Note: Descriptions are shown in the official language in which they were submitted.
CA 02248647 1998-09-09
WO 97/36383 PCTIUS97/04280
"A Rapid-Acquisition Access Channel Scheme for CDMA systems"
Background and Brief Description of Prior Art
Background and Brief Description of Prior Art
Spread spectrum communications is presently being used
for a number of commercial applications and is expected to
proliferate as the demand for untethered communications
increases. One example of this is the IS-95 standard for
cellular telephones. This system uses orthogonal CDMA (OCDMA)
on the outbound links and nonsynchronous CDMA on the inbound
links.
Another example of commercial application of spread
spectrum techniques is the orthogonal CDMA (OCDM) system
discussed for an office PBX system by Magill, et al, in
"Spread-Spectrum Technology for Commercial Applications",
Proc. of the IEEE, June 1994. In this case, the base station
of this star-configured network transmits a set of orthogonal
Walsh functions which are overlaid with a pseudo-noise (PN)
sequence. Each orthogonal function carries voice or data for
a single user. See M. J. E. Golay, IDA Report 108, pg. 110
(1965) which discloses this basic signal format.
The discussion by Magill, et al, is for a short range
system in which it makes sense to provide TDMA time slots on
the return link for network members to transmit a signal for
timing and synchronization purposes. In this manner, an empty
CA 02248647 1998-09-09
WO 97/36383 PCTIUS97/04280
slot is reserved for a member to enter the net at any time.
This technique becomes inefficient and is not useful when the
path lengths are long and the propagation time varies
considerably between users, such as in a satellite system.
A number of consortiums have been formed to develop
satellite based Personal Communications Systems (PCS) with
global coverage. Some examples of these systems include
Globalstar (Globalstar System Application before the FCC by
Loral Cellular Systems, Corp., June 3, 1991) and Odyssey
(Application of TRW Inc. before the FCC to Construct a New
Communications Satellite System "Odyssey," May 31, 1991),
among others. The intent of these systems is that a subscriber
can place telephone calls directly through the satellite
network from almost anywhere on the Earth, using a portable
handset much like the present cellular telephones. Both of
the systems mentioned intend to use spread spectrum CDMA
techniques for a number of reasons.
The return link CDMA signal is generally difficult to
acquire, especially. in satellite systems, due to the
relatively large time and frequency uncertainty of the
received signal and the low received signal power. The
acquisition process may be aided by transmitting an auxiliary
signal in a designated channel. This signal, in conjunction
with a suitable receiver, should be designed so the receiver
can acquire the signal quickly, estimate carrier frequency and
2
CA 02248647 2004-11-03
time-of-arrival accurately, and demodulate data. One means for performing
these functions is
described in Canadian patent application, "Network Entry Channel for CDMA
systems, "No.
2,206,796, Francis D. Natali.
Summary of Invention
1 The present invention relates to a unique signaling and receiver combination
that
makes maximum use of matched filter and noncoherent reception techniques to
achieve very
rapid acquisition and robust performance in the presence of multipath. This
signaling and
receiver combination is called the Return Access Channel (RAC). It is
applicable to both
satellite and terrestrial communication systems.
One aspect of the present invention resides in a spread spectrum CDMA
communication system in which a first station, such as a subscriber station,
transmits a short
synchronization signal to allow a second station, such as a base station, to
(1) recognize that
a signal is present, (2) resolve frequency uncertainty, (3) acquire the
spreading code timing,
(4) acquire frame and symbol synchronization, and (5) demodulate data which
may contain
user identification and other data. The system includes means for generating
and transmitting
a burst signal comprising an Acquisition Field during which the carrier is
biphase modulated
with a pseudo-noise (PN) sequence, a Sync Field during which a synchronization
sequence is
mod-2 added to the PN sequence with one chip of the sync sequence equal to one
or more
periods of the PN sequence, and a data field with m-ary orthogonal code words
mod-2 added
to the PN sequence with one chip of the code word equal to one or more periods
of the PN
sequence. The second station has a receiver which includes a bank of filters
matched to the
PN sequence and covering the frequency uncertainty range of the received
signal. Output
samples over the PN code length are noncoherently combined with samples of the
same code
time phase from previous intervals to form a Signal Detection Table (SDT) of
signal power
versus code phase for each filter in the filter bank. The contents of the SDT
are used for
signal detection and time and frequency bin selection and time offset and
coarse frequency
offset estimates are also based on the contents of the SDT. The receiver also
includes means
to demodulate the m-ary data either noncoherently or coherently.
