Hardware Reference
In-Depth Information
4.4.2 Detection of Rotor Position in Sensorless Drive
It has been mentioned earlier that knowing the rotor position is an essential
requirement for commutation of current and generation of effective torque in
PMAC motors (both BLDC motors and PMSMs). The electronic commuta-
tion means generating proper commutation sequence at the correct position of
the rotor to switch the power electronic devices of the inverter bridge. Physical
sensors such as Hall-effect sensors, resolvers and position encoders can be used
to obtain the rotor position. However, it was explained earlier in section 4.3.1
that the use of physical sensor is not a suitable choice for the HDD application
due to limitation of space, cost restriction, and system reliability. As a conse-
quence, the PMAC motor drive methods that do not need accurate physical
position sensors are receiving wide attention. These methods are commonly
referred to as position sensorless control method, or self sensing method.
The sensorless control methods for PMAC motors presented in different
published articles can be broadly categorized into two groups following the
classi
fi
cation of PMAC motors: (1) the sensorless methods for BLDC motors
and (2) the sensorless methods for PMSMs. The difference between these two
types of methods lies in the way the phases of a motor are energized. In
the sensorless BLDC control methods, only two phases are energized at any
time instant [12], and only a few rotor position signals are required. On the
contrary, all three phases conduct at the same time in the sensorless PMSM
control methods, and a continuous position signals is required [49]. Different
methods can be used to
fi
nd the rotor position of PMAC motors when a drive
control scheme without a position sensor is to be used [104]. A brief summary
of these methods is presented next.
4.4.2.1 Review of Methods for Sensorless Detection of Rotor Position
Some of the sensorless methods used to detect rotor position are,
1. based on voltage and current measurement [49],
2. hypothetical d-q model [139],
3. stochastic
fi
ltering [20],
4. self observers [180],
5. detecting variations in inductance [117], and
6. nonconventional methods such as use of neural network or fuzzy logic or
both [177].
These methods are generally complex and computation intensive, demand-
ing powerful digital signal processors (DSPs) for the motor control system.
