Audible noise and losses in variable speed induction motor drives with IGBT inverters-Influence of design and the switching frequency

| Introduction | Audible Noise | Losses | Motor and Inverter Used | Measuring Set-up | Measurements and Conclusions | Acknowledgement | References |

ABSTRACT
This paper describes a large number of experiments comparing audible noise, losses and efficiency of different induction motors supplied by IGBT inverter. The influence of the switching frequency on the motor behavior is analyzed. As comparison the motor is also supplied directly from the mains. Furthermore two different motors are compared to study the influence of the rotor cage design on the motor behavior. It is shown that the choice of the switching frequency has a large influence on the audible noise produced by induction motors. By taking higher switching frequencies, the time harmonics move to a not audible region. At some higher switching frequencies subharmonics appear, having a bad influence on the audible noise. At higher fundamental frequencies the audible noise of the fan dominates.

• The switching frequencies can cause pure tones that are very disturbing to the human ear.
• The influence of the switching frequency on the efficiency is rather small. Minimal overall losses appear around 3 and 4 kHz.
• The voltage frequency ratio should be taken constant from the point of view of the efficiency.
  

1. INTRODUCTION

The use of variable speed drives has gained increasing importance during the last years. For small and medium sized drives, PWM voltage source inverters are the standard type of equipment nowadays. Until a few years ago, the highest switching frequencies available were offered by Power-MOSFET inverters. However, their rated power was modest. Bipolar transistors could be used for higher power rating, but had a lower stitching frequency, The IGBT based inverters combine the advantages of both power electronic components: High switching frequency and high power ratings. Power ratings up to several hundreds of kVA are available commercially. The switching frequency may be several kHz.

Manufacturers claim that the increased switching frequencies lead to lower additional losses, due to the harmonics of the supplied voltage. Therefore, derating the motor at rated speed would not be required. Furthermore. The low harmonic content would avoid audible noise problems.

The paper describes a large number of experiments comparing audible noise, losses and efficiency of induction motors supplied by an IGBT inverter or directly from the mains. The switching frequency of the IGBT PWM inverter is varied in order to study the influence of this parameter on the motor behavior. Furthermore, two motors are compared. The first motor is a standard induction motor with cast aluminum double squirrel cage rotor. The second motor has a copper cage, with round bars to avoid current redistribution; furthermore the bars are insulated with respect to the rotor iron in order to avoid extra losses, which are difficult to incorporate in the theoretical analysis.

II. AUDIBLE NOISE

Audible noise has been of concern for users and manufacturers of inverter supplied induction motors since their introduction. The first developments using thyristors and high power GTO inverters may be particularly noisy. Using an appropriate switching pattern can resolve part of the problem, but still audible noise increases of 10 to 15 dB(A) or more are noticed, both for small and medium sized drives [1,2,3]. The analysis using rotating field theory shows that the PWM switching pattern clearly influences the audible noise level. At higher fundamental frequencies however, the degree of freedom for choosing the PWM pattern is limited. In order to get the maximum output voltage, the voltage pattern is very close to a rectangular shape, independent of the switching frequency.

The theoretical analysis of the audible noise starting from the design parameters of the motor, especially the number of stator and rotor slots, and the frequency spectrum of the voltage supplied by the inverter, yields its frequency spectrum. It is shown that the transition of the PWM pattern is the most important factor in the audible noise. Furthermore, the analysis shows under which circumstances pure tones in the spectrum of the audible noise may be expected. This is very important as the human ear is very susceptible to these pure tones: the same audible noise level is experienced being 10 dB(A) higher when it contains pure tones.

III. LOSSES

Due to the harmonics in the inverter voltage and current, the losses in the rotor increase as a result of the skin effect [5]. The currents induced by the higher voltage harmonics are concentrated in the upper part of the rotor bars. Therefore, the efficiency decreases due to these losses. In conventional inverter supplied drives using low switching frequencies , these extra losses are high and the manufacturers of drives prescribe generally a 10 % derating of the motor in order to limit excessive heating. However, practical experiments showed that a 10 % derating sometimes is not sufficient [5,6,7].

The higher switching frequency of modem inverters lead to a lower harmonic content, especially at low speeds. Therefore, the increase in the losses is less pronounced at low speed. However, in pump and fan drives, the output power is limited at low speed. Therefore, the efficiency gain is large, without yielding large power or energy savings. If a constant torque as a function of speed has to be delivered to the load, the power is proportional to the speed. The gain is larger at lower speed than in pump and fan drives.

