Vous êtes sur la page 1sur 9

54

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 14, NO. 1, FEBRUARY 2005

Axial-Flux Permanent Magnet Machines for Micropower Generation


Andrew S. Holmes, Member, IEEE, Guodong Hong, and Keith R. Pullen
AbstractThis paper reports on the design, fabrication, and testing of an axial-ux permanent magnet electromagnetic generator. The generator comprises a polymer rotor with embedded permanent magnets sandwiched between two silicon stators with electroplated planar coils. Finite element simulations have been carried out using ANSYS to determine the effects on performance of design parameters such as the number of layers in the stator coils, and the rotor-stator gap. The effect of including soft magnetic pole pieces on the stators has also been studied. A prototype device with a diameter of 7.5 mm has been tested, and shown to deliver an output power of 1.1 mW per stator at a rotation speed of 30 000 rpm. The generator has been integrated with a microfabricated axial-ow microturbine to produce a compact power conversion device for power generation and ow sensing applications. [1265] Index TermsAxial-ow microturbine, energy harvesting, laser micromachining, permanent magnet machine, power generator, power MEMS.

I. INTRODUCTION HE proliferation of portable electronic devices in recent years has led to increasing interest in alternative power sources that might eliminate the need for chemical batteries. The motivation for developing such sources depends on application area and power requirements. For very low power applications, there is the possibility of powering electronics by energy scavenging or harvesting, i.e., extracting energy from the surroundings. The power available is generally low (typically at the level), but there is no need for the periodic recharging and replacement associated with batteries, leading to lower maintenance. At higher power levels, there is interest in replacing batteries by fuel-burning devices that offer longer shelf life and/or higher power density. Energy scavenging power generators based on microelectromechanical systems (MEMS) technology have been researched extensively over the past 10 yr. These have mostly been based on mass-spring-damper systems, where the damper is a transducer that converts mechanical energy to electrical energy. Devices of this type have been developed based on electrostatic [1], electromagnetic [2], [3], and piezoelectric [4] transduction mechanisms. Being resonant, these generators
Manuscript received February 3, 2004; revised June 1, 2004. This work was supported by the U.K. Engineering and Physical Sciences Research Council, by Grant GR/N18895 Microengineered Axial-Flow Pumps and Turbines. Subject Editor E. Obermeier. A. S. Holmes and G. Hong are with the Department of Electrical and Electronic Engineering, Imperial College London, London SW7 2BT, U.K. (e-mail: a.holmes@imperial.ac.uk). K. R. Pullen is with the Department of Mechanical Engineering, Imperial College London, London SW7 2BX, U.K. Digital Object Identier 10.1109/JMEMS.2004.839016

work most efciently when the applied excitation is a low amplitude, reciprocating motion at a frequency close to resonance. This makes them potentially very effective for extracting energy from vibrating machinery. Alternative devices containing only a mass and a damper are also under development. These have been shown to be more efcient in the case of large, low-frequency excitations such as might be experienced by a portable or wearable electronic device [5]. Work on fuel-burning MEMS power generators has been focused mainly on microturbines. The most highly developed work in this area is that of MIT, where there is an extensive gas turbine program [6], [7]. This project aims to develop a fully integrated device that combines compressor, burner, turbine, and electrical generator in a single chip fabricated as a multiwafer stack. Other workers have focused on turbogenerator subsystems. For example, in [8] a radial-ow turbine and an electrical generator were combined on a single chip. Recently at Imperial College, we have been developing a turbine-based power generator that combines an axial-ow turbine with an axial-ux electromagnetic generator, as shown in Fig. 1. The idea of this device is that it should be able to extract power directly from an externally generated gas ow, rather than relying on an internal burner and compressor. The axial-ow geometry is the obvious choice for this kind of generator, because it can operate at low pressure ratios [9]. Axial-ow microturbines are difcult to realize by conventional microfabrication methods, because to produce the necessary curved proles on the rotor blades and guide vanes [see Fig. 1(b)] it is necessary to fabricate structures where the sidewall angle evolves with depth in a controlled manner. We have solved this problem by using excimer laser micromachining with a variable aperture mask to dene the required blade proles in SU8 polymer parts preformed by UV lithography. Applications for our device are envisaged in areas such as wireless sensing, where the turbine could be used to power a remote sensor and a short-range radio transmitter. In the case of a ow sensor or an air-speed sensor, the turbine could also perform the sensing function, with the advantage over other sensor types of generating a frequency output. This paper focuses on the electromagnetic generator section of our device, which is based on an axial-ux, permanent magnet machine. The axial-ux geometry has been widely used in larger scale motors and generators [10], and also more recently in microfabricated motors [11]. It is particularly attractive for MEMS devices because it allows the use of planar coils on the stators, which are easier to fabricate than solenoids. In the following sections, we consider in detail the design and fabrication of the generator. ANSYS simulations are presented

