PWM Strategy to Alleviate Common-Mode Voltage with Minimized Output Harmonic Distortion for Five-Level Cascaded H-Bridge Converters

: High-frequency components of common-mode voltage (CMV) induce the shaft voltage and bearing current, which lead to premature failures in motors. In addition, due to non-zero average CMV, the low-frequency components of CMV, particularly the third-order harmonic component, have been reported to cause difficulties in common-mode filter design. Furthermore, the utilization of distant voltage vectors in the pulse-width modulation (PWM) with reduced CMV magnitudes gives rise to high output harmonic distortion compared to PWM ones without CMV reduction. In an attempt to solve the aforementioned issues, this article presents a PWM strategy that features reduced CMV magnitudes, zero average CMV, and improved output harmonic distortion for a five-level cascaded H-bridge (CHB) converter. In addition, the carrier rotation technique based on the phase-leg redundancy of the CHB topology is also combined with the proposed scheme to achieve equal power loss distribution among power switching devices. Both simulation and experimental results confirm that the proposed strategy produces better output harmonic distortion than that of POD-SPWM and APOD-SPWM under the condition of reduced CMV magnitudes, average CMV, and equal power loss distribution.


Introduction
Multilevel converters have found themselves in many industrial motor-drive applications. Several advantages, when compared to two-level ones, include lower harmonic distortion, higher blocking voltage, a more significant number of redundant switching states, lower electromagnetic interference (EMI), and lower dv/dt [1]. Some prominent multilevel converter topologies are neutral point clamped (NPC), flying capacitor (FC), and cascaded H-bridge (CHB) converters. Among them, CHB converter topologies have been widely used in industrial applications, especially for the number of levels being greater than or equal to 5 (n ≥ 5). Aside from inheriting all the strengths of multilevel converters, they also possess some other merits such as a modular structure [2], a smaller number of switching devices as opposed to NPC converter topologies for the same number of levels [3]. With regard to the DC-link voltage levels among different cells in the same phase leg, the CHB converters can be classified as asymmetric and symmetric ones. The asymmetric CHB topologies aim at reducing the number of isolated DC-link voltages involve modifying PWM strategies. Hence, they are more preferred over the hardware ones.
From the perspective of CMV alleviation, PWM methods can be classified as fullyeliminated-CMV PWM, partially-eliminated-CMV PWM, and non-eliminated-CMV strategies. Under the fully-eliminated-CMV PWM schemes, CMV can be theoretically eliminated to zero by avoiding all switching states that generate CMV magnitudes. Due to having no zero-CMV-magnitude vectors in the two-level space-vector diagram, the complete CMV elimination cannot be achieved in a conventional two-level converter [21] and is only feasible for odd n-level converters. As a result, the research work in [22] proposed a fully-eliminated-CMV strategy in conjunction with switching loss reduction under different power factor conditions for odd n-level converters. The switching loss reduction is realized based on the redundancies of implementation of zero-CMV switching sequences in a carrier period, in which the sequence with double switching commutations is assigned to one of the three phases with maximum absolute current magnitude. However, the total harmonic distortion of the proposed strategy in [22] is significantly higher than that of in-phase disposition sinusoidal PWM (IPD-SPWM) due to the use of distant zero-CMV vectors. The fully-eliminated-CMV scheme proposed in [23] aims to solve the deficiency of high harmonic distortion in [22] by once again utilizing the redundancies of implementation of zero-CMV switching sequences. The analytical harmonic distortion factor then determines which redundancy that leads to minimum harmonic distortion. The research work in [24] further improves the harmonic distortion produced in [23] by modifying the switching sequence in a carrier period. In [25], a marginal improvement of harmonic distortion is observed for a fully-eliminated-CMV PWM strategy with the use of a sawtooth carrier over the one in [24]. Even though the CMV magnitude is theoretically eliminated to zero, the CMV spike