You are on page 1of 6

A COMPARISON OF GEO-LOCATION AND SPECTRUM SENSING IN COGNITIVE RADIO

Stephen J. Shellhammer

Qualcomm Inc.
5775 Morehouse Drive
San Diego, CA 92121
sshellha@qualcomm.com

ABSTRACT terial is required to understand the geo-location and spec-


trum sensing methods. Background on propagation models
A comparison is made between the FCC geo-location and
for electromagnetic field strength are provided in Section 2.1
spectrum sensing requirements in the TV white space. Back-
while how electromagnetic field strength is converted into re-
ground material on propagation models and conversion be-
ceive power at the spectrum sensor is given in Section 2.2.
tween electromagnetic field strength and receive power is pro-
The geo-location requirements for operation in the TV white
vided. The geo-location and spectrum sensing requirements
space are described in Section 3 while the spectrum sensing
are both converted to field strength where they can be com-
requirements are provided in Section 4. A comparison of geo-
pared, and a typical TV broadcast example is provided to
location and spectrum sensing is given in Section 5. After we
allow comparisons in terms of the physical distances of the
understand this comparison, it is interesting to make observa-
protection contour, the geo-location contour and the spectrum
tions about the conclusions in the FCC test report. These ob-
sensing contour. Observations are made about the FCC tests
servations are contained in Section 6. Some alternative meth-
of the TV white space prototype devices and the interpretation
ods of testing spectrum sensing devices are described in Sec-
of those results.
tion 7.
Index Terms— Cognitive Radio, Spectrum Sensing, Geo-
location, TV White Space
2. BACKGROUND
1. INTRODUCTION
In order to appreciate the relationship between geo-location
and spectrum sensing for protection of TV broadcast services
In November 2008 the United States Federal Communication
it is necessary to understand both the relationship between
Commission (FCC) issued a report and order (R&O) [1] pro-
transmit power and electromagnetic field strength at a dis-
viding the rules under which a cognitive radio network can ac-
tance and how that field strength is converted into receive
cess unused TV spectrum, referred to as TV white space. Prior
power. These two concepts are reviewed in this section. It is
to issuing the R&O the FCC performed extensive tests of
more common to combine both of these effects into a single
devices performing spectrum sensing and geo-location with
effect which relates the receive power to the transmit power.
database access. After completing these tests the FCC in-
This relationship is called the path loss. However, there are
cluded requirements in the R&O for both geo-location with
certain advantages to separating the two effects, and this is
database access and spectrum sensing for protection of TV
the method generally employed by the broadcast industry and
broadcast services. The FCC test report includes evaluations
also by the FCC.
of the effectiveness of geo-location with database access and
spectrum sensing. Some of the report’s conclusions com-
pared the performance of spectrum sensing to geo-location. 2.1. Propagation
Since these two techniques perform utilize different methods
of protecting TV broadcast services, it is interesting to under- There are many models for the conversion from transmit power
stand the difference between these two methods. This paper to electromagnetic field strength at a distance. We will be-
provides this comparison between geo-location and spectrum gin with the free space model and then introduce an empiri-
sensing. cal model developed by the International Telecommunication
In order to perform this comparison some background ma- Union (ITU). From [2] we have that for free space propaga-

978-1-4244-4581-3/09/$25.00 ©2009 IEEE


tion that, are typical values for TV transmission in the UHF frequency
PT GT E2 band. In the ITU model the receive antenna is assumed to
= (1)
4π d 2 RFS be 10 meter high, which is typical for a TV receive antenna
Where PT is the transmit power, GT is the transmit antenna on the roof of someone’s home. Figure 1 illustrates the free
gain, d is the distance in meters, E is the electric field strength space propagation and the ITU model, in the UHF frequency
and RFS is impedance of free space and is given by RFS = 120π . band, for these parameters.
Solving for the field strength we get,
Electromagnetic Field Strength
30PT GT 120
E =
2
(2)
d2 100

