You are on page 1of 11

I.

INTRODUCTION

Random Hysteresis PWM Inverter With Robust Spectrum Shaping

B. J. KANG C. M. LIAW, Member, IEEE National Tsing Hua University Taiwan

Hysteresis current-controlled pulsewidth modulation (PWM) scheme is simple and robust in switching control, but it is subject to not constant switching frequency. However, since its harmonic spectrum is not randomly distributed like those yielded by the random PWM (RPWM) scheme, the annoying harmonic effects still exist. A random switching control strategy is proposed here to let the harmonic spectrum of the hysteresis current-controlled PWM inverter be randomly distributed. First, the effect of varying band on the harmonic distribution characteristic of a hysteresis current-controlled PWM scheme is analyzed. Then a random hysteresis PWM scheme and its quantitative design procedure are proposed. Finally, a robust spectrum shaping technique is devised such that the desired harmonic distribution pattern can be achieved. In the resulting frequency spectrum, the magnitudes of lower frequency harmonics are shaped and reduced. Hence the inverter output with better harmonic attribute is obtained. Validity of the proposed random hysteresis PWM scheme is verified by some simulated and measured results.

Manuscript received March 6, 2000; revised November 6, 2000; released for publication December 6, 2000. IEEE Log No. T-AES/37/2/06334. Refereeing of this contribution was handled by W. M. Polivka. This research was supported by the National Science Council, ROC, under Grant NSC 85-2213-E-007-038. Authors addresses: B. J. Kang, Department of Power Mechanical Engineering, National Tsing Hua University, Hsinchu, Taiwan, ROC; C. M. Liaw, Department of Electrical Engineering, National Tsing Hua University, Hsinchu, Taiwan, ROC.

c 2001 IEEE 0018-9251/01/$10.00

Current control is indispensable for a pulsewidth modulation (PWM) inverter-fed electromechanical system having high driving control performance. The conventional current control methods for PWM converters can be broadly classified into the hysteresis, the ramp comparison, and the predictive controls [1, 2]. Among these, the hysteresis current control is very simple and provides adaptive and robust current tracking control performance. Compared with other approaches, the hysteresis current-controlled converter yields varying output harmonic spectrum, which is dependent on the variations of input voltage, back electromotive force (EMF), and load inductance [3, 4]. However, their output harmonic spectrum still gathers around a fixed frequency. It follows that the annoying acoustic noise and the susceptibility of mechanical vibration caused by harmonics still exist. One approach to alleviate these phenomena is to reduce the hysteresis band and thus increase the dominant harmonic frequency range beyond the audible and the mechanical resonant vibration frequency ranges, but the switching loss will be increased significantly. The random pulsewidth modulation (RPWM) technique is an effective approach to let the inverter-fed actuators possess low acoustic noise and mechanical vibration caused by harmonics when lower switching frequency is chosen [5]. The chief reason is that the harmonic spectrum of an RPWM inverter can be dispersedly and continuously distributed. Depending on the ways to let the time positions of switching signals for inverter switches be randomly varied, the RPWM techniques can be roughly classified into randomized switching frequency [6, 7], randomized pulse position [8], and random switching [9]. In the recent years, the research concerning the RPWM has continuously attracted much attention. In [10], Covic and Boys investigated the effects of the random switching frequency on probability density function for RPWM is the harmonic spectrum distributed characteristics. In [11], the analytical expressions have been derived for facilitating the understanding of frequency domain characteristics of a converter employing the random lead-lag pulse position technique and the random switching frequency technique. In [12], the authors proposed a new family of random modulation techniques for the three-phase converters operating at fixed switching frequency, but the pulses are randomly positioned within the switching period. As to the research made in [13], for the ease of obtaining the random signal, a random slope PWM inverter employing the existing system background noise is developed. As to current-controlled PWM inverters, Jacobina, et al. [7] evaluated the feasibility of incorporating the current control in an RPWM inverter and compared
APRIL 2001 619

