Spectrum Occupancy Estimation in Wireless Channels with ...
Spectrum Occupancy Estimation in Wireless Channels with Asymmetric Transmitter Powers John T. MacDonald
Dennis A. Roberson
Sapient Systems, Inc. Suite 34 456 West Frontage Road Northfield, IL 60093
Department of Computer Science Illinois Institute of Technology Chicago, IL 60616
Abstract—In proposed cognitive radio schemes, channel occupancy is often offered as the metric to determine if a channel is free to open a new communication channel. Channel occupancy is the time average of detected transmissions above a certain power level. In many systems the transmission power between the uplink side and down-link side are asymmetric. If a single power threshold is used, then a system may underestimate channel occupancy on the low power down-link side. This results in an increased probability of interference in an existing channel if a listen-before-talking scheme is used. To avoid this hidden terminal problem, cognitive radios need to better understand the receiver properties of existing wireless channels.
I. I NTRODUCTION In cognitive radio systems, a frequency agile radio can scan a broad swath of spectrum and determine where there are unused or under utilized spectra that can be exploited for new communication paths [1]. The means for determining whether a radio frequency (RF) channel is available involves many factors, including regulatory restrictions on the spectrum band, how often in time a primary transmitter is radiating, and the useful range of the radio system. The usual metric offered to determine if a portion of the RF spectrum is available at a particular point in time, at a particular point in space, is spectrum occupancy. Spectrum occupancy is a determination that a frequency channel is either used or unused based on the detection of RF energy above a certain threshold. This is often referred to as the received signal strength indication (RSSI). If a receiver detects energy above a certain threshold, then the channel is “in-use”, otherwise it is assumed to be unoccupied. When employed in wireless networks this is known as the clear channel assessment (CCA). In cognitive radio parlance, this is known as a listen-beforetalking system (LBT) [2]. These systems first determine if the channel is free before transmitting energy into the ether. In wireless networking systems, the RSSI mechanism can give rise to the “hidden terminal” problem, where a mobile system may detect the base station, but fail to detect transmissions from a remote competing transceiver in range of the base station. There is then a possibility that the hidden terminal will suffer interference from a competing spectrum user. An analogy to the hidden terminal problem presents itself in cognitive radio systems that use a listen-before-talking scheme. Many two-way wireless systems exhibit a transmitter
power asymmetry between the up-link and the down-link side. An example is land-mobile communication systems [3]. Cellular telephone systems deploy land based stations that transmit to mobile devices over a fixed geographic area. On this down-link side, the base stations have abundant power supplied by the power grid and can house bulky, but highly sensitive, receiver systems. Mobile phones, on the up-link side, are of necessity small an low powered. In GSM phone systems, the down-link power may be 100 watts while the up-link power is typically 100 milliwatts. Thus the down-link side has 30 dB less transmit power than the up-link side. If a single detection threshold is employed to determine the occupancy of the channel, then there exists a high probability that an up-link transmitter (mobile device) will not be detected even though the down-link side will be detected with high probability. Many different wireless communication systems employ this asymmetry in the transmission power. Most land-mobile systems operate in this manner. Satellite systems employ large powerful earth based stations to communicate with distant low powered satellites. Even radio location systems (i.e. RADAR) fall into this category because a RADAR transmits a large energy signal and receives a very low energy reflected signal. If an indiscriminating energy detection approach is the only criterion employed, then LBT based cognitive radio systems may interfere with many existing wireless systems. Knowledge of the existence of a transmission is insufficient to determine the availability of the channel. Cognitive radio systems must incorporate more knowledge about the receiver properties of existing wireless systems. The paper is organized in the following manner: In Section II we will consider the properties of the power asymmetric wireless channel and setting the detection threshold for spectrum occupancy. In Section III we reveiw the effective range of wireless systems. In Section IV we address the question of how much power an adaptive radio may safely transmit without interfering with competing systems. We conclude in Section V with some suggestions about how problems can be addressed by spectrum regulators. II. S IGNAL D ETECTION AND O CCUPANCY Cognitive radios operate as frequency agile radios that can survey the radio spectrum and determine what channels
are occupied, and which are free for transmission. After the assessment is performed, the radio can then operate in a manner to most efficiently utilize the available spectrum. The simplest method for assessing the state of the radio environment is to scan over a broad band of perspective channels to determine which are occupied and which are free. In such surveys, the metric most often used is spectrum occupancy. One definition for spectrum occupancy is “the event that, during an observation, the signal strength at a monitor will be above a certain threshold.” [4] The undefined parameter is the “certain threshold.” The received signal strength will be a function of the transmitter power, the separation between the transmitter and monitoring receiver, and several other factors including the gain of the transmit and receive antennae, the height of the antennae, the propagation properties of the air at the frequency of interest, and the physical properties of the propagation path. The received signal strength at the monitor can be modeled as SR = ST − L(dT R , θ) + W, dBm
