Franalyzer Anp 10019 Enu

OMICRON Page 1 of 23

Application Note

Understanding Noise Suppression Methods

Author Juan Velasquez | [email protected]

Date 2010

Related OMICRON Product FRAnalyzer

Application Area General

Version v2.0

Document ID ANP_10019_ENU

OMICRON 2010 Page 2 of 23

Content

1 Introduction ..................................................................................................................................... 3

1.1 Conventions and Symbols Used ............................................................................................... 3 1.2 Related Documents .................................................................................................................. 3

2 Sources of Noise in the FRA Measurements ................................................................................. 4

3 Effects of Noise in the FRA Results ............................................................................................... 5 3.1 Effect of the Power Frequency Noise in the FRA Results .......................................................... 5 3.2 Effects of the Noise Floor in the FRA Measurements................................................................. 6

4 Noise Suppression Methods in the FRAnalyzer ............................................................................ 8 4.1 Hardware Based Methods ......................................................................................................... 8

4.1.1 Input Attenuators .........................................................................................................................8 4.1.2 Input Filters..................................................................................................................................9 4.1.3 FRAnalyzer Sweep Profiles ..........................................................................................................9

4.2 Software Based Methods .........................................................................................................16 4.2.1 FRAnalyzer Signal Processing ................................................................................................... 16

4.3 Connection Technique .............................................................................................................20

5 Literature ........................................................................................................................................21

6 Contact Technical Support ............................................................................................................22

Please use this note only in combination with the related product manual which contains several important safety instructions. The user is responsible for every application that makes use of an OMICRON product. OMICRON electronics GmbH including all international branch offices is henceforth referred to as OMICRON. OMICRON 2010. All rights reserved. This application note is a publication of OMICRON.

All rights including translation reserved. Reproduction of any kind, for example, photocopying, microfilming, optical character recognition and/or storage in electronic data processing systems, requires the explicit consent of OMICRON. Reprinting, wholly or in part, is not permitted.

The product information, specifications, and technical data embodied in this application note represent the technical status at the time of writing and are subject to change without prior notice.

We have done our best to ensure that the information given in this application note is useful, accurate and entirely reliable. However, OMICRON does not assume responsibility for any inaccuracies which may be present. OMICRON translates this application note from the source language English into a number of other languages. Any translation of this document is done for local requirements, and in the event of a dispute between the English and a non-English version, the English version of this note shall govern.


1 Introduction

Noise can be defined as unwanted disturbances that may be superimposed upon a useful (desired) signal. Noise tends to obscure the information content of the useful signal and, consequently, its detection and mitigation is necessary. As in any other electrical diagnostic method, the frequency response analysis (FRA) results can also be affected by noise. This Application Note describes the noise suppression methods applied in the FRAnalyzer. Before examining the suppression methods in detail, an explanation of the sources of noise as well as of its detrimental effects is given

1.1 Conventions and Symbols Used

In this document, the following symbols indicate paragraphs with special safety:

1.2 Related Documents


2 Sources of Noise in the FRA Measurements

In a substation environment the noise can be found basically in two forms, as a broadband or as narrowband noise. A typical narrowband noise is the power frequency noise (50 or 60 Hz noise). This kind of noise is called narrowband because its effect can be seen in the FRA plots normally in the frequency range between 30 Hz and 100 Hz. At frequencies higher than 300 Hz is it very unusual to find narrowband noise. In substations with high harmonic pollution, some narrowband noise at frequency multiples of the power frequency could also be present. Another possible source of narrowband noise could be some communication signals in the substation, or noise generated by corona discharges, but such sources occur at higher frequencies and can be found very rarely in the FRA plots. With respect to the broadband noise, there will be always a noise floor that will affect the FRA plots. The presence of this noise is very closely related to the dynamic range of the FRA instrument. In this Application Note, the attention is focused on the two most important sources of noise, that is, the power frequency noise and the noise floor. In this sense, the noise can be characterized in the FRA plots in two regions as shown in Figure 1.

