♦ WaveLAN®-II: A High-Performance Wireless

LAN for the Unlicensed Band

Ad Kamerman and Leo Monteban

In July 1997 the Institute of Electrical and Electronics Engineers (IEEE) completed
standard 802.11 for wireless local area networks (LANs). WaveLAN®-II, to be released
early in 1998, offers compatibility with the IEEE 802.11 standard for operation in the
2.4-GHz band. It is the successor to WaveLAN-I, which has been in the market since
1991. As a next-generation wireless LAN product, WaveLAN-II has many enhance-
ments to improve performance in various areas. An IEEE 802.11 direct sequence
spread spectrum (DSSS) product, WaveLAN-II supports the basic bit rates of 1 and 2
Mb/s, but it can also provide enhanced bit rates as high as 10 Mb/s. This paper dis-
cusses various aspects of the system design of WaveLAN-II and characteristics of its
antenna, radio-frequency (RF) front-end, digital signal processor (DSP) transceiver
chip, and medium access controller (MAC) chip.

Introduction
The first wireless local area network (LAN) prod- nal processing (DSP) application-specific integrated
ucts appeared in the market around 1990, although circuit (ASIC) and the Intel Ethernet controller—
the concept of wireless LANs has existed since the stayed the same as for the original WaveLAN product.
late 1970s.1 The release of the industrial, scientific, The improvements covered such areas as smaller card
and medical (ISM) bands made unlicensed spectrum design, lower power consumption, and increased soft-
available and prompted significant interest in the ware support.
design of these wireless LANs. The next generation of At the beginning of 1998, Lucent Technologies,
these products is being implemented on Personal which played a leading role in driving the Institute of
Computer Memory Card International Association Electrical and Electronics Engineers (IEEE) 802.11 for
(PCMCIA) cards used in laptop computers and wireless LANs, will introduce WaveLAN-II. This prod-
portable devices. The major technical issues sur- uct has taken a big step forward in performance for
rounding wireless LAN systems are size, power con- the small-size, low-power PCMCIA card-based design.
sumption, bit rate, aggregate throughput, range of This paper is organized as follows: In “The IEEE
coverage, and interference robustness. 802.11 Standard,” we define the standard and exam-
After the NCR WaveLAN® was introduced in ine the differences between direct sequence spread
1991 as an Industry Standard Architecture (ISA) card spectrum (DSSS) and frequency hopping spread spec-
for the 915-MHz band, several new versions appeared trum (FHSS) within the context of this standard. Next,
for other computer buses, for the 2.4-GHz band, for in “WaveLAN-II’s Enhanced Bit Rates,” we consider
various network operating systems, and as original these higher bit rates. We look at where the different
equipment manufacturer (OEM) products for compa- functions are implemented in “Distribution of
nies such as DEC, Toshiba, and Solectek. Throughout Functions.” Further, the implementation complexity
the WaveLAN product improvement life cycle, a num- needed for the radio-frequency (RF) and DSP design is
ber of key components—notably the NCR digital sig- compared to the design in the code division multiple

118 Bell Labs Technical Journal ◆ Summer 1997

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 Panel 1. Abbreviations, Acronyms, and Terms
ACK—acknowledgment IEEE 802.11—IEEE standard for CSMA/CA-based
ARF—automatic rate fallback wireless LANs
ASIC—application-specific integrated circuit IF—intermediate frequency
BCA—basic coverage area IFS—interframe spacing
BCPM—Barker code position modulation IS-95—Interim Standard 95 (U.S. standard for
BPSK—binary phase shift keying CDMA-based digital cellular telephone)
CDT—carrier detect threshold (level) ISA—Industry Standard Architecture (specifies ISA
CDMA—code division multiple access bus in computers)
CSIR—co-channel signal to interference ratio ISM—industrial, scientific, medical (frequency
CSMA/CA—carrier sense multiple access with colli- band for these applications)
sion avoidance LAN—local area network
CSMA/CD—carrier sense multiple access with colli- MAC—medium access control (layer)
sion detection MIPS—million instructions per second
CQ—communications quality (WaveLAN-II propri- OEM—original equipment manufacturer
etary quality scale) PCMCIA—Personal Computer Memory Card
CTS—clear to send International Association (specifies PC card
DSP—digital signal processing/processor interface)
DSSS—direct sequence spread spectrum PHY—physical (layer)
DT—defer threshold (level) PN—pseudo noise
FHSS—frequency hopping spread spectrum PPM—pulse position modulation
FSK—frequency shift keying PSK—phase shift keying
GFSK—Gaussian frequency shift keying QAM—quadrature amplitude modulation
GSM—Global System for Mobile Communications QPSK—quadrature phase shift keying
(European standard for digital cellular tele- RAM—random access memory
phone) RF—radio frequency
IEEE—Institute of Electrical and Electronics RTS—request to send
Engineers SCA—shared coverage area
IEEE 802.3—IEEE standard for CSMA/CD-based SNR—signal-to-noise ratio
wired LANs (Ethernet) WMAC—wireless medium access control/controller

access (CDMA)-based cellular telephone systems, United States and Europe, the 2,400- to 2,483.5-MHz
described in “IEEE 802.11 DSSS vs. IS-95 CDMA.” We band has already been allocated, while for some other
also discuss multi-channel roaming and automatic rate countries, such as Japan, another part has been
selection. “Power Management” describes the power assigned, as shown in Table I. The 802.11 standard
consumption and support for the IEEE 802.11 stan- focuses on the medium access control (MAC) and
dard, and “Scalable System” examines configurable physical layer (PHY) protocols for ad-hoc networks
carrier detection, configurable defer behavior, roaming and networks based on access points.
thresholds, and infrastructure density. “Conclusion” In access point-based networks, the stations within a
summarizes the paper and projects how the group or cell can only communicate directly with the
WaveLAN-II design will respond to future upgrades in access point. The access point forwards messages to
data rates. the destination station within the same cell or
through the wired distribution system to another
The IEEE 802.11 Standard access point, from which messages arrive at the desti-
IEEE 802.11 is a standard for wireless systems that nation station. In ad-hoc networks, the stations operate
will operate in the 2.4- to 2.5-GHz ISM band. on a peer-to-peer level with no access point or
Available worldwide, this ISM band allows unlicensed (wired) distribution system.
operation of spread spectrum systems. For both the The 802.11 standard supports DSSS with differen-

