Torque Ripple Minimization in Hybrid Stepper Motors Using Acceleration Measurements * Riccardo Antonello, Angelo Cenedese and Roberto Oboe University of Padova, Department of Management and Bugincering, Vicenza, Tlaly (Tels +39-0444-998844; (e-mail: antonelialigest.anipd.st; angela. cenedese unipd st ‘oberto.oboe usp.) Abstract: Hybrid stepper motors (HSMs) are commonly used in many cost-sensitive industrial and consumer applications. With the use of micro-stepping techniques, they could theoretically achieve a very high resolution in positioning of mochanical loads, even without position sensors. However, it is well known that 1ISMs are affected by a large torque ripple, due to cogging and phase unbalancing. This, in turn, may cause large vibrations on the load, especially in those systems with flexible elements (evg. transmission belts). Several solutions have been proposed toalleviate this problem, but most of them make use ofa load.-side position sensor, by means of Which itis possible to determine a position-cependent torque ripple profile, to be compensated during operations. Introducing @ high resolution sensor on the load side, however, makes the cost of the system higher, thns vanishing the advantage of having a low cast open-loop sctuator. Additionally, it is not always possible to accommodate a new position sensor on an existing mechanical system. In this paper, we propose a new system to compensate for the first two harmonies of the torque ripple in HSMs, based on the use of a load-sice MEMS accelerometer, which can be easily fitted into existing systems, without any major modifications. The automated procedure developed minimizes the torque ripple by acting on the offset and amplitude of the phase currents. Experimontal resulls on systems with and without load elasticity are reported proving the effectiveness of the proposed approach, Keywords: Hybrid stepper motor, MI 1, INTRODUCTION Stepper motors are commonly used in those motion control applications that donot require extremely high positioning accuracy, but instead impose constraints on cost and complexity of the driving eircuts. Hybrid stepper motors (IISMs) are a particular kind of stepper motors operating tinder the combined principles of the permanent magne! (PM) andl variable eeluctance (VR) motors [Kenjo, 1984, ‘The structure of a hybrid stepper motor is depicted in Fig 1. The rotor consists ofa eylindzicaly shaped permazca aagnet core magnetized along the eylinder axis, whose poles are covered by two lauinated soft steel toothed fnd-eaps, Teeth in the tivo end-eaps are misaligned with respect to ench other by half a tooth pitch. The stator has typically eight poles, and each pole has betwoen two and six teeth. The stator tooth pitch is usually equal {o the rotor tooth piteh, but i some motors is alittle lager to improve the positioning accuracy and reduce the detont torque. In n bipolar two-phase hybrid stepper motor (as considered in See. 2), two windings (phases) are alternatively arranged on the stator poles Lo form This work fas been partially supported by the Programa Operative FSB, 20072019 Regione Veneto, Codice Progetto 2105/01 /17/2214/2000 ("Modelizztsone Oriente al Conealo di Stemi Meceatroni). MS accelerometers coils through which the fow of the permanent magnet flux can be enhanced or reduced. Suocessive coils of each phase are wound in opposite sense, so that the same phase current generates oppositely oriented magnetic fields in ‘opposite stator poles, ‘Thanks to the skew arrangement of the rotor end-caps, the rotor moves in steps of a quarter of rotor toot pitch when the two phases are excited with a conventional onc-phase-on scheme (.0. each phase is excited in turn with a positive or negative rated Fig. 1. Blectromechanical structure of hybrid stepper motor (after [Kenjo, 1984))current): hence, a fine angular positioning resolution ean be achieved with a sufficiently high number of rotor teeth. For example, in a motor with 50 rotor teeth (typical specification for several commercial products), the step Tength is 1.