A Complementary Effect in Active Control of Powertrain and Road Noise in the Vehicle Interior

This study shows that a concurrent active noise control strategy for engine harmonics and road noise has a complementary effect. In particular, we found that engine booming noise is additionally attenuated when road noise control is concurrently used with engine harmonics control; an additional attenuation of 2.08 dB and 1.25 dB for the C1.5 and C2.0 orders, respectively, was achieved. A parallel multichannel feedforward controller for non-stationary narrowband engine harmonics and broadband road noise was designed and implemented to reduce noise in all four seats. Two control signals were considered independent because the reference signals, engine revolution speed for the engine harmonic controller, and acceleration signal for the road noise controller are uncorrelated. However, if the reference sensor for the road noise controller is installed along the overlapping transfer path between the engine noise and road noise, the engine noise may also be suppressed by the control signal for the road noise attenuation. Based on transfer path analyses for both engine harmonics and road noise, the optimal positions for the reference sensors were selected. In addition, we identified several overlapping transfer paths between the engine booming noise and road noise. A practical active noise control system combined with a remote microphone technique was implemented for a large six-cylinder sedan using a vehicle audio system to evaluate the noise attenuation performance. The experiments showed that the interior noise from the engine and road excitation was effectively suppressed by the proposed concurrent control strategy.


I. INTRODUCTION
Active noise control (ANC) in vehicles has been widely researched to resolve low-frequency noise problems. In powertrain noise, the engine firing order and its harmonic order have been considered the main targets to be controlled [1]- [4]. After the first implementation of the ANC system, which used two speakers, four microphones, and a tachometer as a reference sensor for mass production by Japanese motor companies [5], considerable research has been conducted on commercialization, mainly by research groups and companies [6]- [8]. In practice, ANC is mainly applied to suppress idle booming, or acceleration engine noise, including lock-up booming [9], [10]. Generally, narrowband feedforward control based on multi-input multi-output (MIMO) filtered reference least mean square (FxLMS) has been used in powertrain noise control because its noise characteristics are mainly tonal [11], [12]. Moreover, the engine revolution speed obtained via controller area network (CAN) bus has been used as a reference signal [13] instead of physical sensors such as microphones or accelerometers to reduce the cost of the ANC system. However, it can lead to performance degradation due to the low data rates and delays [3]. In powertrain noise control, the fast convergence speed of the controller is critical because the engine speed changes rapidly, such as under full-load conditions [14]- [16]. Therefore, many adaptive filters with variable step size have been proposed [17]- [19]. In practical applications, fast convergence and stability are critical because the FxLMS algorithm can lead to instability problems, such as nonlinear distortion and divergence. This occurs when excessive noise levels overload the secondary source [20]. To address this problem, the leaky FxLMS algorithm was proposed [21] and is generally used in practical applications.
This algorithm avoids unconstrained weight overflows and also limits the output power to prevent nonlinear distortion [20]. However, the leaky FxLMS approach can lead to performance degradation in the adaptive filter.
Meanwhile, road noise is also the main target in the interior noise control of a vehicle. Road noise has broadband random noise characteristics that are affected by road surfaces. The algorithm for the active road noise cancellation (RNC) system has generally been used in MIMO FxLMS like powertrain ANC [22]. RNC is more difficult than engine harmonic cancellation because of its complex and unpredictable characteristics [23]. A reference signal that is coherent with interior noise is relatively simple to obtain in the engine noise control system, whereas in a RNC system, the selection of reference signals is less direct [24]. In the practical implementation of RNC, accelerometers are installed as reference sensors in vehicle suspensions or bodies [24], [25]. It has been found that at least six reference signals are required to achieve good control performance because the reference signal should consider the entire road noise source, which significantly contributes to the interior noise [26]. Coherence analysis considering vibro-acoustic paths has been applied to select optimal reference sensor positions [27].