1057-7157/$20.00 2005 IEEE

HOLMES et al.: AXIAL-FLUX PERMANENT MAGNET MACHINES

55

Fig. 2. Schematic of permanent magnet generator, showing design parameters. (a) Plan view. (b) Cross section along arc passing through coil centers.

Fig. 1. Microengineered turbo generator based on axial-ow turbine and axialux permanent magnet generator. (a) Schematic cross section through device. (b) Details of turbine section, showing rotor blade and guide vane shapes.

that show the effects on generator performance of design parameters such as the number of layers in the stator coils and the rotor-stator gap. The effect of including soft-magnetic pole pieces on the stators is also considered. The fabrication process is then described, and experimental results are presented for a prototype generator with a diameter of 7.5 mm. Finally, issues associated with future downscaling of the device are discussed. II. DEVICE DESIGN A. Basic Design Procedure Fig. 2(a) is a plan view of the permanent magnet generator, showing just one of the pole pairs, while Fig. 2(b) is a cross section through the same part of the device. The generator coil is spiral coils arranged in an ana two-layer structure, with nulus in each layer. Other stator geometries based on overlapped coils (see, for example, [12]) could be more efcient in terms of power generation, but were avoided because of potential process yield issues associated with producing many interlayer vias. In our design, each spiral coil is connected at its outer edge to an adjacent coil in the same layer, and by a single via at the center to the coil above or below it. For example, in Fig. 2, coils 1 and 2 and coils 3 and 4 are connected by vias [shown in Fig. 2(b)], while coils 2 and 3 are connected by an in-plane link [shown in Fig. 2(a)]. In this way, all the coils are connected in series, except at one point where the in-plane link is replaced by tracks to external contacts. The various design parameters are dened in

Table I, where the values used in the rst prototype devices are also shown. The overall scale of the rst prototypes was set by the decision to use conventional ball race bearings (BOCA type SMF681, ) and commercial neodymium boron iron (NdBFe) permanent magnets (CERMAG grade N30H). The bearing used had a ange with an outer diameter of 3.8 mm, and the value of was chosen to allow a 200clearance between the outer edge of the bearing ange and the inner radius of the coil annulus. The minimum radius for the magnet trajectory, based on the requirement that the magnets should lie inside the coil annulus for , while the maxmaximum induction, is imum number of pole pairs at this radius is bounded roughly by . When , , and . In terms these expressions give of output power, it is generally advantageous to maximize the number of pole pairs. However, a lower number of pole pairs was used to improve the mechanical robustness of the polymer rotors, and to facilitate rotor assembly. Once and had been dened, the spiral coils were constructed, with the number of turns being set by the interturn pitch and the coil size in the circumferential direction, allowing for a via post at the was chosen for ease of center. A relatively large pitch of 60 value of 12 turns per spiral. processing, giving an B. Scaling Considerations The induced output voltage of any permanent magnet gener, where is the total magnetic ator is given by ux linkage in the stator coils. For our device we can express the maximum ux linkage for each stator in the form (1)

56

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 14, NO. 1, FEBRUARY 2005

TABLE I DESIGN PARAMETERS FOR PERMANENT MAGNET GENERATOR, WITH VALUES USED IN PROTOTYPE DEVICES

where is the remnant ux density of the permanent magnets, is the area enclosed by the boundary of each coil, and is a geometrical factor of order unity that depends on (a) the precise layout of the turns in each coil, (b) the exact form of the ux distribution around each permanent magnet, and (c) the gap between rotor and stator. The factor is simply the number of spiral coils on each stator. Assuming has a harmonic time dependence, as will be approximately true for any well-designed generator, we can then write the rms output voltage as (2) where is the angular rotation speed of the rotor. The available output power per stator is then given by (3) where as is the total winding resistance, which may be expressed

