Noise compensation device and method in a discrete time control system

[0002] The present invention refers to a noise compensation device and method in a discrete time control system.

[0003] It is known that high precision control devices frequently have, in addition to a feedback control system, closed loop systems for compensating external noise. These compensation systems use sensors detecting the presence of external noise and generating compensation signals that are added to the control signals used in the feedback loop.

[0004] In particular, discrete time control systems are required to synchronize the control signals and the compensation signals so that they may be correctly added.

[0005] An application of this compensation technique, to which reference will be made hereinafter, are hard disk read/write devices for controlling the position of reading heads.

[0006] As is known, R/W devices, referred to as hard disk drives (HDDs), normally comprise a set of magnetic disks, rotating all together, a head-actuator assembly, and an electronic control unit. The head-actuator assembly comprises a plurality of arms turning about a same rotation axis, integral with one another and actuated by a motor, and reading heads, each reading head being carried by a respective arm and facing respective surfaces of the disks.

[0007] Control information is stored in preset control sectors on the disks, is picked up by the heads and enables detection of the position of the heads with respect to the tracks on the magnetic disks. In particular, the heads generate an analog signal that is sampled at a preset rate (depending upon the rotation rate of the disk) to generate a numeric position signal. The electronic control unit detects a position error on the basis of the numeric position signal and generates a numeric control signal for controlling the head-actuator assembly and keeping the heads in optimal reading positions.

[0008] In addition, the control system comprises an open loop compensation line for compensating the effects of external disturbance. In particular, an acceleration sensor (for example, a sensor made using MicroElectroMechanicalSystem—MEMS—technology), mounted so as to be integral with the casing of the R/W device, detects any disturbance mechanical vibration and supplies an acceleration signal which is, in turn, sampled at a compensation rate and used by the electronic control unit to generate a numeric compensation signal to be added to the control signal.

[0009] The data supplied by the sensor and the control signal are not, however, synchronous, and thus cannot be immediately summed. To overcome this problem, synchronization techniques normally used carry out a sampling rate conversion.

[0010] In practice, the sequence of data supplied by the sensor is initially expanded by interposing, between two consecutive samples, a first preset integer number L of null samples corresponding to instants comprised within a same sampling interval. The expanded sequence of data is then filtered using a low pass filter to replace the null samples with interpolated samples. Next, a decimator reduces the expanded sequence of data, maintaining one sample every M samples (where M is a second preset integer number) and eliminating all the others. By selecting the numbers L and M so that the condition $F_c=\frac{L+1}{M}{F}_s$

[0011] is satisfied, wherein F_Cis the control sampling rate and F_Sis the sensor sampling rate, the sequence of data at the output of the decimator is synchronous with the control signal and can thus be used to generate the compensation signal.

[0012] Known devices, however, have a number of drawbacks. In fact, performing frequency conversions increases the phase delay of the compensation signal with respect to the control signal. This is particularly disadvantageous because, as is known to those skilled in the art, the phase delay is a critical parameter for the effectiveness of open loop compensation and must therefore be contained as far as possible. In addition, the frequency conversion is carried out by microprograms (firmware) which, on the one hand, require a physical encumbrance as they must be stored in a nonvolatile memory and, on the other hand, cause an increase in the required processing power.

[0013] A further drawback of known devices results from the high frequency noise (approximately 4-5 kHz) normally introduced during the noise measure and the acceleration signal generation. This high frequency noise must be reduced by filtering which, once again, is carried out by microprograms.

[0014] The present invention is embodied in a system and method for noise compensation in a discrete time control system. In an exemplary embodiment, the device comprises closed loop control means for generating a first timing signal, a signal indicative of the quantity to be controlled, and a control signal, which have a control frequency. The device further includes open loop control means for generating a compensation signal synchronous with the control signal and supplied to the closed loop control for correcting the control signal. The open loop control means comprises a sensing means for generating an analog signal correlated to a disturbance quantity, sampling means for receiving the analog signal and generating a sampled signal having a sample frequency correlated to the control frequency and a decimator stage for receiving the sample signal generating the compensation signal.

[0015] In an exemplary application, the noise compensating device may be used in a read/write (R/W) device, such as a disc drive and comprises a rotatable computer-readable media, a R/W head positioned in proximity with the computer-readable media to read data stored on the computer-readable media and write data to the computer-readable media, and an R/W device control circuit. The control circuit comprises a closed loop control circuit and an open loop control circuit. In one embodiment, the R/W device is a hard disc and the signal indicative of the quantity to be controlled is a track position and the disturbance quantity is a disturbance acceleration.

