You are on page 1of 8

Discrete control for a computer hard disk by using a

fractional order hold device

R.Barcena, M. de la Sen, ISagastabeitia and J.M.Collantes

Abstract: It is well known that the type of hold circuit used in a digital control scheme critically
influences the position of the discrete-time zeros. Most digital control systems use a zero-order
hold (ZOH). Nevertheless, there are more convenient alternatives to the use of a ZOH signal
reconstruction device in certain digital control problems. A fractional order hold (FROH) circuit is
proposed in this paper in order to manage a computer hard disk read-write head described as a
second order system. Significant improvements in the transient performance of the closed-loop
system are obtained with respect to the use of a ZOH when using a properly adjusted FROH device.

1 Introduction been extended to sufficiently small finite sampling periods
for zero-free continuous-time plants.
Unstable zeros limit the performance that can be achieved In the present paper, a study of the stability properties of
when controlling a system, [I]. Moreover, several techni- the discrete-time zeros of a computer hard disk read-write
ques for control system design are based on the cancella- head model using a FROH signal reconstruction device is
tion of process zeros. However, such methods cannot be presented. In addition, two different approaches will be
applied when the process has unstable zeros [2-51. Conse- developed here in order to improve the stability degree of
quently, special attention is being paid to the study of the the discrete zeros. The magnitude of the zeros obtained
zeros of sampled-systems in recent years. This subject was with such techniques is down to 50% smaller than the
studied by Astrom et al. [6] (whose work has been recently corresponding ZOH ones. Finally, a classical pole-zero
extended by Blachuta [7] for the case of zero-order hold placement digital control scheme is applied to the system
(ZOH)). Hagiwara et al. [8] carried out a very motivating under study in order to confirm and quantify the reached
comparative study demonstrating that a First order hold improvement on the controlled system performances. Such
(FOH) provides no advantage over a ZOH as far as the improvement can be obtained in two different ways. On the
stability of the zeros of the resulting sampled-data systems one hand, the referred FROH zeros may be cancelled more
is concerned. Passino et al. [9] proposed the fractional efficiently by the controller if required by the design
order hold (FROH) as an alternative to the ZOH. specifications. On the other hand, in some cases the
Ishitobi [ 101 analysed the stability properties of the limit- cancellation of the zeros is unsuitable and the transmission
ing FROH zeros-for sufficiently small or large sampling to the reference model becomes necessary. In such cases,
period. In that paper, it has been proved that the FROH- the zeros of the reference model may be selected freely
type reconstruction leads to inverse stable discrete plants within some range. In that way, the synthesised closed-loop
for a wider range of second-order continuous systems if the system satisfies more tightly a set of given transient
hold adjustable parameter is allowed to be negative. In specifications (e. g. percentage overshoot or settling time)
spite of the fact that a nonlimiting sampling period is used since response performances highly depend on the relative
in the present paper, the obtained results coincide qualita- positions of the closed-loop poles and zeros [12, 131.
tively with those obtained in [IO]. In [ l l ] , theoretical
analysis of the stability of the FROH limiting zeros as
2 General background theory
the sampling period tends to zero for continuous-time
plants of arbitrary pole-zero excess has been developed.