3
CA 02248647 2004-11-03
The invention may also reside, therefore, in an orthogonal CDMA (OCDMA)
communication system including a base station receiver means for receiving a
plurality of
spread spectrum signals from a corresponding plurality of subscriber stations
on a selected
frequency channel, each signal from a subscriber station being composed of
data symbols
overlaid with one set of orthogonal functions and a PN sequence, the signals
from subscriber
stations being sysichronized to arrive at the base station in time and
frequency synchronism;
and a subscriber station Return Access Channel (RAC) comprised of a
transmitter means at
the subscriber stations for transmitting a spread spectrum RAC burst in an
assigned frequency
band.
The invention is intended to serve several functions and objectives. These are
summarized as:
Provide a system for rapidly acquiring the RAC with relatively large
uncertainty in
carrier frequency and time-of-arrival.
Perform robust signal acquisition with relatively low received signal levels.
Provide the user with a high link margin, low data rate channel for net entry
requests.
Provide a system in which the receiver accurately estimates the received
signal time-
of-arrival and carrier frequency.
3a
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
Provide means for accessing an OCDMA network on a
noninterfering basis without prior time and frequency
synchronism.
Allows the hub station to detect and synchronize the user
before assigning an orthogonal function for OCDMA operation.
The invention has the following novel features
1. The RAC waveform and receiver are designed and
configured to take maximum advantage of matched filter and
noncoherent detection techniques. The unique combination of
these techniques results in rapid acquisition, efficient time
ambiguity resolution and symbol synchronization, as well as
efficient data detection even in the presence of multipath.
It further results in a very efficient receiver hardware
implementation.
2. The RAC receiver uses a single shift register with
multiple weighting networks to synthesize a bank of filters
matched to the PN sequence and with center frequencies that
span the initial carrier frequency uncertainty. The filter
outputs are combined noncoherently to increase the probability
of detection without narrowing the matched filter input
bandwidth. The summed outputs for each frequency bin are
stored for all sample times in a code period, The resultant
two dimensional table is scanned for time/frequency bins which
contain an entry large enough to indicate signal presence.
This results in very fast and reliable signal detection.
4
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
3. The above configuration allows multiple signals to be
processed simultaneously, as long as they are separated in
time-of-arrival by more than one PN code chip interval.
4. Frame and symbol sync are achieved by further
noncoherent matched filter processing of the signal
correlation peaks from the input MF. This is a process that
is both rapid and robust in the presence of multipath.
5. Transmission of the data by 8-ary orthogonal code
words allows noncoherent data detection, which is robust in
multipath, while maintaining good power efficiency.
Description of the Drawings
The above and other objects, advantages and features of
the invention will become more clear when considered with
the following specification and accompanying drawings,
wherein:
Figure 1 is a typical satellite radio communication
system incorporating the invention;
Figure 2 illustrates the return access channel (RAC)
burst signal structure;
Figure 3 is a block diagram of a linear shift register
generator for RAC signal length 255 PN sequence;
Figure 4 is a table of RAC signal parameter summary;
Figure 5 is a block diagram of the RAC receiver with
parallel matched filter time and frequency processing,
5
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
Figure 6 is a functional block diagram of the matched
filter (MF) algorithm,
Figure 7 is a graph illustrating the receiver DFT
frequencing bins.
Figure 8 is a partial plot of the signal detection
table for C/No = 36 dB-Hz, and sum of 64 power samples for
each time bin, the signal being centered in frequency bin #3
and time bin #3,
Figure 9 illustrates the matched filter (MF) output for
frequency bin #3 and time bins 1 through 200 with C/No 36
dB-Hz, and sum of 64 power samples for each time bin,
Figure 10 illustrates the real part of the couplex MF
output with a frequency offset 0.1 of the PN code
repetition rate,
Figure 11 is a graph illustrating the real part of the
baseband signal with sync modulation without noise after AFC
frequency correction,
Figure 12 is a graph illustrating sync output power
after noncoherent post detection combining of 4 samples,
Figure 13 is a block diagram of the circuit for
frequency correction of the decimated MF output before data
demodulation.