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 37, NO. 2

the control quality with that in a deterministic PWM inverter. Comparative analysis of the results showed the effective current control in all the investigated cases, but a higher variance of control error was yielded when RPWM was employed. Until now, research of RPWM based on a hysteresis current-controlled PWM scheme is still rarely made. In [14], the concept of a multilevel hysteresis current-source control has been set forth. However, it still belongs to the deterministic approach. In this paper, the study of the random switching for a hysteresis current controller is performed. With adequate tuning and control of the hysteresis band, the harmonic spectrum of inverter output can be controlled and randomly distributed in a desired pattern. To achieve this goal, the influence of randomly varying hysteresis band on the harmonic distribution characteristics is first intuitively analyzed and observed. Then, the quantitative design and implementation of the proposed random hysteresis current-controlled PWM scheme are made. In addition, to get better low-frequency range harmonic distribution characteristic, a robust spectrum-shaping technique is further proposed to let the inverter output have the desired harmonic spectrum distribution profile. While the proposed random hysteresis current-controlled PWM scheme is applicable to all kinds of converters, it is applied here to an inverter-fed electrodynamic shaker [15] as an example. Some simulated and experimental results are provided to demonstrate its effectiveness using both commercial random signal generator and ROM-based (read-only memory) random signal generator. II. SPECTRAL ANALYSIS OF HYSTERESIS CURRENT-CONTROLLED PWM INVERTER

Fig. 1. Hysteresis current-controlled PWM inverter-fed shaker system. (a) System configuration. (b) Typical current and voltage waveforms.

For a PWM inverter-fed actuator, the application of current control in its switching mechanism can lead to simplified system dynamic behavior and better electromagnetic energy conversion performance. For the ease of explanation of the proposed switching control methodology, a hysteresis current-controlled PWM inverter-fed electrodynamic shaker system [15] shown in Fig. 1 is employed as an example. Shaker, a vibration proof testing equipment, is traditionally excited employing linear power amplifier with clean power. However, it possesses the disadvantages of low efficiency and large size and weight. Therefore, it is desirable to use inverter as an alternative [15]. To achieve this goal, sophisticated current control should be applied and an output filter is usually added in the inverter output to obtain lowly distorted exciting current over wide operating frequency range, typically 5 2000 Hz and even higher. For motor drives, which have significant low-pass filtering effect in mechanical load, the output filter usually can be omitted. Generally, the load model of all
620

electromechanical loads can be represented by a serially connected inductance-resistance-back EMF branch as shown in Fig. 1(a). The estimated parameters of the constituted parts of the shaker system are [15]: L0 = 0:32 mH, C0 = 0:8 F, V d = 80 V, mass of shaker table with load m = 0:545 kg, spring constant k = 13143 N/m, damping coefficient c = 3:54 N s/m and force generating constant = 12:3 N/A. The resistance and inductance of the shaker armature winding, which are varied with the operating frequency, are reasonably represented by [15] 1:30 + 0:19 log f ( ), 5 Hz f 45 Hz R= 0:27 + 0:81 log f ( ), 45Hz f 2000 Hz (1) and
L= 2:75 0:97 log f (mH), 0:96 0:26 log f (mH), 5 Hz f 330 Hz

(2) respectively. Owing to the inherent frequency-varying characteristics of R and L, the switching behavior of the current-controlled PWM scheme will be much affected. And moreover, the load resonance is easily excited by the harmonics. The voltage equations of the inverter-fed system shown in Fig. 1(a) can be written as vpwm = Re iL + L0 diL + v0 dt di v0 = Ri0 + L 0 + ve dt (3) (4)
APRIL 2001

