Wireless Communication in Modular Multilevel Converters and Electromagnetic Interference Characterization

The wireless control of modular multilevel converter (MMC) submodules was recently proposed. The success of the control depends on specialized control methods suitable for wireless communication and a properly designed wireless communication network in the MMC valve hall while aiming for low latency and high reliability. The wireless communication in the hall can be affected by the electromagnetic interference (EMI) of MMC submodules, voltage and current transients. In this article, firstly, a wireless communication network based on 5G New Radio is designed for an example full-scale MMC valve hall. After that, the radiated EMI characteristics of the MMC submodules with different voltage and current ratings and two dc circuit breakers are measured. The effects of EMI on wireless communication in the multi-GHz frequency band are tested. The interference from the components is confined below <inline-formula> <tex-math notation="LaTeX">$\mathrm {500 \text {M} \text {Hz} }$ </tex-math></inline-formula>, and the wireless communication with <inline-formula> <tex-math notation="LaTeX">$\mathrm {5825 \text {M} \text {Hz} }$ </tex-math></inline-formula> center frequency is not affected by the interference.


I. INTRODUCTION
The modular multilevel converter (MMC) is extensively used in high-voltage multi-megawatt applications such as high-voltage dc (HVdc) transmission systems [1]. The communication between the MMC central controller and the submodules is commonly realized by optical fiber cables as they are reliable, immune to electromagnetic interference (EMI), and provide galvanic isolation between the connecting entities [2]. On the other hand, the total number of MMC submodules to reach the targeted power ratings might be in the range of thousands [3], and the dimensions of MMC valve halls might rise to hundreds of meters [4] for HVdc transmission applications. Therefore, substantial issues arise from cable-based deployment of large MMC valve halls: 1) Laying out, terminating, and insulation-testing thousands of fiber cables with hundreds of meters in length requires a significant workforce and time during the installation of the MMC. Moreover, each cable needs to The associate editor coordinating the review of this manuscript and approving it for publication was Tariq Masood .
be identified and checked whether it connects the correct terminals during the commissioning of the MMC, which further increases the required workforce and time [5]. The required resources for all these processes contribute to the initial cost of the MMC station, with several other factors mostly specific to the project. Furthermore, in case of a delay in the commissioning of the MMC, significant fines might be incurred to the MMC contractor, which might correspond to or even exceed the profit margin. Examples of voltage-sourced converter overall costs in HVdc transmission applications, which are considered to be mainly MMCs, are available in the literature: The cost of a converter station per MW ranges from 218 ek/MW to 429 ek/MW in [6] for eight different projects with power ratings from 700 MW to 3000 MW. Cigre exemplifies the converter cost as 102 ek/MW and 220 ek/MW for two projects with 1500 MW and 1000 MW ratings [7]. The share of fiber-cable-related costs in the overall costs of an MMC station is not publicly available to the authors' best knowledge. On the other hand, typical commissioning and erection costs for a new HVdc facility are estimated VOLUME 10, 2022 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ as 8 % of the overall costs in [8], while the exact cost structure is unclear.
2) The routing of the thick cable bundles in the MMC valve hall can be cumbersome and requires a resourceintensive mechanical design process. Further, the bundles might complicate the replacement of the (failed) submodules during the maintenance of the MMC, thus increasing the maintenance time. 3) With thousands of submodules and hundreds of meter valve hall dimensions, the total length of optical fiber cables in an MMC valve hall is possibly in the range of hundreds of kilometers. This much of cables lead to an increase in the converter footprint, weight, and volume, which are problematic, especially in off-shore and in-city transmission platforms [9]. 4) Although the insulation of the fiber cables is tested during the installation, the cables present an insulation breakdown risk between the central controller (lowvoltage area) and the submodules (high-voltage area) due to, e.g., accumulation of dust and humidity on them over time. In order to minimize the risk, proper air conditioning and filtration are required in the valve hall, which might increase the costs. 5) The fiber cables risk spreading fire in the MMC valve hall [10]. To minimize the risks, cables made from special material and with coatings are to be used, such as flame-retardant, non-combustible, or selfextinguishing, which increase the cost and might complicate the layout of the bundles due to their poorer mechanical properties. 6) The fiber cables might be subject to mechanical stress and fatigue, which leads to failure. The share of Control and Protection systems, including Control and Protection Telecommunications subsystem, in the Forced Energy Unavailability hours of the HVdc systems worldwide, is 7.8 % between the years 1983-2016 and 7.1 % between 2017-2018 [11]. The two other subsystems included in the figure are Local Control and Protection and Master Control and Protection. Wireless control of MMC submodules is proposed in [12], implemented and verified in [13] to solve the issues above. The control method is advanced in [14] so that the submodules experiencing communication errors that last up to hundreds of wireless transmission cycles (tens of milliseconds) can continue their operation as if no error occurred. The method is advanced further in [15] against longer-lasting communication errors (hundreds of milliseconds and more) so that overvoltages, overcurrents, or component failures in the MMC are avoided. The verification of the proposal and advancements in [13]- [15] are all based on a laboratory-scale single-phase MMC with three half-bridge submodules per arm. The experimental setup does not represent a full-scale MMC and its valve hall in terms of the dimensions, the number of communication nodes, and electrical characteristics. Thus, it is unclear how to implement a wireless communication network in a full-scale MMC valve hall. Moreover, if the radiated electromagnetic interference (EMI) from the MMC components coincides with the wireless transmission frequency, higher transmission errors can be expected, which can complicate the wireless control of the submodules. Highpower transients resulting from, e.g., electrical arcs and circuit breaker operations, can also interfere with wireless communication [16], [17].