(4) In (4), is the conductivity of the coil metal (copper in our case), and and are the width and height of the coil tracks. represents the total track length of each The term stator coil, being a second geometrical factor depending on the coil shape and layout of turns. Using (2)(4), we can establish some basic scaling rules for the generator. In particular, if the entire device is scaled in size without any change in geometry, so that only , and are affected by scaling, then the following rules will apply (5) scaling where is any characteristic linear dimension. The of output power imposes some severe limitations on the minimum useful size for generators of this type. For example, in this paper we demonstrate cm-scale devices with output power levels of about 1 mW. At constant maximum rotation speed,

downscaling of these devices by a factor of only four would re, which is on the limit duce the output power to around 1 of what is useful with current electronics. The detrimental effects of length-scaling can be offset by increasing the rotation speed in smaller devices, but this is not feasible with conventional bearings. Some care must be exercised when using the above equations to explore the scaling effects of parameters other than device size, because the combined effects of , , and may not be as rst expected. Moreover, (2) does not remain valid under all conditions. To illustrate this point we consider the case where the number of poles is varied by adding or removing permanent magnets. In this case, assuming the stator coils are reshaped as varies so as to keep the coil annulus lled, then changes will be largely compensated by changes in , so that the in product will remain roughly constant. Under these condi. In practice, however, changes in tions (2) predicts the shape of the induced voltage waveform as the number of being poles is varied result in the actual scaling index for closer to 1.5. Under the same conditions shows little varia(because ), so the available tion with . power scales roughly as C. Numerical Simulations While (2)(4) are useful in showing the functional dependence of output voltage and power on various design parameters, they cannot yield quantitative information unless the values of and are known. Such quantitative information is key to and the design of miniature power generators, particularly as are likely to be small, and there is a risk that the device may not generate a useful level of output voltage at the design speed. For self-primed operation (i.e., in the absence of a permanent voltage source such as a battery), it is helpful if the generator can produce 1 V pk-pk or more. Assuming the track width and (to maintain conheight are scaled in inverse proportion to stant line/space and aspect ratios), the output voltage and power and respectively, so to maximize scale as that power output we should choose the smallest value of will satisfy any minimum voltage constraint.

HOLMES et al.: AXIAL-FLUX PERMANENT MAGNET MACHINES

57

Fig. 3. Flux distributions for a one-dimensional (1-D) array of permanent magnets, calculated using ANSYS. (a) and (b) Overall distributions with and without soft magnetic pole pieces. (c) and (d) Effect of increasing the gaps between the permanent magnets and the pole pieces.

Calculation of is simply a matter of geometry once the coil shape and turn layout are known. A very approximate value for can also be obtained relatively easily, without solving the magnetic problem, simply by assuming that the magnetic ux in any coil region lying density is uniform and equal to directly over a permanent magnet, and zero elsewhere. is the ux density on the axis at the end of a long cylindrical magnet. Given this assumption, numerical calculation of the ux linkage for the geometry in Fig. 2 is straightforward. This approach is fast, and gives surprisingly good results in spite of the gross nature of the approximations involved. For more accurate calculations, the actual ux density distribution experienced by the stator coils must be known. We have carried out nite element analysis (FEA) simulations using ANSYS software to predict the variation of output voltage with , and to determine the effect on perthe rotor-stator gap formance of the soft magnetic pole pieces. Similar calculations have been reported previously for larger scale axial-ux generators [13]. However, these are typically based on a slightly different geometry, consisting of a single stator with a pair of coupled, permanent magnet rotors placed either side of it. This latter design is more efcient because it avoids eddy current losses in the magnetic materials. However, it is more difcult to realize by microfabrication, because it requires the stator coils to be fabricated on a thin membrane. In qualitative terms, it is expected that the soft magnetic pole piece will tend to reduce the lateral ux leakage from each magnet pole, thereby increasing the induced output voltage. On