Although both ANC technologies have been successfully developed and implemented in practice for their respective targets, concurrent control of powertrain noise and road noise is required to enhance acoustic comfort further. In concurrent control, when selecting the position of the reference sensor in the road noise controller, if an overlapping position is selected among the transfer paths of the two noise sources, the engine harmonics are expected to be controlled through the RNC signal. However, studies focusing on this topic or attempts in practice are scarce. Moreover, no studies are proving this topic practically.
The primary aim of this study is to design and implement concurrent active noise control to reduce both nonstationary narrowband engine harmonics and broadband road noise. Moreover, it also aims to investigate the influence of other control signals during concurrent control. The remainder of this paper is organized as follows. Section 2 presents the engine sound characteristics during acceleration and road noise at a constant speed driving on a road with a rough surface. In Section 3, transfer path analysis (TPA) for engine booming noise and road noise is conducted to derive the significant vibro-acoustic paths and select the proper reference sensor position for the road noise controller. Section 4 presents the control algorithm used in this study. The control system was developed based on a feedforward-filtered reference leastsquares algorithm. For narrowband engine harmonic noise control, the engine revolution speed was used as the reference signal, and for broadband RNC, accelerometers were used as the reference sensor. In addition, the implementation of a control system for the vehicle is described. The ANC system was implemented in a passenger car based on the vehicle audio system. In Section 5, the experimental results are presented and discussed. Finally, several insights obtained in this study are discussed, and the conclusions are presented in Section 6.

II. CHARACTERISTICS OF ENGINE ACCELERATION NOISE AND ROAD NOISE
In this study, ANC experiments were performed on front-andrear-wheel drive (FR) vehicles with a six-cylinder engine large sedan. Acceleration engine sound and road noise were measured and analyzed to set up the ANC target. First, the engine sound during acceleration was measured under a wideopen throttle condition at the third gear. The acceleration sound was recorded at the position of the ear on the window side of each seat using a 1/2 inch microphone. Figure 1 shows the spectrogram in terms of the C-weighted sound pressure level according to the engine revolution speed at each seat. In Fig. 1, (a), (b), (c), and (d) represent the acceleration noise characteristics at the driver, front right-side passenger, rear left passenger, and rear right passenger seat, respectively. It can be seen that the C1.5 order, which is the subharmonic order, and the C2.0 order generates prominent noise near the 120 Hz region. In contrast, broadband noise above 4,200 RPM, as seen from the rear seat, is airflow noise from the muffler. Therefore, in this study, the target of the engine harmonic cancellation (EHC) was set to C1.5 and C2.0 orders below 200 Hz. Moreover, the road noise was measured on a rough surface road at a constant speed of 60 km/h. Noise characteristics between the left and right ear positions are quite different in the interior road noise. Thus, the interior noise for both the left and right ear positions of the entire seat was measured. Figure  2 shows the spectrogram in terms of the A-weighted sound pressure with respect to time. In Fig. 2, (a) and (b) show the noise for the left and right ears of the driver position. In addition, (c) and (d) are indicates for the rear right seat position. In this test vehicle, the booming noise at 80 Hz and 120 Hz, tire cavity resonance noise at 200 Hz, and the rumbling noise in the 300 Hz region are primary structure-borne road noises. Therefore, in this study, the frequency range for RNC was considered to be below 400 Hz.

