Co-Channel Interference Analysis of Point to Point mm-wave Radio Links

Transcription

1 Co-Channel Interference Analysis of Point to Point mm-wave Radio Links Xu Mingdong and Peter Nuechter Research and Advanced Development, HUBER+SUHNER AG, CH-8330 Pfäffikon ZH, Switzerland Now with Communications Laboratory, Dresden University of Technology, 0062 Dresden,Germany Abstract - Point to point mm-wave radio links are being increasingly deployed because of the expansion of commercial wireless services. Providers of wireless services continue to demand systems with higher data rate and higher carrier frequency. In order to fulfill the demand unlicensed mm-wave bands have been investigated for fixed point-to-point outdoor radio applications. An analytical method is developed in this paper to study co-channel inteference of mm-wave devices in a given area. An optimum number of mmwave devices in a certain area may be drawn from the analysis. MatLab programs are also produced to simulate the approach. The simulation results are plotted to give a clear view. It is expected that the interference analysis would help to implement the mm-wave point to point communication systems in outdoor environment. Keywords - point to point radio link, unlicensed mm-wave, co-channel interference, carrier to interference ratio (C/I), optimum link number, spatial efficiency I. INTRODUCTION There exists an ever increasing supply of, and demand for, broadband multimedia applications calling for an ever increasing capacity of point to point wireless communications. In order to achieve this goal, an obvious solution is to resort to the mm-wave band, where bandwidth is abundantly available. The mm-wave band can fulfill this demand with an extremely high transmitting data rate of as much as Gbit/s. In 200 and 2003, FCC released mm-wave frequency bands 57-64GHz and 7/8GHz pair from license regulation [2][3]. These frequency bands have some unique characteristics, which contribute to both benefits and disadvantages. Firstly, oxygen molecules absorb electromagnetic wave energy around 60GHz dramatically (which reaches summit around 6dB per km at 60GHz); secondly, precipitation usually dominates operation link distance due to high sensitivity to rain attenuation; finally, beam width of antenna that operates in these bands is relatively narrow. Although attenuation due to oxygen and rain limits the link coverage; on the other side, accompanied with narrow beamwidth, this high attenuation decreases interference between neighboring links, which makes high link density possible in a certain area. Because of narrow beamwidth, it can also make advantage to use these frequency bands for point to point radio communications []. In order to estimate the attenuation of oxygen an approximate method based on ITU-R recommendation [7] is adopted in this paper. This method is a simplified one for frequency range - 350GHz, and from sea level to an altitude of 5km. For terrestrial paths, or for slightly inclined paths close to the ground, the attenuation rate due to oxygen γ o (db/km), may be expressed as a function of operating frequency f (GHz), atmospheric temperature t ( o C) and atmospheric pressure p (hp a): γ o = γ(f, t, p) () To calculate the loss due to rain an empirical approach based on the approximate relationship between attenuation rate A (db/km) and rain rate R (mm/hour) is followed. In practice, this method has tended to be used most often and with a good result. A = κr a (2) where κ and a are frequency-dependent coefficients and can be obtained from ITU-R recommendation [8]. From ITU-R recommendation [7], it can also be found that in frequency range from 50GHz to 00GHz, attenuation due to water vapor in atmosphere is not greater than 0.4 db/km. It may also be seen that only above 00GHz, the attenuation due to fog and cloud is significant [9]. Because of the short coverage of mmwave applications and relatively small path attenuation compared to free space loss and rain attenuation, attenuation due to water vapor, fog and cloud may be ignored in this paper without a significant error. Interferences from neighboring broadband links, especially the one operating in same frequency band, impair link quality. It causes noise-like interference power to victim receivers and raises their noise floor levels [5]. Usually C/I (Carrier to Interference Ratio) has to fulfill one criteria in any case: C/I greater than T/I (Threshold to Interference Ratio). Threshold to Interference ratio (T/I) is designated as the ratio between static threshold point of carrier level, which can fulfill certain link fidelity, for example, 0 6 BER (Bit Error Rate), and the interference level that would cause db degradation to the threshold of the protected receiver [6]. T/I can be obtained in the following way: first adjust carrier to a threshold level (T)