330 Hz f 2000 Hz

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 37, NO. 2

denotes the current command and Ks is the where iL scaling factor of current sensor. Referring to Fig. 1(b), the instantaneous switching period tsw is equal to t1 + t2 and the instantaneous switching frequency fsw is equal to 1=tsw . Combining (5) and (6) and rearranging, we can obtain " # 2 2 L v di V 1 0 L fsw = d 1 0 + : (7) 2 4L0 h L0 Ks dt V d

where L0 is the inductance of the output filter, Re is the series equivalent resistance of L0 , L and R are, respectively, the armature winding inductance and resistance, ve is the back EMF of the shaker and v0 is the shaker armature terminal voltage. If the output filter is not needed, it is simple to let the parameters Re , L0 , and v0 in (3) be replaced by R , L, and ve , respectively. Fig. 1(b) sketches the current and voltage waveforms of a hysteresis current-controlled PWM scheme with fixed band. Depending on the switch operating status, the following two voltage equations can be written from (3) and Fig. 1(b) in the respective switching intervals t1 and t2 : 8 di+ > < L0 L = V d v0 dt (5) > : diL L0 = V d v0 dt where V d is the dc-link voltage, and the resistance voltage drop is neglected owing to the high frequency switching characteristic. From (5) and Fig. 1(b), we can obtain the relationship between the hysteresis band h and the current changes as follows: 8 di+ 1 diL L > t > < dt 1 K dt t1 = 2h s (6) > > : diL t2 1 diL t2 = 2h dt Ks dt

Fig. 2. Proposed random hysteresis PWM scheme. (a) Block diagram. (b) Randomized-band hysteresis comparator.

spectrum can be dispersedly and continuously distributed. The practical analysis and design of a proposed random hysteresis PWM scheme is introduced in the next section. III. PROPOSED RANDOM HYSTERESIS PWM INVERTER A. Random Hysteresis Current-Controlled PWM Scheme The proposed random hysteresis current-controlled PWM scheme is shown in Fig. 2(a). Therein the hysteresis comparator with randomly varying band is realized in Fig. 2(b). The feedback ratio of the closed-loop comparator is varied by a PWM switching controlled resistor. The control signal vcont for determining the duty ratio is composed of a dc control voltage vcont,0 and a random signal ns (t) with zero mean value. A CMOS switch IC is employed to realize the on-off control. From Fig. 2(b), one can find that the equivalent hysteresis band is related to the duty ratio D as follows: h0 = Kb R2 + (1 D ) R R1 + R2 + (1 D ) R (8)

Observations 1) Equation (7) indicates that, for a fixed hysteresis band h, the switching frequency varies with respect to the changes of input voltage V d , shaker armature terminal voltage v0 and hence the back EMF ve , filter inductance L0 , and slope of current command iL . 2) However, instead of randomly distributed, the harmonic spectrum of a hysteresis current-controlled PWM inverter with fixed band will gather around a fixed frequency for a particular operating condition. This behavior is explained by results provided later. It follows that the annoying harmonic effects still cannot be avoided. In addition, the output filter design is not easy to perform. 3) Equation (7) reveals that by randomly varying the hysteresis band h adequately, the switching frequency fsw and thus the inverter output harmonic

with the duty ratio being randomly changed according to the used random signal ns (t). Where h0 equals to Ks h, and the unit of h0 is volts while h is amperes. The nominal hysteresis band is determined by the dc control signal vcont,0 . The small resistor R2 in Fig. 2(b) is employed for avoiding too small value of the resulted hysteresis band and hence too high switching frequency. Its choice depends on the blanking time and the switching speed of the power switches. The parameter Kb relates to the saturation voltage of the operational amplifier. Equation (8) reveals that the hysteresis band decreases inversely with the increase of the duty ratio D . The maximum and minimum
621

