# 1.

1 Basic concepts of digital filtering
Digital filtering has specific characteristics that you need to pay special attention to. The analog input signal must satisfy certain requirements. Furthermore, on converting an output digital signal into analog form, it is necessary to perform additional signal processing in order to obtain the appropriate result. Figure 1-1 shows the block diagram of digital filtering process.

Figure 1-1. Digital filtering The process of converting an analog signal into digital form is performed by sampling with a finite sampling frequency fs. If an input signal contains frequency components higher than half the sampling frequency (fs/2), it will cause distortion to the original spectrum. This is the reason why it is first necessary to perform filtering of an input signal using a low-pass filter that eliminates high-frequency components from input frequency spectrum. This filter is called anti-aliasing filter as it prevents aliasing. After the process of filtering and sampling, a digital signal is ready for further processing which, in this case, is filtering using the appropriate digital filter. The output signal is also a digital signal which, in some cases, needs to be converted back into analog form. After digital-to-analog conversion, signal contains some frequency components higher than fs/2 that must be eliminated. Again, it is necessary to use a low-pass filter with the sampling frequency fs/2. The specific characteristics of conversion affecting the signal are beyond the scope of this book. Digital filter attenuation is usually expressed in terms of the logarithmic decibel scale (dB). The attenuation measured in decibels can be found using the following expression: a = 20 * log(H(f)) Cut-off frequencies are used for filter specification, which will be discussed later. The cut-off frequency of the passband is a frequency at which the transition of the passband to the transition region occurs. The cut-off frequency of the stopband is a frequency at which the transition of the transition region to the stopband occurs. These two frequencies are equivalent only for the ideal filter which is not possible to realize in practice. In other words, they are always different.

1.2.Types of digital filters

Filters can be classified in several different groups, depending on what criteria are used for classification. The two major types of digital filters are finite impulse response digital filters (FIR filters) and infinite impulse response digital filters (IIR). Both types have some advantages and disadvantages that should be carefully considered when designing a filter. Besides, it is necessary to take into account all fundamental characteristics of a signal to be filtered as these are very important when deciding which filter to use. In most cases, it is only one characteristic that really matters and it is whether it is necessary that filter has linear phase characteristic or not. Speech signal, for example, can be processed in the systems with non-linear phase characteristic. The phase characteristic of a speech signal is not of the essence and as such can be neglected, which results in the possibility to use much wider range of systems for its prosessing. There are also signals for which the phase characteristic is of the essence. A typical example are signals obtained from various sensors in industry. Therefore, it is necessary that a filter has linear phase characteristic to prevent loosing important information. When a signal to be filtered is analysed in this way, it is easy to decide which type of digital filter is best to use. Accordingly, if the phase characteristic is of the essence, FIR filters should be used as they have linear phase characteristic. Such filters are of higher order and more complex, therefore. Otherwise, when it is only frequency response that matters, it is preferable to use IIR digital filters which have far lower order, i.e. are less complex, and thus much easier to realize. The basic characteristics of Finite Impulse Response (FIR) filters are:
  

linear phase characteristic; high filter order (more complex circuits); and stability.

The basic characteristics of Infinite Impulse Response (IIR) are:
  

non-linear phase characteristic; low filter order (less complex circuits); and resulting digital filter has the potential to become unstable

Chapter 2: Finite Impulse Response (FIR) Filter
    

2.1. Introduction 2.2. FIR filter design methods 2.3. Window functions 2.4. Examples 2.5. Finite word-length effects

2.1 Introduction
FIR filters are digital filters with finite impulse response. They are also known as nonrecursive digital filters as they do not have the feedback (a recursive part of a filter), even though recursive algorithms can be used for FIR filter realization.

Figure 2-1-1. Block diagrams of FIR and IIR filters FIR filters can be designed using different methods, but most of them are based on ideal filter approximation. The objective is not to achieve ideal characteristics, as it is impossible anyway, but to achieve sufficiently good characteristics of a filter. The transfer function of FIR filter approaches the ideal as the filter order increases, thus increasing the complexity and amount of time needed for processing input samples of a signal being filtered.