Bell Labs Technical Journal ◆ Summer 1997 119

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 Table I. Frequency bands and power levels for wireless LANs.
Countries Frequency range Maximum RF power level Rules for DSSS and FHSS
U.S., Canada,* and Latin 902–928 MHz 1W (at ERP and maximum DSSS: Receiver processing gain >
America (FCC Part 15,247) 2,400–2,483.5 MHz 6 dBi antenna gain) 10 dB
5,750–5,850 MHz FHSS: 75 hops or more
Europe† 2,400–2,483.5 MHz 100 mW (at EIRP) DSSS: Power spectrum density
(ETS 300 328) maximum 10 mW/MHz
FHSS: 20 hops or more
Japan (MPT Ordinances 2,471–2,497 MHz Not specified DSSS/FHSS: Power spectrum
78 and 79) density maximum 10 mW/MHz
Australia 2,400–2,450 MHz 500 mW

* In Canada, not the 5,750–5,850-MHz band

† In France/Spain, only the 2,445–2,483.5/2,475-MHz band
EIRP – Equivalent isotropically radiated power
ERP – Effectively radiated power
ETS – European Telecommunication Standard
MPT – Ministry of Posts and Telecommunications (in Japan)

tial encoded binary phase shift keying (BPSK) and probability of a collision would occur just after the
quadrature phase shift keying (QPSK), FHSS with medium becomes free, following a busy medium,
Gaussian frequency shift keying (GFSK), and infrared because multiple stations would have been waiting
with pulse position modulation (PPM). The basic for the medium to become available again. Therefore,
medium access behavior allows compatible PHYs to a random back-off arrangement is used to resolve
operate together by using the carrier sense multiple medium contention conflicts, as shown in Figure 1.
access with collision avoidance (CSMA/CA) protocol A very short-duration carrier detect turnaround time
and a random back-off time following a busy medium is fundamental to this random wait characteristic.
condition. In addition, all directed traffic uses immedi- The 802.11 standard DSSS uses a slotted random
ate positive acknowledgment (ACK frame), in which wait behavior based on 20-µs time slots, which cover
the sender schedules a retransmission if it does not the carrier detect turnaround time.
receive an ACK. In addition, the 802.11 MAC defines an option for
While CSMA/CA and Ethernet’s carrier sense reserving a medium using request-to-send/clear-to-
multiple access with collision detection (CSMA/CD) send (RTS/CTS) polling interaction and point coordi-
have similarities, their one fundamental difference is nation (for time-bounded services). 2,3 After a
the way they handle collisions. In wire-based net- busy-medium period, all wireless LAN devices have to
works, such as Ethernet, it is not technically compli- wait during an interframe spacing (IFS) period. After
cated to detect if transmissions from two stations are waiting a random number of time slot intervals, these
colliding. Detecting collisions in wireless systems that devices can attempt to transmit, provided no other
use only one channel is impractical, however, because transmission was detected in the interim.
of the large dynamic range of receive levels. Therefore, Regulations under which radio spectrum may
802.11 has chosen CSMA/CA, which uses a collision be used and frequencies in which wireless LANs can
avoidance scheme. (Panel 2 compares IEEE 802.11 be deployed have made the 2.4-GHz ISM band
DSSS with IS-95 CDMA.) available worldwide. Three similar approval regimes
The 802.11 CSMA/CA protocol is designed to exist: the U.S., Europe, and Japan. As Table I shows,
reduce the collision probability between multiple sta- there are no worldwide regulations for spectrum
tions accessing the medium, at the point in time occupation or power levels. The IEEE 802.11 stan-
where collisions would most likely occur. The highest dard has focused its standardization efforts on the

