Experimental and Theoretical Evaluation of Interference Characteristics between 2.4-GHz ISM-band Wireless LANs

Transcription

1 Experimental and Theoretical Evaluation of Interference Characteristics between 2.4-GHz ISM-band Wireless LANs Kazuhiro Takaya, Yuji Maeda, and Nobuo Kuwabara NTT Multimedia Networks Laboratories 9-11 Midori-Cho, 3-Chome, Musashino-Shi, Tokyo Japan Abstract - Wireless LANs in the 2.4-GHz ISM-band create a new Electromagnetic Compatibility (EMC) problem. We investigated the interference characteristics between such wireless LANs in the case of identical systems, systems with different technical parameters for modulation and demodulation, and using a Gaussian noise source as a disturbance source. Experimental results show that higher throughput is obtained when adjacent wireless LANs use different systems, and that interference characteristics can be evaluated experimentally using a Gaussian noise source. Calculated BER characteristics for the interference agree with experimental measurements, indicating that this calculation method can be used for the design of the wireless LAN network to avoid interference. It is possible to construct an efficient wireless LAN network by combining different wireless LAN systems. INTRODUCTION Wireless LANs based on Spread Spectrum (SS) in the 2.4-GHz band are widely used in the world[l][2]. Since SS systems of several wireless LANs occupy the same frequency range, interference will occur between them. Although there are regulations to avoid interference in many radio communication systems, no such regulations have been specified for 2.4- GHz ISM-band wireless LANs[l]. This should be considered as a new Electromagnetic Compatibility (EMC) problem because transmitters of other wireless LANs represent a kind of disturbance source. Although interference between radio communication systems and other disturbance sources, such as a microwave oven or Gaussian noise, has been studied[3]-[sj, the interfercnce characteristics have not been investigated using real wire- less LANs. It is important to know these characteristics in order to design wireless LAN system in buildings or offices without the interference[6][7]. This paper presents measured interference characteristics of 2.4-GHz band wireless LANs and theoretically analyzes them. As disturbance sources, we used a wireless LAN of the same type, one with different technical parameters for modulation and demodulation, and a Gaussian noise source. WIRELESS LAN SYSTEM The wireless medium access control (MAC) and physical (PHY) specifications of the 2.4-GHz ISM-band wireless LAN are standardized by IEEE A wireless LAN system is composed of some base stations called access points and many personal stations. 2.4-GHz ISM-band wireless LAN systems use Direct Sequence Spread Spectrum (DSSS) or Frequency-Hopping Spread Spectrum (FHSS). We used two kinds of DSSS systems in our measurements. The technical parameters of the wireless LAN systems are summarized in Table 1. We cannot connect their different wireless LAN systems because their transmission frame format, modulation, and correlation process are different. Access to wireless LAN systems is Table 1, Technical parameters of DSSS wireless LAN. Operating frequency range: 2.47 l GHz Modulation : DQPSK Channel data rates: 2 Mbit/s Spreading sequence: Barker sequence (11 chips) Transmit power levels: 10 mw/mhz Access control: CSMAJCA 0-7SO3-3) 13--l/98/$ s IEEE SO

2 Fig. 1 Problem of wireless LAN systems. controlled by Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA). Generally, the range of a wireless LAN is about 50 m for indoor operation. As shown in Fig. 1, if the office walls are made of metal or absorber, the communication signal will penetrate through the walls and reach the neighboring office. Then, there will be an overlap area where signals of both system A and system B arrive and interference will occur between the wireless LAN systems. We investigated the interference characteristics experimentally and theoretically. EXPERIMENT The experimental setup to measure the interference characteristics in a semi-anechoic chamber is shown in Fig. 2. The locations of the disturbance source and the station were fixed, and the distance between the access point and the disturbance source was changed. Interference characteristics were measured for several conditions as shown in Table 2. Some technical parameters (e.g. intermediate frequency of the modulation process) of system B in Table 2 are different from those of system A because the systems came from different manufacturers. Therefore, the designs of those systems are different, and they cannot be connected. System C is an SS system constructed in our laboratory to measure Bit Error Rate (BER). access points of the two systems and a Gaussian noise source were used as a disturbance source. Throughput Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case I Fig. 2 Experimental setup to measure interference characteristics. Table 2. Measurement conditions. Wireless LAN system Svstem B System B System C System C Disturbance source System B Svstem A System B Gaussian noise Gaussian noise Case 8 System C System B A file (5 Mbites) was transmitted by File Transfer Protocol (FI P) from the station to the access point and throughput was measured to evaluate the interference characteristics. Throughput versus distance between the access point and the disturbance source is shown in Fig. 3. The throughput was normalized by that with no disturbance source. When the systems were the same types (cases 3 and 4), the throughput did not change with distance between access point and disturbance source. On the other hand, when the systems were different types (cases 1 and 2), the throughput deteriorated when the distance was less than 4 m, and the throughput for case 2 falled below that for case 4. The relationship between cases 2 and 4 is similar to that between cases 1 and 3. If the distance between the access point and the disturbance source is less than 2 m in case 1 and less than 6 m in case 2, the effect of interference from different system is bigger than that from the snme type of system. However, if the distance between the ac- Sl