The waveform of frequency response depends on the method used in design process as well as on its parameters. Figure 2-1-3 illustrates input and output signals of non-linear phase systems. which is not like IIR filters that will be discussed in Chapter 3. Each filter category has both advantages and disadvantages. FIR filters are not good solution at all.Figure 2-1-2. FIR filters can have linear phase characteristic. as is the case with processing speech signals. FIR filters are the only option available. in such cases when it is necessary to have a linear phase characteristic. This book describes the most popular method for FIR filter design that uses window functions. This is the reason why it is so important to carefully choose category and type of a filter during design process. . If the linear phase characteristic is not necessary. The characteristics of the transfer function as well as its deviation from the ideal frequency response depend on the filter order and window function in use. Obviously. Ideal low-pass filter approximation The resulting frequency response can be a monotone function or an oscillatory function within a certain frequency range.

anti-symmetric impulse response (about its middle element) . Input signal consists of natural frequency ω and one harmonic with the same amplitude at three times that frequency. However. such distortion in the phase of the signal would be unimportant. In order that the phase characteristic of a FIR filter is linear. if the phase characteristic is of importance. It is obvious that these two signals have different waveforms. which is expressed in the following way: h[n] = h[N-n-1] . shows the block diagram of input signal (left) and output signal (right). the impulse response must be symmetric or anti-symmetric. and π radians at three times that frequency. Figure 2-1-3. only the phase of the second harmonic is changed. such a great distortion mustn’t be allowed.Figure 2-1-3. The power of signals is not changed. nor the amplitudes of harmonics. If we assume that the input is a speech signal whose phase characteristic is not of the essence. The effect of non-linear phase characteristic The system introduces a phase shift of 0 radians at the frequency of ω. the system satisfies all necessary requirements. symmetric impulse response (about its middle element) h[n] = -h[N-n-1] . In this case.

The word specification actually refers to the frequency response specification.One of the drawbacks of FIR filters is a high order of designed filter. Figure 2-2-1 illustrates a low-pass digital filter specification. i. Thus. For this reason. but filters designed this way have small attenuation in the stopband. The transform function of a typical FIR filter can be expressed as a polynomial of a complex variable z-¹. This is the reason why it is so important to use FIR filters only when the linear phase characteristic is very important. it is not possible to understand analyses and synthesis of digital filters. The order of FIR filter is remarkably higher compared to an IIR filter with the same frequency response. As we have mentioned above. 2. The filter design process starts with specifications and requirements of the desirable FIR filter. thus making the filter and its implementation more complex. Each of the given methods has its advantages and disadvantages. All the poles of the transfer function are located at the origin. if the filter is assumed to be of order 10. Which method is to be used in the filter design process depends on the filter specifications and implementation.2 Finite impulse response (FIR) filter design methods Most FIR filter design methods are based on ideal filter approximation. A number of delay lines contained in a filter. You should be aware that without being familiar with these concepts. the window method is most commonly used method for designing filters. FIR filters are guaranteed to be stable. whereas IIR filters have potential to become unstable. it is necessay to learn the basic concepts that will be used further in this book. . Due to its simplicity and efficiency. For example. The resulting filter approximates the ideal characteristic as the filter order increases. it means that it is necessary to save 10 input samples preceeding the current sample. it is very important to carefully choose the right method for FIR filter design.e. determines the order of a filter.1 Basic concepts and FIR filter specification First of all. The sampling frequency method is easy to use. All eleven samples will affect the output sample of FIR filter. 2.2. the design process starts with the specification of desirable FIR filter. a number of input samples that should be saved for the purpose of computing the output sample. This chapter discusses the FIR filter design method using window functions.

Low-pass digital filter specification .Figure 2-2-1a.

ap – maximum ripples in the passband. ωs – normalized cut-off frequency in the stopband. Frequency normalization can be expressed as follows: . Low-pass digital filter specification       ωp – normalized cut-off frequency in the passband.Figure2-2-1b. δ2 – minimum attenuation in the stopband [dB]. δ1 – maximum ripples in the passband. and as – minimum attenuation in the stopband [dB].

Specifications for high-pass.where:    fs is a sampling frequency. band-pass and band-stop filters are defined almost the same way as those for low-pass filters. High-pass digital filter specification . Figure 2-2-2a. f is a frequency to normalize. and ω is normalized frequency. Figure 2-2-2 illustrates a high-pass filter specification.