[0016] For a better understanding of the present invention, a preferred embodiment thereof is now described, purely as a non-limiting example, with reference to the attached drawings, wherein:

[0017] FIG. 1 is a simplified top plan view of a R/W system for a hard disk.

[0018] FIG. 2 is a side view of the device of FIG. 1, in a section taken along plane II—II.

[0019] FIG. 3 is a simplified block diagram of a noise compensating device according to the present invention.

[0020] FIG. 4 is a more detailed block diagram of one part of the diagram of FIG. 3.

[0021] FIG. 5 shows the plot of quantities taken in the device according to the invention.

[0022] FIG. 6 is a more detailed block diagram of one part of the diagram of FIG. 4.

[0023] In the embodiment described hereinafter, the invention is applied to a read/write (R/W) device for hard disks, even though the invention may be applied to a generic discrete time control system with open loop noise compensation.

[0024] With reference to FIGS. 1 and 2, a R/W device 1 for hard disks is housed inside a casing 2 and comprises a plurality of magnetic disks 7, a head-actuator assembly 4, and a control card 5.

[0025] The magnetic disks 7 are mounted coaxially and rotationally integral to a shaft 8 of a motor (not shown) and each have a top surface 7aand a bottom surface 7b. In addition, control information is stored, in a per se known manner, in preset sectors of the magnetic disks 7.

[0026] The head-actuator assembly 4 comprises a rotating support 13 (generally referred to as an E-block on account of its shape), a plurality of suspension arms 11, integrally connected to the rotating support 13, and a plurality of R/W heads 12 carried by respective suspension arms 11; in use, each R/W head 12 faces a respective surface 7a, 7bof the magnetic disks 7, for reading and writing. A motor 14 is connected to and actuates the rotating support 13 for positioning the R/W heads 12 on a desired data track 15.

[0027] The control card 5 is carried by the casing 2 and comprises a control unit 16 and a sensor assembly 17. In addition, the control card 5 is connected to the motor 14 through a connector of a known type (not shown).

[0028] As illustrated in FIG. 3, the control unit 16 comprises a logic unit 20, a controller 21, a compensation block 22, a subtracting node 23, and an adding node 24.

[0029] The logic unit 20 generates, in a known way, a reference signal S_REFindicative of a target position for the R/W heads 12. The subtracting node 23 receives the reference signal S_REF, as well as a position signal S_POSissued on a first output of the head-actuator assembly 4, in a known way. The output of the subtracting node 23 is connected to an input of the controller 21. The controller 21 has an output connected to the adding node 24 and supplying a control signal S_CONTR. The head-actuator assembly 4 has an input that is connected to the adding node 24 and a second output that is connected to the sensor assembly 17 and supplies a first timing signal T₁. The control signal S_CONTRand the first timing signal T₁have a first frequency F₁determined by the rotation rate of the disks 7 and by the number of control sectors present. For instance, in a disk having an angular speed of 5400 rpm and 60 control sectors, the first frequency F₁is 5400 Hz.

[0030] The sensor assembly 17, which will be described in detail later on with reference to FIG. 4 and is preferably made in a single integrated device, has an output connected to an input of the compensation block 22. In addition, the sensor assembly 17 supplies, on its own output, an acceleration signal S_ACCsynchronized with the control signal S_CONTRand representative of an acceleration A due to noise forces, such as mechanical vibrations, acting on the R/W device 1. The compensation block 22 has an output connected to the adding node 24 and supplying a compensation signal S_COMP.

[0031] In this way, the sensor assembly 17 and the compensation block 22 form an open loop compensation circuit 27. Furthermore, the controller 21, the head-actuator assembly 4, the subtracting node 23 and adding node 24 form a closed control loop 25.

[0032] The position signal S_POSsupplied by the head-actuator assembly 4 is representative of the distance of the R/W heads 12 from a center-of-track position and is generated when the read heads 12 fly on the sectors of the magnetic disks 7 storing the control information. In particular, the signal S_POSis obtained, in a known way, by sampling an analog signal at the first frequency F₁. The error signal S_ERRand the control signal S_CONTRgenerated within the control loop 25 are synchronous with the position signal S_POS, and hence have the first frequency F_1.

[0033] The subtracting node 23 subtracts the position signal S_POSfrom the reference signal S_REFand generates an error signal S_ERR. The error signal S_ERRis then used in the controller 21 to obtain the control signal S_CONTR, using any one of known control techniques (for example, a PID control technique).

