7/29/2019 10.1.1.125.3124

1/7

Advanced Site Solutions for WiMax

Dr. J. R. Sanford

Cushcraft Corporation5933 Sea Lion Pl, Carlsbad, CA

Introduction

The allure of the WiMax wireless high speed data service is well documented[1]. Significant effort has been

invested in developing a spectrally efficient modulation that is rather multipath impervious. Still, the laws

of physics dictate that we need to get power from pointA to pointB that is above the receiver threshold.

This requires rigorous RF planning and hardware design. Unlike mobile phone systems, we need to providethe receiver with a C/I ratio that allows for the use of a modulation with a high spectral efficiency. In this

paper we compare the properties of a WiMax site to those of a cellular phone site. Like the cellular systems

of the eighties, initial WiMax deployments will be noise limited. However, as the deployments increase thesystem will quickly be interference limited. This capacity limitation is the subject of this paper. Noting the

difference between WiMax and cellular, the known limitations of a capacity limited cellular network is

used to predict what will happen with WiMax deployments. Well tested site planning and transport

configurations that maximize system capacity and reliability are then outlined. While analysis has

indicated that MIMO techniques may be able to increase the spectral efficiency by a factor or two, thetechniques presented here can easily triple the spectral efficiency without introducing cost.

An analysis of a number of different link scenarios shows that filtering and antenna sidelobe rejection are

key to maximizing data throughput. Also apparent is the vast increase in data that a well planned WiMaxsystems can support in comparison to cellular systems. Finally, from the RF transport quality viewpoint,

the hub receiver will likely be the limiting factor. Either interference robust coding schemes or lower order

modulation may be appropriate for the uplink. Let us start by pointing out the main differences between acellular phone site and a WIMAX site.

WiMax Relative to 2G Mobile Phone

Mobile Phone systems are well understood. In order to better understand the issues related to multipoint

distribution systems, we compare the cellular phone environment to that of WiMax. Below are the defining

differences between the two systems.

1. Antenna Directivity- - Phones are omni-directional while WiMax CPEs may use directionalantennas.

2. Mobility - Cell phones must work both indoors and outdoors while WiMax CPEs willgenerally be mounted outside.

3. Batteries - Cell phone run on Batteries (saturated PA) while WiMax CPEs can bepowered (linear PA)

4. Adaptive Modulation - WiMax depends on adaptive modulation to provide robust coverageFor the purposes of this paper, noise and interference are differentiated as follows. Interference is unwanted

signal generated by the WiMax system. Noise is everything that is not interference or the desired signal.

For example, the image signal from a PCS site or internally generated harmonics will both be considered

noise. Interference can be further subdivided into co-channel and cross channel. Co-channel interference

resides at the same frequency as the desired signal and cross channel interference is due to signal on nearbyfrequency bands.

7/29/2019 10.1.1.125.3124

2/7

Figure 1 - The RF Transport Layer

Figure 1 shows a simplified diagram of the WiMax RF transport layer in frequency division duplex (FDD).The antennas and filters provide noise and interference rejection to the front end amplifier. The aggregate

performance of these devices is comparable through both up and down links. The hub antenna will

typically cover a sector ranging from 45 to 120 degrees in azimuth. The customer premise electronics

(CPE) antenna will have a gain ranging from about 7dBi to 25dBi. Higher gain CPE antennas tend to belarger and prohibitively expensive. The spatial filtering of the CPE antenna provides a degree of

interference rejection not available in a cellular system. An externally mounted WiMax CPE antenna can

consistently provide between 25dB and 40dB of additional isolation between cells. In contrast, cellularphones have omni directional antennas that provide no isolation between cell site locations. Hence the

isolation is purely due to the frequency filtering and the propagation characteristics.

The cost constraints on the hub are not as stringent as those on the CPE. Therefore higherQ filters andhigher performance antenna can be employed at the hub end. Because line power is generally available at

both the CPE and hub, both the hub and CPE amplifiers can operate in a linear region. Therefore, spectrally

efficient modulations such as versions of OFDM with QAM are used. The level of the modulations (i.e.

QPSK through 64QAM) will be determined by the acceptable bit error rate (BER) and the signal to noise

ratio (SNR) in a sparsely deployed system or Carrier to Interference (C/I) in a densely deployed system.For simplicity let us assume that we can increase our N-QAM level by a factor of 4 for every 7dB

improvement in SNR. This corresponds to a doubling of the data rate with the same 7dB improvement. Theeffect of improving C/I is more modulation and error correction dependent, but similar trends hold. Hence,

reducing interference increases data capacity so long as we are well above the receiver threshold.

