2nd International Conference on Electrical, Coputer Engineering and Electronics (ICECEE 25 Mitigation of GPS L 2 signal in the H I observation based on NLMS algorith Zhong Danei, a, Wang zhan, a, Cheng zhu, a, Huang Da, a National University of Defense Technology, Changsha 473, China a z823374@63.co Keywords: evaluating indicator, radio astronoy, NLMS, iliary channel, RFI. Abstract. In the radio observation based on the large single antenna, in order to reduce the interference of navigation signal L 2 fro observation of the red-shifted H I spectral line at L-band, we used the an iliary channel and noralized LMS algorith, and proposed new evaluating indicators deduced by theoretical ethod. Finally, we verified the suppression perforance and the indicators utility by ultiple siulations and statistical ethod. Introduction With the rapid developent of the radio astronoy, the technology of the large radio antennas is iproving. However, the radio frequency interference (RFI significantly liit the research of the radio astronoy in China. Usually, radio astronoy signal is very weak and alost the receipt power spectral flux density by radio signal source is less than Jy (Jansky= -26 W/ 2 /Hz. So that the radio signals is susceptible to RFI. In addition, as radio telescopes are high-sensitivity, high-resolution and wide range of operating frequency, they are sensitive to weak interference, such as navigation signal. Even though the iniu level of navigation signals just approxiately -6dB [], it still has a huge ipact on radio astronoy observations. The observation of H I is an iportant issue in radio astronoy observation. As H I is one of ain coponents of the interstellar edius and distribute throughout the galaxy, the study for H I will contribute to galaxy kineatic [2]. The H I without red-shift is usually called 2c spectral line and its center frequency is 42.46MHz. However, the frequency range of in actual observation is fro 4MHz to 4MHz or even wider due to the red-shift which caused by cosic expansion [3]. Moreover, navigation signals alost distribute on L-band as well, such as L 2 (227.6 MHz, L3 (38.5MHz and so on, which observation frequency overlaps red-shift H I. Therefore, we need an effective anti-interference algorith to eliinate the interference and obtain the useful H I. Usually, the ethods to eliinate the interference by navigation signals are classified into two pieces. One is only using the statistical data of radio observation, including threshold processing in tie or frequency doain and the odel reconstruction ethod. The core idea of threshold processing is that we can drop a part of data which is obvious over the variance or other eigenvalues. It is easy and convenience for auto-processing, but it has an influence on the integrity of the radio signal [4, 5]. The odel reconstruction ethod is using priori inforation of navigation signals to odel the interference signal, then we estiate the interference by navigation signals to eliinate it [6]. This siulation ethod can often get obvious effect on anti-interference, even reaching to 2dB, but it need a large nuber prior inforation to odel, which can hardly apply on reality. The second one is adaptive noise cancelling (ANC [7, 8] ethod, which is using the correlation of ain antenna and iliary antenna to eliinate the interference. The ethod perfors very well on kinds of interferences, however, we cannot research the perforance of ANC on radio observation as the liitation of evaluation criteria. To solve those probles proposed before, this paper put forward soe evaluation criteria for perforance of anti-interference in siulating H I and L 2 signal, and this siulation is based on the 25-eter cassegrain antenna in Xinjiang Observatory, and in consideration of the free space propagation loss, atospheric loss, different observing angles at the sae tie [9]. These criteria are radio signal attenuation, G s and interference suppression, G I. This paper apply 25. The authors - Published by Atlantis Press 25