Field Strength dBu


Taking logarithms and multiplying by 10 we have, 80

60
20 log(|E|) = 10 log(PT )+10 log(GT )+10 log(30)−20 log(d)
(3) 40 Free Space
We would like to represent the field strength in dBu (dB mi-
20 ITU Model
crovolts/meter), the transmit power in dBm, and the distance
in kilometers. To make these conversions we let, 0
10 20 50 100 200 500
Distance km
FSdBu = 20 log(|E|) + 120 (4)

PdBm = 10 log(PT ) + 30 (5) Fig. 1. Electromagnetic field strength for free space and ITU
F(50,90) propagation models
20 log(d  ) = 20 log(d) − 60 (6)
This gives the following formula for the electromagnetic field
strength in dBu for free space propagation,
2.2. Receive Power
FSdBu = PdBm + 10 log(GT ) + 10 log(30) + 30 − 20 log(d)
(7) In order to compare geo-location and spectrum sensing we
After simplification we get, need to be able to relate field strength to receive power at
spectrum sensor. Once again we can find what we need in
FSdBu = PdBm + 10 log(GT ) + 44.77 − 20 log(d) (8) [2],
|E|2
Now this formula gives us the field strength for free space PR = AE (9)
RFS
propagation. However, we rarely attain free space propaga-
Where once again RFS is the impedance of free space and is
tion in practice. For this reason, the ITU [3] has developed an
equal to 120π and AE is the effective area of the receive an-
empirical model for propagation which includes a number of
tenna. For a wavelength of λ we can relate the receive antenna
factors including transmit antenna height.
gain to its effective area by,
The ITU model provides a set of propagation curves. They
are all normalized for an effective radiated power (ERP) of 60 4π
dBm. The ERP is the effective power normalized to a dipole GR = AE (10)
λ2
antenna. These sets of propagation curves are identified as
F(X,Y ) curves, where X specifies a spatial percentage and Y Combining we obtain a formula for the receive power,
specifies a temporal percentage. The F(X,Y ) is a two dimen-
|E|2 GR λ 2
sional complimentary cumulative distribution function with PR = (11)
RFS 4π
parameters X and Y . If the propagation value is F(X,Y ) then
this tells you that at X percent of the locations and Y per- Taking logarithms and multiplying by 10 gives,
cent of the time the actual field strength it above that value.
For example, the F(50, 50) is the median field strength. The 10 log(PR ) = 20 log(|E|) + 10 log(GR ) + 20 log(λ )
F(90, 90) is the value at which 90% of the locations and 90% −10 log(RFS ) − 10 log(4π ) (12)
of the time the actual field strength is higher than that value.
We can compare the free space propagation curve with Which simplifies to,
the ITU propagation model. Let us set the ERP at 300 kwatt
(84.77 dBm) and the antenna height of 600 meters. These PdBm = FSdBu + GdB + 20 log(λ ) − 126.76 (13)
f (MHz) λ (m) K (dB) We can see that in a typical deployment example as provided
512 .586 131.4 in Section 2.1 the radius of the protection contour is around
125 km.
605 .496 132.9
The FCC geo-location requirements specify how far out-
698 .430 134.1 side the protection contour that a TV band device must be in
order for it to be permitted to transmit. The distance beyond
Table 1. Constant for Conversion from Field Strength to Re-
the contour depends on the antenna height of the TV band de-
ceive Power
vice, since the higher the TV band device antenna the more
interference it can cause to a TV receiver. For co-channel op-
eration the distance is 6 km for antennas less than 3 meters
Where PdBm is the receive power in dBm, FSdBu is the field high, 8 km for antennas between 3 and 10 meters high, and
strength in dBu, GdB is the receive antenna gain in dB, and λ 14.4 km for antennas between 10 and 30 meters high. So in
is the wavelength in meters. the worst case condition the additional distance beyond the
Now it is useful to evaluate this formula in the TV white protected contour is 14.4 km.
space spectrum. We will focus on the UHF frequency band