120 Bell Labs Technical Journal ◆ Summer 1997

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 2.4-GHz band because it is available worldwide,
despite its “regional” variations. Panel 2. IEEE 802.11 DSSS vs. IS-95 CDMA
IEEE 802.11 DSSS Interim Standard (IS)-95 for mobile cellular tele-
DSSS systems spread the signal energy across a phone systems, the Telecommunication Industry
relatively wide band by increasing the occupied band- Association’s standard, is based on CDMA. This
width. A DSSS transmitter converts a bit stream into a standardized CDMA technique combines DSSS
symbol stream, in which each symbol represents a with a high spreading factor and medium access.
During a call setup at an IS-95 cellular system, a
number of bits, depending on the phase shift keying
full-duplex connection is established, with 28.8
(PSK) modulation technique. The symbol information kb/s in the up link and 19.2 kb/s in the down link.
is converted into a complex-valued signal that is fed to Convolutional coding is applied on top of these
the spreader. The spreader multiplies its input signal bit rates for the embedded speech information
with a pseudo noise (PN) sequence, called a chip (1.2 kb/s to 9.6 kb/s). The 28.8-kb/s and 19.2-kb/s
sequence. This multiplication creates a signal with a match modulation rates over spreading factors of
wider bandwidth. The in-phase and quadrature com- (64/6) × 4 and 64, respectively, which, for both bit
rates, result in a chip rate of 1.288 megachips per
ponents of the spreader output signal are fed to a
second (Mc/s).
quadrature modulator. The transmitter front-end pro-
vides filtering, conversion to a higher RF, and power Power control in the transmitters of mobile sta-
tions reduces the interference at the base station
amplification.
receiver side between the transmitted signals
The 802.11 DSSS, based on the 11-chip Barker from in-cell mobile stations and the interference
sequence +1, -1, +1, +1, -1, +1, +1, +1, -1, -1, -1, at other base stations. Likewise, power control in
has already been used for WaveLAN-I.4 This 11- the base station transmitters reduces interference
chip sequence is used as the PN code sequence, and at mobile station receivers. Such reduction of
the symbol duration corresponds to the time of 11 interference allows more in-band activity and
chip intervals. The 11-chip spreading makes the increases the capacity for a given bandwidth.
occupied bandwidth larger and increases the effec- The 802.11 standard DSSS prescribes a single
tive bandwidth from 1 MHz to 11 MHz. The 802.11 fixed 11-chip spreading code that is used by all
standard specifies two bit rates—1 Mb/s with BPSK stations. The medium access assignment is not
made by a unique code, but rather by a listen-
and 2 Mb/s with QPSK, with a spectrum that looks
before-talk carrier sensing and at a deferred
the same for both bit rates. The channel center fre-
transmission. This transmission can only start
quencies under 802.11 are 2,412 MHz, 2,417 MHz, after waiting a random number of time slots of
2,422 MHz, … 2,462 MHz (for the U.S. and 20 µs. Because the medium access is based on
Europe). In Europe it also supports 2,467 and 2,472 sensing carrier activity above a receive level
MHz, but 802.11 supports only the channel center threshold that matches with the target defer
frequency 2,484 MHz in Japan. range, a transmit power control would conflict
with the harmonious defer behavior within the
IEEE 802.11 FHSS target cell area. The stations and the access point
FHSS systems hop from narrow band to narrow share the single-channel medium and, after get-
band within a wide bandwidth. FHSS wireless LAN ting access to the medium, they can use the full
stations send one or more data packets at one carrier bandwidth of 2 Mb/s for some period. The half-
frequency, then hop to another carrier frequency to duplex operation at 2 Mb/s is based on a symbol
rate of 1 Mbaud and a chip rate of 11 Mc/s. The
send one or more packets, and continue this
medium/channel access, medium/channel occupa-
hop–transmit sequence, called slow frequency hopping. tion time, bit rate, bit error rate requirements,
The time these FHSS radios dwell on each frequency is real-time aspects, and capacity definitions of
fixed. The hopping pattern appears random, but it is packet transmission and circuit-switched tele-
actually a periodic sequence tracked by both the phone services are totally different in nature.
sender and receiver. A FHSS transmitter converts the

Bell Labs Technical Journal ◆ Summer 1997 121

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 IFS Contention window
IFS
Busy medium Back-off window Next frame

Slot time
IFS – Interframe spacing

Figure 1.
Basic CSMA/CA behavior.

bit stream into a symbol stream in which each symbol tions, FHSS will continue to work over a few hop
represents one or more bits. channels a little longer than over the other hop chan-
Two frequencies are applied for binary frequency nels. DSSS, however, still gives reliable links for a dis-
shift keying (FSK) modulation, and four frequencies tance at which very few FHSS hop channels still work.
are applied for quaternary FSK modulation. For collocated networks (access points), DSSS
Frequency hopping is applied to the FSK-modulated gives a higher potential throughput with fewer access
signal. The transmitter front-end supplies conversion points than FHSS, which has more access points. The
to a higher RF and power amplification. The 802.11 smaller number of access points used by DSSS lowers
FHSS uses a GFSK modulation technique with a low the infrastructure cost.
modulation index (Gaussian frequency shaping, BT
[bandwidth – symbol interval] product = 0.5, modu- WaveLAN-II’s Enhanced Bit Rates
lation index h = 0.34 at 1 Mb/s, 0.15 at 2 Mb/s), In addition to the 802.11 bit rates of 1 and 2 Mb/s,
which gives a relatively narrow spectrum and allows a release of WaveLAN-II scheduled for July 1998 will
a higher bit rate in the 1-MHz narrow hop channels. provide bit rates as high as 5, 8, and 10 Mb/s. This
However, these FSK conditions increase the sensitiv- allows interoperability with 802.11 DSSS devices at
ity to noise and other impairments. The 802.11 stan- the lower (fallback) bit rate and coexistence with those
dard defines hops over channel center frequencies devices when the proprietary bit rates are higher. The
according to a periodic sequence (that looks like a PHY training preamble/header of the transmission,
random pattern), within a set of 79 frequencies: which takes 200 µs, looks the same for each bit rate; it
2,402 to 2,480 MHz in the U.S. and Europe. For is always based on the modulation for a bit rate of
Japan, however, 802.11 supports a set of only 23 fre- 1 Mb/s. The data portion—which can take a few mil-
quencies (2,473 to 2,495 MHz). liseconds, depending on the frame size (up to 2,300
DSSS vs. FHSS bytes)—is transmitted at a bit rate of 1, 2, 5, or 8 Mb/s.
We have compared various performance aspects5 Because the preamble/header is recognized by any
of the 802.11 standard DSSS and FHSS systems. DSSS 802.11 DSSS device, such a device then acquires the
has a more robust modulation and a larger coverage correct defer behavior. After that device transfers ini-
range than FHSS, even when FHSS uses twice the tial information at 1 or 2 Mb/s, the interoperating
transmitter power output level. FHSS gives a large devices assess each other’s capability, which may be 5,
number of hop frequencies, but the adjacent channel 8, or 10 Mb/s. A station then knows if it is useful to
interference behavior limits the number of indepen- switch to a higher bit rate.
dently operating collocated systems. Hop time and a The 802.11 standard DSSS uses an 11-chip
smaller packet size introduce more transmission time Barker code sequence to carry 1 or 2 bits of data per
overhead into FHSS, which affects the maximum pulse use, in the BPSK and QPSK formats, respec-
throughput. Although FHSS is less robust, it gives a tively. After the received signal is correlated using a
more graceful degradation in throughput and connec- filter matched to the same 11-chip Barker code
tivity. Under poor channel and interference condi- sequence, the resulting signal gives a main lobe,

