EXHIBIT B
US005282222A
United States Patent [ill Patent Number: 5,282,222
Fattouche et 4.
1 1451 Date of Patent: Jan. 25, 1994
(541 METHOD AND APPARATUS FOR tal Audio Broadcasting (DAB) System, P. Hocher, J.
MULTIPLE ACCESS BETWEEN Hagenauer, E. Offer, Ch. Rapp, H. Schulze, Globe-
TRANSCEIVERS IN WIRELESS comP1, CH298&1/91/0000a040, pp. 0040-OW6.
COMMUNICATIONS USING OFDM The Multitone Channel, Irving Kalet, IEEE Transac-
SPREAD SPECI'RUM tions on Communications, vol. 37, No. 2, Feb. 1989, pp.
[76] Inventors: Michel Fattouche, 156 Hawkwood 119-124.
Blvd. N.W., Calgary, Alberta, Optimized Decision Feedback Equalization versus Op
Canada, T3G 2T2; Hatim waul, timized Orthogonal Frequency Division Multiplexing
-
402 1st Avenue, N.E., Calgary, for High-Speed Data Transmission Over the Local
Alberta, Canada, T2E OB4 Cable Network, Nikolaos A. Zervos and Irving Kalet,
[21] Appl. No.: 861,725 CH2655-9/89/0000-1989 IEEE, p. 1080-1085.
[22] Filed: M u . 31,1992 (List continued on next page.)
[51] lot. (3.5 ............................................... HWK 1/00
[52] US. CI. ........................................... 375/1; 380/34 Primary Examiner-Tod R. Swann
[58] Field of Seuch .............................. 380/34; 375/1; Attorney, Agent, or Finn-Daniel L. Dawes '
36W724.01, 827 [571 ABSTRAC~
[561 References Cited A method for allowing a number of wireless transceiv-
U.S. PATENT DOCUMENTS ers to exchange information (data, voice or video) with
each other. A fvst frame of information is multiplexed
4,601,005 7/1986 Kilvington . over a number of wideband frequency bands at a fmt
4,623,980 11/1986 Vary .................................... 3641724
4,893,266 1/1990 Deem . transceiver, and the information transmitted to a second
4,914,699 4/1990 Dunnet al. ......................... 375/1 X transceiver. The information is received and processed
5,034,911 7/1991 Rachels . at the second transceiver. The information is differen-
5,063,560 11/1991 Ycrbury et al. .................... 375/1 X tially encoded using phase shift keying. In addition,
5,089,982 2/1992 Grm et al. . after a pre-selected time interval, the fvst transceiver
5,151,919 9/1992 Dent ........................................375/1 may transmit again. During the preselected time inter-
OTHER PUBLICATIONS val, the second transceiver may exchange information
with another transceiver in a time duplex fashion. The
Reduction of Multipath Fading Effects in Single Vari- processing of the signal at the second transceiver may
able Modulations, M. A. Poletti and R. G. Vaughan,
ISSPA 90 Signal Processing Theories, Implementations include estimating the phase differential of the transmit-
and Applications, Gold Coast, Australia 27-31 Aug., ted signal and pre-distorting the transmitted signal. A
1990, 672-676. transceiver includes an encoder for encoding informa-
OFDM for Data Communication over Mobile Radio tion, a wideband frequency division multiplexer for
FM Channels; Part 11: Performance Improvement1 by multiplexing the information onto wideband frequency
E. F. Casas and C. Leung, Department of Electrical voice channels, and a local oscillator for upconverting
Engineering University of British Columbia, Vancou- the multiplexed information. The apparatus may in-
ver, B.C., Canada V6T 1W5. clude a processor for applying a Fourier transform to
OFDM for Data Communication Over Mobile Radio the multiplexed information to bring the information
FM Channels-Part I: Analysis and Experimental Re- into the time domain for transmission.
sults, Eduardo F. Casas and Cyril Leung, IEEE Trans-
actions on Communications, vol. 39, No. 5, May 1991,
pp. 783-793.
Performance of an RCPC-Coded OFDM-based Digi-
OTHER PUBLICATIONS Transactions on Communications, vol. Corn-28, No. 1,
Jan. 1980, pp. 73-83.
Advanced Groupband Data Modem Using Orthogo- An Improved Method for Digital SSB-FDM Modula-
nally Multiplexed QAM Technique, Botaro Horosaki, tion and Demodulation, Rikio Maruta and Atsushi
Satoshi Hasegawa, and Akio Sabato, IEEE Transac- Tomozawa, IEEE Transactions on Communications,
tions on Communications, vol. COM-34, No. 6, Jun. vol. Corn-26, No. 5, May 1978.
1986, pp. 587-592. Data Transmission by Frequency-Division Multiplex-
A 19.2 Kbps voiceband data modem based on orthogo- ing Using the Discrete Fourier Transform, S. B. Wein-
nally multiplexed QAM techniques B. Hirosaki, A. stein aad Paul M. Ebcrt, IEEE Transactions on Com-
Yoshida, 0. Tanaka, S. Hasegawa, K. Inoue and K. munication Technology, vol. Com-19, No. 5, Oct.,
Watanabe, CH2175-8/85/-1, IEEE, pp. 1971, pp. 628-634.
661-665. Performance of an Efficient Parallel Data Transmission
System, Burton R. Saltzberg, IEEE Transactions on
Analysis and Simulation of a Digital Mobiie Channel Communication Technology vol. Corn-15, No. 6, Dec.,
Using Orthogonal Frequency Division Multiplexing, 1967, pp- 805-81 1.
Leonard J. Cimini, Jr., IEEE Transactions on Commu- A Theoretical Study of Performance of an Orthogonal
nications vol. Com-33, No. 7, Jul., 1985, pp. 665-675. Multiplexing Data Transmission Scheme, Robert W.
An Orthogonally Multiplexed QAM System Using the Chang, Richard A. Gibby, IEEE Transactionson Corn-
Discrete Fourier Transform, Botaro H i r d , IEEE munication Technology, vol. Com-16, No. 4, Aug.,
Transactions on Communications, vol. Com-29, No. 7, 1968, pp., 529-540.
Jul. 1981, pp. 982-989. u-
Synthesis of Band-Limited Orthogonal Signals for M l
tichannel Data Transmission by Robert W. Chang, The
An Analysis of Automatic Equalizers for Orthogonally Bell System Technical Journal, Dec. 1966, pp.
Multiplexed QAM Systems Botaro Hirosaki, IEEE 1775-1796.
U.S. Patent Jan. 25,1994 Sheet 1 of 23
U.S. Patent Jan. 25,1994 Sheet 2 of 23
U.S. Patent Jan. 25, 1994 Sheet 3 of 23 5,282,222
U.S. Patent Jan. 25, 1994 Sheet 4 of 23 5,282,222
U,S, Patent Jan. 25, 1994 Sheet 5 of 23
U.S. Patent Jan. 25, 1994 Sheet 6 of 23
U.S. Patent Jan. 25,1994 Sheet 7 of 23
U,S, Patent Jan. 25, 1994 Sheet 8 of 23
V)
8 %%
"E"
Repeat right & Left Repeat right & left
w~thout overlap with overlap
followed by a raised followed by a
cosine window rectangular window
(last 2 blocks in processor) (last 2 blocks in de-processor)
Flg. 6c
U.S. Patent Jan. 25,1994 Sheet 10 of 23
..
U S Patent Jan. 25,1994 Sheet 1 of 23
1 5,282,222
at each frequency
and calculate the
amplitude and phase
I
Calculate
A fn(A(9) -
Calculate
t i
Calculate Calculate
o (n) + A w (n) -
o(n) A w (n)
1 1
1
t
Calculate Calculate
P= w,(n) - (c~(n) A w (n))I
+ N =C[w,(n)-(w(n) - Ao (n))]
Yes No ,
t t
Use Use
o (n) + A o (n) -
w (n) A w (n)
to demodulate to demodulate
Fig. 7b
U.S. Patent Jan. 25, 1994 Sheet 12 of 23 5,282,222
Fig. 8a
Fig. 8b
U,S, Patent Jan. 25, 1994 Sheet 13 of 23 5,282,222
US. Patent Jan. 25,1994 Sheet 14 of 23
U.S. Patent Jan. 25,1994 Sheet 15 of 23
U.S. Patent Jan. 25,1994 Sheet 16 of 23
U.S. Patent Jan. 25, 1994 Sheet 17 of 23
U.S. Patent Jan. 25, 1994 Sheet 18 of 23 5,282,222
U.S. Patent Jan. 25, 1994 Sheet 19 of 23
U.S. Patent Jan. 25,1994 sheet 20 of 23 5,282,222
Digital Signal Processors
+
repeat RC
128 128 pt 128 . 384 p window 160 +
samples lFFT 'sample: "9"
& eft samples 0=0.25 ~amples
Processor
Fig. 14a
f
160
right 422 Rect. 128 128pt 128
samples &left samples* wiridow wsampleF FFT a p
om Zl
A .