single-iliary-antenna ANC to eliinate the interference fro signals of ain antenna. On the one hand, we verify and study anti-interference effectiveness of the ethod that applying the LMS algorith in radio observation by those criteria we proposed before. On the other hand, we verify the effectiveness of the criteria itself. Basic principle Antenna Table. The power of each part of received signals Signal (dbw H I L 2 Noise Main antenna rn ( - 9.89-53.32-44.39 Auxiliary antenna r( n - 59.86-78.49-88.38 Fro table, there are the power coparison of the ain and iliary antenna through the processes which aplify, ixing and so on. The navigation signal L 2 will interfere the radio signal H I observation in the 25-eter antenna, according to the table, the interference power is ore than the H I s, but less than the noise s in the ain antenna, as result of the low interference-to-noise ratio (INR, the LMS algorith show the poor perforance. Therefore when navigation satellite pass the ain-lobe, we can obtain a strong L 2. The iliary antenna can iprove the received power of L 2 by rising the antenna aperture, and the data in the table is the result of 5-eter iliary antenna aperture. The power of L 2 is ore than the noise s in iliary antenna, and the power ratio of L 2 to H I reaches 8, therefore the H I in iliary antenna can be ignored. Algorith Signal odel. Fro the above analysis, it can be seen that the received signal r(n of the ain antenna is the su of H I fro ain lobes, L 2 fro the third side lobes (the observing angle is 2 and the Gaussian noise: rn ( sn ( in ( Nn ( ( sn (, in ( are respectively H I signal and L 2 signal, Nn ( is Gaussian noise, all of the are obtained fro the ain antenna receiver IF aplifier. Siilarly, the r ( n fro iliary antenna is as follows: r ( n i ( n N ( n (2 where, i ( n and N ( n express L 2 and the Gaussian noise fro the iliary antenna. Defines the H I in the coputer siulation environent as Gaussian white noise []. Algorith. For the correlation between in ( and i ( n in tie doain, we can use the adaptive algoriths to eliinate the L 2 in ain antenna. There are least squares /recursive least squares (LS/RLS and least ean square (LMS algorith. Copare to the forer algoriths, the latter has a lower convergence rate, but a sipler structure, a saller aount of coputation, a ore stable perforance and easily ipleented. Due to the extreely weak power of signal in ain and iliary antenna, a very large step factor (about orders of agnitude in LMS algorith is required in order to ake the weight vector convergence. The noralized LMS (NLMS is used to figure it out, and increased the convergence speed and stability, the procedure is divided into two steps, and details are as follows: 26
Fig. The scheatic of the NLMS: yn ( is output signal of the FIR filter and evaluated by w ( n and r ( n, the relationship is as follows: T en ( rn ( w ( n r ( n (3 en ( is the difference between rn ( and yn, ( w( n is the weight vector and the length of it represents the filter order L, r ( n is the iliary signal in every circulation process, which is the L nuber data fro r( n, backward extracting the increasing of n step by step. 2 Weight vector adjustent: w ( n is adjusted as follows: w( n w( n r ( ( 2 n e n r ( n is the step factor and noralized by the agnitude of r ( n. (4 Evaluation indicators Define the received signals in ain antenna as follows: rn ( GF ( s sn ( GF ( i in ( Nn ( (5 Where G is the ain antenna gain, F( s F( i express the noralized directivity function of the radio signal and the navigation interference respectively. Denote sn ( in ( Nn ( as radio signals, navigation interference and noise before they enter the antenna. Define r ( n Gr Fr ( i i( n+ n N ( n (6 as the signal in the iliary antenna, whereg represent the iliary antenna gain and F ( is r the noralized directivity function of navigation interference in iliary antenna, n is the tie delay and N( n is the iliary antenna noise. In general, assue that after the anti-interference algorith there s only changes of noralized directivity function and noise, as below: r( n GF( s s( n GF( i i( n N( n (7 assue G sn ( in ( are not changed before and after the algorith, while Nn ( F( s F( i rn ( are turned into N ( n F ( s F ( i r( n separately after the anti-interference algorith,. In the coputer siulation, paraeters sn ( in ( Nn ( N( n G G r F( s F( i and Fr( i are given. As a result, with sn ( rn ( and r ( n we can get two cross-correlation functions: r i 27