still occurs due to the deadtime effect. As a result, the work in [26] was presented to reduce the CMV spike. Nevertheless, the CMV spike is reduced at the expense of high switching loss. Although significant efforts have been made to reduce the harmonic distortion and switching loss in the fully-eliminated-CMV PWM strategies, both harmonic distortion and switching loss are still significantly higher than those of partially-eliminated-CMV PWM and non-eliminated-CMV PWM ones [27], thereby raising the torque pulsation in motors and reducing the overall efficiency of a system. The partially-eliminated-CMV schemes offer a nice balance among the CMV magnitude reduction, the THD of outputs, and the switching loss by neglecting switching states that generate high CMV magnitudes. In the conventional two-level converters, owing to the nonexistence of zero-CMV voltage vectors, significant attention has been paid to reduce the CMV magnitudes by avoiding the two zero voltage vectors. Several PWM strategies include remote state PWM (RSPWM) [28], active zero-state PWM (AZPWM) [29], and near state PWM (NSPWM) [30]. A comparative analysis of harmonic distortion for the above-mentioned PWM strategies is included in [31]. As demonstrated in [31], a common drawback of these strategies is the deteriorated output quality compared to the non-eliminated-CMV strategy such as SPWM. Therefore, the demand for both CMV-magnitude reduction and improved output quality gives rise to the adoption of multilevel converters. Owing to the complex control of NP voltage balancing for n ≥ 5 and uneven power loss distribution in an NPC converter, a CHB converter topology is generally preferred.
The partially-eliminated-CMV PWM strategy typically used in a CHB converter is phase opposition disposition SPWM (POD-SPWM), whereas the non-eliminated-CMV-PWM ones are IPD-SPWM and alternate phase opposition disposition SPWM (APOD-SPWM). Regarding the output quality, as demonstrated in [32], IPD-SPWM produces the best output harmonic distortion performance compared to APOD-SPWM and POD-SPWM due to the use of three nearest vectors. Unfortunately, the CMV magnitude given by IPD-SPWM is twice as high as that of POD-SPWM. Although the CMV magnitude reduction is achieved in POD-SPWM, its total harmonic distortion is noticeably higher than IPD-SPWM, as indicated in [32].
As a result, this article presents a PWM strategy with reduced CMV magnitude and zero average CMV in conjunction with minimized output harmonic distortion for a fivelevel CHB converter. The key contributions of the article include: (1) A PWM strategy featuring reduced CMV magnitudes, zero average CMV, and minimum output harmonic distortion is extended from a three-level NPC [33] to a fivelevel CHB converter; (2) A carrier-rotation technique [7] is combined with the proposed scheme for even power loss distribution in a five-level CHB converter, thereby enhancing the efficiency and reliability over the long term compared to a five-level NPC one; (3) A power loss analysis conducted in a PLECS thermal module is presented to confirm the even distribution of power loss, and to establish the condition of equivalent switching loss power under which the harmonic distortion of different PWM strategies is fairly compared against one another; (4) A detailed comparative analysis of output harmonic distortion on the basis of comparable switching loss power is conducted in both simulation and experiment for the proposed strategy, POD-SPWM, APOD-SPWM, and IPD-SPWM.
Finally, the article is organized as follows: Section 2.1 presents a mathematical model of a five-level CHB converter for a seamless extension of the proposed PWM strategy. Section 2.2 further enhances the performance of the proposed one by employing the carrier-rotation technique unique to a CHB topology for even power loss distribution. Section 3 illustrates some fundamental performance criteria for evaluating any PWM strategy, such as power loss (conduction and switching loss), total harmonic distortion, weighted total harmonic distortion, and RMS value of phase current ripple. Section 4 details a comparative analysis of output harmonic distortion, CMV magnitudes, and power loss in the simulation for the proposed one, POD-SPWM, APOD-SPWM, and IPD-SPWM. Section 5 demonstrates a comparative analysis of harmonic distortion in the experiment for the strategies mentioned above. Section 6 includes key remarks on the performance of the proposed scheme over POD-SPWM, APOD-SPWM, and IPD-SPWM.