(1)
where ST is the transmitter power, L(dT R , θ) is the path loss from the transmitter to the receiver, and W is a lognormal fading function to account for variations in the terrain and medium. The path loss L is primarily a function of the separation between the transmitter and the montoring receiver, dT R . The vector, θ, includes other paramters that effect the path loss. The ability to detect a signal SR that would indicate the occupancy of the channel depends on many factors of the transmitter. However, in cognitive radio systems, we are interested not only in avoiding the energy of other transmitters (that may interfere with our communications,) but avoiding interfering with existing receivers. In that regard, choosing the detection threshold becomes critical: If we chose the power threshold too low, then we may make a very conservative occupancy estimate that decides that the channel is in use even though the signal levels are too low for the intended receivers to demodulate. The worse case is if we set the power threshold too high. In that case we would determine that the channel is unoccupied and free to fill with transmissions, even though their may be equipment operating perfectly well at low power with highly sensitive receivers. Clearly, the proper determination of spectrum occupancy depends on the sensitivity of the receivers intended for operation in that particular spectrum allocation, at that particular location. It is apparent from (1) that with a measure of the received signal strength we could estimate the distance to the transmitter if we know the transmitter power a priori. Or we could estimate the power of the transmitter given the location of the transmitter. Neither of these numbers assists in answering the question: Is there a receiver in the vicinity of our transmitter that may suffer interference were we to transmit in the channel? To answer that question we need to know what type of receivers are intended to operate in the channel. Passively sensing the environment will not help in that determination.
Properly setting the power threshold for an occupancy evaluation requires knowing the minimum receiver sensitivity of the equipment intended to operate in each channel. In detection theory [5], there is a trade-off between probability of detection and probability of a false alarm. The hypothesis in occupancy testing is whether the signal power is sufficient for a receiver to demodulate. To maximize the probability of detection, the detection threshold should be set to the minimum receiver sensitivity, SDT = SRS ,
(2)
where SRS is the receiver sensitivity of the intended users of the allocated spectrum band. If that allocated band is a digital television signal, the receiver sensitivity may be as high as -80 dBm. If the allocated band is a cellular phone up-link channel, the receiver sensitivity may be -110 dBm. If the allocated band is for a satellite down-link channel, the receiver sensitivity may be as low as -200 dBm. The determination of the threshold must be made with some prior knowledge of the intended users in the band. Quite often, these spectrum occupancy surveys are performed with frequency scanning radios (spectrum analyzers) that detect the power of the received signal at the monitor. The threshold is often set just above the noise level of the test instrument [6]. Consider the spectrum occupancy plot of Figure 1, that covers the PCS up-link and down-link frequencies (this is a cell phone system similar to Europe’s GSM [7].) The occupancy survey was conducted at 10 West 35th Street in Chicago Illinois on November 16, 2005. The top plot shows the max-hold and min-hold traces of the spectrum analyzer along with the occupancy threshold. The middle plot shows the spectrograph of channel usage in the frequency bands over time. The dark pixels indicate points in time and frequency where energy was detected above the detection threshold. The bottom plot shows the spectrum occupancy as a percentage of time. The upper portion of the band, from 19301990 MHz represents the down-link channels, from the fixed base station to the mobile devices. These channels typically operate at nearly 100 Watts and the spectrum survey indicates that these frequencies are nearly entirely occupied. The up-link channels, in which mobile units transmit to the base stations, are located on the lower end of the band between 1850 and 1910 MHz. These typically operate at much lower transmit powers, perhaps 100 milliwatts and hence they fall below the noise level of the spectrum analyzer. The conclusion in the occupancy report was that the up-link band was nearly unoccupied. Mobile phone users talk as much as they listen [3], so the occupancy figures between the up-link and downlink sides should be identical, but a poor choice in power thresholds can easily overestimate the occupancy on the downlink side and underestimate the occupancy on the up-link side. III. T HE E FFECTIVE R ANGE
OF
W IRELESS S YSTEMS
Spectrum occupancy is a determination of whether or not a transmission is present in a frequency band. From the perspective of cognitive radio systems, a better determination
is whether the radio channel is utilized in the immediate vicinity. To do so we must consider the effective range of radio systems. In (1) the received power from a transmitter is a function of the transmitter power, the path loss (which is a function of the distance between the transmitter and the receiver), and fading noise. In order to operate as a useful communication system, the received power, SR , must be greater than the minimum sensitivity, SRS of the receiver. The receiver sensitivity, together with the transmit power limits the effective range of the system. SRS SRS
≤ ≤
SR ST − L(dRT , θ)
(3) (4)
dRT
≤
L−1 (ST − SRS )
(5)
