What NLOS Means in Terms of Wave Propagation

The object of this book is to go light on RF theory and heavy on practicality. However, to understand
even the rudiments of NLOS, one must know something about how physical objects
affect radio wave propagation.
When a radio wave encounters a physical object—and by that I mean anything from an air
molecule to a mountain—it can behave in one of three ways. It can give up some of its energy
to the object in the form of heat, a process known as absorption. It can rebound from the object
without surrendering appreciable energy to it, a process called reflection. Or it can bend
around the object, a process known as diffraction. These three processes, by the way, are
notmutually exclusive. A reflected signal, for instance, may immediately be diffracted as it
encounters a different contour of the object reflecting it, and in every case where reflection or
diffraction occurs, some energy will be absorbed as well.
Absorption
Pure absorption does not change the direction of the radio wave but robs it of energy and
reduces the fade margin. Since the signal steadily loses energy simply by being propagated over free space, ultimately limiting its useful range, the effect of absorption of energy by physical
structures such as walls or trees is to further reduce the distance over which a reliable
connection may be maintained, and often very appreciably at that. Indeed, absorption losses
can, if significantly severe, interrupt the signal entirely. A good example of this is provided by a
large park full of high trees where anyone attempting to blast through the treetops with a
microwave signal to reach buildings on the other side will be completely unable to establish an
airlink even with the most advanced NLOS equipment. Incidentally, as a good rule of thumb, a
direct signal path through dense foliage will result in a loss of signal strength of approximately
1 decibel per meter.
Reflection
The effect of reflection depends heavily on whether the main lobe or the side lobes are
reflected. If only the side lobes are reflected, multipath distortion will occur in the main lobe,
compromising signal integrity and resulting in loss of data but not interrupting the signal
entirely. If, on the other hand, the main lobe is reflected by an obstruction standing directly in
its path, then almost no energy from that lobe will appear at the receiver’s antenna. In such
instances, reflected energy, most likely from a side lobe, may reach the receiver at a sufficient
level to provide a usable signal, although the fade margin will be vastly reduced over what it
would be with a direct signal. The problem here, however, is more significant than a simple
reduction in signal level because now the receiver is operating entirely in the multipath environment,
and it is likely to be subject to not one but multiple reflections, each of which will
interfere severely with one another, further reduce fade margin, and substantially increase the
bit error rate.
To get a better idea of how multipath occurs, refer to Figure 5-5 for a schematized depiction
of multipath reflections. The severity of mulitpath will depend on the frequency of the
transmission, the distribution of reflective surfaces in the area separating the transmitter from
the receiver, and the directivity characteristics of the transmitter receiver. Now it is entirely possible to set up a deliberate reflection in an effort to reach an
obstructed receiver, but again one is likely to be operating in a multipath environment because
the side lobes will be encountering their own obstructions, possibly the same obstruction
affecting the main lobe, and will likely be reflected into the path of the main lobe. In addition,
the direct signal will lose energy to whatever object is reflecting it.
A number of radios on the market do have the ability to function in a pure multipath environment,
but the important thing to remember is that they do not function nearly as well in  such a regime. Thus, the presence of some degree of NLOS capability in a receiver does not
mean that one can ignore line-of-sight considerations.
Diffraction
Diffraction, which occurs when radio waves bend around the edges of an object, results in a
transmitted beam becoming off-axis in relationship to the receiver antenna. Since radio waves
are vastly greater in wavelength than visible light waves, diffraction can occur when there is
still optical line of sight between two radio terminals. The effect of diffraction is to reduce signal
strength substantially at the receiver and also to introduce a possible source of unwanted
early reflections. Here again, a diffracted signal is not necessarily useless, but it is certainly less
useful than a direct signal.
NLOS: A Truer Conception
It should be obvious by now that, absent line of sight, no radio functions optimally, and the
mere fact that a radio has some measure of NLOS capability does not mean that the network
operator can place base stations and subscriber haphazardly. NLOS technologies are partial
remedies (misleading might be a better word) at best. They do not rewrite the laws of physics.