The steady-state operation of full-scale MMCs, which mainly represents the switching of the semiconductors in the submodules, is observed to generate interference up to 1.4 GHz but has significant effects up to 400 MHz in and out of the MMC hall [18], [19]. In another study, the interference in the MMC hall is measured up to 600 MHz [20].
The contributions of this article are twofold. First, a wireless network design is provided for a full-scale MMC based on the wireless control method given in [13]. A theoretical case study is carried out on an example MMC used in HVdc transmission. 5G New Radio (NR) is investigated as the wireless communication solution in the MMC valve hall. Also, antenna placement options in the hall and the related communication link budget and reliability are discussed. Wireless control of the submodules can be enlarged to full-scale MMCs using the proposed wireless network. As the second contribution, the article presents a more realistic radiated EMI environment of a full-scale MMC compared to that in [13] and observes its effects on wireless communication. To this end, firstly, the voltage and current rise rates of the state-ofthe-art high-power semiconductors are determined. Then, the radiated EMI (henceforth called just EMI) of two different laboratory-scale submodules are measured while having voltage rise rates similar to that of high-power semiconductors. Also, the EMI of two different dc circuit breakers (DCCBs) are measured with 1 kA peak interrupted current. Lastly, wireless communication packet loss characteristics are measured in the close vicinity of a submodule to observe whether the EMI from the submodule affects the communication. The measurements show that the EMI from the components are confined below 500 MHz (parallel to the results in [18]- [20]) and, thus, a wireless transmission in the multi-GHz range is not affected by the EMI.
The structure of the article is as follows. In Section II, a wireless network is designed for an example full-scale MMC valve hall. In Section III, the EMI spectrum of MMC components is analyzed in a laboratory environment. Section IV contains wireless communication packet loss characteristics when an MMC submodule operates nearby and not. Conclusions are given in Section V.

II. WIRELESS NETWORK DESIGN CASE STUDY IN AN MMC VALVE HALL
In this section, the wireless control method for MMC submodules implemented in [13] and [14] is taken as the basis, and a wireless network is designed for an example fullscale MMC. In the mentioned implementations, the central controller periodically broadcasts the real-time control data to all the submodules in a single data package in the 38190 VOLUME 10, 2022  downlink (DL). The communication from the submodules to the central controller, i.e., uplink (UL), is not included in the discussions below as communication in UL is not required to be real-time with the proposed control method in [13]. The conceptual wireless communication network is illustrated in Fig. 1.
The MMC stations used in the HVdc electrical interconnection project between Baixas, France and Santa Llogaia, Spain are chosen as the example application to design the wireless network [4]. The outside and inside views of the MMC station valve hall in Baixas are shown in Fig. 2. Considering other photos of the halls, the hall building looks like a big Faraday cage that attenuates the electromagnetic interference from outside to the valve hall and vice versa. The plant parameters to implement the wireless network are shown in Table 1 and the application requirements are shown in Table 2. Parameters in Table 2 are taken from the laboratory-scale wireless control implementations in [13] and [14], except that 48 % of the payload in Table 2 is spare compared to the mentioned implementations. The transmission cycle, which is set by the cycle time of the central controller algorithms, and the wireless payload in Table 2 result in a 2.56 Mbit s −1 raw data rate in the DL.

