On Spectral Approach to the Synthesis of Shaping Filters
Konstantin A. Rybakov · 2025
Read On Spectral Approach to the Synthesis of Shaping Filters by Konstantin A. Rybakov (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-15174.txt.
Introduction
Konstantin A Rybakov writes on how to build a shaping filter for a random process, in a paper from two thousand twenty-five. The aim is a process with a given power spectral density, stated as a rational form. Such filters let a known input be shaped into noise with the right spread of power across bands.
He lays out several routes to the task, with most weight on the spectral form of the maths. In that form, the link for the filter comes from a transfer function with no extra work. The paper gives all the relations a reader needs to build the filter from that spectral form.
Catalog record
| CREATOR | Konstantin A. Rybakov |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-15174
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
On Spectral Approach to the Synthesis of Shaping Filters Konstantin A. Rybakov Abstract [email protected] Abstract. This paper describes various approaches to modeling a random process with a given rational power spectral density. The main attention is paid to the spectral form of mathematical description, which allows one to obtain a relation for the shaping filter using a transfer function without any additional calculations. The paper provides all necessary relations for the implementation of the shaping filter based on the spectral form of mathematical description. Keywords: random process, power spectral density, spectral form of mathematical description, statistical modeling, shaping filter MSC: 60G15; 60H35 1. Introduction The paper considers various approaches to modeling a random process with a given power spectral density that is specified as a rational function, and the distribution of this random process is assumed to be Gaussian. Thus, we can speak about the problem of synthesizing a shaping filter, i.e., a linear continuous system whose input signal is the Gaussian white noise and whose output signal is a random process with desired probability characteristics. The description of shaping filters is traditionally based on power spectral densities or stochastic differential equations [ 1 ] . The problem of synthesizing a shaping filter is relevant and in demand in a number of applications. We can emphasize problems of information processing in navigation [ 1 ] , problems of radio engineering [ 2 ] , and many others. It is needed for statistical modeling of complex engineering systems subject to random effects, e.g., in problems of aircraft or rotorcraft flight dynamics in turbulence [ 3 , 4 ] . The last example, namely the Dryden turbulence model, is considered in the recent paper [ 5 ] , which proposes a new method to obtain a system of stochastic differential equations for a given power spectral density (strictly speaking, for a transfer function that determines the power spectral density), i.e., one of the methods for synthesizing a shaping filter. There are also other methods to obtain a system of stochastic differential equations [ 6 ] . However, it is interesting to note that despite the availability of such methods with a simple implementation, the standard [ 7 ] recommends for modeling the turbulence to use the numerical scheme corresponding to first-order stochastic differential equations, although only two of six power spectral densities in the Dryden turbulence model correspond to first-order equations. The remaining power spectral densities correspond to second- and third-order equations. The shaping filter can be constructively defined not only by stochastic differential equations. In general, for linear continuous systems with deterministic or random input signals, various forms of mathematical description can be used, on the basis of which methods and algorithms for solving analysis, synthesis, and identification problems are proposed. Four forms are distinguished [ 8 , 9 ] : (1) description by linear differential equations; (2) description by response functions; (3) description by integral transforms; (4) description by spectral transforms. The description of linear discrete systems assumes similar forms, of course, taking into account the discreteness of time, e.g., linear difference equations are used instead of linear differential equations. The aim of this paper is to complement the paper [ 5 ] and to give a