the other hand, increasing the rotor-stator gap is expected to reduce the output voltage, because the ux density will decay with axial distance from the magnet poles. These effects are illustrated in Fig. 3, which shows magnetic ux plots for an array of permanent magnets (a) without a pole piece, and (b-d) with a pole piece at different distances. These plots were produced by a two-dimensional (2-D) simulation, but a threedimensional (3-D) simulation would produce similar results in qualitative terms. The plots in Fig. 3 neglect the effects of induced eddy currents in the pole pieces. Such currents will inevitably arise whenever the rotor is moving if the pole pieces are conducting. Eddy currents lead to power dissipation, reducing the generator efciency, so conducting pole pieces should be avoided if maximum efciency is to be achieved. Eddy currents also produce an additional magnetic eld that modies the ux distribution both inside the pole piece and, to a lesser extent, in the rotor-stator gap. However, the change in the output voltage amplitude due to this effect is less than 1% for our device at a rotation speed of 30 000 rpm. Consequently we have ignored eddy currents in our output voltage simulations. Fig. 4 shows the ANSYS model set up for full 3-D simulation. Only the permanent magnets and the pole pieces were included in the model; all other parts of the structurestator silicon, SU8 polymer layers, and copper coilswere assumed and modeled as air. Based on the manufacto have turers data, the permanent magnets were assumed to have a value of 1160 mT, and a coercivity of 880 kA/m, while the

58

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 14, NO. 1, FEBRUARY 2005

Fig. 4. Physical model of permanent magnet generator used in ANSYS calculations. Silicon, copper and SU8 are assumed to have  = 1, and are treated as air. Generator output is assumed to be open circuit.

the rate of this decay is not affected signicantly by the presence of the soft magnetic pole piece. The pole piece enhances , correthe output voltage by roughly 30% at a gap of 100 sponding to a 70% increase in output power. This effect gradually becomes less pronounced as the gap increases. The output voltage per stator at a speed of 30 000 rpm (nominal shaft speed for prototype turbine) for a generator with a pole piece and a is expected to be 1.88 V pk-pk. With rotor-stator gap of 120 a stator winding resistance of 45 , this corresponds to an available power of 2.45 mW, which is more than adequate for power generation in many low-power applications. The above winding resistance value is calculated using the parameters in Table I, as. suming copper coils with a conductivity of
Fig. 5. Calculated variation of output voltage with rotor-stator gap, with and without pole pieces. Y axis is peak-to-peak induced voltage in mV for one stator, divided by rotation speed in rad=s.

III. FABRICATION In this section, we briey describe the processes used to fabricate the rotors and stators for the rst prototype generators. These rely on a combination silicon deep reactive ion etching (DRIE), multilevel electroplating, SU8 processing, and laser micromachining. A. Stator Fabrication The stators were produced by double-sided processing of 4 -dia, 500- -thick silicon substrates. First, 300- -deep annular cavities were etched in the backside of the wafer by DRIE using a photoresist mask. After reoxidation and sputter deposition of a Cr/Cu seed layer, the cavities were lled with nickel by electroplating [see Fig. 6(a)]. A dry-lm photoresist mask was used to conne the electrodeposition to the cavities. In a second process step, two-layer planar coils were fabricated on the front side of the wafer by multilayer UV lithography and copper electroplating [Fig. 6(b)]. The insulation layer between the two layers of each coil, and a top protection layer, were made by SU8 photolithography. When dening the coils,

soft magnetic material was assumed to be linear with a relative permeability of 600 (Nickel). The magnetic problem was solved using the Vector Potential method [14]. This calculation was done once for each rotor-stator gap. The ux linkage for a single coil was then determined for a series of angular positions, and differentiated numerically to determine the rate of change of ux linkage with angle. From this function the output voltage for the complete stator was calculated. The simulation results are shown in Fig. 5, where the peak-topeak output voltage per stator has been plotted as a function of rotor-stator gap, with and without a soft magnetic pole piece. Note the output voltage is normalized to unit angular rate, and . The corresponding values, expressed in units of inferred from (2) (assuming a sinusoidal waveform), are shown on the right-hand y axis. It is clear from Fig. 5 that the output voltage falls off steeply with increasing rotor-stator gap, and that

HOLMES et al.: AXIAL-FLUX PERMANENT MAGNET MACHINES

59

Fig. 6. Stator cross sections. (a) After backside processing. (b) After fabrication of two-layer coils. (c) After through-wafer etching for uid channel.