122 Bell Labs Technical Journal ◆ Summer 1997

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 DSP WMAC Protocol stack
RF Drivers
ASIC ASIC and applications
IC
“Theseus” “Hermes”

Antenna

IF
RAM FlashROM Utilities Standard software
IC
interfaces

802.11 Standard hardware

air interface interfaces

ASIC – Application-specific integrated circuit IF – Intermediate frequency

DSP – Digital signal processor RAM – Random access memory
FlashROM – Flash read-only memory WMAC – Wireless medium access controller
IC – Integrated circuit

Figure 2.
Major WaveLAN-II functions.

which is 1 chip wide. The 10 additional chip positions are 5 Mb/s (QPSK + joint BCPM) and 8 Mb/s (QPSK
within the symbol period give side lobe values 11 + disjoint BCPM). The spectra of the transmitted sig-
times smaller than the main lobe. The position of the nals at 1, 2, 5, and 8 Mb/s are all the same, because
11-chip Barker code sequence within each symbol of the spreading of the identical 11-chip Barker code
interval could be modulated in addition to the modu- sequence. The differences between these four bit
lating polarity of the Barker code sequences in the rates with respect to the required signal-to-noise
in-phase and quadrature channels. ratio (SNR) for reliable receiver operation are limited
Bell Labs Research and the Wireless Com- to 6 dB, while the bit rate increases eight times.
munications & Networking Division in The However, the constraints for the transceiver imple-
Netherlands have jointly investigated modulating the mentation with respect to linearity and compensa-
position of the code sequence; this effort has resulted tion for channel degradation are more severe for
in higher bit rates when no extreme channel degrada- operation at 5 and 8 Mb/s.
tion exists. If we allow eight positions for the main Likewise, as described above, we can combine
lobe within the 11-chip symbol interval, and these multilevel modulation with BCPM. At quadrature
main lobe positions are chosen independently in the amplitude modulation (QAM)-16 and disjoint BCPM,
in-phase and quadrature channel, then the informa- the bit rate will be 10 Mb/s.
tion rate is increased by 2 times 3 bits.6 In this way we
can achieve an increase from 2 bits per symbol (QPSK) Distribution of Functions
to 8 bits per symbol (QPSK + disjoint Barker Code The WaveLAN-II network card and a set of soft-
position modulation [BCPM], or independent BCPM ware modules cooperate to offer an 802.11 LAN con-
in the I and Q channels). nection for a PC. At the highest software interfaces,
In a similar way we can increase the information the 802.11 LAN connection is equivalent to a tradi-
rate from 2 bits per symbol (QPSK) to 5 bits per sym- tional Ethernet (802.3) LAN. At the air interface, the
bol (QPSK + joint BCPM); then the BCPM bit rates 802.11 RF/baseband modulation and protocols are

Bell Labs Technical Journal ◆ Summer 1997 123

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 chip, which is a special development derived
from a Global System for Mobile
Communications (GSM) chip.
• DSP ASIC “Theseus.” The DSP ASIC, which per-
forms the 802.11 DSSS, was developed by
Lucent Technologies. It is the heart of the scal-
able radio section. (See Panel 3 for a compari-
son of RF and DSP components.)
Figure 3.
• WMAC ASIC “Hermes.” The “Hermes” is a pow-
WaveLAN-II PCMCIA card with internal antenna.
erful processor that runs the 802.11 wireless
MAC (WMAC) protocol firmware. Beyond the
used. Figure 2 gives a schematic overview of the basic frame transmit and receive, some of the
major functional elements of the hardware and soft- firmware functions it carries out are:
ware and describes how the various functions are dis- – Power management,
tributed over these elements. – Automatic rate fallback (ARF), and
The WaveLAN-II card, shown in Figure 3, is used – Multi-channel roaming and handover man-
in laptop computers, its right-hand side extending agement.
from the laptop cabinet. The transceiver front-end is • FlashROM. The FlashROM provides nonvolatile
mounted in a plastic cover. The slightly thicker part of memory for storing permanent and config-
the card contains the internal antenna, which has urable information to control the behavior of
excellent properties. The antenna automatically firmware and hardware operation. For exam-
switches from internal to external when connected to ple, it holds values used in roaming and scal-
an external antenna cable. able system design.
Network cards will be available for a number of • RAM. The random access memory (RAM)
standard hardware interfaces like PCMCIA and ISA. volatile memory is used for message buffering
Figure 2 shows the hardware elements on the net- and internal firmware administration.
work card, including the: The network card’s software elements include:
• Antenna. The WaveLAN-II card incorporates • Drivers. A set of LAN drivers directly control
two L-shaped inverted-F antennas. They are the data flow through the network card.
placed in mutually orthogonal fashion (cross- Different drivers offer a variety of high-level
polarized) and spaced at a wavelength of 0.25 software interfaces, such as Microsoft’s
to 0.5 from each other, which provides Network Driver Interface Specification,
antenna diversity. This antenna represents a Novell’s Open Data-Link Interface, and Packet
“leaky” resonator at 2.45 GHz (+/-100 MHz). Driver.
Shaping it in such a manner that none of its • Utilities. This set of programs, provided for net-
dimensions exceeds 1/8 of a wavelength work planning and card diagnosis, uses the
makes this antenna “electrically small” and network card to collect statistics and to moni-
consequently very omnidirectional. tor the RF medium.
• RF front-end. The radio-frequency–intermediate
frequency (RF–IF) up/down conversion is Single- and Multi-channel Roaming
based on a dedicated chip design for the 2.4- WaveLAN-II provides roaming within the cover-
GHz ISM band. This chip includes a low-noise age boundaries of a set of access points that are inter-
amplifier. The IF frequency is 352 MHz. connected by a wired distribution system. The access
• IF transceiver. The IF-baseband up/down con- points send beacon messages at regular intervals.
version is made by a modulator/demodulator Stations can track the conditions at which the bea-

