less busy. I looked back and refined what I wanted out of this motor drive.

Requirements
  • Have precise control of position, speed, and torque of an AC motor
  • No torque ripple in output
  • Controller should be suitable for a wide range of PMSM motors
Goals
  • Small PCB footprint (~50 mm squared)
  • Develop software to tune controller profiles and gains
Constraints
  • Design drive for permanent magnet synchronous machines (PMSM) with surface mounted magnets (non-salient motor)
  • Use field oriented control (FOC) to modulate phase currents
  • End board should be <$100 CAD
To clarify, these PMSMs are hobby grade motors used for drones/UAVs, however the goal is to make it work with any sinusoidally wound brushless motor.

Controlling Current

To drive an AC motor, the three phase currents must be applied in a controlled fashion. There are several ways: one way is 6-step commutation, which is good for high speed applications however it produces a noticeable torque ripple. Another is sinusoidal commutation which gets rid of the torque ripple, however performs worse at high speeds. The solution chose (and mentioned in the constraints) is to implement Field Oriented Control (FOC) as it mitigates all these issues - maximum torque with no ripple, and precise control over a wide range of speeds. 

I won't explian FOC in-depth here, however here were some resources that helped me understand it.
  1. Measure incoming phase currents and rotor position
  2. Transform these three phase currents into two phase currents using the Park Transform
  3. Transform the two phase currents to a stationary reference frame using the Clarke Transform
  4. Now the currents can be viewed as DC variables - apply a PI controller to regulate these two currents ($i_q$ and $i_d$ for quadrature and direct axes respectively).
  5. These PI controllers produce a correction voltage, which is still in the DC frame, so use the Inverse Clark, then Inverse Park transforms to turn the correction voltages to AC
  6. Using PWM, modulate these signals onto the legs of a three phase inverter to spin the motor.

FOC Simulation - Current Loop

I simulated the current loop in Simulink (view the files here on Guthib). I am using a default PMSM motor model however I as simulating using a Teknic-2310P for motor constants (i.e. resistance and flux values). I will go more in-depth later on tuning the PI regulators, however there is a deterministic way of picking them (view Dave Wilson's blog on TI).

Here is the top level model:




Here is the step response for the quadrature axis current for a 1 A step:


This response is pretty high bandwidth which is needed to follow the cascaded speed and position loops to come. Also the three phase currents at steady state are all sinusoidal shifted by 120 degrees, which is expected.





Comments

Post a Comment

Popular posts from this blog

AS5048A Magnetic Encoder and STM32

Image
This past week I began to write the controller firmware - mainly playing with the AS5048A angular position sensor from AMS. Here is the datasheet .  This sensor will be used to get the rotor position and then compute the resulting electrical angle for the Clarke/Park transforms and also it will be used to find rotor speed for the out loop.  I will be using an STM32F446 board to run the loop as it has a high clock speed. The AS5048A sends its measurements over SPI. Below is the read package sent from the sensor: We get 14 bits of position data - aka 16384 ticks per revolution, or a resolution of 0.0219 degrees. I wrote this read() function that sends the SPI command package and reads the resulting response: uint16_t read(uint16_t registerAddress) { uint16_t command = 0x4000; // PAR=0 R/W=R uint8_t RxBuff[2]; uint8_t TxBuff[2]; command = command | registerAddress; command |= (uint16_t) (spiCalcEvenParity(command) << 0xF); TxBuff[0] = command >> 8; TxBuff[1] = co
Read more

Pseudo-FOC Controller - Simulations and Inital

Image
As a part of my FOC controller, I am developing an intermediate project - the  Pseudo-FOC Controller. This post will introduce the theory, simulations, and an initial prototype of the project. At its core, this controller uses sinusoidal commutation, which is being implemented through the inverse Park and Clarke transforms. Project Goals The goals of this project are listed below: Control torque, speed, and position of a PMSM - specifically for gimbal and drone grade motors Develop a PCB for the controller Develop auto-tuning algorithms for controller gains Develop an auto-calibration sequence for the rotor position sensor BLDC vs PMSM A key point of confusion is motor classification of permanent magnet machines - specifically between Brushless DC Motors  (BLDCs) and Permanent Magnet Synchronous Motors (PMSMs). In my experience/readings, the differing factors is in their back-emf profile: BLDCs having a trapezoidal BEMF and PMSMs having a sinusoidal BEMF. It is also important to note t
Read more

Revised CNC Mill Design

Image
In the last post I fixed some of the clearance and parts issues for the x-axis assembly for the CNC mill. I since applied these same changes to all axes and dialed in most of the components. Here is a image of the full system:  Version 2 of the CNC mill design. The entire assembly weighs 48 kg which is on the lighter side. To fix this I plan on casting epoxy granite around the y-axis base to add both mass and damping properties.  The work area is: X travel: 240 mm Y travel: 180 mm Z travel: 110 mm Parts Each of the axes use SFU1204 ball screws, the only difference being the length: 500 mm on x, 400 mm on y, and 300 mm on z. Similar to this, all axes use HGR25 linear rails (a pair) and 4 HGH25-CA linear bearings (2 per rail). Again the only difference is the lengths: 500 mm on x, 400 mm on y, and 300 mm on z. For the spindle I plan on using a 1.5 kW 110 V air cooled spindle and a VFD ( found here ). I modelled the dimensions in Solidworks (as seen above). The motors are just NEMA 27s, h
Read more