LOW VOLTAGE PWM INVERTER-FED MOTOR INSULATION ISSUES

LOW VOLTAGE PWM INVERTER-FED MOTOR INSULATION ISSUES Copyright Material IEEE Paper No. PCIC-4-15 RAPS-1433 Abstract - The topic of how low voltage IGBT-based PWM inverters create additional insulation stress through voltage reflections (due to high dv/dt) has been documented for several years [1-5]. This paper first explains and summarizes the basic causes and effects of the high dv/dt environment. Second, the methodologies of various approaches to solutions are critically reviewed. Finally, standards (such as NEMA MG1, parts 3, 31) [6] have attempted to partially address this issue. However, there are still some areas which require important clarification. This paper makes the case for what can be done to achieve the needed clarity. Testing of motor insulation systems to verify compliance with the proposed requirements is also covered. I. INTRODUCTION As low voltage inverters have evolved over the past 5 years, current waveforms (Fig. 1) have become markedly more sinusoidal. As a result of the better current waveform, the additional stator IR losses experienced by the motor have been minimized. The major source of the improvement in the current waveform has been the use of pulse width modulated (PWM) techniques with higher switching frequencies. One of the enabling technologies for the use of higher carrier frequencies has been the development of transistors with very low switching loss. A technique by which the low switching loss is achieved is through minimization of the transition time (turn-on and turn-off). This lower switching loss characteristic is employed to achieve both higher switching frequencies and physically more compact inverters. 4 3 amps 1-1.5.1.15. - -3-4 (a) motor insulation system and attempted to address appropriate restrictions and expectations for both general purpose and definite purpose (inverter duty) motors [6]. (b) Fig 1: Motor Current Waveform from (a) Six-Step Variable Voltage Inverter, (b) Modern PWM Inverter While faster transistor turn-on helps minimize the switching loss in the device, it also, by definition, creates a voltage pulse with increased dv/dt (Fig. ). Standards such as NEMA MG1 recognized that high dv/dt creates additional stresses on the Fig : II. NEMA MG1-3 Graphic Depicting Voltage Wavefront Appearing at an Inverter-Fed Motor MOTOR ISSUES RELATED TO HIGH dv/dt A. Capacitive Current Coupling Normally we do not think of AC induction motors as having a characteristic capacitance associated with them. At frequencies that are relatively low (certainly at 5-6 Hz, for example), there are in fact negligible capacitive effects. However, the fast voltage transitions of today s PWM inverters represent high enough frequencies (high dv/dt) that capacitive effects are very real. Any two conductors that are separated by insulation can be thought of as making up a capacitor. Windings within an AC motor, or even the leads connecting the inverter output to the motor can provide such capacitance. This capacitance appears both phase to phase (line to line) as well as phase to ground. A basic equation relating to the current carried through a capacitive element is as follows: I = C x dv/dt (Eq. 1)

Even though these stray (unintended) capacitors in motors or leads may be of fairly small magnitude, a high enough dv/dt (Eq. 1) can result in a significant current flow. Figure 3 shows two dv/dt events (turn-on and turn-off) and the associated ground currents which appear during the periods of voltage transitions. It is interesting to note in Figure 3 that for this case, the turn off is a noticeably faster transient than the turn on. This results in more capacitive current flow at the turn off transition than at the turn on. 5 4 3 Line-to-neutral Voltage 1 8 6 4 1. -1.3.4 - - -3 Capacitively-coupled Ground Current Fig 3: Two Voltage Transients and the Associated Capacitively-Coupled Currents, µsec/div horizontal, V/div, Amps/div vertical -4-6 Fig 4: Line-to-line Voltage at Motor Terminals, 5 V/div, 5 µsec/div At Inverter B. Peak Voltage Overshoot When a voltage changes from one value to another in 5 1 nanoseconds, the leads feeding that voltage from point A to point B can have a significant impact on what is seen downstream. [1] This can be thought of as a transmission line effect. A motor will frequently have a surge impedance (high frequency impedance) which is much higher than the characteristic high frequency impedance of the feeder leads. This mismatch of impedances will have the voltage overshoot effect as shown in Figures 4 and 5. In the trace of Fig 4, each PWM transition (turn-off as well as turn-on) as received at the motor terminals has an overshoot beyond the quasi steady-state value. Rather than reaching the 8 VDC bus level, for example, the waveform first overshoots to 146 V peak, before ringing and eventually settling at 8 V. In the traces of Fig 5, the time base has been set fast enough (1 nsec/div) to allow the dv/dt to be quantified. At the inverter output, the dv/dt is seen to be 98 V/µsec. With the relatively nominal lead length of 1 feet for this case, it is interesting that the peak overshoot at the motor is 167% of the steady-state value. While the phenomenon of peak voltage overshoot due to transmission line effects is often considered a long lead issue, given fast enough transitions, even 1 feet can be considered long leads. At Motor Fig 5: Transient Voltage with 5 nsec turn-on and 1 foot leads, Ch 1 = Line to Neutral Volts at Motor, Ch = Line to Neutral Volts at Inverter, V/div, 1 nsec/div In fact, the length of cable required to provide a full x ring up of the PWM voltage can be seen to be very strongly influenced by the risetime (Figure 6). Only about 3 feet of cable is required to get a full x peak voltage ring up when the risetime is as fast as 5 nanoseconds. If the risetime is slowed down to nanoseconds, the lead length required for a full x peak voltage is extended beyond 15 feet. With substantially longer leads, the transmission line effects can move from the simple reflected wave ring-up seen above to the move complex (and more damaging) phenomena covered in Section II D below.