KANG & LIAW: RANDOM HYSTERESIS PWM INVERTER WITH ROBUST SPECTRUM SHAPING

each particular instant, the instantaneous harmonic spectrum gathers around a fixed frequency. Thus with the band varied by the random signal, the resulting equivalent harmonic frequency distribution can be predicted in Fig. 3(c). From Fig. 3, one can find that too large extreme values of ns (t) leads to the following. 1) Too large hysteresis band h (due to ns ) will result in harmonic components distributing in too low frequency range that will excite the undesired mechanical resonance. 2) Too small hysteresis band h (due to ns ) may lead to too high switching frequency that will significantly increase the switching losses. As a result, the values of ns and ns should be properly limited. This issue is made via intuitive and practical way. Before making the design, the general requirements for a particular application are first given: 1) the highest fundamental frequency is f 1 ; 2) apart from f 1 there are no significant harmonics within a frequency band of f" ; and 3) the lowest switching frequency f sw is beyond the audible frequency range. One can find from (7) that as f sw , V d , L0 , and iL are prescribed, the maximum value of hysteresis band h can be estimated. Then, the duty ratio D of the PWM scheme shown in Fig. 2(b) can be calculated with predetermined R1 , R2 , and R . Subsequently, the maximum allowable value of control signal vcont can also be determined. Similarly, using the analysis procedures given previously, the minimum allowable value of control signal vcont can also be obtained based on the given maximum switching frequency f sw . It follows from the prescribed extreme values of vcont , the dc control signal vcont,0 and the extreme values ns and ns of the random signal ns (t) can be determined correspondingly. In the use of a random signal as the tuning function, in addition to its amplitude, the effects of its bandwidth on the inverter output harmonic distributed characteristics should also be considered. The bandwidth fn or the varying speed of the random signal ns (t) may affect the uniform distribution characteristic of the harmonic spectrum between f sw and f sw designed above. However many simulations show that its effect is not significant. Generally, the following rule of thumb is sufficient fn > f1, f n < f sw (10) where f n and f n denote the lowest and highest frequency components of ns (t), respectively, and fn = f n f n .
APRIL 2001

Fig. 3. (a) Sketched hysteresis band variations due to randomization. (b) Sketched harmonic spectra with respect to (a). (c) Sketched equivalent instantaneous harmonic distributed characteristics of proposed random hysteresis PWM.

hysteresis bands corresponding to D = 0 and D = 1, respectively, can be expressed as


0 hmax = Kb 0 hmin

R2 + R , R1 + R2 + R

R2 = Kb : R1 + R2

(9)

Since the quantitative analysis and design of the random switching scheme are rather difficult, an intuitive approach is proposed here to accomplish these tasks. B. Intuitive Spectral Analysis The random signal ns (t) employed for varying the hysteresis band and the resulted varying bands around the current command are sketched in Fig. 3(a). The nominal and the extreme value of the hysteresis bands h0 , h, and h are corresponding to the control signals vcont = vcont,0 , vcont = vcont,0 + ns and vcont = vcont,0 + ns shown in Fig. 2(b), respectively. And the harmonic spectra with respect to these different hysteresis bands are sketched in Fig. 3(b). As shown in Fig. 3(b), the smaller the hysteresis band, the frequency distribution of its harmonic spectrum is wider and toward the higher frequency region. Fig. 3(b) indicates that for
622

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 37, NO. 2

C. Design of Random Switching Scheme According to the previous spectral analysis, the design procedure of the proposed random hysteresis PWM scheme is described as follows. Step 1 The operating frequency range of the actuator is specified: f 1 f1 f 1 , taking the shaker as an example, f 1 = 5 Hz and f 1 = 2000 Hz. For the maximum amplitude of a sinusoidal output current I s = 8 A, it is reasonable to choose the nominal hysteresis band as h = 1 A. Consequently, the possible switching frequency range, corresponding to the nominal hysteresis band, can be approximately estimated from (7) to be 39 kHz fsw 63 kHz. It is known from the previous analysis that this switching frequency range may vary with the random varying band. Considering the constraints of the operating frequency range of shaker, the audible frequency range and the switching losses, the resulting extreme switching frequencies of the proposed random hystersis PWM inverter are selected as f sw = 30 kHz and f sw = 200 kHz. Step 2 After determining f sw and f sw , the corresponding extreme band values h and h of the random hysteresis comparator can be approximately estimated from (7) to be h = 1:73 A and h = 0:057 A. 0 These correspond to h = 0:43 V and h0 = 0:014 V, respectively, for the adopted current sensor scaling factor Ks = 0:25 V/A. Step 3 In the hysteresis comparator shown in Fig. 2(b), the parameters are preset as Kb = 13:5 V, R1 = 30 k , R2 = 20 , and R = 2400 . From (8), the required extreme values of duty ratio D and 0 D corresponding to h0 and h can be determined. Then, based on the modulating characteristics of the PWM-controlled resistor and the known amplitude of triangular wave (= 1 V), one can find that the dc control voltage is vcont,0 = 0:59 V and the extreme values of the random signal ns (t) are ns = 0:4 V and ns = 0:4 V. The generation of random signal ns (t) with suitable magnitude and bandwidth is indispensable in the realization of random switching scheme. While the commercial white noise generator is the best choice, the ROM-based random signal generator [13] is a practical and cost-effective alternative. In this paper, the white noise generator manufactured by Ando Company followed by a second-order low-pass filter with suitable cut-off frequency is first employed to generate the desired random signal. According to the previous analysis and the guideline suggested in (10), the cut-off frequency of low-pass filter is set at 20 kHz. The low-pass filtered random signal is then properly amplified and limited to yield the random signal with its amplitude varying between ns = 0:4 V and ns = 0:4 V. According to the spectrum characteristics of random signal generated by the approach described above, a ROM-based random