IV. MOTOR AND INVERTER USED

The first motor used is a standard 22 kW machine with a double cage rotor. The rotor has double bar construction. The rotor bars are made in cast aluminum. These bars are not insulated with respect to the rotor iron, leading to supplementary losses due to interbar currents. This type of rotor construction is required in order to increase the starting torque when connecting the motor directly to the supply. The rotor slots are skewed.

As these starting problems are not present in inverter supplied machines, a motor with no current redistribution is no real problem and would lead to less additional losses if the motor is supplied with a non sinusoidal voltage. In order to assess the influence of the current redistribution in the rotor, a special rotor for a standard 13.5 kW machine is built with round rotor bars. The area of the rotor bars is chosen in such a way that the torque and the losses at rated speed are the same as in the standard design. Copper is used as the material for the bars. Interbar currents are avoided by insulating the bars with respect to the rotor iron. The number of rotor bars is kept the same and the stator is a standard construction. No skewing is applied. Both standard rotor design and new rotor are accounted for in the analysis. The inverter is a standard inverter with IGBTS. The switching frequency of the components can be chosen from I to 12 kHz. The tests are done in the normal operation mode, i.e. keeping the Volts/Hertz ratio constant. As a comparison the motors are directly coupled to the mains. The use of a pure sinusoidal supply yields an absolute target for inverter manufacturers, as this is the ultimate optimum regarding frequency content. Using these combinations the influence of the most important parameters on both losses and audible noise are studied.

V. MEASURING SET-UP

The measuring set-up consists of an inverter supplied induction motor, with combinations as discussed above. The machine is loaded with an eddy current brake that can be controlled accurately in order to load the motor as prescribed. If a pump, fan or compressor characteristic is to be implemented, the mechanical output torque is proportional to the square of the speed, otherwise the torque is kept constant with respect to the speed (elevator or crane drive). The overall lay-out and most important components of the measuring set-up are shown on Fig. 1.

The audible noise was measured in a semi-anechoic chamber, avoiding interference of the frequency inverter audible noise. The load has a very low audible noise level due to special bearings and water cooling. The audible noise is measured using a condenser microphone. Both the overall noise level in dBlin or dB(A), and its frequency spectrum are registered.

A torque transducer is used to couple both machines. The transducer contains strain gauges, that are linked with the stationary recording equipment using a frequency modulated transmitter system. These measurements yield a very high accuracy of the torque, far better than the reaction torque often used as an alternative. Also non-steady state torque components and torsional vibrations may be accounted for. In the torque transducer a speed transducer is incorporated, yielding the mechanical power by multiplication. The accuracy of the output power is better the 0.5 %.

The input -power is measured with an accuracy of 0.2 % using special power, voltage and current measuring equipment

For the assessment of the electrical power between the inverter and the motor, special attention has to be paid to the equipment used. The harmonics in both current and voltage prevent the use of standard power measuring .equipment The frequency spectrum of the power measuring equipment goes up to 200 kHz, in order to account for the power produced by the higher harmonics. The accuracy of the power between inverter and motor is 0.5 %.

All efficiency measurements are carried out after thermal equilibrium is reached in the motor. Therefore the motor is loaded during more than one hour with rated power at rated speed before the experiments are started. In the specially designed machine the temperature is monitored and measured using Pt-100 (Platina resistors) sensors.

At rated speed and load, the temperature when supplying the motor with the inverter is compared with the temperature when the motor is coupled directly with the supply, in both cases loaded with rated power. The temperature gives a direct indication of the required derating, if any.

VI MEASUREMENTS AND CONCLUSIONS

A Overall audible noise level
The overall audible noise level was measured for the 22 kW motor loaded with 100 Nm. In Fig. 2 the overall audible noise level is presented as a function of the switching frequency and this for different fundamental frequencies (15-25-40-45-50-70 Hz).