In addition, the results for limiting zeros when T+ 0 have Suppose that the state space equation of a nth order time-
invariant single-input single-output controllable and obser-
vable system is expressed as:
0 IEE, 2001
IEE Proceedings online no. 20010196
DOI; 10.1049/ip-cta:20010196
Paper first received 10th February and in revised form 12th December where u(t) andy(t) are the input and the output scalars, and
x ( t ) is the state vector. To design a digital control scheme,
R. Barcena is with the Departamento de Electronica Telecomunicaciones,
Universidad del Pais Vasco, E.U. Ingenieria Tca. Industrial. Plaza de La
we are interested in the discrete-time system composed of a
Casilla, 3. Aptdo. 48012, Bilbao, Spain hold device, the linear continuous-time system and a
E-mail: sampler in series. When the FROH signal reconstruction
M. de la Sen is with the Instituto de Investigacion y Desarrollo de device is used, the input is described by:
Procesos, Universidad del Pais Vasco, Facultad de Ciencias, Leioa
(Bizkaia), Aptdo. 644 de Bilbao, Spain
I. Sagastabeitia and J.M. Collantes are with the Departamento de
Electricidad y Electrbnica, Universidad del Pais Vasco, Facultad de
Ciencias, Leioa (Bizkaia), Aptdo. 644 de Bilbao, Spain KT 5 t < ( K + 1)T (2)
IEE Proc.-Control Theory Appl., Vol. 148, No. 2, March 2001 1I7

u(KT) . for example. equivalently. the sampled system is [9. like analogue signal.. the ZOH and the FOH will be considered used by the control scheme in order to generate the FROH- as particular cases of the FROH when /3 = 0 and p = 1 . Vol. IO]: (7) b x(KT+ T ) 0 I[xl(KT+T)l=[O y. 148. March 2##I . Tis the sampling period respectively.switches are controlled by the digital inputs and would be implemented using. S has z det[zI . and b is the adjustable gain of the device. The .where K is a non-negative integer. field effect transistors. It has been proved [14] that the "out . 5 duty cycle << r). 1 shows a very simple and intuitive design of an and 0 is an nth row vector with all zero elements. a modified integral operator. In doing so.(z) = (z .(KT)l+[)' rP N. The FROH device that implements the u(t) function given by sampled transfer function is: eqn.I way that: b(z . notice that. +zdet[ ZI -@ i'] In [I41 an approximate method is proposed as an alternative to the electronic implementation of the FROH device. 1 Fractional order hold (FROH) circuit 118 ILE Pro. Such a method utilises a ZOH device with a sampling period T I smaller than the sampling period T In the following. Fig. Fig. 5.-Control Theory Appl.@] (5) a period T an11 manages the discharges of the C .1) det digitally generated sinusoidal function with f = 1 kHz.assuming that the reconstructed signal is a No(z) = b(z . The circuit coasists of two resistor ladder networks.x(KT) O][x.. In addition. the zeros of the discrete-time system are the roots of: the FROH circuit operation for various values of the parameter /?. the location of the discrete-time poles does not depend on fl. 2.- Fig. Then. an adder and several electro- nic switches. adjusting the integrator gain in such . respectively.C. input resistance is controlled by B. 1)det 0 + z det -- 'Ref RI"tl"C1 p[.u((K .1)T) T 1 Both A and B are calculated by the digital controller. The digital inputs are the binary numbers (4) A and B (only 4-bits for the simplicity of the figure) and the pulse function S. 2. according to eqn. 2 illustrates an example of Thus. No. A represents u(K) and the integrator or.1)det + z det (8) (3) where NTO(Z)and NT1(z)are the numerators of the sampled where transfer funclions with ZOH and FOH. capacitor (R.

