How much is it worth to let us do it

[David Farber <farber () linc cis upenn edu>]

     This rule turns the conventional wisdom of microwaves upside
down.  For example, it is true that microwaves don't travel far
in the atmosphere.  You don't want to use them to transmit 50,000
watts of Rush Limbaugh over 10 midwestern states, but to
accommodate 200 million two-way communicators will require small
cells; you don't want the waves to travel far.  It is true that
microwaves will not penetrate most buildings and other obstacles,
but with lots of small cells, you don't want the waves to
penetrate walls to adjacent offices.
     
     Microwaves require high-power systems to transmit, but only
if you want to send them long distances.  Wattage at the receiver
drops off in proportion to the fourth power of the distance from
the transmitter.  Reducing cell sizes as you move up the spectrum
lowers power needs far more than higher frequencies increase
them.  Just as important, mobile systems must be small and light.
The higher the frequency, the smaller the antenna and the lighter
the system can be.
     
     All this high-frequency gear once was prohibitively
expensive.  Any functions over two gigahertz require gallium
arsenide chips, which are complex and costly.  Yet the cost of
gallium arsenide devices is dropping every day as their market
expands.  Meanwhile, laboratory teams are now tweaking microwaves
out of silicon.  In the world of electronics - where prices drop
by a third with every doubling of accumulated sales - any
ubiquitous product will soon be cheap.
     
     The law of the telecosm dictates that the higher the
frequency, the shorter the wavelength, the wider the bandwidth,
the smaller the antenna, the slimmer the cell and ultimately, the
cheaper and better the communication.  The working of this law
will render obsolete the entire idea of scarce spectrum and
launch an era of advances in telecommunications comparable to the
recent gains in computing.  Transforming the computer and phone
industries, the converging spirits of Maxwell, Shannon and
Shockley even pose a serious challenge to the current
revolutionaries in cellular telephony.






The New PC Revolution: PCN


     
     Many observers herald the huge coming impact of wireless on
the computer industry, and they are right.  But this impact will
be dwarfed by the impact of computers on wireless.
     
     In personal communications networks (PCN), the cellular
industry today is about to experience its own personal computer
revolution.  Just as the personal computer led to systems
thousands of times more efficient in MIPS per dollar than the
mainframes and minicomputers that preceded it, PCNs will bring an
exponential plunge of costs.  These networks will be based on
microcells often measured in hundreds of meters rather than in
tens of miles and will interlink smart digital appliances,
draining power in milliwatts rather than dumb phones using watts.
When the convulsion ends later this decade, this new digital
cellular phone will stand as the world's most pervasive PC.  As
mobile as a watch and as personal as a wallet, these PICOs will
recognize speech, navigate streets, take notes, keep schedules,
collect mail, manage money, open the door and start the car,
among other computer functions we cannot imagine today.
     
     Like the computer establishment before it, current cellular
providers often seem unprepared for this next computer
revolution.  They still live in a world of long and strong - high-
powered systems at relatively low frequencies and with short-
lived batteries - rather than in a PCN world of low-power systems
at microwave frequencies and with batteries that last for days.
     
     Ready or not, though, the revolution will happen anyway, and
it will transform the landscape over the next five years.  We can
guess the pattern by considering the precedents.  In computers,
the revolution took 10 years.  It began in 1977 when large
centralized systems with attached dumb terminals commanded nearly
100 percent of the world's computer power and ended in 1987 with
such large systems commanding less than one percent of the
world's computer power.  The pace of progress in digital
electronics has accelerated sharply since the early 1980s.
Remember yesterday, when digital signal processing (DSP) - the
use of specialized computers to convert, compress, shape and
shuffle digital signals in real time - constituted an exorbitant
million-dollar obstacle to all-digital communications? Many
current attitudes toward wireless stem from that time, which was
some five years ago.  Today, digital signal processors are the
fastest-moving technology in all computing.  Made on single chips
or multichip modules, DSPs are increasing their cost-
effectiveness tenfold every two years.  As radio pioneer Donald
Steinbrecher says, "That changes wireless from a radio business
to a computer business."
     