Conclusions:

• At a switching frequency of 2 or 4 kHz, the overall audible noise level is higher than expected in the case of a fundamental frequency of 40 Hz. This is due to the interference between the natural frequency of the stator and the frequency of one of the vibrations produced by the electromagnetic forces caused by the switching frequency. These frequencies should be avoided. At higher frequencies these phenomena do not occur because of the difference between the set-up switching frequency and the real switching frequency used in the inverter.
• Generally the overall audible noise level decreases if the switching frequency increases. Particularly up to 6 kHz there is a huge drop in the audible noise. Once the switching frequency is above 6kHz, the overall audible noise level is stabilized because the number of switching harmonics in the audible noise region strongly decreases.
• At higher fundamental frequencies thus at higher motor speeds the overall audible noise level decreases less if the switching frequency increases because the fan audible noise dominates the motor audible noise.
• At lower motor speeds the overall audible noise is strongly reduced by applying higher switching frequencies.
• Generally the audible noise increases with the increasing fundamental frequency. But at 50Hz because the motor produces less audible noise than when supplying the motor with 45Hz because the motor design is optimized for a 50 Hz supply by an appropriate choice of the number of stator and rotor bars.
  When comparing 50 Hz sinus main supply and 50Hz inverter supply there is only a difference of 1 dB(A) due to the harmonics.

B. Audible Noise Spectrum
The audible noise spectrum is analyzed to see if there are any pure tones in the spectrum as the human ear is very susceptible for such tones.

Several audible noise spectra have been analyzed for both motors. Only two spectra will be discussed here: the spectrum for the 22 kW motor at a fundamental frequency of 5Hz, a switching frequency of 6kHz and a load of I 00 Nm (Fig. 3) and the spectrum for the13.5 kW motor at a fundamental frequency of 45.05 Hz, a switching frequency of 5 kHz and a load of 75 Nm (Fig. 4).

Figure 3. Audible noise spectrum of the 22kW motor, fundamental frequency = 50Hz, switching frequency = 6kHz, load = 100Nm.

The switching frequency is set-up to be 6 kHz but the real switching frequency used in the inverter is 8400 Hz, as can be clearly seen in Fig. 4 where there is an increase in audible noise of 25 dB(A) around 8400 Hz
There is also a component of 70 dB(A) at 300 Hz. This pure tone is caused by the fifth harmonic of the current at 250 Hz. The frequency difference between the resulting audible noise at 300 Hz and the cause at 250 Hz can be explained as follows. The fifth harmonic of the current has a velocity of 250 Hz in the opposite direction of the velocity of the rotating field at 50 Hz. The relative velocity between the fifth harmonic and fundamental of the rotating field is 300 Hz. This gives torque pulsations at 300 Hz and thus audible noise is produced at 300 Hz. It should be mentioned that this pure tone only appears at a voltage of 380 V.

Around 5 kHz also some large components appear in the audible noise. They are caused by space harmonics of the rotating field. These space harmonics cause vibrations and audible noise. If the frequency of the vibrations is the same as one of the natural frequencies of the stator, there is a large audible noise production.

Figure 4. Voltage and audible noise spectrum of the 13.5kW motor, fundamental frequency = 45.05 Hz, switching frequency = 5 kHz, load = 75 Nm.

The switching frequency is set-up to be 5 kHz but the real switching frequency used by the inverter is .5406 Hz as can be sea in the voltage spectrum.

Because the motor bars are not skewed, the rotor harmonics are not suppressed so they yield a lot of vibrations. The frequencies of these vibrations may be calculated [8]:

In this case the motor has 28 rotor bars, a speed of 1318 t/min and 4 pole pairs. At a fundamental frequency of 45.05 Hz the motor has a slip of 2.48%. Applying (1) and (2) for g2 = *9 one can calculate that there should be vibrations at 5626 Hz, 5445 Hz and 5563 Hz.

At 5536 Hz there is also a resonance frequency of the stator. Therefore, at a fundamental frequency of 45.05 Hz and a speed of 1318 t/min there is a spike of 78 dB at 5536 Hz at all switching frequencies. At the switching frequency of 5 kHz as shown in Fig. this spike is even 86 dB because of the interaction of a vibration caused by a winding harmonic and a vibration caused by a time harmonic of the voltage spectrum.

The harmonics of the voltage system cause also vibrating forces, yielding audible noise. In the voltage spectrum of Fig. 4 there is a time harmonic at 5488 Hz. As the fundamental frequency is 45 Hz, the vibration frequency is 5533 Hz- Taking the accuracy of the measurements into consideration, this frequency equals to the resonance frequency of the stator and with a vibration frequency caused by a winding harmonic.

The pure whistle tone of 86 dB in the audible noise spectrum of Fig. 4 is caused by the accidental coincidence of 3 effects at 5536 Hz:

stator resonance;
vibrations due to winding harmonics;
vibrations due to switching harmonics.