K = 10 N m/rad and Ki = 0. 0.1 -0. C = 0.5 time. zero-free transfer function: FROH (I( = d 0. 2 Signal reconstruction with ZOH (i. 2. 4.8 0 0. 5).62 ~ ~ a ZOH ([I = 0) i(s) ~ ( s ) 1s2 h FOH (b= 1) c FROH (b= 0. Taking is commonly defined as I zI 5 0.01 -1. . A simplified but can be used to provide accurate positioning of the read- appropriate model for the read-write head (see Fig. Therefore. The zeros of the later category are called discretisation zeros and are the kind of zeros that The advantages of the FROH signal reconstruction device will appear when discretising this zero-free system. I ] a n dTe[O. No. The system will be discretised by using a FROH device with B E [-I. 148. This transfer function may be discretised obtained results are as advantageous as those obtained by under the use of a sampling and hold device.K. fi = 0).p solenoid Fig. << 7: gories: those which correspond to the zeros of the contin- uous-time system. FOH (i.0 -0. I ] and a sampling I8 + Cb + KO = Kii period T E [0.01 Kg m2.5) and sampling period 0.03 0.O.004 N m s/rad. In doing so.0 the motor torque constant and i is the input be built (Fig. a b I I .o 1.Is] 0 0.(s + 0 . The zeros the accurate electronic circuit when the sampling period T of a discrete-time system can be classified into two cate- tends to zero and T . are verified through the design of a computer hard disk The subsequent task is to design a digital controller that read-write head position digital controller.a v -1.2 0. the first step consists of discretising described by the following differential equation [ 121: the continuous plant.5) (1 0 ) d FROH (fi = -0.2 FROH porometer .0 Fig. 3 Computer hard disk drive FROH zero IEE Proc.09 n disk pioten z- 0 ._ o L 0 5a 0. 2 ) ~+ 3 1.e. a contour plot composed of lines of constant damping coefficient of the bearings.05 0. the system is described by a second order Fig.5 1. Vol. angular zeros are suitable to be cancelled by the digital controller- velocity and the position of the head respectively. March 2001 I19 . C the viscous of this zero.5) when H(s) and i(s) are the respective Laplace transforms of H ( t ) and i(t). 3) is write head.05Nm/A. 6. R are the angular acceleration.06 0. To study the stability where I is the inertia of the head assembly..6 0. 4 Magnitude of most unstable FROH discrete zero of GB(z) for I ( ~ [ .e. The magnitude of the most unstable zero is presented in Fig.1 SI. K is the return spring magnitude values obtained from the previous surface can constant. 5 Contour plot of constant magnitude values of most unstable Fig.6 -0. 101.1 0.8 1.1 ms of the digitallj~ 5 generated junction g(t) = sin(2nP) V with f = I kHz G(s) = = B(s) = Kl - + Cs + K . and those introduced by the discretisation method which do no have any continuous- 3 Zeros of sampled system time counterparts [8.02 0.-Control Theory Appl. The stability region-where the discrete current.01 5 VI 0.9 in order to enhance the the Laplace transform with zero initial conditions and using the usual numerical values I = 0. the discretisation zeros (9) can be determined by eqn. 0. fi = I).l .5 and fi = -0. 0.07 0.08 0.

the starting and termi- nating points remains identical. (ii) The contour lines limiting each region are very abrupt when negative values of p are used (i.9960) + 0. Then. Once the p gain of the that can be reached by means of a suitable choice of the FROH reconstruction circuit is tuned to the value of bopt parameter p is shown. 7 shows. Ihe terminating points of the root locus of value of p. by adjusting p to a suitable value magnitude provided by the FROH (Izoptl= 0. Besides.The CRL is depicted in Fig. 7 Two brcakaway points are located on the real axis at z .005s ZOH discretisation (IzZOHI = 0.005 s. Therefore. However. The generalised root locus method will be used to study the (i) For any given sampling period 7: it is always possible to location of the discretization FROH zeros in the complex obtain FROH discretisation zeros that are more stable than plane as a function of /3. ZOH (p = 0)l discretisation ( lzZOHI = 0. = 0.0 are the roots Of: electronic circuitry. p discretisation zero magnitude ( lzoptl= 0.54% smaller than the magnitude corresponding to the obtained.4999 and marked as zOpt.. the following properties can be derived from these FROH zeros are located inside the stability region.. where fl E ( . Both the ZOH ( p = O ) and FOH the most stable zeros are obtained for this plant and this ( p = 1) are not able to provide discrete-time zeros into the sampling period. 