broader description of methods for modeling a random process with a given power spectral density, as well as their relation in the context of different forms of mathematical description for linear continuous systems. The main attention is paid to the spectral form of mathematical description, which allows one to obtain a relation for the shaping filter using a transfer function without any additional calculations. And this relation forms the basis for approximate modeling of a random process with desired probability characteristics. This approach allows one to obtain an approximation of a random process in continuous time and to calculate the mean square approximation error exactly. The paper contains an addition to [ 5 ] in terms of obtaining a system of stochastic differential equations that defines the shaping filter. The rest of this paper is organized as follows. Section 2 discusses the description of the shaping filter by differential equations. Section 3 defines the shaping filter by the impulse response function. A new approach for the description of the shaping filter using the spectral form is proposed in Section 4 , and corresponding examples are given in Section 5 . The paper is summarized in Section 6 . Appendix contains all necessary relations for the implementation of the shaping filter based on the spectral form of mathematical description. 2. Description of the Shaping Filter by Differential Equations We assume that the power spectral density S(\omega) of the centered stationary random process x(t) is given; it is defined by the Fourier transform \mathbb{F} of the covariance function R(\eta) , i.e., \mathrm{E}x(t)=0,\ \ \ \mathrm{E}x(t)x(t+\eta)=R(\eta),\ \ \ S(\omega)=\mathbb{F}[R(\eta)]=\int_{-\infty}^{+\infty}R(\eta)\mathrm{e}^{-\mathrm{i}\omega\eta}d\eta, and this power spectral density is represented as S(\omega)=|H(\mathrm{i}\omega)|^{2}. (1) In the above formulae, t and \eta denote the time, \omega is the frequency, \mathrm{i} is the imaginary unit, H(s) is the rational function of a complex variable s : H(s)=\frac{M(s)}{D(s)}=\frac{b_{m}s^{m}+\ldots+b_{1}s+b_{0}}{a_{n}s^{n}+\ldots+a_{1}s+a_{0}}, (2) a_{0},a_{1},\dots,a_{n} and b_{0},b_{1},\dots,b_{m} are real numbers, a_{n},b_{m}\neq 0 , n>m , and \mathrm{E} means the mathematical expectation. Then x(t) is the output signal of a one-dimensional linear stationary system called the shaping filter [ 6 , 1 ] whose input signal g(t) is the white noise, H(s) is the transfer function, and H(\mathrm{i}\omega) is the frequency response function [ 9 ] . The random process x(t) is assumed to be Gaussian, i.e., g(t) is the standard Gaussian white noise. The transfer function H(s) corresponds to the following differential equation [ 9 ] : a_{n}x^{(n)}(t)+\ldots+a_{1}x^{\prime}(t)+a_{0}x(t)=b_{m}g^{(m)}(t)+\ldots+b_{1}g^{\prime}(t)+b_{0}g(t). (3) If we assume that g(t) is a sufficiently smooth function, i.e., its derivatives up to order m exist and are continuous, then the (ordinary) differential equation does not raise any questions, and its solution is strictly defined. But if g(t) is the standard Gaussian white noise, then the (stochastic) differential equation can be understood only formally even for m=0 . Sometimes the correctness of typical white noise definitions is questioned [ 10 ] ; however, here not only the white noise is used but also its derivatives. The main result of [ 5 ] is a new method to obtain a system of linear stochastic differential equations that do not contain white noise derivatives: \bar{x}^{\prime}(t)=A\bar{x}(t)+Bg(t), (4) where \bar{x}(t) is the vector-valued random process of size n^{*} , A is the square matrix of size n^{*}\times n^{*} , and B is the column matrix of size n^{*}\times 1 , n^{*}\geqslant n . A strict meaning can be given to this equation if we rewrite it as the stochastic integral equation using one of the stochastic integral definitions: \bar{x}(t)=\bar{x}_{0}+\int_{0}^{t}A\bar{x}(\tau)d\tau+\int_{0}^{t}Bdw(\tau)=\bar{x}_{0}+A\int_{0}^{t}\bar{x}(\tau)d\tau+Bw(\tau), (5) where \bar{x}_{0} is the initial random vector of size n^{*} , and w(t) is the standard Wiener process. Since the white noise is additive, this equation can be understood in different ways: in the sense of Itô, Stratonovich, or Ogawa [ 15 , 16 ] . The solutions to such equations are stochastically equivalent. The random process x(t) is expressed as follows: x(t)=C\bar{x}(t), (6) where C is the row matrix of size 1\times n^{*} . Equations ( 4 ) and ( 6