Noise suppression and elimination techniques in pre-testing

With the rapid development of electronic technology, people are enjoying more and more convenience brought by scientific and technological progress, but the development of science and technology has also caused some problems, and it is precisely these problems that may bring a lot of trouble, such as environmental noise. The noise here refers to the electromagnetic noise existing in the environment or background. From the perspective of electromagnetic field, it is no exaggeration to say that the world we live in now is full of various noises. If you turn on the RF spectrum analyzer, you will find that this is not alarmist. Artificial noise accounts for a large proportion. Because of its existence, it will directly affect the measurement. If it is in an environment without any shielding, it will even completely submerge the useful signal and cause the measurement to fail. Therefore, we must pay special attention to the problem of environmental noise.
In fact, the debate on how to eliminate and avoid the influence of environmental (background) noise on the test has been going on for a long time, but so far, both sides of the debate have no conclusive evidence to fully explain its correctness. This article will use some simple experiments to illustrate how to solve this problem that has troubled people for a long time, and it is very clear and thorough. Noise elimination technology is mostly used in the following two situations, namely verification and measurement. The verification technology mentioned here refers to the positioning of the radiation of the object under test in an environment with high noise levels. It is well known that if some environmental noise suppression technology is not adopted, the positioning of the radiation of the object under test is very inaccurate. The noise suppression in measurement technology is higher than the requirements in verification, because the purpose of measurement is to obtain the accurate radiation level of the object under test. This article will focus on noise suppression in measurement technology and explain various strict requirements for noise suppression in order to achieve the purpose of reasonable noise suppression.
The instruments and equipment provided by EMC test equipment manufacturers generally use two background noise suppression and elimination technologies. The first method is to measure the background noise first, then turn on the object under test (EUT) and perform a second measurement, and then use the second measurement result to subtract the first measurement result in the software to eliminate the background noise (i.e., differential technology); the second method is to use a dual-channel analyzer, which can adopt two schemes. One is to use a near-field probe for measurement, and use the test data of the far-field antenna to correct the data of the near-field probe, and assume that any meaningful radiation can be detected in the near-field area of ​​the EUT, and the near-field probe is insensitive to environmental noise. The second is to use two far-field antennas at the same time, and place one of them at a meaningful distance from the test unit, and finally use differential technology to extract the radiation of the EUT.
The above two methods can be said to have their own advantages, but neither is perfect! For example, method 1 will encounter great troubles because the background noise is surging (in fact, the noise is surging in many frequency bands), but the potential advantage of this method is that any standard EMC receiver or analyzer can be used; in contrast, both solutions in method 2 require the use of special dual-channel EMC analyzers, and there is an obvious problem in solution 1 that it assumes that the near-field probe is insensitive to background noise. In fact, very strong environmental noise will be induced through the cable connected to the EUT, and the effect is as if a near-field signal has generated radiation. The most important thing is that both method 1 and the second solution in method 2 use differential technology. Whether the application is reasonable is the key to these two methods and is also the subject of this article.
The credibility or effectiveness of using multiple differential methods to calculate the radiation level of the EUT in a noisy environment is based on the assumption that the field strength will increase when two or more radiation sources are present at the same time. This seems obvious and has been demonstrated in open field (OATS) testing. It is well known that in a standard test environment, the addition of a directly incident signal and a signal reflected from the ground plane will increase the field strength by about 6 dB, which means a doubling in linear coordinates. Of course, the intensity of the radiating sources is basically the same at this time, and these radiating sources are strictly related and should be in phase. If this phase relationship is changed (for example, changing the height of the antenna), the effect will be completely different. Therefore, the phase issue is very important. Specifically, what is the impact on the measurement of the EUT's radiation and environmental noise signals? The following experiments will study this issue.
Experiment:
In this experiment, we create a known background noise and a known EUT signal to study the impact of the simultaneous presence of both on the field strength measurement using an EMC analyzer. When creating the signal, the signal type is taken into account, including narrowband and/or broadband signals. These two types of signals are very common, both in the EUT and in the environment. We should not assume that any cancellation technique will achieve good (or poor) processing results for the combination of these two signals.
Table 1 lists the possible combination modes and their corresponding characteristics

Possible combination patterns	Narrowband signal in background noise	Broadband signal in background noise
Narrowband Signals in the EUT	If the frequency interval is smaller than the resolution bandwidth of the receiver, the peaks will merge (peak envelope), and the phase problem of the signal will increase the complexity, which is relatively complicated.
Broadband Signals in the EUT	By definition, broadband radiation has relatively flat characteristics, so broadband radiation can be clearly observed in the presence of narrowband noise.	This is also a more complicated situation.