7 and 8 into the ZOH and FOH ones by means of a right choice of eqn. 6 as a function of parameter p.(z) = 0 as /? + fCO are the roots of being as stable as possible for a particular continuous-time plant and sampling period. the FROH discretisation zeros zero as a function of p i s obtained from eqn. always moving away from the origin.4999 and z2= . Besides. as Fig.8 -0. due to the relative positions of the the computer hard disk read-write head for a given value of starting and tlxminating points on the real axis. Therefore. (a) The most general approach consists of obtaining the The classical root locus procedure ensures that the root roots of eqn.1976 x lo3.3334 and &=-2. =-0. If the implementation is made with and the starting points of N.1 -0. except the fixed one at z=O.9993). This fact agrees with generalised root locus with the parameter fl as its general- the theoretical properties of limiting FROH zeros-for ised gain: sufficiently small (T-+ 0) limiting sampling periods- given by [ lo]. Substituting eqns. an inverse stable discrete plant can be 48. -1.0 -0. In this case. there exists a range of 1.6 -0.8 1. the previous observations: Finally. the zeros of Nlc(z)can be calculated as p varies as a negative values of the parameter fl.4999) that is Fig. However. This negative rangc: goes from 0 to a value pmin= . = -0. 12 and 13 to the system described by 4 Improving the stability properties of the zeros Two techniques are now presented in order to determine the optimum value of p as far as the stability degree of the eqn. the FROH I20 IE'E Proc. 6 and plotted with j > 0 are always real and of less convergence abscissa in Fig. the value that provides the FROH zeros N. Vol. = 0. Note that the sampling period T. Therefore. but poorly damped. the magnitude changes very quickly as fl varies). 10. 2. 11.March 2003 . one 82btains: ' "[ + (Z . are also digital control system.0 popt= P I = -0. Notice that the most unstable zero stability region. Th'is implies that the FROH discretisation zeros (b) The second method consists of applying the classical with p ranging from 0 to plus infinity evolve leftwards- root locus method with fi being the generalised gain.I)(z Z(Z 1- + 1.=-0. Besides. 11. thi: second approach is described here. lamin 1 where Bmin depends on the continuous-time plant parameters and the current sampling period Tc...2 0. the optimum In eqn. while the departure and arrival angles are opposite.1 0.9970.1. 6 Mugnittide of most unstable FROH zero for snmpling period of 49.3332 in our case). namely: First let us study the direct generalised root locus (GRL). the optimum value of the parameter p corresponds to the breakaway point located at zI =-0. the value locus on the real axis always lies in a section of the real of is adjusted in order to minimise the largest root axis to the lelt of an odd number of starting and terminat- magnitude by using approximated numerical recipes.on the negative real Now.robustness of the design by avoiding the cancellations of obtained by searching the minimum of that function stable.2 0 0.2 negative values of p for which the FROH zeros are located closer to the origin than the ZOH and FOH ones.e.3334. then /3 has to be fixed as close as possible to its optimum value according to the achievable technical performances of the components.97% smaller than the magnitude corresponding to the digital control system tinder study T.&) = 0 as p+. Thus. Applying eqns. 6. must be calculated as exactly as possible in the case when using algorithm implementations of the FROH device by using the approximate method proposed in [14]. Note that this FROH provides a FROH porameter.6 0.9993). 148. that is well suited for our the starting p'oints. The corresponding values of p are f l . No.5 142) is in the FROH. the optimum value of p can be easily than the ZOH ones. ing points.CO. the two above approaches are applied to the case of axis as 13 increases.9993) =0 (14) zeros of resulting sampled systems is concerned. plant zeros. 6. The magnitude of the most unstable the ZOH zeros. Then.-Control Theory App/.Lr locus (CRL). in the case of complementary generalised root 1. 01. that is. The improvement (flop.