The higher the transmit power, the larger the effective range of the wireless system. The greater the receiver sensitivity (reflected by a more negative sensitivity measure SRS ), the greater the effective range. To determine if the spectrum band is utilized, a fair question to ask is whether a receiver can detect its intended transmission. For that reason, the detection threshold should be set to the best sensitivity of the intended receiver. If the received signal threshold is set higher than the sensitivity of the receiver, then we will tend to underestimate the effective range of a system and under report the spectrum utilization. If the received signal threshold is set lower than the sensitivity of the receiver, then we will tend to overestimate the range of the system. In cognitive radio systems, it may be better to err on the side of over-estimating, rather than underestimating, the range of the allocated legacy system. Spectrum regulators do not set limits on the sensitivity of receivers, but rather on the power of the transmitters. They assign licenses in certain geographic areas in the hopes that signal spaces will not overlap. Manufacturers set the limits of the receiver sensitivity of communication devices based on the demands of the market. While the majority will want inexpensive devices with adequate sensitivity, there are always users who want additional range for there communication devices and are willing to pay for it. Users may invest in higher gain antennas, or higher quality receivers, to extend the range of their systems. Setting a reasonable threshold of the minimum receiver sensitivity becomes an statistical task of estimating the quality of receivers employed for a particular function. Wherever the threshold is set, a certain number of users may fall on the wrong side of the estimate. In systems with asymmetric transmitter powers, the transmit powers are very different on the down-link side than the uplink side. The effective range on the up-link and down-link side are however similar because there is an equal asymmetry between the up-link and down-link receiver sensitivities. Because the up-link channel needs the same effective range as the down-link channel, (ST U − SRSU ) = (ST D − SRSD )
(6)
where the U and D subscripts correspond to the up-link and down-link characteristics respectively. If a single power
U
M
D
Fig. 2. Detecting an assymetric power channel. The up-link tranceiver, at position U, and the down-link tranceiver at position D, are at the limits of their respective ranges. A monitor, at position M, with a single power threshold, will miss the up-link tranceiver which will appear to be out of range.
threshold is set between the two, then the down-link utilization will be over estimated, and the up-link utilization will be overestimated. This is illustrated in Figure 2. The up-link and and down-link tranceivers are at the outer limits of there equal ranges reflected by the circles. A monitor is positioned directly between then but it underestimates the range of the up-link side and may give the false indication that the channel is unutilized. IV. E STIMATING AVAILABLE P OWER M ARGIN Beyond simply determining the presence of other competing transmitters in a channel, a cognitive radio should be able to determine just how much power it may transmit into a channel without impacting the existing allocated users. Here again, knowing the receiver properties of the primary channel systems, is key to avoiding interference with allocated spectrum users. If our occupancy test shows that there is a transmitter within range of the intended receivers, then an opportunistic transmitter would have to operate below the receiver sensitivity of the the legacy radio equipment. If our the occupancy test shows that the received signal strength is below the receiver sensitivity of the victim receivers, then we can safely transmit power without interfering with non-existent signals [2]. The estimate of the received signal power of the legacy transmitter, together with our knowledge of the receiver sensitivity of the intended receivers gives a measure of the power margin that we may introduce into the channel without causing interference with the legacy systems, SP M = SRS − min(SRM , SRD ),
(7)
where again, SRS is the receiver sensitivity of the victim receivers, SRD is the detected power of the most powerful transmitter detected, and SRM is the detection sensitivity (the noise floor) of the monitor device. From (7), if the received power from the detected transmitter, SRD , is greater than the sensitivity of the intended receiver, SRS , then the channel is essentially occupied and any additional power would have to operate below the receiver sensitivity to avoid interference. This would constitute an “underlay” system. [8].
C
T A
B
Fig. 3. A opportunistic transmitter, T, operating between in the gap beyound the effective ranges of three allocted transmitters at locations A, B, and C. The power of the transmitter would be limited by the effective range of the legacy transmitters.
If the received power from the detected transmitter, SRD , is less than the sensitivity of the intended receiver, SRS , then we can transmit with power up to the limiting range of the legacy transmitter. In this case we can transmit in a power range that outside of the range of the legacy system and any potential victim receivers. This case of operating an opportunistic transmitter within a limited power margin is illustrated in Figure 3. Here the cognitive radio system, C, operates between two allocated transmitters at positions A and B. Because it falls outside of the useful range of the legacy transmitters, (as determined by the transmission power and receiver sensitivity of the legacy system) it may transmit with power up to the range limit imposed by the range of the legacy system. The third case of (7) is where no transmitter is detected, and we can transmit freely. In this case, we must err on the side of caution because the transmitter may operate below the noise floor of our monitoring instruments. Since we can not detect the presence of a transmitter with a detected power below the noise floor of the monitor, SRM , we must allow for the existence of such a transmitter and limit the transmission power, SP M , accordingly. To achieve the greatest range and flexibility of a cognitive radio system, one would seek to maximize the receiver sensitivity in the design of a cognitive radio [9]. V. C ONCLUSION Cognitive radios must operate with some awareness of the state of the spectrum space that it intends to occupy. This requires that they understand competitors that may interfere with their own communications, and the nature of legacy radios with which they may interfere.