124 Bell Labs Technical Journal ◆ Summer 1997

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 cons are received per individual access point. The
running average of these receive conditions is deter-
Panel 3. RF and DSP Components
mined by a communications quality (CQ) indicator, The WaveLAN-II card has miniaturized RF circuitry
as shown in Figure 4. The different zones within the for 2.4 GHz and contains ASICs for DSP and MAC.
As a reference, Lucent’s IS-95 design for the digital
full-range CQ scale refer to various states of activity
cellular telephone is based on RF circuitry for 850
at which a station tracks or tries to find an access MHz, the Lucent DSP 1629 chip, and an Intel
point. When the CQ is poor, the station has to microcontroller.
expend more effort to quickly find another access
WaveLAN-II provides half-duplex operation for a
point that gives a better CQ. 2-Mb/s bit rate with a 1-Mbaud symbol rate and
The access points are interconnected by a wired a chip rate of 11 Mc/s with 22-MHz sampling (a
distribution system, shown in Figure 5. They can use factor of 2 oversampling). The receive mode
channel frequencies from a basic set of all supported requires much more DSP power than the trans-
channel frequencies within 802.11 DSSS. A station mit mode. The receiver processing has to be com-
can search for an access point giving a better CQ by
pleted in a 1-µs symbol interval, including the
correlator (22-MHz complex-valued signal sam-
looking at all channel frequencies selected for the
ples), adaptive matched filter, phase error con-
interconnected access points. The searching station can trol, and decision unit.
initiate an active mode by sending a so-called probe
The current WaveLAN-II DSP implementation
request message referencing the target set of intercon-
requires more than 2,000 multiply–add operations
nected access points. Each access point will respond to per microsecond, which far exceeds the processing
a probe request with a probe response message. This power—in million instructions per second
will serve as a “solicited” beacon. (MIPS)—of present and next-generation general-
When a station’s CQ decreases with respect to its purpose DSP chips. Therefore, the DSP ASIC
associated access point, the station starts searching
“Theseus” is used for WaveLAN-II as a “horse-
power” DSP. This “Theseus” chip integrates two
more actively. After it has found a second access
analog-to-digital and two digital-to-analog con-
point that gives a sufficiently good CQ, the station verters, which both run at 22 MHz and have a 6-
moves into a handover state and reassociates with bit digital representation.
this second access point. The access points deploy an
inter-access point protocol to inform each other of from the OFF state. The card must be able to make a
station handovers and to correct any intermediate transition from the DOZE state to the fully operational
MAC bridge filter tables. receive (AWAKE) state in just 250 µs. A transition
from the OFF to the AWAKE state will take much
Power Management
more time.
For battery-powered PC devices, the power con-
Power management allows a station to spend
sumption of a LAN card is a critical factor. WaveLAN-II
most of its idle time in the DOZE state, while it is still
supports power management for these environments.
connected to the rest of the network to receive unso-
The 802.11 standard defines power management pro-
licited messages. For the latter requirement, the other
tocols that can be used by stations. Power manage-
stations or the access point must temporarily buffer
ment schemes result in a lower consumption of
the messages that are destined for a station operating
(battery) power compared to traditional operation,
in a power management scheme. Such a station must
where a station is always monitoring the medium dur-
ing idle periods. To achieve savings in power con- “wake up” at regular intervals to check on whether
sumption, a LAN card in a station must have a special any messages are buffered for it.
low-power state of operation called the DOZE state. In 802.11 Power Management Support
this state the LAN card will not monitor the medium Each frame that is transmitted by a station con-
and will be unable to receive a frame. This state differs tains one so-called PM bit to indicate the station’s

Bell Labs Technical Journal ◆ Summer 1997 125

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 Highest quality

No cell search zone

Stop cell search threshold

Stop cell search zone

Regular cell search threshold

Communications
Quality Regular cell search zone

Fast cell search threshold

Fast cell search zone

Dead spot threshold

Dead zone
Lowest quality

Figure 4.
CQ scale and cell search zones.