2 to achieve required BER; then increase this carrier level db; inject interference increasingly until the required BER is recovered, record the interference level (I) at this time; finally draw the ratio of T/I. Value of T/I is roughly 6 db greater than the theoretical threshold value of carrier to noise ratio (C/N) if the interferer produces a thermal-noise-like interference with a bandwidth less than or equal to that of the desired signal. To remain the availability of the affected receiver acceptable despite the interference, for the range of carrier power levels between the clear-air (unfaded) value and the fullyfaded static threshold value, in no case shall interference cause C/I to be less than T/I, unless it can be shown that the availability would still be acceptable under the interference [6]. The rest of the paper is organized as follows: in Part II an analytical approach to investigate co-channel interference in a given area is discussed. In part III simulations based on the analysis are carried out and results are plotted. Finally some conclusions are drawn in part IV. Fig.. Sketch of Scenario II. INTERFERENCE ANALYSIS This section aims to theoretically investigate intereference of unlicensed mm-wave devices with co-channel interferers existing. Then the optimum number of links in a given area can be drawn from the analysis results. In mm-wave LOS (Line of Sight) propagation, multipath fading, which is primary issue in lower frequency applications, is not significant; attenuations due to precipitation and atmospheric oxygen are more substantial. In order to investigate the mm-wave communication system a link budget has to been developed. Only co-channel intereference is investigated in paper; all links are assumed to have same system parameters such as link distance, transmitted power, central frequency, bandwidth and antenna radiation pattern. ATPC (Automatic Transmitter Power Control) is adopted in all links. Here we study a circular area with radius of R. As we are investigating the worst case, only interference suffered by the device at the center of the circular area is considered. Also because of the general nature of this investigation, a minimum distance is assumed, which defines a small circular area around the victim receiver to be kept free from interferers. Deployment of devices in such a close neighborhood requires a special co-cite design, which is out of scope of this paper. Moreover, in order to fulfill far-field condition (r 2D 2 /λ, D is the maximum dimension of antenna), a reasonable value of around 0 meters of is recommended. Since main beams of an existing mm-wave device and a new device should also try to be avoided to point to each other in deploying a new device, the interferers would be distributed uniformly in the fan-shaped area with a radius of R and angles from to 2π, excluding a small circular area (interferer clearance area) of radius. is the angle from the boresight of receiving antennas (Figure. ) and typically is chosen to exceed antenna main beam. The probability density functions of a random interfering device in this fanshaped area in terms of distance and degree are given respectively as: p(r) = p(α) = { 2r/(R 2 R 2 min ) r R 0 others { /(2π 2α0 ) α 2π 0 others (3) (4) The interference level from an interferer at a distance of r may be obtained as: I(r, α, α 2 ) = P t G(α )G(α 2 ) L(r) (5) where P t is transmitted power, G(α ) and G(α 2 ) are gains of interfering and victim antennas as functions of angles between the interfering path to each main beam respectively, L(r) is the loss of interfering signal (figure. ), which includes free space loss and oxygen attenuation and is expressed as: L(r) = L free (r) L oxygen (r) (6) where L free is free space loss and L oxygen is path loss due to oxygen []. The expectation value of a randomly-located cochannel interferer in this fan-shaped area is: I = I(r, α, α 2 )p(r)p(α )p(α 2 )

