Strategies Using Time-Domain Measurements for Radiated Emissions Testing in Harsh Environments

Performing in-situ radiated emissions measurements, that is, in locations different from a standard test site, can be a challenging task because of the high electromagnetic noise levels in the ambient. A harsh electromagnetic environment characterizes such sites, and it usually results in difficulties when discerning between emissions from the equipment under test (EUT) and electromagnetic fields generated by surrounding devices. Moreover, communication signals from broadcasting services are generally significantly higher than the standard emission limits, making it even harder to determine compliance. In this article, we present different techniques leveraging the advantages of time-domain measurement systems to provide effective and practical solutions to mitigate ambient noise’s effect on radiated electromagnetic interference measurements. First, the test method used is described, and pragmatic considerations are given to ensure reliable and repeatable measurements. Multichannel time-domain measurement systems are introduced as the fundamental tool for the proposed strategies. Subsequently, different study cases are evaluated with real test examples, highlighting several criteria intended to reduce the impact of ambient noise on the actual emissions measures produced by the EUT. Finally, a real application of those strategies for measuring a photovoltaic system is described. Overall, the methods employed and the main advantages of using full-time-domain FFT-based receivers are reviewed. In addition, the possibility of incorporating this article’s outcomes into forthcoming electromagnetic standards about in-situ radiated emission measurements is also debated.

be tested in an EMC laboratory facility is photovoltaic (PV) systems [3].
As the installed solar PV power generation capacity grows for the sake of more sustainable energy generation, the deployment of PV systems has brought challenges from the EMC standpoint. This is because the elements of the PV systems, mainly the optimizer and the inverter, have switching power electronics that may generate significant levels of electromagnetic noise. When such electromagnetic noise triggers the malfunctioning of other electronic equipment or degrades radio communications, it is called an electromagnetic interference (EMI).
As mentioned, such large and high power systems must be EMC tested on-site, that is, installed in their definitive location. However, experience indicates this is often not correctly performed, leading to actual EMI problems and noncompliances with the EMC Directive [4]. In some instances, this has resulted in the halt of sales, market recalls, or product banning due to electromagnetic problems [5]. Such a situation is directly linked to a lack of specific measurement methods and test procedures that allow checking the compliance with harmonized EMC standards while addressing the described scenarios in a reasonable way. Due to the severity of the problem, the Norwegian National Committee of the International Special Committee on Radio Interference (CISPR) requested to start working on a technical report containing information and guidance to be used by the stakeholders and their respective installers for the assessment of the EMC conditions before and after the installation of PV systems [6].
In this regard, the most influential inconvenience during in-situ radiated emissions assessments is the effect of uncontrollable and unpredictable ambient noise conditions. Therefore, detecting disturbances produced by the equipment under test (EUT) can be complicated in harsh electromagnetic environments. To reach a meaningful test conclusion, it is crucial to distinguish between both ambient field levels and the EUT emissions. Of course, this is even more cumbersome if we consider a large number of test points (different antenna polarization, multiple measurement axis and distances, etc.), conditions and operating modes that must be included in the test plan. All this implies lengthy and costly test campaigns.
To answer the current need for practical and reliable radiated emissions measurements in the in-situ context, in this work, we present different strategies. These strategies take advantage of time-domain measurement systems to provide conclusive EMI assessments even in harsh electromagnetic environments.
This article is structured as follows: Section II describes the main challenges of in-situ radiated emissions measures as well as how to determine whether or not an EUT is compliant. Section III presents an introduction to the range of possibilities offered by EMI measurement systems in the time domain and the methodology, measurement procedure, and practical considerations when performing in-situ campaigns. Subsequently, Section IV discusses five strategies to enhance the distinction between ambient noise and EUT emissions. Then, Section V covers a real application for the in-situ radiated emissions measurements of a PV inverter. Finally, Section VI concludes with a brief review of the lessons learned and how they could be applied to future standards for in-situ measurements.

