1. GP 1.30 (0553) NARROW BAND NOISE ATTENUATION FOR FMCW SOUNDING Alexey A. Kolchev, Andrey O. Shiriy Mari State University Lenin sq., 1 Yoshkar-Ola, Mari El, Russia E-mail: kolchevaa@ mail.ru

2. Abstract The way of the attenuation of the narrow band noise on an output of the chirpsounder receiver is out forward. Detection of the noise is based on difference of noise statistical characteristics and a receiver signal after compression in frequency area. Attenuation of the signal element struck with the concentrated handicap, is carried out on the basis of criterion of gross blunders detection in experimental measurements. Results of numerical and natural experiments are presented according to efficiency of the offered method.

3. Introduction The processing of a LFM signal of the chirpsounder by a method of compression in frequency area consists in multiplication of the received signal to a heterodyne signal, complex-conjugated to a transmitted signal. As a result of demodulation the noise concentrated in frequency area becomes pulse noise, allowing to use specific methods of pulse noise reduction. The operation of the noise reduction system will be effective at significant excess of noise energy over energy of a signal [1]. Therefore the signal on an output of the receiver can be considered within the framework of a mix model of two distributions with different dispersions. Identification of distribution form has been realized on the basis of estimations use of the kurtosis and entropy factor. Basic Equations De-chirped signal from the receiver A(t) is represented by the sum of two symmetric distributions with various dispersions: where - f(x;  ) distribution density of a signal A(t); f 1 (x;   ) - distribution density of a chirp signal sum and wide band noise; f 2 (x;   ) - distribution density of a narrow band noise; h 1,h 2 - densities of distribution in general set;  2,   ,    - corresponding dispersions (    h 2 ). For allocation of the concentrated handicap the signal A(t) was broken on K elements. Estimations of a standard deviation   k were found for each element of a signal.

4. The factor of entropy under the histogram of distribution   k is calculated as where h – width of a column histogram, n – volume of sample, m – number of column histogram, n j - number of sample in j- column of a histogram. If samples of a signal element of two-modal distribution was considered, it speaks about the existence noise and attenuation of this element is made. The indication of the sample belonging to the noise is the big distance of this sample from distribution center. The importance of difference between samples is estimated on the base of criterion finding of the blunders in experimental measurements justified for a great number of distribution laws [2]. If inequality is carried out for the value of   k, where E - factor of an excess, there are no grounds to consider this value to be distinguished considerably from   s. If the inequality for some element is not carried out and   k >   s, this element of a signal is attenuated. Efficiency of attenuation procedure was estimated on value of size n=10lg (P1/P2) (P1 – the attitude signal / noise after attenuation, P2 – the initial attitude signal / noise).

5. Applications to Observations The data received in experiments were used for the analysis, carried out on radiolines Inskip(England) - Nizhny Novgorod and Irkutsk - Nizhny Novgorod. On fig. 1 the fragment of ionogram before (а) and after (б) applications of noise attenuation procedure is given. Fig. 1. Inskip(England) - Nizhny Novgorod, 16.12.2003, Time: 13:52(LT)

6. Separate oscilogramma and spectra of a signal before (a,в) and after (c, d) attenuation of the noise are shown on fig.2. Application of procedure attenuation enables to allocate a signal on frequency of 700 Hz which cannot be allocated in an initial spectrum. Fig. 2. Oscilogramma and spectra of the signal. Inskip(England) - Nizhny Novgorod, 16.12.2003, 10:52 (LT). From 1500 spectra of a signal processed in carried out experiments the noise concentrated in frequency area has been revealed in 60 % spectra. Thus average value of the cut out band has made 14 kHz at a band of a signal in 100 kHz. The average value was equal 4.1 dB, maximal - 13 dB. It is established experimentally, that the value n depends on the attitude P2. If an average value for 5dB <P2 <20dB is equal 4.7 dB (maximal - 13 dB), for 20dB <P2 <40dB - 3 dB (maximal 9 dB).

7. Thus application of the offered technique is especially effective at a small signal on a background of significant noise. Application of estimations of the kurtosis and factor of entropy for identification of the form distribution has increased probability of correct noise detection from 76 % up to 95 %. The factor of correlation between n and the entropy factor in the carried out experiments is equal -0,8, i.e. the increase in energy of noise leads to reduction of the entropy factor. Conclusion Application of the offered method of the narrow band noise attenuation will allow to increase reliability of ionosphere parameters definition, at work low-power FMCW ionosonde, in research of a high order signal mode and weak round-the-world signals. References [1] Goldberg A.P. Complex rejection of the narrow band noise (in Russia) Radiotekhnica, Moscow. 1978, no.4, pp.3-9. [2] Novitskij P.V., Zograf I.A. Errors estimation of measurements results (in Russia), Leningrad. Energoatomizdat. 1991. – 304p.