     Thus, we can expect the cellular telephone establishment to
reach a crisis more quickly than the mainframe establishment did.
The existing cellular infrastructure will persist for vehicular
use.
     
     As the intelligence in networks migrates to microcells, the
networks themselves must become dumb.  A complex network, loaded
up with millions of lines of software code, cannot keep up with
the efflorescent diversity and creativity among ever more
intelligent digital devices on its periphery.  This rule is true
for the broadband wire links of fiber optics, as intelligent
switching systems give way to passive all-optical networks.  It
is also true of cellular systems.
     
     Nick Kauser, McCaw Cellular Communications' executive vice-
president and chief of technology, faced this problem early in
1991 when the company decided to create a North American Cellular
Network for transparent roaming throughout the regions of
Cellular One.  "The manufacturers always want to sell switches
that do more and more.  But complex switches take so long to
program that you end up doing less and less," says Wayne Perry,
McCaw vice-chairman.  Each time Kauser tried to change software
code in one of McCaw's Ericsson switches, it might have taken six
months.  Each time he wanted to add customer names above a 64,000
limit, Ericsson tried to persuade him to buy a new switch.  The
Ericsson switches, commented one McCaw engineer, offer a huge
engine but a tiny gas tank.  The problem is not peculiar to
Ericsson, however; it is basic to the very idea of complex switch-
based services on any supplier's equipment.
     
     When McCaw voiced frustration, one of the regional Bell
operating companies offered to take over the entire problem at a
cost of some $200 million.  Instead, Kauser created a Signaling
System 7 (SS-7) network plus an intelligent database on four
Tandem fault-tolerant computers, for some $15 million.  Kauser
maintains that the current services offered by North American
Cellular could not be duplicated for 10 times that amount, if at
all, in a switch-based system.  Creating a dumb network and off-
loading the intelligence on computer servers saved McCaw hundreds
of millions of dollars.
     
     The law of the microcosm is a centrifuge, inexorably pushing
intelligence to the edges of networks.  Telecom equipment
suppliers can no more trap it in the central switch than IBM
could monopolize it in mainframes.
     
     Kauser should recognize that this rule applies to McCaw no
less than to Ericsson.  His large standardized systems with 30-
mile cells and relatively dumb, high-powered phones resemble big
proprietary mainframe networks.  In the computer industry, these
standardized architectures gave way to a mad proliferation of
diverse personal computer nets restricted to small areas and
interlinked by hubs and routers.  The same pattern will develop
in cellular.






Could "Charles" Upend McCaw?


     
     Together with GTE and the regional Bell operating company
cellular divisions, McCaw is now in the position of DEC in 1977.
With its new ally, AT&T, McCaw is brilliantly attacking the
mainframe establishment of the wire-line phone companies.  But
the mainframe establishment of wires is not McCaw's real
competition.  Not stopping at central switches, the law of the
microcosm is about to subvert the foundations of conventional
cellular technology as well.  Unless McCaw and the other cellular
providers come to terms with the new PC networks that go by the
name of PCNs, they will soon suffer the fate of the minicomputer
firms of the last decade.  McCaw could well be upended by its
founder's original vision of his company - a PICO he called
"Charles."
     
     Just as in the computer industry in the late 1970s, the
fight for the future is already under way.  Complicating the
conflict is the influence of European and Japanese forces
protecting the past in the name of progress.  Under pressure from
EEC industrial politicians working with the guidance of engineers
from Ericsson, the Europeans have adopted a new digital cellular
system called Groupe Speciale Mobile (GSM) after the commission
that conceived it.
     
