A Low-Cost Directional Antenna for IEEE 802.

11b/g
Greg Berman, 41624939 Tim Everhardt, 80452656 Adam Knapp, 80282400 Noah Liebman, 80170631
University of Michigan, Ann Arbor 1236 EECS Building 1301 Beal Avenue Ann Arbor, MI 48109 Jamie D. Phillips, Advisor

A Low-Cost Directional Antenna for IEEE 802.11b/g

Contents
1 Introduction 2 Theory 3 Construction 3.1 3.2 Parts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 3 3 4 4 4 5 6 7 8 8 10 11 11

4 Testing 4.1 4.2 4.3 Network Analyzer Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Anechoic Chamber Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Field Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5 Experimental Results 6 Recommendations 7 Conclusion 8 Appendix - Field Testing Parameters Acknowledgments References

Abstract Though often built by enthusiasts, the body of scientically conducted research and analysis of the tin can antenna, sometimes referred to as a cantenna, is quite small. We set out to remedy this by providing theoretical analysis of the design, and conducting both laboratory testing and eld testing of the application: IEEE 802.11b-based wireless internet. Our design was successful, resulting in a gain of over 10 dB, and reasonable impedance match.

Introduction

Wireless computer networks have proliferated in recent years, allowing increased mobility and exibility for computer users. In addition to the necessary software, a reliable physical communications link must be established for the devices to function properly. A properly congured antenna is an important part of this link. While the built-in antennas of many devices are sucient for networking over short distances, their range is very limited. Commercial external directional antennas are available, but are very expensive, costing four to eight times as much as the wireless devices themselves. As a solution to this problem, amateur enthusiasts have experimented with a variety of antennas. They found that a cylindrical metallic antenna, commonly known as a cantenna, works well and can be made out of common parts [4][2]. However, their design is not based on rigorous theory. The purpose of this study is to examine and document the theory on which the antenna operates, test its operating characteristics in a laboratory setting, and evaluate its performance in the eld compared to commercially-available solutions.

Theory

Many know how to construct a cantenna, but not nearly as many understand the theory behind its operation. From our research, we were able to derive conditions from resources to give us a theoretical design basis and expected operational characteristics.[1][3] The basic design of the antenna is quite simple. It consists of a monopole ground-plane exciter that generates or receives waves, which are then guided through the cylindrical metallic waveguide. The waves then diract at the circular aperture, creating the antenna pattern. The rst step was to derive the guidance condition for the cylindrical metallic waveguide. To do this, Bessel functions were needed. The Bessel function is the solution to the radial component of the wave equation and hence describes the -dependent behavior of waves in a cylindrical geometry. The mathematical expression of a Bessel function of the rst kind is as follows: Jm (u) = j
m 2

ej(mv+u cos(v))
0

(1)

The guidance condition for the waveguide is determined by the boundary conditions on the electric eld. All components of the electric eld tangent to the inside of the can must approach zero at the metallic boundary, as is required for a metallic waveguide. For the cylindrical geometry, this condition is reached when the derivative of the Bessel function is set to 0 for the so-called transverse electric (TE) waves. For the so-called transverse magnetic (TM) waves, the condition is met when the Bessel function itself is set to 0. Symbolically, Jm (k a) = 0, where a corresponds to the interior radius of the waveguide and k is the component of the wave number, for the TE modes. First, the wave number equation is stated in cylindrical coordinates as kz =
2 k 2 k . Second, the

guidance condition gives the value of k a, which will now be referred to as , for TE modes and 1

for TM modes. After evaluating the Bessel function, it is found that value of for the rst-order Bessel mode is 1.8, which corresponds to the TE11 mode. The next mode has an value of 2.4, which corresponds to mode TM01 . Combining the above equations, the wave number equation becomes kz = 2 2 , with kz being a 2 in the axial direction. The axial wavenumber then gives us the guide wavelength, kz = g , which will become important later. A cuto frequency equation in terms of the mode and the radius of the waveguide can now be expressed. fc = g = c 2a 2 kz (2) (3)
2

A fascinating characteristic of the TE11 mode is that a plane wave forms at the center axis of the cylinder, which explains the linear polarization of the wave in the waveguide despite the guides physical geometry, as shown in Figure 1.

Figure 1: TE11 mode in cross-section of waveguide[3] Knowing that the operating frequency of a standard wireless network (IEEE 802.11b) is around 2.40 GHz and that only the single TE11 mode should propagate, the proper radius of the can for use in the antenna can be determined. We eventually found a can with a radius of 0.047 m, which is nearly ideal for TE11 mode propagation. The cuto frequency for the TE11 mode evaluates to 1.828 GHz, and the cuto frequency of the TM01 mode is 2.44 GHz. Having determined these parameters, we validated the widely available design for the antenna shown in Figure 2.