With CMV Alleviation and Improved Output Harmonic Distortion
This section presents a PWM method for CMV alleviation, which can be achieved by reducing CMV magnitudes and maintaining zero average CMV in one sampling period. In addition, the output harmonic distortion is also minimized, thereby enhancing the overall efficiency of a system.
The PWM strategy for a 5-level CHB converter is derived based upon the equal DClink voltage condition, which can be obtained by multi-pulse diode rectifiers [1]. Hence, the instantaneous pole voltages of phase leg A ( ), phase leg B ( ), and phase leg C ( ) as shown in Figure 1a can be expressed as: where , ∈ {1,2,3,4} ∈ { , , } is regarded as a switching state of SWix, which assumes a value of 0 or 1.
The virtual instantaneous normalized switching voltage of phase leg A ( ), phase leg B ( ), and phase leg C ( ) can then be written as: As for 5-level NPC converters, there is a restriction over the switching states constituting the normalized switching voltage , i.e., ≤ ≤ ≤ [22]. However, there is no such restriction in 5-level CHB converters. This characteristic is referred to as phaseleg redundancy on which the carrier-rotation techniques [7][8][9][10] are based to distribute the power loss among switching devices equally. The instantaneous normalized switching voltage of phase leg A ( ) in (4), phase leg B ( ) in (5), and phase leg C ( ) in (6) can be decoupled into 2 components.
where is the active switching state of phase leg ∈ { , , } which takes a value of 0 or 1.
is the base voltage of phase leg ∈ { , , }.
The instantaneous CMV can be defined as [22][23][24][25][26]: The average value of , , in one sampling period can be written as: where is the reference voltage signal of phase ∈ { , , }. is the offset voltage. is the average active switching state of phase ∈ { , , }. is the base voltage of phase ∈ { , , }. and returns the smallest integer value of . For ease of understanding, the instantaneous, and average voltage models of threephase legs are illustrated in Figure 2. The average CMV in one sampling period can be then written as: The three-phase reference voltage signals in (5) can be expressed as: The condition of zero average CMV dictates the limit of linear modulation region, i.e., 0 ≤ ≤ √3 2 . This condition also leads to in (17) being equal to 2. Let us define , , and as total switching voltage, total average active switching voltage, and total base voltage, respectively.

= +
= + + = + + The zero average CMV in one sampling period results in = 6 in (21). Figure 1b shows that the red-line hexagon is the region where the average CMV is zero. This region consists of six sub-regions corresponding to two cases of = 4 and = 5, as shown in Figure 3. The reference vector ⃗ in the five-level space vector diagram can be synthesized by ⃗ in the two-level one with the aid of the base voltage vector ⃗ , as illustrated in Figure   1b. Under the condition of zero average CMV and reduced CMV magnitude, three possible switching sequences can be used to synthesize ⃗ , as shown in Figure 3b, for = 4 and ⃗ in the area (1). Among these, there is only one switching sequence that yields the best output harmonic distortion, i.e., 011→111→110→ 100 →110→ 111 → 011 [33]. The dwell time calculation of which is also shown in Figure 4b, along with the flowchart featuring switching sequences for the areas (1)-(6) which also gives the minimum output harmonic distortion. In this article, they are referred to as RCMV1-PWM.

With Equal Loss Distribution
The presented RCMV1-PWM strategy ensures the zero average CMV, reduced CMV magnitude, and minimum output harmonic distortion. In addition to those, equal power loss distribution among switching devices is also realized by cyclically changing the switching states after a fundamental period of as per Table 1. In the latter section, RCMV1-PWM is evaluated thoroughly against other methods to confirm the equal distribution of power loss indicated in this section.

Performance Criteria for Evaluation
This section presents two important criteria, including the power loss and the harmonic distortion, to evaluate PWM strategies. The power loss estimation is used to verify the power loss balance of the proposed strategy among switching devices, as mentioned in the previous section. In addition, it establishes the condition of comparable power loss on which the harmonic distortion calculation is based to fairly compare across different PWM strategies.

Power Loss
The power loss, which typically consists of the conduction and switching loss, can be readily estimated by the thermal module in the PLECS simulation program [34]. The principle of power loss calculation is explained in detail in [35]. The power switching devices (IGBTs) with an antiparallel diode used in the five-level CHB inverter are IKW15N120H3, of which the thermal characteristics can be obtained from the manufacturer's website [36] and imported into the thermal module. For comparison, the power loss distribution of a five-level NPC inverter is also included. The power switching devices IGBTs used in the five-level NPC topology are the same as those in the five-level CHB one and the clamping diode type is IDP30E120, of which the datasheet and thermal characteristics can be found in [37]. For easy reference, some main parameters of IGBT IKW15N120H3 and clamping diode IPD30E120 are tabulated in Tables 2 and 3.