Figure 1: Noise sources in a typical FRA


3 Effects of Noise in the FRA Results

3.1 Effect of the Power Frequency Noise in the FRA Results

Power frequency noise is only present when measurements on-site in substations with high electromagnetic fields are carried out (typically in substations with live bus-bars with rated voltages above 380 kV). The presence of this kind of noise also depends on the size of the windings, because the signal attenuation increases with the inductance making the response signal more sensitive to noise. The effects of the power frequency noise fortunately take place in the frequency range in which the linear behavior of the magnetizing inductance predominates the response. From the point of view of the result interpretation, this noise is not so harmful since the useful information for the diagnosis remains intact. The first resonance points which take normally place above 130 Hz are not affected by the noise which allows a reliable assessment even when some 50 Hz noise is present. Nevertheless it is worth to mention that some windings failure modes are visible at low frequencies, such as short-circuits between turns and openedcircuits, what makes the suppression of the power frequency noise very advisable. Noise at harmonic frequencies are not so usual but can also be present in the FRA plot. In substations with rated voltages higher than 400 kV the likelihood of finding this noise is higher. As an example, Figure 2 shows the FRA plots measured in the 500 kV windings of a power transformer. From 30 Hz to 100 Hz, the effects of the noise are significant while at harmonic frequencies (250 Hz and 350 Hz) the effects are slighter but still present.

Figure 2: Frequency response of a 500 kV winding


3.2 Effects of the Noise Floor in the FRA Measurements

In signal theory, the noise floor is the measure of the signal created from the sum of all the noise sources and unwanted signals within a measurement system. The noise floor of an FRA instrument is defined by its dynamic range. The dynamic range is normally defined as the absolute value of the negative measurement range plus 20 dB. This definition is graphically illustrated in Figure 3. According to the CIGR report 342 [2], a measurement range of -100 dB to +20 dB should be enough to cover all the cases. The only FRA standard available at present [1] also states that the FRA instrument shall have a dynamic range of -100 dB ~ 20 dB.

Figure 3: Illustration of the dynamic range concept

The dynamic range of one FRA instrument can easily be measured by connecting the instrument as shown in Figure 4. The dynamic range measured in the OMICRON FRAnalyzer is presented in Figure 5. As can be observed, the instrument has a dynamic range of at least 130 dB (-110 dB+20 dB).

Figure 4: Connection scheme for measuring the dynamic range of FRA instruments

Figure 5: Dynamic range of the OMCIRON FRAnalyzer


Note: The dynamic range test should be performed at fixed settings for the receiver bandwidth and attenuators. It is recommended to set the bandwidth to 10 Hz and the attenuator to 0 dB. This noise is normally present in the FRA response measured in transformers with high magnetizing inductance, in windings connected in delta and in capacitive inter-winding FRA measurements. Some examples of the effect of the noise floor in the frequency response in delta windings measured with FRA instruments with a limited dynamic range are illustrated in Figure 6 and Figure 7. In both cases it can be seen that the frequency response is highly affected by the noise floor below -80 dB. Figure 6 also shows that in contrast to the power frequency noise, the effect of the noise floor is very harmful. The presence of this noise makes the assessment of the results difficult since several resonance points which contain very important information for the diagnosis are obscured by the noise.

Figure 6: Example of the simultaneous effect of noise floor in FRA measurements in delta connected windings

Figure 7: Example of the effect of noise floor in FRA measurements in delta connected windings


4 Noise Suppression Methods in the FRAnalyzer

The FRAnalyzer uses different methods to suppress the effects of noise in the FRA results. These methods can be divided into three groups: hardware based methods, software based methods and connection technique.

4.1 Hardware Based Methods

Hardware based methods rely on the capability of the FRA instrument to minimize the presence of noise during the measurements. The two key factors defined in the technical specifications of the FRA instruments that are close related to their capabilities of suppressing noise are the output voltage of the instrument and the input filters. The FRAnalyzer suppresses noise by setting the bandwidth of the input filters and the input attenuators properly.