Vpk at Motor / Vdcbus. Turn-on times: 5 ns, 5 ns, 1 ns, ns, 4 ns, 8 ns 1.8 1.6 1.4 1. 1 1 1 1 1 Inverter to Motor Cable Length (ft) The effect of this higher localized stress (even though it is brief in duration) is that insulation degradation may occur if the insulation system is not designed to withstand this stress. One of the potential modes of degradation is through partial discharge (also referred to as corona). Figure 7 shows the impact of partial discharge degradation on a motor winding exposed to a localized stress which exceeded the corona inception voltage. The white color on the magnet wire film is due to damage from many individual partial discharges. These discharges could occur each time the pwm voltage waveform has a transient exceeding the insulation corona inception voltage (CIV). The motor of Fig 7 was wound with a magnet wire that was intended to resist degradation due to partial discharge, and was exposed to test conditions that imposed voltage transients that would create a high dielectric gradient. By building this test sample without resin impregnation, the region of high voltage gradient would be able to produce partial discharges. In addition, the winding could then (without resin) be easily disassembled for failure analysis without obscuring the damage itself. Fig 6: Voltage Overshoot as a Function of Lead Length for Various Transition (turn-on or turn-off) Times C. Voltage Gradients Within Windings In addition to the fact that high dv/dt results in capacitivelycoupled current as well as the opportunity for voltages to ring up / overshoot, it also provides some challenges due to the sheer speed of the transition. Most of us would normally think of voltages as being an instantaneous event in terms of propagation times. This is due to the fact that a voltage will travel through a conductor at a significant fraction of the speed of light. With dv/dt levels in the range of 1, V/µsec, however, the risetime and the propagation time through a motor winding are rather comparable figures. It is easy to have a length of wire within a single motor coil (multiple turns within one slot) which is on the order of 1 meters (about 3 feet) in length. If the propagation speed is even 1/3 the speed of light, that still represents a.1 µsec propagation time from one end of the coil to the other. At dv/dt = 1, V/µsec, that same.1 µsec time would correspond to a change in voltage of 1, volts. The result is that the lead end of a motor coil could have risen up by 1, volts before the other end of the same coil experiences even the start of this transition. That results in there being a transient voltage in the range of 1, volts which briefly would appear between two specific points of a single coil. While the overall dielectric strength of the motor insulation system (phase-to-phase and phase-to-ground) may have sufficient capability compared to the 1, volt level, the fact that this voltage transient is appearing across parts of a single coil represents a much higher localized stress (turn to turn or coil to coil) than would normally occur without the high dv/dt transitions. Fig 7: Partial Discharge Damage to a Motor Winding in an Area of High Transient Voltage Gradient D. Peak Voltages Beyond x V bus In addition to the effects reviewed in Section II B, there are also opportunities for the peak line to line voltage at the motor to significantly exceed twice the inverter dc bus voltage magnitude. This situation is normally associated with either very long leads, very high switching frequencies, modulation schemes without minimum deadtime between transitions, or a combination of these. One result of the circumstances is the occurrence of double pulsing. This can allow motor line to line voltage to achieve the levels seen in Figure 8. This essentially amounts to another transistor transition while the ringing from the prior one transition still has not dampened out. Another case which can occur is a polarity reversal as seen in Figure 9. Both of these situations are quite damaging to motor insulation systems.