$°. Unfortunately, the generated torque is not smooth, being affected by undesired pulsations (harmonic components) that produce an irregular motion of the Toad. Nou-smoothness of the generated torque becomes feven more problematic in case of a resonant load, where the torque ripple could excite the load resonances, thus producing unacceptably large oscillations. Therefore, the roduetion of torque ripple is highly desirable in many ‘motion control applications. The adoption of a sine-cosine microstepping excitation scheme, in whieh the two phases are excited with two sinusoidal quadrature currents, is perhaps the most. con- ventionel strategy tsed to alleviate (but not definitely re- move) the torque ripple. Several other methods have been proposed in literature for reducing or compensating the torque ripple in permanent magnet synchronous motors - see [Jahns and Soong, 1996] and the references therein. However, the majority of the proposed methods requires ‘ high resolution encoder for measuring the rotor angular position on the load side, This requirement is obviously counterintuitive in stepper motor applications, because of the intrinsic open-loop nature of the actuator. In this paper, the problem of minimizing the torque ripple induced by uncalibrated offsets and amplitudes of the phase currents in a microstepping excitation scheme is eon sidered. The proposed procedure is specifically conceived for nding the optiraal ofiset and amplitude compensation values by relying only on the acceleration measurements provided by a los--cast MEMS accelerometer placed on the Joad side. Compated to other existing solutions, the one proposed in this paper has the advantage of being inexpen= sive and easily adaptable to existing systems. The paper is organized as follows. In Sec. 2 a model of the torque ripple jgenerated by an HSM excited with a (uniceal) microstep- ping scheme is presented; such model is then used in Sec. 5 to describe the proposed minimization procedure. The ‘experimental resulls obtained with both an inertial and a resonant load are reported in See. 4. Finally, conclusions and futuze research directions are reported in See. 5. 2. TISM TORQUE RIPPLE, A bipolar two-phase permanent magnet (PM) hybeid step. per motor (ISM) is considered in the following. An ana- Iptical expression of the generated torque can be obtained by applying the principle of energy conservation; under the simplifying assumption that no magnetic satarstion ecurs in the magnetic eireuit (.0. the magnetic fnx is proportional to the stator currents andi independent of the Internal mageet), ican be proved [Khorrami etal, 2003) that tho motor Yorque Ty, wrt. the rotor position 6, is equal to 1 rab tam SE 0 where L is a 3 x3 symmmetsie positive definite inductance snatrix Ly La Lay Le [is La i) 2 Lay Lay Ley and & ~ [insta ip], with i, #2 the the winding currents in, respectively, phase 1 and phase 2, and iy is the constant fictitious field current used to deseribe the magnetic ficld produced by the permanent magnet. In the inductance matrix, La: and Lap ate the selfinductances of windings Vand 2, Liz is the mutual induetance between the two windings, Lay and Lag are the mutual inductances between the two windings and the fictitious rotor winding, and Ly is the selEindnctance of the fetitions rotor winding ‘The Lorque expression (1) can be expanded as follows ins, a fn (Matis Matis) a, alms, bw, ,.) , 1aLy 2 ip Sang eas) 4 Sg (3) ‘The first term in round brackets is the torque component generated by the interaction of the maguetic fields pro= duced by the stator windings and the permanent magnet, ‘The second term is the reluctance torque and depends fon the variations in self and mutual inductance of the ‘windings (mainly due to geometric imperfections and rotor anisotropy). The last term is the cogging (or detent) torque fand is mainly due to variations in seleinductance of the fictitious rotor winding caused by the presence of slots in the stator (ininimum reluctance path). The cogging torque is present even in absence of winding currents; its periodicity with respect to the rotor position is equal to ‘the periodicity at which the slots are located along the stator All the entries of the inductance matrix are periodic func tions of the rotor position 6, and their basic frequencies ean be deduced from the syrametries of the motor. Denoting, with 2, the number of rotor tet, holds that Lnx(8) ~ bo + ba c0s(2N8) o Lnp{0) = Lo ~ Ly 008(2N,9) (5) Ti3(8) = + Ty sin(2N,8) 6) Lsj(8) = Eno So foms c0969%8) Lay(0) = Lmo + S> Linj sin jN-O) 8 Lyp(8) = Lyn + Yo by, 00858) (9) where n ig the number of harmonies considered in the expression of the motor torque. Higher order hariuonies (u > 2) in the mutual inductance terms Lyy and Lay ‘model the nonsinusoidal (with respect to the fotor angle) flux distribution in the airgap Jn a typical (ideal) hybrid stepper motor, the variation of phase