De-Processor
flg. 14b
Repeat right & Left Repeat right & left
w~thout overlap with overlap
followed by a raised followed by a
cosine window rectangular window
(lad 2 bkd
Fig. 14c
Estimate Group Delay o(n)
aa
128
estimate A'(n) Arithmetic
'
from Operation ransform of m(n)
Deprocessor -.
Fig. 15
5,282,222
1 2
cochannel interference and little or no intersymbol in-
METHOD AND APPARATUS FOR MULTIPLE terference.
ACCESS BETWEEN TRANSCEIVERS IN The new access technique can offer up to 38 times the
WIRELESS COMMUNICATIONS USING OFDM capacity of analog FM.It includes in one aspect wide-
SPREAD SPECIlRUM 5 band orthogonal frequency division multiplexing of the
information to be exchanged, and may include slow
FIELD OF THE INVENTION Frequency Hopping 0. The technique is imple-
This invention relates to voice md data transmission &
mented Using D &Signal Processors @SP) repiac-
in wireless communications, and particularly between ing conventional analog devices. The system operates
fixed and portable transmitters and receivers. 10 with relatively small cells. In other aspects, dynamic
channel allocation and voice activation may be used to
CLAIM TO COPYRIGHT improve the capacity of the system.
A portion of the disclosure of this patent document Advantages of the present invention include:
contains material which is subject to copyright protec- 1. It can be used indoors as well as outdoors using the
tion. The copyright owner has no objection to the fac- l5 same transceivers. If data is to be exchanged, as op-
simile reproduction by anyone of the patent document, posed to voice, the transceiver preferably contains an
as it appears in the Patent and Trademark Ofice patent estimator to allow pre-distortion and post-distortion
file or records, but otherwise reserves all copyright of the transmitted signal.
rights whatsoever. Software for carrying out of 2. The system, as compared with prior art systems omits
the method described in this patent document has been 2O the clock or c A e r recovery, automatic gain control
filed with the Patent and Trademark Oflice in the form or passband limiter, power amplfier, an equalizer or
of a microfiche and includes 55 frames including a title interlurV,-dehterlmr, and therefore has low
frame. complexity.
BACKGROUND AND SUMMARY OF THE 3. The system offers good speech quality, as well as low
25
INVENTION probabilities of dropped and blocked calls. It is robust
against Doppler and multipath shifts. It is also robust
This patent document presents a new against both impulse noise and narrowband interfer-
technique for Personal Communication Networks ence.
(PCN)- communication networks Ire net- 4. The system is flexible, such that at the expense of
works that allow individuals and equipment to ex- 3 0
change information with each other anywhere at any- increased complexity of the DSP it be
time through voice, data or video. PCN typically in- applied over noncontiguous bands. This is accom-
clude a number of transceivers, each capable of trans- plished by dividing a 100 MHz (in one of the exem-
and receiving infomation (voice, data or video) plary embodiments described band several
in the form of electromagneticsignals. The transceivers 35 subbands accommodating integer number
may be fued or portable, and may be identical or one or
more of them may be more complex. 5. The system offers low frame delay (less than 26.2 m s
system must allow the transceivers to access in the cellular embodiment described
-h other to enable the exchange of information. When here)+The transceiver requires low average transmit-
there are a number of transceivers, multiple access, that 40 ted power (of the order 20 pw in the
is, access by more than one transceiver to another trans- cellular embodiment described here) which means
cciver, must be allowed. power saving as well as enhanced biological safety.
One of the constraints of designing a PCN is that a 6. The system offers up to a 38 fold increase in capacity
transceiver, or portable radio unit, must be small in size. over the North American Advanced Mobile Phone
The smaller the unit, the better for portability. The 45 System (AMPS) which uses analog frequency modu-
small size of the units means only small and light-weight lation.
power sources be used. ~f the portable is to be used Operation of the system in accordance with the tech-
for any length of time, it must therefore consume mini- niques described in this disclosure may permit compli-
ma1 power. ance with technical requirements for spread spectrum
Also, to allow use of the radio frequency spectrum 50 systems.
without obtaining a license in North America, the sys- There is therefore disclosed in one aspect of the in-
tern must use a sprurd spectrum and satisfy federal regu- vention a method for allowing a number of wireless
htions. In part, these regulations impose limits on the transceivers to exchange information (data, voice or
power and the frequency spread of the signals ex- video) with each other. In the method, a first frame of
changed between the transceivers. An object of an as- 55 information is multiplexed over a number of frequency
pect of this invention is to satisfy those requirements. bands at a first transceiver, and the information trans-
Also, t r w v e r s talk to each other over a fmed mitted to a second transceiver. In a cellular implements-
bandwidth. &cause of the limited availability of the R F tion, the second transceiver may be a base station with
spectrum, the system must be bandwidth efficientyet at capacity to exchange information with several other
the same time maintain high quality exchange of infor- 60 transceivers. The information is received and processed
mation at a l times in one of the most hostile channels
l at the second transceiver. The frequency bands are
known in communication. The new multiple access selected to occupy a wideband and are preferably con-
technique proposed here addresses all these issues. tiguous, with the information being differentially m-
The new access technique has a low Bit Error Proba- coded using phase shift keying.
bility (BER) as well as a low probability of dropped and 65 A signal may then be sent from the second trans-
blocked calls. This is due to the fact that the access ceiver to the first transceiver and de-processed at the
technique is robust against multipath, Doppler shifts, first transceiver. In addition, after a preselected time
impulse noise and narrowband interference. It has a low interval, the first transceiver transmits again. During
5,282,222
3 4
the preselected time interval, the second transceiver FIG. 7a is a schematic showing the structure and
may exchange information with another transceiver in a function of the channel estimator in FIG. 56;
time duplex fashion. FIG. 7b is a flow chart showing the operation of the
The processing of the signal at the second transceiver channel estimator of FIGS. Sb and 7a;
ma include estimating the phase differential of the trans- 5 FIGS. &, 8b and 8c are respectively schematics of
mitted signal and predistorting the transmitted signal. 126, 63 and 7 cell reuse patterns;
The time intervals used by the transceivers may be FIGS. 9a and 96 are schematics showing transmit
assigned so that a plurality of time intervals are made protocols according to one aspect of the invention;
available to the fmt transceiver for each time interval FIG. 10 is a schematic showing the use of the avail-
made available to the second transceiver while the ftrst 10 able frequencies according to another aspect of the
transceiver is transmitting, and for a plurality of time invention for use with local area network applications;
intervals to be made available to the second transceiver FIG. 1 0is a schematic showing an idealized pulse
1
for each time interval made available to the first trans- for transmission over a local network system;
ceiver otherwise. Frequencies may also be borrowed by FIG. 118 is a schematic showing a modified version
one base station from an adjacent base station. Thus if 15 of the pulse shown in FIG. 110;
one base station has available a first set of frequencies, FIG. l l c is a schematic showing a further modified
and another a second set of distinct frequencies, then a version of the pulse shown in FIG. llo;
portion of the frequencies in the first set may be tempo- FIG. 12 is a schematic showing a preferred protocol
rarily re-assigned to the second base station. for local area network communication;
In an implementation of the invention for a local area 20 FIG. 1& is a block diagram showing the structure
network, each transceiver may be made identical except and function of an embodiment of the transmimr of a
for its address. local area network transceiver according to the inven-
Apparatus for carryingdout the method of the inven- tion;
tion is also described here. The basic apparatus is a FIG. 13b is a block diagram showing the structure
transceiver which will include an encoder for encoding 25 d function of an embodiment of a further local area
,
information, a wideband frequency division multiplexer network transceiver according to the invention;
for multiplexing the information onto wideband fre- FIG. 13c is a block diagram showing the structure
quency voice channels, and a local oscillator for upcon- and function ofan embodiment ofthe receiver ofa local
verting the multiplexed information. The apparatus may area network transceiver according to the invention;
include a processor for applying a Fourier transform to 30 FIG, is a flow diagram showing the function of
the multiplexed information to bring the information the processor in either of FIGS. or 13b;
into the time domain for transmission. FIG. 146 is a schematic showing the function of the
BRIEF DESCRIPTION OF THE DRAWINGS deprocessor in either of FIGS. 13; or 1%;
FIG. 14c is a schematic further illustrating the opera-
There will now be described a preferred embodiment 35 tion of the processor and deprocessor shown in FIGS.
of the invention, with reference to the drawings, by and 14b; and
way of illustration, in which like numerals denote like FIG. is a =hematic showing the stmcture and
elements and in which: function of the channel estimator in FIG. 13b.
