Está en la página 1de 5

CSOR - Carrier Sensing On Reception

Tim Esemann
Lbeck University of Applied Sciences Mnkhofer Weg 239 Lbeck, Germany

Horst Hellbrck
Lbeck University of Applied Sciences Mnkhofer Weg 239 Lbeck, Germany

esemann@fh-luebeck.de

hellbrueck@fh-luebeck.de

ABSTRACT
Since the 1990s the number of wireless devices increases and new areas of applications evolved. Therefore, frequency spectrum has become a scarce resource with no free frequencies left all over the world and interference between transmissions sharing the same frequency band is started to become one of the major problems in wireless transmission. The ISM bands become crowded with various standards sharing the same frequency band. One solution to the problem is to use frequency bands that are rarely used by the licensed users like TV channels, where in some region specic channels are not used at all. As a result, we need to develop adaptive systems that search for unused spectrum, use it as long as the band is free and shift to other frequency bands if there is a risk to interfere with a primary user. Such systems being aware of their radio environment are called Cognitive Radios. To search for unused frequencies and detect primary users approaches that listen before talk or ecient carrier/spectrum sensing algorithms have been presented in the past. These mechanisms are incomplete and one of the drawbacks of todays wireless transmissions is that communication partners do not detect interference reliable during an ongoing transmission. In this paper we suggest a cross layer approach Carrier Sense on Reception (CSOR) that extends the functionality of the physical layer of a transceiver to be able to detect interference while receiving data. We introduce the idea, describe the concept and give rst evaluation results as a proof of concept based on real measurements.

Keywords
Interference Mitigation, Spectrum Sensing, Cognitive Radio

1.

INTRODUCTION

Categories and Subject Descriptors


C.2 [Computer Systems Organization]: Computer Communication Networks

General Terms
Algorithms, Design, Experimentation, Measurement

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for prot or commercial advantage and that copies bear this notice and the full citation on the rst page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specic permission and/or a fee. CogART 11, October 26-29, Barcelona, Spain Copyright 2011 ACM ISBN 978-1-4503-0912-7/11/10 ...$10.00.

20 years ago rst wireless transmission systems became digital. Since then this eld has developed in an enormous speed. Today, digital wireless systems cover all elds of application reaching from short range personal area communication via Bluetooth IEEE 802.15 to LANs IEEE 802.11 up to metropolitan area network size with IEEE 802.16 and nally in wide area applications like GSM, 3G/4G Systems. Many standards including IEEE 802.11/802.15 operate in so called ISM-Bands reserved for industry, science and medical applications. Thereby, the ISM bands become crowded and interference gets a severe reliability problem. Systems and products tested in the lab or in special environment are not guaranteed to work as in their intended workplaces especially in the future when the density of interference continues to grow. A eld of application with high potential to replace cables by wireless communication is medicine or e-Health. Wherever humans are involved cables hinder processes and create complications especially at intensive care units or during surgery. Therefore, many medical companies start to introduce products in the market based on todays wireless technologies or proprietary solutions. However, medical applications require high reliability for the communication between wireless partners. In this paper we will rst provide a summary of investigations including measurements to demonstrate that existing devices do not provide appropriate measures against interference from other systems. Next we will present mechanisms recently developed for cognitive radios to detect and avoid interference. We will demonstrate that these mechanisms are incomplete and identify room for improvement. One of the drawbacks of todays wireless transmissions is that communication partners are insensitive for detecting interference during transmission. We will introduce our cross layer approach for a so called Carrier Sense on Reception (CSOR) that enables sensing of interference while actually receiving data. To the best of our knowledge this is the rst time that such an approach has been presented. We have nished a rst implementation of CSORs basic functionality and present rst results in this paper. The rest of the paper is organized as follows. Section 2 will present related work and an analysis and investigation of the shortcomings. The concept will be presented in Section 3 followed by a description of the implementation in

Section 4. In Section 5 we will evaluate the performance of the implementation. We will conclude the paper with a summary and a future outlook.

called CSOR concept will be presented in the next section.

3.

CSOR CONCEPT

2.

