Spectrum Survey in Singapore: Occupancy Measurements and Analyses

Md Habibul Islam , Choo Leng Koh , Ser Wah Oh , Xianming Qing , Yoke Yong Lai , Cavin Wang , Ying-Chang Liang , Bee Eng Toh , Francois Chin , Geok Leng Tan , and William Toh
Institute for Infocomm Research (I2 R), 21 Heng Mui Keng Terrace, Singapore 119613 e-mail: {habibul,clkoh,swoh,qingxm,ycliang,chinfrancois}@i2r.a-star.edu.sg Info-communication Development Authority (IDA) of Singapore, 8 Temasek Boulevard, Singapore 038988 e-mail: {LAI Yoke Yong,Cavin WANG,TOH Bee Eng,TAN Geok Leng,William TOH}@ida.gov.sg Invited Paper

Abstract We study the 24-hour spectrum usage pattern in Singapore in the frequency bands ranging from 80 MHz to 5850 MHz. The objectives are to nd how the scarce radio spectrum allocated to different services is utilized in Singapore and identify the bands that could be accessed for future opportunistic use due to their low or no active utilization. The results from the spectrum measurements taken over 12 weekday periods reveal that a signicant amount of spectrum in Singapore has very low occupancy all the time. The occupancy is quantied as the amount of spectrum detected above a certain received power threshold. The outcome of this study suggests that Singapore has a great potential for employing emerging spectrum sharing technology such as the cognitive radio technology to accommodate enormous demands for future wireless services. However, this study of spectrum survey is preliminary in its nature and future long term studies need to be performed to determine any potential secondary usage on those channels that have low or no active utilization.

I. I NTRODUCTION The emergence of new technologies and the phenomenal growth of wireless communication services have created an ever increasing demand for the radio frequency spectrum. The radio frequency spectrum is a scarce natural resource and to meet the growing demands of all these new bandwidth hungry services, effective and efcient utilization of radio spectrum is essential. Here comes the issue of introducing a new and better spectrum management policy. The traditional policy of xed spectrum allocation came into question by the shocking ndings from a number of measurement studies of spectrum utilization by the FCC Spectrum Policy task force [1]. These studies reveal that a large portion of the allocated spectrum is used sporadically and geographical variations in the utilization of assigned spectrum ranges from 15% to 85% with a high variation in time [2]. At the same time, it is observed in some other related studies that most of the spectrum, in most of the places is completely unused most of the times. The results of these studies indicate that the xed spectrum allocation policy is no more a viable option to meet the increasing demands for radio spectrum for future generation wireless services. Underutilization of scarce radio spectrum has given birth to the promising idea of allowing unused part of spectrum by the primary license holder to become available temporarily to secondary users.

In many countries, government policy initiatives are expanding unlicensed spectral bands facilitating dynamic spectrum access technologies such as cognitive radios and considering how to afford access to the white spaces in the television bands [3]. However, without proper understanding of the current and projected spectrum usage pattern, all these initiatives and investments might not produce the expected results. Spectrum monitoring/spectrum survey is one of the essential tools of spectrum management which provides the policy makers with the necessary information about the frequency usage pattern of different services in different bands. Several broadband spectrum surveys were conducted at San Francisco [4], Denver [5], and San Diego [6]. These survey reports show the maximum, minium and average measured power levels of received signals in different bands. Although the maximum, minimum and average curves from these studies can be used to qualitatively assess the relative density of channel occupancy on a band by band basis, these data cannot be used to infer the statistical percentage of time that channels are occupied. A series of more recent studies by Shared Spectrum Company (SSC) [3], [7] [12] provide the needed temporal spectrum use information and could be used to estimate spectrum white space. In this work, we conduct a similar spectrum survey in Singapore. The objective is to nd the spectrum usage pattern in frequencies from 80 MHz to 5.85 GHz and identify the bands that have low or no active utilization. The 24-hour measurement has been taken at the roof top of Institute for Infocomm Researchs (I2 Rs) building over 12 weekday periods. It should be noted that compared to different cities in USA, where the previous spectrum surveys were conducted, Singapore is unique in the sense that spectrum usage in neighboring countries might affect the occupancy pattern in some bands in Singapore. The rest of the paper is organized as follows. In Section II, we provide the detail of the measurement equipment. A description of data collection and data calibration methodology is given in Section III. In Section IV, we describe our methodology to analyze the data. Occupancy results of this survey are presented in Section V. In Section VI, we make some notes regarding the major observation of this study. Finally, in Section VII,