Figure 5.
Wired infrastructure between access points and multi-channel operation with three different “colored” channel
frequencies.

mode of operation. PM-bit = PS indicates that the sta- points will learn from the PM-bit whether a station is
tion is operating in the power saving mode, and in active or power saving mode and will buffer unicast
PM-bit = A indicates that the station is in the active messages as required by the station. Access points will
mode (continuously active). send beacon frames on a regular basis (for example,
In an access point-based network, the access every 100 ms). In each beacon frame, the access point

126 Bell Labs Technical Journal ◆ Summer 1997

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 Higher-layer High-level
protocol stack WMAC protocols

Multicast Unicast Unicast Management

frame frame/ frame/PM control
no PM frame

PM-bit = PS

PM-bit = A
AP sleep AP sleep
queue for queue for
Multicast multicasts unicasts
announcement (per-station)
sent PS-POLL
received

Normal Priority
transmit transmit
queue queue

Frame transfer

AP – Access point
RF medium
PM – Power management
PS – Power saving
RF – Radio frequency
WMAC – Wireless medium access control

Figure 6.
The 802.11 power management scheme for an access point.

will announce the station whose messages it has Queue Structure in an Access Point
buffered. The stations using a power management To support the power management schemes of
scheme will wake up with high accuracy just before a 802.11, a WaveLAN-II access point will deploy the
beacon transmission and determine, via the beacon, queuing structure shown in Figure 6.
whether any messages are buffered. If no messages are The frame transfer function, shown at the bottom
buffered, the station will go into the DOZE state until of Figure 6, will interface with the RF medium. This
the next beacon arrives. When one or more messages frame transfer block has two input queues, a normal
transmit queue and a priority transmit queue. The pri-
are buffered, the station will stay in the AWAKE state
ority queue schedules internal management and con-
and will poll the access point for transmission of the
trol protocol frames; the normal queue is used for data
buffered messages. At a regular interval, which is a
frames that come from the higher-layer protocol stack.
multiple of the beacon transmission interval, another
There is a single multicast sleep queue and multiple
information field is included in the beacon to
unicast sleep queues, one for each station under con-
announce whether any multicast messages are trol of this access point.
buffered. These messages are transmitted directly fol- When a station polls for a buffered message, the
lowing their announcement. Therefore, the stations buffered message at the head of the unicast sleep
that need to receive these buffered multicast messages queue for that station is passed directly to frame trans-
must stay in the AWAKE state following the multicast fer for highest priority transmission. When a station
announcing beacon. that was in the PS mode sends a message to the access

Bell Labs Technical Journal ◆ Summer 1997 127

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 All unicasts received

No transmit or receive
Sleeping in power for holdover period
save mode
Transmit request

Beacon received,
nothing buffered
Next beacon due
All multicasts Beacon received, Holdover in
received, no unicasts buffered active mode
Polled receiving
unicasts buffered Listen for beacon
buffered unicast
messages
Beacon received,
multicasts buffered All multicasts received,
still unicasts buffered
Receiving buffered
multicast messages

Figure 7.
State diagram for two different power management schemes.

point with the PM-bit set to “A”, all buffered messages directs the receiver to stay in the AWAKE state and
for that station are forwarded to the normal transmit either wait for the multicast messages or actively poll
queue. When a station that was in the active mode the access point for unicasts. After all buffered mes-
sends a message with the PM-bit set to “PS”, all mes- sages are received, the receiver reenters the “sleep-
sages for that station in the normal transmit queue are ing” state.
moved to the unicast sleep queue. The main difference between the two schemes is
Power Management Schemes for a Station in an Access that the enhanced scheme has a special “holdover”
Point-Based Network state, which is entered when either a transmission is
The stations in an access point-based network can requested or when a buffered unicast message is
use the 802.11 power management provisions in a detected. In this “holdover” state, the station tem-
number of schemes. Figure 7 shows two such porarily switches operation from the power save
schemes—the “standard” 802.11 scheme and the mode to active mode and transmits this information
“enhanced” scheme, developed especially for to the access point via the PM bit. The access point
WaveLAN-II. When not involved in actual reception, then forwards all buffered messages immediately and
the receiver will switch between the “sleeping” state will stop sleep queue buffering as long as this state
and the “listen for beacon” state. An accurate timer lasts. The state will last until there is a designated
will wake up the receiver just in time for the next bea- continuous period—about 0.5 to 3 seconds—of no
con frame; when nothing is buffered, the receiver transmit or receive activity for this station, called the
immediately returns to the DOZE state. This power holdover period.
management provision, used in both schemes, yields The enhanced scheme was developed for the fol-
an efficiency rate of more than 99% of idle time spent lowing reasons:
sleeping. The “standard” 802.11 scheme is shown in • In many LAN applications, the exchange of
solid color on the left side and dark gray on the right messages between a station and a server takes
side. The “enhanced” scheme is shown in solid color place in bursts. Multiple transmits and receives
on the left side and light gray on the right side. occur over a short interval, followed by a rela-
When the standard scheme analyzes the beacon tively long period of inactivity.
and detects the presence of buffered messages, it • Keeping the receiver in the active mode prevents

128 Bell Labs Technical Journal ◆ Summer 1997

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 Cell 1 Cell 2

x x = Access point

Carrier level
observed
when access Carrier detect
point transmits threshold 2

Carrier detect
threshold 1

–R1 –R2 0 +R2 +R1

Distance from
access point

Figure 8.
The effect of carrier detect threshold on cell size.

the formation of queues at the access point, WaveLAN-II has three modes of operation for
which fill up during large message transfers (for both 3.3V and 5V supplies: the transmit mode, with a
example, when a file is downloaded to a station). used current draw of 300 mA; the receive mode, with
• The PS-POLL mechanism, which polls each a draw of 250 mA; and the DOZE mode, with a draw
message individually, is inefficient. of 9 mA. Power management can reduce current con-
• When a long sleep time occurs in transac- sumption significantly.
tion-oriented systems, it is useful to keep
the receiver awake long enough after a Scalable System
transmission to cater to the typical applica- The WaveLAN-II system will be deployable in a
tion or protocol stack response time. broad variety of situations, imposing different and
For example, if the application always often conflicting requirements on the system behavior.
responds to a transaction request within In particular, a stand-alone single cell network will
(say) 3 seconds, then the holdover can be have different requirements for the transmit and
set at 3 seconds, ensuring that the station receive behavior than an infrastructure network,
will still be awake to receive the response. which contains multiple overlapping cells. To accom-