Harmonic Distortion
The harmonic distortion performance of each PWM strategy is evaluated by the THD of the voltage and current outputs, the weighted THD (WTHD) of the voltage output, and the root mean square (RMS) of current ripple in Ampere. The formulas of these performance metrics are given as follows [33]: where: and are the fundamental magnitude and the n th harmonic magnitude of line voltage output, respectively. and are the fundamental magnitude and the n th harmonic magnitude of phase current output, respectively. and are the measured and the fundamental component of phase current, respectively.

Simulation Results
The simulation is conducted in MATLAB/Simulink environment under the condition of = 200 V , = 50 Hz , = 5 KHz , = = = 33.3 Ω , = = = 2.7 mH , = 4700 μF for a five-level CHB converter. The comparative analysis of RCMV1-PWM will be conducted against POD-SPWM, APOD-SPWM, and IPD-SPWM in terms of the performance criteria presented in the previous section.

Harmonic Distortion Evaluation
On the basis of the comparable power loss set forth in Section 4.2, the output harmonic distortion performance of RCMV1-PWM is compared against that of other methods. The THD and WTHD of voltage and the THD of current, as well as the RMS value of current ripple, are evaluated with regard to the modulation index for four strategies. Figures 5 and 6 shows that the IPD-SPWM gives the best output quality for a whole modulation range due to the three nearest vectors principle. However, the magnitude of CMV is twice as high as that of other strategies, as illustrated in Figure 7a. Hence, it is unfavorable in motor-drive applications due to adverse effects of high CMV magnitude, as mentioned earlier. As demonstrated in Figures 5 and 6, the proposed RCMV1-PWM produces better output harmonic performance than the POD-SPWM over a whole modulation range. For example, at m = 0.3, the THD values of voltage and current outputs are 61.78% and 22.19%, respectively, for RCMV1-PWM as opposed to 73.55% and 26.65%, respectively, for POD-SPWM. This translates to a more than 10% improvement of THD of voltage output and a more than 4% improvement of THD of current output. Likewise, at m = 0.3, the WTHD of voltage output, and the RMS value of phase current ripple, are 0.61% and 0.13 A, respectively, for RCMV1-PWM in comparison to 0.73% and 0.16 A, respectively, for POD-SPWM. Both RCMV1-PWM and POD-SPWM produce a reduced magnitude of CMV of /3 for a whole modulation range, as shown in Figure   7a. As a result, under the reduced-CMV-magnitude condition, RCMV1-PWM is more favorable than POD-SPWM in motor-drive applications due to the lower output harmonic distortion. As for APOD-SPWM, its THD and WTHD of both voltage and current outputs, as well as RMS value of current ripple, are inferior to those of RCMV1-PWM for ≤ 0.45 and ≥ 0.7. For these two modulation ranges, both APOD-SPWM and RCMV1-PWM obtain a reduced CMV magnitude of /3 as demonstrated in Figure 7a. For 0.45 ≤ ≤ 0.7, the output harmonic performance of APOD-SPWM is better than that of RCMV1-PWM. However, the CMV magnitude of APOD-SPWM in this modulation range is the same as that of IPD-SPWM, i.e., 2 /3 as shown in Figure 7a. Hence, the better output harmonic performance of APOD-SPWM is achieved at the expense of higher CMV magnitude. Therefore, under the reduced-CMV-magnitude condition, the proposed RCMV1-PWM gives the most optimum output harmonic performance.