It is insufficient to survey the landscape for spectrum occupancy without some knowledge of the systems that may be negatively impacted by new transmissions. Cognitive radios must also posses some knowledge of the operational parameters of the radio equipment that they will compete with. One case mentioned was the asymmetric power channel. A cognitive radio could conceivably inter-operate with a GSM down-link channel at a higher power than it could inter-operate with a GSM up-link channel. That is because of the nature of the GSM radio equipment where the down-link receivers (on the mobile units) have a much higher receiver sensitivity than the down-link receivers (the fixed base stations.) For that reason, broadband spectrum occupancy surveys that operate with a threshold set to the noise floor of the instrument will tend to underestimate spectrum utilization. With the knowledge of the capabilities of existing wireless systems, cognitive radios can determine a power margin in which they can operate without impacting existing receivers. From the perspective of spectrum policy makers, perhaps it may be necessary to define the parameters of fixed and flexible radios that are permitted to operate within the same allocated band. They may need to define maximum transmitter powers, maximum channel bandwidth, and minimum receiver sensitivity so that all users of a spectrum band can operate with an equal set of ground rules. It would be best to operate without making unnecessary assumptions so that smart radios can be idiot proof. ACKNOWLEDGMENT The authors would like to thank Shared Spectrum, Inc., for providing the spectrum occupancy plots. R EFERENCES [1] S. Haykin, “Cognitive radio: Brain empowered wireless communications,” IEEE Journal on Selected Areas in Communications, vol. 23, no. 2, p. 201, February 2005. [2] A. Leu, M. McHenry, and B. Mark, “Modeling and analysis of interference in listen-before-talk spectrum access schemes,” International Journal of Network Management, vol. 16, no. 2, March 2006. [3] G. Hagn and T. Dayharsh, “Land mobile radio communications channel occupancy, waiting time, and spectrum saturation,” IEEE Transactions on Electromagnetic Compatibility, vol. 19, no. 3, p. 281, 1977. [4] A.D.Spaulding and G. Hagn, “On the definition and estimation of spectrum occupancy,” IEEE Transactions on Electromagnetic Compatibility, vol. 19, no. 3, p. 269, 1977. [5] H. Poor, An Introduction to Signal Detection and Estimation. SpringerVerlag, 1988. [6] M. McHenry, D. McCloskey, D. Roberson, and J. MacDonald, “Chicago spectrum occupancy measurements november 2005,” Report for the WIL of IIT, 2006. [7] B. Z. Kobb, Wireless Spectrum Finder - Telecommunications, Government, and Scientific Radio Frequency Allocations in the U.S. 30 MHz 300 GHz. McGraw-Hill, 2001. [8] O.Popescu and C.Rose, “Water filling may not good neighbors make,” Proceedings of the IEEE Global Communications Conference, GLOBECON2003, 2003. [9] R.Rubenstein, “Radios get smart,” IEEE Spectrum, vol. 44, no. 2, p. 46, February 2007. [10] D. A. Roberson, C. S. Hood, J. L. Locicero, and J. T. MacDonald, “Spectral occupancy and interference studies in support of cognitive radio technology deployment,” Proceedings of the IEEE Workshop on Networking Technologies for Software Defined Radio, November 2006.
Power Spectrum -60
dBm
-80 -100 -120 -140 1840
1860
1880
1900
1920 freq MHz
1940
1960
1980
2000
1940
1960
1980
2000
1940
1960
1980
2000
Duty 350
time sec
300 250 200 150 100 50 0 1840
1860
1880
1900
1920 freq MHz Occupancy
100
percent
80 60 40 20 0 1840
1860
1880
1900
Start Freq. Stop Freq. Resolution Attenuation Sample Points Threshold Occupancy
1920 freq MHz
1850 1990 25 0 5600 -95 35
MHz MHz KHz dB dBm Percent
Fig. 1. Spectrum occupancy in the range of 1850-1990 MHz., the PCS band. The up-link channels are in the range of 1850 to 1910 MHz. The down-link channels are in the range of 1930 to 1990 MHz. The detection threshold is set to -95 dBm which tends to underestimate the occupancy of the up-link channels. The top graph shows the spectral power detected over a five minute period. The middle graph shows the time at which energy was detected for the frequency range. The bottom graph shows the occupancy as a percentage of time over the frequency range.
