MAIN MENUFront MatterTechnical ProgramAuthor IndexSearch CD-ROMSearch ResultsPrintView Full PageZoom InZoom OutGo To Previous DocumentCD-ROM HelpDesign and Analysis of a Dual-Stage Disk Drive Servo SystemUsing an Instrumented SuspensionXinghui Huang, Ryozo Nagamune, Roberto Horowitz and Yunfeng LiAbstract—This paper presents the design and analysis of atrack-following controller using a mixed-objective optimizationtechnique for dual-stage servo systems in hard disk drives(HDD). The objective of minimizing tracking error in thepresence of plant uncertainties and operational variations isformulated into a framework of multi-objective minimization.The tracking error minimization is reasonably formulated asan H2norm minimization problem, while the robust stabilityissue is addressed by some H∞norm bounds. These normminimization or constraints are then translated into a set ofparametric feasibility conditions using linear matrix inequal-ities (LMI) which are readily solved by convex optimizationsolvers. To enhance tracking performance and stability ro-bustness, attenuation of airflow excited suspension structuralvibration is also explicitly taken into consideration by aninner-loop fast-rate damping and compensation controllerutilizing a vibration sensor on the suspension surface. Analysisand simulation results show a noticeable improvement intracking performance over a previous designs, while retainingacceptable robust stability under certain multiplicative andparameter uncertainties.I. INTRODUCTIONDual-stage actuation, which combines a normal voicecoil motor (VCM) actuator and a secondary microactu-ator (MA) placed close to the head, has been studiedintensively as a means of achieving higher track densities,hence higher data capacity in HDDs, by increasing theservo bandwidth. The design and optimization of track-following controllers have been studied by many researchersover the past years. These works vary from decoupled orsequential single-input-single-output (SISO) classical fre-quency shaping design techniques, such as the master-slavemethod [1], the PQ method [2], and the sensitivity transferfunction decoupling method [3], to those methodologiesthat explicitly account for the coupling effects betweenthe VCM and MA actuators and that utilize multivariableoptimal control design techniques, such as LQG/LTR [4], µ-synthesis [5]. Most of those works optimize the performanceby modelling the system in detail and the informationof various sources of disturbances is also brought intoconsideration. A problem related to detailed parametricmodelling is stability robustness. Plant uncertainties orvariations may deteriorate the claimed performance throughThis work was supported by the Information Storage Industry Consor-tium (INSIC) and the Computer Mechanics Laboratory (CML) of U.C.Berkeley.X. Huang, R. Nagamune and R. Horowitz are with the Departmentof Mechanical Engineering, U.C. Berkeley, CA 94720 {xhhuang,ryozo, horowitz}@me.berkeley.edu.Y. Li is with the Maxtor Corporation, 500 McCarthy Blvd, Milpitas,CA 95035, yunfeng [email protected] parametric modelling. In the worst case, the systemmay be unstable. A disk servo controller should performwell to meet performance specifications over a huge batchof production drives while providing internal stability underplant uncertainties and operational variations from driveto drive. The µ-synthesis technique incorporates stabilityrobustness in design explicitly through properly modelleduncertainty dynamics or estimated parametric uncertainties.Other design methodologies can only consider robustnessimplicitly when optimizing system performance.In this paper, we discuss an optimization method forthe dual-stage track-following controller design in harddisk drives. This design methodology formulates multi-ple objectives as a problem of some norm optimizationor norm constraints that can be expressed as a set ofLMIs, which are then solved through convex optimization.Quantitative information on the track runout spectrum andon the windage to suspension structural vibrations, andalso plant uncertainties are accounted for explicitly in thisdesign. In the context of disk drive servo design, this designmethodology was first applied in [6] for a single-stagesystem. The design presented in this paper is for a dual-stage servo system with a secondary MEMS MA locatedbetween the suspension tip and the slider. Realistic modelsfor the VCM, MA, windage, and track runout, are obtainedfrom either experimental tests or finite element analysis.Furthermore, airflow excited structural vibration attenuationis explicitly accounted for by an inner-loop damping andcompensation controller, utilizing a strain sensor signalfrom the suspension surface. This inner controller can be runat a higher rate than that of the basic servo loop since, unlikethe position error signal (PES), this strain signal does nothave such a physical limitation on its sampling rate. Somedistinct issues on robust stability arising from the dual-stageconfiguration are circumvented properly. Meanwhile, animproved algorithm of norm characterization and controllerparametrization has been employed in this design whichexhibits less conservativeness and is expected to yield betteroverall performance [7] over the original algorithm [8].This paper is organized as follows: Section 2 describesthe problem formulation and controller design of the dual-stage servo system. Simulation results and analyses arepresented in Section 3. Section 4 concludes this paper.II. MULTI-OBJECTIVE TRACK-FOLLOWING CONTROLDESIGNThe dual-stage servo system consists of two actuators: amain VCM actuator, and a secondary actuator MEMS MA,Proceeding of the 2004 American Control ConferenceBoston, Massachusetts June 30 - July 2, 20040-7803-8335-4/04/$17.00 ©2004 AACCWeA16.5535which is sandwiched between the gimbal and the slider.It can generate translational motion of the slider relativeto the suspension tip and therefore provide the potentialfor achieving higher servo bandwidth. The magnitude fre-quency responses of the VCM and MEMS MA are shown inFig. 1. With proper design and assembling, the MEMS MAhas a moderately damped resonance mode around 2 kHzand no other appreciable structural resonance modes up to40 kHz. Therefore, it can be modelled as a simple mass-spring-damper second-order system. Capacitive sensing canbe incorporated into the MA such that the relative motionoutput of the MA, RP ES, can be measured. On the