Pulse at Inverter Second transition while the ringing from the prior transition has not decayed Higher transient peak after second transition (17 V) Second, the ability of the motor to successfully operate in the presence of these waveforms is a function of the entire motor design and insulation system, including: a) winding connection, b) coil pattern, c) end turn separation, d) magnet wire insulation, e) resin material, and f) impregnation process. B. Component Approach Pulse at Motor Fig 8: Double Pulsing, Line to Line Voltage at Inverter and Motor, 5 V/div, µsec/div It would be desirable to have a single corrective action that would mitigate the reflected wave issue before it ever reaches the motor terminals. Sloping off the dv/dt with various filter circuits, e.g. output line reactors, might be proposed, but this adds cost, hardware, voltage drop, as well as efficiency loss. The degree to which the dv/dt must be mitigated would also have to be determined in order to design the filter. This would involve establishing the level of peak voltage and dv/dt that can be tolerated at the motor terminals, as well as taking into account the effects of the cabling (length and type). 5 V/Div 1 V/Div 8 Vpk Inverter Volta ge Motor Volta ge In terms of component solutions at the motor, since turn to turn and coil to coil are potential failure modes, it is interesting to look at insulation components such as the resin and the magnet wire. In considering resins, data such as that of Figure 1 might be considered. This shows a typical 3 kvac (rms) voltage endurance test at 155 o C, with the results shown as hours of life. Since these tests are typically run with twisted pair samples of wire, the varnish penetration is essentially perfect, and the test is to a large degree a measure of the varnish build. In actual stator windings, however, the process by which the stator is impregnated with the resin along with the resin properties determines to a high degree the effective penetration to eliminate any voids. 4 Fig 9: Polarity Reversal, Line to Line Voltage at Inverter and Motor, 1 µsec/div, Vertical scales as indicated III. APPROACHES TO DEAL WITH HIGH dv/dt A. Problem Identification TIME (HR) 35 3 5 15 It is often the case that the approach to a technical problem seeks root cause failure analysis and silver bullet solutions. For motors exposed to high dv/dt waveforms, there is a need for a systems approach to both failure analysis (and causes) as well as the development of solutions. The primary reasons for this are twofold. First, the waveform that appears at the motor windings is a function of a) inverter hardware, b) inverter modulation scheme and operating point, c) cabling from the inverter to the motor (length, gage, and configuration), and d) the motor design. 1 5 none A B C D E RESIN TYPE Fig 1: Voltage Endurance of a Specific Magnet Wire with Various Resins, 3 kvac (rms) at 155 o C Similarly, twisted pair samples of magnet wire can be subjected to actual inverter output waveforms, including the peak voltage overshoot effects due to lead length. Figure 11 shows samples that were tested with and without a resin coating. The sample in Fig 11(b) can be seen to have deteriorating film insulation (lighter color) due to partial

discharge activity. It is interesting that the sample in Fig 11(b) is a magnet wire marketed specifically for its resistance to degradation from partial discharge. (a) (b) Fig 11: Test samples of varnished (a), and unvarnished (b) magnet wire (corona-resistant) twisted pairs exposed to PWM inverter voltage waveforms If the magnet wire itself is looked to for a solution, one approach would be to increase the voltage endurance by reducing the rate of degradation in the presence of partial discharge activity. Such magnet wire is available, often marketed with terms such as corona resistant or spike resistant. Depending on the degree of partial discharge activity, these wires may increase life by anywhere from a small percentage to a factor of four to six. It is only for the cases of extreme discharge activity that the larger improvements are seen (as a percentage). The actual life for either wire type would be quite short in the presence of high levels of corona. If the magnet wire is looked at to prevent partial discharge, rather than just to last longer in the corona environment, then usually this means heavier build of insulation on the wire. Such a heavier build often implies that less cross section of copper can be wound into the stator slots, resulting in reduced efficiency. C. Holistic / System Approach Because of the multiple influencing factors in both the creation of, as well as the withstand of, the voltage transients, it is not optimal to simply blame one component as the cause, nor to look to one component as providing the cure. Across a range of power ratings and construction types, it is more appropriate to look at the system interactions of the components and to design optimal solutions. These solutions may utilize mitigation at the source, reduced deleterious effects of the cabling, or various design aspects within the motor