Fig. 1: Measurement systems.

we conclude our report and provide some recommendations for future study. II. M EASUREMENT S YSTEMS The measurement has been taken on the roof top (6th oor) of the Insitute for Infocomm Research (I2 R) located at 21 Heng Mui Keng Terrace, Singapore. The equipment used for the measurement in this study consist of a BiConiLog (hybrid biconical and log periodic) directional antenna of model 3149 made by ETS-Lindgren. The antenna has a frequency range of 80 MHz to 6 GHz and it accepts a maximum of 750 W continuous power input at low frequency of operation. It is mounted on the roof top using vertical polarizaton. The antenna is connected by a low loss cable to Agilents spectrum analyzer of model E4407B which has a operating frequency range of 9 KHz to 40 GHz, measurement range from -150 dBm to 30 dBm, 1 dB overall amplitude accuracy and a maximum sensitivity of -150 dBm. The spectrum analyzer is connected to a desktop computer by general purpose interface bus (GPIB) and these two equipment are kept in a cabinet with ventilaton. The LabVIEW8.2 software, installed in the desktop computer, is used for automatically setting the equipment for measurement and recording the captured data. Note that the antennas used in previous spectrum surveys were both omni-directional, e.g. [3] and directional with automatic rotator, e.g. [7]. Moreover, the antennas in these studies were connected to pre-amps and lters. In our study, the directional antenna does not have any automatic rotator and no pre-amps and lters are used. Therefore, we rotate the directional antenna manually to get the omni-directional effect of the signals and we rely on spectrum analyzers lter. The whole measurement system is shown in Fig. 1. III. DATA C ALIBRATION The spectrum survey is performed over 12 weekday periods across a frequency range of 80 MHz to 5.85 GHz. To get the omnidirectional effect, the antenna is rotated manually each day by 30o . The overall frequency range is divided into several bands each having a frequency span of 60 MHz. Each 60 MHz frequency band has 401 frequency points, i.e. the separation between two consecutive frequency points is 150 KHz.

Fig. 2: Antenna gain

For each frequency point, the raw data represent the received signal power level measured at the spectrum analyzer for that frequency. For each 60 MHz frequency band, the measurement is carried out around every 13.8 minutes. Therefore, for each frequency point, we have 104 received data samples from 24hour measurement and a total of 104 12 = 1248 samples for the whole observation period. During the measurement, the resolution bandwidth (RBW) and the video bandwidth (VBW) of the spectrum analyzer are set to 10 KHz and 100 KHz, respectively. Due to antenna gains and cable losses, there is a difference between the power level at the spectrum analyzer input and the power level at the antenna input. Antenna gains and cable losses across different frequencies are shown in Fig. 2 and Fig. 3, respectively. Let y, x, g, and l denote the power level at the spectrum analyzer input, power level at the antenna input, antenna gain, and cable loss, respectiveley. Then, the calibrated data x is obtained from the following relation y (dBm) = x (dBm) + g (dB) + l (dB) . IV. M ETHODOLOGY FOR DATA A NALYSES Spectrum occupancy is the most often used metric to determine which frequency bands are occupied and which are free. One way to dene the occupancy is the event that, during an observation, the received signal strength at the measuring equipment is above a certain threshold. However, for opportunistic use of an unoccupied licensed channel, choosing the threshold is critical. Selecting the threshold too low would result in a very conservative occupancy estimate that decides that channel is in use due to the presence of ambient noise and make the opportunistic use limited even though the signal levels are too low for the incumbent receivers to demodulate. On the other hand, if the threshold is chosen too high, some channels may be declared as unoccupied even though there may be some incumbent devices operating perfectly well at low power with highly sensitive receivers. In the existing literature on spectrum (1)

Outdoor cable loss 1 2 3

50 Power (dBm) 20 30 40

4 dB 5 6

60 70 80 90

7 8 9

100 110 120 1000 2000 3000 4000 Freqeuency (MHz) 5000

500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 Frequency (MHz)

Fig. 3: Cable loss

Fig. 4: Received power vs frequency.