p = 0 ) and the model reference control scheme of Fig. the equivalent the controlled system. Now. 8 Unit-step response of open-loop continuous plant parameter p to popt= -0. 151.5394 . Therefore. The ZOH signal reconstruction device and.6.e. Z(Z~ + 1. 9 is applied. oscillatory (see Fig. a well-known digital control scheme is used and zeros of the Gz.. the system has a very oscillatory step response (Fig.0. inverse stable plant. The second part of the next Section damping coefficient of the closed-loop system is about 0. The classical model reference control tageous since the associated procedure can be strongly scheme of pole-zero placement proposed by [2. when the control objective is reached. 8). lo]) since the controller possesses a alternatively. it is close to be critically stable). this method could be difficult to apply in The desired closed loop performance specifications are other cases where the generalised CRL is very complex.5.05 (17) the reference model from a transmission of those of the plant. 10-5(z 0. fairly. 9 Pole-zero placement digital control scheme IEE Proc. samples Fig.9 0 1000 2000 3000 4000 5000 6000 time.e. 6.Ho. The design objective is to achieve that the controlled system meet certain perfor- Rernark 2: The second approach is. 10p5(z 0. The closed-loop step response is very Section 3. The open-loop discrete transfer function when a ZOH device is used ( D = O ) is: BZOHOJZ) . (by the cancellation of such zeros). it may A.L(z) .9979) - precompensotor On the one hand the poles (z = 0.9732 0..6 is devoted to such a design philosophy. the most advan- mance specifications.0101 1 2. theoretically. the performance of the controlled system could be significantly improved [ 1. the proposed FROH circuit are used here to poorly damped (almost critically stable) pole (which corre- reconstruct the control signal. 148. On the other hand. No. (z) plant are cancelled by the for accurate positioning of the readlwrite head described in digital controller. An illustrative example Assume that a ZOH is used (i. GFROHOL(z) = AFROHOL(Z) . 2. the discrete open-loop transfer function is then: zeros are inside the stability region. so that they can be cancelled by means of a digital controller. improving.9979)+ (16) Rernark I : The small discrepancies found with respect to the results obtained using the first approach are due to the where.March 2001 121 . Nevertheless. whose zeros are still unspe- be pointed out that the previous study in the evolution in cified: the complex plane of the FROH zeros with /3 may also be used to choose appropriately the position of the zeros of A M ( z )= z2 + 0.(z) of the reference model.. the transient performances of Thus.-Control Theov Appl. step reponse 0. described by the roots of the characteristic polynomial Remark 3: As mentioned in the introductory section. in this way. Fig.9993)+ reference model GZOHOL = + + AZOH. + BFROHOL(Z) .2328 . Then.4999)2 In this way. the performances of the controlled system in both cases are compared. 5 Computer hard disk read/write head digital 5. and the transient response is quite fast. 9 7 3 ~ 0.3332.155j) have very signal rec - y(t) light equivalent damping (equiv.987 f0. the open-loop system performance is studied when feedbock the FROH is used. 161 (see supported by the use of classical graphic techniques. the discrete zero is located almost on the unit circle (i.22 + 0. Vol. By using the optimum adjustment of the Fig. damping = 0. the discrete zeros are stable enough to finite accuracy of the numerical recipes used to find the allow the free selection of the complete reference model minimum of the function depicted in Fig. All the poles In this section.00632) and circuit continuous plont are near the unit circle. 9) is used here.1 Cancellations of the discrete plant zeros control design.(z* 1 .