FIGS. l a and 1b are schematicsof a prior art receiver
and transmitter respectively; 40 DETAILED DESCRIPTION OF PREFERRED
FIG. 2 is a schematic showing the use of the available EMBODIMENTS
frequencies according to one aspect of the invention for Introduction
use with cellular applications;
FIG. 30 is a schematic showing an idealized pulse for The benefits of the invention can be readily appreci-
transmission over a cellular system; 45 ated with reference to FIG. 1, which shows a prior art
FIG. 3b is a schematic showing a modified version of transmitterheceiver configuration for a portable Unit.
the pulse shown in FIG. 3a; The transmitter includes a vocoder 110, an interleaver
FIG. & is a schematic showing a further modified 112, a modulator 114, a filter 116, local oscillator 118,
version of the pulse shown in FIG. 3a; power amplifier (PA) 120 and antenna 122. The re-
FIG. 4 is a schematic showing an exemplary protocol so ceiver includes an LNA 124, a local oscillator 126, a
for cellular communication; filter 128, automatic gain control (AGC) 130 with an
FIG. Sa is a block diagram showing the structure and associated passband hardlimiter not separately shown,
function of an embodiment of the transmitter of a cellu- carrier recovery 132, sampler 134, clock recovery 136,
lar portable in accordance with the invention; adaptive (or fued) equalizer 138, demodulator 140,
FIG. Sb is a block diagram showing the structure and 55 deinterleaver 142 and decoder 144. With implementa-
function of an embodiment of the transmitter and re- tion of the present invention, several of the blocks
ceiver of a cellular base station in accordance with the shown in FIG. 1are not required. Thest are the inter-
invention; leaver 112, deinterleaver 142, power amplifier 120, au-
FIG. 5 is a block diagram showing the structure and tomatic gain control 130 with passband hard-limiter,
function of an embodiment of the receiver of a celIular 60 clock recovery 136 and carrier recovery 132, and the
portable in accordance with the invention; equalizer 138. It will now be explained how the pro-
FIG. 6a is a flow diagram showing the function of the posed system obtains the omission of these blocks with-
processor in either of FIGS. 5a or 5b; out impairing the quality and capacity of the system.
FIG. 6b is a schematic showing the function of the In this disclosure there will be described two systems
deprocessor in either of FIGS. Sb or Sc; 65 as examples of the implementation of the invention. The
FIG. 6c is a schematic further illustrating the opera- system described first here will apply to a cellular sys-
tion of the processor and deprocessor shown in FIGS. tem with a number of portable transceivers and base
60 and 6b; stations (BS). Then will be described a local area net-
5,282,222
5 6
work implementation. A local area network will typi- is duration of one time domain sample and x is any real
cally be a system of equal transceivers. The invention value, the shift is equal to 2aAfxT. Hence, T causes a
may also be implemented with combinations of cellular shift in the phase difference between adjacent symbols
and local area networks, or to a system with a number of value 27rxAC1 since T is equal to l/(KlAf). By dou-
of equal transceivers and a master or controlling trans- 5 bling the number of symbols from KI to 2K1 the shift in
ceiver. "4ual" as used here means that the transceivers the phase difference is reduced by half from 2ax/X1 to
have more or less the same processing equipment and ax/Kl. Thus, the effect of the clock error on the BER
processing capabilities. The system described here is is reduced by incrming K.
primarily for the exchange of voice information. (2) When there is relative motion between the trans-
Link set-up and termination protocols between trans- 10 mitting transceiver and the receiving transceiver, a
ceivers, and the equipment required to implement them, ~~~~l~~ shift occurs with a maximum value
are understood in the art as as the basic strut- IV/AI where V is the relative velocity between the two
ture of radio transceivers that may be used to implement transceivers and is the wavelength of the travelling
the invention. Hence these elements are not described wave corresponding to the frequency fc(i.e. f,is
here. What is described here are the novel operational, 15 the frequency corresponding to the middle point in the
functional and structural elements that constitute the frame). Such a Doppler shift a ssmpling error in
invention.
the frequency domain of the same amount, or cquiva-
Cellular Implementation of Wideband Modulation lently, it causes a sampling error of V/(AAf) relative to
The present invention proposes in one embodiment a 20 one symbol sample. Thus, the effect of the Doppler shift
On the BER reduced by increasing Af'
wideband modulation scheme for exchange of informa-
tion between transceivers such as portables and base (3) When a between the Lo in the
stations. transmitter and the one in the receiver occurs with a
Wideband in this patent document is described in the value fh it causes a sampling error in the frequency
context of Wideband-Orthogonal Frequency Domain 25 domain of the %me amount* equivalently,it causes a or
Modulation (W-OFDM or wideband OFDM). In sampling error of fdAf relative to one symbol sample.
OFDM, the entire available bandwidth B is divided into Thus1 the effect on the BER of the f r e q u m c ~ offset
a number of points K, where adjacent points are =pa- between the LO in transmitter and the one in the re-
rated by a frequency band Af, that is B=KAf. The K ceiver is reduced by increasing Af.
points are grouped into a frame of K1 points and two tail 30 In summary, OFDM with a K and a Af large enough
o +
slots of K2 points each, s that K =K1 2K2. The frame to be able to achieve a specific throughput and large
cames the information intended for transmission under enough to be able to avoid using either a clock or a
the form of multilevel differential phase shift keying camer recovery device without substantially affecting
(MDPSK) symbols or differential quadrature amplitude the BER is referred to here as Wideband-OFDM. As an
modulated (DQAM) symbols. Thus each point in the 35 example, let us assume that MDPSK is used in an
frame corresponds to one information symbol. The two OFDM system with the number M of levels, with a
tail slots act as guard bands to ensure that the out-of- carrier frequency f,, with a raised cosine pulx of roll-
band signal is below a certain power level. For example, off P, with the LO at the receiver having a frequency
(
when a pulse P 0 is selected for pulse shaping and the offset forelative to the LO at the transmitter (so that the
out-of-band signal has to be ydB or less relative to the frequency offxt betwen the carrier frequencies in the
in-band signal, K2 is selected such that first and second transceivers of the multiplexed infor-
mation is fo), with a given maximum expected clock
2O.loglol P(f)/F(O) I B y for fZK2Af.
error 7=xT at the receivi~g transceiver, where T is the
duration of one time domain sample, and with a maxi-
When the pulse is a raised-cosine pulse with a 6 q5 mum expected relative velocity V betwetn the trans-
and when the number of levels each 'yrnbol can take 's
M the bit rate is equal to Kllog2M/(fjt+(l +B)/A9
Y ceivers. Thus, in order to ensure that the out-of-band
where (1 +B)/Af is the duration of the frame and 6t is signal is ydB or less to the in-band signal and to
the guard time required to take into account the delay be able to avoid using either a clock or a carrier rccov-
of arrival and the delay spread due to multipath. In this ery device without substantially affecting the BER we
case, the bandwidth eff~ciency, which is defined as the rn have to:
ratio between the bit rate and the bandwidth, is equal to: Find the Af to
one symbol sample, which does not substantially
lomM/((I+B+~~A~KI+~K~/KI)) affect the BER. This can be done using the following
rules:
In wideband-OFDM, both K and Af are selected 55
sufficiently large to achieve a high throughput as well When 0.2SBS0.3, A f =7.50%
as to reduce the effects on the BER of the clock error, When 0.3SBS0.4, A f =10.0%
the Doppler shift and the frequency offset between the
LO in the transmitter and the one in the receiver. To Whm 0.4ZBS0.5, A f = 12.5%
show what is meant by "K and Af are selected SUE- 60
ciently large", consider the effect of increasing K and when 0.55BS0.6, ~f = 15.0%
Af on (1) the clock error, (2) the Doppler shift and (3)
the frequency offset between the LO in the transmitter 2. Find Af such that:
and the LO in the receiver.