A. WIRELESS NETWORK IN THE VALVE HALL
MMC submodules control is a time-critical process. It requires low end-to-end latency and high reliability in the control system. The typical sampling period in the control system of the MMCs (and similar power electronics applications) is in the range of 100 µs [13]. The packet error rate (PER) requirement in the application layer for power electronics applications is in the 10 −8 to 10 −9 range in [22], which is not reachable with the mentioned sampling (cycle) periods using the current or upcoming wireless communication technology. In order to realize the wireless control of submodules, it is essential to adjust the submodules control such that it can tolerate higher PER values. Using the control method in [14], increasing the suitable PER value at least up to 10 −2 range is possible, which can be reached by the contemporary low-latency, high-reliability wireless communication solutions without compromising the performance of the MMC. In this sense, 5G NR, which is standardized with 5G release 15, is a promising ultra-reliable low-latency communication (URLLC) solution [23], [24]. The latency bound, packet size, and reliability requirements in a 5G system are set as 1 ms, 32 B, and 99.999 %, respectively, by ITU-R in the IMT-2020 specifications [25]. The packet size and reliability requirements are compatible with the application requirements given in Table 2; however, the latency performance of 5G NR should be investigated further to check whether it can be lowered to the sub-millisecond range.
5G NR operates in frequency bands in two frequency ranges: Frequency Range 1 (FR1) for bands within 410 MHz to 7125 MHz, and Frequency Range 2 (FR2) for bands within 24 250 MHz to 52 600 MHz [26]. 5G has a 10 ms long frame structure in the physical layer. Each frame comprises ten subframes with a 1 ms length, and each sub-frame is made up of scalable transmission time intervals (TTIs). One or more consecutive slots allocated to either DL or UL make up a TTI.
Each slot contains 14 orthogonal frequency-division multiplexing (OFDM) symbols, and the length of each slot can be in the range of 125 µs to 1 ms. The length of each OFDM symbol and thus the slot depends on the employed sub-carrier spacing (SCS) frequency, which is 15 kHz (corresponding to 1 ms slot length), 30 kHz, or 60 kHz for operation in FR1, and 60 kHz or 120 kHz (corresponding to 125 µs slot length) for operation in FR2. Hence, the shortest achievable slot duration, T slot , for data transmission with the full 14 OFDM symbols is 250 µs for the operation in FR1 and 125 µs for the operation in FR2. Consequently, 5G NR with a full slot structure cannot respond to the cycle time given in Table 2. On the other hand, 5G NR introduces mini-slot transmissions with 2, 4, or 7 OFDM symbols in the DL. Considering that a single OFDM symbol duration is 17.84 µs and 8.92 µs with SCS equal to 60 kHz and 120 kHz, using mini-slots consisting of 2 OFDM symbols, the shortest achievable T slot can be lowered to 36 µs and 18 µs in the DL with the operation in FR1 and FR2, respectively.
The total physical layer latency, T L , includes the data processing time intervals in the transmitter (central controller), T ptx , and receivers (submodule controllers), T prx , on top of T slot as [27] with no re-transmission or acknowledgment. The processing time depends on several scheduling factors, which are out of the scope of this article, and the processing capabilities of the transceivers. In [28], as a worst-case scenario, T ptx is assumed equal to five times T slot for SCS equal to 60 kHz or 120 kHz for mini-slots with 2 OFDM symbols. Similarly, T prx is assumed equal to five and ten times T slot for SCS equal to 60 kHz and 120 kHz, respectively. The mentioned T ptx and T prx values with the aforementioned T slot end up T L being equal to roughly 400 µs and 300 µs for the operation in FR1 and FR2, respectively, using frequency division duplex (FDD) mode. (The FDD mode promises shorter processing delay than time division duplex mode as it has shorter alignment delay, i.e., the time from when data becomes available until the next transmission slot starts.) The mentioned T ptx value includes some UL related factors, which are not needed in our scenario. Also, the T prx assumption for 120 kHz SCS is for receiver hardware with basic processing speed. Thus, lower T ptx and T prx values can be obtained. In [29], for example, T ptx and T prx are assumed equal to T slot , which makes T L equal to 108 µs and 54 µs for SCS equal to 60 kHz and 120 kHz, respectively. Hence, the exact T L depends on the specific application. Although it does not seem like a straightforward target, the authors do not rule out the possibility of reaching 100 µs transmission cycle with 5G NR, especially with the operation in FR2 with SCS equal to 120 kHz. The key to reaching the target is to reduce T ptx and T rtx to the same range of T slot , which may require advanced transceivers.
In case 100 µs transmission cycle is not feasible (e.g., the required hardware costs too much), a workaround would be increasing the cycle time of the central controller algorithms and consequently the transmission cycle, e.g., to 200 µs.