It is not difficult to see that the most severe challenge faced by any cancellation technology is the situation where the signal of the EUT and the environmental noise signal are of the same type. Therefore, our experiment will study these two situations.
Narrowband experiment
In this experiment, two Laplace Instrumen Ltd radiation reference sources (ERS) will be used, one of which will simulate the background (environmental) field and the other will simulate the EUT. Laplace Instrumen Ltd radiation reference sources (ERS) are very versatile instruments, especially narrowband radiation sources with continuous (time domain) radiation output. They can generate radiation signals with a spacing of 2MHz, and the frequency consistency when using two instruments is very good, reaching 40ppm, which is guaranteed to be within the resolution bandwidth of the analyzer, so the first complex situation mentioned above is almost completely reproduced. The good frequency consistency eliminates the possibility of separating the signals by frequency resolution techniques, and it is impossible to use averaging techniques. The actual radiation levels generated by the two ERS are very similar, but they are placed at different locations in the test unit (laboratory), so the signals received by the measurement antenna are different. This is because the attenuation at different locations in the test unit is different. The frequency range we use is 350-450MHz. We choose this frequency band because the background signal is relatively clean in this range. First, each ERS is measured separately. The one with the lower value (through the measurement antenna) is used to represent the EUT, which we call device A. The other ERS (device B) is used to represent the environment. Then, the following series of experiments are used to simulate the real EMC test.
To calculate the noise signal that causes the field strength to increase from XdBuV/m to YdBuV/m, we cannot simply subtract the test data to get the difference. Directly subtracting two logarithmic values ​​is equivalent to division, so the logarithmic value should first be transformed into a linear value, that is, X(lin) = 10(X(log)/20), and then calculate the difference Z(lin) = X(lin) –Y(lin), and finally transform it into a logarithmic value, ZdBuV/m = 20log(Z(lin)). From the comparison between the evaluation value and the measured value, we can see that the consistency is very good. For example, at 380MHz, the difference between the evaluation value obtained by the differential method and the measured value is within 4dB, and the radiation generated by the EUT at this time is 15dB lower than the ambient noise, which shows that our method is very effective. In the above experiment we used two independent sources, so the signals are not correlated, that is, they are not phase locked. This is also true in many EMC applications when background noise signals are to be eliminated. If there is a phase correlation between the EUT and each background noise source, this means that the resultant field will surge at certain frequencies. That is, a 253,456,789Hz noise signal mixed with a 253,456,800Hz EUT signal will have a resultant or surge frequency of 211Hz. If a quasi-peak detector with a time constant of 550ms is used, these surges will obviously not affect the resultant field. In fact, it is only possible to distinguish any phase effect if the two frequencies are within a few Hz of each other, but this is extremely unreliable, especially for signals that exceed a certain period of time. The only thing that will affect the results is unstable environmental noise, but there is a way to stabilize this noise, that is, use the average or peak hold technique when measuring the noise line and testing the EUT and the noise line at the same time. Generally speaking, no matter which technique is used, as long as the number of sweeps reaches more than 8, stable results can be obtained. The differential result is relatively clean and only reflects the radiation of the EUT. This is very useful for noise signals with a certain time pattern. These signals appear to be the radiation of the EUT, but can be excluded by the above.
Broadband experiment
In this experiment, the corresponding steps in the narrowband experiment will be repeated, but two broadband radiation sources, namely York Electromagnetics' CNE (control noise emitter), are used. They can produce a relatively flat output spectrum and are pulsating noise sources with a wide bandwidth. Each radiation noise source is also measured separately, and then one of the devices is turned on first, and then both devices are turned on at the same time and tested separately. The differential trace contains a series of interval drifts that reach negative infinity, which looks like the logarithm of zero! The reason for this phenomenon is that at some frequencies, the ambient noise is almost the same as the ambient noise plus EUT, or even smaller.
The last figure shows the comparison between the actual EUT radiation level obtained by the narrowband experiment (orange trace) and the difference trace obtained by calculation. From the results, it can be found that a rough assessment of the EUT radiation level cannot be obtained using the method in the narrowband experiment. Obviously, the noise signal and the EUT signal are not related to each other as in the narrowband source. In order to try to improve it, the voltage value is used instead of the power value when performing the subtraction operation. The pink curve shows that the result is improved, but it is still far from the actual situation.