RELATED WORK

Wireless standards for unlicensed ISM-band introduced in the previous section already provide some schemes for interference mitigation. These schemes can be categorized into two classes, Spread Spectrum and Spectrum Sensing. Direct Sequence Spread Spectrum (DSSS) mitigates interference by spreading the signal in the frequency domain. It is used by IEEE standards 802.15.4 [2] and 802.11 [1]. Bluetooth [3] implements Frequency Hopping Spread Spectrum (FHSS). Both schemes can tolerate some interference by other systems. The other class, Spectrum Sensing, as the name suggests senses the spectrum prior the transmission to avoid collision or interference with other wireless transmissions. IEEE 802.15.4 and 802.11 perform this in a simple scheme called Carrier Sensing or Clear Channel Assessment (CCA). For a reliable transmission all wireless standards use acknowledgement to conrm a correct reception. In the past numerous investigations and measurements were conducted to evaluate the performance of the interference mitigation schemes of the wireless standards [10, 15, 16, 5, 7]. All papers agreed that increasing occupation of the limited frequency bandwidth decreases signicantly the throughput due to transmission errors caused by interference. In [7] we have shown in previous work that even a single IEEE 802.11 transmission can cause a packet loss for a Bluetooth transfer larger than 10%. In the eld of cognitive radios new approaches for interference mitigation and spectrum sensing were developed. Ycek and Arslan give a good overview of the suggested u spectrum sensing algorithms in [19]. The performance of detection and identication of other parallel transmissions is improved extensively. Some examples for new algorithms are cyclostationary feature detection [9, 17, 14] and wavelet based sensing [18]. However, these techniques cannot detect and identify interference occuring during transmission. In our new approach we will extend the concept of so called soft bit information. Soft bit information are exploited for error correction [12] and reliable wireless transmission on multipath and fading channels [6]. In the concept of soft bits the decoder adds a probability in the sense of maximum likelihood to each decoded bit. In our concept we do not exploit soft information to detect or correct errors but to detect and in a future step identify interference. Therefore, we need to investigate series of soft bits and do not consider single soft bit information. From this section we summarize that a major drawback of state of the art in wireless communication is that during transmission and reception interference cannot be detected. That applies to both communication partners: sender and receiver. In our concept we tackle these shortcomings. For the transmitter there is no chance to detect any interference as a consequence of the dierent power levels for transmitting and receiving which is in many cases more than 70dB. The challenge for the wireless transmitter is comparable for a human to hear someone whispering near to you while shouting to a far distant receiver at the same time. However, receivers that are insensitive today can be extended to detect interfering signals by analyzing the received signal in more depth than in the state of the art. This so

Todays receivers work in a chain of blocks (RF-Frontend, Demodulator and Decoder). The received signal is amplied, transferred into baseband and demodulated there. The result of this process is a digital signal that we call digital samples for the rest of this paper. Finally the decision device and a demapper decide whether Bit 0 or Bit 1 is received. This process is called decoding. Todays demodulators are so called I/Q Demodulators and we illustrate the basic principle in Figure 1.

I-Channel Filter recv(t) 90 Filter


LO

demod(t) F

Q-Channel

Figure 1: I/Q Demodulator The amplied signal recv(t) is demodulated into two signals I and Q. A very common illustration for I and Q samples is the constellation diagram. Figure 2 shows two constellation diagrams for a MSK modulation that we will use for the evaluation later in this paper.

(a) Without (ideal)

Interference (b) With Interference at the Receiver side

Figure 2: Constellation Diagram for MSK Modulation The rst diagram shows the ideal constellation points whereas the second diagram shows resulting constellation points for a receiver when some kind of interfering signal adds to the original signal. Interference in this context means noise or another parallel transmission. To decode bits the receiver decides for each pair of received I and Q samples in which area (separated by the diagonal lines) the resulting measured constellation point resides. Additionally, in the right lower corner in Figure 2 (b) a bit error occurs where 11 has been sent and an I and Q sample for 10 has been demodulated. In general, I and Q signals are combined by a function F resulting in a digital signal digital samples that will be fed to the decision device and demapper for decoding the bits. In the rst step of this work we take a closer look at the signals before the decoder. Therefore, we compare two in-