B. ANTENNA PLACEMENT IN THE VALVE HALL AND COMMUNICATION LINK BUDGET
The wireless communication eliminates the workload and costs related to laying out, terminating, and testing the optical-fiber cables between the submodules and the central controller. On the other hand, the placement of the antennas for the central controller and the submodules is introduced as a new task. The reliability of wireless communication depends on the signal-to-interference-plus-noise ratio (SINR) at the receivers, which is dependent on the placement of the transmitter and receiver antennas. Positioning the antennas, the type of antenna to use, and connections between the antenna and the transceiver circuitry require a comprehensive engineering study to minimize the PER and not to present any threat to electrical safety. The critical points for the antenna placement are listed below: 1) The antenna placement should maximize the received radio power considering the other constraints. A lineof-sight (LOS) is desirable.
2) The propagation environment, i.e., MMC valve hall, is complex to model with various conductive elements and metal bars. The assessment of the antenna placement would require detailed design, simulations, and several tests in the valve hall considering the PER.
3) The antennas should not present an electrical safety risk in the high-voltage environment. Hence rod-like and pointy antennas should be avoided. The antennas cannot protrude outside the valve assembly envelope. The electric field redistribution in the hall should be taken into account.
The control room of the MMC in Fig. 2 is placed towards the corner of the rectangular valve hall. The central controller is proposed to have multiple antennas in the hall placed on the ceiling and, if required depending on the antenna placement options below, on the sidewalls of the hall building. The antennas should be placed such that preferably all the submodule antennas have a LOS to a central controller antenna.
An enclosure around the submodule control units is common to provide electromagnetic immunity from nearby interference sources. If the submodule antennas are placed inside the enclosure, the received radio power will be attenuated by the enclosure depending on its material and thickness. We propose placing the antennas outside the enclosures to avoid attenuation.
Three antenna placements options for the submodules and the central controller are considered in this article, where the central controller and submodule antennas are referred to as the transmitter and receiver antennas, respectively: 1) One (set of) antenna per submodule within the submodule structure; one (set of) transmitter antenna on the building ceiling for submodules that are placed faceto-face in two adjoining valve towers and one (set of) transmitter antenna on the building sidewalls for the submodules of a valve tower that are facing towards the walls; 2) One (set of) antenna per submodule on the valve tower ceiling, the antenna (set) is connected to its submodule over a cable; one (set of) transmitter antenna per tower on the building ceiling; 3) One (set of) antenna per valve tower on the tower ceiling connected to all the submodules in the tower over cables; one (set of) transmitter antenna per tower on the building ceiling. The options are illustrated in Fig. 3 and compared from different perspectives in Table 3. It is considered that option 3 should be avoided as it presents a single point of communication failure in the valve tower. Option 2 (similar to option 3) has a LOS with a short distance between the transmitter and receiver antennas. However, the electric field in the valve tower might induce high potential difference and harmful currents between the submodule controller and its antenna on the ceiling of the valve tower over the cable connecting them. In order to avoid these, the submodule controller and the antenna need to be galvanically isolated by, e.g., electrical to optical conversion on both ends of the cable, which adds extra latency in the control system. For option 1, submodules facing the building walls can have a LOS to transmitter antennas placed on the walls. Also, for submodules placed face-to-face in two adjoining valve towers, LOS to a transmitter antenna placed diagonally on the ceiling seems possible. However, a detailed investigation is required to ensure LOS for all the submodules in the valve tower. Considering Fig. 2 and the dimensions of the valve hall in Table 1, the distances between the transmitter and receiver antennas are estimated at around 5 m in options 2 and 3. The maximum distance is estimated at around 10 m for option 1.
5G NR operates in frequency bands in two frequency ranges, FR1 and FR2, as expressed in the previous subsection. A great majority of the specified bands are reserved for 5G NR operation. Thus, interference from the ISM bands, which are reserved internationally for industrial, scientific, and medical (ISM) purposes and usually unlicensed, is highly unlikely for 5G NR unless it operates in one of the few frequency bands that coincide with the ISM bands, e.g., n53.