4.1.1 Input Attenuators

The input attenuators are used for controlling the loading conditions of the analogue to digital converters (ADC) of the instrument. The resolution of the ADCs is 24 bits. At full scale of the ADC using no input attenuator (0 dB), the input voltage of the ADCs is 100 mVRMS (100 mV range). Depending on the device under test (DUT), the attenuation of the DUT might take low values, potentially leading to voltages higher than 100 mV at the input of ADCs. The input voltage of 100 mVRMS corresponds to the ADCs load of 100 %. If the voltage at the input of the ADCs is higher than 100 mV, the ADC is overloaded. It this case, additional input attenuators prevent overloads of the ADCs. The noise performance of the FRAnalyzer can be optimized by setting the input attenuator according to the input signal level. For low input signals, no attenuator (0 dB) is used, so the signal level at the ADC will be maximum. The noise of the amplifiers and ADCs will have less influence, resulting in high signal to noise ratio. For high input signals, attenuators (10 dB, 20 dB, 30 dB) are necessary to avoid overloading of the ADCs. Since the signal level is high, there is no problem concerning noise. The loading condition of the ADCs can be monitored in real time during the execution of the measurements as shown in Figure 8.

Figure 8: Real time monitoring of the loading conditions of the ADCs

When the impedance of the device under test is higher than -30 dB, it is recommended to set the input attenuators to 0 dB to improve the voltage resolution of the ADCs and hence improve the signal to noise ratio. As described in section 4.1.3, the sweep profiles of the FRAnalyzer can perform dynamic setting of the attenuators automatically during a measurement.


4.1.2 Input Filters

Normally, at the inputs of the FRA instruments there are filters for suppression of noise. The noise suppression capabilities of such filters can be controlled by the bandwidth. The selection of the bandwidth is a compromise between the measurement time and noise. An optimal setting of the bandwidth can also be achieved by adapting the receiver bandwidth automatically as function of the attenuation of the signal. This concept is illustrated in Figure 9. If a sample at the frequency "if" is to be measured, the instrument will be able to measure any signal whose frequency is within the bandwidth BW. Any perturbation with frequency out of the BW range is not measured by the instrument. This shows how the noise suppression capabilities can be enhanced by reducing the receiver band width (RBW). Note that the FRAnalyzer takes only a single measurement point for each frequency. Under this filtering technology, it is not necessary to take several cycles of the signal for suppressing noise by means of averaging. It should also be highlighted that the reduction of the RBW increases the measurement time, since with decreasing RBW higher transient time is necessary to wait for measuring a point in steady state.

Figure 9: Illustration of the bandwidth concept

4.1.3 FRAnalyzer Sweep Profiles

The FRAnalyzer has different sweep profiles which can be chosen according to the noise suppression requirements. The available sweep profiles are a compromise between measurement time and noise suppression capabilities as shown in Figure 10.

Figure 10: FRAnalyzer sweep profiles


OMICRON (FRAnalyzer 2.1) By default, when a new test is created in the FRAnalyzer, the sweep profile OMICRON (FRAnalyzer 2.1) is defined. A screenshot of this profile is shown in Figure 11. Under this profile a total of 1000 points are logarithmically measured in four intervals. A detailed explanation about the RBW settings and attenuator follows in this section.

Figure 11: Screenshot of the sweep profile FRAnalyzer 2.1

As illustrated in Figure 12, under this profile the receiver bandwidth is automatically set to different values depending on the attenuation and frequency of the signal. Taking into account that the power frequency noise used to be present when the attenuation of the signal is higher than 40 dB in the range between 30 Hz and 100 Hz, in this range the RBW is reduced from 30 Hz to 10 Hz between 30 Hz and 45 Hz and from 10 Hz to 3 Hz between 45 Hz and 65 Hz. Then, from 65 Hz till 100 Hz the RBW is set again to 10 Hz. At frequencies higher than 100 Hz, the power frequency noise does not affect the noise. Only in exceptional cases where the measurements are performed in a substation with high harmonics pollution it could be possible to see the effects of power frequency noise. Due to this, the RBW is increased to 30 Hz in order to make the sweep faster. Statistically it can be demonstrated that at frequencies higher than 3 kHz the attenuation of the signal is not so high and, consequently, the signal is strong and therefore not so vulnerable to noise. Moreover, at these frequencies there are usually no perturbations. This sweep profiles takes advantage of this for speeding up the sweep by increasing the RBW. The black arrows depicted in Figure 12 indicate the logic used for the automatic setting of the RBW at frequencies higher than 3 kHz. Two cases are considered. The case A corresponds to increasing the attenuation of the signal (from top to bottom), while the case B corresponds to decreasing the attenuation (from bottom to top).