Bell Labs Technical Journal ◆ Summer 1997 129

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 Cell 2

x x = Access point

Carrier level
observed
when access Carrier detect
point transmits threshold 2

Defer
threshold

–R2 0 +R2

Distance from
access point

Figure 9.
Ideal relation between the defer threshold and the carrier detect threshold.

modate these variations in operation, the WaveLAN-II Setting the CDT to a lower value will result in a num-
products have a number of built-in provisions to cre- ber of meaningless receive attempts with a high failure
ate scalable systems optimized for environmental and rate. Configurable carrier detection allows the
network usage needs. WaveLAN-II cards to be configured at smaller cell sizes
The power level of a WaveLAN-II transmitter is than the receiver is capable of handling. Small cell
15 dBm (30 mW). When an access point transmits, the sizes play an important role in reusing the same chan-
carrier signal levels observed by a station will decrease nel in a relatively small area.
with distance. Figure 8 illustrates the typical curve for Configurable Defer Behavior
the signal level in two opposite directions. This curve The 802.11 medium access rules (CSMA/CA) are
forms the basis for the scalability control elements. based on the defer and random back-off behavior of all
Configurable Carrier Detection stations within range of each other. The defer decision
Figure 8 shows two values of the so-called carrier is based on a configuration entity called the defer
detect threshold (CDT), defined as “the carrier signal threshold (DT). When a carrier signal level is observed
level, below which the WaveLAN-II receiver will not above the DT level, the WaveLAN-II card holds up a
do a receive.”7 The CDT at level 1 crosses the curves of pending transmission request.
Figure 8 at distances -R1 and +R1, which implies an If we consider the example of CDT level 2 (Cell 2)
associated cell size for this CDT value with radius R1. from Figure 8, the ideal DT value produces a double
This is shown as a bright color circle above the curves. radius, shown in Figure 9. A station on one edge of
The smaller cell, Cell 2 (radius R2), is higher (less sen- the cell defers to a station at the farthest edge. We can
sitive) than level 1. The range for meaningful CDT lev- show that by plotting the curve for one edge station
els has a lower boundary, determined by the and ensuring that the DT level crosses this curve at the
sensitivity of the WaveLAN-II receiver circuitry. other cell edge. Choosing this relation between CDT

130 Bell Labs Technical Journal ◆ Summer 1997

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 Cell 1 Cell 2

x x = Access point

Carrier level
observed
when access
point transmits

Carrier detect
threshold Defer
threshold

–R –R’ 0 +R’ +R

Distance from
access point

Figure 10.
Large cell characteristics.

and DT levels produces a cell in which all stations the area where mutual deferral occurs has a smaller
defer to each other, and where each station can com- size, shown as the smaller color circle of radius R′. The
municate with the access point, thereby avoiding hid- outer area of the cell (radius R) will not have guaran-
den station problems within the group of stations teed deferral. This combination of thresholds can be
belonging to this cell. used with the RTS/CTS medium reservation mecha-
The range for the DT level has a lower boundary, nism to prevent the hidden station phenomenon.
determined by the sensitivity of the WaveLAN-II car- The total cell is referred to as the basic coverage
rier detect circuitry. Below a certain level, the signal area (BCA). In the smaller color area, called the
will not be detected and no defer will take place. The
shared coverage area (SCA), the 802.11 medium-
ideal relation, shown in Figure 9, cannot be achieved if
sharing rules will be in effect. In the ideal setup, the
the CDT is set to the lowest (and most sensitive) level.
SCA is equal to the BCA.
In that case the lowest meaningful DT will not guaran-
tee the wanted deferral between two “edge stations,” Roaming Thresholds
as illustrated in Figure 10. Creating a cellular infrastructure system with the
When we choose a low CDT value, we create a above-defined thresholds for the low-level receiver
large cell size with radius R, shown as the large color and transmitter control requires a proper balance with
circle. When the lowest meaningful DT level is plotted, the roaming thresholds described earlier. The CDT and

Bell Labs Technical Journal ◆ Summer 1997 131

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 1 2 3 1 2
Automatic rate
2 Mb/s
fallback boundary
2 3 1 2
3 4
1 2 3
1 Mb/s

(a) (b)

Figure 11. Figure 12.

High-density versus low-density infrastructure. The relation between data rate and cell regions.

DT determine the transmit/receive behavior of stations Case B in Figure 11 illustrates a maximum cell size
and access points that belong to the same cell, and the setup. Here the SCA is smaller than the BCA. The
roaming thresholds determine the moments for decid- overall capacity is four times that of a single cell (or
ing to start or stop participation in a cell. A station roughly two times with a single channel). This lower
should base its handover decisions on the currently capacity lowers the cost of the total system. Only four
configured capabilities of the receiver. In particular, access points need to be installed versus the ten
when small cell sizes are required, the roaming thresh- required for case A.
olds must be set to ensure that stations will start
searching for a new (better) access point before the Automatic Rate Fallback
receiver becomes physically incapable of receiving The different modulation techniques used for the
messages from the current access point. different data rates of WaveLAN-II can be character-
Infrastructure Density ized by more robust communication at the lower rate.
The ability to define scalable cell sizes translates This translates into different reliable communication
directly into the ability to control the density of cells ranges for the different rates, 1 Mb/s giving the largest
(access points) that cover a certain area. Increasing range. When a WaveLAN-II system is dimensioned for
the number of small cells within an area will the largest cell size, the 1 Mb/s range is chosen.
increase the use of the same channel, thereby maxi- Stations moving around in such a large cell will be
mizing the overall throughput that can be obtained capable of higher-speed operation in the inner regions
by fewer large cells. of the cell. To ensure usage of the highest practicable
Case A in Figure 11 illustrates a high-density cov- data rate at each moment, WaveLAN-II includes an
erage setup. In such a setup, the choice of thresholds is ARF algorithm. This algorithm causes a fallback to the
such that the SCA and BCA of the cells coincide. At lower rate when a station wanders to the outer
this level of the DT, no defer occurs for traffic at two regions and an upgrade to the higher rate when it
stations on the same channel one cell diameter apart. moves back to the inner region. Figure 12 shows the
If we use three independent channels, the roster can two cell regions associated with the two data rates. The
be filled completely. In the example, the raw through- ARF functions come into play when the ARF bound-
put capacity is 10 times that of a single cell. At 2-Mb/s ary is crossed in either direction.
operation, this gives a 20-Mb/s composite capacity. Besides enlarging the range, the lower rates will
When a single channel is used (for instance, forced by also be more robust against other interfering condi-
regulations in Japan), the addition does not work this tions like high path loss, high background noise, and
way because of the influence of adjacent cells. If we extreme multipath effects. The ARF scheme will per-
use the same example, the overall raw capacity would form a temporary fallback when such conditions
be on the order of three to four times the single cell. appear and an upgrade when they disappear.