inductance is as small as a few percent, nnd its contribution to the total torqne is negligible; moreover, the surfaces of the rotor and stator teeth are shaped so tat the magnetic fx in the airgap is almost sinusoial By setting 0 522 (10)Jn the self and mua inductances (4)—(9}, and replacing then in the torque components, (8) yekls Tym = Avi N in( NO) ~ gmt Ny cos(N,0} (11) fed Teo = 5 Do bss Nesin(iNeO} (a2) and tor =0, One way to generate an ideally constant torque consists of adopting a taicrastepping driving technique, in whic two Sinusoidal quadrature currents ate imposed into the stator windings: fy = Leos(iVeO), in = Tsin(NBu) (13) ‘where isthe angular position of the stator fx. In fact, replacing (13) in (11) yields: pm = ism NeL sin (NCO — 8) (4) Hence, the torque tym is constant whenever the angle 0 = 6, —8, called torque (or load) angle, is constant. In practice, due to unbalancing and imperfections of the driving amplifiers, the phase currents will always be al fected. by some offset, and their amplitude will be un- matched. By taking into account the offsets and gain imbalancing in the expressions of phase currents i. in Lao + fax cos(NB4) i= Fao + fou sin(Ny6,) (15) and replacing them in (11) yields: Fm = irl Ne {li + ta) sine (04 —9)] — hasin(N,@) + Fao cos(.N-8) — feta tad sinin(6 +00) } (6) For a constant torque angle p = 0, — 8, two harmonic components show up in the torque tym, in addition to a constant term. The first harmonic component of the torque ripple in 7pm has the same frequency of the phase currents depending on the current offsets Lig nd component has twice the frequency of the phase currents and an amplitude depending on the unbalancing of the phase current amplitudes Jy) and fy 3. TORQUE RIPPLE MINIMIZATION IN HSMS. Several metiods have been proposed in literature for re- ducing or compensating the torque ripple in permanent ‘magnet synchronous motors; a rather complete review is reported in [Jabns and Soong, 1996. The methods can be broadly classified in two major categories: 1) methods based on the improvement of the machine design and the torque generation mechanism, and 2) methods based, fon the improvement of the excitation scheme, possibly exploiting open or closed control loop scaemes for com pensating the nonideal characteristies of tie machine. A well known design expedient to reduce the torque ripple in hybrid stepper motors consists of skewing the two end- ‘caps of the rotor by half the rotor tooth pitch. A review fof many classical and innovative design techniques for smooth torque generation in PM motors are reported in [Bianchi and Bologsan!, 2002), Regarding the second category of torque ripple minimiza- tion methods, the most typical example is perhaps the so-called sine-cosine microstepping excitation scheme, in ‘which two simsoidal quadrature currents are imposed Lo the stator windings. With such method, the magnetic flux generated by the stator rotates smoothly, thus reduc ing torque ripple and rotor oscillations, Nevertheless, the residual torque ripple could be still too large, especially in presence of large cogging torque. In order to further reduce the torque ripple, a broad range of alternative techniques, based cither on’ feedforward or fecdback compensation schemes, have boon proposed so far in literature, ‘The usual feedforward compessation scheme consists of using programmed excitation wavelorms for the phase currents to cancel the pulsating Lorque components. Ia the rajority ofthe eases, the phase current protles ae shaped by injecting suitable harmonic components that selectively climinate the torque ripple components, similarly to what originally proposed in Le-lHuy et aly 1986), In [Favre ot al., 1903}, the compensating harmonics are deterrsined boy employing an iterative procedure: in [Hung and Ding, 1903), [lanselman, 1094) and, more recently [Wea and Chapman, 2005), tho current profiles for eliminating the torque ripple componosts up to.a given harmonie order are compiated in closet form by employing merical optitaizae tion techniques, once the back em waveforms are knowa, In errett et al, 1998, compact model forthe pulsating torque in PM niotors is used to perionn a feedforward compensetion on the poxition controller output; the une knowin parameters of the model are iientified by per‘orm. ing simple closed-loop motion experiments More advance compensation schemes exploit alaptive (Chen and Paden, 1903) or nonlinear [Taylor, 1994, [Bodson et a, 1003) feedback