Fig. 4. Proposed robust random hysteresis PWM scheme. (a) Desired harmonic spectra relationship. (b) Robust randomized hysteresis PWM scheme configuration. (c) Simplified block diagram of (b).

signal generation scheme [13] with suitable low-pass filtering is designed and employed for the proposed random hysteresis PWM inverter in real operation. Rather similar results obtained by employing these two types of random signals can be observed from the results presented in Section V. IV. PROPOSED ROBUST SPECTRUM SHAPING APPROACH Through applying the proposed random switching control technique, the output harmonic spectrum of a hysteresis current-controlled PWM inverter can be randomized to distribute dispersedly. However, for some special applications, it is more desirable to have smaller magnitude of harmonics within lower frequency range. To achieve this goal, a robust spectrum shaping approach is devised here to modify the distribution characteristics of harmonic spectrum. The philosophy of the proposed robust control approach can be understood with the help of the block diagrams shown in Figs. 4(a)4(c). As the equivalent system and sketched spectrum shown in Fig. 4(a), the 0 output current iL of the proposed random hysteresis
623

KANG & LIAW: RANDOM HYSTERESIS PWM INVERTER WITH ROBUST SPECTRUM SHAPING

PWM inverter can be regarded to be composed of three components as


0 0 0 = Ks iL = iL + irL + i"L = iL + Ks irL + Ks i"L iL

(11)

where it is assumed that the fundamental component 0 will closely follow its command iL , i"L and of iL irL denote the current harmonics within low and high frequency ranges, respectively. In the modified spectrum shown in Fig. 4(a), which is yielded by the proposed robust shaping control, i"L is significantly attenuated to {"L , while irL is only slightly affected ( {rL irL ). The proposed robust spectrum-shaping controller is 0 0 0 shown in Fig. 4(b). The tracking error i" = (irL + i"L ) is low-pass filtered by HLp (s) to extract the 0 low-frequency component "I = i"L . This error signal "I is then passed through a weighting factor to generate a compensation control signal iLc . Through carefully rearranging, one can find that Fig. 4(b) can be approximately simplified to those shown in Fig. 4(c). Obviously, the low-frequency harmonic 0 current component i"L has been approximately reduced by a factor of (1 w). Although the ideal compensation control can be obtained by letting w = 1, the effect of infinite loop gain [16] on the random hysteresis PWM switching behavior should be considered. It follows that a suitable value of w (0 < w < 1) must be chosen. The trial and error with the help of simulation is adopted here to make this choice. It should be noted that the other specific spectrum shaping characteristics could also be achieved by just choosing proper transfer functions of HLp (s). V. SIMULATED AND EXPERIMENTAL RESULTS

Fig. 5. Simulated current and acceleration at 2000 Hz of fixed-band hysteresis PWM inverter. (a) Waveforms. (b) Spectra.