Power Loss Comparison
The conduction and switching loss of eight IGBTs of the phase leg A for RCMV1-PWM at m = 0.3 and m = 0.8 under the power factor of 0.99 in the five-level CHB inverter are shown in Figure 8. As shown in Figure 8, the equal power loss distribution among switching devices is achieved for the proposed strategy, as indicated in Section 2. By contrast, the NPC converter, which has the issue of unequal power loss distribution [11] illustrated in Figure 9 for a five-level NPC inverter, possibly leads to failures in some switching devices that have to experience a considerable amount of power loss over a long-term operation, thereby reducing the efficiency and reliability of a system, as well as the overall switch utilization [11]. The active neutral point clamped (ANPC) converter was proposed in [11] to resolve this issue. However, the increase in the number of switching devices results in higher power loss, and the control strategy is more complicated when compared to the CHB converter at the same number of levels. Therefore, the CHB converter topology with the proposed strategy featuring lower CMV magnitude, zero average CMV, and improved output harmonic distortion, as well as equal power loss distribution among switching devices, becomes a topology of choice due to higher efficiency and reliability over a long-term operation as opposed to the NPC one for the number of levels being greater than or equal to 5 (n ≥ 5). In order to further evaluate the power loss performance of RCMV1-PWM, the comparison of its power loss is made against that of POD-SPWM, APOD-SPWM, and IPD-SPWM at m = 0.2, 0.5, and 0.8, as illustrated in Figure 10. As can be seen from Figure 10, both the total switching loss and the total conduction loss of RCMV1-PWM are equivalent to those of other strategies at three modulation indices.

Experimental Results
The experiment is conducted on a five-level CHB converter with a no-load three-  Figure 11a-c, respectively, for RCMV1-PWM, POD-SPWM, and APOD-SPWM at f0 = 36 Hz (m = 0.623) under the no-load condition. In addition, to demonstrate the output harmonic distortion performance, the experimental THD of line voltage, and the phase current, as well as the RMS values of phase current ripple of RCMV1-PWM, are compared against those of other strategies. As can be observed from Figure 12a, the THD of line voltage of RCMV1-PWM is better than that of POD-SPWM for a whole fundamental frequency range. This is consistent with the simulated one, as shown in Figure 8a. The most noticeable reduction in THD of voltage happens at f0 = 14 Hz, at which it is 88.9% for RCMV1-PWM, as opposed to 106.1% for POD-SPWM. These results in an 18.87% improvement of THD of voltage for RCMV1-PWM over POD-SPWM. The frequency spectra of line voltage at f0 = 36 Hz are shown in Figure 13a,b to further illustrate the THD reduction in line voltage of RCMV1-PWM over POD-SPWM. As can be observed from Figure 13a,b, the harmonic components around the switching frequency (f = 5 KHz) of RCMV1-PWM are significantly less than those of POD-SPWM. In particular, they are less than 14 V in RCMV1-PWM as opposed to more than 25 V in POD-SPWM. With regard to APOD-SPWM, the THD of voltage of RCMV1-PWM is lower than that of APOD-SPWM for f0 < 26 Hz (m < 0.45) and f0 > 40 Hz (m > 0.69). The most significant improvement of the THD of line voltage happens at f0 = 10 Hz, at which it is 127.59% for RCMV1-PWM compared to 147.32% for APOD-SPWM. This translates to a 19.72% improvement of the THD of the voltage of RCMV1-PWM over APOD-SPWM. For 26 Hz ≤ f ≤ 40 Hz, the THD value of the voltage of APOD-SPWM is better than that of RCMV1-PWM. However, this frequency range 26 Hz ≤ f ≤ 40 Hz (0.45 ≤ ≤ 0.69) coincides with the higher CMV magnitude (2 /3) of APOD-SPWM than that ( /3) of RCMV1-PWM and POD-SPWM, as demonstrated in Figure 10a. As can be observed from Figure 12a, IPD-SPWM still produces the best output harmonic distortion. However, its magnitude of CMV is twice as high as that of RCMV1-PWM and POD-SPWM for a whole frequency range and is the same as that of APOD-SPWM for f ≥ 26 Hz ( ≥ 0.45).
200V/div 10ms/div 30V/div 10ms/div 700mA/div 10ms/div (a) 200V/div 10ms/div 30V/div 10ms/div 700mA/div 10ms/div (b) 200V/div 10ms/div 30V/div 10ms/div 700mA/div 10ms/div (c) With regard to the current, the THD of current and the RMS values of current ripple of RCMV1-PWM are consistently better than those of POD-SPWM for a whole fundamental frequency range, as illustrated in Figure 12b,c, respectively. This is further supported by the frequency spectra of current of both RCMV1-PWM and POD-SPWM as shown in Figure 14a,b at f0 = 36 Hz. As can be observed from Figure 14a,b, the harmonic components around the switching frequency (f = 5 KHz) are significantly less than 6 mA in RCMV1-PWM, as opposed to more than 10 mA in POD-SPWM. Similar to the THD of voltage, the THD of current and the RMS values of current ripple of APOD-SPWM are lower than those of RCMV1-PWM for 26 Hz ≤ f ≤ 36 Hz (0.45 ≤ ≤ 0.62), which corresponds to a higher CMV magnitude of APOD-SPWM. As for f0 < 26 Hz (m < 0.45) and f0 > 36 Hz (m > 0.62), the THD of current, as well as the RMS value of current ripple of APOD-SPWM are worse than those of RCMV1-PWM, as evidenced in Figure 12b,c. The harmonic distortion calculations, including THD and WTHD in both simulation and experiment, are calculated up to the 200th harmonic of the fundamental frequency.
As for the CMV reduction performance, both RCMV1-PWM and POD-SPWM produce a reduced magnitude of CMV of /3 for a whole fundamental frequency range.
In particular, the CMV waveforms and their frequency spectra are shown in Figure 11a,b and 15a,b, respectively, at f0 = 36 Hz (m = 0.623). The CMV reduction performance is demonstrated by the reduction in the magnitude of CMV components around the switching frequency (f = 5 KHz). Specifically, they are less than 5 V for both RCMV1-PWM and POD-SPWM compared to more than 10 V for APOD-SPWM at f0 = 36 Hz (m = 0.623), as shown in Figure 15. Moreover, the 3rd harmonic component of CMV, as shown in Figure  15, is well less than 1 V, which accounts for less than 2% of Vdc (Vdc = 50 V), thereby ensuring the zero average CMV in one sampling period for RCMV1-PWM. Therefore, under the condition of the reduced magnitude of CMV, the zero average CMV, and the equal distribution of power loss, the proposed strategy produces the most optimum output harmonic performance when compared to POD-SPWM, and APOD-SPWM.