For our example from Section 2.1 the TV band device
where both fixed and portable devices can operate. A similar
must be approximately 140 km (≈ 125 + 14.4) from the TV
analysis can be performed in the VHF frequency band. Ac-
broadcast antenna. The field strength at that distance is ap-
cording to the FCC R&O [1] the channels that permit both
proximately 36 dBu. So we see that the field strength is at
fixed and portable devices are Channels 21-51. Now a few
around 5 dB weaker, for the worst cast distance of 14.4 km.
of those channels are excluded, like Channel 37, but we are
For the 6 km distance the field strength is only about 2 dB
just interested in the frequency range. Channels 21-51 span
weaker, than at the protection contour.
the UHF frequency band from 512 MHz to 698 MHz. We
will consider both ends of this band of channels and also the
middle of the band, which is at 605 MHz.
4. SPECTRUM SENSING REQUIREMENTS
Let use define the conversion constant between field strength
and receive power for a 0 dBi antenna as,
Unlike the geo-location requirements the spectrum sensing
K = 126.78 − 20 log(λ ) (14) requirements are in terms of the receive power of the signal,
not the field strength. For digital TV, analog TV and wireless
So for a 0 dBi receive antenna the receiver power is then just microphones the spectrum sensing requirement is to sense
the field strength minus K. For these three frequencies we down to −114 dBm, utilizing a 0 dBi receive antenna. This re-
tabulate the values of K in Table 1. ceive power is well below the noise floor of a typical receiver.
So we see over the frequency range of interest the constant The thermal noise in 6 MHz bandwidth is −106.2 dBm. If
varies from 121.4 dB to 134.1 dB, with a mean value around we assume a conservative receiver noise figure of 10 dB, then
133 dB, which we will use in this paper. the noise power is approximately −96 dBm. Thus the receive
power of −114 dBm is approximately 18 dB below the noise
floor. Sensing at this power level can be challenging; how-
3. GEO-LOCATION REQUIREMENTS
ever, it is possible to sense for ATSC at this power level [4].
One of the methods for protecting licensed services required The FCC does not specify the probability of detection re-
by the FCC R&O [1] is geo-location with database lookup. quired at the −114 dBm but it is expected that a detection
In this paper we will focus on the digital TV (ATSC) require- probability of at least 0.99 will be required when the FCC
ment in the UHF frequency band. In the US all high-power completes their test plan.
TV broadcasts are converting to digital TV, so it makes most In order to ensure detection at −114 dBm it is likely that at
sense to focus on that case. spectrum sensor will need to detect even lower, say to −116
The FCC defines a protection contour for ATSC around dBm or lower, to alow some margin to compensate for the
a TV tower as a contour of locations at which a propaga- effects of additive white Gaussian noise (AWGN) and multi-
tion model predicts that the F(50,90) field strength is 41 dBu. path fading. One cannot build a detector that detects perfectly
There are a number of such propagation models, but in this down to −114 dBm and not at all below −114 dBm due to the
paper we will use the ITU model [3] described earlier in Sec- effects of AWGN and multipath fading. In this paper we do
tion 2.1. This contour is a circle with a given radius centered not consider a specific type of spectrum sensing technique.
around the TV tower. The radius is given by the distance at For a survey of the spectrum sensing techniques for ATSC
which the F(50, 90) curve gives a field strength of 41 dBu. considered by the IEEE 802.22 working group, see [5].
Field Strength (dBu) Distance (km)
Protection 41 125
Contour Protection Contour
Geo-location 36 ∼ 39 131 ∼ 140
Contour
Geo-location Contour
Spectrum 19 218
Sensing
Contour