Figure 2: Schematic diagram of nished can

The exciter is a quarter-wave monopole which, using image theory, can be approximated as a halfwave dipole due to the conducting plane made up of the can below it. The quarter-wave monopole is also a good solution because of its length relative to the diameter of the can. The last aspect of the design is the placement of this exciter; it is placed a quarter guide-wavelength from the closed 2

end of the can. This placement ensures proper phase matching so the reection from the can provides constructive interference. Ideally, the distance from the exciter to the aperture should be n + where 4 4 n is an integer, so the aperture acts as a matched load. A larger n ensures that higher-order modes are attenuated. However, cans that meet these specications are not readily available. The last aspect of theory to be considered is what happens to the wave once it leaves the waveguide. This problem would be simple if the circular aperture was uniformly excited; however, as shown earlier in Figure 1, this is not the case across the whole cross-section. In the absence of advanced mathematical techniques and numerical methods to solve such a problem, we chose to calculate the far-eld radiation pattern assuming uniform excitation, and to later compare it to actual measurements to see if it was a valid assumption. The far-eld radiation pattern of a uniformly excited circular aperture is [1] 2J1 (ka sin()) ka sin()
2

S() = I0

(4)

Figure 3: Theoretical diraction from a uniformly-excited circular aperture For uniform excitation, the diraction pattern has nearly-negligible sidelobes, as shown in Figure 3. This is important because it implies that the antenna pattern is highly directional. The directive far-eld pattern implies that the antenna should have a gain that is relatively high. Now that the theory behind our design has been described, we are prepared to actually build the antenna, and to see if it will perform as well as theory predicts.

Construction

One of the primary benets of the cantenna is that it is both inexpensive and relatively easy to build. The necessary parts, cost, and a general approach to construction are outlined below.

3.1

Parts

Below is a list of the parts necessary for construction and their approximate cost: 1. Pepperidge Farm Pirouette Tin - $5.25

2. N-Female Adaptor (screw type, not surface mount) - $5.00 3. Short piece of 12-gauge copper wire (4-5cm) - $1.00 4. Pigtail to connect wireless card to antenna - $13.00 Total Price: $24.25

3.2

Assembly

Below is a general description of the assembly process. Reference Figure 2 for a side-cut drawing of the nished product. 1. Gather all necessary parts 2. Measure the radius of the can to be used as the waveguide. (If a Pirouette can is used, the radius is 0.047m.) 3. Using the radius and Equation 3 from Theory, calculate the quarter guide wavelength. The free-space wavelength is xed for a given frequency. For our frequency, 0 = 12.5 cm so 0 = 4 3.125 cm. For the Pirouette can, g = 4.81 cm. 4. Carefully solder the piece of wire into the open end of the N-female connector. Make sure the wire is as close to perpendicular with the surface of the connector as possible. Take care that the solder ows properly, and that the dielectric material of the connector itself does not melt. 5. Cut the newly-attached wire so that the wire and connector extend a total of one quarter freespace wavelength from the dielectric material of the connector (3.125 cm). 6. Place a large X-mark on the can a distance of one quarter guide wavelength from the back of the can. Make a hole centered at the mark large enough to securely insert the N-Female connector. During the construction of our can, we used a drill to create a hole in the can, then used a mill to enlarge it to the proper size. To eliminate the remaining dangerously-sharp edges of the hole, we used a le to smooth o the edges. 7. Screw or push the N-female connector and exciter into the hole in the can. If a screw-type connector is used, fasten the nut to the connector from the inside of the can.

Testing

The testing of the antenna can be divided into three primary types of tests: bench testing with a network analyzer, pattern testing with an anechoic chamber, and eld testing using an infrastructure wireless network.

4.1

Network Analyzer Testing

The rst test to be performed is an analysis of the antenna with a network analyzer. This testing is to provide basic information about the antennas operating characteristics and to ensure that it will not cause harm to the wireless networking device to which it is attached. The following procedure was followed in the testing of the antenna: 4

1. Gather necessary equipment (a) Network analyzer capable of measuring S11 parameter for a frequency range that includes 2400-2480 MHz (b) Proper cabling to connect the network analyzer to the N-Female adapter of the antenna 2. Calibrate network analyzer to the end of the cable to be used during the test. This will ensure that the length of the cable or any minor imperfections in it will not interfere with the test results. 3. Very securely attach the cable to the N-female connector on the antenna. In our tests, an improperly-tightened cable increased the observed SWR parameter of the antenna by over 25%. 4. Measure and record both the log-scale magnitude of the reection coecient and SWR of the S11 parameter of the antenna for the operating range of the antenna. 5. Turn o and properly return all testing equipment.

4.2

Anechoic Chamber Testing

The next test to be performed involves the use of an anechoic chamber designed for determining the eld patterns of antennas. Its results will give information about both the gain and directivity of the antenna. Our setup is shown in Figure 4.