II. RADIATED EMISSIONS TESTING IN HARSH ENVIRONMENTS
The CISPR 11 [7] standard has been widely applied for in-situ EMC assessments of industrial, scientific, and medical equipment, as it includes some clauses about this topic. However, this standard only provides limits and minimum number of antenna positions. CISPR 11 refers to CISPR 16-2-3 [8], which provides measurement requirements and, on its Annex A, limited guidance to deal with the high levels of ambient noise. Nonetheless, no former EMC standard is sufficient to eliminate the impact of ambient noise from in-situ measurements.
To address the problem of radiated emissions measurement in harsh electromagnetic environment conditions, let us start by formally describing the problem considering the hypothetical ambient noise spectrum profile as shown in Fig. 1.
In such a hypothetical scenario, several distinct situations can be observed depending on the relationship between EUT emissions, E(f i ), and ambient noise, N(f i ), at each i-th frequency, f i , with respect to the corresponding limits, L(f i ), with a certain margin, M = 6 dB. In summary, we can reduce the problem according to the flowchart in Fig. 2.
This decision tree should be applied for each frequency in the measured range. Once non compliance has been detected, this is sufficient to consider that the EUT has failed. In cases when EUT does not fail at f i , the next frequency, f i+1 , shall be evaluated. If the compliance is uncertain at some frequency, specific methodologies or strategies to determine the conformity must be employed. For example, previous research has tackled this problem through ambient noise cancelation methods [9], [10], [11], [12]. The cited studies involve complex mathematics and processing algorithms, which are not feasible for a standard test method because explaining them in generalized and plain terms is impossible. In Section IV, we present some techniques for assessing radiated emission measurements when compliance is uncertain due to high ambient noise. These techniques include prescanning, which has previously been used in EMI assessment in laboratories, and other new signal processing techniques that are only possible due to employing time-domain analysis.

A. MEASUREMENT SYSTEM
Time-domain EMI receiver implementations have been developed for measuring and analyzing electromagnetic emissions from complex systems in [13] and [14]. However, most research on in-situ measurements is based on frequency swept receivers, specially when dealing with the radiated emissions up to 1 GHz [9], [15]. The reason for this is that the superheterodyne architecture of such receivers offers better sensitivity and oscilloscopes capable of measuring radiated emissions appropriately are relatively novel. Commonly, these complex systems have behavior and modes of operation that changes over time [16], [17]. Due to this, the emission signature of the system being tested can vary during measurement, making the frequency-swept measurement method unreliable. Moreover, when assessing emissions of atypical equipment in-situ, the signals and the radio-frequency noise in the environment add further uncertainty to measurements, making it even more relevant to consider the time-domain information as part of the emissions assessment [18]. Because of these reasons, it can be concluded that time-domain EMI measurement systems are well suited for in-situ emissions assessment. Advanced EMI receivers with multidomain and multichannel capabilities can be used (Fig. 3) to cope with the above-mentioned challenges. In this context, multidomain means EMI measurements are processed and analyzed in time, frequency, and statistical domains, either independently or simultaneously. This allows for accurate spectral estimations according to standard specification [19], [20], [21] and complementary analysis, including waveform measurements, spectrograms, waterfall/persistence plots, and probabilistic information about the interference, e.g., the amplitude probability distribution function. Moreover, multichannel means the instrument has several inputs. Therefore, it can perform the acquisitions synchronously, given a triggering event. Those analysis capabilities beyond the standard emission requirements have proven highly valuable for interference evaluation, for instance, in wireless systems [22], [23].
For the measurement performed in the following sections, we used a multichannel time-domain EMI measurement system compatible with flexible resolution deep memory oscilloscopes to capture and process EMI waveforms. Welch's method is used for spectral estimation, and adaptive windowing functions are used to set any required resolution bandwidth. A 4-channel Digital Real-Time Oscilloscope Tektronix 5104B was employed for E-field measurements. It has a maximum sampling rate of 10 GSa/s and a bandwidth of 1 GHz with frequency response correction. Employing this instrument, CISPR 16-1-1 baseline requirements are met for CISPR bands A and B [14], [19].

B. EXPERIMENTAL SETUP
The proof of concept experiments for the proposed measurement strategies were performed in the presence of uncontrolled electromagnetic ambient noise. For experimental validation and control, some tests were also conducted in a fully anechoic room (FAR) with a 3-m measurement distance, d meas . This is convenient for the experiments since it allows for higher disturbance levels, usually above the instrument noise floor. However, when evaluating emissions with respect to standard limits, the measurement distance employed should be coherent with the limit specified at  a reference distance, d ref .
When d meas = d ref , then, distance correction factors must be applied to the measurements before comparing them with the required limits [8].
For the sake of simplicity of the test cases presented, the radiated emissions were measured in terms of either magnetic or electric field using a loop (9 kHz-30 MHz), biconical (30-200 MHz) and a log-periodic (200 MHz-1 GHz) antenna depending on the frequency range. Given the high ambient field levels, for E-field measurements, it is preferable to use two separate antennas instead of a single, ultrawideband, bilog (hybrid) antenna. This prevents unwanted saturation effects or the need for additional attenuation at the receiver input. The antenna height is 2 m. Complementary, we used a 30-dB low noise preamplifier for better sensitivity.
To evaluate the emissions of the EUT, several measurements are executed with the equipment "ON" and with the equipment switched "OFF" for subsequent comparison. The purpose of taking several measurements for the same condition is to decide whether the EUT produces any detectable disturbancethat is not merely a sporadic or random event.