Figure 2-2-2b. The same values are defined in both cases with the difference that in the later case the passband is substituded by the stopband and vice versa. High-pass digital filter specification Comparing these two figures 2-2-1 and 2-2-2. it is obvious that low-pass and high-pass filters have similar specifications. . Figure 2-2-3 illustrates a band-pass filter specification.

Band-pass digital filter specification .Figure 2-2-3a.

. Band-pass digital filter specification Figure 2-2-4 illustrates a band-stop digital filter specification.Figure 2-2-3b.

Band-stop digital filter specification .Figure 2-2-4a.

Example: .Figure 2-2-4b. Band-stop digital filter specification 2. It is very suitable for analysing discrete time-domain signals and systems. It converts a discrete timedomain signal into a complex frequency-domain representation.2 Z-transform The Z-transform is performed upon discrete-time signals. The Z-transform is defined as: where z is the complex number. The z-transform is derived from the Fourier discrete time-domain transformation and is considered the basic operation in digital filter design process.2.

4. x(n)={1.5. It further becomes: Figure 2-2-5 illustrates the frequency spectrum of the given signal.4.1} .2.Assume that samples of a discrete-time signal x(n) are known.3. 0 ≤ n ≤ 8 z-transform is defined as follows: It becomes: The last expression represents the z-transform of the given signal.2. The Fourier transform can be found by rewriting the previous expression in terms of z as z=ejω. It is necessary to transform this signal through the ztransform and Fourier fransform.3. .

it is easy to notice some similarities between them: In polar coordinate system. the complex number z may be expressed as: The two last expressions lead us to the conclusion that the Fourier transform is just a special form of the z-transform for r=1. In the z plane. which can be seen in Figure 2-2-6 below.Figure 2-2-5. . the Fourier transform is represented as a unit circle. Frequency spectrum of the given signal Comparing Z and Fourier transforms.

it is necessary to consider its zeros and poles in the z plane.2.Figure 2-2-6.1 Transform function of discrete-time systems The Z-transform is primarily used for finding the transfer function of linear discrete-time systems.2. Fourier transform in the z plane 2. When the transfer function is found. The transfer function of discrete-time systems is defined as: .

FIR filters do not have this recursive part of the transfer function. H0 is a constant. . As we previously mentioned. so the expression above can be simplified in the following way: The impulse response of discrete-time system is obtained from inverse z-transform of the transfer function i. qi are the zeros of the transfer function. the transfer function of discrete-time system is actually the Ztransform of impulse response: where h(n) is the impulse response of discrete-time system. The recursive part of the transfer function is actually a feedback of discrete-time system.e.where:      bi are the feedforward filter coefficients (non-recursive part). aj are the feedback filter coefficients (recursive part). and pj are the poles of the transfer function.

discrete-time system design. i.Figure 2-2-7. the discrete-time system. Example: The impulse response of a 10-th order FIR filter designed using the Hamming window (discussed in the next chapter) is: . can also be expressed as the convolution of the input signal x(n) and the impulse response h(n) of the system: In the frequency domain. the discrete-time system. can be expressed as the multiplication of the Z-transform input signal X(z) and the transfer function H(z) of the system: which further gives: The first way of representing discrete-time systems is more suitable for software implementation itself. shown in Figure 2-2-7.e. whereas the later is more suitable for analyse. shown in Figure 2-2-7. Block diagram of a linear discrete-time system In the time domain. hardware implementation (described later) and synthesis.

h(n) = {0.0638.2π then: . If for example ω = 0.0127.2761.0.2761. 0. 0. 0.0127. 0. . .0638. 0.0248. 0.0. .0. 0} The transfer function of this filter is found via the ztransform of impulse response: Using the following expression: it is possible to yield the transfer function of the fixed normalized frequency.0248.4.

.One example of hardware realization of this filter is illustrated in Figure 2-2-8.