Before I leave this discussion, I should mention the term near-NLOS, industry jargon that
surpasses simple NLOS in its sheer ambiguity. As the term is most commonly employed, it
refers to radio equipment capable of dealing with consequences of obstructions that occur
within the Fresnel zone but do not block optical line of sight. What is really being claimed here
is not the ability to reach completely obstructed sites but the ability to cope with multipath
with a high degree of effectiveness.
NLOS Technologies
Radios are able to function in NLOS environments by using a number of technologies. These
involve the basic functioning of the radio, the modulation system employed, the design of the
antenna, and the use of certain extraneous signal processing techniques normally involving
multiple antennas.
Basic Radio Performance and NLOS
Increasing radio sensitivity and channel selectivity while providing the radio front end with
high overload capabilities and a very low noise circuitry will in and of itself endow the receiver
with an enhanced ability to use the weakened signals upon which one is forced to rely on a
NLOS installation. In other words, basic high quality RF engineering and circuit implementation,
most of which, incidentally, is still analog, will go a long way toward the aiding in the
recovery of a marginal signal.
Modulation Technique and NLOS
Modulation technique can also play a major role in enabling links where line of sight is not
present. All of the spread spectrum techniques in common use in broadband wireless, including
frequency hopping, direct sequence, CDMA, and OFDM, can immunize the signal to some
extent from multipath, with frequency hopping perhaps performing best in this regard. However,
interestingly frequency hopping figures in the IEEE standard only for WLANs and 802.11, not in the 802.16 standard. OFDM also has very good immunity from multipath, and OFDM
radios can generally operate in the presence of weaker signals than is the case with most other
modulation techniques, thus permitting the radio to retrieve a usable signal in marginal settings.
Nevertheless, OFDM, in and of itself, will not allow a link to be maintained in the face of
complete line-of-sight obstructions, as is the case when a target terminal is located behind a
hill, a large masonry building, or a dense grove of trees. It is actually better suited to near-NLOS
conditions.
OFDM is the only advanced modulation technique specified in the 802.16 standard, and
only in the 802.16a amendment. Single-carrier modulation is permitted within the 802.16a
standard as well, but no company currently active in wireless broadband within the lower
microwave region is making a single-carrier system any longer except for wireless bridges.
OFDM
Since OFDM figures so prominently in the standard, and because its advocates make so many
claims in its behalf, the network operator should acquire at least a rudimentary understanding
of how the technology works, of its real strengths and weaknesses (and it has weaknesses), and
the degree to which it truly advances the art.
OFDM has a lengthy patent history and a long incubation period, but in its present form it
is quite a recent development. Indeed, prior to 1990 the basic technology that would permit the
concept to be realized scarcely existed.
In simplest terms, OFDM is a technique by which a message is assigned to a number of
narrowband subcarriers, usually numbering in the hundreds or thousands, simultaneously. In
802.16a the specified number of subcarriers is 256. The same information is not replicated on
all of the subcarriers; rather, it is parceled out, with some bits going to one subcarrier and some
to another and to another. Some redundancy will normally be present; that is, some bits will be
shared among some sets of subcarriers, but, in the main, the data will be widely scattered over
a considerable expanse of spectrum.
Incidentally, OFDM may not be considered a modulation technique in the strictest sense
because there is nothing in the technology itself that determines how a signal is impressed
on a carrier wave, and OFDM can be—and in fact always is—combined with amplitude modulation,
phase modulation, or combinations thereof. It can also be combined with direct
sequence or CDMA, both of which involve an initial modulation of phase to impress the signal
on a carrier and then remodulation of the resulting signal to impose an additional coding
sequence upon it. OFMD is in fact combined with frequency hopping in Flarion Technologies’
Flash OFDM system. processors capable of performing digital fast Fourier transforms (FFTs), covered shortly, on
very high frequency waveforms, and such digital signal processors (DSPs) did not exist until as
recently as a few years ago.
To explain the significance of digital FFTs to OFDM, you need first to look at the design
goals in engineering an OFDM radio, as follows. First, one wants to pack available spectrum
very efficiently; second, one wants to achieve a very high immunity to interference and multipath;
and third, one wants to reuse spectrum aggressively. OFDM does all of these things
thanks to a DSP programmed to perform the FFT.