IV. MEASUREMENT STRATEGIES A. PRESCAN AT SHORTER DISTANCES
The first case consists of performing a prescan at uncertain frequencies at a distance nearer to the equipment than the specified distance of the standard. By measuring at a shorter distance, we can determine that emissions are being produced by the EUT if the contribution at a certain frequency increases. Although we might be measuring in near-field conditions and we cannot compare directly with the limits of the standard, with this test we can determine if the EUT is producing disturbances at those frequencies [24].
In Fig. 4, we present a real test of an in-situ measurement campaign for a customer who develops industrial equipment. The test was conducted on a warehouse with only the EUT active. No other machinery or electronic equipment were active when conducting the measurements. It was possible to ensure a measure distance of up to 30 m.  When the device is powered on and measuring at the standard distance of 30 m, no noticeable differences are observed between the ambient measurement and the measurement with the active EUT. Same measure was repeated at a distance of 10 m with the same outcomes (Fig. 5).
However, if we approach to a distance of 3 m, we can see a significant difference between the measurement with the active EUT and the ambient measurement, specifically in the bands of 50-60 MHz and 200-300 MHz (Fig. 6).
It is also worth mentioning that the ambient noise measured with the EUT turned off is practically identical in both figures. In summary, we can say that the equipment is radiating at the frequencies mentioned above. However, as in the standard distance, these frequencies do not exceed the specified limit, there is no problem. In the other frequencies, no contribution from the equipment is detected at a distance of 3 m, so we can conclude that all other disturbances exceeding the limit are produced by external sources. While these measurements could have been made with a traditional frequency-domain receiver, by using time-domain measurements, we can compensate the time needed for the increased number of points to be measured due to the prescan. In this way, we keep the campaigns efficient in terms of duration.

B. MULTI-RBW ANALYSIS
In this experiment (Fig. 7), the analysis with multiple resolution bandwidth (RBW) is described. By reducing the RBW, we can differentiate narrow bands produced by the EUT that remain overlapped when the normative RBW is applied. The experiment carried out consists of an antenna transmitting a tone at 100.2 MHz. The antenna is placed inside an anechoic chamber of a standard EMC laboratory. A receiving antenna of similar characteristics is placed and the door of the chamber is left open in order to have an uncontrollable ambient noise. Note that the generated tone is masked by the FM radio broadcasting services.
A measurement is made with the tone active, another with the tone off, and finally one with the tone active but with the door closed in order to eliminate the ambient noise (Fig. 8). The spectrum is computed using a RBW 1 = 120 kHz, the bandwidth established in the CISPR16-1-1 standard. At first glance, no difference is observed between the case with the equipment active and the equipment off.
If we recalculate the spectrum using a smaller RBW, in this case RBW 2 = 5 kHz, a clear difference can be appreciated between a radio station centered at 100 MHz and the tone at 100.2 MHz generated by our "EUT" (Fig. 9). In contrast, using the normative RBW, we are unable to discriminate between the two interferences (black-dashed line). Furthermore, it should be emphasized that, when recalculating the spectrum with the smaller RBW, the measure with ambient noise and the measure with the door closed match perfectly.
This experiment is feasible thanks to keeping the data in the time domain. This allows to compute the spectrum for the same capture varying any desired parameter, including the resolution bandwidth. Although a measurement at a lower RBW could be performed with a traditional receiver, it would be too time consuming. Furthermore, the verification measure at a lower RBW and the measure at a standard RBW would correspond to two different realizations, whereby the environmental noises or other conditions could have changed.

C. TIME GATING
The third experiment consists of eliminating transients in the ambient noise using time gating of impulsive components. Time gating consists of applying a rectangular windows via software to the useful time-domain data and then recomputing the spectrum. It is useful for eliminating high-spikes in the measurement waveforms due to sporadic events not correlated with the EUT operation. Such events may not be present in the measures with the EUT off but may be present in the measures with the active EUT; hence, we are interested in removing them. To support this approach, two different experiments are carried out.