FIR filter realization Software realization requires a buffer of minimum length 9. which in this case means that the circular buffer is of length 16=2^4. Buffers are usually circular and their length can be expressed as 2^n.Figure 2-2-8. .

detect round-off errors made due to software implementation of a filter as well as errors in the coefficients encountered during hardware implementation of a filter. it ocurrs due to impossibility of representing the coefficients with absolute accuracy. This property is particularly typical of highorder filters because their zeros are very close each other. However. In software implementation. The finite word-length effect on the transform function of a FIR filter is clearly marked. . Such errors cause frequency deviation of discrete-time system designed.2 Effect of the poles and zeros of the transfer function The location of zeros and poles of the transfer function is very important for discretetime system analyses and synthesis. If this requirement is not satisfied. Assume that a 50-th order low-pass FIR filter with normalized cut-off frequency of 0. This property of FIR filters actually represents their essential adventage. slight errors in coefficient representation may cause large frequency deviations. which is very dangerous. According to their position it is possible to test stability of a discrete-time system. In order that a discrete-time system is stable. which further means that the transfer function has no poles.2.25 Hz is designed using the Hann window. as shown in Figure 2-2-6.2. an error is triggered by the finite wordlength effect. Frequency deviation depends on the spacing between the zeros of the FIR filter transfer function. Recalling that FIR flters do not have a feedback. whereas in hardware implementation. An error in coefficient representation is always produced due to software and hardware implementation. The result in both cases is that the value of coefficients differs from their value obtained in design process. FIR filter coefficient error affects more the frequency characteristic as the spacing between the zeros of the transfer function narrows. This causes a FIR filter to be always stable. Filter stability will be discussed in more details along with IIR filters which have potential to become unstable because of the feedback they have. From now on. The location of zeros doesn’t affect the stabilty of discrete-time systems. Figure 2-2-9 illustrates the required and obtained frequency characteristic of a FIR filter. the system becomes unstable. only the zeros of the transfer function will be discussed in this chapter.2. all poles of the discrete-time system transfer function must be located inside the unit circle.

As the ideal filter frequency response is infinite. while more affecting the resulting frequency characteristic. even though it is very large at certain points. . it is easy to produce sampling errors.Figure 2-2-9. The objective is to compute the ideal filter samples. so that such deviation is acceptable.3 Ideal filter approximation The ideal filter frequency response is used when designing FIR filters using window functions. FIR filters have finite impulse response. it is necessary to be very careful when designing high order filters because the transfer function zeros get closer. Figure 2-2-10 illustrates the transfer functions of four standard ideal filters.2. which means the ideal filter frequency sampling must be performed in a finite number of points. The error is less as the filter order increases. 2. as this is not a common case. However. Deviation from required frequency characteristic The frequency deviation shown in Figure 2-2-9 is basically slight deviation. The minimum attenuation and the width of transition region of the resulting IIR filter remain unchanged.

. Transfer functions of four standard ideal filters The ideal filter frequency response can be computed via inverse Fourier transform.Figure 2-2-10. The four standard ideal filters frequency responses are contained in the table 2-2-1 below.

If M is not an integer. the expression for inverse Fourier transform must be used: . N can be expressed as N = 2M. Since the variable n ranges between 0 and N. The constant M is an integer if the filter order N is even.Table 2-2-1. which is not the case with odd order filters. If it is needed to find frequency response of a non-standard ideal filter. where N is the filter order. A constant M can be expressed as M = N / 2. the ideal filter frequency response is symmetric about its Mth sample which is found via expression shown in the table 2-2-1 above. the ideal filter frequency response has N+1 sample. but not about some frequency response sample. Equivalently. the ideal filter frequency response is still symmetric. The frequency responses of four standard ideal filters The value of variable n ranges between 0 and N. If M is an integer (even filter order).

that is. If the resulting filter has too wide or too narrow transition region.3). ωc2). it is necessary to change the filter order by increasing or decreasing it according to needs. Specifying a window function according to the filter specifications.Non-standard filters are rarely used. Computing the window function coefficients.3. After this step. Computing the ideal filter coefficients according to the filter order. it is possible to calculate the window function coefficients w[n] using the formula for the specified window function. This issue is also covered in the next chapter.2. Accordingly. 6. . ωc1. The window function and filter order are both specified according to these parameters. The final objective of this step is to obtain the coefficients hd[n]. it is necessary to find the ideal filter frequency samples. This point will be discussed im more detail in the next chapter (2.3 under Ideal filter approximation. The expressions used for computing these samples are discussed in section 2. After estimating the window function coefficients. the integral above must be computed via various numerical methodes.4 FIR filter design using window functions The FIR filter design process via window functions can be split into several steps: 1. when the window function is known. One of the techniques for computing is provided in chapter 2. 5. When both the window function and filter order are known. Computing the filter order required for a given set of specifications. The final objective of defining filter specifications is to find the desired normalized frequencies (ωc. 2. Defining filter specifications. However. the selected window function must satisfy the given specifications. Two sequencies w[n] and hd[n] have the same number of elements. 5 and 6 are iterated as many times as needed. 4. transition width and stopband attenuation. and after that steps 4. 2. we can compute the filter order required for a given set of specifications.2. 7. if there is a need to use some of them. 3. Computing FIR filter coefficients according to the obtained window function and ideal filter coefficients.