Consequently, we can potentially double the revenue generating potential of the base station if we chooseRF hardware that improves the C/Iby 7dB.

In their most basic configuration, cellular phone systems are sectorized into three-cell configurations. The

motivation for this is frequency reuse. In a single cell, the frequency band might be broken into three equal

sub-bands with one sub band per sector. This in itself does not provide any frequency reuse over an omni

configuration. However, when the same scheme is used in a multiple cell configuration, the result becomesa reuse of three. The cost of the additional hub equipment required for the sectorization is small in

comparison to the over all system cost. This scheme keeps the C/I ratio good enough for phone traffic

(CDMA or GMSK) over the bulk of the multicell coverage area and hence maximizes the capacity, whichin this case corresponds to the number of users. The cell phone modulations allow the transmitters to work

in a saturated mode that increase the power amplifiers efficiency and thus prolong battery life. The tradeoff

to these import features is spectrally efficient. GSMs GMSK modulation can have a "spectral efficiency"

of up to approximately 1 bit per symbol or 1 bit per hertz of bandwidth while 64QAM over OFDM mighthave a spectral efficiency of 6 bits per symbol. The word "approximately" is used because there are several

different ways to measure the bandwidth of a signal. The method used for GMSK signals in GSM is to find

the bandwidth that contains about 99% of the radio signal power. Gaussian Minimum Shift Keying

(GMSK) was developed specifically for GSM by the COST (Council on Science and Technology and later

ETSI). GMSK is a type of minimum frequency shift (MFS) modulation that achieves an approximatelyoptimum compromise between the amount of power out of the desired bandwidth of the modulated carrier

signal (only about 1% for the type used in GSM) and still allows the binary data to be accuratelydemodulated from the received radio signal in the presence of noise. GMSK is one of a number of well

PA

LNA

PA

LNA

modem

modem

LO

LO

LOLO

Pup

Pdown

7/29/2019 10.1.1.125.3124

3/7

thought-out modulations appropriate for mobile phone systems. While great for mobile telephony, it is not

the modulation of choice for broadband data due to the low spectral efficiency.

In order to transmit data quickly in a limited bandwidth an alternative modulation was chosen. QAM allows

an inherently high level of spectral efficiency. When transmitted over OFDM the system becomes tolerantto inter-symbol interference. Relative to basic QAM, there is a small loss of bandwidth due to the guard

time between carriers and the system requires good frequency synchronization. Still, due to the use of FFT

technology the modulation remains relatively simple to implement in hardware. The biggest impact tosystem planning is peak to average power requirement (PAP). This will need to be dealt with through

prudent RF chain design and antenna selection. Table 1 indicates the spectral efficiency of the various

modulation levels.

Table 1 Typical Data Rate in a 7MHz channel

We can now speculate on how we might want to plan a WiMax site relative to a mobile phone site. Wewould like to optimize the C/I over the bulk of the site rather than settling for good enough. Further, we

would like to stay well above the receiver threshold and well below the transmitter saturation point to

account for the extra requirement on PAP. Clearly, the site planning methodology for mobile phone

systems is not appropriate for wireless broadband system that requires a linear radio and employs adaptive

modulation.

In order to better understand the issues confronting a WiMax site planner we look at two scenarios. First we

consider CPEs at various distances from the WiMax cell site. After, we will look at CPEs in differentsectors.

Figure 2 Effect of Antenna Directivity

Range Effects

Let us assume we are deploying WiMax, an OFDM system that employs adaptive modulation ranging from

QPSK to 64QAM with optimized coding rates. Let us first look at a system with a hub and three CPEs. TheCPEs are at various distances from the hub (i.e. 100m, 1Km and 5Km). Perhaps the biggest advantage

WiMax has over cellular is the antenna directivity of the CPE unit. Besides adding link margin it reduces

interference into adjacent cells. Higher gain antennas can be used towards the cell periphery for link margin

Modulation/Code

Rate

QPSK

1/2

QPSK

16 QAM

1/2

16 QAM

3/4

64 QAM

2/3

64 QAM

3/4

7MHz 4.1MBit/s8.7MBit/s 11.6MBit/s17.5MBit/s23.8MBit/s26.2MBit/s

ShapedEffect to

this line

No Shaping

to this line

Uniform coverage

r2 roll-

Down

Potential

7/29/2019 10.1.1.125.3124

4/7

balancing. Normally, higher gain antennas can provide a higher front-to-back ratio than lower gain

antennas. The systems at the cell edge employ the better front to back ratio for reduction of interference

into and from the adjacent cells. The combination of the shaped beam hub antenna together with the use of

CPE antennas with various gains allows the majority of the CPE receiver dynamic range to be used by the

modem rather than for power control and link balancing. Likewise, it also adds to the dynamic range in theCPE transmitter. The same argument applies to the hub receiver and transmitter. This virtual dynamic range

increase corresponds to higher data capacity. Hub antennas with shaped elevation patterns are readily

available. The shape of the pattern follows a cscelectric field distribution from the antennas 3dB point toapproximately 45 degrees in elevation. An example is shown in figure 3.