The 5G NR link budget for the antenna placement option 1 is calculated in an online tool [30]. (Options 2 and 3 have a shorter distance between the transmitter and receiver antennas.) The parameters for the calculations are given in Table 4. Additionally, Indoor Hotspot 3D-InH LOS + NLOS propagation model is chosen. For FR1 operation, the 5G frequency band with FDD mode and the highest transmission center frequency -n7-is selected. The resulting SINRs are 50 dB and 30 dB for transmission in FR1 and FR2. The SINR values are so high that the resulting PER is low enough to satisfy the 10 −2 to 10 −3 requirement in Table 2 for almost any modulation and coding scheme [28], [31]. As a result, option 1 seems like a better alternative considering all the factors mentioned above and given in Table 3. The definite choice would require detailed analyses and tests to verify the SINR calculations, PER, and whether the errors can be compensated by other measures like the autonomous submodule controllers [14].

III. EMI SPECTRUM ANALYSES OF MMC COMPONENTS
In the previous section, a 5G NR communication system operating in the multi-GHz range is evaluated for wireless communication in the MMC valve hall. This section analyzes the EMI of two different submodules and two different DCCBs to identify whether they coincide with the 5G NR transmission frequency bands. For this purpose, measurements are performed in a laboratory environment. In all the measurements, the following infrastructure is shared. The electromagnetic (EM) spectrum is analyzed with a Rohde&Schwarz FSH8 spectrum analyzer. A BicoLOG 5070 X biconical antenna and an MVG QH800 open boundary quad-ridge horn antenna are used to measure EMI in 50 MHz to 700 MHz and 0.8 GHz to 6 GHz frequency ranges, respectively. The spectrum analyzer is set to record the maximum measured signal during the measurements, and the measurements remained for 10 minutes for the submodules.
EMI from power electronics applications is dependent on the power semiconductor switching characteristics. Table 5 shows the dynamic performance test ratings and rise rates of the voltage and current (dv/dt and di/dt) for the stateof-the-art insulated-gate bipolar transistors (IGBTs) used in high-power applications. The current and voltage rise rates  are calculated using the rise and fall times of the current, respectively, stated in the datasheets of the IGBTs with the given test conditions [32]. It is considered that the table gives the correct order of magnitude for the rates, even if not the correct values, for the submodules in full-scale MMCs.

A. FULL-BRIDGE SUBMODULE WITH SI MOSFETs
The experimental setup is shown in Fig. 4. The full-bridge submodule comprises IRFP4768PbF silicon MOSFETs (the same MOSFETs used in the experimental study in [13]) and two parallel-connected 30 µF film capacitors on the dc-side.
The submodule electrical circuit and the controller are assembled on a single printed circuit board. The controller is based on the Xilinx Zynq-7000 system-on-chip. The submodule is connected to a voltage source on the dc-side by a twisted pair of cables. The ac-side is connected to variable resistors by a twisted pair of cables. Measurements are taken when the submodule is enclosed in an aluminum 19-inch rack and not enclosed. When the submodule is enclosed, the dc-and ac-side cable pairs are shielded, and the shields are earthed on both ends. Otherwise, no shield is used. When used, the aluminum rack is earthed via the cabinet. The experiments are performed with 100 kHz switching frequency and unipolar pulse-width modulation with 0.95 amplitude modulation index. The voltage measurements are taken directly from the MOSFET terminals. During the experiments, no other highpower equipment is operated nearby.

1) RESULTS
Combinations of two dc-side voltages, 100 V and 150 V, and four ac-side peak currents, 0 A, 10 A, 20 A, and 30 A, are experimented. The results are tagged in the figures as  volts/amperes according to the voltage and peak current level used in the experiment. For instance, 100/10 stands for the experiment run with 100 V dc-side voltage and 10 A peak acside current. R-NoSw is for the reference experiment in which the converter and the controller are not energized. R-Sw is for the reference experiment in which the controller is energized, and the MOSFETs are switching, but the converter has 0 V dc-side voltage. The experiments with nonzero dc-side voltage are referred to as active cases hereafter. Fig. 5 shows the submodule ac-side voltage v ac of the active cases 100/0 and 150/0 right after the turn-on of the respective MOSFETs that result in positive ac-side voltage. 100/0 and 150/0 have smooth waveforms with ≈ 0.5 V/ns and ≈ 0.6 V/ns voltage rise rates, respectively. No snubber is used for the MOSFETs.