contro! techniques to generate smoother torques Tm any case, all the aforementioned solutions share the common drawback of requiring the measurement of te rotor position: this isa rather restrictive assumption for Stepper motor applications, where the actuator is specific cally chosen to perform accurate positioning in open-loop, ‘without requiring any load-side position sensor (eg. high resolution encoder) 9.1 Proposed scheme Differently from most of the solutions available in ltera~ ‘vate, the procedure for torque ripple reduction proposed in tis paper does not require any expensive high-resolution encoder for measuring the rotor angular position; instead, it makes use of a low-cost MEMS accelerometer to detect the vibration induced by the torque tipple on the load-side Apact from cost considerations, another advantage of the proposed solution is that it ean be easily Btted in existing equipments, without requiring substantial interventions, ‘The proposed procedure is conceived to minimize the torque ripple contribution induced by imperfections in the generation of the current profiles when using @ mi crostepping excitation scheme. As already mentioned in Sec. 2, two imperfections are considered, namely ande- sired nonzero offsets in phase currents and unbalanced gains in the driving amplifiers of the two phases. The compensation scheme is based on the observation Ut, under the assumption made in See. 2, Ue nonzero eurrent offsets produce a torque harmonic disturbance component ‘at the same frequency of the driving currents, while theamplitude unbalancing between the two current phases xonerates e second harmonic al twice the froqueney of the driving currents. With reference to (16), it ean be noted that the two torque ripple harmonies have magnitudes \omal = iplimiNe YT + By ay 1 lomal = iim Ne 5 Ui — Jo as) i.e. the squared magnitude of the first harmonic depends quadratically on the current offsets Tig and Lap, while the squared inagnitude of the second harmonic depends quadratically on the amplitude unbalancing Ji, — Iz). This observation allows to formulate a multi-step minimization procedure, in which some current offsets and gain unbal- ancing are deliberately introduced in the phase currents, and then varied to seek for a minimum of the amplitude of the frst two harmonies of the acceleration ripple measured by the MEMS accelerometer. ‘The procedure can be summarized in the following stops: (2) doa frst experiment in which the ideal microstepping ‘excitation scheme (15) is modified as follows sin( NC) where the offset Zip is slowly varied over a specified range (inaction ofthe rated custent 1), while the otor roves with almost constant velocity (ie. while 8y is slowly linearly inezeased). Dy using a conventional synchronous aruplitude de modulation seheme, extract the amplitude of the firs harmonic of the acceleration ripple measured by the MEMS accelerometer Then, use the (weighted) least squares method to ba parabola to the data consisting of the squared magnitude of the fast harmonic vs. the values of the offset fo: the minimum of the parabolie ft gives the ‘optimal offiet compensation value ftp for the phase ‘current § {0 a second experiment in which the previous step is ropeated for the second phase current, Use the modified microstepping excitation scheme Ou). tx = hao + Fsin(N,8,) (20) where the offst Zap is slowly varied over a spociiod range. Obtain the optimal offset. compensation value ii by ceponting the procedure deseribed in the pre- vious step. (8) co'a thied experiment in whieh the ideal mierostop- pity excitation scheme (15) is modified as follows ¥ fieos(N4u) 2) ly + fa n(N Pu) (2) ig = hip + Feos(NeBu}, ta a9) 2 iy = fig + eos whore iy is slowly varied over a specified range {across the rated current 1), while fay = 20 — hy. ‘The last condition gusrantoos that the averege torque provided Lo the load during the whole experiment is feepl constant, despite the unbalancing of the phase current amplitudes - see equation (16). By using a conventional synchronous amplitude de modulation scheme, extract the amplitude of the sec- cond harmonic of the acceleration ripple measured by the MEMS accelerometer ‘Then, use the (weighted) least squares method to fit a parabola to the data cousisting of the squared rmagaitude of the second harmonie vs. the values of the current amplitude fy: the minimum of the parabolic Gt gives the optimal amplitude value Jj, for the phase current iy (and, indivectly, the optional amplitude value £5, for the pase currext i) In principle, the tree steps described above can be iter ‘ated many times to refine the estimation of the optimal compensation values; however, it has been noted that a single iteration is generally sufficient for attaining the ‘optimal values with a satisfactory accuracy level. 