Some fundamental questions about InstaSPIN FOC

1. Yes, absolutely, one of the key features of the algorithm. We call this "Flying Start".  We have demonstrated this multiple times on different motors/applications (fans, washing machines, starter/generators, etc).  Here is a post that discusses the high level capability

http://e2e.ti.com/support/microcontrollers/c2000/f/902/t/253662.aspx

2. Yes, you will have to design this for your application though.  FAST can track the rotor within the normal boundaries - which is motor/inverter/sense dependent but typically ~1Hz - so you will have to add some system logic to understand what speed you are at and the associated behavior.  For example, in a boundary where FAST is tracking well you will make sure the ForceAngle option is DISABLED, and you will need to set your speed controller so it is synchronized in to the current speed.  Another option that actually works better is to take over in torque mode with an inital command of Iq = 0A, and hold this 0A command for a certain amount of time (will have to test with your application), at which you switch to speed mode with a new reference (either the current speed or your new target speed).  This will guarantee a synchronized start.

You will also want to make sure that your current controller is stiffened for this application, so instead of using the default Kp values (which are the bandwidth *0.25 StiffnessFactor you want to use the full bandwidth, or a *1.0 factor). This is because when the inverter is first enabled, a 50% duty cycle is at the output, which can generate current flowing, so the current controller needs to get that back to zero. 

This is the how this is done using the gMotorVars interface in all of the MotorWare projects

      CTRL_setKp(handle,CTRL_Type_PID_Id,gMotorVars.Kp_Idq);
      CTRL_setKp(handle,CTRL_Type_PID_Iq,gMotorVars.Kp_Idq);

Of you can set them individually in case you are using different Ls_d and Ls_q in your user.h [this also shows you the calcualtions made for the Kp gain of the Iq/Id controllers: StiffnessFactor * Ls * Current / Ts * Voltage

  _iq Kp_Id = _IQ((1.0*Ls_d*fullScaleCurrent)/(ctrlPeriod_sec*fullScaleVoltage));

  _iq Kp_Iq = _IQ((1.0*Ls_q*fullScaleCurrent)/(ctrlPeriod_sec*fullScaleVoltage));

 3. Absolutely. FAST will be tracking and locked on to the signal (assuming the rotor speed isn't so slow that the feedback is corrupted, in this case enable ForceAngle) so you just need to manage the powering of the inverter and making sure any PI controllers aren't producing an immediate output that will cause an unwanted response in the system. You are immediately able to drive a torque command to the motor, so in your case you may have some logic that if the speed is over a certain threshold you will set a new target speed with certain acceleration (deceleration) to drive the motor as fast as possible to a new speed.  In other cases you may want to brake the motor by turning on all high side or low side switches.

4. Yes, it is excellent.

5. 2 or 3 current measurements has nothing to do with the flying start capability.  You decide 2 or 3 in your system based on the maximum bus voltage utiliization. If you want to use over modulation (beyond SVPWM to full Trapezoidal BLDC) you must have 3 current measurements (3 shunts or 2 phase and calculate the third). With only 2 shunts you must limit your maximum duty cycle/modulation to insure you always have valid sampling windowns on the two lower legs. This is also discussed on this forum in several places.

Here are some plots at some of the boundary conditions.

Using DRV8312 + included motor, running at 4K RPM to start, turning off the inverter, and then telling the application to catch the rotor at 2 Hz (15 RPM) and keep it running at 2 Hz.

Here is similar with TMDSHVMTRINSPIN usiagain catching and driving at 2 Hz

And here is using TMDSHVMTRINSPIN with a direct drive Washing Machine, catching at 0.1 Hz (0.25 RPM) and then driving to 1 Hz speed reference right away so it’s not stuck below 1 Hz (2.5 RPM). It is repeatable and it never goes backwards.  For this motor, control at 1 Hz (2.5 RPM) is not so good, but this is ok as a transitional speed and keeps the estimator stable.  In this application you typically never need to motor below 10-20 Hz.

Yes, lab4 in MotorWare shows you how to do this, it's quite simple.

    // Dis-able the Library internal PI.  Iq has no reference now
    CTRL_setFlag_enableSpeedCtrl(ctrlHandle, false);

  // get the speed estimate
  gMotorVars.Speed_krpm = EST_getSpeed_krpm(obj->estHandle);

  // set the speed reference so that the forced angle rotates in the correct direction
  // for startup.
  if(_IQabs(gMotorVars.Speed_krpm) < _IQ(0.01))
  {
      if(iq_ref < 0)
          CTRL_setSpd_ref_krpm(handle,_IQ(-0.01));
      else if(iq_ref > 0)
          CTRL_setSpd_ref_krpm(handle,_IQ(0.01));
  }

  // Set the Iq reference that use to come out of the PI speed control
  CTRL_setIq_ref_pu(handle, iq_ref);