III. TRANSFER PATH ANALYSIS TO CONSIDER THE POSITION OF REFERENE SENSORS IN THE ROAD NOISE CONTROLLER
The engine revolution speed has high coherence with the interior noise in the ANC for powertrain noise. Thus, in EHC, the engine revolution speed obtained through CAN bus was used as a reference signal. Meanwhile, acceleration signals are generally used as reference signals for RNC. Thus, the location of the accelerometers is very significant to ensure the high performance of noise suppression. Moreover, because this study aims to reduce the engine harmonic noise with the road noise controller, finding an overlapped transfer path between two independent noise sources is necessary. As a method for selecting the position of the reference sensors, coherence analysis has been generally applied [27]. Accelerometers were installed at various locations on the chassis and body of the vehicle to obtain a highly coherent reference signal. The position of the reference sensor was determined from the results of the coherence function calculation. However, it is challenging to select a sensor location corresponding to an overlapped transfer path of the two noise sources, that is, engine booming noise and road noise, through this approach. Therefore, in this study, TPA was conducted for engine and road noise to determine the optimal reference sensor position. In TPA, interior noise can be expressed as follows: where ( ) is the interior sound pressure, / is the vibroacoustic transfer function, and # ( ) is the operational force along the )* path. In addition, / implies the acousticacoustic frequency response function, and ( ( ) is the operational pressure along the )* path. In this study, only structure-borne noise was required to determine the reference sensor locations. Thus, the air-borne noise term was not considered. In conventional TPA, the matrix inversion method is generally used to obtain the operational force, as shown in equation (2) [28].
where 3/ $ is the measured frequency response function and ̈3 +!,denotes the operational acceleration from the measurement. Thus, structure-borne noise can be calculated using the operational force and vibro-acoustic transfer function (VATF), as shown in equation (1).

A. TRANSFER PATH ANALYSIS FOR ENGINE HARMONIC NOISE
First, TPA was conducted for structure-borne acceleration booming noise at the left ear position of the driver and rear right passenger was conducted. Figure 3 shows the geometry of the TPA and the typical measurement positions for acquiring the VATF to the primary vehicle body input point. The TPA geometry, as shown in Fig. 3, was constructed considering the primary structural transfer path such as engine (ENG) mount, transmission (TM) mount, subframe mount, spring, shock absorber, center bearing, rear crossmember (X/MBR) mount, and exhaust muffler hanger. The VATF was measured with the condition of the source removed. The interior acceleration noise was calculated using the operational force and VATF. Recalling the analysis of accelerated noise in Section 2, peak noise was found at approximately 120 Hz. Therefore, the top five positions that were highly ranked in path contribution and VATF were analyzed at 120 Hz. The TPA results are presented in Table 1 of the appendix. In Table  1, the path contribution means the partial noise for each path, and VATF is the transfer function of the corresponding path.
Analysis of the path contribution through TPA revealed that the path contribution of the rear connection point of the rear crossmember was significant for the booming noise at the position of the driver and the rear right passenger. Moreover, according to the VATF analysis, it can be seen that the center bearing was significant at both the driver seat and the rear seat positions.

B. TRANSFER PATH ANALYSIS FOR ROAD NOISE
Next, TPA for the structure-borne noise of the interior road noise was performed. Road noise of the left ear position of the driver and the center position of the rear seat were considered for TPA. Figure 4 illustrates the structural transfer path of the road noise and the typical measurement points for the measurement of operational data. Structure-borne road noise is transferred to the vehicle cabin through the chassis and body of the vehicle, as shown in Fig. 4. Therefore, TPA was conducted by considering the body interface. The interior road noise was calculated using operational data, which was measured at a constant speed of 60 km/h on a road with a rough surface and VATF in a manner similar to that described in Section 3.A.
Tables 2 to 5 in the appendix present the results of path contribution and VATF according to the rankings, in high order for booming, tire cavity resonance, and rumbling noise, respectively. In Section 2, the peak of road booming noise was found at approximately 80 to 120 Hz. Thus, the path contribution and VATF were analyzed for the noise in the 70 to 80 Hz and 110 to120 Hz region, as shown in Table 2 and  Table 3, in the drive and rear right seat, respectively.