Figure 14 is a curve of the quadrature fit frequency
discriminator.
6
CA 02248647 1998-09-09
WO 97/36383 PCTIUS97/04280
Figure 15 is a plot of the AFC loop pull-in and
tracking performance with C/No 36 dB-Hz, BL=32 Hz, dfz-107
Hz, and
Figure 16 is a functional block diagram of an RAC
transmitter incorporating the invention.
Detailed Description of the Invention
The embodiment as discussed here relates to the return
link of a star configured spread spectrum satellite network
of the type shown in Figure 1, but it can also be applied to
appropriate terrestrial systems. The satellite receives the
user signal from the ground and transponds it to a hub
ground station (GS). The return link signal structure
described below incorporates CDMA with a separate Return
Access Channel (RAC).
The Subscriber Terminal (ST) transmits an Return Access
Channel (RAC) signal to alert the Earth Station (ES) that he
wishes to place or receive a call. This signal is acquired
by the ES which makes an estimate of the ST time and
frequency offsets as well as demodulating RAC data which
includes a temporary Identification (ID) word as well as a
priority designation for 911 use. In the case of an
Orthogonal CDMA (OCDMA) system, appropriate time and
frequency corrections are computed at the ES and transmitted
back to the ST on the Forward Link (F/L) control channel,
7
CA 02248647 1998-09-09
WO 97/36383 PCTIUS97/04280
along with a traffic channel assignment. The ST can then
switch to the Return Link (R/L) traffic channel and begin
call set up. The description given assumes parameters
values which were chosen for illustrative purposes.
RAC Signal Description
The RAC signal is a spread spectrum burst signal which
consists of three parts as shown in Figure 2, the
Acquisition Field, the Frame and Symbol Sync Field, and the
Data Field.
The Acquisition Field
The Acquisition Field is a length 255 PN m-Sequence
which bi-phase modulates the carrier. The chipping rate is
272 kcps and the code is repeated 192 times, corresponding
to 180 ms. The signal parameters are summarized in the
Table shown in Figure 3. The PN sequence is characterized
by the primitive polynomial p(x) = X8 + X4 + X3 + X2 + 1 and
is generated by the linear shift register generator of
Figure 3.
Frame and Symbol Sync Field
The Frame and Symbol Sync Field is distinguished from
the Acquisition Field in that a length 8 Neuman-Hofman
synchronization word is mod(2) added to the PN spreading
sequence before bi-phase modulation of the signal. One
symbol of the sync word is exactly a PN sequence long. The
8
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
spreading sequence and sync word transitions are
synchronized to occur at the same time. The Sync Field is
30 ms long, which corresponds to 32 repetitions of the PN
code and 4 repetitions of the sync word. The sync sequence
is 00011101. These parameters are summarized in Figure 4.
Data Field
The Data Field modulation is bi-phase using 8-ary orthogonal
waveforms which are spread with the same spreading sequence
as described above. A data symbol is exactly 8 PN code
periods long, which gives a symbol rate of 133.33 sps. Each
chip of the orthogonal code word is exactly one PN sequence
long. The orthogonal code word set is shown in the Table
below. The Data Field is 160 ms long, which corresponds to
data symbols or 60 data bits. These parameters are
15 summarized in the Table shown in Figure 4.
TABLE
1 1 1 1 1 1 1 1 1
2 1 -1 1 1 -1 1 -1 -1
3 1 1 1. -1 1 -1 -1 -1
4 1 1 -1 1 -1 -1 -1 1
5 1 -1 1 -1 -1 -1 1 1
~, 1 1 -1 -1 -1 1 1 -1
7 1 -1 -'i -1 1 1 -1 1
g 1 -1 -1 1 1 -1 1 -1
9
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
Receiver Architecture
The RAC receiver described here, and shown in Figure 5,
is conceptually a bank of 16 parallel matched filters
(matched to the PN sequence) which span the initial
'5 frequency uncertainty of 3 kHz. The matched filter (MF)
outputs are noncoherently combined by a sliding accumulator
(64 samples long) for each time bin, resulting in a two-
dimensional array of power measurements which covers 16
frequencies and 1020 time offsets (corresponding to 4
samples per PN chip). Signal detection is performed based
on the data in this table (called the Signal Detection
Table). Detection will occur within about 70 ms of signal
reception, Note that multiple signals can be processed in
parallel.