Fig. 8. Rotor fabrication process. (a) After deposition and patterning of Cu sacricial layer. (b) After two-layer SU8 lithography. (c) After sacricial release.

second level of coil plating to be carried out on a surface with relatively extreme topography, but eliminated the need for an SU8 plasma etching step. Both processes were shown to be capable of fabricating viable two-layer stator coils. B. Rotor Fabrication The rotors were fabricated by a combination of SU8 lithography and laser micromachining. The lithography stage was performed on 4 -dia silicon wafers on which a 10- -thick patterned copper sacricial layer had previously been dened [Fig. 8(a)]. Two layers of SU8, each 500- -thick after processing, were deposited and patterned to produce a 1-mm-thick rotor structure with cylindrical cavities for the permanent magnets and a stepped cavity at the center to accommodate the shaft [Fig. 8(b)]. After the second layer of lithography, the rotors were released from the wafer by wet etching of the sacricial layer [Fig. 8(c)]. The turbine rotor blades, which required curved proles, were produced by excimer laser micromachining with a dynamically variable mask. Crosslinked SU8 is well suited to machining by laser ablation at 248 nm wavelength, producing relatively clean structures with good surface nish. With the setup available, the blades had to be machined individually, with two machining operations being required for each blade, one for the convex side and another for the concave side. A xed mask was used to protect other regions of the rotor from exposure, and a moving mask was translated, under computer control, in front of the xed mask so that different parts of the blade received different exposures. This process has been described in detail elsewhere [21]. Fig. 9 shows an SEM image of a rotor after the laser machining process. Following laser machining, permanent magnets were manually inserted into the cavities in the rotor, and secured in place with SU8 which was applied using a needle and then cured by UV exposure and heating. For rotors to be used in turbo generator tests, a precision-machined steel shaft was also attached by a similar method. Each turbo generator was assembled on a metal jig with integral pins to align the two stators. An SU8 spacer was used to dene the gap between the stators, as shown in Fig. 1(a).

Fig. 7. SEM photograph showing top layer of a stator coil prior to deposition of the protective SU8 layer.

an overplating method based on thin photoresist masks was used to form mushroom-shaped tracks. These were coated by evaporation with Cr to improve the adhesion of subsequent SU8 high, with a width layers. The coil tracks were nominally 12 and a pitch of 60 . Fig. 7 shows an SEM photoof 30 graph of a nished coil prior to deposition of the top protection layer. MEMS coil fabrication processes have been reported previously for numerous other devices such as magnetic sensors [15], [16], magnetic actuators [17], micromotors [18], inductors, and transformers [19]. The problem was complicated in our work by the fact that parts of the wafer surface had to be left free of SU8 at the end of the coil fabrication process to allow through-wafer etching by DRIE. This was essential for the intended integration of the generator with an axial-ow turbine. In early prototypes, windows were opened in the SU8 following coil fabrication by plasma etching using an electroplated nickel mask [20]. Subsequently, however, a more efcient process was developed in which the windows were dened lithographically in each successive layer of SU8 during coil fabrication. This required the

60

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 14, NO. 1, FEBRUARY 2005

Fig. 9. SEM photograph showing an SU8 rotor after laser micromachining of the turbine blades.

Fig. 11.

Oscilloscope trace showing output waveform from a single stator.

Fig. 10. Test setup used to measure the variation of output voltage with rotor-stator gap.

IV. CHARACTERIZATION Fig. 10 illustrates the setup used for characterization of prototype generators. For these tests, the rotor was driven by a small dc motor. A soft magnetic disc was attached to the motor shaft, and magnetic forces due to the embedded permanent magnets were used to hold the rotor in position against the disc. The motor was mounted with its axis horizontal, and positioned adjacent to a vertically mounted stator. The axial position of the stator could be adjusted by means of a micropositioner. Fig. 11 is a captured oscilloscope trace showing the time variation of the generator output voltage at a rotation speed of around 7200 rpm. The time variation is close to sinusoidal, with very low harmonic distortion, conrming that the assumptions made in deriving (2) are valid for our device. Fig. 12 compares the measured and simulated variations of output voltage with rotation speed for stators with and without soft magnetic pole pieces. The rotor-stator gap was 120 throughout. In all cases the output voltage is proportional to rotation speed, as expected from (2). Furthermore, the enhancement due to the pole piece is similar in the experimental and simulated cases, being 33% in the former and 27% in the latter. However, the simulated voltages are consistently higher than the experimental ones, the ratios of simulated to measured voltages being 1.53 and 1.61, respectively, for the stators with and without pole pieces. This discrepancy is attributed primarily to lower than expected magnetization in the permanent magnets. Evidence for this comes from remnance measurements made on a random sample of ten magnets taken from the same batch
Fig. 12. Comparison between simulated and measured variations of output voltage with rotation speed for stators with and without soft magnetic pole pieces. Rotor-stator gap is 120  in all cases.