In the booming noise at 70 to 80 Hz region, the path contributions to the rear shock absorber and rear crossmember were relatively high for the driver position. Moreover, from a VATF perspective, the rear crossmember was found to be important. In addition, in the position of the rear right passenger, the rear crossmember and rear spring were significant from both the path contribution and VATF perspective. In the booming noise at 110 to 120 Hz, it can be seen that the path contribution to the front strut and left-back and right-front of the subframe was relatively high for the driver position. Moreover, the rear crossmember and subframe were significant in terms of VATF. In addition, in the position of the rear right passenger, the front strut, subframe, and rear crossmember were significant from the path contribution perspective, and the rear crossmember and subframe were found to be important in terms of VATF.
Next, the results for tire cavity resonance noise of approximately 200 Hz are shown in Table 4 for the driver and rear right seat positions. The subframe was the main transfer path in the driver's position. And the subframe and rear crossmember were significant from the VATF perspective. Furthermore, in the rear seat position, the transfer paths of the rear crossmember and rear spring were primary, and the magnitude of VATF of the rear shock absorber and rear crossmember was high. Finally, the results for rumbling noise around 350 Hz are shown in Table 5 for the drive and rear right seats. In the driver's seat position, the path contribution of the rear crossmember, subframe, and rear spring was relatively high. Moreover, the subframe and rear crossmember were significant from the VATF perspective. Furthermore, in the rear seat position, the rear spring and rear crossmember were important transfer paths. Moreover, the rear shock absorber and rear crossmember were significant from the VATF perspective.
Through the analysis of the transfer path for the engine booming noise and road noise, it can be seen that the rear crossmember is an overlapped transfer path. Therefore, the engine booming noise can be controlled by the road noise controller if the reference signal is obtained at the rear crossmember position.

IV. CONCURRENT CONTROL ALGORITHM FOR NARROWBAND ENGINE HARMONICS AND BROADBAND ROAD NOISE
In this study, a feedforward adaptive controller, which was combined with a remote microphone technique, was designed and implemented [11]. The recursive least square (RLS) and least mean square (LMS) algorithms have been widely used for adaptive noise control [20]. The LMS algorithm is more advantageous for vehicle noise control because the reference signals are nonstationary, depending on the driving conditions [29]. Moreover, the FxLMS algorithm is a suitable approach for controlling vehicle interior noise considering the secondary path [20], [25]. Thus, in this study, the FxLMS algorithm was designed for both nonstationary narrowband and broadband control. Figure 5 illustrates the block diagram of the algorithm applied in this study.
The narrowband engine harmonics controller and broadband road noise controller were combined in parallel. In this configuration, the output signal of the other controller can act as an additional control signal under conditions that are combined with acceleration booming noise and road noise. Therefore, it is necessary to consider the influence of additional signals from the viewpoint of each controller. If the position of the reference signal for the road noise controller coincides with the transfer path of the engine vibration, then the reference signal is coherent with the engine harmonic.
Thus, the control signal of the road noise controller will have a positive influence on the engine harmonic controller as an additional control signal. In contrast, from the viewpoint of road noise controllers, it is difficult to expect positive effects via the EHC signal. There was no correlation between the error signal, that is, road noise, and the reference signal, that is, engine harmonics. In Fig. 5, the upper inside dashed block represents the engine harmonic controller, and the lower one denotes the road noise controller. ( ) and 4 ( ) are the disturbance signals in which a reference signal, ( ), passes through real and virtual primary paths, ( ) and 4 ( ).
In addition, ( ) and 4 ( ) are output signals that ( ) filters the corresponding real and virtual secondary paths, ( ) and Similarly, a virtual error signal can be converted as 4 ( ) = 4 ( ) + 4 ( ). The FxLMS algorithm updates the filter coefficient vector ( ), to minimize the quadratic cost function given by the mean square error. The instantaneous gradient of with respect to ( ) can be expressed as [11]: where I( ) is a reference signal vector filtered by the secondary path model, ( ). Thus, the adaptation algorithm becomes: where is a convergence coefficient.