Figure 4: Anechoic Chamber Testing Setup

1. Pattern Testing (a) Setup the spectrum analyzer at the throat of the anechoic chamber to scan for signals and report the strength of the maximum signal in the 2400-2040 MHz range, using a bandwidth of 100 MHz to avoid excessively long testing times. (b) Setup the signal generator on the rotating Styrofoam pedestal, ensuring that it will generate a signal at a valid wireless networking frequency, and that all modulation is turned o. In our case, we used a transmit power of 10 dBm. (c) Securely connect the antenna to the signal generator using N-female cables. (d) For the rst test, the H-plane or azimuthal plane, align the antenna so its exciter is oriented vertically. Align the horn antenna on the receiving end so that it is like-polarized with the transmitting antenna. 5

(e) Direct the transmitting antenna at the receiving antenna and adjust its direction until a maximum signal is received. This ensures that the 0 degree measure references the boresight of the antenna. (f) Using the anechoic chambers software, record data for the antenna at intervals throughout the plane. Take measurements at least every 2 degrees around an entire 360-degree rotation. (g) Next, the E-plane, or elevation plane, will be tested. Align the antenna so its exciter is oriented horizontally. Align the horn antenna on the receiving end so it is like-polarized with the transmitting antenna. (h) Again, direct the transmitting antenna at the receiving antenna and adjust its direction until a maximum signal is received to ensure that boresight is at 0 degrees. (i) Using the anechoic chambers software, record data for the antenna at intervals throughout the plane. Take measurements at least every 2 degrees around an entire 360-degree rotation. 2. Gain Testing (a) Setup the signal generator on the rotating Styrofoam pedestal, ensuring that it will generate a signal at a valid wireless networking frequency, and that all modulation is turned o. In our case, we used a transmit power of 10 dBm. (b) Securely attach an antenna of known gain to the signal generator and point its maximum radiation direction at the throat of the anechoic chamber. (c) Attach an antenna, preferably of known gain, to the spectrum analyzer at the throat of the chamber, pointed at the transmitting antenna. (d) Record the magnitude of the received signal in dBm. (e) Replace the antenna attached to the spectrum analyzer on the receiving end of the chamber with the antenna to be tested. (f) Record the magnitude of the received signal in dBm. (g) Turn o and properly return all testing equipment.

4.3

Field Testing

While laboratory measurements can provide good insight into the behavior of an antenna, its performance in a typical environment is the true test. Described below is the general method used to test our antenna in the eld, attempting to minimize interference eects. Specics about our eld testing can be found in Appendix A. 1. Select a test site that is in a at area with relatively few objects that may interfere with the testing. We used a straight gravel road surrounded by large elds. 2. Use a standard 802.11b wireless base station as the transmitter. Disable all security features and encryption on the base station. The eectiveness of these features depends on the brand of hardware and software used. Though they will not aect signal strength, they may make testing more dicult due to the incompatabilities and added latencies within the system. 3. Measure signal strength, SNR, and any other applicable parameters through the range of distances where a reliable connection can be established. 4. For comparison purposes, perform the above procedure for a variety of wireless networking equipment. Take care to note the type, location, and orientation of antennas, as well as the transmit and receive power of all devices used in the testing. 6

Experimental Results

The results of our battery of tests were quite pleasing. The impedance and SWR of the antenna were obtained from the network analyzer, while the anechoic chamber resulted in both numerical data and graphical insight into the pattern of the antenna. Values for reection coecients and SWR were both measured using the network analyzer, though the value for SWR proved to be a much more useful performance characteristic. The reection coecient values were primarily used to prove the self-consistency of the SWR measurements through the relation 1+|| SWR = 1|| . SWR values were constant across the band, at approximately 2.1:1. The anechoic chamber tests resulted in two sets of data: E-plane and H-plane. The two sets of data are from the two dierent geometries in which the test was run. In its normal conguration, that is, with the exciter vertically oriented, the test platforms rotation about its vertical axis results in the azimuth pattern. When both the transmitting and receiving antennas are rotated 90 degrees about the boresite axis, the same rotation of the test platform will generate the elevation pattern. These two sets of data, while similar, bare some striking dierences. These dierences are manifestations of the physical asymmetry of the can. The azimuth plane is symmetrical about boresight because the left and right halves of the antenna are symmetrical; the elevation plane is not because the bottom of the can contains the connector and exciter, while the top of the can does not. (Figure 5)