When a signal is detected, the sample time/frequency
bin corresponding to a correlation peak is selected arid the
complex samples are routed to the demodulator. A separate
demodulator path is required for each signal to be
processed,
An estimate of time offset is performed based on the
data of the Signal Detection Table. The ambiguity
resolution capability of the PN code is approximately + 0.5
ms. An increased ambiguity resolution capability of 4 ms
is obtained With the symbol sync operation as is described
later herein.
CA 02248647 1998-09-09
WO 97/36383 PCTIUS97/04280
A coarse estimate of frequency offset is also performed
on the detected signal using a quadratic polynomial fit to
the power in three adjacent frequency bins of the signal
detection table. This frequency estimate is used to
reduce the baseband frequency offset of the received signal.
A cross product AFC loop is then enabled to perform a
"fine" frequency estimate and further reduce the frequency
offset. This loop has a bandwidth of about 30 HZ and
settles in less than 50 ms. The combined results of the
"coarse" and "fine" frequency estimates are processed to
derive an accurate frequency estimate of frequency offset.
The frequency corrected samples corresponding to the MF
correlation peaks for a particular signal are routed to the
frame and symbol sync detector. This operation is performed
by an 8-bit matched filter followed by noncoherent combining
(4 samples).
Once frame and symbol sync are established, data
demodulation can begin. This is accomplished by processing
the received signal samples in a band of abit filters
matched to the orthogonal code word set. Each of these
functions is examined in more detail below.
Signal Detection
The received signal is applied to the equivalent of a
bank of 16 matched filters, each of length 1020
(corresponding to 4 samples per chip and 255 chips), as
11
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
shown in Figure 6. These filters are implemented with a
single shift register whose outputs are weighted by the PN
sequence as well as the sinusoid values corresponding to the
desired matched filter center frequency.
The filters are spaced at 533.3 Hz (one-half the PN
code repetition rate) and overlap as shown in Figure 7. The
magnitude squared of each filter output is accumulated over
64 code periods for each time bin offset, This noncoherent
combining serves a dual purpose. First, it limits the
length of the digital MF while still providing good
detection performance at low signal levels. Secondly, the
frequency bin bandwidth would be greatly reduced, thus
requiring many more bins if the filter were lengthened in a
coherent fashion. The MF is updated at a
1.088 mHz rate (4 times per chip).
The summed MF output powers are used to construct a
two-dimensional Signal Detection Table (SDT) with a total of
16,320 time/frequency bins corresponding to 16 frequencies
and 1020 time offsets. A partial plot of the Signal
Detection Table (SDI-) derived by computer simulation is
shown in Figure 8. Note that the correlation peak occurs is
frequency bin #3, and time bin #3. A cross section of this
plot for frequency bin #3 is shown in Figure B.
All of the entries in the SDT are updated each PN code
period (approximately 1 ms). The contents of the table are
12
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
examined for the largest entry that has not been designated
as a signal within the previous 360 ms. Signal detection is
declared if this entry exceeds the detection threshold. The
frequency bin number of a bin where detection has been
declared corresponds to one of the MF outputs shown in
Figure 6, while the ti4ne bin corresponds to samples
associated with a particular reference clock timing. 'Thus,
once a signal is detected, the associated signal samples can
be routed to the data demodulator as shown in Figure 4.
This results in an effective decimation by a factor of 1020
in the number of samples to be processed. The MF output
before decimation is shown in Figure 10. The uncorrected
frequency offset results in a sinusoid at baseband which
modulates the correlation peaks. Note that correlation peaks
occur every 1020 samples.
Frequency Estimate
The RAC receiver performs three frequency estimates is
order to reduce the frequency offset sufficiently to allow
data demodulation with little degradation as well as to
facilitate traffic channel entry. The estimates are
performed sequentially by:
1. The signal detection frequency bin number
indicates frequency offset with an accuracy of
approximately 1 kHz accuracy.