survey, the threshold is often set a certain dB above the noise level of the measurement equipment. The drawbacks of this method is that if the equipment noise is more than the ambient noise, we come up with a threshold too high to declare the observed channel as unoccupied most of the time. Thus this method is prone to the sensitivity of the measuring equipment. The drawback of this method can be overcome by setting the threshold a certain dB above the ambient noise. One of the ITU recommendations [13, page 168] suggests setting the threshold 10 dB above the ambient noise. Although the theoretical value of the ambient noise can be calculated easily, the actual ambient noise is hard to determine. Observing that the noise level at our measurement equipment is a little higher than the theoretical ambient noise, in this work, we set the threshold as 6 dB above the minimum received signal power recorded in an observed band during 24 hours over 12 days. V. S PECTRUM O CCUPANCY R ESULTS In this section, we present the band by band spectrum occupancy results, where each of the spectrum occupancy plots consists of three subplots. The upper sub-plot depicts the maximum received power at the antenna versus frequency measured during the whole 12-day period. For each frequency point, the maximum received power is determined by taking maximum of all 1248 received samples of that frequency point. The middle sub-plot shows the spectrum occupancy versus time and frequency. The occupancy is determined as follows: for each frequency point over 104 time instances, the maximum value of 12 received samples is taken; if this maximum value is greater than the threshold, then that frequency band is declared as occupied and it is shown by a red dot in the middle-plot. Note that, the threshold for a particular band is derived from the minimum of the received power levels in all frequency points of that band. The lower sub-plot shows the duty cycle versus frequency, where the duty cycle is dened as the fraction of time the signal is above the threshold. The average duty cycle

of a band is dened as average duty cycle = N 104 Nf

where N is the total number of frequency points (based on the middle sub-plot) with the maximum received signal exceeding the threshold in a 24-hour period and Nf is the total number of frequency points in the observed band. Before moving to our band by band spectrum occupancy analyses, in Fig. 4, we show the received power versus frequency plot for the whole frequency range of the measurement study (80 MHz to 5850 MHz). This plot show that in many bands, the maximum received power level is even below the equipment noise level which is around -103 dBm. Although, it indicates that overall spectrum utilization in the whole frequency range of our study might be very low, it does not give us a detail picture of how spectrum is utilized in different bands allocated to different services. Therefore, for a better view of the band by band occupancy pattern, we zoom into some selected bands. Fig. 5 shows occupancy for 80 to 174 MHz band. This band is mainly allocated to FM radio, xed/mobile (land), primary and secondary radars, and maritime navigation [14]. Average duty cycle of this band is 34.84%. In Fig. 6 and Fig. 7, we show the occupancy in the bands 174 to 230 MHz and 490 to 614 MHz. Both bands are allocated to broadcasting (analogue TV, digital TV, HDTV, DAB) services [14] and as can be seen, they are most heavily utilized bands observed in this study with average duty cycle 49.05% and 52.35%, respectively. Notice that these two bands have some spectrums where signals coming from some TV channels, e.g. ch6, ch9, ch10, ch11 and ch12, are very weak. If we consider 174 to 230 MHz band, we can easily see that two third of this band is unoccupied due to the weak reception of the signal broadcasting from those TV stations. Also note that two local TV channels namely Suria (ch12) and Central (ch5) provide some opportunities for secondary usage since Suria is closed from 12 am to 9 am and Central is closed from 12 am to 6 am. Another broadcasting band, which is mainly allocated to TV band V (ch39 to ch60) is highly

PMR

FM Radio Band II

Aeronautical Mobile/Radionav

Maritime, Amateur, PMR & Paging

Unused Unused

Broadcasting
DVB-T Unused

90

100

110

120 130 140 Freqeuency (MHz)

150

160

170

Power (dBm)

-20 -40 -60 -80 -100 -120 80

Power (dBm)

Ch 24 Central

TV Mobile

Ch 26

Ch 28 Ch U

Ch 30 Ch I

Ch 31

Ch 32 CNA

Ch 33

Ch 34

Ch 35

threshold = -94.5 dBm

90

100

110

120 130 140 Frequency (MHz)

150

160

170

Time Instants

06:00h 00:00h 18:00h 12:00h 06:00h 80 1

Time Instants

-20 -40 -60 -80 -100 -120

Ch 38

threshold = - 101.5 dBm

500 520 540 560 Freqeuency (MHz) 580 600

06:00h 00:00h 18:00h 12:00h 06:00h

500

520

Duty Cycle

540 560 Frequency (MHz)

580

600

Duty Cycle

0.5 0 80

average duty cycle = 0.3484

90 100 110 120 130 140 Frequency (MHz) 150 160 170

1 0.5 0

average duty cycle = 0.5235

500 520 540 560 Frequency (MHz) 580 600

Fig. 5: 80 MHz to 174 MHz.