     GSM is a very conservative digital system that multiplies
the number of users in each cellular channel by a factor of
three.  GSM uses an access method called time-division multiple
access (TDMA).  Suggestive of the time-sharing methods used by
minicomputers and mainframes to accommodate large numbers of
users on centralized computers, TDMA stems from the time-division
multiplexing employed by phone companies around the world to put
more than one phone call on each digital line.  Thus, both the
telephone and the computer establishments are comfortable with
time division.
     
     Under pressure from European firms eager to sell equipment
in America, the U.S. Telephone Industry Association two years ago
adopted a TDMA standard similar to the European GSM.  Rather than
creating a wholly new system exploiting the distributed powers of
the computer revolution, the TIA favored a TDMA overlay on the
existing analog infrastructure.  Under the influence of Ericsson,
McCaw and some of the RBOCs took the TDMA bait.
     
     Thus, it was in the name of competitiveness and
technological progress, and of keeping up with the Europeans and
Japanese, that the U.S. moved to embrace an obsolescent cellular
system.  It made no difference that the Europeans and Japanese
were technologically well in our wake.  Just as in the earlier
case of analog HDTV, however, the entrepreneurial creativity of
the U.S. digital electronics industry is launching an array of
compelling alternatives just in time.
     
     Infusing cellular telephony with the full powers of wide and
weak - combining Shannon's vision with computer advances - are
two groups of engineers from MIT who spun out to launch new
companies.  Qualcomm Inc. of San Diego, is led by former
professor Irwin Jacobs and telecom pioneer Andrew Viterbi.  A
Shannon disciple whose eponymous algorithm is widely used in
digital wire-line telephony, Viterbi now is leading an effort to
transform digital wireless telephony.  The other firm,
Steinbrecher Corp. of Woburn, Mass., is led by an inventor from
the MIT Radio Astronomy Lab named Donald Steinbrecher.
     
     Like Bernie Bossard and Vahak Hovnanian, the leaders of
Qualcomm and Steinbrecher received the ultimate accolade for an
innovator: They were all told their breakthroughs were
impossible.  Indeed, the leaders at Qualcomm were still
contending that Steinbrecher's system would not work just weeks
ago when PacTel pushed the two firms together.  Now they provide
the foundations for a radical new regime in distributed wireless
computer telephony.






Signals in Pseudonoise


     
     Ten years ago at Linkabit, the current leaders of Qualcomm
conceived and patented the TDMA technology adopted as the U.S.
standard by the Telephone Industry Association.  Like analog
HDTV, it was a powerful advance for its time.  But even then,
Viterbi and Jacobs were experimenting with a Shannonesque
technology.
     
     A classic example of the efficacy of wide and weak, CDMA
exploits the resemblance between noise and information.  The
system began in the military as an effort to avoid jamming or air-
tapping of combat messages.  Qualcomm brings CDMA to the
challenge of communications on the battlefronts of big-city
cellular.
     
     Rather than compressing each call into between three and 10
tiny TDMA time slots in a 30-kilohertz cellular channel,
Qualcomm's CDMA spreads a signal across a comparatively huge 1.
25-megahertz swath of the cellular spectrum.  This allows many
users to share the same spectrum space at one time.  Each phone
is programmed with a specific pseudonoise code, which is used to
stretch a low-powered signal over a wide frequency band.  The
base station uses the same code in inverted form to "despread"
and reconstitute the original signal.  All other codes remain
spread out, indistinguishable from background noise.
     
     Jacobs compares TDMA and CDMA to different strategies of
communication at a cocktail party.  In the TDMA analogy, each
person would restrict his or her talk to a specific time slot
while everyone else remains silent.  This system would work well
as long as the party was managed by a dictator who controlled all
conversations by complex rules and a rigid clock.  In CDMA, on
the other hand, everyone can talk at once but in different
languages.  Each person listens for messages in his or her own
language or code and ignores all other sounds as background
noise.  Although this system allows each person to speak freely,
it requires constant control of the volume of the speakers.  A
speaker who begins yelling can drown out surrounding messages and
drastically reduce the total number of conversations that can be
sustained.
     