put signals for the decision device in the time domain as illustrated in Figure 3 and Figure 4. Figure 3 shows a clean received signal whereas Figure 4 illustrates the case of an additional interfering signal. The main assumption in this work is that although demodulation can be a nonlinear transformation (depending on the applied modulation scheme), interfering signals will create an overlaying pattern that can be distinguished from noise. Although the pattern looks random in Figure 4 with help of appropriate signal processing it can be distinguished from noise. In the future we strive for more detailed identication of interference.
20 15 10 Amplitude 5 0 -5 -10 -15 -20 0 100 200 Time 300 400 500

4.

IMPLEMENTATION

For the rst basic evaluation of the approach we decided to implement a CSOR-receiver for Minimum Shift Keying MSK due to the following reasons. MSK is used in various IEEE standards like Bluetooth [3] and 802.15.4 [2] (in a modied form) that are especially of interest to our chosen eld: medical application. Bluetooth especially was recommended by the Continua Healthcare Alliance [4] and we have shown that there are some severe shortcomings to be solved which can be improved by CSOR. The chosen platform for implementation is GNU Radio [11] in combination with Universal Software Radio Peripheral USRP 2 [8] by the company Ettus Research LLC. The advantage is that Software Dened Radio (SDR) like GNU Radio allow us to implement the major signal processing part of the receiver in software. Additionally, the hardware (USRPs) is comparably cheap which is important as we need three of them at a time for the evaluation.
Receiver in GNU Radio Rx Signal RF Frontend I/Q Demod Digital Samples Data File Decision Device Data

CSOR in Matlab Chi 2 Test CSOR Result

Figure 3: Digital Samples (without Interference)

Figure 5: MSK Receiver with CSOR


20 15 10 Amplitude 5 0 -5 -10 -15 -20 0 100 200 Time 300 400 500

Figure 4: Digital Samples (with Interference)

Our approach is inspired by the soft bit method that was already introduced in Section 2. However, our target is completely dierent. We process the digital I and Q samples in order to detect interference. Therefore, we extend the receiver by a processing block (see Figure 5 for details) that provides the functionality of Carrier Sensing on Reception. To detect interference we perform statistical analysis with the received I and Q samples in order to identify patterns. Single I and Q sample values do not reveal any useful information. Only a sequence of digital samples allows interpretation of the cause of interference by performing statistical analysis. As a wireless frame consists of hundreds of bits, we always have a large sequence of digital samples at hand for further analysis. We will present more details of our implementation in the next section.

The MSK receiver illustrated in Figure 5 consists of the I/Q demodulator as explained in Section 3 together with several Phase Lock Loops (PLLs) and synchronization blocks that are not displayed here. In this stage of the work the detection of interference is not yet integrated into GNU Radio, so statistical analysis is done oine. Therefore, the existing implementation of the receiver stores I and Q samples in a le for further processing with Matlab. Our goal in this paper is to distinguish between random variations through noise and other signal patterns that result from carriers of other transmissions. To identify noise or randomness there are established methods and tests developed in the eld of evaluation of random noise or random number generators that we will apply here. Here, a simple decision if the signal is considered random or not is sucient although we strive for a more detailed classication of interference in the future of this project. We decided in this work to use the well-known Chi-SquareTest [13] that is established and ecient. We need to calibrate the Chi-Square-Test for this purpose in the following steps: 1. Chi-Square-Test needs to be parameterized for the expected values primarily the range of the digital samples and the number of intervals. This needs careful choice, so that the result achieves the required precision. 2. In the case of noise as interference the observed values follow a Gaussian distribution. Therefore we will calculate the theoretical expected frequency of occurrence in the interval with a Gaussian distribution as a hypothesis.

Frequency of Occurence

The described calibration process will be applied in the evaluation in Section 5.

5 10 15 Magnitude of I and Q

20

Frequency of Occurence

3. Perform the Chi-Square-Test with the magnitude of the received I and Q samples and compare the result of this test with a threshold to decide between interfering signal and noise.

70 60 50 40 30 20 10 0

70 60 50 40 30 20 10 0

5 10 15 Magnitude of I and Q

20

5.