Broadcasting
Ch 5 Ch 6 DAB Ch 8 Ch 9 Ch 10 Ch 11 Ch 12

Fig. 7: 490 MHz to 614 MHz.

Broadcasting (from foreign stations) Broadcasting TV BandGovernmentCh 60 V: Ch 39 to Services

threshold = -98.9 dBm

175 180 185 190 195 200 205 210 Freqeuency (MHz) 215 220 225 230

Power (dBm)

-20 -40 -60 -80 -100 -120

Power (dBm)

-20 -40 -60 -80 -100 -120

threshold = -102.7 dBm

620 640 660 680 700 720 Freqeuency (MHz) 740 760 780

Time Instants

06:00h

175

180

185

190

195 200 205 210 Frequency (MHz)

215

220

225

230

Time Instants

00:00h 18:00h 12:00h 06:00h

06:00h 00:00h 18:00h 12:00h 06:00h

620

640

660

Duty Cycle

1 0.5 0

680 700 720 Frequency (MHz)

740

760

780

average duty cycle = 0.4905

175 180 185 190 195 200 205 210 Frequency (MHz) 215 220 225 230

Duty Cycle

1 0.5 0

average duty cycle = 0.0881

620 640 660 680 700 720 Frequency (MHz) 740 760 780

Fig. 6: 174 MHz to 230 MHz.

Fig. 8: 614 MHz to 790 MHz.

unoccupied. From Fig 8, we can see that average duty cycle of this band is 8.81% and from 720 to 790 MHz, this band has almost zero utilization. Next, we observe the spectrum usage of the bands allocated for digital cellular services, GSM and 3G services. Fig. 9 shows the spectrum usage of 824 to 890 MHz band allocated for digital cellular services, e.g. trunked radio services and mobile data services [14]. Average duty cycle of this band is 27.24%. Fig. 10 and Fig. 11 show the occupancy for the GSM900 and GSM1800 bands in Singapore [14], respectively. These two bands have average occupancy of 38.11% and 19.82%, respectively. Note that, the occupancy gures in Fig. 10 and Fig. 11 between the uplink and downlink sides are not identical. Similar occupancy pattern has been observed in the uplink and downlink bands of 3G services as can be seen in Fig. 12. This result can be explanined as follows. The control channels for GSM 900, GSM 1800 and WCDMA are constantly being broadcasted by the base stations on the DL thus the DL for these frequencies seems fully occupied as they are always transmitting with relatively high

power. The uplink, on the other hand, for cellular systems is based on active users communicating through the network. If there is no active communication, there are still some periodic short-pulse transmissions on the uplink for location updating procedures which are too short to be picked up by the analyzer. Also note that GSM 900 mobile stations have a higher transmit power than on GSM 1800, which probably explains the higher power picked up by the analyzer. From Fig. 12, it is observed that 3G uplink is totally unoccupied. The reason behind this outcome is that WCDMA is a spread spectrum system where the signal is modulated over a larger bandwidth to give a very low output transmission power, which might not be detectable with the analyzer and thus does not show any occupancy on the 3G uplink. Therefore, from the captured results, we cannot conclude that no 3G user is active during the measurement period. Fig. 13 shows the spectrum usage by well-known unlicensed Industrial, Scientic and Medical (ISM) band (2400 to 2500 MHz) and wireless broadband access (WBA) (2500 to 2700 MHz) services. Looking at Fig. 13, it appears that the whole

DECT & TDD

Digital Cellular Services Digital Cellular Services

3G

Fixed/Mobile

3G

Fixed/Mobile

WBA

Power (dBm)