as those used in the prototypes. The ten magnets analyzed had a mean Br value of 767 mT with a standard deviation of 46 mT. The mean value is lower than the manufacturers stated Br of 1160 mT by a ratio of 1.51, which is very close to the simulated-to-measured voltage ratio of 1.53 obtained for the device with pole pieces. The small difference between the simulated-to-measured voltage ratios observed with and without pole pieces is probably due to differences in the rotor-stator gap, . which was measured only to within Fig. 13 compares measured and simulated variations of the normalized output voltage with rotor-stator gap for a stator with a soft magnetic pole piece. In each case, the voltage has been . normalized to give a value of 1 at the minimum gap of 120 The extremely close agreement between the simulated and experimental trends lends further support to the suggestion that the discrepancies in the absolute voltage levels in Fig. 12 are due to value rather than, for example, geometrical discrepa lower ancies between the simulated and the experimental devices. In addition to the generator tests reported here, preliminary tests of completed turbo generators have been carried out, using a test jig for measuring the induced voltage and pressure drop as a function of nitrogen ow rate. Operation at pressure drops down to a few mbar has been demonstrated, indicating that the

HOLMES et al.: AXIAL-FLUX PERMANENT MAGNET MACHINES

61

Fig. 15. Fig. 13. Comparison between simulated and measured variations of normalized output voltage with rotor-stator gap for a stator with a soft-magnetic pole piece.

two layers.

Grc = 120 m and Gcc = 25 m. Y axis is normalized to output power for

Variation of output power with number of coil layers, assuming

Fig. 14. Measured variations of rotation rate and output voltage for a prototype turbo generator driven by nitrogen.

devices are well suited to extraction of power from ambient gas ows. The results of these tests have been reported in detail in a separate paper [22]. Fig. 14 shows the measured variation with nitrogen ow rate of the rotation speed and output voltage for a typical device. This particular turbine achieves a rotation rate of 30 000 rpm at a pressure drop of around 8 mbar and a ow rate of about 35 L/min. At 30 000 rpm the open-circuit voltage per stator is 1.19 V pk-pk, corresponding to an available power of 1.1 mW per stator into a matched (40 ) load. This generator performance is marginally worse than the best achieved in the generator tests (290 mV at 7000 rpm, corresponding to 1.24 V at 30 000 rpm) because the assembly method used in the com. pleted devices did not allow rotor-stator gaps down to 120 V. DISCUSSION This paper has presented design calculations, simulations and initial experimental results for a mm-scale axial-ux electromagnetic generator aimed at easy integration with an axial-ow microturbine. A fabrication process has been developed that involves deep silicon etching and multilayer electroplating for the stator parts, and SU8 processing for the rotors. Experimental measurements on prototype generators show close agreement

with ANSYS simulations in terms of trends, although the measured output voltages are consistently lower than the simulated ones by around 35%. This discrepancy is due to the average remnant ux density of the permanent magnets being lower than stated in the manufacturers data. The mW power levels generated by the prototype devices are more than adequate for many remote sensing applications, and the output voltages are sufcient to allow self-primed operation. Higher output powers could be achieved from devices of the same size and basic design by a number of routes, for example by reducing the rotor-stator gap, enhancing the coil ll-factor (ratio of copper to dielectric in the stator coils), increasing the number of poles, or increasing the number of coil layers on each stator. All of these changes would make fabrication or assembly of the device more difcult, and each would afford only a modest increase in output power. However, the combined effect could be signicant. For example, a device sim(compared to ilar to ours but with a rotor-stator gap of 25 ), a coil ll factor of 50% (compared to 25%) and eight 120 poles (compared to 5) would generate around 9 mW per stator at 30 000 rpm. Increasing the number of coil layers while maintaining the same interlayer coil spacing could achieve a further threefold increase in power, as illustrated in Fig. 15. In the absence of ux leakage, the output power would be proportional to the number of coil layers. In practice, however, beyond about eight coil layers the gap between any additional coils and the rotor is so large that ux leakage renders them ineffective. Thus, a device with an output power of around 25 mW per stator at 30 000 rpm should be achievable without changing the basic conguration. The main factor limiting the output power of such a device would be the maximum operating speed of the ball bearings. Commercial permanent magnets and bearings were used in the rst prototypes to shorten the development cycle. In future work, we will investigate the use of permanent magnets that can be deposited at wafer level prior to sacricial release of the rotor. This will eliminate the manual assembly and bonding operations associated with the current design; no other changes to the rotor fabrication process will be required. Printed magnets based on a powdered magnetic material in a polymer binder will probably be used here [23], as these appear to be capable of higher