A Fourier transform, for those lacking a grounding in the mathematics of wave theory, is a
computational method for deriving the single-frequency constituents of a complex waveform,
and it enables a radio to tune to various frequencies entirely in the digital domain since a
microprocessor is actually running the computations. This is in marked contrast to the operation
of a traditional analog radio where electrical filters separate individual frequencies and no
computing functions are performed.
A Fourier transform requires a powerful processor, which adds considerable expense to
the system, but it can disentangle transmissions that are so close together in frequency as to
overlap, and it eliminates the expedient of empty guard bands between channels that are
required in all analog radio systems (most so-called digital radios used in cell phones and
WLANs still use conventional analog front-end tuning circuitry).
The end result is that the FFT-based OFDM system can pack the available spectrum with a
great multitude of overlapping channels and tease them out of a blizzard of interference by
finding the repetitive patterns that signify a carrier wave while simply ignoring extraneous
information. Because the resulting system is at once so robust and so spectrally efficient, it lets
the network operator reuse subcarriers very aggressively and tolerates degrees of multipath
that would cripple most other systems. It also allows the receiver to recover a weaker signal
than is possible with most other modulation techniques and thus to operate in the presence of
obstructions that would render other radios useless.
The fact that the bit rate for individual subcarriers is relatively low also contributes to
the system’s high immunity from multipath. Because individual bits endure for relatively
long durations, antiphase reflections are less likely to result in complete cancellations of an
individual bit.
The drawbacks? The main problem with OFDM is electrical inefficiency. The DSP gobbles
current, and the subcarriers are always transmitting regardless of whether any data is assigned
to them. For this reason, OFDM has never been used in a commercial mobile system to date
and is not favored by most of the proponents of 802.20. Nevertheless, OFDM is favored by the
architects of the fourth-generation cellular telephone system and may see deployment there
by the end of this decade.
Another limitation with OFDM is that the technology is not really applicable above a few
gigahertz. Existing logic circuits simply cannot switch fast enough to deal with signals in the
millimeter microwave region, and even if they could, it is not clear what OFDM would buy
the operator there since multipath is not a problem, and signal attenuation through the lightest
obstructions is so severe that no modulation technique is going to be much help in
preventing it.
Nevertheless, OFDM will probably gain ever-wider acceptance in the future and may
even find its way into mobile networks. The distant future is difficult to predict, though.
OFDM faces challenges, at least within the 802.20 standard, from the older technology wideband
CDMA, and may face challenges in all broadband wireless applications from the rapidly evolving ultrawideband RF technology. What I can say with utter certainty is that OFDM does
provide a means of building useful NLOS links at least in some circumstances, and when combined
with other technologies, such as adaptive array antennas (covered next), can enable high
degrees of synergy. Multiple Antennas: Diversity Antennas, Phased Arrays, and Smart Antennas
For the worst-case scenarios, something more is required than advanced modulation, though
even that something may not be enough in all cases. The following sections refer to multiple
antennas, conjoined in most cases with a sophisticated signal processor that can sample the
different antenna feeds and construct useful signals where none may be present with a single
antenna element.
Diversity Antennas The simplest kind of multiantenna array is what is known as a diversity
antenna system. Here two or more antennas, generally simple omnidirectional rods spaced
some distance apart, are deployed. These work essentially by choosing between or among different
samples of the same signal.
At the wavelengths used for broadband wireless services, the signal quality may differ
considerably from one antenna to another. The radio behind the diversity antenna will have
circuitry for detecting the signal least afflicted with multipath distortion and will select that
antenna receiving such a signal. In cases of changes over time in the incidence of multipath
from one antenna element to another, the circuitry will simply choose again.
Phased Arrays A much more sophisticated way of using multiple antenna elements is to construct
what is called a phased array. Here the outputs or inputs of the various elements are
constructively or destructively merged to form beams of almost any desired shape, and this
can be done at both the transmitter and receiver. Beams can even be steered over many
degrees of arc to direct energy off to one side.