Mobile Fixed

Mobile Data, PMR & Trunking

-20 -40 -60 -80 -100 -120

Power (dBm)

threshold = -101.3 dBm

830 840 850 860 Freqeuency (MHz) 870 880 890

-20 -40 -60 -80 -100 -120

threshold = -100.2 dBm

1900 1950 2000 2050 2100 2150 2200 2250 Freqeuency (MHz) 2300 2350 2400

06:00h 00:00h 18:00h 12:00h 06:00h

Time Instants

06:00h 00:00h 18:00h 12:00h 06:00h

Time Instants

1900

1950

2000 2050

830

840

850 860 Frequency (MHz)

870

880

890

2100 2150 2200 2250 Frequency (MHz)

2300

2350

2400

Duty Cycle

1 0.5 0

Duty Cycle

1 0.5 0

average duty cycle = 0.2724

830 840 850 860 Frequency (MHz) 870 880 890

average duty cycle = 0.0355

1900 1950 2000 2050 2100 2150 2200 2250 Frequency (MHz) 2300 2350 2400

Fig. 9: 824 MHz to 890 MHz.

Fig. 12: 1880 MHz to 2400 MHz.

ISM WBA

RFID

GSM900 uplink
-20 -40 -60 -80 -100 -120 890 Power (dBm)

GSM900 downlink

-20 -40 -60 -80 -100 -120 2400

Power (dBm)

threshold = -98.5 dBm

2450 2500 2550 2600 Freqeuency (MHz) 2650 2700

900

910

920 930 Freqeuency (MHz)

940

950

960

Duty Cycle

06:00h 00:00h 18:00h 12:00h 06:00h 890 1 0.5 0 890

Time Instants

Time Instants

threshold = -101.3 dBm

06:00h 00:00h 18:00h 12:00h 06:00h 2400 1 0.5 0 2400

2450

2500

2550 2600 Frequency (MHz)

2650

2700

900

910

920 930 Frequency (MHz)

940

950

960

average duty cycle = 0.0000

2450 2500 2550 2600 Frequency (MHz) 2650 2700

Duty Cycle

average duty cycle = 0.3811

900 910 920 930 Frequency (MHz) 940 950 960

Fig. 13: 2400 MHz to 2700 MHz.

Fig. 10: 890 MHz to 960 MHz.

GSM 1800
-20 -40 -60 -80 -100 -120 Power (dBm)

GSM 1800

threshold = -100.1 dBm

1720 1740 1760 1780 1800 1820 Freqeuency (MHz) 1840 1860 1880

06:00h 00:00h 18:00h 12:00h 06:00h

Time Instants

band of 2400 to 2700 MHz is completely unused at this point. Similar occupancy pattern has been observed in the xed satellite band from 3400 to 4200 MHz (Fig. 14) and the ISM band from 5755 to 5875 MHz (Fig. 15). The occupancy estimates of these bands might not be the representatives of the actual occupancy. These results could be explained as follows: 1) WBA signals might not be detected at the measurement point if the access points are not close enough, 2) satellite signal power might be much lower than the ambient noise when it reaches the ground, and 3) short wavelength of ISM signals cannot penetrate through walls. VI. M AJOR O BSERVATIONS In this section, we summarize the major observations from the occupancy results of this study and identify the channels for long-term studies in order to provide the policy makers with the necessary information for taking proper initiative to facilitate dynamic spectrum access technologies such as cognitive radio. A graphical presentation of the band by band average spectrum occupancy as well as the average spectrum occupancy for the

1720

1740

1760

1780 1800 1820 Frequency (MHz)

1840

1860

1880

Duty Cycle

1 0.5 0

average duty cycle = 0.1982

1720 1740 1760 1780 1800 1820 Frequency (MHz) 1840 1860 1880

Fig. 11: 1710 MHz to 1880 MHz.