In FxLMS with the remote microphone technique, the virtual error signal, 4 ( ), cannot be directly obtained in realtime control because it is only used for modeling of acoustic paths. The estimated signal 4( ) thus used [11], and it is given by: where O 4 ( ) is an estimated virtual disturbance signal in which the virtual disturbance signal, 4 ( ), passes through the primary path difference model, Q ∆ . Thus, equation (6) can be written as [11]: where I( ) is an estimate of the output signal vector and I 4 ( ) is the estimated virtual output signal. In addition, Q ∆ implies a vector of the impulse response difference between the physical and virtual primary paths. By inserting I 4 ( ) and 4( ) into equation (5), the renewed adaptation algorithm can be obtained as: where is the convergence coefficient and I 4 ( ) is a filtered reference signal vector that passes through the virtual secondary path model O 4 ( ) [31].
Next, the implementation of a concurrent active noise control system based on a vehicle audio system is presented. The real-time adaptive controller was implemented on the digital signal processors (DSP) in the external audio amplifier. The sampling frequency for ANC is 1,500 Hz. The controller for ANC was used the DRA 751 (Texas Instruments) with two 750 MHz C66x DSP. Control signals were delivered to the external audio amplifier to control both the engine harmonics and road noise, as shown in Fig. 6. The total latency should be within approximately 3.0 milliseconds [32], considering the transmission time of the road noise from the vibration source to the ear in the vehicle cabin and secondary path. In our experiments, the total latency was 1.48 milliseconds. Therefore, it can be confirmed that the total latency is within the controllable range. In Fig. 6, A indicates an accelerometer, EM is a physical error microphone, and VM is a virtual microphone. Microelectromechanical system (MEMS)-type physical microphones (SPG08P4HM4H-1, Knowles) [25] for measuring residual noise were installed on the headliner in all seat positions. Virtual microphones, which were used only in the tuning stage, were located at the left and right ear positions of each corresponding seat. Among the audio speakers, a total of five speakers, a 16-cm woofer of each door, and a 25-cm subwoofer were used as noise control actuators. For narrowband powertrain noise control, the engine revolution speed from the engine management system electronic control unit (EMS ECU) via high-speed CAN bus was used as the reference signal. Two engine orders, that is, C1.5 and C2.0, which are dominant for acceleration booming noise, were considered as control targets. The frequency range of ANC for engine harmonics was approximately 17 to 200 Hz, equivalent to 1,000 to 6,000 rpm corresponding to the engine orders. Since the EHC targets the booming noise due to the powertrain in the low frequency range, only one error microphone was used for each seat to simplify the control system. Among the microphones on the left and right at each seat, the microphone on the window side, which has a higher noise level, was used for acquiring the error signals. Thus, four microphones, EM1, EM4, EM5, and EM8, were selected to acquire the error signals. In addition, four virtual microphones corresponding to each physical microphone were used. For broadband road noise control, four three-axis accelerometers were used as the reference sensor to acquire a coherent signal with error microphones at the vehicle cabin, according to the TPA results presented in Section 3. Accelerometers, which are MEMS-type sensors, (ADXL316, ANALOG DEVICES) were installed at the subframe and rear crossmember.
Furthermore, the microphone and accelerometer were connected to the controller via an automotive audio bus (A2B) interface (AD2425, ANALOG DEVICES). The frequency range of noise control was approximately 20 to 400 Hz, which is the structure-borne noise region of road noise. In RNC, a zone of quiet is naturally reduced than EHC because the control frequency is higher than EHC [30]. Thus, eight error microphones and eight virtual microphones were used, considering both left and right ear at the entire seat. Four error microphones, i.e., EM1, EM4, EM5, and EM8, and four virtual microphones corresponding to each error microphone were shared with EHC. Other microphones, EM2, EM3, EM6, and EM7, were used only for RNC. Figure 7 illustrates the layout of the error microphones and accelerometers in the test vehicle. Figure 8 presents the multiple coherence function between the virtual error signal and accelerometer signals. It can be seen that the coherence between the reference signal and the error signal is high within the control frequencies.