Clearly such a situation should be avoided by all means.
Generally for all motors one can say that if there are voltage and current harmonics then audible noise harmonics will be generated. If the spectrum is optimized for instance by taking higher switching frequencies or by taking another pulswidth modulation pattern, the audible noise level can be reduced. The fundamental wave can also cause in some cases resonance and thus high audible noise levels, even if a sinusoidal supply is available. The switching frequency should be chosen in such a way that time harmonics do not yield vibration forces that match with mechanical natural frequencies of the stator.

C. Losses and efficiency

The higher the chosen switching frequency, the better the spectrum supplied by the inverter. The harmonic currents induced in the motor become smaller. The motor voltage gets more and more sinusoidal. Therefore, the losses in the motor decrease. As the losses in the inverter are defined by the number of times the power electronic components have to operate, the losses in the inverter on the other hand increase with a higher switching frequency. Therefore an optimum switching frequency will exist.

The type of motor used is also important: motors experiencing a large amount of current redistribution in the rotor during when supplied from the mains cause extra losses with inverter supply as the high frequency currents are concentrated in the upper portion of the rotor bars. To analyze the influence of the rotor lay out two motors of the same rated power arc compared (one with an aluminum skewed double cage rotor and one with a rotor having non skewed copper bars.

1) Definition of the losses and efficiencies
The following losses and efficiency definitions are used:




2) Influence of the switching frequency on the efficiency

The temperature is kept constant during the measurements. Again several measurements were done, but only one measurement reported in detail. The efficiencies and the losses as a function of the switching frequencies for the 13.5 kW motor supplied with a fundamental frequency of 35 Hz and loaded with 70 Nm, are shown in respectively in fig. 5 and table I

Fig. 5. Efficiencies as a function of the switching frequencies, for the 13.5 kW motor supplied with a fundamental frequency of 35 Hz and loaded with 70 Nm.

TABLE I
LOSSES AT 35 Hz AND 70 Nm

The following conclusions may be drawn:

• The motor losses are independent of the switching frequency.
• The inverter losses increase with increasing switching frequency.
• The harmonic losses decrease with increasing switching frequency. From 10 kHz on subharmonics appear and give rise to a slight increase of this loss component.
• The total losses reach a minimum around a frequency of 3 kHz.

In other operating points (different torque and different fundamental frequency) the overall losses are also minimum around 3 to 4 kHz.

A comparison is also made between the two different motors of 13.5 kW and between mains supply and inverter supply at 5OHz- The overall losses are represented in table II. In fig. 6 the different losses are represented as a function of the switching frequency.

TABLE II
Comparison Between the Losses of the Standard Motor and the New Motor (13.5kW)

Fig. 6 Losses as a function of the switching frequencies, for the 13.5 kW motor supplied with a fundamental frequency of 50 Hz and loaded with 75 Nm
a) Standard rotor
b) New motor

The following conclusions can be made:

• Both motors have almost the same harmonic losses. At lower switching frequencies ( < 3 kHz) the standard motor has slightly higher losses.
• The inverter losses are in both cases comparable.
• The motor losses of the standard motor are smaller d= the motor losses of the new motor. This is due to the larger space harmonics in the air gap caused by missing skewing
• With a main supply the overall losses are in both cases smaller then when an inverter is used

The same measurements were repeated for the 22 kW standard motor and again the optimal switching frequency concerning the motor losses is found to be 4 kHz.

3) Influence of the U/f ratio on the efficiency

There are several possibilities to set-up the voltage/ frequency ratio. In this paper two of them are compared. U/f is a constant and U/f is dependent of the load (automatic)(Fig. 7).


Fig. 7 U/f at T = 60 Nm and f_switch = 6 Hz

The lower voltage in the second case decreases the flux level and therefore less audible noise is produced. The harmonic losses are also lower because of the improved spectrum of the voltage but the motor losses are larger because the currents have to increase to match the required power. The inverter losses also increase. Other disadvantages are a decrease in speed and less dynamic response for torque variations.

D. Motor temperature when using an inverter

The manufacturers of the IGBT-inverters supply allowable torque versus frequency curves (fig. 8)

Fig. 8 Load curves for continuous load of IEC 34 motors Switching frequency 3 kHz.

These curves are valid if the fundamental frequency at the point of field weakening is 50 Hz. If the fundamental frequency is higher than 50 Hz, the torque decreases as the voltage is kept constant as the frequency rises. Therefor the flux and the torque decrease. At lower fundamental frequencies the torque decreases due to a lack of cooling.

Therefore with the inverter used between a fundamental frequency between 40 Hz and 50 Hz, the motor can deliver its rated torque. The temperature rise at a fundamental frequency should not be higher than with the main supply.