Figure 12: Automatic settings of the RBW under the OMCIRON (FRAnalyzer 2.1) sweep profile

Under this profile, in order to increase the dynamic range of the instrument and, consequently, to reduce the effects of the noise floor, the attenuator is set automatically. The rules for setting the attenuator are shown in Figure 13. Here again we consider two cases, the case of increasing attenuation (case A) and the case of decreasing attenuation (case B). As can be appreciated, for the case A, when the attenuation level of the signal is higher than -45 dB, the attenuator is set to 0 dB. Changing the attenuation to 0 dB has the benefit of reducing the noise generated by the amplifications and ADCs, thus enhancing the suppression of the floor noise.

Figure 13: Automatic settings of the attenuator under the OMICRON (FRAnalyzer 2.1) sweep profile

In summary, this sweep profile has several enhanced features with respect to his predecessor: better noise suppression of power frequency noise, higher number of points (1000), shorter measurement time and better stability. In view of these advantages, the OMICRON (FRAnalyzer 2.1) sweep profile is the default profile of the FRAnalyzer software. Notice that after upgrading the FRAnalyzer software from 2.0 to FRAnalyzer 2.1, the sweep profile FRAnalyzer 2.1 will be set as default.


OMICRON (FRAnalyzer 1.1) This profile provides a good compromise between the noise suppression and measurement time. In the most cases, this profile provides satisfactory noise suppression. However, in delta connected transformers or in very big transformers, this profile could result in noising FRA plots due to the limited dynamic range of the instrument. As presented in Figure 14, this profile uses a bandwidth of 30 Hz and an attenuation of 20 dB which is equivalent to a dynamic range of approximately -100 dB.

Figure 14: OMICRON (FRAnalyzer 1.1) sweep profile

OMICRON (FRAnalyzer 2.0) (adapt automatically) A screenshot of this profile is shown in Figure 15.

Figure 15: OMICRON (FRAnalyzer 2.0) sweep profile

The main characteristic of this profile is an enhanced distribution of the measurement points in different subbands as illustrated in Figure 16. Compared to the conventional logarithmic distribution of the points, with this profile you can record more points in the subbands containing the most important information for the diagnosis (see Figure 17).


Figure 16: Optimal distribution of the measurement points

Figure 17: Comparison of the measurement points recorded with the OMICRON (FRAnalyzer 2.0) sweep profile and the conventional logarithmic distribution

The sweep profile OMICRON (FRAnalyzer 2.0) adapts automatically the settings of both input attenuators and the bandwidth. The attenuators are set as in the profile FRAnalyzer 2.1, while the bandwidth is set as shown in Figure 18. In this case, the settings are adjusted independently of the frequency.

Figure 18: Automatic settings of the RBW under the OMICRON (FRAnalyzer 2.0) sweep profile


The logic of the algorithm is designed for two cases. The first case corresponds to increasing attenuation. Between 0 dB and -70 dB, the receiver has a constant bandwidth of 30 Hz but for attenuation higher than -70 dB, the bandwidth is reduced to 10 Hz, 3 Hz and even 1 Hz depending on the attenuation. The second case corresponds to a decreasing attenuation. For attenuation lower than -100 dB, the bandwidth changes from 1 Hz to 3 Hz, when the attenuation reaches -80 dB, the bandwidth changes to 10 Hz and so on. This algorithm relies basically on the settings of the bandwidth and input attenuators as function of the attenuation of the device under test. This algorithm reduces the broadband noise (noise floor) well but does not suppress the narrowband noise (power frequency noise) that sometimes takes place during on-site measurements between 30 Hz and 100 Hz. OMICRON (FRAnalyzer 2.0) adapt frequency dependently A screenshot of this profile is shown in Figure 19.

Figure 19: OMICRON (FRAnalyzer 2.0) adapt frequency dependently sweep profile

In exceptional cases when the power frequency noise is found even in the range between 0 dB and 35 dB, this profile can be used for suppressing the noise, since under this profile the RBW is set to 3 Hz in the range from 30 Hz to 100 Hz independently of the attenuation of the signal as illustrated in Figure 20. Additionally, this profile also makes use of RBW reduction to increase the dynamic range of the instrument. As a consequence, the measurement time is longer. The automatic settings of the attenuator of this profile are as in the profile FRAnalyzer 2.1.

