Advanced Controller Designs for Head Positioning and Tension Regulation in Tape Drive Systems

Hua Zhong, University of Colorado at Boulder


High density digital tape systems provide a cost-efficient and reliable solution for massive data backup. To maintain a competitive position against other data storage technologies, the tape industry aims to enable higher track densities and shorter data access time. The servomechanisms of the tape transport system and the head positioning system need to be improved for these purposes. This thesis investigates combined feedforward and feedback control in the tape transport dynamics and the head actuator, respectively. As the tape winds from one reel to the other, the reel radii vary due to the winding of the tape and the reel runouts. These variations can cause tension errors. As the tape is very thin, the variations between two consecutive revolutions can be considered negligible. A feedforward-inspired control scheme is developed to take into account the time-varying radius to reduce tape tension error caused by the variations in radii. In order to better align the write/read head with the desired data track, a combined feedforward/feedback control scheme to improve the tape head positioning servo system in the presence of lateral tape motion is also developed. Approximate dynamic inversion techniques and model matching methods are investigated to design the feedforward controller. The feedforward control is computed based upon an estimate of the lateral tape motion displacements (LTMDs) at the head that is calculated using measurements of the upstream and downstream lateral tape motion displacements near the head. To find the correlation between successive LTMDs, linear and polynomial regression methods are first studied. The nonlinear partial differential equation modeling tape motion in the lateral direction is then explored to solve the displacements for a short segment of tape surrounding the head using measurements of the upstream and downstream displacements as boundary conditions. The equation is discretized in the spatial domain using the finite element method (FEM) and integrated using the Wilson theta approximation. Performance and effectiveness of the controllers are demonstrated in simulations. Feasibility analysis of implementing the feedforward controller on an actual tape drive is also conducted.