Dennis A. Roberson
Sapient Systems, Inc. Suite 34 456 West Frontage Road Northfield, IL 60093
Department of Computer Science Illinois Institute of Technology Chicago, IL 60616
Abstract—In proposed cognitive radio schemes, channel occupancy is often offered as the metric to determine if a channel is free to open a new communication channel. Channel occupancy is the time average of detected transmissions above a certain power level. In many systems the transmission power between the uplink side and down-link side are asymmetric. If a single power threshold is used, then a system may underestimate channel occupancy on the low power down-link side. This results in an increased probability of interference in an existing channel if a listen-before-talking scheme is used. To avoid this hidden terminal problem, cognitive radios need to better understand the receiver properties of existing wireless channels.
I. I NTRODUCTION In cognitive radio systems, a frequency agile radio can scan a broad swath of spectrum and determine where there are unused or under utilized spectra that can be exploited for new communication paths [1]. The means for determining whether a radio frequency (RF) channel is available involves many factors, including regulatory restrictions on the spectrum band, how often in time a primary transmitter is radiating, and the useful range of the radio system. The usual metric offered to determine if a portion of the RF spectrum is available at a particular point in time, at a particular point in space, is spectrum occupancy. Spectrum occupancy is a determination that a frequency channel is either used or unused based on the detection of RF energy above a certain threshold. This is often referred to as the received signal strength indication (RSSI). If a receiver detects energy above a certain threshold, then the channel is “in-use”, otherwise it is assumed to be unoccupied. When employed in wireless networks this is known as the clear channel assessment (CCA). In cognitive radio parlance, this is known as a listen-beforetalking system (LBT) [2]. These systems first determine if the channel is free before transmitting energy into the ether. In wireless networking systems, the RSSI mechanism can give rise to the “hidden terminal” problem, where a mobile system may detect the base station, but fail to detect transmissions from a remote competing transceiver in range of the base station. There is then a possibility that the hidden terminal will suffer interference from a competing spectrum user. An analogy to the hidden terminal problem presents itself in cognitive radio systems that use a listen-before-talking scheme. Many two-way wireless systems exhibit a transmitter
power asymmetry between the up-link and the down-link side. An example is land-mobile communication systems [3]. Cellular telephone systems deploy land based stations that transmit to mobile devices over a fixed geographic area. On this down-link side, the base stations have abundant power supplied by the power grid and can house bulky, but highly sensitive, receiver systems. Mobile phones, on the up-link side, are of necessity small an low powered. In GSM phone systems, the down-link power may be 100 watts while the up-link power is typically 100 milliwatts. Thus the down-link side has 30 dB less transmit power than the up-link side. If a single detection threshold is employed to determine the occupancy of the channel, then there exists a high probability that an up-link transmitter (mobile device) will not be detected even though the down-link side will be detected with high probability. Many different wireless communication systems employ this asymmetry in the transmission power. Most land-mobile systems operate in this manner. Satellite systems employ large powerful earth based stations to communicate with distant low powered satellites. Even radio location systems (i.e. RADAR) fall into this category because a RADAR transmits a large energy signal and receives a very low energy reflected signal. If an indiscriminating energy detection approach is the only criterion employed, then LBT based cognitive radio systems may interfere with many existing wireless systems. Knowledge of the existence of a transmission is insufficient to determine the availability of the channel. Cognitive radio systems must incorporate more knowledge about the receiver properties of existing wireless systems. The paper is organized in the following manner: In Section II we will consider the properties of the power asymmetric wireless channel and setting the detection threshold for spectrum occupancy. In Section III we reveiw the effective range of wireless systems. In Section IV we address the question of how much power an adaptive radio may safely transmit without interfering with competing systems. We conclude in Section V with some suggestions about how problems can be addressed by spectrum regulators. II. S IGNAL D ETECTION AND O CCUPANCY Cognitive radios operate as frequency agile radios that can survey the radio spectrum and determine what channels
are occupied, and which are free for transmission. After the assessment is performed, the radio can then operate in a manner to most efficiently utilize the available spectrum. The simplest method for assessing the state of the radio environment is to scan over a broad band of perspective channels to determine which are occupied and which are free. In such surveys, the metric most often used is spectrum occupancy. One definition for spectrum occupancy is “the event that, during an observation, the signal strength at a monitor will be above a certain threshold.” [4] The undefined parameter is the “certain threshold.” The received signal strength will be a function of the transmitter power, the separation between the transmitter and monitoring receiver, and several other factors including the gain of the transmit and receive antennae, the height of the antennae, the propagation properties of the air at the frequency of interest, and the physical properties of the propagation path. The received signal strength at the monitor can be modeled as SR = ST − L(dT R , θ) + W, dBm
(1)