Phased arrays have existed for decades, but until rather recently they were manually configured,
and the phase relationships were relatively fixed. Much more recently, adaptive array
antennas, or smart antennas, have been developed that utilize a computing engine to shape
beams dynamically on a channel-by-channel basis in order to concentrate energy in whatever
direction and at whatever intensity is desired. Some such antennas can even steer the beam
dynamically and track a moving object, thus providing the receptor terminal with a strong
direct signal at all times.
Adaptive Array “Smart Antennas” Essentially two types of adaptive array antennas have been
developed: the switched-beam antenna and the beam-steering type. The switched-beam
antennas can combine only the beams from the different element in a finite number of juxtapositions.
The beam can assume a few fixed widths and a few fixed angles and is not infinitely
variable. The beam-steering type is, on the other hand, infinitely variable and is far more flexible.
It also requires far more processing power to operate effectively.
Smart antennas excel in NLOS applications for a number of reasons, not all of which are
pertinent to all designs. Most significantly—and this does apply to all design variants—the
antenna array has the ability to focus a beam very tightly toward each subscriber unit on a
packet-by-packet basis. The RF energy in that beam is not dispersed through the atmosphere  as in a normal broadbeam transmission and instead is delivered almost in its entirety to the
subscriber site, where it can blast through considerable obstructions and still provide a usable
signal. By concentrating energy in such tight patterns on a channel-by-channel or even packetby-
packet basis, the adaptive array can also increase frequency reuse, theoretically up to
several times within a single cell, while still adhering to regulatory limits on transmitter power.
This in turn reduces the need for sectorization.
All adaptive array antennas provide signal diversity; that is, the mere fact that several
spaced antenna elements are exposed simultaneously to the transmission practically guarantees
that signal quality will vary from element to element. The system can then select the best
signal, and in an NLOS situation, there is a far higher likelihood that an array will find a usable
signal than a single element. In some systems the adaptive-antenna array can even take multipath
reflections impinging on the various elements and phase-align them so as to construct a
single, coherent, high-strength signal.
Other systems can combine various signals and then use vector-cancellation strategies to
reduce interference. In one experimental system developed by Lucent but never marketed, the
smart antenna system could simultaneously receive several signals over the same channel,
separate them on the basis of time of arrival and multipath signature, and then reconstruct
usable signals out of all of the interfering transmissions, thus reusing spectrum at virtually the
same point in space! Figure 5-6 shows an ArrayComm adaptive array antenna. Adaptive array antennas are most beneficial when used in both the base station and the
subscriber terminal. Such double-ended systems are known as multiple-in, multiple-out
(MIMO) links (the multiples refer to the antenna arrays). Currently there is no explicit MIMO
standard for 802.16, though a standard is in preparation for 802.11.
I recall first reporting on such intelligent phased arrays back in 1996 and hearing such
capabilities discussed even then, and, predicting—erroneously, as it turned out—that intelligent
adaptive array antennas would sweep the wireless industry. I also attended a number of
demonstrations of advanced adaptive array technology and can attest that it brings real benefits
that should be appealing to network operators.
Why, then, has it not in fact become established in the marketplace?
In a word, price. Adaptive array antennas require multiple radios, one for each antenna
element, and heavy-duty processors. To date, most have been based on expensive field programmable
gate array circuitry rather than application-specific integrated circuits (ASICs),
which further drives up the price. Although the capabilities of the network are most enhanced
when adaptive array antennas are used at both ends of the airlink, the high price of such antennas
has ruled out their use in subscriber terminals thus far, and thus the full potential of the
technology has seldom been realized. At least one company in the broadband wireless business,
IP Wireless, promises to introduce moderately priced adaptive array antennas for
subscriber units some time in 2004, but, if it occurs, this will be a first.
At the time of this writing only a handful of companies manufacture equipment utilizing
adaptive array antennas intended for fixed broadband wireless deployments. These include
the aforementioned IP Wireless, Navini, BeamReach, Redline, Orthogon, Vivato, and Array-
Comm. Of these, only Orthogon makes a MIMO system, and that system is intended solely
for point-to-point wireless bridges. Vivato’s equipment is really designed only for 802.11 networks,
and ArrayComm makes a component antenna module, not a complete system. Thus,
current choices for the network operator seeking a networkwide single-vendor deployment of
adaptive array antennas really only number four. Interestingly, of that four, two, Navini and IP
Wireless, make products supporting mobility.