Comparing is the closed-loop discrete transfer function when the ZOH Fig.020 0.030 0. 17. the resulting closed-loop discrete 5. It turns out that a zero cancellation with the ZOH device and then these zeros are transmitted to the reference model. considering the position of the roots zero is much more stable (see Fig.-Control Theory Appl.06 time.03 s (six sample times). The maximum overshoot is around 3% and the output settles in 0.05 0. e. 13 (see also Table l).06 0 0. 148.3332.3332 is used.3332. an alternative solution is proposed consisting in z = -0. synthesising an appropriate pole-zero relative positions in Now.‘EE Proc. the FROH with the optimal parameter the reference model by selecting a pertinent FROH hold /jopt= -0. 12 and 13 are presented in Figs.01 0. controlled continuous-time system and the closed-loop The control signals referred to the system step responses discrete transfer function GZoHcL(2).step response of compensated system by using FROH Fig. However. when discrete zero is transmitted to with /j = bop.030 0.8- ( I I I 0 0.s time. 12 Unit-step response of compensated system by using ZOH device. since important to place the FROH zeros with respect to the poles of A M ( z ) oscillatory effects appear in the system output between the was the obtaining of a fast response with a small percen- sampling instants. when plant zero is cancelled by digital controller sponds with the cancelled plant zero located in Now.025 0.01 1.= -0. 14 and shoot of the transient response is approximately 14% and 15 for the ZOH and FROH case.010 0 0.s Fig.01 0. the controllced system performance when a properly Fig.v using FROH device device with := pop. 2.015 0. 12 with Fig. [13].9993). 11). the improvement on circuit is used. cancellation of the process zeros [2. of the characteristic polynomial AM(z) given by eqn.= -0. 13.8 1. In such a case.2 - *T 1. the cancellation of the The basic criterion observed in this particular case in order FROH zero is not a good choice either.. The maximum over.2 Transmission of the reference model zeros transfer function is then: Evidently.005 0. it was found that the best closed-loop response again corresponds to identical p. 131. MarcJi 2001 .s Fig.01 0.01 0. Vol. respectively.035 0. Observe that settles in about 0. Thus.No. the FROH zeros are placed closed-loop system is much better since now the cancelled properly by using p. The obtained response of the device pararneter j.03 0.s time.02 0. see. displayed in Figs.03 0. 13 Unit. 10 Unit-step response of compensated system by using ZOH Fig. After exhaustive simulation. is highly unsuitable.2 c OV I 0 0.020 0.0 - 0.g. no attempt must be made to cancel the zero of G.005 0.010 0. popt= -0. First.3332. The step responses of the previous system are presented on Fig.025 s (five sample times).015 0. when plant zero is cancelled by digital controller reference model 122 .OHOL(z).025 0.010 0.040 time.035 0. 12 describes the unit-step responses of the adjusted FROH device is used can be shown. 2.02 0. when discrete plant zero is transmitted to reference model device. 11 Unit-step response of closed-loop system 6. Such well known effects are due to the tage overshoot for the controlled system. 1.05 0.

. zeros 1z/=0.. pp. Automatica.W. to obtain the optimum value of the parameter p. and WITTENMARK.: ‘Inverse stable sampled low- pass systems’. B. Fig. although the read/write head regular ZOH. K.9993 (21= 0. 2. R. 4 5 6 4 6 4 IEE Proc. the the control scheme in Fig. % Max.: ‘Zeros of sampled systems’. ( S ) . I99 I . M. Improvement of FROH (p=&.H. 1985. a discrete inverse stable trapsfer function ‘(Le. (4). M. K. Theory and design’ (Prentice-Hall. 15 Plant input rejerred to Fig. 1905-1913 10 ISHITOBI.: ‘Stability of zeros of samoled systems with fractional order hold’.D. the FROH device allows the . 14 Plant input referred to Fig. 58. it is unexpected for the above numerical results to be highly different from those to be obtained in a ZOH ( p = O ) FROH Improvement. 28 1-292 2 ASTROM.. 1 Control. K.) - 3 ASTROM. 147.M.s Fig. IEE Proc.020 0. 1984.015 0..: ‘Self-tuning control of nonminimum-phase systems’.and First-Order Holds’. & m o l Theory Appl. (6). Englewood Cliffs. Automgtica. and ANTSAKLIS.005 0. overshoot 14% 3% 78. 296- 300 0 0. 7 Acknowledgment The authors would like to thank the anonymous referees for their useful comments concerning improvements of the --1.. 1996. 14 and 15). 1229-1234 8 HAGIWARA. the maximum amplitude of the FROH analogue signal is AS a result of this study. more exigent for the electronic circuitry since the power Furthermore. (I). experiment. 1999.. 1325-1346 9 PASSINO.5 Settle time 0.: ‘Improv- time. two approaches have been used.005 0. Int. when suitably adjusted. M. (6).020 0. New J&sey. In addition.Table 1.. it has been shown that the sampled system zeros can be located inside’ the stability region not increased with respect to that associated with the ZOH device. 