3 dα 2 dα dr (7) where p(r), p(α ) and p(α 2 ) are probability density functions represented as equation (3) and (4). Substitution of equation (5) into (7) results in: I = P t p(r) L(r) dr G(α )p(α )dα G(α 2 )p(α 2 ) dα 2 (8) The expectation value of interfering antenna gain and victim antenna gain (they are same) along the interfering path is given as: G e = G(α) p(α) dα (9) where G(α) is the antenna gain as a function of path angle. This integral excludes the range from to to avoid antenna main beam. Please note that the discussion above is based on the assumption that all links are parallel to the ground and in same height. However if only 2-D deployment is considered, would restrict the available link number. So deployment in 3-D space is also taken into consideration in our paper, the equation (9) can be used as an upper bound of mean gain when heights of mm-wave terminals are slightly different and links have small tilt angles, in which cases link capacity may be increased dramatically. Here we define mean path loss of a randomlydistributed interferer : = p(r) dr (0) L(r) Then the expectation value of interfering signal level from the random interferer is expressed as: I = P tg 2 e () The interference level from k randomly-distributed interferers is: I k = k I (2) The interference level should not cause the C/I (Carrier to Interference Ratio) to be less than T/I (Threshold to Interference Ratio). Since T/I is approximately equal to C/N+6dB, (C/N is threshold Carrier to Noise ratio to fulfill certain link quality requirement, for instance 3.5dB for QPSK and 0 6 BER), total interference should not be greater than N-6dBm. The optimum number of devices k opt in this area, which makes C/I equal T/I, may be drawn from the following equation: We may also find that: k opt P t G 2 e/ = N/4 (3) k opt = N 4P t G 2 e = k 0T 0 BF 4P t G 2 e (4) where k 0 is Boltzmann s constant, T 0 is the standard temperature 290K. F is receiver s noise figure. B is receiver bandwidth. Equation (4) simply states, that systems with higher noise level, larger path attenuation, lower transmitted power and lower interference due to antenna sidelobes allow a greater number of interferers. This confirms the basic concept of mm-wave communications at high atmospheric attenuation with pencil beam antennas. Furthermore, when ATPC is adopted, P t should have an equation as following: P t G 2 0 L d = (C/N) N = (C/N) k 0 T 0 BF (5) where G 0 is the maximum gain of transmitting and receiving antennas, L d is the path loss of the desired signal and C/N is threshold carrier to noise ratio. When we combine equation (4) and (5), we may get: We also know that: k opt = 4 G2 0 G 2 Le e L d C/N (6) C N = E b N 0 η B (7) where E b /N 0 is the threshold ratio of signal energy per bit to noise power density required at the receiver input for a certain probability of error (say 0 6 ); η B is the spectral efficiency, which is defined as the data rate per bandwidth [4]. E b /N 0 is introduced here because it is used more often than C/N. From equation (6) and (7), a new equation as following may be drawn: k opt = 4 G2 0 G 2 Le (8) e L d E b /N 0 η B From the equation above, we may find that the optimum link number in a given area doesn t depend on system bandwidth when ATPC is adopted in all links, since with bandwidth raising, transmitted power also increases proportionally. From later simulations, we also find that, in 60GHz band the optimum link number is approximately proportional to the area we investigate when radius R is large enough (say greater than km). The reason for this phenomenon is that only mm-wave devices inside a circular area of radius R=km play a dominant role in interference, since around 98% of total signal energy is absorded by oxygen within first km distance at 60GHz. We define a concept spatial efficiency the optimum link number per square km, which is left side of the following equation. k opt S = 4πR 2 G2 0 G 2 Le e L d E b /N 0 η B (9) According equation (0), (6) and (), may be expressed as: = 2r/(R 2 Rmin 2 ) (4πr/λ) 2 c r dr (20) λ is the wavelength in meter and c is 0 γ0/0000, where γ 0 is oxygen attenuation rate expressed in db/km. When substituting into equation (9), considering

4 that is much smaller than R, spatial efficiency becomes: k opt S 2π λ 2 G2 0 G 2 e L d E b /N 0 η B (2) dr rc r When R is large enough, rc becomes r incomplete gamma function - Γ(0, ln c) [2]; therefore, spatial efficiency is not dependent on R and approximately becomes a constant. With system parameters as Table. and a reference antenna radiation pattern from ITU-R (see Part III), Γ(0, ln c) is obtained as and G e as 4.30dB. These result in a spatial efficiency k opt /S of 08/km 2 and an optimum link number of 339 at R=km, which are in accordance with the later simulations. In this paper, only clear sky condidion is considered. Actually, both desired link and interfering link suffer precipitation attenuation; since most interferers are supposed to locate at a distance longer than desired link, attenuation of interfering signal may be larger than that of desired signal. Therefore our analysis results are still available in rain condition in most cases. Furthermore the obstacles such as vegetation and buildings along interfering path are not included in our analysis. Because these obstacles