EVALUATION

(a) Noise, SNR=12dB

The evaluation setup is illustrated in Figure 6 and consists of three USRPs. The USRPs are shown as single blocks in the gure. The rst USRP works as a sender. The second USRP serves as source of interference.
Sender Interference Receiver Data File Chi2 Test Matlab

(b) MSK SIR=12dB

Interference,

Figure 8: Histogram of the Magnitudes of I and Q for one selected Symbol

Figure 6: Block Diagram of Evaluation Setup

Both source signals are added and propagate to the third USRP where we implemented our CSOR-Approach in the MSK-Receiver extended by CSOR components. Before the quantitative performance evaluation we present in Figure 7 two constellation diagrams for the symbol 11. In the rst diagram background noise with a signal to noise ratio (SNR) of 12dB is added to the original sender. In the second diagram an MSK signal with an oset-frequency of 11kHz to the original sender with a signal to interference ratio (SIR) of 12dB is added.
10 10 Q Q

5 I 0 0 -5 5 10 15 20

5 I 0 0 -5 5 10 15 20

In order to validate that CSOR can distinguish between noise and an interfering signal and to evaluate the performance of the algorithm we run two series of experiments. One series with a frame size set according to 1-slot Bluetooth packets which are 366 bits the other series with and 3-slot Bluetooth packets that are 1622 bits accordingly. We expect that with longer series of bits the CSOR will detect the interference more accurately. In both series of measurements we add an interfering signal with SIR of 6dB to the original sender. In the next step we vary the background noise level from SNR of 6dB to 38dB. These measurements were repeated with interfering signals of dierent strength SIR of 12dB and 20dB. For each test we transmitted 300 frames and performed a Chi-Square Test with the received digital samples and a Gaussian distribution hypothesis. The threshold to decide between an interfering signal and noise has been taken from a table for quantiles of a Chi-Squared distribution. According to the given degree of freedom the threshold has been set to 40 for 1622 bits and to 35 for 366 bits. As a result of these measurements we present the observed detection probability for an interfering signal in Figure 9 for 366 bit frame length and Figure 10 for 1622 bit frame length. Both gures show three curves with the same SIR and the x axis gives the SNR in dB. For each curve the signal level of the MSK interfering signal is xed and the noise oor is decreased successively.
100 Detection Probability 80 60 40 20 0

-10

-10

(a) Noise, SNR=12dB

(b) MSK SIR=12dB

Interference,

Figure 7: Constellation Diagram of one Symbol As one can see with background noise the distribution of samples is around a center with a high density at this point which is typical for a normal distribution. With the MSK interfering signal however the distribution is more uniform in a circular region which clearly identies the dierence between both kinds of interference. Figure 8 shows the histograms of the magnitudes for this set of I and Q samples. It can be seen that the distribution with noise follows the typical Gaussian curve. With an interfering MSK signal the shape of the curve is dierent. Consequently, values are not normal distributed. Therefore, a Chi-Square Test with these two dierent sets of I and Q samples will result in two dierent values.

10

15

20 SNR

25

30

35

SIR=6dB

SIR=12dB

SIR=20dB

Figure 9: Detection Probability with 366 Bits

At the beginning of each curve the interfering signal is completely covered by the noise oor. The higher the SNR gets the better is the detection probability. The results show

100 Detection Probability 80 60 40 20 0

7.

REFERENCES

10

15

20 SNR

25

30

35

SIR=6dB

SIR=12dB

SIR=20dB

Figure 10: Detection Probability with 1622 Bits

that with a frame length of 1622 samples an interfering signal can be distinguished from an increased background noise level. As soon as the SNR is larger than the SIR the detection probability is larger than 75% for all three curves. For a SIR of 6dB the detection probability reaches 100% directly. With a frame length of just 366 samples the detection probability is larger than 75% if the interfering signal is approximately 16dB larger than the noise level. These measurement results demonstrate that interfering signal can be clearly distinguished by noise with our CSOR approach. These measurement results are very promising and although more a proof of concept but keep in mind based on real measured data demonstrate the eectiveness of this approach.