The temperature in the stator winding is measured with embedded Pt - 100 temperature sensors. Table III gives the temperate rise when the motor is operating at a fundamental frequency of 50 Hz, supplied by the inverter at a switching frequency of 3 Hz and 12 kHz and directly by the mains, at a load torque of 70 Nm

TABLE III
TEMPERATURE RISE WITH INVERTER SUPPLY AND MAIN SUPPLY AT 50 Hz AND 70 Nm

The following conclusions can be drawn:

• The temperature of the motor is independent of the switching frequency at a fundamental frequency of 50 Hz.
• The temperature of the motor directly connected to the mains does not differ much from the motor temperature when fed by a inverter.
• The optimum switching frequency of 3 kHz gives a slightly lower temperature rise.

The same measurements have also been done for other loads, yielding comparable results. It may be concluded that for the motor and inverter combination used is these tests derating is not necessary. It is dangerous to generalize this conclusion for all motor-IGBT inverter combinations because only the temperature in the stator is measured. Furthermore these measurements were carried out on a single squirrel-cage rotor motor with a specialty built rotor.

VII GENERAL CONCLUSIONS

The choice of the switching frequency has a large influence on the audible noise produced by induction motors. By taking higher switching frequencies, the time harmonics move to a not audible region. At some higher switching frequencies subharmonics appear, having a bad influence on the audible noise and therefore, should be avoided. At higher fundamental frequencies the audible noise of the fan dominates and the use of higher switching frequencies has practically no effect on the audible noise.
The switching frequencies can cause pure tones that are very disturbing to the human ear. Fundamental waves can also cause vibrations and audible noise. One should be careful that the vibration frequencies from fundamental waves and switching frequencies do not match the natural frequencies of the stator as otherwise the audible noise increases a lot.
The influence of the switching frequency on the efficiency is rather small. Minimal overall losses appear around 3 and 4 kHz.
The voltage-frequency ratio should be taken constant from the point of view of the efficiency.

VII ACKNOWLEDGMENT

The authors wish to thank the council of Belgian National Science Foundation for granting a project sponsoring the research.
Thanks is also given to ABB-Belgium (Zaventem and Antwerp) for supplying material making the measurements possible.

IX REFERENCES

[1) R.Belmans, L.DHondt, A.Vandenput, W.Geysen: "Analysis of the audible noise of three phase squirrel cage induction motors supplied by inverters," IEEE Trans. on Industry Applications, September-/October 1987, Vol.IA-23, pp.842-847.

[2) R.Belmans, D.Verdyck, W.Geysen, R.Findlay: "Electromechanical analysis of the audible noise of an inverter-fed squirrel-cage induction motor," IEEE Transactions on Industry Applications, Vol-27, No.3, May/June 1991, pp.539-544.

[3] R.Belmans, W-Geysen, G.Bailly, P-K.Sattler: TheoreticaI and experimental analysis of the audible noise of an inverter fed squirrel cage induction motor," International Conference on Electrical Machines (ICEM), Cambridge, Massachusetts, USA., August 13-15, 1990, pp.485-490.

[4] R.Belmans, D.Vcrdyck, T.-B.Johansson, W.Geysen: "Comparison of starting conditions of induction motors fed from an infinite bus and current source inverter using finite element calculations," Proceedings of the Fourth European Conference on Power Electronics and Applications, Firenze, Italy, September 3-6, 1991, pp.2.375-2.378.

[5] R.Belmans, D.Vermeulen, A.Vandenput, W.Geysen: "Economy of the introduction of adjustable speed drives for pumps, fans and compressors," Conference record on the 1986 IEEE-IAS Annual Meeting, September 28-October 3, 1986, Denver, U.S.A., pp.321-327.

[6] R. Belmans, D.Vermeulen, A.Vandenput, W.Geysen: "Techno-economical analysis of inverter fed pump drives," Second European Conference on Power Electronics and Applications, EPE, 22-24 September 1987, pp. 1003-1008

[7] E.Lajoie-Maznec, D.Pratmarty: "Etat de l'offre europeene de convertisseurs de frequence pour machines asynchrones", Electricite de France, Direction des Etudes et Recherches, Mars 1993

[8] P.L.Alger, "Induction Machines, Their Behavior and Uses," Gordon and Branch Science Publishers, New York,
London, Paris, 1970

Copyright 1999 Rockwell International Corporation For more information call 216-266-7000.

Document D-7168-1