Figure 20: Automatic settings of the RBW under the OMICRON (FRAnalyzer 2.0) adapt frequency dependently sweep profile


Caution: Rarely, during the measurement with the signal attenuation lower than 35 dB, a wrong point might be recorded under this profile as shown in Figure 21 because only very few points are taken between 20 Hz and 100 Hz(10 points). If you detect this problem, repeat the measurement.

Figure 21: Example of recording a wrong measurement point

In Figure 22 and Figure 23, a comparison of the performance of the sweep profiles in a transformer with very high attenuation (higher than -100 dB) is presented. In Figure 22 it can be observed that the FRA plot measured with the profile OMICRON (FRAnalyzer 1.1) has some noise below -80 dB and below -90 dB the plot is significantly distorted. In contrast, the plot measured with the profiles OMICRON (FRAnalyzer 2.0) is noise-free. On the other side, Figure 22 shows that both profiles OMICRON 2.0 and OMICRON 2.1 have an excellent performance with respect to the noise floor suppression.

Figure 22: Comparison of the sweep profiles OMCIRON 2.0 (adapt automatically) and OMICRON (FRAnalyzer 1.1)


Figure 23: Comparison of the sweep profiles OMICRON 2.0 and OMICRON FRAnalyzer 2.1

4.2 Software Based Methods

Software based methods relay on removing the noise of an already measured FRA plot by using signal processing techniques. Some methods can be found in the literature for the detection and suppression of noisy signals. The method described in [6] is based on the variance analysis developed for detecting and quantifying the nonlinear distortions and the disturbing noise in frequency response measurements. Other examples include the application of the Wavelet transformation that was used in [7] for suppressing noise in the time domain FRA measurements and the application of Kalman filters [8] for eliminating the narrowband noise and broadband noise from the FRA plots. Other methods of easier implementation are based on typical averaging filters such as the moving average filter, exponential weighted moving average filter, and so on. Applications of the fitting algorithms such as the vector fitting [9] can also be used for reconstructing noisy FRA plots.

4.2.1 FRAnalyzer Signal Processing

In the FRAnalyzer software some signal processing tools are implemented for removing noise in FRA plots. Averaging Filter To use the averaging filter:

1. Right-click a selected trace either in the Trace section of the Data Management Window or on the Test View window, and then click Filter and Noise Blanker (see Figure 24). Alternatively, you can select a trace, and then click Filter and Noise Blanker on the Trace menu (see Figure 25).

2. In the Filter and Noise Blanker window, select with the cursors the frequency range where you want to remove noise.

3. Use the averaging filter for removing noise in the nonlinear regions of the plot by clicking the Apply filter to selected region button as shown in Figure 26. You can also change the number of points used for averaging through the filter intensity. Move the intensity to the right to increase the number of points and move the intensity to the left to reduce the number of points.

4. The default intensity value usually provides proper results. Nevertheless, if you observe that the resulting signal gets distorted after applying the filtering, you can remove the changes by using the button at the bottom of the window and decrease the intensity.


Noise Blanker In many cases, the power frequency noise that affects the FRA plot between 30 Hz and 100 Hz corresponds to a linear decay of the attenuation. Due to the linear behavior in this frequency range, a linear interpolation can be applied, herein defined as noise blanker. To use the noise blanker:

1. Follow steps 1 and 2 described earlier in this section.

2. Click the Blank selected region button as shown in Figure 26.

3. After you applied the interpolation, inspect the resulting new plot visually to ensure that the interpolation is applicable to the case under study.

After a successful application of both the averaging filter and the noise blanker, a FRA plot affected by power frequency noise and noise floor can be cleaned up as shown in Figure 27.

Figure 24: Filter and noise blanker function from the Test View window


Figure 25: Filter and noise blanker function from the Trace menu of the toolbar

Figure 26: Noise suppression in FRA plots by using average filtering


Figure 27: Comparison of the FRA plot before and after noise suppression

Magnitude Adjustment You can shift the FRA plots vertically by using the magnitude adjustment function. This can be useful for analysis of the vertical displacements of FRA plots. You can also use this feature for correcting a wrong measurement point in the plot as shown in Figure 28.