(1) When a clock error at a transceiver of value T 65 v/(~if)+f~fziaf
occurs in the time domain, it causes a shift in the phase
difference between adjacent symbols in the frequency 3. Find K2 such that
domain of value 2aAfr. When T is equal to XTwhere T
5,282,222
7 8
? . o l l ( ) P O I dy for fZK2af
OlgoPf/() to ensure a transmission delay to allow one transceiver
to communicate with other transceivers at the same
4. Find K1 such that time, but must not be so long that the delay becomes
unacceptable to the user. Delays longer than about 40
2nx/P1
5 ms are too great for voice, and it is preferable to be
In this case, we refer to OFDM as Widebmd-OFDM. lower' For data, the may be longer and be
Element 4 is a necessary condition for wideband
O m M , and given a sampling error, the emor In the exemplary embodiment described here,
may be corrected with the methods described in this bit rates are considered for the vocoder: 18.77 Kbps,
patent document. 1 9.16 Kbps and 6.1 8 Kbps. Table I displays the structure
0
To implement wideband modulation, orthogonal of a VC slot and the number N of VC for each vocoder
Frequency Division Multiplexing (OFDM) is preferred rate. The control symbols in each vc slot are required
in which the information, for example encoded speech, for handoff and power control. FIG. 2 shows that N vc
is multiplexed over a number of contiguous frequency can be transmitted simultaneousl~.This is known as
bands. Wideband OFDM forces the channel to be fie- l5 Frequency Division Multiple Access. FIG. 3 shows c
quency selective and causes two types of linear distor- that 126 full duplex frames can be transmitted every
tion: amplitude distortion and phase distortion. To re- 13.104 ms in a Time Division Multiple Access fashion
duce the effect of amplitude distortion the modulation is (TDMA). The total number of Full Duplex voice chan-
preferably phase modulation, while the effect of phase nels (FDvc) is therefore 126X N and is shown in Table
distortion is reduced by employing differential phase 20 I.
modulation. Hence the modulation may be referred to To ensure that the channel is slowly fading, a Time
as Differential OFDM (DOFDM). Unlike in other pro- Division Duplex protocol for exchange of information
posed schemes, neither pilot tones nor diversity are between the portable and the base station is proposed as
required in DOFDM. Possibly, quadrature amplitude illustrated in FIG. 4. The protocol is as follows:
modulation might be used, but amplitude modulation 25 1. The portable transmits a frame 410 over one vc slot.
makes it difficult to equalize the distorting effects of the See the discussion in relation to FIG. 50 below.
channel on the signal. 2. The Base Station (BS) receives the frame 410 from
T o implement wideband modulation in a cellular the portable and processes (analyzes) it as shown and
system with a plurality of portables and one or more discussed in relation to FIG. 5b below.
base stations, a 100 MHz band is divided into 4096 30 3. ~~d on the received signal, the BS prediStOrtS a
points, as shown in FIG. 2, plus two tail slots of 195.3 frame 420 and transmits it to the portable over the
KHz each. The 4096 points represent channels same vc slot, 520 p or some other~uitable time inter-
(vc). Adjacent points are 24.414 KHz and val later in which the channel does not &ange sub-
each point represents a Differential eight Phase Shift ,-tially. The time interval will depend on factors
@8PSK) di(")* where 35 such as the frequency, speed of the transceiver and
+
f;b- +(n) ~ ( ~4(n) takes One the eight
1 . other environmental factors.
. ..
10, 2 ~ / 8 , 4 ~ / 8 . 14m/8) with equal probability for
n= 1, 2, .. . ,4096 and +(0) takes an arbitrary value. 4. The portable receives the frame from the BS.
discussion in relation to FIG. Sc below.
the
X(n) also takes an arbitrary value. ~ ( nmay be used as
)
a security key and will be known only to the transmitter 40 5. Steps 1 through 4 are repeated, as for example by the
and receiver. in the form of output bits transmission of the next frame 430, every 13.104 ms
from a vocoder are mapped onto +(n). Vocoders are the is terminated'
well known in the art and do not need to be described 520 pS9a portable Outdoor at loo
in detail here. The focus here is to transmit the bits with km/hr cm9 which leaves the Outdoor
an acceptable Bit Error Rate, i.e. with a BER 5 10-2 45 largely unchanged. Indoors, a portable moving at 2
for voice and S 10-8 for data. m/s moves 0.1 cm again leaving the channel un-
ideally, 40.96 ps (= 1/24.414 K H ~ ) the minimum
is changed. Assuming that the channel is reciprocal and
duration required for one frame to be transmitted with- Over 520 ps* a predistorted signal, transmit-
out frequency domain intersymbol interference. This ted by the BS9should reach the portable undistorted.
can be achieved using a Raised Cosine (RC) pulse with 50 49 One can see that the portable trans-
zero roll-off, as shown in FIG. 3a. FIG. 3a illustrates a mits/receives one FDvc every 13.104 ms, while the BS
rectangular (time domain) window corresponding to p
Can transmit/receive U to 21 frames or equivalently up
the RC (frequency domain) pulse. Such a pulse, how- to 21 X N FDvc every 13.104 ms- The frames 440 la-
ever, requires an infinite frequency band. TO alleviate belled frame 2 . . . frame 21 are frames that may be
such a requirement, an RC pulse with a 20% roll-off 5 transmitted to other portables. This implies that while
5
(i.e. fi =0.2) may be used as shown in FIG. 3b. The one BS is processing its data over 520 ps, six other BS
frame duration has increased by 20% to 49.152 ps. The can communicate to their corresponding portables in a
two tail slots of 195.3 KHz each (i.e. 8 points each) Time Division Multiple Access (TDMA) fashion using
ensure that the signal outside the entire band of 100.39 the same frequency bands. Also, during the 13.104 ms,
MHz is below -50 dB. T o allow the frame to spread 60 or such other preselected time interval that is suitable,
over the time as a consequence of the multipath nature the BS may communicate with one or more other porta-
of the channel, an excess frame duration of 2.848 p is bles.
provided as shown in FIG. 3c, making the frame dura- When a portable is stationary during a call, it is possi-
tion 52 ps in total. ble with high probability to have the transmitted signal
Since the frame duration is 52 ps, the frame rate is 252 65 centered with several deep (frequency domain) nulls,
frames per 13.104 ms or equivalently, 126 full duplex hence, causing speech degradation. Also, narrowband
frames may be transmitted/received every 13.104 ms. interference over the vc slot can deteriorate the speech.