62

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 14, NO. 1, FEBRUARY 2005

remnant ux densities than electroplated materials [24]. We will also explore the possibility of replacing the ball bearings with microfabricated aerodynamic bearings. Signicant progress on bearings of this type has been reported by other workers in recent years [7]. With these two changes, downscaling of the devices will become a possibility. As noted earlier, scaling laws do not favor a signicant reduction in size. However, a device smaller by a factor of 2 or 3 (in all linear dimensions) could still yield output power levels useful for some sensing applications, particularly if downscaling could be accompanied by an increase in rotation speed. ACKNOWLEDGMENT The authors are grateful to Dr. J. Stagg of Imperial College for his help with DRIE. REFERENCES
[1] S. Meninger, J. O. Mur-Miranda, R. Amirtharajah, A. P. Chandrakasan, and J. H. Lang, Vibration-to-electric energy conversion, IEEE Trans. Very Large Scale Integr. (VLSI) Syst., vol. 9, no. 1, pp. 6476, 2001. [2] C. B. Williams, C. Shearwood, M. A. Harradine, P. H. Mellor, T. S. Birch, and R. B. Yates, Development of an electromagnetic micro-generator, IEE Proc. Circuit Device Syst., vol. 148, no. 6, 2001. [3] J. M. H. Lee, S. C. L. Yuen, W. J. Li, and P. H. W. Leong, Development of an AA size transducer with micro resonators, in Proc. ISCAS 2003, vol. 4, Bangkok, Thailand, May 2528, 2003, pp. 876879. [4] P. Glynne-Jones, S. P. Beeby, and N. M. White, Toward a piezoelectric vibration-powered microgenerator, IEE Proc. Sci. Meas. Technol., vol. 148, no. 2, 2001. [5] P. D. Mitcheson, P. Miao, B. H. Stark, A. S. Holmes, E. M. Yeatman, and T. C. Green, Analysis and optimization of a MEMS electrostatic on-chip power supply for self-powering of slow-moving sensors, in Proc. Eurosensors XVII, Guimaraes, Portugal, Sep. 2124, 2003, pp. 492495. [6] A. Mehra, X. Zhang, A. A. Ayn, I. A. Waitz, M. A. Schmidt, and C. M. Spadaccini, A six-wafer combustion systems for a silicon micro gas turbine, J. Microelectromech. Syst., vol. 9, no. 4, pp. 517527, 2000. [7] A. H. Epstein, Millimeter-scale, MEMS gas turbine engines, in Proc. ASME Turbo Expo 2003, Power for Land, Sea and Air, Atlanta, GA, Jun. 1619, 2003, pp. 128. [8] T. G. Wiegele, Micro-turbo-generator design and fabrication: a preliminary study, in Proc. IECE 96, vol. 4, 1996, pp. 23082313. [9] B. S. Massey, Mechanics of Fluids, 6th ed: Routledge, 1989. [10] A. Cavagnino, M. Lazzari, F. Profumo, and A. Tenconi, A comparison between the axial-ux and the radial ux structures for PM synchronous motors, IEEE Trans. Ind. Appl., vol. 38, no. 6, pp. 15171524, 2002. [11] N. Achotte, H. Raisigel, J. Delamare, and O. Cugat, Characterization of planar brushless micro-motors/micro-generators, in Proc. 6th JapanFrance and 4th Asia-Europe Congress on Mechatronics, Saitama, Japan, Sep. 912, 2003, pp. 179183. [12] P.-A. Gilles, C. Locatelli, O. Cugat, and J. Delamare, Planar double-layer coil for brushless micromotor, in Proc. Mecatronics 01, Besanon, France, Oct. 911, 2001, pp. 377382. [13] T. S. El-Hasan, P. C. K. Luk, F. S. Bhinder, and M. S. Ebaid, Modular design of high-speed permanent-magnet axial-ux generators, IEEE Trans. Magnetics, vol. 36, no. 5, pp. 35583561, 2000. [14] K. Hameyer and R. Belmans, Numerical Modelling and Design of Electrical Machines and Devices. Southampton, U.K.: WIT Press. [15] R. Puers, G. Vandevoorde, and D. De Bruyker, Electrodeposited copper inductors for intraocular pressure telemetry, J. Micromech. Microeng., vol. 10, no. 2, pp. 124129, 2000. [16] V. Seidemann and S. Buettgenbach, Fabrication technology for closely coupled micro coils with integrated ux guidance and their application to proximity and magnetoelastic force sensors, IEEE Sensors J., vol. 3, no. 5, pp. 615621, 2003. [17] L. K. Lagorce, O. Brand, and M. G. Allen, Magnetic microactuators based on polymer magnets, J. Microelectromech. Syst., vol. 8, no. 1, pp. 29, 1999. [18] Y. Watanabe, M. Edo, H. Nakazawa, and E. Yonezawa, A new fabrication process of a planar coil using photosensitive polyimide and electroplating, Sens. Actuators, vol. 54, pp. 733738, 1996.