[0034] The adding node 24 receives the control signal S_CONTRand the compensation signal S_COMP(which are synchronized with one another) and adds them, to generate an actuation signal S_ATTsupplied to the head-actuator assembly 4 to correct the position of the read heads 12.

[0035] With reference to FIG. 4, the structure of the sensor assembly 17 is now described, the output signal whereof (acceleration signal S_ACC) is synchronized with the control signal S_CONTR.

[0036] The sensor assembly 17 comprises a sensing element 30, an acquisition stage 31, a synchronization stage 32, and an interface circuit 35.

[0037] The sensing element 30, preferably an acceleration sensor of the rotational type, made using the MEMS technology, detects the rotational component of the acceleration A acting onto the R/W device 1, for example on account of mechanical vibrations, and has an output supplying a measure analog signal V_Mcorrelated to the acceleration A.

[0038] The acquisition stage 31 is connected between the sensing element 30 and the interface 35, and comprises an analog-to-digital converter 37, preferably of the sigma-delta type, a decimator 38, a numeric filter 39, a correction block 40, and an output register 41, cascade-connected together. In addition, the A/D converter 37 has a control output 37dconnected to a control input of the sensing element 30. Thereby, the sensing element 30 and the A/D converter 37 form an acquisition loop of the sigma-delta type, in itself known and described, for example, in M. Lemkin and B. Boser, “A Three Axis Micromachined Accelerometer with a CMOS Position-Sense Interface and Digital Offset-Trim Electronics”, IEEE Journal of Solid State Circuits, Vol. 34, No. 4.

[0039] The synchronization stage 32 comprises a phase locked loop (PLL) 42 and a phase generator circuit 43. The PLL 42 has an input 42areceiving the first timing signal T₁, a first output 42bconnected to a first input 37aof the A/D converter 37 and supplying a second timing signal T₂, and a second output 42cconnected to a timing input 38aof the decimator 38 and to a timing input 39bof the numeric filter 39 and supplying a third timing signal T₃. In the PLL 42, the second timing signal T₂and the third timing signal T₃are generated as described below, starting from the first timing signal T₁.

[0040] The phase generator circuit 43 has inputs 43aand 43breceiving the first timing signal T₁and, respectively, the third timing signal T₃, and an output 43cconnected to a decimation input 38bof the decimator 38 and to an activation input 39cof the numeric filter 39. The output 43cof the phase generator circuit 43 supplies decimation pulses T_DECat the first frequency F₁and with a preset delay with respect to the first timing signal T_1.

[0041] The second timing signal T₂has a second frequency F₂not lower than the first frequency F₁, and the third timing signal T₃has a third frequency F₃comprised between the second frequency F₂and the first frequency F₁. In particular, the following relationship apply: F₂=M*F₁, and F₃=N*F₁, where M and N are two integers, with M≧N (for example, M=224 and N=32). Consequently, since F₂=(M/N)*F₃, the second frequency F₂is greater than or equal to the third frequency F₃. In addition, the decimation pulses T_DEChave the first frequency F₁of the control signal S_CONTR.

[0042] During operation of the R/W device 1, the PLL 42 locks to the first timing signal T₁of the control loop 25 and generates the second timing signal T₂and the third timing signal T₃. In turn, the phase generator circuit 43 generates decimation pulses T_DECwith a preset delay K with respect to the edges of the first timing signal T₁. The delay K is representative of the sample to be selected by the decimator 38. Those skilled in the art will appreciate that the decimator 38 selects a predetermined number of subsamples, but is not limited the selection of every tenth sample.

[0043] The A/D converter 37, preferably of the sigma-delta type, receives, on a second input 37b, the analog measure signal V_Mand, using the second timing signal T₂as internal timing signal, issues, on an output 37c, a sampled signal BS formed by a flow of bits oversampled with respect to the first frequency F₁. In particular, since the A/D converter 37 is of the sigma-delta type, the individual bits of the sampled signal BS are produced by the A/D converter 37 at the third frequency F₃.

[0044] In a per se known manner, the decimator 38 filters the sampled signal BS received from the A/D converter 37 in order to reconstruct an oversampled numeric signal, wherein every sample includes a preset number of bits, and in order to eliminate any high frequency components introduced by the analog measure signal V_M. Indeed, if these component are not eliminated, decimation would cause a spectrum folding that, as known, causes distortions. In addition, at each decimation pulse T_DEC, the decimator 38 selects and supplies a sample to a data input 39aof the numeric filter 39, while the other samples are eliminated. For example, at each cycle the K-th sample following the rising edge of the first timing signal T₁is selected, as illustrated in FIG. 5, where the delay K is an integer number comprised between 0 and N−1. Since a decimation pulse T_DECis generated at each cycle of the first timing signal T₁, the decimator 38 outputs samples that have the same frequency F₁as the first timing signal T₁of the control loop 25.