The reason for pre-selecting an interval of 13.104 ms is In order to avoid both situations, the signal is preferably
5,282,222
9 10
frequency hopped into a new vc slot within the same low, the transmitted power is increased and if the aver-
(frequency domain) frame. This frequency hopping is age power is too high, the transmitted power is de-
ordered by the BS which is constantly monitoring the creased. The power controller 525 may also be used in
channel frequency response. Monitoring techniques, as frquency hopping to monitor the average power of the
well as frequency hopping, arc known in the art, and 5 received frequency and determine when frquency
not described here further. When an unacceptable hopping need take place.
speech degradation is first noticed by the BS a proba- FIG. Sc correspondsto step 4, and shows the receiver
ti0n period i initiated and maintained for at least 10
s of the portable, which is the same as the receiver in the
cycles 6.e. lox 1 . 0 ms) unless speech degradation
314 BS except it does not include an estimator or a power
ceased-In other words, the probation period is lo controller. These are not required in the portable on the
terminated if speech degradation hasceased. Frequency assumption that the BS will carry out the phase estima-
hopping is ordered at the end of the probation tion and the power control. However, if desired, the
period. The period of 1 cycles is long enough to indi-
0 portable may include these functions.
the portable stationarit~and is enough to FIGS. 6a,6b and 6c illustrate the function and-struc-
dow s ~ h betwen unacceptable l5 ture of the processor and the respectively
speech hence maintaining good s h
~ qdity. in the transmitter and receiver. Software for modelling
As known in the art, the BS ensures that no collisions the function of the processor in a general purpose
take place bctwem hopping portables. puter has been filed with the Patent and Trademark
Digital Signal Processing Office as frames 3 to 26 of the microfiche appendix and
20 for modelling the function of the deprocessor has been
The transrnitter/receiver block diagrams correspond- filed with the Patent and Trademark Ofice as frames
ing to the protocol in FIG. 4 are shown in FIGS. So, 5b 27-41 of the microfiche appendix.
and &7 step in the protocol FIG. 6a shows that the processor is a DSP implcmen-
above- 'peech is provided to a vowtier tation of an RC inverse Fourier transform. nK roll-off,
where the speech is digitized and coded to create bits of 25 followed by an pulse shaping filter with a 20% proces-
information. The bits are provided to the modulator 512
which turns them into D8PSK symbols, with three bits sor first inverse Fourier transforms the 4096 D8PSK
per symbol. me DgPSK symbols are then processed in Output the The
the processor 514 which is described in more detail in transformed symbols are then triplicated as a group so
FIG. 6a. The output from the processor is then filtered 30 that the number -pits is trip1ed* with three
in low pass filter 516, upconverted to RF frequencies consecutive groups each the 4096 trans-
using local oscillator 518 and transmitted by antenna formed symbols. The triplication of the signal is illus-
520. Figure Sb corresponds to steps 2 and 3. trated in FIG. 6c, where the symbols are shown as first
In FIG. 56,the received signal at the base station is and added Next, as shown in FIGS.
filtered in a bandpass fdter 522,and down converted by 35 and 6cpthe groups are by a Raised
mixing with the output of a local 524. ~h~ Cosine window with a roll-off of 0.2 centered in the
average power of the downcoverted signal is monitored the three groups- In other words*the pr--
by a power controller 525 that adjusts the average takes D8PSK in* pulse shapes them and
power to the specificationsrequired by the sampler 526. invqse Fourier transforms them. Onthe hands the
The adjusted downconverted bits are then sampled in 4 de~rocessorundoes what the processor did, i.e. it re-
sampler 526 to produce bits of information. The bits are m o v e the pulse shaping~ then Fourier transforms the
then in the deprocessor ~ a , described in received signal to obtain the original D8PSK symbols.
more detail in FIG. An estimate of the phase differ- The first two blocks in FIG. 6b are similar to the second
ential of the received signal is taken in the channel =ti- two blocks i FIG. 6a except for two differences. The
n
mator 530, as described in more detail in relation to 45 two differences are as follows. In the first block of the
FIG. 7a and 76 below, and the estimated phase differen- r~
d e ~ r m m the repeated groups of symbols are Par-
tial is supplied to a decoder-demodulator 532 to correct tially overlapped as shown in FIG. & on the right hand
the received bits. The estimated phase differential is also side. In the second block, a rectangular window is used
supplied to a pre-distorter 534 in the transmitter. At the instead of the Raised Cosine. In the preferred imple-
transmitter in the B= Station, the m e blocks are 50 mentation, the blocks are repeated three times but other
incorporated as in the portable transmitter except that a numbers of repetition may be used.
predistorter is used to alter the phase of the D8PSK FIGS. 6a, 6b and 6~ show that the DSP blocks used
symbols to make the channel appear Gaussian (ideal) as in the Processor are identical to the ones used in the
opposed to a fading channel. The predistorter 534 re- d e p r m s o r , except for a small change in the two trans-
ceives a signal corresponding to the estimated phase 55 forms and a small change in the shapes of the two win-
differential of the channel. On the (believed reasonable) dows. Thus the m e hardware can be used by both the
assumption that the channel is reciprocal, the signal processor and the deprocessor.
beiig transmitted is predistorted with the estimated FIG. 7a shows a block diagram of an example of a
p h w difllerential s that the received signal at the por-
o preferred channel estimator, and FIG. 7b is a flow chart
table with which the BS is communicating will be cor- 60 showing the operation of the phase estimator. Each of
rected for any phase distortion over the channel. The the steps is carried out in a computing means that may
advantage of rendering the channel Gaussian is a large be a special purpose computer or a general purpose
saving in the power required to achieve an acceptable computer programmed to carry out the digital signal
BER. The initial power control 525 also sends a signal processing described here, as for example with the soft-
to the predistorter 534 to adjust the transmitted power 55 ware that has been filed with the Patent and Trademark
to an appropriate signal level for the sampler 526 in the Office as frames 42-55 of the microfiche appendix.
portable's receiver depending on the average power of Other methods of estimating the channel may be used
the received signal. Thus if the average power is too that obtain an estimate of the channel group delay or
5,282,222
11 12
phase differential of the transmitted symbols. However, N=X Io{o(n)-Ao(n)) I are calculated. If P
a preferred implementation is described here. o(n)+ Ao(n) is used to correct the signal, and if not then
The first block in FIG. 7a estimates the envelope o(n)-Ao(n) is used to correct the signal.
A(n) for n= I, . .. , 4096 of the (frequency domain) For simplicity of the estimator, the determination of
samples transmitted over the fading channel as output 5 the sign need only be carried out for phase differentials
from the deprocessor. The estimate A1(n) is the square- greater than a predetermined threshold. will be in
root of the sum of the squares of the quadrature (Q) and the vicinity of a fade and may be accomplished by seg-
inphase samples output from the de~rocessor which menting the data record into a segment in which the
may be filtered in aceoxlance with known phase differential is larger than a selected threshold and
before Or after estimation of the envelope. The second 10 setting the of the data record to zero.
block performs the operation: computation may be camed out with a simple discrimi-
Aln(~'(t)) (A'(t)) =(A1(n)-A'(n - l))/A'(n), for n=2,
= nation circuit or equivalent computing ms - in the
..., estimator.
where A'(n) is the esthate of A(n). The third Th,bias of the group delay is =t-ted
prforms a
H[Aln(A8(t))I on the result of the m n d block.
'peration by averaging Aot(n) over n for n= 1, ... ,4096 where
A ~ ~is (the rnWured value of ~ ~ ( The estimates
~ ) ~ 1 .
HIA1n(A'(t))l is an estimate of 1 Ar(n)l for n=2, 9
At(.) and are uped directly in the pr&istoflion
4096, where Ao(n) is the phase differential of the trans- fdter in FIG. 5b, while the estimates Ao(n) md So of
signal (Wis the phase of the The the channel group delay and of the bias of'the
transform is P ~ carried~Out taLing the dis- 20~
~ ~ ~group delay rerpectively Y
~ ~ in the de-
Crete fast Fourier transform of the data record, multi-
plying the positive frequency spectrum of the transform modulator.
by -i (square root -I), and the negative frequency The complexity of the processordeprocessorchan-
spectrum of the transform by i, and taking the inverse 1
nel estimator is displayed in Table 1 . Complexity is
discrete fast Fourier transform. The result is a set of 25 measured in Mega Instructions Per Second NIPS)
symbols representing an estimate of the phase differen- where one instruction is defined as one complex addi-
tial of the received signal, as determined from its tion, one complex multiplication and a storage of one
pled amplitude envelope. complex number. It does not include overhead.
Instead of a Hilbert transform, a different estimation The the processor-deprocessOrchan-
may be made to estimate the phase differential. ln this 30 nel estimator in the BS is computed from the complex-
case, firstly, after the electromagnetic signal has been ity Of the Inverse Fast Fourier Transform(IFFT)mmt
sampled, a series of data frames of a number of consecu- Transform (FFT)milbert Transform. The
tive amplitude samples (A(t)) of the electromagnetic complexity is 4096x12X4X21/13.104 ms for the BS-
signal are constructed. These data frames are then seg- For the portable, it is computed from the complexity of
mented into segments [tl,t2],where the amplitude of the 35 the FFTfiFFT per vc: (32X 5 +64+ 128
electromagnetic signal is at least a predetermined num- +256+512+ 1024+2048+4@36)2/13.104 ms for the
ber of dB less than its running mean, for example, IWB. portable with a 6.18 Kbps vocoder. Such a complexity
The following calculation is then applied to these seg- assumes that the A/D converter operates at 100 M H z
ments of the amplitude samples: with 12 bit precision. As seen in Table 11, the portable
40 has smaller complexity due to the fact that the portable
transmits/receives one vc in 13.104 ms and the BS trans-
Ao(t) = l / t o-'
1 + V/&
mits/receives up to 2 1X N vc in 13.104 ms.