The system parameters of this experimental shaker system have been listed in Section II. In the following tests, the acceleration control loop is opened. Before implementation, the inverter-fed shaker shown in Fig. 1 using the proposed random hysteresis PWM scheme is simulated to test its validity. For comparison, the simulated waveforms and spectra of iL , iL , and acceleration a of hysteresis PWM inverter with fixed-band (band h0 = 0:25 V) at operating frequency of 2 kHz are shown in Fig. 5. The results reveal that the output harmonic spectrum is concentrated in a specific narrow frequency range with significant magnitude. Then, shaker is powered by the proposed random hysteresis PWM inverter using the random signal produced from commercial white noise generator. First, let the robust spectrum-shaping controller be not activated; Figs. 6 and 7 show the simulated waveforms and spectra of currents and acceleration at operating frequencies of 5 Hz and 2 kHz, respectively. All the results show that
624

Fig. 6. Simulated current and acceleration at 5 Hz of random hysteresis PWM scheme. (a) Waveforms. (b) Spectra.

the harmonic spectra yielded by the proposed random hysteresis PWM scheme are dispersedly distributed. As those mentioned previously, although the output harmonic spectra of the proposed random hysteresis PWM inverter can be randomly distributed,
APRIL 2001

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 37, NO. 2

Fig. 7. Simulated current and acceleration at 2000 Hz of random hysteresis PWM scheme. (a) Waveforms. (b) Spectra.

Fig. 8. Simulated current and acceleration at 5 Hz of random hysteresis PWM scheme and robust control. (a) Waveforms. (b) Spectra.

it is more desired to reduce the magnitude of low-frequency harmonics as far as possible, while the uniform distribution of high-frequency harmonic spectra is still preserved. Now let the proposed robust spectrum-shaping controller be added. The weighting factor and the spectrum-shaping low-pass filter transfer function HLp (s) shown in Fig. 4(b) are chosen to be w = 0:9 and HLp (s) = 3:55 1010 s2 + 3:77 105 s + 3:55 1010 (12)

where the cut-off frequency of HLp (s) is 30 kHz. Figs. 8 and 9 show the simulated results at the same conditions as those in Figs. 6 and 7. Significant reduction of harmonic magnitude in the frequency region between f 1 and f sw can be clearly observed from the comparison between the results in Figs. 69. Specifically, Figs. 6(b) and 8(b) indicate that the magnitude reduction of low-frequency harmonics around f1 (= 5 Hz) by applying the proposed method is up to one order. This leads to the great reduction of acceleration ripple as shown in Figs. 6(a) and 8(a). The measured results of the fixed-band hysteresis PWM inverter at the same conditions of Fig. 5 are plotted in Fig. 10. The results show that they are roughly close to those yielded by simulation. The circuit implementation of the proposed scheme shown in Figs. 2 and 4 is performed. The analog switch used in the tunable-band hysteresis comparator is realized using CMOS switch IC CD4066, and other parts are realized using OP amplifiers. Using the random

Fig. 9. Simulated current and acceleration at 2000 Hz of random hysteresis PWM scheme and robust control. (a) Waveforms. (b) Spectra.

signal generated by the commercial Ando white noise generator and letting the robust spectrum-shaping controller be not activated, Figs. 11 and 12 show
625

KANG & LIAW: RANDOM HYSTERESIS PWM INVERTER WITH ROBUST SPECTRUM SHAPING

Fig. 10. Measured current and acceleration at 2000 Hz of fixed-band hysteresis PWM inverter. (a) Waveforms. (b) Spectra.

Fig. 11. Measured current and acceleration at 5 Hz of proposed random hysteresis PWM inverter without robust spectrum shaping. (a) Waveforms. (b) Spectra.