Table 2. Comparison of Protection, Geo-location and Spec- Spectrum Sensing Contour

trum Sensing Contours

Fig. 2. Comparison of Protection, Permitted Transmission


5. COMPARISON and Spectrum Sensing Contours

In order to make a comparison between the geo-location and


spectrum sensing requirements we must convert between field
strength and receive power. Since the propagation curves pro- contour. Thus it is expected that spectrum sensing will detect
vide field strength it is most natural to convert everything to an ATSC signal well beyond the geo-location contour.
field strength. The geo-location requirement is already in field The other comparison between geo-location and spectrum
strength so all we need to do is to convert the spectrum sens- sensing is that geo-location is based on a propagation model
ing requirement from receive power to field strength and then for the field strength while spectrum sensing is based on the
we can make our comparison. actual field strength. Thus beyond the large differences in
The -114 dBm spectrum sensing requirement can be con- the field strengths associated with geo-location and spectrum
verted to a field strength requirement using the conversion sensing, there is also the differences between a propagation
provided in Section 2.2. The receive power is given by, model and the actual propagation.
The final comparison between the geo-location protection
PdBm = FSdBu − K (15) contour and the spectrum sensing contour is that of variation.
The geo-location accuracy requirement is 50 meters which is
where we use a 0 dBi receive antenna and K is given in Ta- very accurate. The spectrum sensing variation is measured
ble 1. If we use the typical value of K = 133 then the receive in dB and not in meters. The variation depends on specific
power of -114 dBm corresponds to a field strength of 19 dBu. sensing technique employed. For a given sensing technique
We compare this to the field strength at the protection con- there are two factors that lead to variation in the sensing: ad-
tour of 41 dBu and the field strength at the geo-location con- ditive white Gaussian noise (AWGN) and multipath fading.
tour (contour within which the TV band can not transmit co- The effect of AWGN can be reduced by using longer sensing
channel) of between 36-39 dBu. If we consider the example times. The effect of multipath fading can be reduced by using
given in Section 2.1 the distance at which the field strength spatial diversity. However, there is always some variation in
is 19 dBu according to the F(50, 90) propagation curve is sensing due to these effects. A good sensing technique may
218 km. In Table 2 we summarize the field strength and cor- be able to limit this variation to a few dB, say between 2 and
responding distance for the protection contour, geo-location 5 dB. It is not likely that the sensing variation can be made
contour and the spectrum sensing contour. to be much less than 2 dB. Of course, to make more precise
We see that the spectrum sensing requirement requires the statement about the variation in sensing one needs to consider
sensor to detect the TV transmission at approximately 175% a specific sensing technique. In order to compare the varia-
of the radius of the protection contour. If we consider that tion in geo-location and sensing we can once again relate it to
the spectrum sensor is likely to detect the ATSC signal below variation in the field strength. The 50 meter accuracy in the
-114 dBm since there is likely at least several dB of margin geo-location contour equates to much less than a 1 dB varia-
built into the design we can expect the spectrum sensor to tion in field strength. So the variation in geo-location contour
detect up to twice as far as the protection contour. The size of is much less than the 2 to 5 dB variation in the sensing con-
these contours is illustrated in Figure 2. tour.
We see that the spectrum sensing contour is significantly Another summary of the comparison between the protec-
larger than either the protection contour or the geo-location tion contour, the geo-location contour and the sensing con-
Field Field Variation false alarms. Given our understanding of the relationship be-
Strength Strength (dB) tween the protected contour, the geo-location contour and the
Source (dBu) spectrum sensing contour we can see that detections outside
Protection Propagation 41 1 the protected contour and within the spectrum sensing con-
Contour Model tour are in face true detections and not false alarms.
Geo-location Propagation 36 ∼ 39 1
Contour Model 7. ALTERNATIVE METHODS OF TESTING
Spectrum Actual Field 19 2∼5 SPECTRUM SENSING DEVICES
Sensing Strength
Having commented on the FCC testing of the spectrum sens-
Contour
ing prototypes it useful to consider alternative methods of
Table 3. Final Comparison of Protection, Geo-location and testing spectrum sensing devices. The challenge in testing
Spectrum Sensing Contours these devices on over-the-air signal is that it is difficult to
know the signal power when the power level is well below
the noise floor, as it is at a power level of −114 dBm. If
one can estimate the receive power then one can measure the