132 Bell Labs Technical Journal ◆ Summer 1997

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.
 The ARF scheme used in WaveLAN-II for IEEE 2. R. O. LaMaire, A. Krishna, P. Bhagwat, and
802.11 operation, which alternates between 2- and J. Panian, “Wireless LANs and Mobile Net-
working: Standards and Future Directions,”
1-Mb/s operation, is based on keeping track of a tim-
IEEE Communications Magazine, Vol. 34, No. 8,
ing function and missed acknowledgment (ACK) Aug. 1996, pp. 86–94.
frames. Operation at 2 Mb/s is considered the default. 3. H. S. Chhaya and S. Gupta, “Performance of
When an ACK is missed for the first time following Asynchronous Data Transfer Methods of IEEE
earlier successful transmissions, the first retry trans- 802.11 MAC Protocol,” IEEE Personal
Communications, Vol. 3, No. 5, Oct. 1996,
mission is still performed at the same (2 Mb/s) rate. pp. 8–14.
When the ACK is missed again, the second retry and 4. B. Tuch, “Development of WaveLAN, an ISM
subsequent transmissions are also performed at the Wireless LAN,” AT&T Technical Journal, Vol. 72,
fallback (1 Mb/s) rate. A timer is started to track good No. 4, July/Aug. 1993, pp. 27–37.
5. A. Kamerman, “Spread Spectrum Schemes for
ACKs and missed ACKs. When either the timer
Microwave-frequency WLANs,” Microwave
expires or the number of successively received good Journal, Vol. 40, No. 2, Feb. 1997, pp. 80–90.
ACKs reaches 10, the device attempts to upgrade the 6. I. Bar-David and R. Krishnamoorthy, “Barker
rate. A new transmission is sent at the 2 Mb/s rate. Code Position Modulation for High-Rate
When this fails (missed ACK), the system immedi- Communication in the ISM Bands,” Bell Labs
Technical Journal, Vol. 1, No. 2, Autumn 1996,
ately reenters the fallback condition and resumes pp. 21–40.
normal operation. 7. A. Kamerman, “Throughput Density
If higher data rates (5, 8, or 10 Mb/s) are used, the Constraints for Wireless LANs Based on DSSS,”
system follows an enhanced ARF scheme, which dis- Proc. IEEE Fourth Int. Symp. on Spread Spectrum
Techniques and Applications (ISSSTA ‘96), Mainz,
tinguishes multiple regions and aims at optimum data
Germany, Sept. 1996, pp. 1344–1350.
rates per region.
(Manuscript approved August 1997)
Conclusion
WaveLAN-II will be one of the early product AD KAMERMAN is a member of technical staff assigned
implementations of the IEEE 802.11 wireless LAN to the WaveLAN-II System Team in the
Wireless Communication and Networking
standard, offering the broad suite of functions defined
Division of Lucent’s Network Systems in The
in the standard. The modem design and the choice for Netherlands. He works on wireless LANs,
the DSSS technology will enable robust, efficient sys- designing their system performance as it
tem design and seamless roaming facilities for multi- relates to the radio-frequency environment, signal pro-
channel wireless LANs. The added features of the cessing, and network throughput. Mr. Kamerman
received a B.S. and M.S. in electrical engineering from
product, such as radio scalability, automatic rate fall-
Twente University of Technology in The Netherlands.
back, and enhanced power management will further
differentiate the product.
LEO MONTEBAN received a B.S. in mathematical engi-
A key aspect of the WaveLAN-II design is the abil-
neering and an M.S. in mathematical engi-
ity to upgrade its radio to the higher data rates of 5, 8,
neering and computer science from Delft
and 10 Mb/s without major redesign. This design will University of Technology in The Netherlands.
allow the introduction of a new class of high-speed A member of technical staff and part of the
radio LANs that can fully coexist with stations follow- WaveLAN-II System Team in the Wireless
ing the IEEE 802.11 standard, and that can interoper- Communication and Networking Division of Lucent’s
Network Systems in The Netherlands, Mr. Monteban
ate with such stations through access points.
develops software for the WaveLAN-II product family
References and is responsible for its overall system architecture. ◆
1. K. Pahlavan, T. H. Probert, and M. E. Chase,
“Trends in Local Wireless Networks,” IEEE
Communications Magazine, Vol. 33, No. 3,
Mar. 1995, pp. 88–95.

Bell Labs Technical Journal ◆ Summer 1997 133

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 25,2020 at 04:54:12 UTC from IEEE Xplore. Restrictions apply.