the filter order can be decreased for the purpose of optimizing hardware and/or software resources. multiply them in order to obtain the frequency response of designed filter and reestimate the transfer function as well. It is also necessary to reestimate the filter frequency coefficients after that.2. The transfer function can be found via the z- . For the sake of precise estimates. it is necessary to increase the filter order.5 FIR filter realization FIR filter transfer function can be expressed as: The frequency response realized in the time domain is of more interest for FIR filter realization (both hardware and software). the filter order should be decreased or increased by 1. If the transition region is narrower than needed. reestimate the window function coefficients and ideal filter frequency samples.The next step is to compute the frequency response of designed filter h[n] using the following expression: Lastly. the transfer function of designed filter will be found by transforming impulse response via Fourier transform: or via Z-transform: If the transition region of designed filter is wider than needed. 2.

This chapter covers direct. and y[n] are FIR filter output samples. FIR filter output samples can be computed using the following expression: where:    x[k] are FIR filter input samples. . As for the software implementation.2.1 Direct realization Direct realization of FIR filter is based on the direct implementation of this expression: Direct realization is also known as a transversal filter. Figure 2-2-11 illustrates the block diagram describing the hardware direct realization of a FIR filter.5. 2. There are several types of FIR filter realization. direct and optimized realizations will be discussed in this book. A good property of FIR filters is that they are less sensitive to the accuracy of constants than IIR filters of the same order. direct transpose and cascade realizations which are very convenient for the hardware implementation of a filter.transform of a FIR filter frequency response. h[k] are the coefficients of FIR filter frequency response.

As for software direct realization of the FIR filter. Accordingly: .e. it is necessary to provide a buffer for minimum N samples. where N is the number of FIR filter coefficients. the multiplication constants are the same as the transfer function coefficients. For its simplicity and speed. FIR filter direct realization For direct realization structure. i. The value of the constant k is a minimum value for which the expression N ≤ 2^k is valid. the FIR filter frequency response coefficients.Figure 2-2-11. most commonly used buffer is so called circular buffer the length of which can be expressed as 2^k.

Figure 2-2-12. and 3. The buffer length is obtained in the following way: . Reading the samples of a signal being filtered. Since the buffer is 16 bits wide. Design this filter using direct realization with circular buffer. resulting in a FIR filter output sample. circular buffer addressing is performed using addressing mode 16: Example: Assuming that a filter used in this example is a 5th order FIR filter. Circular buffer of length 16 = 2^4 The algorithm used for software direct realization of FIR filter consists of: 1. Performing a convolution operation upon filter coefficients (frequency response coefficients).where the operator represents rounding down to a less value. Storing a new sample on the first available location. 2. The buffer length needs to be 2^k.

where N is the number of FIR filter coefficients. The contents of the buffer after receiving the first 10 samples is shown in the table 2-2-2. Figure 2-2-13 illustrates the block diagram describing hardware direct transpose realization of a FIR filter. the direct transpose realization must also have a buffer of minimum length N. addr. addr. filtering of input samples is performed as per formula below: 2.2. addr. Input circular buffer after receiving 10 samples For software realization. . Step 0 1 2 3 4 5 6 7 8 9 10 addr. Speaking about software implementation. 7 6 5 4 3 2 1 0 x[0] x[0] x[0] x[0] x[0] x[0] x[0] x[0] x[8] x[8] x[7] x[7] x[7] x[6] x[6] x[6] x[6] x[5] x[5] x[5] x[5] x[5] x[4] x[4] x[4] x[4] x[4] x[4] x[3] x[3] x[3] x[3] x[3] x[3] x[3] x[2] x[2] x[2] x[2] x[2] x[2] x[2] x[2] x[1] x[1] x[1] x[1] x[1] x[1] x[1] x[1] x[9] Table 2-2-2. addr.2 Direct transpose realization Direct transpose realization is similar to direct realization in many ways. addr.This means that the minimum length of circular buffer is 2^3 = 8.5. addr. addr. Input samples are denoted by x[n]. whereas the shaded cells denote buffer locations that have been changed.