where ST is the transmitter power, L(dT R , θ) is the path loss from the transmitter to the receiver, and W is a lognormal fading function to account for variations in the terrain and medium. The path loss L is primarily a function of the separation between the transmitter and the montoring receiver, dT R . The vector, θ, includes other paramters that effect the path loss. The ability to detect a signal SR that would indicate the occupancy of the channel depends on many factors of the transmitter. However, in cognitive radio systems, we are interested not only in avoiding the energy of other transmitters (that may interfere with our communications,) but avoiding interfering with existing receivers. In that regard, choosing the detection threshold becomes critical: If we chose the power threshold too low, then we may make a very conservative occupancy estimate that decides that the channel is in use even though the signal levels are too low for the intended receivers to demodulate. The worse case is if we set the power threshold too high. In that case we would determine that the channel is unoccupied and free to fill with transmissions, even though their may be equipment operating perfectly well at low power with highly sensitive receivers. Clearly, the proper determination of spectrum occupancy depends on the sensitivity of the receivers intended for operation in that particular spectrum allocation, at that particular location. It is apparent from (1) that with a measure of the received signal strength we could estimate the distance to the transmitter if we know the transmitter power a priori. Or we could estimate the power of the transmitter given the location of the transmitter. Neither of these numbers assists in answering the question: Is there a receiver in the vicinity of our transmitter that may suffer interference were we to transmit in the channel? To answer that question we need to know what type of receivers are intended to operate in the channel. Passively sensing the environment will not help in that determination.
Properly setting the power threshold for an occupancy evaluation requires knowing the minimum receiver sensitivity of the equipment intended to operate in each channel. In detection theory [5], there is a trade-off between probability of detection and probability of a false alarm. The hypothesis in occupancy testing is whether the signal power is sufficient for a receiver to demodulate. To maximize the probability of detection, the detection threshold should be set to the minimum receiver sensitivity, SDT = SRS ,
(2)
where SRS is the receiver sensitivity of the intended users of the allocated spectrum band. If that allocated band is a digital television signal, the receiver sensitivity may be as high as -80 dBm. If the allocated band is a cellular phone up-link channel, the receiver sensitivity may be -110 dBm. If the allocated band is for a satellite down-link channel, the receiver sensitivity may be as low as -200 dBm. The determination of the threshold must be made with some prior knowledge of the intended users in the band. Quite often, these spectrum occupancy surveys are performed with frequency scanning radios (spectrum analyzers) that detect the power of the received signal at the monitor. The threshold is often set just above the noise level of the test instrument [6]. Consider the spectrum occupancy plot of Figure 1, that covers the PCS up-link and down-link frequencies (this is a cell phone system similar to Europe’s GSM [7].) The occupancy survey was conducted at 10 West 35th Street in Chicago Illinois on November 16, 2005. The top plot shows the max-hold and min-hold traces of the spectrum analyzer along with the occupancy threshold. The middle plot shows the spectrograph of channel usage in the frequency bands over time. The dark pixels indicate points in time and frequency where energy was detected above the detection threshold. The bottom plot shows the spectrum occupancy as a percentage of time. The upper portion of the band, from 19301990 MHz represents the down-link channels, from the fixed base station to the mobile devices. These channels typically operate at nearly 100 Watts and the spectrum survey indicates that these frequencies are nearly entirely occupied. The up-link channels, in which mobile units transmit to the base stations, are located on the lower end of the band between 1850 and 1910 MHz. These typically operate at much lower transmit powers, perhaps 100 milliwatts and hence they fall below the noise level of the spectrum analyzer. The conclusion in the occupancy report was that the up-link band was nearly unoccupied. Mobile phone users talk as much as they listen [3], so the occupancy figures between the up-link and downlink sides should be identical, but a poor choice in power thresholds can easily overestimate the occupancy on the downlink side and underestimate the occupancy on the up-link side. III. T HE E FFECTIVE R ANGE
OF
W IRELESS S YSTEMS
Spectrum occupancy is a determination of whether or not a transmission is present in a frequency band. From the perspective of cognitive radio systems, a better determination
is whether the radio channel is utilized in the immediate vicinity. To do so we must consider the effective range of radio systems. In (1) the received power from a transmitter is a function of the transmitter power, the path loss (which is a function of the distance between the transmitter and the receiver), and fading noise. In order to operate as a useful communication system, the received power, SR , must be greater than the minimum sensitivity, SRS of the receiver. The receiver sensitivity, together with the transmit power limits the effective range of the system. SRS SRS
≤ ≤
SR ST − L(dRT , θ)
(3) (4)
dRT
≤
L−1 (ST − SRS )
(5)