To distinguish the EMI of different test cases, the differences in the EMI of the active cases from the reference R-NoSw are shown in Figs. 6 and 7 for the 150 V dc-side voltage active cases with and without the aluminum rack, respectively. The figures reveal that as soon as the switching starts in the submodule, even with zero dc-side voltage, EMI in the low-frequency range from 50 MHz to 300 MHz is 10 dB to 20 dB higher than R-NoSw. Having a nonzero dc-side voltage or higher ac-side current does not affect the EMI particularly. Compared to the R-NoSw reference, the higher EMI is mainly related to the switchings and the driver circuit. Comparing Figs. 6.(a) and 7, the aluminum rack decreases the interference from 100 MHz to 300 MHz up to 10 dB. In the high-frequency range, there are some outliers from   Fig. 6.(b). The legend of the figure is the same as in Fig. 6. 800 MHz to 900 MHz, 2.4 GHz to 2.7 GHz, and 5.2 GHz to 5.6 GHz; however, no continuous frequency range has higher interference than the reference. The outliers in the 800 MHz to 900 MHz range are considered sourced from GSM, 3G, and 4G networks, and those in 2.5 GHz to 2.7 GHz are from 4G networks in Sweden [33]. The outliers in the 2.4 GHz to 2.5 GHz and 5.2 GHz to 5.6 GHz ranges are considered sourced from the nearby WLAN networks. Consequently, the experimented full-bridge submodule does not affect the EM spectrum above 300 MHz. It should be noted that the voltage rise rates in the experiments are an order of magnitude smaller than those in Table 5.

B. HALF-BRIDGE SUBMODULE WITH SiC MOSFETs
The experimental setup is shown in Fig. 4. The halfbridge submodule comprises a CAS300M12BM2 half-bridge silicon-carbide MOSFET module and two parallel-connected 680 µF film capacitors. The submodule is enclosed in an aluminum enclosure. The half-bridge driver is placed in the enclosure, and it has optical fiber switching pulse inputs from the submodule controller, which is placed out of the enclosure. The controller is based on the Xilinx Zynq-7000 systemon-chip. The submodule and the controller are galvanically isolated. The submodule is connected to a voltage source on VOLUME 10, 2022

1) RESULTS
Combinations of two dc-side voltages, 200 V and 400 V, and four ac-side peak current levels, 0 A, 10 A, 20 A, and 30 A, are experimented. The results are tagged in the figures as explained in the previous subsection. Test case R stands for the reference case in which the converter has 0 V dc-side voltage, but all the control circuits and auxiliary setup are up and running. In Fig. 8, the ac-side voltage enlarged waveform of 200/0 and 400/0 measurements right after the low-side MOSFET turn-off are compared. The active cases with 200 V and 400 V dc-side voltages have ≈ 14 V/ns and ≈ 19 V/ns voltage rise rates, respectively. The values are higher than those in Table 5. Oscillations in the waveforms are observed with peak-to-peak voltages of 220 V and 125 V for 200/0 and 400/0 active cases, respectively. The oscillations are fundamentally due to the resonant behavior of the parasitic inductance on the current path of the MOSFETs and the parasitic capacitance of the MOSFETs. The current rise rates of 200/30 and 400/30 active cases are 0.37 A/µs and 0.52 A/µs, respectively. They are much smaller than those in Table 5. The current rise rates refer that the experimental setup has a parasitic inductance in the hundreds of µH range.
The differences in the EMI of the active cases from the reference case R are shown in Figs. 9 and 10 for the 200 V and 400 V dc-side voltages, respectively. The low-frequency spectra from 50 MHz to 700 MHz present more diversity than the high-frequency spectra from 0.8 GHz to 6 GHz in both figures. A detailed investigation reveals that the differences are mostly individual low-bandwidth outliers in a single active case. The number and amplitudes of outliers in the 200 V cases are more than in the 400 V cases. This difference can be verified by comparing the variances of the active cases shown at the top of the figures. The higher peak-to-peak amplitude of the switching oscillations in 200 V active cases is considered to cause more and higher amplitude EMI outliers. There are a few noticeable frequencies around  which both the 200 V and 400 V cases have increased interference levels altogether by more than 5 dB. This behavior can be seen around 50 MHz and 470 MHz, but there is no relation between the increased ac-side current amplitude and the interference level. It is important to note that the LTE450 network is present at around 470 MHz [33]. In the highfrequency spectra, there are several outliers in the 800 MHz to 900 MHz and 2.4 GHz to 2.7 GHz ranges and around 5.5 GHz, which are considered originating from the GSM, 3G, 4G, and WLAN networks as detailed previously.