the measured waveforms and spectra of iL , iL , and acceleration a at operating frequencies of 5 Hz and 2 kHz, respectively. Then, the proposed robust spectrum-shaping controller is added with weighting factor w = 0:9. Figs. 13 and 14 show the measured results at the same conditions as those in Figs. 11 and 12. The results shown in Figs. 1114 indicate that they are rather close to those shown in Figs. 69 by simulation. Similar to those by simulation, the inverter output harmonic spectra are dispersedly distributed by the proposed random switching technique, and by using the robust spectrum shaping control, the magnitudes of low-frequency harmonics around the operating frequency are further suppressed to lower levels. Practically, instead of employing the commercial white noise generator, an economic way for generating the random signal is needed. One of the simplest digital types of random signal is perhaps the ROM-based pseudorandom binary sequence (PRBS). Here, a 1024 data set of 10-bits PRBS is first generated off-line using digital operation of shift register and NOR gate, then it is stored in a ROM. In real operation, a voltage-controlled oscillator (VCO) IC XR-2206 provides the adequate clock signal to extract the ROM-based PRBS random signal. Then it is scaled, limited, and passed through a second-order low-pass filter with cut-off frequency at 20 kHz and Q = 3 to obtain the desired random signal. The spectra (not shown here) of the random signals

Fig. 12. Measured current and acceleration at 2000 Hz of proposed random hysteresis PWM inverter without robust spectrum shaping. (a) Waveforms. (b) Spectra. APRIL 2001

626

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 37, NO. 2

Fig. 15. Measured current spectra of proposed random hysteresis PWM inverter with robust spectrum shaping using filtered ROM-based random signal at: (a) 5 Hz, (b) 2000 Hz.

Fig. 13. Measured current and acceleration at 5 Hz of proposed random hysteresis PWM inverter with robust spectrum shaping. (a) Waveforms. (b) Spectra.

generated by using Ando white noise generator and ROM-based generator are very close. By applying this ROM-based generating random signal to the proposed random hysteresis PWM inverter with robust spectrum-shaping controller, the measured spectra of iL at operating frequencies of 5 Hz and 2 kHz are shown in Figs. 15(a) and 15(b), respectively. All the measured current spectra show that they are very close to those shown in Figs. 13 and 14, which are yielded using the commercial random signal generator. It is worth mentioning that if a more economic approach for generating the desired random signal is needed, one can obtain it by extracting and properly processing the systems inherent existing random signal [13]. VI. CONCLUSIONS

Fig. 14. Measured current and acceleration at 2000 Hz of proposed random hysteresis PWM inverter with robust spectrum shaping. (a) Waveforms. (b) Spectra.

This paper has present a random switching control strategy to let the harmonic spectrum of the hysteresis current-controlled PWM inverter be randomly distributed. The study of the effect of hysteresis band on the harmonic distribution characteristic of a hysteresis current-controlled PWM scheme shows that the band should be randomly modulated such that its harmonic spectrum can be dispersedly distributed. To achieve this goal, a randomized hysteresis band switching control scheme is devised. Detailed analysis, design, and implementation of the proposed switching scheme have been performed, in particular, the quantitative design and the generation of random signal, which significantly affects the harmonic spectrum randomly distributed characteristic. Moreover, since generally, it is desired that the low-frequency harmonics have lower magnitudes, a robust spectrum-shaping technique is further devised. In the resulted inverter output harmonic spectrum, the magnitudes of lower frequency harmonics are
627

KANG & LIAW: RANDOM HYSTERESIS PWM INVERTER WITH ROBUST SPECTRUM SHAPING

significantly reduced by robust control, hence the inverter output with better harmonic attribute is obtained. The simulated and measured results have confirmed the validity of the proposed robust random hysteresis PWM scheme using both the commercial and the ROM-based random signal generators.
REFERENCES [1] Holtz, J. (1992) Pulse modulation: A survey. IEEE Transactions on Industrial Electronics, 39 (Oct. 1992), 410420. Kazmierkowski, M. P., and Malesani, L. (1998) Current control techniques for three-phase voltage-source PWM converters: A survey. IEEE Transactions on Industrial Electronics, 45 (Oct. 1998), 691703. Tilli, A., and Tonielli, A. (1998) Sequential design of hysteresis current controller for three-phase inverter. IEEE Transactions on Industrial Electronics, 45 (Oct. 1998), 771781. Dawande, M., and Dubey, G. K. (1999) Bang-bang current control with predecided switching frequency for switch-mode rectifiers. IEEE Transactions on Industrial Electronics, 46 (Feb. 1999), 6166. Trzynadlowski, A. M., Blaabjerg, F., Pedersen, J. K., Kirlin, R. L., and Legowski, S. (1994) Random pulse width modulation techniques for converter-fed drive systemsA review. IEEE Transactions on Industry Applications, 30 (Sep. 1994), 11661175. Habetler, T. G., and Divan, D. M. (1991) Acoustic noise reduction in sinusoidal PWM drives using a randomly modulated carrier. IEEE Transactions on Power Electronics, 6 (May. 1991), 356363. Jacobina, C. B., Lima, A. M. N., Da Silva, E. R. C., and Trzynadlowski, A. M. (1998) Current control for induction motor drive using random PWM. IEEE Transactions on Industrial Electronics, 45 (Oct. 1998), 704712.