Conclusions
This article presents a reduced-CMV PWM strategy with zero average CMV and minimum output harmonic distortion for a five-level CHB converter. The CMV magnitude reduction aims to alleviate the shaft voltage and bearing current problems that lead to premature failures in motors while the zero average CMV is maintained to avoid the complications in CM filter design. The switching loss analysis conducted in the PLECS thermal module confirms that the proposed strategy, namely RCMV1-PWM, gives the comparable switching loss to that of POD-SPWM, APOD-SPWM, and IPD-SPWM. On the basis of equal switching loss power, the CMV reduction and the output harmonic distortion are compared for different PWM strategies. In terms of the CMV magnitude reduction, both RCMV1-PWM and POD-SPWM produce the same reduced CMV magnitude while the CMV magnitude of IPD-SPWM is twice as high as that of RCMV1-PWM and POD-SPWM over a whole range of modulation. Concerning output harmonic distortion, RCMV1-PWM gives better output harmonic distortion than that of POD-SPWM over a whole range of modulation under the condition of the reduced CMV magnitude. Moreover, in the modulation range that the CMV magnitude of APOD-SPWM is the same as that of RCMV1-PWM and POD-SPWM, the output harmonic distortion of APOD-SPWM is inferior to that of RCMV1-PWM. Therefore, under the condition of the equivalent switching loss power, the reduced CMV magnitude, and the zero average CMV, the proposed scheme, namely RCMV1-PWM, gives the most optimum harmonic distortion as opposed to that of POD-SPWM and APOD-SPWM, thereby reducing the torque vibration attributed to high harmonic distortion, as well as adverse effects caused by the high CMV magnitude in motor-drive applications. Finally, thanks to equal power loss distribution, the proposed strategy, when applied to a five-level CHB converter, yields higher efficiency and reliability than a five-level NPC counterpart.