Reducing Analog Complexity
where tl=t-tm;,,, tmin is the time in [tl, t2] when A(t)
reaches its minimum, t is the time from the beginning of Comparing FIG. 1 (prior art) and FIG. 5, it will be
the segment, and to is the time from the instant the am- seen that several conventional blocks are not used in the
plitude of the electromagnetic signal reaches its mini- present invention, namely the interleaverdeinterleaver,
mum during the segment until the amplitude reaches the Power Amplifier (PA), both the clock and the car-
double its minimum during the segment. In other rier recovery, both the AGC with its associated Pass-
words, the phase differential may be calculated from band hard limiter, as well as the equalizer.
From the BS point of view, the interleaver-deinter-
A d t ) = -td
before transmission forcing the received samples to be
The polarity of Ao(n) is extracted using the last block independent. From the portable point of view, the inter-
shown in FIG. 7a. The estimate so calculated does not 55 leaver-deinterleaver is not required as a separate entity
provide the sign of the differential.This may be deter- from the vocoder due to the fact that the channel is
mined by known techniques, for example by adding the highly frequency selective, hence the interleavinddein-
phase differential to and subtracting the phase differen- terleaving can be applied implicitly in the vocoder over
tial from the received phase (tan-1 (QA)) and taking one vc, without a need for a separate time domain inter-
the sign to be positive if the addition results in the 60 leaver/deinterleaver. This eliminates excess speech de-
smaller Euclidean distance to the expected value and lays associated with interleaving/deinterleaving be-
negative if the subtraction results in the smaller Euclid- tween frames.
ean distance to the expected value. The PA is not required since the cells can have, as
Equivalently, for each sample n, the ideal phase clos- shown later, a radius of up to at least 250 m outdoors
est to w(n)+Ao(n) is determined and labelled w+(n), 65 and 30 m indoors, if the transmitted power is up to 6
and the ideal phase closest to o(n)-Ao(n) is deter- dBm. Such a power can be generated by the Local
mined and labelled o-(n). The two sums P = Oscillator (LO) without a need for a PA. It is important
2 Io+(n)-Co(n)+Ao(n)) I and to avoid using a PA since DOFDM generates a time
domain signal with non constant envelope. A power the BS over one vc slot is (6 dBm - 10logl$UdB) while
efficient class C PA cannot be used without distorting the signal power transmitted by the portable over one
the signal. A c h A PA can be used at the expense of vc slot is 0 dBm. Also, since the noise power over a 100
power efficiency. MHz band is -94 dBm, it is (-94 dBm - lO1ogloN dB)
A clock recovery device is not required since a sam- 5 over one vc. A typical noise figure at the receiver is 7
pling error in the time domain is equivalent to a phase dB. The penalty for not using a matched filter in the
shift in the frequency domain. The phase shift is a linear receiver is 1 dB. Combining together the above figures
function of frequency. It contributes to the bias in the provides the portable with an (92 dB - path loss in dB)
channel group delay. Such a bias can be easily estimated received signal to noise ratio (SNR), while it provides
+
and removed as mentioned previously by averaging 10 the BS with an (86 dB lOlogl$U dB - path loss in
of(n) over n in the frequency domain. Such an estimate dB) received SNR.
is accurate as long as the sampling error is less than 0.2 For a path loss of 75 dB, the radius of the urban cell
ps or equivalently less than 20 samples (since in this can be 250 m while it can be 30 m for the indoor cell.
case, the correspondingphase shift is less than n),and as Such a path loss provides the portable with a 17 dB
long as the number of points in one vc is large enough 15 received SNR, while it provides the BS with an (11 dB
as it is here. + lOlogloN dB) received SNR. From the portable
A carrier recovery device is not required since a point of view, the channel can be modeled approxi-
d e r offset in the time domain is equivalent to a sam- mately as an ideal AWGN channel, hence the 17 dB
pling error in the frequency domain. For the chosen RC received SNR results in a 2X 10-3 BER. On the other
pulse, a sampling error of up to 10% of the duration of 20 hand, the channel can be pessimistically modeled as a
one pulse is acceptable. Rayleigh fading channel from the BS point of view. The
This implies that a frequency offset of up to 2.414 corresponding BER are displayed in Table I11 which
KHz is acceptable regardless whether it is due to carrier shows that the achieved BER is 54x10-3. A BER
offset as low as 1part in a million, i.e. as low as 1 KHz S 10-2 is acceptable for voice.
per 1 GHz. When a carrier frequency higher than 2.414 25
GHz is required, one can decrease in FIG. 2 the number Cell Pattern Reuse
of points per 100 MHz or one can use an RC pulse with From Table I, the number of Full Duplex voice chan-
a rolloff larger than 20%. nels (FDvc) that can be transmitted/rcceived per frame
Neither an AGC nor a Passband hard-limiter are is 136 over 100 MHz, for a 6.18 Kbps vocoder. If the
required since the level of the received power may be 3 bandwidth is halved to 50 MHz, the number of FDvc
0
controlled constantly. This is achieved as follows: The per frame is reduced to 68, the noise floor is reduced by
portable transmits a frame. The BS receives the frame 3 dB and the number of full duplex frames that a BS can
and predistorts a frame intended for transmission ac- transmit/receive is doubled to 42, leaving the frame
cordingly, assuming that the channel is reciprocal and duration, the number of frames per 13.104 ms and the
stationary over 520 ps. This includes controlling the 35 processor/deprocessor complexity unchanged.
transmitted power according to the received power. Reducing the available bandwidth directly affects the
The BS transmits the predistorted frame and simulta- cell pattern reuse. This can be explained as follows,
neously orders the portable to control its power. The assuming that we are required to offer a minimum of
order is conveyed using the control symbol in the vc 136 FDvc per cell, that the vocoder rate is 6.18 Kbps
slot (See table I). The degree of power control may be 40 and that the cell radius is fixed at 250 m outdoors and 30
determined using the power controller 525, and the m indoors. For a 100 MHz band, we assign one frame
instruction for the inclusion of a power control symbol per cell and offer 136 FDvc per cell. In this csse,the
in the vc may be sent from the power controller 525 to cell pattern reuse consists of 126 cells as shown in FIG.
the predistortcr 534. 80 which displays a seven layer structure. For a 50 MHz
One advantage of wideband modulation over nar- 45 band, we assign two frames per cell and offer 136 FDvc
rowband modulation is that the wideband signal does per cell, hence reducing our cell pattern reuse to a 63
not experience short term fading the same way the cell pattern as shown in FIG. 8b which displays a five
narrowband one does. The wideband signal is mainly layer structure. If the available bandwidth is as low as
affected by shadowing and other long term effects 5.86 MHz, we have 8 vc per frame. Hence we have to
which vary slowly and arc easily monitored from one M assign 18 frames per cell in order to offer the minimum
frame to the other as long as the same vc slot is used by required number of FDvc per cell. This reduces the a l l
the portable to transmit and receive (i.e. as long as TDD pattern reuse to as low as a 7 cell pattern as shown in
is employed). FIG. 8c which displays a two Iayer structure.
Finally, conventional equalization, whether it is lin- In FIGS. 8a b and c, a shaded area is shown around
ear or nonlinear, is not required simply because there is 55 the center of the pattern, indicating 19, 38 and 126 N 1
little or no ISI. Also, from the portable point of view, duplex frames that the central BS can transmit/receive
each received vc is predistorted by the BS. Hence, the respectively. Tables IVa, b and c show the number of
channel can be modeled approximately as an ideal mem- cell layers in each cell pattern reuse, the coverage area
oryless Additive White Noise Gaussian (AWGN) chan- in Km2 of the pattern reuse for both the indoor and the
nel, assuming channel reciprocity and stationarity over 60 urban environments, as well as the carrier to interfer-
520 p.From the BS point of view, since the received ence ratio (CIR) in dB, for the 100 MHz, M MHz and
signal is not predistorted by the portable prior transmis- 5.96 MHz bands, respectively. In all cases, the CIR is
sion, the channel estimator is used to reduce the effect large enough to sustain a toll quality speech.
of the channel group delay.