2nd edn.. observe that. 31-38 7 BLACHUTA. HAGANDER. pp.l990.: ‘Computer controlled - systems. and SAGASTABEITIA. the ZOH control signal turns out to be when the ZOH (/3 = 0) and FOH (fl= 1) fails to do so. Control. 20. by the FROH signal reconstruction device has been ‘studied for the case of a digitally controlled hard disk readjwrite head. these exists a dissipation in the ZOH case is higher than that related to range of negative values of /3 for which the discretisation the FROH reconstruction (compare Figs. No.J..035 0.-Conlrol Theory Appl. 1 1 3 4 2 3 6 ASTROM.035 0. 143. I. the size of the function v(t) of period T.6 The stability properties of the zeros introduced. the discussion has been extended model used in eqn. K.0LO original version of the manuscript.J..: ‘Adaptive control’ (Addison- Wesley.025 0. 1989) 4 CLARKE. (2). Int. 2000.015 0. J.01 O‘I- 0 U 0.. pp. March 2001 123 . In that way.030 0. 148. 2. R. 501-517 5 M’SAAD.. 1988. In the case of the hard disk read-write head. IEE Proc. Automatica.: ‘Adaptive controllers for discrete-time systems with arbitrary zeros: an overview’. Hence.s ing the stability properties of the zeros of sampled systems with Fractional order hold’.106-EB8235/2000 and DGES (Project DPI2000-0244).03 s 0. by using suitable negative values of the FROH parameter p.025 s 16. 12 1 MIDDLETON. B. D. 171 like a correct device description. and LANDAU. R. pp. Control. avoiding related saturations. 1993. that is. Therefore. pp. Finally.application of a wider range of digital control schemes and enhances notably the performance of the closed-loop controlled system. and WITTENMARK.. the designer must magnitude of the zeros obtained with. pp. pp.: ‘Stabilitv ofthe limiting zeros of sampled-data systems with Zero. the control signal demands transfer function with stable numerator polynomial) has will be accomplished with. ZOH Therefore.. Thus. P. a design of Fig. the size of the driving current depends on the value of the FROH gain that provides the most stable zeros magnitude of the required angular displacement of the for a given continuous-time plant’ and a given sampling read/write head.: ‘Trade-offs in linear control system design’. it is highly assumed to the design of improved reference models by the trans- mission of the handled zeros of the resulting discrete plant.J. pp.. in order signal improvement with respect to the ZOH case. T. Vol. 44. such techniques is up consider the maximum displacement of the head in order to to a 50% smaller than the corresponding ZOH’ones.J. 8 References time. that is. I. 13 (4). 1984. ORTEGA. P. pp. 20. Moreover.030 0.. for a given sampling period’ T. Automatica. 9 is very simple..J.97 6 Conclusions Max.025 0.t) compared to in the literature [12. (3).040 11 BARCENA. 27. Autoin. YUASA.4999 49. (6). 21. FROH zeros are up to 50% more stable than the ZOH and no apparent significant price has been paid for the output FOH ones.010 0. T.. Control Theory Appl.: ‘On zeros of pulse transfer function’. In this choose appropriately the electronic components in the way.010 0. IEEE Trans. the Evidently. M. This work has been partially supported by the University of the Basque Coun- try (UPV/EHU) through Project 106. DE LA SEN. been obtained while unstable zeros were generated by the Finally. and ARAKI. and STERNBY. M. mag. 47. 9.

: ‘Modern control systems’ (Prentice 698-702 Hall. pp. IEEE International 127. R. 1980. pp..H.. 2. and WORKMAN. Int. 2nd edn. and H A M . No. 2000. 17 DORF.) tary sensitivitli functions singles input-output digital control systems’. 1992. 1.: ‘Digital control systems’ (Saunders College Publishing.L. 2429-2439 1992. H. (3).F.I. G.: ‘Digital 15 SUNG. 120-1 30 Conference on Intelligent processing systems.F Proc. B.. Vol. L). pp.. 13 KUO. 1988. March 2001 . 148.2nd edn. M. New York. R.) 16 ASTROM.-Control T h e o y Appl. Control Theory and Appl..12 FRANKLIN. 9th edn. and WITTENMARK.D. 48. IEE Pmc. J. POWELL. B. (6).K..: ‘Propeities of sensitivity and complemen- control of dynamic systems’ (Addison-Wesley. and BISHOP...: ‘Zeros ofsampled systems with fractional on pole-zero placement’.) 124 IE. M. 1997. Q. order hold implemented by zero order hold’. and ZHU.C. 1 Control. K.: ‘Self-tuning controllers based 14 ISHITOBI. S.C..