winding and insulation systems, or a combination. Solutions may involve motor winding patterns, or additional end turn separation controls. Or, they may involve combinations of magnet wire, resin, and impregnation systems. They can also involve filtering at the inverter output or impedance matching at the motor terminals to avoid peak voltage overshoot. Of course, a combination of two or more of these items will commonly be appropriate for a given system. In order to provide a system that gives full motor life, with no degradation due to the waveform dv/dt, the goal should naturally be to completely eliminate any partial discharge activity. Doing this in an optimal way for a range of motor sizes should be expected to involve various approaches, but always with the goal of corona avoidance (not just resistance) as the remediation methodology. IV. TESTS OF SAMPLE MOTORS Corona inception voltage (CIV) measurements can provide a measure of the insulation system effectiveness as well as its repeatability. CIV tests of ten sample motors from one manufacturer are shown in Figure 11. These were motors rated for operation on 46 VAC inverters. It can be seen from the histogram that this group of motors had quite low CIV compared to the potential twice bus voltage overshoots (~13 V peak) that would be expected from 46 V inverters. There are also many ways to rate magnet wire in terms of voltage or dielectric capability. A declaration that a specific wire has a voltage rating of xxxx Volts is not meaningful, without describing what test was used to establish that voltage rating. Table 1 below illustrates both that different magnet wires can have voltage ratings different than each other, and that one specific magnet wire can have voltage ratings that are grossly different depending on the test that was used. The rating which is clearly most critical for inverter-fed motors is the corona inception voltage. Frequency 3 1 TABLE 1 Magnet Wire Voltage Ratings (rms Volts) Wire Type Test Type X Y Single Surge (.µsec) 18, V 5, V 36, Surges 5, V 13, V Corona Inception 5 V 1, V 9 95 1 15 11 115 1 15 More PEAK CIV Fig 1: Peak CIV Histogram from Ten Sample 46 V Motors from One Manufacturer

The data shown in Figure 13, in contrast to that of Figure 1, shows some higher CIV levels, but also shows a larger spread of data. Three of the motors are well below the levels needed for corona-free operation on 46 V inverters, while the other seven are substantially higher. Neither the group of motors in Figure 1 nor those in Figure 13 would be particularly well suited for utilization on modern PWM inverters at 46 VAC. Frequency 3 1 8 9 1 11 1 13 14 PEAK CIV 15 16 17 18 More Fig 13: Peak CIV Histogram from Ten Sample 46 V Motors from a Second Manufacturer V. RECOMMENDATIONS Users should be cautious of one size fits all solutions or those that take a component approach, rather than considering the total system interactions. Avoid band-aid techniques that may only prolong the inevitable and not provide a true solution. Use corona inception as a criteria to determine if an insulation system can withstand a specific level of peak transient voltage and provide full insulation life expectancy. Clarify NEMA MG1 3, paragraph 31.4.4. to utilize an absence of partial discharge activity as the methodology to determine if a motor has indeed been designed to operate at a specific peak voltage. ability of the system to preclude the possibility of partial discharge is a function of many factors within the system, including interactions between these factors. Verification of suitability of inverter-fed motor systems needs to be based on a systems rather than a component approach. VII. REFERENCES [1] L. Saunders, G. Skibinski, S. Evon, D. Kempkes, Riding the Reflected Wave IGBT Drive Technology Demands New Motor and Cable Considerations, IEEE PCIC Conference, Sept., 1996. [] M. Melfi, G. Skibinski, S. Bell, J. Sung, Effect of Surge Voltage Risetime on the Insulation of Low Voltage Machines Fed by PWM Converters, IEEE Industry Application Society Conference, Oct., 1997. [3] E. Persson, Transient Effects in Application of PWM Inverters to Induction Motors, IEEE Trans. Ind. Applications, vol. 8, pp. 195 111, Sept./Oct. 199. [4] A. H. Bonnett, Analysis of the Impact of Pulse Width Modulated Inverter Voltage Waveforms on AC Induction Motors, IEEE Pulp and Paper Industry Conf., 1994, pp. 68 75. [5] S. Bell, J. Sung, Will Your Motor Insulation Survive a New Adjustable-Frequency Drive, IEEE PCIC Conference, Sept., 1996. [6] NEMA Standards Publication MG 1, National Electrical Manufacturers Association, Washington, DC, 3. VI. CONCLUSIONS Low voltage PWM inverter-fed motors are exposed to elevated dv/dt levels thousands of times per second during operation. The ability of the system to operate with full, reliable motor life is dependent on preventing the occurrence of partial discharge activity within the motor insulation system. Such activity would degrade the insulation system life, and would combine with the other insulation stresses (thermal, chemical, mechanical, etc) to result in unsatisfactory life. The