Let us assume the hub is mounted 50m above the CPEs and that it is used to cover a 10Km diameter cell.

The CPE at 100m resides at the 30degree point in the elevation pattern while the CPE at 1Km resides at the

3-degree point in the elevation pattern. The antenna has a very small electrical down tilt such that the CPE

at 10Km is at the main beam peak. The energy available at the locations of the inner two CPEs is

approximately the same due to the pattern shaping. We will assume that physically small 11dBi antennasare used at these locations. The CPE at 5Km has at least 14dB more path attenuation than the unit at 1Km

and the hub has 3dB more gain in this direction. Here we make up the 11dB difference by using a 22dBi

gain antenna. Now, at both the hub and the CPE we have approximately the same link characteristics. Thehub then receives three channels of comparable strength and demodulates each of them. Each CPE receives

three channels of comparable strength but only demodulate one of the three channels. This all happens

without sacrificing dynamic range on either the up or down link. Hence, the signals can be transmitted attheir maximum data rates. Clearly, this is somewhat of an oversimplification, however it does illustrate the

goal: maximize the available dynamic range to allow the highest order modulation to be used.

In order to minimize noise at the hub a horizontal polarized antenna is chosen. This reduces the noise

generated by vertically polarized PCS base station transmitters. The WiMax hub has better than 40dB of

cross polar rejection. However, typical base station antennas typically only have about 20dB rejection inthe sideline region. Hence, there is about a 20dB improvement in PCS noise rejection over using a

vertically polarized antenna.

Figure 3 Elevation Pattern of a Shaped Beam Hub Antenna

Optimal

7/29/2019 10.1.1.125.3124

5/7

Sectorization

Without sectorization, adjacent cells will interfere with one another and the frequency reuse factor will be

one. However, an overly sectorized cell will interfere with itself and will not efficiently reduce the

interference into adjacent cells. The optimal sectorization from an interference standpoint is one that

balances the effects of the cells self interference with the interference from adjacent cells. This normallyworks out to about six sectors. The frequency reuse factor is more difficult to understand. In cellular

systems the problem was simpler due to the fixed data rate. For voice communications there was no point

in improving beyond a given threshold. However, advanced WiMax systems will employ dynamic rateallocations based on the quality of the communications channel. Adding more sectors increases the system

cost. However, site acquisition and infrastructure normally out weigh the antenna and RF equipment cost.

Let us first assume the same equipment used in scenario 1 is again used. Due to the link balancing, theproposed equipment should reduce the spectral re-growth of cross channel interference to negligible level.

For simplicity a four sector configuration is assumed and can be extrapolated to higher orders of

sectorization. Recently, antennas with a high degree of sidelobe rejection have been developed for

WIMAX. Figure 4 depicts the carrier and the interference power. The blue (solid) line depicts frequency set1 and the red (dashed) line depicts frequency set 2. Here we see that the co-channel interference is around -

33dB while the worst cross channel interference is 0dB(the cell cross over points). Hence, frequency

filtering is relied-on for the cross channel interference. In order to maximize the rejection in the modem,frequency set one might consist of odd channels while frequency set two would correspond to even

channels. Based on QAM threshold level, this level of co channel interference would allow 256 QAM witha 10-6 bit error rate (BER). This corresponds to nearly 8 bits/s/Hz in a sector. However, 64 QAM tends to

be the more common choice. This would provide a potential data thru-put of approximately 3x26.2 or 78.6MBits/s in a 7 MHz channel or a whopping 11 MBits/Hz/s per base station.

Figure 4 Frequency Reuse of Three

Antennas providing this type of reuse are quickly becoming popular. Cushcrafts dual polarized WiMax

hub antenna (J33016D00) is the example used for this analysis. Some operators emplo the dual polarizationfeature for additional isolation from sector to sector while other use polarization for additional isolation

between base station. The original intent of the dual polarization feature was to enable Alamouti and other

Space-Time Code techniques at the base station in order to provide additional fade margin. The approach

will likely be key in enhancing the robustness of next generation WiMax base stations.