The measurements are repeated with an unmounted enclosure lid. The differences in the EMI of the active cases from the reference case are shown in Fig. 11 for 400 V dc-side voltage. All the active cases have up to 20 dB higher interference than the reference in 50 MHz to 350 MHz and 400 MHz to 500 MHz ranges. The interference amplitude decreases with the frequency. When Figs. 10 and 11 are compared, the interference from the submodule is confined mostly below 500 MHz. In the high-frequency range, individual outliers are seen in the active cases in 0.8 GHz to 1 GHz, 1.8 GHz to 2.2 GHz, 2.4 GHz to 2.7 GHz ranges, and around 5.5 GHz. Interference in the 1.8 GHz to 2.2 GHz range is considered to be originating from GSM, 3G, and 4G networks [33]. The other frequency ranges were explained previously. These results are promising indications that the wireless communication in the multi-GHz frequency band would function properly with no harmful interference from the submodule switchings even without proper EM shielding.

C. DC CIRCUIT BREAKERS
MMCs can be used to form HVdc grids. The reliable operation of HVdc grids requires protection against dc-side faults. MMC-HVdc terminals constructed with half-bridge submodules cannot control and limit the dc fault current and requires DCCBs for the protection [34]. AC circuit breakers rely on the current zero-crossing of the ac fault current to interrupt. However, dc fault currents do not have a natural current zerocrossing, and specialized DCCBs are needed to protect dc systems. A solution is the general DCCB structure depicted in Fig. 12. As shown in [35], the implementation of the general DCCB structure only affects the internal commutation processes between the parallel paths for the current, as in Fig. 12. Still, the internal commutation processes can cause different EMI throughout the operation of the DCCB because of the underlying physical phenomena and the switching transients during the commutations. Although it is unlikely that the DCCBs are placed directly in MMC valve halls and affect the  EM spectrum in the halls significantly, the EMI of DCCBs are still considered of interest as they would be placed in the same switchyard of the MMC station [36].
Two DCCBs, which operate similarly to the two major types of DCCBs, are used in dc interruption tests: a solidstate DCCB based on an integrated gate-commutated thyristor without snubber and an enhanced active resonant (EAR) DCCB based on a triggered vacuum gap as described in [37]. As in ac circuit breakers, the core component of EAR DCCBs, the triggered vacuum gap, operates with an electric arc discharge enclosed by the device. The experimental setup is shown in Fig. 13. The current i dc and voltage v dc as seen by the dc system at the terminals of the DCCB measured in experiments at 1 kA are given in Fig. 14. The switching transient of v dc indicates that injection-type DCCBs like the EAR DCCB commutate slower than semiconductor-based DCCBs. The maximum di/dt and dv/dt observed, considering the internal currents measured, are 61.0 A/µs and 14.3 V/µs and 80.7 A/µs and 137.8 V/µs for the EAR and solid-state DCCB, respectively. The EMI differences after the operation of DCCBs than before the operation of DCCBs are shown in Fig. 15. In both cases, the interference generated by the operation of the DCCBs is marginal, and it is observed that the primary source of EMI is the auxiliary circuits such as switching dc power supplies.

D. DISCUSSION
The EMI measurements of MMC submodules with voltage rise rates up to 19 V/ns and current rise rates up to 0.52 A/µs VOLUME 10, 2022 resulted in the interference from the submodules being confined below 500 MHz. The measurements with solid-state and EAR DCCBs have a marginal effect in the sub-GHz range and above. It is hypothesized that a wireless communication link with a transmission center frequency in the multi-GHz range, as proposed in Section II-A and as in [13], would not suffer from the EMI generated by the MMC submodules and DCCBs operating with similar parameters to those experimented in the previous subsections. The hypothesis will be experimented with in the next section.

IV. WIRELESS COMMUNICATION PACKET LOSS CHARACTERISTICS
In order to verify the hypothesis raised in Section III-D, packet losses of a wireless communication link are measured in the presence of the EMI generated by the specified halfbridge submodule in Section III-B, and the results are compared to a reference measurement. The measurement setup and the surrounding environment are sketched in Fig. 16.   the ceiling, and the floor material is semiconducting. Power panels with metal covers are placed on the sidewalls. All the surroundings present a unique EM wave propagation environment, which probably does not match a full-scale MMC hall. However, some resemblance can be argued with the metal frames, pipes, and power panels on the walls. It is considered that in an MMC hall, antenna placements with LOS and five to ten meters of distance might be realistic, as proposed in Section II-B. From the electrical rating point-of-view, the voltage rise rate of the employed submodule is higher than that of the considered full-scale MMCs, but the current rise rate is much smaller, as expressed in Section III-B1.