6.

CONCLUSION AND FUTURE WORK

In this paper we have motivated the need for further sensing mechanisms for cognitive radios. We presented our approach CSOR (Carrier Sense on Reception) and described a rst proof of concept implementation of the well-established modulation scheme MSK extended by CSOR components. The rst evaluation results are very promising. CSOR detects interfering MSK transmissions even with a notable noise level and out of tune with the original transmission. The detection sensitivity is reached when the interfering transmission reaches the noise level. With 6dB above the noise level the interference can be detected with a success rate of at least 75% with a frame length of 1622 bits. For a frame length of 366 bits the interference needs to be 16dB above the noise level to be detected with same probability. In the future, we strive for a detailed classication of the interfering signal to identify the primary user and distinguish from other sources of interference. For a completion of the evaluation we will nish our implementation of the whole detection processing in GNU Radio and measure the computational overhead. Therefore, we will consider the computational complexity of the detection algorithm and optimize the CSOR implementation as a tradeo between computational overhead and identication quality.

Acknowledgment
This work has been supported by the Federal Ministry of Education and Research of Germany: Frderkennzeichen o 17N3809, SoFT.

[1] Wireless LAN medium access control (MAC) and physical layer (PHY) specication, 1997. [2] Wireless medium access control (mac) and physical layer (phy) specications for low-rate wireless personal area networks (lr-wpans), 2003. [3] Specication of the bluetooth system version 4.0 [vol 0], 2009. [4] C. H. Alliance. The Continua Version One Design Guidelines, 2009. [5] B. Azimi-Sadjad, D. Sexton, P. Liu, and M. Mahony. Interference eect on IEEE 802.15.4 performance. In Proc. 3rd International Conference on Networked Sensing Systems. [6] M. Benthin and K. Kammeyer. Extraction of soft bit information with the use of m-ary modulation, 1995. [7] T. Esemann and H. Hellbrck. Limitations of u frequency hopping in 2.4 GHz ism-band for medical applications due to interference. In Proc. IEEE CCNC Conference, 2011. [8] Ettus Research LLC. USRP2 Universal Software Radio Peripheral, 2009. [9] U. Gardner. Exploitation of spectral redundancy in cyclostationary signals. IEEE Signal Processing Magazine, 8(2):1436, 1991. [10] N. Golmie, N. Chevrollier, and O. Rebala. Bluetooth and WLAN coexistence: Challenges and solutions. IEEE Wireless Communications Magazine, 10:2229, 2003. [11] Gnu radio. [online] http://gnuradio.org. [12] S. L. Howard, V. C. Gaudet, and C. Schlegel. Soft-bit decoding of regular low-density parity-check codes. [13] R. Jain. The art of computer systems performance analysis. Wiley, 1991. [14] K. Kim, I. A. Akbar, K. K. Bae, J.-S. Um, C. M. Spooner, and J. H. Reed. Cyclostationary approaches to signal detection and classication in cognitive radio,. In Proc. IEEE Int. Symposium on New Frontiers in Dynamic Spectrum Access Networks, 2007. [15] A. Mathew, N. Chandrababu, K. Elleithy, S. Rizv, and L. Almazaydeh. Interference of 802.11b WLAN and Bluetooth: Analysis and performance evaluation. International Journal of Computer Networks & Communications (IJCNC), 2:140150, 2010. [16] S. Shin, J. S. Kang, and H. S. Park. Packet error rate analysis of ZigBee under interferences of multiple Bluetooth piconets. In Proc. IEEE 69th Vehicular Technology Conference, Barcelona, Spain, 2009. [17] P. D. Sutton, K. E. Nolan, and L. E. Doyle. Cyclostationary signatures in practical cognitive radio applications. IEEE Journal on Selected Areas in Communications, 26:1324, 2008. [18] Y. Tsaig and D. L. Donoho. Compressed sensing. IEEE Trans. Inform. Theory, 52:12891306, 2006. [19] T. Ycek and H. Arslan. A survey of spectrum sensing u algorithms for cognitive radio applications. IEEE Communications Survey & Tutorials, 11(1):116130, 2009.

También podría gustarte