A. ROAD NOISE CONTROL
Two experiments, only RNC and concurrent control, were conducted to compare the noise attenuation performance. Concurrent control means the combination of RNC and EHC. The experimental results for each control are illustrated in Fig.  9. In Fig. 9, the gray solid, blue dashed, and red dotted lines indicate the A-weighted sound pressure level before control, RNC (EHC off, RNC on), and both EHC and RNC (EHC on, RNC on) result with respect to frequency, respectively. The sound pressure level was measured using 1/2-inch microphones at both the left and right ear positions of the four seats, and the left-hand and right-hand sides of the front and rear seats. The interior noise was measured at a constant speed of 60 km/h on a rough road surface for ten seconds. As shown in Fig. 9, in both experiments, broadband booming noise below 200 Hz and tire resonance noise of approximately 200 Hz were dramatically attenuated over almost the entire frequency region at all microphone positions. The peak noises around 74 Hz, 114 Hz, and 176 Hz of the road booming noise region were significantly attenuated by approximately 5 to 10 dB. The mean of the attenuation for each seat in the control frequency region was 3.49 dB and 3.59 dB in only RNC and concurrent control, respectively. Therefore, it seems that the performance of noise attenuation in the two experiments was almost similar within the structural road noise frequency region, 20 to 400 Hz. Next, we analyzed whether the EHC signal had an adverse effect on the same control frequency of the road noise combined with the engine harmonic noise. During the measurement of the interior noise, the engine revolution speed was 1,750 rpm with the fifth gear. Thus, the frequencies of the C1.5 and C2.0 orders were 43.8 Hz and 58.3 Hz, respectively. Figure 10 shows the comparison of noise attenuation for frequencies corresponding to the C1.5 and C2.0 orders in the two experimental results. In Fig. 10, the blue solid and red dashed lines indicate the attenuation at only RNC and concurrent control, respectively. Measurement positions 1 to 8 are the same as the positions of the virtual microphones. The average noise attenuations in the seats were 2.12 dB and 3.88 dB for the C1.5 and C2.0 orders in the RNC experiment, respectively. Furthermore, the average noise attenuations in the seats were 2.60 dB and 4.43 dB for the C1.5 and C2.0 orders in concurrent control, respectively.
Although the attenuation of the C1.5 order was slightly reduced by 0.1 to 0.2 dB at the rear left seat, that is, measurement position numbers 5 and 6, during concurrent control, it can be noticed that the attenuation of both the C1.5 and C2.0 orders overall had a slightly increase by concurrent control than when only the road noise was controlled. Figure  11 shows the difference in attenuation between the two experiments. The attenuation difference was calculated by subtracting the attenuation of the RNC from the attenuation of the concurrent control. The average attenuation at all measurement points of the C1.5 order showed a slight increase of approximately 0.48 dB (standard deviation 0.52 dB) in a concurrent control experiment. In addition, for the C2.0 order, attenuation slightly increased by an average of 0.55 dB (standard deviation 0.44 dB).

B. ENGINE HARMONIC CONTROL
Three experiments, only RNC, only EHC, and concurrent control, were conducted to compare the noise attenuation performance. The interior noise was measured at a wide-open throttle with a third gear condition on the road with the general surface. The sound pressure level was measured using a 1/2inch microphone at the position of the ear on the window side of the entire seat. That means the left and right ear positions for the driver's seat and the front row passenger seat, respectively. Figures 12 and 13 present the experimental results, where the gray thin solid, blue dashed, green dotted, and solid red lines indicate the C-weighted sound pressure level before control, after Exp. 1 (EHC off, RNC on), Exp. 2 (EHC on, RNC off), and Exp. 3 (EHC on, RNC on) with respect to the engine revolution speed.