Despite the small number of vendors today, adaptive array antenna technology is almost
inevitable since it permits much more efficient use of spectrum than any passive antenna system.
By the end of this decade it will be the norm, and equipment manufacturers without the
technology will not be competitive.
At the same time, I am not suggesting that any of the existing systems featuring adaptive
array antenna systems are necessarily the best choice for any given network operator in any
given market. Adaptive array technology comes at a cost premium, and it is for the network
operator to decide whether the increased access to potential customers will result in enough
additional revenues to balance the higher equipment costs. One should also keep in mind that
NLOS capabilities are only one attribute to consider in choosing infrastructure equipment.
One also has to consider the routing and/or switching options built into the base station, additional
support for quality of service (QoS) beyond the 802.16 standard, built-in support for
voice, and so on.
My advice is that any broadband operator should look at adaptive array antennas closely
and attempt to test equipment. In other words, it is a choice that should definitely be explored.
Whether it is to be embraced at this time is another matter. Mesh Networks and NLOS
Chapter 3 has already discussed the principles of mesh network operation fairly extensively.
This section focuses on the NLOS capabilities of meshes and how a mesh architecture relates
to the previously mentioned techniques.
Using a mesh topology does not exclude any of the methods for facilitating NLOS mentioned
earlier. A mesh network can use OFDM and/or adaptive array antennas, and in fact one
system manufactured by Radiant Networks (now out of business) did utilize a rather crude
mechanically steered antenna that was the functional equivalent of a smart array in certain
respects. Indeed, if one were to combine a smart antenna with a mesh, one would achieve a
powerful synergy, but so far no manufacturer has attempted this.
By itself a mesh offers a compelling solution to NLOS problems in that it allows the network
to route around obstacles—provided enough subscriber terminals are present to permit
circuitous routing paths. The drawback, of course, is that the multiple hops required to
describe such paths rob the network of capacity and can introduce latencies that compromise
time-sensitive applications.
As indicated earlier, the future of the mesh in public networks remains in doubt. The companies
advocating the architecture are, with one exception, all very small and may not be able
to survive in the overall broadband wireless marketplace where there have been too few
deployments to date to support a large and diversified population of equipment manufacturers.
And there are also issues involving the installation of the terminals that could compromise
the effectiveness of what is essentially a sound networking strategy.
In a mesh, as you have seen, the subscriber terminal and the base station are one and the
same. This means that the radio and antenna are likely to be self-installed and the antenna
placed indoors at low elevation. The result will be that the individual radio will be provided
with a number of possible transmission paths but none that are likely to be particularly good,
especially, if, as is the case with most current mesh systems, the antennas are omnidirectional.
Such antennas also give rise to a generally high level of background interference because the
signal is being propagated everywhere. But without adaptive array antennas, or physically
steered antenna such as Radiant makes, an omni is the only way to ensure that the signal can
reach multiple nodes in the network.
The choice of a mesh over a point-to-multipoint architecture is fundamental and will
affect every aspect of network performance and evolution afterward. It is a choice not to be
undertaken lightly. My view is that mesh products are less proven than point-to-multipoint but
may be indicated in certain circumstances. If, for instance, the network is being planned for a
dense urban core where tall buildings ring every conceivable base station location and prevent
the establishment of links with choice customer locations, then one may want to explore what
may be done with a mesh, in particular whether line-of-sight connections can be established
among initial nodes that would enable future prime customers to be easily reached. Unfortunately,
this exercise will have to be done manually, at least in part, because not many software
tools are designed for planning meshes. And, in any case, the fact that mesh equipment itself is
designed to be self-organizing rather than operator configurable limits the degree to which the
network can be planned.
As is the case with any other radio, the operator would want to examine the mesh radio in
terms of its other capabilities—can it support QoS, voice telephony, legacy protocols such as
Asynchronous Transfer Mode (ATM), and so on? NLOS is of little use if one cannot offer the services
and reliability the customer expects.