3 ,.,..,...,,,,,..,.,.....I., I,... yo;!jjzej-[ Fig. 3 Throughput characteristics in case of the interference between two wireless LAN systems. cess point and the disturbance source is more than 4 m in case 1 and more than 6 m in case 2, the affect of the interference is smaller than that from the same system. The communication signal propagation characteristics of the access points are shown in Fig. 4. The electric field strength was measured at the height of 1.O m in a semi-anechoic chamber and normalized by the maximum value. The transmitting antenna of the access points were monopole antennas and the received antenna was a horn antenna. The results show the difference in transmitting power levels between the wireless LANs. This difference affects the results of cases 1 and 2 in Fig. 3. The relationships between throughput and ratio of desired signal to undesired signal (D/U ratio) are shown in Fig Fig. 5 Throughput characteristics versus D/U ratio. to evaluate the throughput characteristics without any difference in transmitting levels. As shown in Fig. 5, the throughput characteristics of cases 1 and 2 are proportional to the D/U ratio, and are higher than those of cases 3 and 4 when the D/U ratio was above 15 db. This means that it is possible to create an efficient wireless LAN network without any throughput deterioration by combining systems using different technical parameters and by considering the D/U ratio. White Gaussian Noise As a simple method for evaluating interference, white Gaussian noise is widely used [4][5]. We examined the valid- I I,llll 1 10 Distance (m) I -j Fig. 4 Communication signal propagation characteristics of access points. Fig.6 Throughput characteristics in case of Gaussinn noise and communication signal. 82

4 ity of the Gaussian approximation for the interference between wireless LAN systems. For system A, we compared two disturbance sources; the access point of system B (case 1) and a Gaussian noise source (case 5). The Gaussian noise used in the measurement had bandwidth of 60 MHz and the center frequency matched the wireless LAN system. Thus, the Gaussian noise had a wider bandwidth than the communication signal of the wireless LAN system. The relationships between throughput and D/U ratio are shown in Fig. 6. The differences in the throughput between cases 1 and 5 were less than 10 percent, which shows that it is possible to estimate the throughput characteristics between wireless LAN having different modulation parameters by using Gaussian noise. Bit Error Rate As throughput characteristics of wireless LAN systems include the effects of error correction, we used the Bit Error Rate (BER) to evaluate the interference characteristics without error correction. However, it is difficult to measure the BER of a real wireless LAN system. Therefore, we constructed an SS system (system C) to measure BER characteristics and to evaluate the interference on the physical layer. Modulation parameters and other technical parameters of System C are shown in Table 3. The BER of System C was measured when disturbance source was system A, system B, or a Gaussian noise source. The relationships between BER and D/U ratio are shown in Fig. 7. The results show that the communication signals of systems A and B had the same effect on the BER of system C, and that the BER for Gaussian noise (case 6) was higher than those for systems A and B for the same D/U ratio. This means that the communication signal modulated by the same spreading sequence has the same effect on BER characteristics of an SS system, and that hardly any SS systems have a noise margin against Gaussian noise. If the SS system is disturbed by the 10-l m ,, Fig. 7 BER characteristics of system C. communication signal modulated by the same spreading sequence, the BER characteristic can be regarded as being almost equal to that for the case of a Gaussian noise. THEORETICAL ANALYSIS We calculated the BER characteristics to investigate the mechanism of the interference, and examined the validity of the calculated results. Figure 8 shows the analytical model of the interference used for Digital Signal Processing (DSP) analysis. The DSP model of system A, system B, or a Gaussian noise source were used as the disturbance source. BER can be obtained as follows: First, the communication signal is modulated by the transmitter and a disturbance noise is added. Second, the synthetic signal demodulated by the receiver is compared with the transmission signal, and the levels of disturbance noise are changed to evaluate various D/U ratios. Table 3 Modulation parameters and other technical parameters of System C. Communication sipl Error + h 011t-b meter Operating frequency range: 2.47 l GHz Modulation : CSK Channel data rates: 200 Kbit/s Disturbance Spreading sequence: M sequence (127 chips) source Tmnsmit power levels: 10 mw/mhz Fig. 8 DSP analysis model for the interference. 83