It can be seen that for the C1.5 and C2.0 orders, the road noise controller overall contributes to noise attenuation over almost the entire RPM range because the accelerometer was installed on the overlapping transfer path of the powertrain induced and road-induced booming noise frequency region. For the C1.5 order, the average attenuation from 1,000 to 6,000 rpm was 1.41 dB, 2.11 dB, 1.16 dB, and 4.11 dB, whereas for the C2.0 order, the average attenuation was 0.63 dB, 2.11 dB, 2.35 dB, and 1.67 dB at the positions of the driver, front passenger, rear left, and rear right seat, respectively. The average attenuation in the four seats was 2.20 dB and 1.69 dB for C1.5 and C2.0 order, respectively.
Moreover, the attenuation of EHC was more significant than that of RNC. For harmonic control, attenuations of up to 7 dB and 12 dB were found in the C1.5 and C2.0 orders, respectively. For the C1.5 order, the control performance was somewhat insufficient in the range below 3,500 rpm. However, the average attenuations from 1,000 to 6,000 rpm were 3.66 dB, 4.32 dB, 5.77 dB, and 3.75 dB, and for the C2.0 order they were 3.55 dB, 4.14 dB, 6.62 dB, and 4.91 dB at the positions of the driver, front passenger, rear left, and rear right seat, respectively. The average attenuation in the four seats was 4.38 dB and 4.81 dB for C1.5 and C2.0 order, respectively.
Furthermore, for concurrent control, it was found that the attenuation performance was higher than that of the EHC. Because the vibration signal from the powertrain is input to the reference signal of the road noise controller under the acceleration condition, the control signal in the road noise controller is positively added to the suppression of the engine harmonic noise in the concurrent control.
For the C1.5 order, the average attenuations from 1,000 to 6,00 rpm were 5.90 dB, 6.11 dB, 9.03 dB, and 4.79 dB, whereas they were 3.41 dB, 5.33 dB, 8.99 dB, and 6.47 dB for the C2.0 order at the position of the driver, front passenger, rear left, and rear right seat, respectively. The average attenuation in the four seats was 6.46 dB and 6.05 dB for C1.5 and C2.0 orders, respectively.
A comparison of the noise attenuation according to the measurement position is illustrated in Fig. 14. For the C1.5 order, additional attenuations of 2.24 dB, 1.79 dB, 3.26 dB, and 1.04 dB were observed for the C1.5 order, and -0.14 dB, 1.19 dB, 2.37 dB, and 1.56 dB for the C2.0 order, compared to the EHC, at front left, front right, rear left, and rear right seat positions, respectively. The average additional attenuation in the four seats was 2.08 dB and 1.25 dB for the C1.5 and C2.0 orders, respectively. Although a slight ineffectiveness was observed in the front left seat position, for concurrent control, the RNC had a positive effect on engine noise control, thereby improving the performance of the EHC.

C. DISCUSSION
Active noise control can mainly be categorized into feedforward, feedback, and hybrid control. Hybrid control combined feedforward and feedback control. Many researchers have researched and improved on hybrid control to control narrowband and broadband noise in a vehicle [4], [33]- [35]. Among the results related to the hybrid control, a recent study [4] showed that the noise reduction result of 4.67 dB was derived applying the hybrid control when the simulation was performed at a constant speed of 60 km/h. They presented an improvement of 0.38 dB from the 4.29 dB reduction in only broadband control by applying the method of separating the error signal of the broadband and the narrowband controller. Although a new approach was attempted from the control algorithm aspect, the study had limitations that were not experimentally validated.
On the other hand, this study has investigated the concurrent implementation of feedforward controllers with different noise reduction goals, rather than a general hybrid control technique in which feedforward control and feedback control are combined. It is a more practical approach for vehicle ANC because it can effectively reduce powertrain noise and road noise, respectively, and can lead to a complementary effect if the strategy of positioning the reference sensor presented in this study is followed.