R ( E[ r( n s ( n ] G F( E[ s( n s ( n ] (8 rs s R ( E[ r( n s ( n ] G F( E[ s( n s ( n ] (9 rs 2 2 s 2 Make tie delays = 2, we get signal power attenuation G s : R ( F ( s rs Gs Rrs ( F ( s ( In the sae way, use in ( rn ( and r( n we get another two cross-correlation functions: R ( n E[ r( n i ( nn ] G F( E[ i( n i ( n n ] ( ri i R ( n E[ r( n i ( nn ] G F( E[ i( n i ( n n ] (2 ri 2 2 i 2 Make n= n2 and we get interference power suppression G I : G I R ( F( i ri R ( F ( (3 ri i It can be seen that G s and G I relate to the changes of noralized directivity functions only, thus, they are able to reflect the changes of radio signal power and interference power respectively correctly before and after the anti-interference algorith, oreover they can be used to verify the perforance of the algorith at the sae tie. Results of Matlab eulation L is the filter order, through ultiple siulations, when L =2, =. the inhibitory effective of L 2 is rather better. Before processed After processed 3 4 25 35 3 2 25 Aplitude 5 Aplitude 2 5 5 5 8 6 24 32 4 Frequency/MHz 8 6 24 32 4 Frequency/MHz Fig 2. The power spectru of received signal before and after the NLMS Fig 2.depicts the power spectru of received signal in ain antenna in the left picture and the average value of power spectru disposed the interference by the Monte-Carlo ethod in the right. The siulation result shows that G s =.db, G I =2.55dB..2 Radio signal attenuation Gs/dB. 2 3 4 5 6 Main antenna INR/dB 5 Interference cancellation Gi/dB 4 3 2 3 4 5 6 Main antenna INR/dB Fig 3. Trends of G s and G I with the ain antenna INR increasing Fig 3.respectively shows the change of G s and G I with the step-by-step increasing of ain antenna INR, obviously, G s is basically constant. Coparing to G s, G I is iproving with the increasing INR value, and when INR reaches 4dB, G I is basically stable. 28
Radio signal attenuation Gs/dB.5 2 3 4 5 6 Auxiliary antenna INR/dB 4 Interference cancellation Gi/dB 3 Fig 4. Trends of G s and G I with the iliary antenna INR increasing Fig 4. respectively shows the change of G s and G I with the step-by-step increasing of iliary antenna INR, relatively, G s is basically unchangeable, G I is raising with INR, and when INR approaches 4dB, G I is generally stable. Conclusions 2 2 3 4 5 6 Auxiliary antenna INR/dB This paper proposed soe new evaluation criteria, G s and G I by odeling the processing of received signal fro the 25-eter cassegrain antenna in Xinjiang Observatory with single-iliary-antenna ANC. It verified cancellation perforance of the ethod that apply NLMS algorith on navigation signal, and also showed that the algorith has a lower daage on radio signal, but the suppression perforance for interference is reduced as sall INR of signal fro ain antenna. Therefore, we can iprove the perforance of the algorith by slightly increasing the INR of the iliary antenna. References [] GLONASS GLOBAL NAVIGATION SATTELITE SYSTEM GLONASS INTERFERENCE CONTROL DOCUMENT. Edition 5..28. [2] K.Rohlfs T.L.Wilson, Biwei Jiang. SHE DIAN TIAN WEN GONG JU [M]. Beijing Noral University Publishing House. Apri, 2. [3] Xuelei Chen, Huli Shi. Tianlai project: Radio detection of dark energy and square kiloeter array (SKA [J]. Physical. 42. Vol. 23 No., P 2-p. [4] G.Hellbourg,2, T.Trainini3,R.Weber,4, E.Moreau3, C.Capdessus4, A.J.Boonstra2. RFI SUBSPACE ESTIMATION TECHNIQUES FOR NEW GENERATION RADIOTELESCOPES 2th European Signal Processing Conference. 22. p2-p24. [5] Andre Gilloire, Herve Sizun. RFI itigation of GNSS signals for radio astronoy: probles and current techniques [J].Ann.Telecoun.29.p625-p638. [6] Chowdhury M.R. Shahriar. MITIGATION OF INTERFERENCE FROM IRIDIUMSATELLITES BY PARAMETRIC ESTIMATION AND SUBTRACTION.26. [7] ITU-R P.676-9 Recoendation: radio wave attenuate in atospheric gases. [8] Brian D. Jeffs and Karl F. Warnick Spatial Array Processing Methods for Radio Astronoy RFI Mitigation 23 IEEE. [9] ITU-R RA.226-: Techniques for itigation of radio frequency interference in radio astronoy. 23. P-P. [] Michael Eler, Brian D. Jeffs; Bea Forer Design Methods for Phase Array Feeds. International Workshop on Phased Array Antenna Systes for Radio Astronoy. May 4, 2. Provo, Utah, USA. 29