[0045] The numeric filter 39, which is preferably a second order self-regression filter (or infinite impulse response IIR filter) of a known type, filters the data received from the decimator 38 to eliminate the high frequency noise components and carry out a phase equalization in the frequency band of interest (for example, from 0 Hz to 800 Hz). This enables the phase delay introduced by the decimator 38 to be compensated and reduced to the minimum.

[0046] The correction block 40 compensates the presence of offsets and the gain losses introduced by the acquisition stage 31 by adding an offset term to the samples received and, respectively, by multiplying the samples by a preset gain factor.

[0047] The signal at the output of the correction block 40 is the acceleration signal S′_ACC, which is stored in the FIFO type output register 41 to be read preferably in an immediately successive cycle of the first timing signal T₁. In practice, each sample of the acceleration signal S_ACCis representative of the acceleration detected by the sensing element 30 during the cycle that precedes the first timing signal T₁and is made available through the interface 35 to be used by the compensation block 22 from the start of a cycle that immediately follows (as represented by the arrow in FIG. 5). In addition, thanks to the use of the timing signals T₂, T₃and the decimation pulses, respectively generated by the PLL 42 and by the phase generator circuit 43 starting from the first timing signal T₁, the acceleration signal S_ACChas the first frequency F₁and is therefore synchronous with the signals present in the control loop 25 (in particular, it is synchronous with the control signal S_CONTR).

[0048] As shown in FIG. 6, the phase generator circuit 43 comprises an edge detector 45, a counter 46, a synchronization register 47, and a comparator 48.

[0049] The edge detector 45 is connected to the first input 43aof the phase generator circuit 43 (fed with the first timing signal T₁) and supplies reset pulses R at rising edges of the first timing signal T₁. The counter 46 has a reset input 46areceiving the reset pulses R, is connected to the second input 43bof the phase generator circuit 43 (fed with the third timing signal T₃) and has an output supplying a counting signal C which is incremented at each cycle of the third timing signal T₃and hence has the third frequency F₃.

[0050] The comparator 48 has a first input connected to the output of the counter 46, and a second input connected to an output of the synchronization register 42, which supplies the delay K. When the counting signal C is equal to the delay K, an output 43cof the comparator 48 switches, supplying a decimation pulse T_DEC. Preferably, the delay K is my be programmed by writing the synchronization register 47.

[0051] Consequently, the present noise compensating device generates a sampled signal BS having a frequency correlated to the signals present in the control loop 25. Thereby, an acceleration signal S_ACCmay be generated, which represents accelerations due to disturbance forces acting on the R/W device 1 and is synchronous with the control signal S_CONTR, without need of sampling frequency conversions. Moreover, the samples of the acceleration signal S_ACCmay be generated immediately before being used by the compensation block 22. As a consequence, the phase of the compensation signal S_COMPis not degraded. Rather, the use of the numeric filter 39 to perform an equalization, enables minimization of the phase delay, and thus ensures a high effectiveness of the open loop noise compensation.

[0052] Synchronization is moreover performed using a hardware circuit instead of a firmware procedure. The design phase is thus simplified, and it is possible to reduce both the size of the nonvolatile memory storing the firmware and the required processing power.

[0053] In addition, the numeric filter 39 eliminates of any noise outside of the frequency band, and in particular high frequency noise. Furthermore, the fact that the sensor assembly 17 is represented by a single integrated device means that the noise affecting the acceleration signal S_ACCis very contained.

[0054] Finally, it is clear that numerous variations and modifications may be made to the device described herein, without departing from the scope of the present invention.

[0055] In particular, as already emphasized, the invention may be advantageously applied in all cases in which it is necessary to perform high precision open loop compensation in a discrete time control system.

[0056] In addition, the numeric filter 39 may be of a programmable type and may be modified at any moment through the interface circuit 35. In this way, it is more versatile and is suitable for creating adaptive control systems. Finally, the sensor element may be of a different type and may comprise, for instance, one or more linear MEMS sensors. In this latter case, the sensors may be oriented along different axes in such a way as to detect different acceleration components. The signals supplied by each sensor may then be combined in a per se known manner for calculating the overall acceleration to which the R/W device 1 is subjected as a result of the disturbance forces.