Fixed-Satellite
-20 -40 -60 -80 -100 -120 3400

threshold = -101.5 dBm

3500 3600 3700 3800 3900 Freqeuency (MHz) 4000 4100 4200

06:00h 00:00h 18:00h 12:00h 06:00h 3400 1 0.5 0 3400

3500

3600

3700 3800 3900 Frequency (MHz)

4000

4100

4200

Duty Cycle

average duty cycle = 0.0000

3500 3600 3700 3800 3900 Frequency (MHz) 4000 4100 4200

Aero Nav, ISM, others: 50005850 MHz Aero Nav, Fixed Sat: 42005000 MHz Fixed satellite: 34004200 MHz Aero Nav, Radiolocation: 27003400 MHz ISM, WBA: 24002700 MHz DECT, TDD, 3G, Fixed/Mobile, WBA: 18802400 MHz GSM1800, Fixed: 17101880 MHz Mobile Sat, Met Sat, Aero Nav: 15251710 MHz Fixed/Mobile, DAB: 14291525 MHz Aero Nav, Radar: 9601429 MHz GSM900, RFID: 890960 MHz Digital Cellular, Mobile Data, PMR, Trunk: 824890 MHz Mobile Data, PMR, Trunk, Fixed, others: 790824 MHz TV: 614790 MHz TV, DVBT: 490614 MHz Mobile Data, PMR & Trunk: 406490 MHz Fixed/Mobile, PMR, Aero, others: 230406 MHz TV, DAB: 174230 MHz FM Radio, Aero, Fixed/Mobile, others: 80174 MHz The whole bandwith of study: 805850 MHz 0 25 50 75 100 Average occupancy (%)

Time Instants

Power (dBm)

Fig. 14: 3400 MHz to 4200 MHz.

Fixed/ Radiolocation

Fig. 16: Band by band average spectrum occupancy in Singapore.

ISM

Aeronautical Radionavigation

Aeronautical Radionav

Maritime Radionavigation

-20 -40 -60 -80 -100 -120 5000

Power (dBm)

threshold = -100.6 dBm

5100 5200 5300 5400 5500 Freqeuency (MHz) 5600 5700 5800

06:00h 00:00h 18:00h 12:00h 06:00h 5000 1 0.5 0 5000

5100

5200

5300 5400 5500 Frequency (MHz)

5600

5700

5800

average duty cycle = 0.0008

5100 5200 5300 5400 5500 Frequency (MHz) 5600 5700 5800

Fig. 15: 5000 MHz to 5850 MHz.

whole band of study are shown in Fig. 16. Note that due to space constraint, we do not provide the occupancy plots for some bands that are shown in Fig. 16, where the average occupancy for the whole 80 MHz to 5850 MHz band is determined as follows. First, we determine the average spectrum usage (in MHz) of each band shown in Fig. 16 by mulitplying the average occupancy of each band by its corresponding bandwidth. For example, average spectrum usage of 80 to 174 MHz band is 0.3484 (174 80) = 32.7496 MHz. Then, summing the spectrum usage of all bands and dividing it by the total availabe bandwidth 585080 = 5770 MHz, we get the average spectrum occupancy of 4.54% for the whole frequency bands of study. Now using Fig. 16 and the results of Section IV, the following useful information can be extracted: Highest occupancy has been observed in the broadcasting bands and GSM900 bands. The bands allocated for GSM1800, radar, and digital cellular services (paging, mobile data, trunked radio) have moderate occupancy.

Low occupancy has been observed in the bands allocated for xed/mobile services, and primary and secondary radar. The bands allocated for aeronautical radionavigation, xed satellite, WBA and ISM appear to have no utilization. Frequencies above 1GHz are relatively underutilized except for the cellular bands. In some bands, e.g. 174 to 230 MHz, frequency utilization is normally higher during the day time compared to the night time due to the end of transmission from some TV channels during the night time. Although some frequencies in some bands are appeared to be used, the duty cycle is quite low. From the above observation, we can easily identify some channels that might be of interest for future long-term studies. All the bands starting from 980 MHz to 5850 MHz except for the cellular bands are of particualr interest. Among them, satellite, WBA and ISM bands might require the measurement equipment, methodology and place to be changed to determine more accurate occupancy estimate. However, the bands which are allocated for aeronautical radionavigation, radiolocation, primary radar, and secondary radar can be identied as opportunity channels. In higher frequency bands, most of these services are allocated with the bands 980-1429 MHz, 2700-3400 MHz, 4200-5000 MHz, and 5000-5850 MHz. In lower frequency bands, these services operate at the bands from 230 MHz to 490 MHz. Some other spectrum of interest might be 174-230 MHz and 614-790 MHz bands. As we can see, 2/3 of the spectrum of 174-230 MHz band and almost all of 614-790 band are unoccupied most of the times and could be considered for opportunistic use. Finally, the spectrum allocated to Singapore TV channels Central and Suria are not occupied during the night time and could be considered for other types of services during those periods.