4, EXPERIMENTAL RESULTS. ‘The proposed procedure for minimizing the torque ripple in bybrid stepper motors has been tested on a commercial bipolar, two-phase HSM with 8° stop angle (50 rotor wets) driving either an inertial or a resonant loud. The inertial load hss been chosen as a stl disk with relatively small inertia, mounted directly on the motor sha‘t; as for the resonant load, at elastic transtnission system based on a pair of reduction gears (reduction ratio Ng = 1/7) cone nected by a toothed belt has been considered (see Fig.2). Such transimission system is part of the mechanism used for generating the pan and tilt motions in positioning uit for surveillance earseras (st Fig.) Fig. 2. Elastic transmission system composed of reduction gears and a toothed belt Since the experimental results are similar for the inertial ‘and the resonant loads, only the latter are discussed in the remaining part of this section, due to space constraints. ‘The proposed procedure has been applied to reduce the vibration induced by the motor torque ripple on the pan ‘motion of the positioning unit, Therefore, a MEMS linear accelerometer has been placed on top of the eamera, as shown in Fig. 4, to measure the tangential acceleration of the unit; then, the angular acceleration can be derived bby measuring the distance of the accelerometer from the rotation axis ‘Two sinusoidal phase currents with a nominal amplitude of I = 1A and a frequency of 20H (corresponding to an angular velocity of = 21°/s for the pan motion)Fig. 4. MEMS acocleromet placement on top of the Ihave boon used in the experiments. The driving frequency Inns been selected to avoid the excitation of the load resonances, thus generating a more regular motion in ‘whieh the acceleration variations produced by varying the Current offseis/ammplitudes can be highlighted more leary. Due to an inaccurate intial calibration of the offsets and gains ofthe linear ampliiers used to drive the two motor phases, the phase currents are affected by some offact and fain unbalancing, as shown in the top plot of Fig. 5 ‘The application of the procedure described in the previous section yiekis the results collectively reported in Fig. 6. The two current offsets have been varied in the range 0.5.4 +0. A (corresponding to =50% of the nominal rated current 1 = 1A); the minimization of the frst harmonic amplitude has been achieved with the ollset values Ff) = —0.1214 and fj) = 055A (see the top plots of Fig. 6). The phase current amplitudes have ‘been unbalanced np to 30% of the nominal rated current (ie. the amplitude of one phase enrrent has been varied Jn the range 0.7 A + 1.3.4, while the remaining one has bbeen adjusted to keep the sum constant to 2/ — 2A), ‘The minimization of the second harmonic aanplitude is obtained with the amplitude values /f, = 0.847 A and dg, =21 =I, = 1159. tis worth noticing that the measured square magnitudes of the first two acceleration harmonies depend quadrati- cally on the two current offsets and the amplitude unbal- ancing, thus legitimating the assumptions made in Sec. 2 lor deriving the torque expression (16). However, differ ently from what could be argued by using (17) and (18), the measurements of the compensated currents reported. in the bottom plot of Fig, 6 show that the torque ripple minimization is attained with some residual small ofscts and gain unbalancing, ‘This discrepancy between the predicted and experimental rosulls is explained by the inductanee terms neglected in the derivation of (16). especially those describing the nonsiausoidal Dux distribution inthe airgap (higher order hhaemonies inthe mutual inductance terms Lyy and Loy). ‘The benefits of the proposed compensation procedure in minimizing the first two harmonies of the torque ripple Anan Ac tie mpton re) Fig. 5. Phase currents measurements before aud after the compensation procedure pein i 01 amp fit 055 & Bae ove hina Fig. 6. Determination of the optimal phase current offsets (top plots) and amplitude unbalancing (bottom plot): Lo rege [ Fig. 7. Acceleration ripple (loae-side) spectrum before andi alter the compensation procedure, can be also appreciated in the acceleration spectrum measurements reported in Fig, 7. The compensation of the current offsets drastically reduces the amplitude of the first harmonie (at 20 /12); the minimization of the second harmonic amplitude is less evident (at 40 112), perhaps due to the fact that the initial gains of the driving amplifiers were almost balanced before applying the compensation procedure. 