Using quasi-peak and average detectors also does not provide significant improvement. To try to figure out why the real signal and the processed signal are so different, we will actually test the EUT signal in the time domain using the receiving antenna.

It is obvious that these sources have strong pulsation characteristics. Through Fourier analysis, we found that in the frequency domain, there is a very flat spectrum, so in the time domain, there must be transient characteristics, so the signal is actually completely random, just like the noise source we imagine.

It can be seen that the pulsation frequency has doubled, but the peak amplitude has not been affected. Therefore, using a peak detector will also maintain the original peak for such a strong source, that is, it will not be affected by any pulsation peak with a lower level. This assumes that the spectrum bandwidth of the two sources overlaps. If this is not the case, it is easy to identify the radiation of the EUT. The signal level in each waveform is calculated by adding the absolute value of all DSO sampling data in each frame.
Considering only the ambient noise, it is 0.3696V, source
one is 3.7744V
, source two is 3.6467V,
and both sources are 5.1472V
. This shows that the signal in the time domain has indeed been expected to increase. Initially, we may think that using average or quasi-peak values ​​should improve the performance of differential technology. In fact, this is not the case. If we carefully examine the definition of average and detector in CISPR16, we will know why. The output of these detectors is strictly related to the repetition frequency of the input pulse signal. The time constant of the detector corresponds to a repetition frequency above 10KHz. At this time, the output of the detector is relative to a peak detector. In other words, once the repetition rate exceeds 10KHz, it will not have any effect. The study of the waveforms shows that the meaningful pulsation signals occur at medium intervals, about 300 nanoseconds, which is equivalent to a repetition rate of 3.3MHz, which is much higher than 10KHz. Obviously, the above analysis shows the authenticity of the noise source we used. Other noise sources with different characteristics may not be the case. For example, some noise sources in real situations are generated by main frequency switching devices (such as phase angle controllers). They have pulsation signals with a repetition rate of about 100 or 120Hz. Introducing a second source will double this repetition rate, which will lead to an increase of 3dB in the QP level and an increase of about 6dB in the average level. In theory, although differential techniques can also be used, coefficients such as relative time (relative phase angle) and duty cycle will affect the final result.
In summary, our experiment shows the real world situation, that is, the background noise and EUT radiation are broadband and have overlapping spectra, and the characteristics of the noise source are unknown (this is an absolutely real background situation). If possible, no differential techniques should be applied. In practice, however, it can be successfully used to detect the common broadband emissions of the EUT, even in the presence of high background noise emissions, by using a modified differential technique that involves an average sweep with a peak detector.
* With the EUT off, the analyzer is used to perform the test and the average level at each frequency point is calculated by performing multiple sweeps.
* When the result is obtained, the sweep is stopped and the obtained data is used as the background.
* After the EUT is turned on, the average sweep process is repeated until a result is obtained.
* The difference between the two results is taken.
Although this method is not recommended for accurate measurements, it can still be used in pre-tests to evaluate the emissions of the EUT.
Summary:
Even if the noise and EUT radiation sources are of different signal types (narrowband and broadband), it is possible to measure the EUT by using different measures. Differential techniques can be used in the narrowband/narrowband case to stabilize the noise. Even for unstable noise signals, some additional techniques have been proven to be very effective. When both the ambient noise and the EUT radiation are broadband, the measurement becomes unreliable, and the differential technique will be inaccurate even if it provides an approximate level of radiation from the EUT. In any case, the characteristics of the radiation source used in the experiment may not be typical of the real world. However, it has been demonstrated experimentally that the differential technique can provide useful guidance for predicting the radiation of the EUT even in very bad conditions. The bad conditions that Leary refers to are those that often cause broadband radiation with very low repetition frequency noise.