Time Instants

Duty Cycle

VII. C ONCLUSION In this work, we studied the spectrum usage pattern in Singapore for the frequency bands ranging from 80 MHz to 5850 MHz. Our measurement results suggest that except for

the frequency bands allocated for broadcasting and cell phones, most of the allocated frequencies are heavily underutilized. The average occupancy for the whole range of frequency of this study was found to be only 4.54%. However, the spectrum survey measurements contained in this report cannot be solely used to assess the feasibility of using alternative services or systems in a band. To assess the feasibility of using alternative services, further occupancy studies need to be performed, especially in those bands which are identied as less utilized. These long term studies would enable us to indentify the seasonal trends and potentially longer term trends in band by band spectrum utilization. For the case of bands where the signal were not detected due to the low power transmission, the sensitivity of the receivers that operate in those bands should be taken into account to determine the spectrum occupancy threshold to get more accurate occupancy estimate. In order to assess the variations in spectrum usage in the environments with different user proles, different population densities and different geographic characteristics, parallel measurements should carried out in different places both indoor and outdoor over an extended period of time. These future studies would help us to more accurately identify the frequency bands with low or no active utilization and open the door for employing new spectrum sharing technologies such as cognitive radio technology. R EFERENCES
[1] Federal Communications Commission, Spectrum Policy Task Force, nov 2002, rep. ET Docket no. 02-135. [2] I. F. Akyildiz, W.-Y. Lee, M. C. Vuran, and S. Mohanty, NeXt generation/dynamic spectrum access/cognitive radio wireless networks: A survey, Computer Networks Journal. [3] M. A. McHenry, P. A. Tenhula, D. McClosky, D. A. Roberson, and C. S. Hood, Chicago spectrum occupancy measurements & analysis and a longterm studies proposal, in First International Workshop on Technology and Policy for Accessing Spectrum (TAPAS06), Boston, MA, Aug. 2006. [4] F. H. Sanders, B. J. Ramsey, and V. S. Lawrance, Broadband spectrum survey at San Francisco, CA, NTIA, May 1995, NTIA Report 99-367. [5] F. H. Sanders and V. S. Lawrance, Broadband spectrum survey at Denver, Colorado, NTIA, Sept. 1995, NTIA Report 95-321. [6] F. H. Sanders, B. J. Ramsey, and V. S. Lawrance, Broadband spectrum survey at San Diego, CA, NTIA, Dec. 1996, NTIA Report 97-334. [7] M. A. McHenry and K. Steadman, Spectrum occupancy measurements, Location 1 of 6: Riverbend Park, Great Falls, Virginia, April 7, 2004, Shared Spectrum Company, Aug. 2005. [8] , Spectrum occupancy measurements, Location 2 of 6: Tysons Square, Center, Vienna, Virginia, April 9, 2004, Shared Spectrum Company, Aug. 2005. [9] M. A. McHenry and S. Chunduri, Spectrum occupancy measurements, Location 3 of 6: National Science Foundation Building Roof, April 16, 2004, Revision 2, Shared Spectrum Company, Aug. 2005. [10] M. A. McHenry, D. McCloskey, and G. Lane Roberts, Spectrum occupancy measurements, Location 4 of 6: Republican National Convention, New, York City, New York, August 30, 2004 - September 3, 2004, Revision 2, Shared Spectrum Company, Aug. 2005. [11] M. A. McHenry and K. Steadman, Spectrum occupancy measurements, Location 5 of 6: National Radio Observatory (NRAO), Green Bank, West Virginia, October 10 - 11, 2004, Revision 3, Shared Spectrum Company, Aug. 2005. [12] M. A. McHenry, D. McCloskey, and J. Bates, Spectrum occupancy measurements, Location 6 of 6: Shared Spectrum Building Roof, Vienna, Virginia, December 15 - 16, 2004, Shared Spectrum Company, Aug. 2005. [13] Radiocommunication Bureau, HANDBOOK Spectrum Monitoring, International Telecommunication Union (ITU), 2002. [14] Spectrum Management Handbook, Info-communication Development Authority (IDA) of Singapore, Feb. 2002, Issue 1 Rev 2.