5. CONCLUSIONS ‘This paper has presented a procedure for minimizing the torque ripple induced in HSMs by uncalibrated offsets and amplitudes of the phase currents generated with @ mie crostepping driver. The proposed method relies exclusively on the acceleration measurement provided by a low-cost MEMS accelerometer placed on the load-side: no high- resolution, expensive encoders are needed for measuring, the rotor position, as usually roquirod in many comper sation schomes available in literature. Tho effectiveness of the proposed solution has beon experimentally proved, in bboth the eases of an incrtial and a resonant load. Some (experimentally validated) working assumptions hhave been considered for the definition of te compensa tion procedure: in particular, the effects of the reluctance torque and the nonsinusoidal ux distribution in the air- zaps have been almost neglected. A future development of the present’ work consists of ‘understanding how the neglected terms (such ax the re- Inctance torque and high order harmonic components in the cogging torque) affect the torque ripple, and how they cant be effectively included in the minimization procedure, 6. ACKNOWLEDGEMENTS ‘The authors would like to thank Videotes (Schio, Vicenza, Italy) for providing (and allow us to hack) the camera positioning units, and Andrea Parisolto, Daniele Pelizeer, Marco Perwz20 and Mattia Sesiesazo for their valuable contribution to the experimental activity REFERENCES N. Bianchi and $. Bolognani. Design techniques for reducing the cogging torque in surface-mounted: PM motors, Industry Applications, IBEB Transactions on, '38(5):1259 ~ 1265, sep. 2002." ISSN 0093-9994, dot 10.1109/TTA.2002.802089. M. Boron, J.N. Chiaston, RIT. Novotnak, and R-B. Rekowski. High-performance noslincar feedback control fof a permanent magnot stepper motor. Control Systems Technology, IBEE Transactions on, 1(1):5 —M4, mar. 1993, D. Chen and B. Paden, Adaptive linearization of hybrid step motors: stability analysis. Automatic Control, IEEE Transactions on, 38(6):874 ~887, jun. 1993. ISSN ‘018-9286. doi: 10.1109/9.222300. 5, Favre, L. Cardoletti, and M, Jufer. Permanent-magnet synchronous motors: a comprehensive approach to cog- ging torque suppression. Industry Applications, IEEE ‘Transactions on, 29(6):1141 ~1149, nov. 1993." ISSN (0093-9994. doi: 10-1109/28.259725. G. Ferretti, G. Magnani, and P. Roceo. Modeling, identifi- cation, and compensation of pulsating torque in perma- nent magnet AC motors. Industrial Electronics, IEEE Drunsactions on, 45(6):912 ~820, dec. 1998. ISSN 0278- 046, doi: 10.1109/41.785385, D.C. Hanselman. Minimum torque ripple, maximum eff ciency excitation of brushless permanent’ magnet mo- tors. Industrial Blectronies, IEEE Transactions on, 41(3):292 B00, jun. 1994." ISSN 275-0046, doi 10,1109/41.293808, LLY, Hung and Z, Ding. Design of currents to roduce torque ripple in brushless permanent magnet motors. Electric Power Applications, IEE Proceedings B, 140(4):260 ~ 266, jul. 1993. ISSN 0143-7038 "T.M. Jahns and WL. Soong, Pulsating torque minimica- tion techniques for permanent magnet AC motor drives- a review. Industrial Electronics, IEEE Transactions on, 43(2):321 ~330, apr. 1996. ISSN 0278-0046. doi 10.1109/41.491356. 'T. Kenjo. Stepping Motors and Their Microprocessor Controls. Clarendon Press, Oxford, 1984. PF. Khorrami, P. Krishnamurthy, and I. Melkote. Mod. ling and adaptive nonlinear control of electric motors. Springer-Verlag, Berlin, 2003. Hoang Lo-Huy, Robert Perret, and Rene Feuillet. “Min. Imization of Torque Ripple in Brushless DC’ Motor Drives, Industry Applications, IEEE Transactions on, TA22(4)-748 ~795, jul, 1986," ISSN 0093-9994, do: 10.1109/TIA.1986.450478". D.G. Taylor. Nonlinear control of electric machines: an overview. Control Systems Magazine, IEEE, 14(6):41 — S51, dec. 1994, ISSN 0272-1708. doi: 10.1109/37.354414 AP.’ Wa and P.L. Chapman. Simple expressions for optimal current waveforms for permanent-magnet syn- ehronous machine drives. Energy Conversion, IEEE ‘Transactions on, 20(1):151— 157, mar. 2005, ISSN O885- 8069, doi: 10.1109/TEC.2004,587209,