65 Transmission/Reception Protocol
Smaller cells Since the number of FDvc a portable can transmit/-
As mentioned previously, the LO generates a 6 dBm receive is one, while the number of FDvc a BS can
average power, hence the signal power transmitted by transmit/receive is much larger as shown in Table V for
each of the three vocoder rates, we have chosen the adjacent cell. It can use the original channel as long as
following transmission/reception protocol: the level of CIR is acceptable. If on the other hand, a
1. The portable transmits a frame over a vc. portable wants to initiate a call in cell Y where d l preas-
2. Seven adjacent BS receive the frame from the porta- signed channels are used, BS Y can borrow a channel
ble. 5 from an adjacent cell up to a limit of 64 channels per
3. One BS transmits to the portable, depending for ex- cell.
ample on the strength of the received signal by each The main advantage of DCA over Fixed Channel
of the BS. Allocation (FCA) is the increase in traffic handling
The control of this protocol may use any of several capability For FCA, a 7 cell pattern each with a preas-
known techniques. For example, the commonly used 10 signed 144 Fdvc can carry a total traffic of 880.81 Er-
technique is to have the portable monitor the channel lang at 0.01 Blocking Probability (BP). For DCA, a 7
and determine which of several base stations it is closest cell pattern consists of 6 cells each with 80 FDvc that
to. It can then order the nearest BS to communicate can carry a total traffic of 392.17 Erlang, combined with
with it. Another technique is to use a master control one cell with 528 FDvc that can carry 501.74 Erlang.
which receives information about the strength of the 15 The total traffic is therefore 893.91 Erlang. This in-
signal on the channel used by the portable and controls crease appears to be marginal (1.5%). However, if
the BS accordingly. Such techniques in themselves are 501.74 Erlang are actually offered to one celI in the
known and do not form part of the invention. FCA system (with 14 FDvc/cell), while the six other
Such aprotocol has several advantages. For instance, cells carry 392.17/6=65.36 Erlang per cell, the BP at
the location of the portable can be determined with high 20 that busy cell 0.714 while it is negligible at the six other
accuracy based on the received vc at the seven adjacent cells. The total blocked traffic (i.e. lost traffrc) in the
BS. Locating the portable can assist in the BS hand-off. FCA system is then equal to
A BS hand-off and a portable hand-off do not necessar- 1
(6 X 65.36~0.0+ X0.714X 501.24) 358.24 Erlang. This
ily occur simultaneously, contrary to other prior art represents a 0.4 average BP. If the DCA is allowed such
systems. In the present invention, when a portable 25 a loss, its traffic handling capacity would increase to
roams from one cell X to an adjacent cell Y, a new vc 1768.04 Erlang which represents a 100% increase in
is not required immediately. What is required is a BS traffic handling capacity over the FCA system, or
hand-off, meaning that BS Y (associated with cell Y) equivalently a 160% increase in the number of available
must initiate transmission to the portable over the same FDvc. The DCA system thus represents a marked im-
vc, while the BS X (associated with cell X) must termi- 30 provement over the FCA system.
nate its transmission to the portable.
A BS hand-off occurs without the knowledge of the Voice Activation
portable and can occur several times before a portable Voice activation is controlled by the BS according to
hand-off is required. A portable hand-off is required techniques known in the art. At any instant during a
only when the CIR is below a certain level. In this case, 35 conversation between a BS and a portable, there are
the Mobile Telephone Switching Office (not shown) four possibilities:
calls for a portable hand-off in accordance with known 1. BS talks while the portable listens.
procedures. Reducing the portable hand-off rate re- 2. BS listens while the portable talks.
duces the probability of dropped calls. This is because a 3. BS and portable talk simultaneously.
dropped call occurs either because the portable hand- 40 4. BS and portable listen simultaneously.
off is not successful or because there are no available The BS controls the voice activation procedure by
channels in cell Y. allocating in cases 1, 3 and 4 three slots (frames 1.1, 1.2
The present invention allows the use of post-detec- and 1.3) to the BS and one slot the portable (frame 1)
tion diversity at the BS, and the use of dynamic channel every four slots as shown in FIG. 9a. Likewise up to 21
allocation (DCA). 45 portables may communicate with the base station in like
fashion.
Dynamic Channel Allocation in case 2, on receiving a signal from the portable, the
DCA is made possible by each BS having capability BS allocates three slots (frames 1.1, 1.2 and 1.3) to the
to transmit/receive more than the number of FDvc portable and one slot (frame 1) to the BS every four
allocated to its cell, namely seven times the number of 50 slots as shown in FIG. 9b. Likewise, up to 21 other
FDvc for a 5.86 MHz band and up to twenty-one times portables may communicate with the base station in like
the number of FDvc for a 100 MHz as well as a 50 MHz fashion. Consequently, instead of transmitting two full
band. The DCA protocol simply consists of borrowing duplex voice frames over four slots as in FIG. 4, voice
as many FDvc as needed from the adjacent cells, up to activation allows us to transmit three full duplex voice
a certain limit. The limit for the case when we employ 55 frames over four slots. Hence, voice activation provides
a 6.18 Kbps vocoder, a 5.86 MHz band and 18 frames a 50% increase in the number of available FDvc at the
per cell is obtained as follows. The cell reuse pattern expense of increasing DSP complexity.
consists of 7 cells. Each cell is preassigned 144 FDvc.
Assuming that at peak hours, 75 FDvc are used on the Capacity
average and 5 FDvc are reserved at all times, then we 60 The capacity of Code Division Multiple Access
are left with 64 idle channels which represent the limit (CDMA) may be defmed as the number of half duplex
on the number of FDvc one can borrow from the cell. voice channels (HDvc) effectively available over a 1.25
One should distinguish between the limit on the chan- MHz band per cell. Based on such a definition, Table 1V
nels borrowed and the limit on the nonpreassigned displays the capacity of analog FM and of the present
channels a BS can use. For instance, if a call originates 65 system with a 6.18 Kbps vocoder, 5.86 MHz band, 1
in cell X and the portable roams into an adjacent cell Y frame per cell and DCA. As shown in Table IV, the
where no preassigned cells are available, BS Y does not capacity of analog FM is 6 HDvcA.25 MHz/cell while
need to borrow immediately a new channel from an for the present system it is 150 HDvd1.25 MHz/cell.
The 6.25 MHz band wnsists of 5.86 MHz plus two ti al sampler 826 to produce bits of information. The bits are
slots. When voice activation is used, the capacity of the then processed in the deprocessor 828, described in
present system is increased by 1.5 times to 225 more detail in FIG. 14b.An estimate of the phase differ-
HDvd1.25 MHz/cell, a 38 fold increase over analog ential is taken in the channel estimator 830, as described
FM. 5 in more detail in relation to FIG. 7 above, and the esti-
mated phase differential is supplied to a decoder/-
L c l Area Networks
oa demodulator 832 to correct the received bits. The esti-
The invention may also be applied to produce a 48 mated phase differential is also supplied to a predis-
Mbps wireless LAN, which also satisfies the technical torter 834 in the transmitter. At the transmitter in the
requirements for spread spectrum. 10 Base Station, the same blocks are incorporated as in the
For wireless LAN, wideband differential orthogonal portable transmitter except that a pre-distorter is used to
frequency division multiplexing is again employed. The alter the envelope and phase of the D8PSK symbols to
LAN will incorporate a plurality of transceivers, all make the channel appear Gaussian (ideal) as opposed to
more or less equal in terms of processing complexity, a fading channel. The initial power control 825 also
and possibly with identical components except for ad- 15 sends a signal to the pre-distorter 834 to adjust the trans-
dresses. mitted power to an appropriate signal level for the
To implement wideband modulation for a LAN,a 26 sampler 826 in the first transceiver. It will be apprcci-
MHz band is divided into 128 points, as shown in FIG. ated that a predistorter will be included in the first
10, plus two tail slots of 1.48 MHz each within the 26 transceiver's transmitter but that it will not be operable,
MHz band. Adjacent points are separated by 180 KHz 20 except when the first transceiver is operating as a base
and each point, as with the application described above station.