1) EXPERIMENT A
The setup of this experiment is shown in Fig. 10. It is performed in an anechoic chamber with no external ambient noise, except that an electrical fast transient generator is placed inside, which triggers every 1 s. Due to the high rise/fall of the transients, multiple wide bands with high levels appears on the spectrum, overlapping all the useful information from the EUT.
The acquired waveform in the time domain is depicted in Fig. 11. Notice that a transient appears at t = 310 µs. In our particular case, we are interested in excluding this section of the data, so the spectrum is recomputed by applying a window between t = 500 µs and t = 1 ms, in which no transient event occurs. Fig. 12 shows the spectrum calculated by taking all the data and the spectrum computed with the gated time-domain window. In addition, a measure was performed with no transient generator in order to get a purely real measure of the EUT. Two clear conclusions can be drawn from the figure presented. The first one is that the calculated spectrum using the gated time window and the spectrum of the measurement only with the EUT match perfectly. The second one is that the fact of a transient appearing during the dwell time results in a nonclean spectrum with a high level which can easily exceed the limit and mislead to wrong conclusions if we cannot distinguish it from the EUT emissions.

2) EXPERIMENT B
In the second experiment, we evaluate a more realistic case. As EUT, we consider a switched-mode power supply (SMPS) located in an open room (Fig. 13).
Nearby, there is an apparatus that randomly produces sporadic transients, which yield to a spectrum with a very high level. In this case, we observe when performing several measurements that the spectrum varies due to this external device. Using the spectrogram (Fig. 14), we are able to notice the appearance of different transients up to three different times (t 1 = 25 ms, t 1 = 26.5 ms, and t 1 = 93 ms).
The spectrum is recalculated with a time window from t = 35 ms to t = 85 ms. Three cases are shown in Fig. 15. First, the spectrum with the original time-domain data (green plot). In this chart, a very high level spectrum with little information can be visualized. Then, the spectrum with the time gated (orange graph), in which the emission produced by the SMPS is clearly visible and which practically matches with the laboratory measurement of the same EUT. Finally, an ambient measurement without the SMPS and at a moment when the external equipment was not producing transients (blue plot).
This technique is beneficial for environments where sporadic transients may occur frequently. By using time-domain measurement instruments, we can easily detect the appearance of these events, and we can suppress them. A positive point of such methods is the off-site post-processing, which can allow for more efficient execution of the measurement campaign and more extensive and detailed post-processing analysis.

D. TIME DECOMPOSITION
This experiment is based on the premise of prior knowledge of the operation of the EUT. In this case, the same SMPS as in the previous experiments is used, which produces impulsive components due to the internal switching of the power transistor. The power supply is placed in a noncontrollable environment (Fig. 16) and the measurements are performed.
We automatically retrieve the impulsive part from the time-domain data (Fig. 17) exclusively. The rest of the data is set to 0 V and then the spectrum is recomputed. Fig. 18 shows a zoom of one of the impulsive components produced.  The spectrum obtained is compared in both measurements in Fig. 19: spectrum computed with the original time-domain data and then with the purely impulsive part.
The clear contribution of SMPS is observed while the stationary part of ambient noise (mainly FM broadcasting) is canceled. This technique is specially useful to eliminate the stationary ambient noise in equipment with a pulse repetition frequency (PRF) lower than the RBW used.

E. AMPLITUDE PROBABILITY DISTRIBUTION ANALYSIS
In the last criterion, a transmitting antenna is placed inside an anechoic chamber with the door open (Fig. 20). In previous experiments, a 100.2 MHz tone was injected. However, this time a 100 MHz tone will be generated, which coincides exactly with the carrier frequency of an FM radio station. When the measurements are performed, the spectrum is identical either with the signal generator ON or OFF (only ambient noise). Even using criterion B (RBW reduction), it is not possible to discern the injected tone from the ambient noise. However, the field amplitude distribution represented as a probability density function (pdf) is useful for differentiating pure ambient noise levels from the situation where there are significant emissions that are completely overlapped in both time and frequency domains. In order to estimate the pdf, the spectrogram is computed for an acquisition time of t = 100 ms. The distribution of the different values of the spectrum at the study frequency, f c = 100 MHz, are shown, as well as the expected value of the measuring detectors. If the complete spectrum is identical in the ambient measure and in the EUT measure, but the amplitude distribution is different, that means that the EUT is producing some disturbance in that specified frequency. Fig. 21 shows the results from the ambient noise measure and the measure with the EUT active.
As expected, both distributions differs. This means that the EUT is producing emissions at f c = 100 MHz, which could not be observed directly in the spectrum.