tour are provided in Table 3, which consider the whether the probability of detection at a know receive power level.
field strength is based on a propagation model or the actual There are two methods that can be used to estimate the
field strength, and also summarizes the accuracy of the field signal power. The first method is to use an additional di-
strength for each of the contours. rectional antenna at the same location as the spectrum sens-
ing device and to point the directional antenna directly at the
6. OBSERVATIONS ON FCC TESTS transmitter location. If the antenna gain is large enough and
is a know value then the output power of the antenna can be
When performing testing on prototype devices the FCC com- brought above the noise level. Also, a low noise spectrum
pared the performance of the spectrum sensing in the proto- analyzer can be used to measure the signal power. Then the
types to geo-location. In this section we will make a few ob- estimate of the receive power in the spectrum sensing device
servations on those test results based on what we know based is just the signal power measured by the spectrum analyzer
on our comparison of geo-location and spectrum sensing. The attached to the high gain antenna minus the antenna gain.
results of the FCC tests can be found in [6]. The second method of estimating the signal power at the
The FCC office of engineering and technology (OET) per- spectrum sensor is to make the measurement at a location at
formed testing on the spectrum sensing prototypes both inside which the signal power is above the noise floor and so the sig-
and outside the protected contour (service contour). In addi- nal power can be estimated accurately. Then an attenuator is
tion they tested to see if the TV signal could be viewed us- placed between the antenna and the spectrum sensing device
ing a digital TV receiver. Within the protection contour the to lower the signal power seen by the spectrum sensing de-
TV signal is not always viewable and outside the contour the vice to the desired power level. By using a know attenuation
signal is sometime viewable. This is due to several factors. level the signal power at the spectrum sensing device can be
One factor is that the actual propagation is not the same as accurately estimated even a signal levels well below the noise
the propagation model. The other factor is there is margin floor.
included in the protected contour. If one uses a variable attenuator it is also possible to mea-
The performance of the various prototypes differed based sure the probability of detection as a function of the signal
on their design. Within the protected contour some of the power level. Since the FCC requirement is to detect at −114
prototypes performed quite well with probability of detection dBm one can plot the probability of detection versus signal
as high as 100%. However, the devices with high probabil- power over the range from −124 dBm to −104 dBm and pro-
ity of detection within the protected contour, when tested out- vide a probability of detection curve using over-the-air sig-
side the protected contour, also detected the ATSC signal with nals. In this way it is possible to evaluate the accuracy of the
high probability. The FCC classified these detections as false spectrum sensing device on actual over-the-air signals.
alarms. We do not know how far outside the protected con-
tour the FCC tested but if they tested outside the protected 8. CONCLUSIONS
contour but within the spectrum sensing contour it is not sur-
prising that the prototypes had a high probability of detection. We have developed a comparison of the TV protected con-
What is curious is that the FCC labeled those detections as tour, the geo-location contour and the spectrum sensing con-
tour. This comparison enables us to see the vast differences
in the FCC geo-location and spectrum sensing requirements.
We have observed that detection of the ATSC signal by a
spectrum sensor outside the protection contour but within the
spectrum sensing contour is in fact a true detection and not a
false alarm.

9. REFERENCES

[1] Federal Communication Commision, Second Report and


Order and Memorandum Opinion and Order In the Mat-
ter of Unlicensed Operation in the TV Broadcast Bands,
Additional Spectrum for Unlicensed Devices Below 900
MHz and in the 3 GHz Band, November 14, 2008, Docu-
ment 08-260.

[2] T. Rappaport, Wireless Communications: Principals and


Practice, Prentice Hall,, 1996.

[3] International Telecommunications Union, Method for


point-to-area prediction for terrestrial services in the fre-
quency range 30 MHz to 3000 MHz, 2003, Recomenda-
tion ITU-R P.1546-1.

[4] S. J. Shellhammer, A. K. Sadek, and W. Zhang, “Tech-


nical challenges for cognitive radio is TV white space
spectum,” UCSD Information Theory and Applications
Workshop, January 2009.

[5] S. J. Shellhammer, “Spectrum sensing in IEEE 802.22,”


Cognitive Information Processing Workshop, June 2008.

[6] Technical Research Branch Laboratory Division Office


of Engineering and Technology Federal Communications
Commission, Evaluation of the Performance of Proto-
type TV- Band White Space Devices Phase II, October
15, 2008, FCC/OET 08-TR-1005.

You might also like