In general, the adaptive notch filter has been applied for EHC. The noise suppression performance of the adaptive notch filter significantly depends on the frequency accuracy of reference signals [16]. The performance of EHC can be deteriorated according to the frequency change of reference signals [36]. In modern active noise control for vehicles, the engine speed is obtained via CAN bus instead of physical sensors [13] to reduce the cost. It leads to performance degradation due to the frequency inaccuracy of reference signals resulting from low data rates and delays. In particular, it may be occurred more significantly in rapid acceleration conditions than in the condition of slow acceleration or constant speed driving. In this study, we have arranged reference sensors of the RNC controller to the overlapped transfer path of engine harmonics and road noise. Therefore, the reference signals of the RNC controller can pick up the vibration for not only road noise but also engine harmonics, which leads to RNC suppressing the engine harmonics. Thus, in concurrent control, RNC can supplement the performance of EHC, which is degraded by applying CAN bus.
Meanwhile, for concurrent control, the previous study presented error microphone positioning [37] to identify the optimal layout of error microphones, leading to optimal noise suppression in EHC and RNC. In that study, the ANC performance was evaluated by constructing various microphone layouts. It was experimentally shown that the correlation between microphone layout and noise attenuation differed in EOC and RNC. In that study, a feedforward-based concurrent control system similar to this study was constructed. However, only the performance of each controller according to the position of error microphones was evaluated. Although the effect of concurrent control has not been evaluated, it is possible to determine the appropriate position of the error microphone in concurrent control through their result. Furthermore, the acoustic mode in the vehicle cabin should also be considered, although not covered in this study. Nodal lines should be avoided as much as possible within the control frequency region. EHC has been commercialized in the automobile field and is now being applied as a general technology. However, RNC has been commercially available since 2020 [38], [39], and its application is expanding along with the development of digital signal processing technology. Therefore, it is expected that the demand for concurrent control to reduce both engine booming noise and road noise is expected to increase in internal combustion engine (ICE) vehicles to improve the NVH performance. If the strategy proposed in this study is applied to concurrent control, a complementary effect on engine harmonics and road noise can be expected.

VI. CONCLUSION
In this study, concurrent control, which can suppress the engine harmonic noise and road noise, was investigated to achieve acoustic comfort inside a vehicle cabin. Moreover, a practical implementation was developed to further suppress engine harmonic noise with control signals from the road noise controller through concurrent control. The same is true for the opposite case. For this purpose, the noise characteristics of engine boom noise during acceleration and road noise were investigated. Furthermore, the path contribution and VATF were analyzed by structural TPA to select the proper position of accelerometers as the reference signal for RNC. Through the TPA result, a transfer path that can cover two noise sources, engine boom and road boom noises, was identified, and through this, the location of the reference sensor of the road noise controller was established. Accelerometers were installed around the body and chassis connection points, that is, the subframe and rear crossmember. To validate the performance of the developed control system, a practical adaptive feedforward active noise control system was implemented using a vehicle audio system.
For RNC, the experimental results presented an average attenuation of 3.49 dB and 3.59 dB in the entire seat position for frequencies up to 400 Hz in RNC and concurrent control, respectively. Moreover, in terms of road noise, the control signal from the engine harmonic controller only suppressed the noise corresponding to the engine harmonic order. Furthermore, it does not consider significant adverse effects that act as additional disturbances. Although an additional noise suppression effect was found by the EHC signal, this was not significant on the entire side of the road noise. Meanwhile, for EHC, the experimental results indicated an average attenuation in the four seats of 4.38 dB and 4.81 dB for C1.5 and C2.0 orders, respectively. Furthermore, it was experimentally determined that the engine harmonic noise can be additionally controlled through RNC in concurrent control. The experiment showed an additional average attenuation of 2.08 dB and 1.25 dB of the C1.5 and C2.0 orders. It was found that if the position of the reference sensor was selected in the overlapping path of engine booming noise and road noise, the RNC signal was effective for additional control of the engine harmonic noise.