     For years, this problem of the stentorian guest crippled
CDMA as a method of increasing the capacity of cellular systems.
Spread spectrum had many military uses because its unlocalized
signal and cryptic codes made it very difficult to jam or
overhear.  In a cellular environment, however, where cars
continually move in and out from behind trucks, buildings and
other obstacles, causing huge variations in power, CDMA systems
would be regularly swamped by stentorian guests.  Similarly,
nearby cars would tend to dominate faraway vehicles.  This was
termed the near-far problem.  When you compound this challenge
with a static of multipath signals causing hundreds of 10,000-to-
1 gyrations in power for every foot traveled by the mobile unit -
so-called Rayleigh interference pits and spikes - you can
comprehend the general incredulity toward CDMA among cellular
cognoscenti.  Indeed, as recently as 1991, leading experts at
Bell Labs, Stanford University and Bellcore confidently told me
the problem was a show-stopper; it could not be overcome.
     
     Radio experts, however, underestimate the power of the
microcosm.  Using digital signal processing, error correction and
other microcosmic tools, wattage spikes and pits 100 times a
second can be regulated by electronic circuitry that adjusts the
power at a rate of more than 800 times a second.
     
     To achieve this result, Qualcomm uses two layers of
controls.  First is a relatively crude top layer that employs the
automatic gain control device on handsets to constantly adjust
the power sent by the handset to the level of power received by
it from the base station.  This rough adjustment does not come
near to solving the problem, but it brings a solution into reach
by using more complex and refined techniques.
     
     In the second power-control step, the base station measures
the handset's signal-to-noise and bit-error ratios once every
1.25 milliseconds (800 times a second).  Depending on whether
these ratios are above or below a constantly recomputed
threshold, the base station sends a positive or negative pulse,
either raising or lowering the power some 25 percent.






Dynamic Cells


     
     Passing elaborate field tests with flying colors, this power-
control mechanism has the further effect of dynamically changing
the size of cells.  In a congested cell, the power of all phones
rises to overcome mutual interference.  On the margin, these high-
powered transmissions overflow into neighboring cells where they
may be picked up by adjacent base station equipment.  In a quiet
cell, power is so low that the cell effectively shrinks,
transmitting no interference at all to neighboring cells and
improving their performance.  This kind of dynamic adjustment of
cell sizes is impossible in a TDMA system, where adjacent cells
use completely different frequencies and fringe handsets may
begin to chirp like Elmer Fudd.
     
     Once the stentorian voice could be instantly abated, power
control changed from a crippling weakness of CDMA into a
commanding asset.  Power usage is a major obstacle to the PCN
future.  All market tests show that either heavy or short-lived
batteries greatly reduce the attractiveness of the system.
Because the Qualcomm feedback system keeps power always at the
lowest feasible level, batteries in CDMA phones actually are
lasting far longer than in TDMA phones.  CDMA phones transmit at
an average of two milliwatts, compared with 600 milliwatts and
higher for most other cellular systems.
     
     A further advantage of wide and weak comes in handling
multipath signals, which bounce off obstacles and arrive at
different times at the receiver.  Multipath just adds to the
accuracy of CDMA.  The Qualcomm system combines the three
strongest signals into one.  Called a rake receiver and co-
invented by Paul Green, currently at IBM and author of Fiber
Optic Networks (Prentice Hall, 1992), this combining function
works even on signals from different cells and thus facilitates
hand-offs.  In TDMA, signals arriving at the wrong time are pure
interference in someone else's time slot; in CDMA, they
strengthen the message.
     
     Finally, CDMA allows simple and soft hand-offs.  Because all
the phones are using the same spectrum space, moving from one
cell to another is easy.  CDMA avoids all the frequency juggling
of TDMA systems as they shuffle calls among cells and time slots.
As the era of PCN microcells approaches, this advantage will