The higher the transmit power, the larger the effective range of the wireless system. The greater the receiver sensitivity (reflected by a more negative sensitivity measure SRS ), the greater the effective range. To determine if the spectrum band is utilized, a fair question to ask is whether a receiver can detect its intended transmission. For that reason, the detection threshold should be set to the best sensitivity of the intended receiver. If the received signal threshold is set higher than the sensitivity of the receiver, then we will tend to underestimate the effective range of a system and under report the spectrum utilization. If the received signal threshold is set lower than the sensitivity of the receiver, then we will tend to overestimate the range of the system. In cognitive radio systems, it may be better to err on the side of over-estimating, rather than underestimating, the range of the allocated legacy system. Spectrum regulators do not set limits on the sensitivity of receivers, but rather on the power of the transmitters. They assign licenses in certain geographic areas in the hopes that signal spaces will not overlap. Manufacturers set the limits of the receiver sensitivity of communication devices based on the demands of the market. While the majority will want inexpensive devices with adequate sensitivity, there are always users who want additional range for there communication devices and are willing to pay for it. Users may invest in higher gain antennas, or higher quality receivers, to extend the range of their systems. Setting a reasonable threshold of the minimum receiver sensitivity becomes an statistical task of estimating the quality of receivers employed for a particular function. Wherever the threshold is set, a certain number of users may fall on the wrong side of the estimate. In systems with asymmetric transmitter powers, the transmit powers are very different on the down-link side than the uplink side. The effective range on the up-link and down-link side are however similar because there is an equal asymmetry between the up-link and down-link receiver sensitivities. Because the up-link channel needs the same effective range as the down-link channel, (ST U − SRSU ) = (ST D − SRSD )
(6)
where the U and D subscripts correspond to the up-link and down-link characteristics respectively. If a single power
U
M
D
Fig. 2. Detecting an assymetric power channel. The up-link tranceiver, at position U, and the down-link tranceiver at position D, are at the limits of their respective ranges. A monitor, at position M, with a single power threshold, will miss the up-link tranceiver which will appear to be out of range.
threshold is set between the two, then the down-link utilization will be over estimated, and the up-link utilization will be overestimated. This is illustrated in Figure 2. The up-link and and down-link tranceivers are at the outer limits of there equal ranges reflected by the circles. A monitor is positioned directly between then but it underestimates the range of the up-link side and may give the false indication that the channel is unutilized. IV. E STIMATING AVAILABLE P OWER M ARGIN Beyond simply determining the presence of other competing transmitters in a channel, a cognitive radio should be able to determine just how much power it may transmit into a channel without impacting the existing allocated users. Here again, knowing the receiver properties of the primary channel systems, is key to avoiding interference with allocated spectrum users. If our occupancy test shows that there is a transmitter within range of the intended receivers, then an opportunistic transmitter would have to operate below the receiver sensitivity of the the legacy radio equipment. If our the occupancy test shows that the received signal strength is below the receiver sensitivity of the victim receivers, then we can safely transmit power without interfering with non-existent signals [2]. The estimate of the received signal power of the legacy transmitter, together with our knowledge of the receiver sensitivity of the intended receivers gives a measure of the power margin that we may introduce into the channel without causing interference with the legacy systems, SP M = SRS − min(SRM , SRD ),
(7)
where again, SRS is the receiver sensitivity of the victim receivers, SRD is the detected power of the most powerful transmitter detected, and SRM is the detection sensitivity (the noise floor) of the monitor device. From (7), if the received power from the detected transmitter, SRD , is greater than the sensitivity of the intended receiver, SRS , then the channel is essentially occupied and any additional power would have to operate below the receiver sensitivity to avoid interference. This would constitute an “underlay” system. [8].
C
T A
B
Fig. 3. A opportunistic transmitter, T, operating between in the gap beyound the effective ranges of three allocted transmitters at locations A, B, and C. The power of the transmitter would be limited by the effective range of the legacy transmitters.
If the received power from the detected transmitter, SRD , is less than the sensitivity of the intended receiver, SRS , then we can transmit with power up to the limiting range of the legacy transmitter. In this case we can transmit in a power range that outside of the range of the legacy system and any potential victim receivers. This case of operating an opportunistic transmitter within a limited power margin is illustrated in Figure 3. Here the cognitive radio system, C, operates between two allocated transmitters at positions A and B. Because it falls outside of the useful range of the legacy transmitters, (as determined by the transmission power and receiver sensitivity of the legacy system) it may transmit with power up to the range limit imposed by the range of the legacy system. The third case of (7) is where no transmitter is detected, and we can transmit freely. In this case, we must err on the side of caution because the transmitter may operate below the noise floor of our monitoring instruments. Since we can not detect the presence of a transmitter with a detected power below the noise floor of the monitor, SRM , we must allow for the existence of such a transmitter and limit the transmission power, SP M , accordingly. To achieve the greatest range and flexibility of a cognitive radio system, one would seek to maximize the receiver sensitivity in the design of a cognitive radio [9]. V. C ONCLUSION Cognitive radios must operate with some awareness of the state of the spectrum space that it intends to occupy. This requires that they understand competitors that may interfere with their own communications, and the nature of legacy radios with which they may interfere.