Figure 5: View of the antenna from boresight Despite their dierences, both planes are rather directional. As seen in Figure 6, each has a two-sided 3 dB beamwidth of 72 degrees, and has its peak signal strength at boresight. The results of our gain testing showed that our antenna has a gain of over 10 dB. While lab testing provided useful data, eld testing provided the most dramatic results. In its rst informal test, performed inside an apartment building, the antenna could reliably detect approximately fteen unique wireless networks in each of the cardinal directions as opposed to only four networks visible with the computers wireless network cards built-in antenna. More formalized testing was also performed that showed dramatic improvement over original equipment manufacturer (OEM) antennas. A head-to-head study was performed. A laptop with a Cabletron wireless network card was tested using its built-in antenna, a quarter-wavelength ground plane monopole, and our cantenna. The recieved signal and noise levels were compared under the same conditions. The results are show in Figure 7. Using the above setup, the default antenna stopped reliably connecting at 75-90 m, the quarter-wave monopole at 125 m, and the cantenna at 400 m. Furthermore, it can be noted that the use of the cantenna does not introduce a disproportionately large amount of noise into the received signal. Another more informal test was performed under the same setup as the formal test to determine the maximum distance at which a reliable connection could be eastablished for some commonly available 7

Azimuth (HPlane) Pattern 0o 30

Elevation (EPlane) Pattern 0o 30

30

30o

60o

60o

60o

60o

90o

90o

90o

90o

120o

120o

120o

120o

150o

150o

150o

150o

180o

180

(a) Azimuth (H-plane) Pattern

(b) Elevation (E-plane) Pattern

Figure 6: Measured antenna pattern wireless network cards. The test showed that the maximum range of the built-in wireless capability of an Apple Powerbook is 60-90 m. A Cisco Wireless PCMCIA card was also tested, and its maximum range was 90-120 m. For comparison, a Belkin desktop card was also tested using the cantenna. The desktop card retained 100% Signal Strength and 100% link quality at 300 m as reported by its internal software. At a distance of 300 m, the cable connecting our antenna to the wireless adapter was inadvertantly pulled sharply and a connector on the wireless card broke. Therefore, we were not able to determine the maximum range of the desktop card with our antenna.

Recommendations

Overall, our antenna works reasonably well, but there are some optimizations that can me made. First, the length of the exciting monopole can be trimmed to give a better impedance match. Fringing eects around the end of the conductor may result in an electrical length dierent than the physical length of the element. Second, the length of the can could be extended to further attenuate higher order modes. The can length can also be optimized to reduce signal reection at the aperture. Finally, there are some geometric non-idealities to the can which could be removed. A slight ridge exists near the aperture, and the closed end of the can is not perfectly at.

Conclusion

We successfully designed, tested, and measured a low-cost directional antenna for IEEE 802.11b/g. The design is widely available and simple to fabricate. Its cost is kept low by the use of over-thecounter and commercially available parts. The antenna pattern is directional, with minimal sidelobes and gain of 10 dB. It is reasonably well matched to a standard 50 system. The antenna is excited 8

(a) Signal Strength vs. Distance

(b) SNR vs. Distance

Figure 7: Results of eld testing by a standard quarter-wave monopole element. The signal is then directed and polarized by a can, which acts as a circular waveguide. The signal then diracts from the aperture of the can to form a directional antenna pattern. The proper size and positioning of all elements of the antenna are dependent upon the frequency. When eld tested with an actual 802.11b network, our antenna had a line of sight range of over 400m, compared to 90m for the same wireless cards included antenna.

Appendix - Field Testing Parameters

For all eld tests, a Netgear MR814v2 router was used as the base station to create the wireless networking environment. All security and encryption was turned o to minimize the impact of the software. The output power and signal thresholds are not available for this particular model; however, it has a singe antenna with a gain of 2 dBi. The testing was performed along the side of a long, straight dirt road in the middle of a plowed corneld. The Netgear base station was powered using an APC uninterruptible power supply and placed on top of the supply, which had a plastic case. The unit was then placed on a raised area at the side of the road. Signal measurements were taken using a Cabletorn Roamabout 802.11b wireless card connected to a Sony Vaio laptop. The Roamabout Client Utility was used to record signal strength, noise, and SNR values at every fty feet (15.24 m). Distance was measured using a surveyors measuring wheel. Table 8 shows the available parameters for equipment used in eld testing. Product Cabletron Roamabout 802.11b Transmit Power (mW) 32 Signal Threshold (dBm) -82

Table 1: Field Test Equipment Parameters

10

Acknowledgments
This paper would not have been possible without help from members of our university community.

References
[1] Hecht, Eugene. Optics. 3rd ed. Reading, MA: Addison-Wesley, 1998. [2] Ilmajoki, Martti Palomaki. Wlan Waveguide Antenni. 29 Sept. 2002. 27 Apr. 2006 <http://www.saunalahti./elepal/antenna2.html>. [3] Kong, Jin Au. Electromagnetic Wave Theory. Cambridge, MA: EMW Pub., 2000. [4] Rehm, Gregory. How to Build a Tin Can Waveguide WiFi Antenna. 27 Apr. 2006 <http://www.turnpoint.net/wireless/cantennahowto.html>.

11