ChannelSet 2

ChannelSet 1

ChannelSet 2

ChannelSet 1

ChannelSet 2

7/29/2019 10.1.1.125.3124

6/7

Figure 5 Dual-Polarization Low Sidelobe Shaped-Beam WiMax Antenna

Diversity, Smart Antennas and MIMO

There are a range of techniques focused on enhancing the capacity or robustness of the WiMax system.

Some of these techniques advertise non line of site performance (NLOS). No technique can effectivelycombat the absorption by tree foliage. However, for the operators willing to pay for it diversity, smart

antennas and MIMO system can each enhance the performance of the system assuming neither the RF

hardware nor the site planning is compromised. As in the mobile telephone industry, diversity techniques

are the most likely to be widely deployed in the early going. Generally operators are charged for the site bythe number of antennas. As such, the use of dual polarization antennas tends to be the most cost effective

solution for diversity. Where a technique such as maximum ratio combining is used, slant 45 degree dual

polarized antennas have been shown to have the most desirable performance. Generally the C/I ratios are

similar for each of the ports. This works out to be beneficial for maximum ratio combining. However, insituation where a space-time combining technique is used, vertical-horizontal dual polarized antennas tend

to be optimal. Most scatterers tend to be horizontally oriented, the decorrelation the scattering environment

of the two polarizations has less correlation and thus enhances the diversity gain. Such a product is shownin figure 6.

One advantage of diversity techniques is ease of use. They allow the described sectorization techniques to

be employed without additional thought. The dramatic spectral efficiently gains of these sectorizationtechniques come without the cost and complexity overhead of MIMO and Smart Antenna (SDMA)

systems. However, a hybrid sectored smart antenna system with high performance antenna has the potential

to add even more capacity by further improving C/Is on a channel by channel basis. At this time it is still

unclear how MIMO based system will perform in terms of spectral efficiency in a fully loaded point to

multipoint environment.

Conclusions

There are multiplicity of techniques to approach wireless access and cell planning. However, in all casesthe capacity of a system is a function of the quality of the RF transport layer. The WiMax systems transport

layer discussed here will provide much higher C/I ratios than cellular systems. Antenna directivity and

sidelobe suppression enable high order adaptive modulation schemes to be a defining advantage. When

7/29/2019 10.1.1.125.3124

7/7

combined with high performance antennas and sectorization the system can provide data rates greater than

an order of magnitude than that of todays mobile phone systems. These techniques are relatively simple

and cost effective when compared to MIMO based solutions.

References:

Telephony, Complete Guide to WiMax, May 31, 2004.

Intel Technical Journal, Multiple-Input Multiple-Output for Throughput and Range, Vol. 08, Issue 3,

Aug 20, 2004.

Dr. John R. Sanford, is currently CTO of Cushcraft Corporation. In July of 2004, Cushcraft acquiredOptimal RF, a startup company active in the area of next generation antenna design where Dr. Sanfordserved as President. Prior to this he was Chief Technical Officer ofREMEC, Inc. where he oversaw the

companys technology roadmap, IP development and strategic partnerships. AtREMEC, Dr. Sanford held a

number of positions including head of engineering of Northern California operations and GM of the Fixed

Wireless Group. Prior to that, he was founder and President ofSmartwavesInternational, which wasacquired by REMEC in February, 1999. At both locations Dr. Sanford developed microwave transceivers

and related products including antennas, filters and amplifiers. Prior to that he was a Professor at Chalmers

University of Technology where he taught courses and conducted research related to Electromagnetics,Antennas and Array Signal Processing. From 1988 through 1993 Dr. Sanford headed the Mobile Tower

Top Group atHuber & Suhner AG. There he introduced the first base station patch antenna, switched beamantenna along with a variety of unique filters, EMP and optical products. From 1985 to 1988 he was a

research engineer and group manager with the Georgia Tech Research Institute where he designed militaryradar and communication systems.

Dr. Sanford has published more than sixty technical papers, research reports, and has taught short courses

and given seminars in the USA, Germany, Switzerland, Sweden, and Finland. He has been a guest lecturer

at theRoyal Institute of Stockholm, The Technical University of Florence, Stanford University, San Jose

State and The Technical University of Karlsrue. He holds numerous patents and has been a two time

Chairman of chapters of the IEEE MTT and AP Societies. In 1995 Dr. Sanford was granted a Docent fromChalmers University. In 1992 he received his Doctorate in Electromagnetics from The Federal University

of Switzerland in Lausanne, Switzerland. He received his MS in Physics in 1988 from Georgia Tech and

his BSEE from Syracuse University in 1984 in the USA.