ADAPTIVE TIME OPTIMAL CONTROL OF A DISK DRIVE VOICE COIL MOTOR ACTUATOR

Abstract

An Adaptive Time-Optimal Control (ATOC) scheme is developed for the hard disk drive servomechanism. The velocity profile of the proposed controller is adjusted based on the actual plant parameters, which are implicitly identified by measuring the position overshoot of two pre-determined seeks. The overshoot information is fed into a neural network regulator to determine necessary adjustment on the Acceleration Discount Factor (ADP). The algorithm works for any combinatorial variation of the individual plant parameters within specified statistical limits. Simulations of the ATOC and a fixed parameter Proximate Time-Optimal Servomechanism (PTOS) controller show up to 20% reduction in access time for the former, depending on the degree of plant parameter variations and the length of the seek. One notable advantage of the proposed scheme is s that it could be easily implemented within the existing disk drive Digital Signal Processor (DSP). The scheme could also be used to calibrate the servo system upon initial drive evaluation and at subsequent intervals as and when required. Control adaptation is achieved through the tuning of a single controller parameter, the ADF. The neural network is small enough that it could be implemented in DSP software. The proposed adaptation steps are designed to reduce the dependency on ADF to compensate for parameter uncertainties of the plant/controller model. In the event of other common uncertainties such as unmodeled dynamics and discrete time implementation, the optimal ADF can be suitably discounted to absorb their effects, much like in PTOS. This improves the seek times with robustness comparable to a PTOS scheme. A test stand has been built for experiments on various disk drive servo control schemes. It includes a disk drive Voice Coil Motor (VCM), a current amplifier, and a laser interferometry based position feedback system. The head position signal is obtained from a retro-reflector attached to the actuator. The servo loop is similar to that of a servo-writer. However, as servo writers do not operate in seek mode, the focus here is on one track step response. Linear strategies, such as proportional plus derivative (PD) controllers, are designed to obtain the desired step response, including overshoot and settling time, etc. The closed-loop system is identified in the frequency domain through a Fourier analyzer. The obtained frequency responses are used both as a guideline in controller tuning and system performance verifying.