The wireless transceivers are based on WARP v3 kits. WARP project offers a reference IEEE 802.11 orthogonal frequency-division multiplexing physical layer (PHY) and a distributed coordination function medium access control (MAC) design [38]. In this study, carrier-sense and random backoff mechanisms of the MAC layer are disabled, and PHY is used for wireless communication. The wireless transmitter is connected to a controller serving as the central controller of an MMC. The wireless receiver is connected to the submodule controller as in Section III-B. The submodule controller is reached remotely during the measurements to observe the number of packet losses. The measurements are conducted in a stationary propagation environment. The employed wireless communication parameters are listed in Table 6. The transmission center frequency is towards the end of the 5 GHz band, and it is the same frequency employed in the previous experiments in [13] and [14]. Nevertheless, a majority of the frequency bands in FR1 are defined below 3 GHz for 5G NR [39].
The submodule is energized and de-energized consecutively for 10-minute periods during the measurements. When energized, the active case 400/10 is implemented. When de-energized, the reference R is repeated with the additional presence and operation of wireless transceivers. Table 7 shows the number of packet losses with specified loss lengths in consecutive packets for each 400/10 and R measurement with 8 m distance between the wireless transmitter and receiver. The other tables corresponding to 2, 4, and 16 m distances are not shown for brevity. The number of packet losses for each loss length and the PER are similar for the 400/10 and the R measurements in each campaign. The mean and standard deviation of the packet losses are close to each other. The packet losses are dependent on the distance between the transceivers, as expected. However, considering each campaign in itself, there is no indication that the EMI from the submodule causes more packet losses. Thus, the measurement results confirm that the EMI from the submodule does not deteriorate wireless communication held in the multi-GHz range.

V. CONCLUSION
This article first exemplifies a wireless network design on a full-scale MMC used for HVdc transmission. Considering the size of the MMC valve hall, the number of submodules, and MMC control system parameters, the 5G NR solution seems promising to provide the required broadcast cycle. Three options for wireless antenna placements in the MMC hall are envisioned and compared in different aspects. Central controller antennas on the ceiling and sidewalls of the hall building and one (set of) antenna per submodule integrated with the submodule structure might be preferential, but the definite choice requires more detailed analysis and experiments. The second part of the article measures the EMI of two different laboratory-scale MMC submodules and DCCBs. The interference from the submodules in the steady-state operation is confined below 500 MHz. Interference from the experimented solid-state and EAR DCCBs is marginal in the whole frequency spectrum. Later, the wireless communication errors are characterized in the vicinity of the halfbridge submodule that operates with a higher voltage rise rate than a full-scale MMC submodule. The results verified that the generated EMI of the submodule does not interfere with the employed wireless communication with 5825 MHz transmission center frequency. This is a promising result to expect similar behavior from the full-scale MMC submodules. Several other points require further investigation to ensure that the EMI from the MMC components does not deteriorate wireless communication. Some of them are the submodule measurements with higher current and rise rates, the EMI from electrical arcs in open space, and shootthrough faults, i.e., the faults over a low resistance path between the submodule capacitor positive and negative terminals. Also, the placement of components in a full-scale MMC hall presents a unique propagation environment and should be engineered accordingly to provide proper wireless communication.
TIM AUGUSTIN was born in Berlin, Germany. He received the Diplom-Ingenieur degree in electrical engineering from the Dresden University of Technology, Germany, in 2015, and the Ph.D. degree from the KTH Royal Institute of Technology, Stockholm, Sweden, in 2021. He conducted research on dc circuit breakers at the KTH Royal Institute of Technology. His research interests include power electronics, switching devices, and electrical machines. From 1994 to 2011, he was a Development Engineer with ABB, Västerås, Sweden, in various power-electronics-related areas such as railway traction systems and converters for HVdc power transmission systems. He is currently an Associate Professor in power electronics with KTH. He is the Inventor or Co-Inventor of more than ten granted patents and has authored or coauthored more than 100 scientific papers published at international conferences or journals. His research interests include power electronics and its applications in power grids, renewables, and electric vehicles.
HANS-PETER NEE (Fellow, IEEE) was born in Västerås, Sweden, in 1963. He received the M.Sc., Licenciate, and Ph.D. degrees in electrical engineering from the KTH Royal Institute of Technology, Stockholm, Sweden, in 1987, 1992, and 1996, respectively. Since 1999, he has been a Professor of power electronics with the Department of Electrical Engineering, KTH Royal Institute of Technology. His research interests include power electronic converters, semiconductor components, and control aspects of utility applications, such as FACTS and high-voltage direct-current transmission, and variable-speed drives. Dr. Nee was a member of the Board of the IEEE Sweden Section for many years and was the Chair of the Board from 2002 to 2003. He is also a member of the European Power Electronics and Drives Association and is involved with its Executive Council and International Steering Committee. VOLUME 10, 2022