.

FIR filter cascade realization The number of multipliers. Figure 2-2-14. otherwise very convenient for . is commonly used for FIR filter hardware realization. a filter is divided into several low-order sections. 2. ak2 are the multiplication coefficients of section k. Figure 2-2-14 illustrates the block diagram describing the hardware cascade realization of a FIR filter. very convenient for its modular structure. Direct transpose realization of a FIR filter There are no significant differences between direct and direct transpose realizations. The main advantage of the cascade realization is its modularity. The second-order sections are most commonly used. When using this realization.2.5. The cascade realization is normally used for higher order filter realization. Individual sections are mostly in direct form realization. Both structures have the same number of delay elements. The transfer function of the cascade realization looks as follows: where:   M is the number of sections.3 Cascade realization Cascade realization. adders and delays is the same as for direct realization. the same number of multipliers and the same coefficients to perform multiplication upon. and ak1.Figure 2-2-13. although they can also be in direct transpose form realization.

2. There are also anti-symmetric FIR filters that are beyond the scope of this book.4 Optimized realization Optimized realization has less. whereas Figure 2-2-16 illustrates that for odd N. the optimized realization may be used for these filters as well. because a reduction in the number of multipliers enhances the process of convolution (samples filtering process). but more demanding multipliers for realization.hardware implementation. This realization is most commonly used for software implementation of FIR filters. 2. . Optimized realization utilizes the symmetry of frequency response coefficients. Anyway.5. The cascade of second order sections is important for the realization of the filters of arbitrary order. The symmetry of the coefficients of FIR filter frequency response can be expressed by equation below: This symmetry makes it possible for the transfer function to be expressed as follows: Figure 2-2-15 illustrates the block diagram of optimized realization for even N.

Figure 2-2-15. Optimized realization for even frequency response 2. Optimized realization for odd frequency response Figure 2-2-16.3 Window functions .

A window is a finite array consisting of coefficients selected to satisfy the desirable requirements. The simplicity of design process makes this method very popular. . and The higher suppression of undesirable spectrum. i. The point is to find these coefficients denoted by w[n]. Each function is a kind of compromise between the two following requirements:   The higher the selectivity. This chapter provides a few methods for estimating coefficients and basic characteristics of the window itself as well as the result filters designed using these coefficients. and The filter order according to the required specifications (selectivity and stopband attenuation).The window method is most commonly used method for designing FIR filters.e. When designing digital FIR filters using window functions it is necessary to specify:   A window function to be used. These two requirements are interrelated. the higher the stopband attenuation.e. Table 2-3-1 below contains all window functions mentioned in this chapter and briefly compares their selectivity and stopband attenuation. i. the narrower the transition region.

.#19074:95:98.25043901789.250841.30/3901443.970.43.25088:239.250  \$3.4011.308..943:5431907.088.//70883850714720/:83.94341#1907.:.:.4:9434507..//70883 24/0  .090-:11078-98/0 .  .3/   !0714723..9439.1907:80/3980.:..-04.4792:80/1478419.947/07#1907 083 981907:83/70.70/70.43889841   #0.7.947 705708039874:3/3/4394.0398 1706:03.83.70854380 ..7-:110741039) %0.970.7.:0  :70   7.7-:1107.9.7-:1107 %0-:1107039300/894-0)  %0-:110703984-9.4011..2508../3908. 070904507.943.0398 708:933.-0319070/   \$9473.