[8]

[9]

[10]

[2]

[11]

[3]

[12]

[4]

[13]

[5]

[14]

[6]

[15]

[7]

[16]

Kirlin, R. L., Kwok, S., Legowski, S., and Tryznadlowski, A. M. (1994) Power spectra of a PWM inverter with randomized pulse position. IEEE Transactions on Power Electronics, 9 (Sept. 1994), 463472. Hui, S. Y. R., Sathiakumar, S., and Sung, K. K. (1997) Novel random PWM schemes with weighted switching decision. IEEE Transactions on Power Electronics, 12 (Dec. 1997), 945951. Covic, G. A., and Boys, J. T. (1998) Noise quieting with random PWM AC drives. IEE Proceedings, Pt. B, 145 (Jan. 1998), 110. Bech, M. M., Pederson, J. K., Blaabjerg, F., and Tryznadlowski, A. M. (1999) A methodology for true comparison of analysis and measured frequency domain spectra in random PWM converters. IEEE Transactions on Power Electronics, 14 (May 1999), 578585. Bech, M. M., Blaabjerg, F., and Pederson, J. K. (2000) Random modulation techniques with fixed switching frequency for three-phase power converters. IEEE Transactions on Power Electronics, 15 (July 2000), 753761. Liaw, C. M., and Lin, Y. M. (2000) Random slope PWM inverter using existing system background noise: Analysis, design and implementation. IEE Proceedings, Pt. B, 147 (Jan. 2000), 4554. Corzine, K. A. (2000) A hysteresis current-regulated control for multi-level drives. IEEE Transactions on Energy Conversion, 15 (June 2000), 169175. Chen, T. H., and Liaw, C. M. (1999) Vibration acceleration control of an inverter-fed electrodynamic shaker. IEEE/ASME Transactions on Mechatronics, 4 (Mar. 1999), 6070. Liaw, C. M., and Chiang, S. J. (1993) Robust control of multi-module current-mode controlled converters. IEEE Transactions on Power Electronics, 8 (July 1993), 455465.

628

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 37, NO. 2

APRIL 2001

Bor-Jehng Kang was born in Taiwan, R.O.C., on August 20, 1961. He received the B.S. and M.S. degrees in power mechanical engineering from National Tsing Hua University, Hsinchu, Taiwan, in 1983 and 1985, respectively. From 1985 to 1997, he was an Assistant Researcher in Chung-Shan Institute of Science and Technology. He is currently working towards the Ph.D. degree at the Department of Power Mechanical Engineering, National Tsing Hua Univeresity. His areas of research interest are servo control, motor drives, and mechatronics.

Chang-Ming Liaw (S88M89) was born in Taichung, Taiwan, R.O.C., on June 19, 1951. He received the B.S. degree in electronic engineering from Tamkang College of Arts and Sciences, Taipei, Taiwan, in 1979, and the M.S. and Ph.D. degrees in electrical engineering from National Tsing Hua University, Hsinchu, Taiwan, in 1981 and 1988, respectively. In 1988, he joined the faculty of National Tsing Hua University as an Associate Professor in electrical engineering. Since 1993, he has been a Professor in the Department of Electrical Engineering. His areas of research interest are power electronics, motor drive, and electric machine control. Dr. Liaw is a life member of the CIEE.
KANG & LIAW: RANDOM HYSTERESIS PWM INVERTER WITH ROBUST SPECTRUM SHAPING 629

You might also like