for a portable-base station, represents a D8PSK symbol. FIG. 13c shows the functional blocks of the receiver
The transmitter components will be the same as shown of the first transceiver, which is the same as the receiver
in FIG. Sb, with suitable modifications as described in in the second transceiver except it does not include an
the following, and will include an encoder. The output 25 estimator. The processor is illustrated in FIG. 14u and
bits from the encoder are mapped onto the D8PSK 14c and the deprocessor in FIG. l4b and 14c. The pro-
symbols. cessor frst inverse Fourier transforms the 128 D8PSK
The frame duration for the symbols is illustrated in symbols output from the modulator. The transformed
FIG. 11. A rectangular time domain window corre- symbols are then triplicated as a group so that the total
sponding to a RC frquency domain pulse has a 5.55 ps 30 number of samples is tripled (see the left side of FIG.
duration, and includes a 25% roll-off and excess frame 44, with three consecutive groups each consisting of
duration of 0.26 ps, making a total 7.2 ps duration for the 128transformed symbols. Next, the three groups are
the frame. windowed by a Raised Cosine window with a roll-off of
For such a wireless local area network (LAN), in 0.25 centered in the middle of the three groups. In other
which the transceivers are equal, the T i e Division 35 words, the processor takes D8PSK symbols in, pulse
Duplex protocol is as illustrated in FIG. 12 (assuming shapes them and inverse Fourier transforms them. On
there are at least a pair of transceivers): the other hand, the deprocessor undoes what the pro-
1. A first transceiver transmits a signal (frame 0) over cessor did, i.e. it removes the pulse shaping, then Fou-
the entire frame. rier transforms the received signal to obtain the original
2. A second transceiver receives the signal from the first 40 DBPSK symbols. The first two blocks in FIG. 14b are
transceiver and processes (analyzes) it. similar to the second two blocks in FIG. 14u except for
3. Based on the received signal, the second transceiver two differences as follows. In the first block shown in
predistorts and transmits nine frames (frames 1-9) to FIG. 146,the repeated groups of symbols are partially
the first transceiver immediately. overlapped, as shown in FIG. 14c.In the second block,
Each transceiver has transmitter components similar 45 a rectangular window is used instead of the Raised
to those illustrated in FIG. 56, with suitable rnodifica- Cosine to produce 128 output samples corresponding to
tions to the internal structure to allow the use of the the 416 input samples.
particular frequency band and frame duration em- The phase estimator is the same as that shown in FIG.
ployed. 7, except that there are only 128 input samples, and the
The transmitter/receiver functional and structural 50 same description applies.
block diagrams are shown in FIGS. 130,13band 13c for For both the LAN and cellular networks, the present
the exchange of data. Data is provided to an encoder system is designed to operate as a spread spectrum sys-
810 where the data is digitized and coded to create bits tem preferably over such bands as are permitted, which
of infonnation. The bits are provided to the modulator at present are the 902-928 MHz band, 2.4-2.4835
812 which turns them into D8PSK symbols, with three 55 GHz and 5.725-5.85 MHz. The camer frequency in
bits per symbol. The D8PSK symbols are then pro- the local oscillator shown in FIGS. Sa, b and c w l then
il
cesscd in the processor 814 which is described in more be 915 MHz in the case of the 902-928 MHz band, and
detail in FIG. 14u. The output from the processor is the frequencies used for modulation will be centered on
then filtered in low pass filter 816,upconverted to RF this carrier frequency.
frequencies using loch oscillator 818 &d transmitted by 60
antenna 830. Alternative Embodiments
In FIG. 136, the received signal at the base station is A person skilled in the art could make immaterial
filtered in a bandpass filter 822,and down converted by modifications to the invention described and claimed in
mixing with the output of a local oscillator 824. The this patent without departing from the essence of the
average power of the downcoverted signal is monitored 65 invention.
by an initial power control 825 that adjusts the average For example, a system may consist of one or more
power to the specificationsrequired by the sampler 826. central controllers (comparable to the Base Stations in
The adjusted downwnverted signal is then sampled in the exemplary cellular system described) and some
5,282,222
19 20
slave units (comparable to the portables). The slave unit 5. The transceiver of claim 4 in which the power
executes the commands it receives from the central controller is also connected to the pre-distorter for
controller. The commands may be requesting the slave controlling the power of the signal to be transmitted.
unit to transmit a receive acknowledge, a status code or 6. The transceiver of claim 1 further including: means
information that the slave has as - to. The command 5 to modify the received signal with one or both of the
may also be to relay the command or the infomation to estimated amplitude and phase differential respectively.
another slave unit. 7. A method for allowing a number of wireless trans-
We claim: ceiver to exchange frames of information, the method
1 A transceiver including a transmitter for transmit-
. comprising the steps
ting electromagneticsignals and a receiver for receiving 10 multiplexing a first frame of information over a num-
electromagnetic signals having amplitude and phase ber of frequencies within a frequency band at 8 first
transceiver to produce multiplexed information;
differential characteristics, the transmitter comprising: processing the multiplexed information at the fmt
an encoder for encoding information; transceiver,
a wideband frequency division multiplexer or multi- transmitting the processed information to a second
plexing the information onto wideband frequency transceiver using a carrier frequency fc;
channels; receiving the processed information at the second
a low pass filter; transceiver; and
a local oscillator for upconverting the multiplexed processing the information at the second
information for transmission; 20 transceiver during a first time interval;
a Processor for applying a fourier tranSf0x-m to the in which the frequency band is formed from a first set
multiplexed information to bring the information of K1 points and a pair of tall slots each having K2
into the time domain for transmission; points, each of the points being separated by a
further including, in the receiver of the transceiver; frequency range of Af, the second transceiver has a
a bandpass filter for filtering the received electromag- 25 maximum expected clock error xT, where T is the
netic signals; duration of one time domain sample, the informa-
a local oscillator for downconverting the received tion is multiplexed over a number M of levels, and
electromagnetic signals to produce output; K1 selected such that 2 ~ x / K l < ? r / M ,whereby
a sampler for sampling the output of the local oscilla- the width of the frequency band is chosen so that
tor to produce sampled signals to the channel esti- 30 neither carrier nor clock recovery is required at the
mator; second transceiver.
a channel estimator for estimating one or both of the 8. The method claim 7 further including transmitting
a second frame of information from the second trans-
amplitude
=
the phase differentialof the
signals to produce output one or both of an
estimated amplitude and an estimated phase differ- 35
ceiver to the first transceiver within the same frequency
ential respectively; and 9. The method of claim 7 in which K2 is selected s o
a decoder for producing signals from the sampled that the out of band signal is less than a given level.
10. The method of claim 7 in which the fmt and
signals and the output from the channel estimator.
second transceivers have an expected maximum relative
2. The transceiver of claim 1further including, in the velocity V, the first and second transceivers have car-
receiver of the transceiver: rier frequencies with a frequency offset from each other
a de~rocessOrfor an inverse Fourier trans-' of of' the c a d e r frequency has a corresponding travel-
form to the samples output from the sampler. ling wavelength h and Af is selected so that
3. The transceiver of claim 2 further including, in the [V/(hAf) + of,^^ is less than or equal to a preselect+d
receiver of the transceiver: 45 sampling error.
a power controller before the sampler for monitoring 11. ~h~ method of claim 7 in the
the power of the received signal and for control- multiplexed information at the second transceiver fur-
ling the power of the signal. ther includes calculating the mean of the phase shift due
4. The transceiver of claim 3 further including, in the to sampling error by summing an estimated phase differ-
transmitter of the transceiver: 50 ential of the received signal.
a predistorter before the processor, the pre-distorter 12. The method of claim 1 in which the mean of the
1
being connected to the channel estimator, for pre- phase shift due to sampling error is divided by K1 and
distorting a signal to be transmitted with one or the result removed from the phase differential of the
both of the estimated amplitude or the estimated received signal.
phase differential. 55 + L + + +