./.2508 478419.43903984190-:1107..-0   35:9.//7  .997.0-003.0.3./0/.94389.702502039.7..//7  .943 90/70.38548070.250888433909.08/03490-:11074.3901789 8.:..8908.970.3.38548070.389.-:110741 232:2039 0708903:2-0741#1907.85071472:.70/70.997.7.//7  .190770.2508850714720/.70/03490/-3( 070..//7  .94341.7/.3 8..30/  .385480 70.0.-4:9 8419.9432:89.190770.7-:1107.943 190734135:98.90890-4.84...0398  :70  :897.4011.794/70.943882.-0    35:98.997.//7  .2508.9 .//7  ..:.2/08.0.//7                     (         (  (        ( (  (       ( ( (  (      ( ( ( (  (     ( ( ( ( (  (    ( ( ( ( ( (  (   ( ( ( ( ( ( (  (  ( ( ( ( ( ( ( (   ( ( ( ( ( ( ( ( \$905 %.//7  .%820.-04   % ^WU``^SZ_\[_W^WS S`[Z 70.7-:11078)  %0.#1907  .7-3.94332.8 \$50.7070.7.990232:203941.

 .

03039147 .#1907  :70   #1907./0/39480.385480147270.702489.943 %03:2-07412:95078 ...//078.94:90./:.9438.8.43/ 47/0780.203:2-0741/0..943448.3/ 908..43.7070.7/.9:70 8.8.3/ ..7024893/70..84-03/70.9.7897:.203:2-07412:95078.08-09003/70.943:543   % S_USVW^WS S`[Z ./070...42243:80/ 3/.9438.7/.2/08.94389824/:.07.943  :70  :897./070..3.970.9438../070.3.7.39/110703..38548070.7-390./070.943 .94341 .997.39./070.030391479824/:.42243:80/147# 1907.943 %0.9438 49 897:..:70   70.43.7034831..8147/70.9147270..../..8.943  %097....0020398 908. .8.90890-4. 80../.81448  070 O O 8903:2-074180..20.03989450714722:95.9434190..943 .9:708.38548070.07.3//70.70..997.79 490780...:80/1470747/071907 70.20.8.19078/.04190.94341..9 97.943 .70902:95../070.4011.8.07..943..0908..9438 %080.3//0.943 03:839870.03984180.94383472.381071:3.4011.943 %0 2.. 4 47/0780.#1907 %070.88908.

3914790 70.943 %0..7-97./04180.8..43/47/0780..7/.9438825479.747/07   %%.702502039.94341901907841..

9431470.943-04  %88220972..91474//  ..0895488-01479097.39 822097.3-00570880/- 06:.4:943 8.94394-00570880/.0890574.4011.702502039.70854380.80  %08220974190.43..03 070.81448 :70  :897.381071:3..84.039841#19071706:03..70-043/908.943 %8 70.08841. 70/:. fifr | Low Pass Filter | Control Theory

 WV^WS S`[Z 5920/70.24145920/70.90890-4.8088 -:92470/02.70854380.0398 %070.42243:80/1478419.-0:80/147908019078.3/32:9507814770.3.9432.7. 90 45920/70./.9089.#190789.9433903:2-07412:9507803.4011....94341#19078 -0.4504198-44 3.2508 19073574.8 :70  :897.943.:80.94382489.9.70 .088  5920/70.943:90890822097411706:03.

:70    5920/70..031706:03.9431474//1706:03.9438 .9431470.70854380  3/41:3.70854380 :70    5920/70..

9903:.43889341..0882.42243:80/2094/147/0833#19078 %0 825.890708:919078/0830/:839080 .94394-0:80/..94389830.3/ %0190747/07.70390770..3/-.1:3.3/-701 ..4011.97:2  0 9007908945-.389437043./08.5907574.1390.-0850.1 O O 3/41:3.102094/81470892..42574280-09003 90941443706:7020398 O O %00790800..-0  -04.3/41:3.3/8945-.943  %08094706:7020398.4011.9903:.#19078:833/41:3.4011.9438 800.3/ %0078:5570884341:3/087.90789.841903/49801.794850.9903:.425.3/41..80.3/.7.439.0398 %0543989413/9080.%03/42094/82489.774079097.0398800.7  3/48.943   .07545:. .941/083574.9.8..3/ .94382039430/398.93.90/ .3/8945-.9.4011.088.47/39490706:70/850.9.9438.0398.9  0 903.3/ .90/948.08982094/.77.38.98190/087.9.5907.9.0398/03490/-3(  03/0833/9.708907800.-0 706:7020398 %8.1.943  %..