[19] M. Brunet, T. ODonell, J. OBrien, P. McCloskey, and S. C. Mathuna, Thick photoresist development for the fabrication of high aspect ratio magnetic coils, J. Micromech. Microeng., vol. 12, no. 4, pp. 444449, 2002. [20] G. Hong, A. S. Holmes, and M. E. Heaton, SU8 resist plasma etching and its optimization, in Proc. Symp. Design, Test, Integr. Packaging MEMS/MOEMS, MandelieuLa Napoule, France, May 57, 2003, pp. 268271. [21] A. S. Holmes, M. E. Heaton, G. Hong, K. R. Pullen, and P. T. Rumsby, Laser proling of 3D microturbine blades, in Proc. LPM 2003, vol. 5063, SPIE, Munich, Germany, Jun. 2124, 2003, pp. 152156. [22] A. S. Holmes, G. Hong, K. R. Pullen, and K. R. Buffard, Axial-ow microturbine with electromagnetic generator: design, CFD simulation, and prototype demonstration, in Proc. 2004 IEEE MEMS Conf., Maastricht, The Netherlands, Jan. 2529, 2004, pp. 568571. [23] M. J. Kramer, Y. Xu, K. W. Dennis, I. E. Anderson, and R. W. McCallum, Development of improved powder for bonded permanent magnets, IEEE Trans. Magnetics, vol. 39, no. 5, pp. 29712973, 2003. [24] H. J. Cho and C. H. Ahn, A bidirectional magnetic microactuator using electroplated permanent magnet arrays, J. Microelectromech. Syst., vol. 11, no. 1, pp. 7884, 2002.

Andrew S. Holmes (M02) received the B.A. degree in natural sciences from Cambridge University, U.K., in 1987, and the Ph.D. degree in electrical engineering from Imperial College London, U.K., in 1992. He is currently a Reader with the Optical and Semiconductor Devices Group, Department of Electrical and Electronic Engineering, Imperial College London. His research interests are in the areas of micropower generation and conversion, MEMS devices for microwave applications, and laser processing for MEMS manufacture.

Guodong Hong received the B.Eng. and M.Eng. degrees in material engineering from Shandong University, China, in 1984 and 1989, respectively, and the Ph.D. degree in mechanical engineering from Tsinghua University, China, in 1998. He is currently a Research Associate with the Optical and Semiconductor Devices Group, Department of Electrical and Electronic Engineering, Imperial College London. His areas of research interests include the micropower generation and conversion, micromoulding, rapid prototyping, and manufacturing for microelectromechanical systems.

Keith R. Pullen received the rst degree in 1987 and the Ph.D. degree in 1990 from Imperial College London, U.K., the latter being funded on full salary by Rolls-Royce plc for investigations into small high-speed turbines and generators. He joined Noble Denton in 1990, where he was in the eld of offshore engineering working in various locations, including Norway. In 1992 he returned to Imperial College as a Lecturer to continue research into small high-speed electrical machines and was a cofounding member of the Turbo Genset Company being employed as Chief of Design. He has undertaken engineering consultancy work for more than 15 companies including BP, BOC Edwards, Honeywell, and Nortel Networks in turbomachinery. He is an author of 50 papers in international journals and conferences and has 21 patents granted or pending. He is currently a Senior Lecturer and has recently been appointed Director of Engineering for Hydroventuri Ltd. As well as teaching subjects such as gas turbine technology, he manages a group of eight researchers and an experimental laboratory of 250 , including several large facilities. Dr. Pullen earned three prizes, including top undergraduate sponsored and graduate, while with Rolls-Royce.

Vous aimerez peut-être aussi