It is insufficient to survey the landscape for spectrum occupancy without some knowledge of the systems that may be negatively impacted by new transmissions. Cognitive radios must also posses some knowledge of the operational parameters of the radio equipment that they will compete with. One case mentioned was the asymmetric power channel. A cognitive radio could conceivably inter-operate with a GSM down-link channel at a higher power than it could inter-operate with a GSM up-link channel. That is because of the nature of the GSM radio equipment where the down-link receivers (on the mobile units) have a much higher receiver sensitivity than the down-link receivers (the fixed base stations.) For that reason, broadband spectrum occupancy surveys that operate with a threshold set to the noise floor of the instrument will tend to underestimate spectrum utilization. With the knowledge of the capabilities of existing wireless systems, cognitive radios can determine a power margin in which they can operate without impacting existing receivers. From the perspective of spectrum policy makers, perhaps it may be necessary to define the parameters of fixed and flexible radios that are permitted to operate within the same allocated band. They may need to define maximum transmitter powers, maximum channel bandwidth, and minimum receiver sensitivity so that all users of a spectrum band can operate with an equal set of ground rules. It would be best to operate without making unnecessary assumptions so that smart radios can be idiot proof. ACKNOWLEDGMENT The authors would like to thank Shared Spectrum, Inc., for providing the spectrum occupancy plots. R EFERENCES [1] S. Haykin, “Cognitive radio: Brain empowered wireless communications,” IEEE Journal on Selected Areas in Communications, vol. 23, no. 2, p. 201, February 2005. [2] A. Leu, M. McHenry, and B. Mark, “Modeling and analysis of interference in listen-before-talk spectrum access schemes,” International Journal of Network Management, vol. 16, no. 2, March 2006. [3] G. Hagn and T. Dayharsh, “Land mobile radio communications channel occupancy, waiting time, and spectrum saturation,” IEEE Transactions on Electromagnetic Compatibility, vol. 19, no. 3, p. 281, 1977. [4] A.D.Spaulding and G. Hagn, “On the definition and estimation of spectrum occupancy,” IEEE Transactions on Electromagnetic Compatibility, vol. 19, no. 3, p. 269, 1977. [5] H. Poor, An Introduction to Signal Detection and Estimation. SpringerVerlag, 1988. [6] M. McHenry, D. McCloskey, D. Roberson, and J. MacDonald, “Chicago spectrum occupancy measurements november 2005,” Report for the WIL of IIT, 2006. [7] B. Z. Kobb, Wireless Spectrum Finder - Telecommunications, Government, and Scientific Radio Frequency Allocations in the U.S. 30 MHz 300 GHz. McGraw-Hill, 2001. [8] O.Popescu and C.Rose, “Water filling may not good neighbors make,” Proceedings of the IEEE Global Communications Conference, GLOBECON2003, 2003. [9] R.Rubenstein, “Radios get smart,” IEEE Spectrum, vol. 44, no. 2, p. 46, February 2007. [10] D. A. Roberson, C. S. Hood, J. L. Locicero, and J. T. MacDonald, “Spectral occupancy and interference studies in support of cognitive radio technology deployment,” Proceedings of the IEEE Workshop on Networking Technologies for Software Defined Radio, November 2006.
Power Spectrum -60
dBm
-80 -100 -120 -140 1840
1860
1880
1900
1920 freq MHz
1940
1960
1980
2000
1940
1960
1980
2000
1940
1960
1980
2000
Duty 350
time sec
300 250 200 150 100 50 0 1840
1860
1880
1900
1920 freq MHz Occupancy
100
percent
80 60 40 20 0 1840
1860
1880
1900
Start Freq. Stop Freq. Resolution Attenuation Sample Points Threshold Occupancy
1920 freq MHz
1850 1990 25 0 5600 -95 35
MHz MHz KHz dB dBm Percent
Fig. 1. Spectrum occupancy in the range of 1850-1990 MHz., the PCS band. The up-link channels are in the range of 1850 to 1910 MHz. The down-link channels are in the range of 1930 to 1990 MHz. The detection threshold is set to -95 dBm which tends to underestimate the occupancy of the up-link channels. The top graph shows the spectral power detected over a five minute period. The middle graph shows the time at which energy was detected for the frequency range. The bottom graph shows the occupancy as a percentage of time over the frequency range.