5 - - Calculated (Cases 7,8) Measured (Case 7) Measured (Case 8) - Calculated (Case 6)..... Measured (Case 6) Fig. 9 Calculated and measured BER characteristics. Numerical Analysis Figure 9 shows the calculated results and the measured data for cases 6, 7, and 8 in Table 2. The calculated and measured BER are offset by only a few db. These differences arise because the disturbance noise was presented in a semi-anechoic chamber for the measurement, but given directly for the calculation. Therefore, The interference characteristics of wireless LANs can be calculated by using the DSP analysis, and we can simulate the BER characteristics using the model in Fig. 8. Theoretical Discussion Experimental and numerical results shown in Fig. 9 are considered theoretically as follows. Process gain G,, of the SS system is written as G,,=BWIR=T,,lTc=N (1) where BW is the bandwidth (Hz) of the spread spectrum, R is the information rate (bit/set), T, is the signal duration, Tc is the chip duration. and N is the spreading sequence length (T,=N* TJ [8]. The noise margin Mj is defined as M, is the residual robustness that the system has against a disturbance. (S/N),, is the signal-to-noise ratio of the receiver output and LsYs represents other losses[8]. Among practical systems; almost all SS systems have equal (Z#Qul and LsYs if they use the same radio frequency bandwidth. Therefore, the noise margin of an SS system is proportional tothe spreading sequence length. If the disturbance source is an SS system using the same spreading sequence, then the noise margin of the SS system will be smaller than when the disturbance source is one using a different spreading sequence. In,other words, the interference characteristics for a different system are similar to the interference characteristics for Gaussian noise. On the other hand, if the disturbance source is an SS system using a different spreading sequence, the noise margin of the SS system becomes effective. For example, as shown in Fig. 7, the BER characteristics for the interference between systems A and B is smaller than that for the interference caused by a Gaussian noise at the same D/U ratio. This means that there is a difference in the spreading sequence length, and which can approximated as 10. log (Mj,,, /M,,,) = 10.6 (db) (3) where Mj,, is the noise margin of the SS system using the 11 chips spreading sequence, and Mj,,, is that using the 127 chips spreading sequence. The difference of 10 db agrees with the difference between experimental and calculated results shown in Fig. 9. Interference between real wireless LAN systems We calculated the BER characteristics for the interference between real wireless LAN systems by using the DSP analysis. The interference characteristics relate to the signal collision timing because the real wireless LANs have the same spreading sequence. Therefore, BER characteristics were calculated by changing the initial offset r of undesired signal. The calculated interference characteristics between wireless LAN systems are shown in Fig. 10, where r is the time delay of the undesired signal for the desired signal and T< is the chip duration. Here, the BER?haracteristics were changed in steps of the difference in the collision timing between desired and undes- (2)

6 that the throughput of a LAN system was improved by combining systems using different technical parameters for modulation and demodulation. The calculated BER characteristics by DSP analysis agreed with the measured data, so the interference characteristics can be calculated by the mode1 used in this paper. In the future, we will calculate the throughput characteristics for the interference and design wireless LAN network that avoids such interference. REFERENCES I Fig. 10 Calculated interference between wireless LAN systems. ired signals. Assume that the collision between desired and undesired signals is caused randomly in the interference between real wireless LANs. Then the probability is l/l 1 if r is less than the chip duration Tc, because the spreading sequence is the length of 11 chips. Accordingly, average BER Pay, is approximated as where PI and P,are the BER of a wireless LAN system when r is above and below Tc, respectively. From Equation (4), the BER for the interference between wireless LAN systems Pa, is calculated as shown in Fig. 10. The BER characteristics are similar to those for Gaussian noise, and this result agrees with the experimental result in Fig. 6. Therefore, the interference characteristics between wireless LAN systems having different technical parameters for the modulation and demodulation can be evaluated by measuring with a Gaussian noise. Moreover, it is possible to design the wireless LAN network without interference by using this interference characteristic analysis and radio signal propagation analysis. CONCLUSIONS (4) [I] IEEE standard , Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, IEEE, Inc., 345 East 47th Street, New York, NY USA, [2] B. Tuch, Development of WaveLAN, an ISM Band Wireless LAN, AT&T TECHNICAL JOURNAL, July/August, [3] S. Miyamoto, Y. Yamanaka, T. Shinozuka, and N. Morinaga, A Study on the Effect of Microwave Oven Interference to the Performance of Digital Radio Communications Systems, (in Japanese) Trans. IEICE, vol. J79-BII, pp , November, [4] M. B. Pursley, Performance Evaluation for Phase-Coded Spread-Spectrum Multiple-Access Communication -Part I: system analysis, IEEE Trans. Commun., vol. COM-25, August, [5] J. Lehnert, and M. B. Pursley, Error Probabilities for Binary Direct-Sequence Spread-Spectrum Communications with Random Signature Sequence, IEEE Trans. Commun., vol. COM-35, No. 1, January, [6] H. Komori, and Y. Konishi, Wide Band Electromagnetic Wave Absorber with Thin Magnetic Layers, leee Trans. Broadcast., vol. 40, No.4, December, [7] S. Y. Seidel, and T. S. Rappaport, 914 MHz Path Loss Prediction Models for Indoor Wireless Communication in Multifloored Buildings, IEEE Trans. ACCESS POINT., ~01.40, No.2, February, [8] R. L. Pickholtz, D. L. Schilling, L. B. Milstein, Theory of Spread-Spectrum Communications-A Tutorial, IEEE Trans. Commun., vol. COM-30, pp , May, The inte;ference characteristics between different wireless LAN systems were investigated, and the results showed