may block interfering signal and therefore reduce interference level, the link capacity in a given area may be greater than our simulation results in practice. dr III. SIMULATION AND RESULTS MatLab programs are developed to simulate the analysis. QPSK is chosen, which gives threshold E b /N 0 of 0.5dB at 0 6 BER and theoretical η B of 3dB; a reference antenna radiation pattern from ITU-R recommendation [0] is adopted, where the gain along each direction may be expressed in terms of direction angle, operation frequency, diameter of antenna and maximum gain along main beam. First we investigate how the link distance behaves in interference level. Except link distance, other parameters are same as Table.. The link distance d is set to 00, 200, 300 meters respectively. The optimum link number can be given at the cross point of C/I and T/I curves. It may be found that longer link distance produces less link numbers because it means higher transmitting power which therefore introduces higher interference level. Furthermore, with input parameters as Table., we change, which determine the value of mean gain (G e ), and plot the simulation result as Figure.3. The simulation result clearly shows that narrow decreases link number greatly. Nevertheless wide angle means difficult deployment of mm-wave devices if the number of devices in a given area becomes large. In practice, different elevation angles or heights of links may mitigate this difficulty. Finally, since mean path loss ( ) of a random interferer depends on the values of R and, MatLab programs are produced to simulate Fig. 2. TABLE I INPUT PARAMETERS Parameters Name Values Modulation Scheme QPSK Link Distance d (meter) 200 System Bandwidth (MHz) 500 Angle of (degree) 4 Radius of Circular Area R (m) 000 Radius of Interferer Clearance Area (m) 0 Operation Frequency (GHz) 60 Maximu Gain of Antenna (db) 25 Diameter of Antenna (m) 0.5 Atmospheric Temperature ( 0 C) 5 Atmospheric Pressure (hpa) 03 Receiver s Noise figure (db) 0 C/I versus number of devices at different d their effects on optimum link number. With other input parameters in Table., simulation results are as Figure.4 and 5. From the simulation, it is clearly shown that the optimum link number may decrease dramatically with approaching 0, which can also be proved by incomplete gamma function; however when becomes large (say 0m), the link number is not as sensive as before. Figure.5 proves the conclusion drawn in previous part that optimum link number is approximately proportional to the area. IV. CONCLUSIONS Co-channel interference from neighboring links limits the capacity of mm-wave links in a given area. The interference behavior is investigated such that the opimum link number in an area may be drawn approximately from the analysis. From our analysis, it is found that optimum link number in a large area is proportional to the area we investigate. According to this conlusion, a new concept spatial efficiency, is introduced to define the optimum link density. Simulations based on this approach are carried out to give a clear view of the impact of some important parameters.

5 V. ACKNOWLEDGEMENTS The authors would like to thank Professor Claes Beckman of University of Gävle in Sweden for his support and kind help. REFERENCES Fig. 3. C/I versus number of devices at different [] Xu Mingdong, A.Rahim, P. Nuechter Investigation of System Parameters of mm-wave Point to Point Radios IEEE IFIP ICI 2005 Bishkek, Sep [2] FCC Regulations Part 5 [3] FCC Regulations Part 0 [4] R.E. Ziemer, R.L.Peterson Introduction to Digital Communication Second Edition, Prentice Hall, 2004 [5] TIA/EIA ZSB0-F Telecommunication System Bulletin- Interference Criteria for Microwave Systems, Jan,2004 [6] WCA-PCG Path Coordination Guide for the 7-76 and 8-86GHz Millimeter Wave Bands Jan [7] Recommendation ITU-R P Attenuation by atmospheric gases, 200 [8] Recommendation ITU-R P Specification attenuation model for rain for use in predivtion methods, 2003 [9] Recommendation ITU-R P Attenuation due to clouds and fog, 999 [0] Recommendation ITU-R F Reference radiation patterns for fixed wireless system antennas for use in coordination studies and interference assessment in the frequency range 00 MHz to about 70GHz 2003 [] ETSI TR v.2. Fixed Radio Systems; Point-to point equipment; Derivation of receiver interference parameters useful for planning fixed service point-to-point systems operating different equipment classes and/or capacities [2] Milton Abramowitz and Iren A. Stegun Handbook of Mathematical Functions, With Formulas, Graphs, and Mathematical Tables June 974 Fig. 4. C/I versus number of devices at different Fig. 5. C/I versus number of devices at different R