Figure 28: Illustration of the magnitude adjustment function


4.3 Connection Technique

The proper connection technique can also help you reduce the noise. Figure 29 compares the FRA plots measured in a substation with very strong noise interferences when using aluminum braids and wire as connection technique. The figure shows that the aluminum braids also reduces the noise.

Figure 29: Noise suppression for different connection techniques


5 Literature

[1] Frequency Response Analysis on Winding Deformation of Power Transformers, The Electric Power Industry Standard of Peoples Republic of China, Std. DL/T911-2004, ICS27.100, F24, Document No. 15182-2005, June 1st, 2005. [2] CIGRE Report 342 WG A2.26, "Mechanical condition-assessment of transformer windings using Frequency Response Analysis, April 2008. [3] IEEE PC57.149/D6, Draft Trial Use Guide for the Application and Interpretation of Frequency Response Analysis for Oil Immersed Transformers April. 2009. [4] E. Brasseur, "How does a receiver work?", 1997, http://www.4p8.com/eric.brasseur/receiv.html [5] N.D. Cogger, R.V.Webb, Frequency Response Analysis, Solartron Analytical, Technical Report 10, 1997. [6] Tom Dhaene, Rik Pintelon, Johan Schoukens, Els Van Gheem, 'Variance Analysis of Frequency Response Function Measurements Using Periodic Excitations', IEEE Transactions on Instrumentation and Measurement, Vol. 54, No. 4, August 2005. [7] R. Wimmer, K. Feser, S. Tenbohlen, M. Krger, 'Erhhung der Reproduzierbarkeit von FRA-Messungen durch Standardisierung', HV Simposium 2006. [8] A.R. Moniri, S. Farshad, 'Modeling the Frequency Response Movements in PowerTransformers for Predicting Purposes'Iranian Journal of Electrical&Electronical Engineering, Vol. 2, No. 1, Jan. 2006', pp. 26-33. [9] B. Gustavsen, A. Semlyen: Rational Approximation of Frequency Domain Responses by Vector Fitting. IEEE Transactions on Power Delivery, 14:1052{1061, 1999.


6 Contact Technical Support

In case of further questions, please contact OMICRON's technical support: Europe/Middle East/Africa [email protected]

Phone: +43 5523-507-333 Fax: +43 5523-507-7333

North and South America [email protected] Phone: +1 713 830-4660 or 1 800 OMICRON Fax: +1 713 830+4661

Asia/Pacific [email protected] Phone: +852 2634 0377 Fax: +852 2634 0390

OMICRON is an international company serving the electrical power

industry with innovative testing and diagnostic solutions. The application of

OMICRON products provides users with the highest level of confidence in

the condition assessment of primary and secondary equipment on their

systems. Services offered in the area of consulting, commissioning,

testing, diagnosis, and training make the product range complete.

Customers in more than 130 countries rely on the company's ability to

supply leading edge technology of excellent quality. Broad application

knowledge and extraordinary customer support provided by offices in

North America, Europe, South and East Asia, and the Middle East,

together with a worldwide network of distributors and representatives,

make the company a market leader in its sector.

Europe, Middle East, Africa

OMICRON electronics GmbH

Oberes Ried 1

6833 Klaus, Austria

Phone: +43 5523 507-0

Fax: +43 5523 507-999

[email protected]

Asia-Pacific

OMICRON electronics Asia Limited

Suite 2006, 20/F, Tower 2

The Gateway, Harbour City

Kowloon, Hong Kong S.A.R.

Phone: +852 2634 0377

Fax: +852 2634 0390

[email protected]

Americas

OMICRON electronics Corp. USA

12 Greenway Plaza, Suite 1510

Houston, TX 77046, USA

Phone: +1 713 830-4660

+1 800-OMICRON

Fax: +1 713 830-4661

[email protected]

For addresses of OMICRON offices with customer service

centers, regional sales offices or offices for training,

consulting and commissioning please visit our web site. www.omicron.at www.omicronusa.com

Franalyzer Anp 10019 Enu

Documents

Transcript of Franalyzer Anp 10019 Enu