V. APPLICATION
The described criteria have been applied in a real case of an in-situ measurement campaign of an inverter of a PV system (609 solar panels). This fixed installation is located in Barcelona, Spain, and it is part of the renewable energy generation infrastructure of Universitat Politècnica de Catalunya. The test site and the specific measurement area are displayed on Figs. 22 and 23, respectively.  The PV system inverter (EUT) has a rated maximum power of 230 kWp and its switching frequency is f sw = 16 kHz. The EUT is shown in Fig. 24. The reference measurement distance is 10 m over the 45 • axis. The electric field and magnetic field limits of the CISPR 37 draft [25], a new standard project currently in development for in-situ testing, have been applied as a reference. This standard draft is being developed by the CIS/B Working Group 7 (WG7).
In the vertical polarization of the E-field measurement (Fig. 25), the exceedance of the radiated emissions with respect to the limit (red line) cannot be reliably confirmed in certain frequency ranges. Most notably, the 32-34 MHz band and the 66-70 MHz. Both of these bands are notably higher than the ambient noise level.
It is decided to apply Criterion A (prescan) employing a distance of 5 m. When approaching the EUT, the 32-34 MHz band level decreases, so it can be concluded that this wideband is produced by an external source.
Nonetheless, in the 66-70 MHz band, the level increases when approaching to the EUT, so it is highly probable that this specific wide band that exceeds the limit is caused by it. This behavior is also observed in the other measurement axis.
Another critical interval in which it is difficult to discern EUT emissions from ambient noise is the FM broadcasting band (80-108 MHz). After conducting the measurements, an inspection is carried out by comparing EUT and ambient levels with a reduced resolution bandwidth (Criterion B: multi-RBW analysis). It is reduced from the standard RBW 1 = 120 kHz to RBW 2 = 12 kHz. We focus on the points where, after reducing the RBW, new components appear that were not present in the ambient noise. For example, in Fig. 26, we can detect a narrow-band component centered at a multiple of f sw . This component is 25 dB higher than the ambient measurement level. Based on these findings, we can consider that the increase of field magnitude is caused by the EUT.
The last example is about the magnetic field emissions (Fig. 27). It is seen that the highest peak corresponds to f sw = 16 kHz. In this case, it has been decided to study the peak appearing at 576 kHz, which is a multiple of the f sw and matches an AM broadcasting station. In both ambient and EUT active measurements, the peak is above the limit. VOLUME 2, 2023 2000211 The analysis will be carried out by studying the amplitude probability distribution (Criterion D) at this frequency. Results are depicted in Fig. 28. In this case, we can see that the shape of both distributions has a strong similitude but is centered on a different magnetic field level. From these results, it can be stated that the magnetic field level difference is produced by a change in the broadcasting service and is not caused by the EUT.

VI. CONCLUSION
Different approaches have been proposed in order to distinguish the emissions of electronic equipment with respect to high electromagnetic ambient noise levels. Such criteria are possible and feasible through the use of time-domain-based instruments.
The first criterion consists of the prescan at shorter distances, which, thanks to time-domain measurements, allows to perform fast acquisitions at a large number of points while maintaining time-efficient measurement campaigns. The second one is the multiresolution bandwidth analysis, which allows unmasking overlapped EUT and ambient noise frequency components at the CISPR RBW. Moreover, the measurement speed and efficiency characterizing FFTbased receivers allow real-time analysis independently of the selected RBW, therefore, they are ideal for analyzing emissions in dynamic and time-sensitive environments. Another key advantage, from now on, is the possibility to perform off-site reprocessing, maintaining the efficiency of the measurements during the campaign and allowing the detailed analysis of each case separately. The third is time-gating to remove sporadic and transient events from the final results. And the fourth is time decomposition to exclude stationary ambient noise between impulsive events generated by the EUT. Obviously, these two criteria are only feasible if the spectrum can be recalculated once the disturbances have been identified in a spectrogram or plot of the acquisition performed. Finally, we discussed the amplitude probability distribution analysis, which allows us to unveil the nonevident behavior of EUT emissions even with very high ambient noise levels.
The testing strategies presented and discussed in this article could be seen as a toolbox of handy methods for facing the complexities of in-situ radiated emissions measurement. In our experience, they are effective for improving measurement campaigns and obtaining better, more reliable, results. In consequence, they might be considered for the forthcoming CISPR 37 standard as part of an informative annex.
To sum up, it is worth noting that the successful implementation of these strategies is reliant on the utilization of a time-domain instrument. This instrument not only reduces costs but also enables fast measurements, the viability of off-site processing, and the acquisition of extraordinary information that would be otherwise unattainable with traditional frequency-swept instruments. The integration of these advantages enhances the practicality and significance of the strategies presented in addressing the challenges of in-situ radiated emissions measurement.