13
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
2. A "coarse" frequency estimate is obtained by using
a quadratic fit to entries in the SDT as described
below. This gives an accuracy of about 30 Hz,
3. "Fine" frequency tracking is enabled with a cross
product AFC loop.
The frequency estimates are used to remove the
frequency error of the baseband signal before routing to the
data demodulator as shown in Figure 13.
Coarse Frequency Estimate
The coarse estimate is performed as follows:
= 7he coefficients of the quadratic equation y cj' 2+ bf + a are choseri
to c-ive the best fit to the SDT entry corresponding to a signal detection
arici its two adjacent freqtrency bin eritries.
= The signal center frequency is then estirnated as f= - h.
2 e
This algorithm gives the discriminator curve shown in
Figure 14. Note that there is a small systematic error
which cat) be removed if desired. The expected accuracy is
aJ=27 Hz at minimum signal level as discussed hereafter.
14
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
Cross-Product AFC
The received baseband signal is sampled at the output of
the MF so that only those samples corresponding to a
correlation peak are routed for further processing and data
demodulation as discussed above. This decimated baseband
signal is frequency corrected by the coarse frequency
estimate as shown in Figure 12. The AFC loop is then
activated to further reduce frequency error.
The AFC is implemented using the cross product AFC
(CPAFC) algorithm which is described in detail in Natalie,
F. D., AFC Tracking Algorithms", IEEE Transactions on
Communications, August, 1984. The discriminator curve is
given by the equation:
p, (c))=A 2 sine2 ( AG)T-L ~ )sinAruT.
where A is the signal amplitude at the detector input, and
TL is the PN code period. The loop has a pull-in range of
about 0.210wTL or 224 Hz. Typical tracking performance with
a loop bandwidth of 32 Hz is shown in Figure 14. Note that
the loop settles in about 40 ms.
Frame and Word Synchronization
As mentioned above, the Frame and Word Sync Field is
characterized by a length 8 Neuman-Hofman synchronization word
which is mod(2) added to the PN spreading sequence before bi-
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
phase modulation of the signal. One symbol of the sync word
is exactly a PN sequence long. The spreading sequence and
sync word transitions are synchronized to occur at the same
time. The Sync Field is 30 ms long, which corresponds to 32
repetitions of the PN code and 4 repetitions of the sync word.
Figure 11 shows the real part of the baseband signal with sync
modulation after frequency correction by the AFC loop when no
noise is present. This plot depicts the baseband waveform
after sample decimation so that there is one sample per PN
code period (0.9375 ms). The first part of the time history
shows the transient due to the AFC pull-in during the
Acquisition Field. The phase transitions due to the sync word
modulation are present during the Sync Field.
The sync sequence is detected with a MF that is 8 symbols
(one sync word) long. Post detection noncoherent combining of
the output of this filter is accomplished with a sliding
accumulator which sums 4 samples spaced by the sync word
length. Note that the MF has a frequency response with first
nulls at the inverse of the filter length, i.e. 133.3 Hz.
There will be little degradation due to frequency offset
because the AFC is small compared with the filter bandwidth.
Increasing the filter length before combining, while giving
better noise performance, would cause the filter to become
unacceptably narrow.
16
CA 02248647 1998-09-09
WO 97/36383 PCT/US97/04280
The sync filter power output waveform is shown in Figure
12. Note that there are a number of subsidiary correlation
peaks which must be discriminated against in order to avoid a
false sync. Detection of the correct sync position is
important since it gives frame sync and data code word sync.,
Data Demodulation
Data is transmitted using an 8-ary orthogonal code word set
which is bi-phase modulated on the carrier as discussed
earlier. Correct detection of the sync field gives the
necessary timing, as discussed above.
Noncoherent data detection is implemented by observing the
power output of a bank of 8 FIR filters matched to the 8 code
words. The code word corresponding to the filter with the
largest output power is declared the winner and the
appropriate data bits are output by the data demodulator.
Transmitter Description
A functional block diagram of the RAC signal transmitter is
shown in Figure 16. During the Acquisition Field both the
data and sync sources are gated off, and only the PN code is
modulated onto the carrier. The sync word is gated on and
mod-2 added to the PN code during the Sync Field and then
gated off. The data source is gated on and mod-2 added to the
PN code during the data field. The resultant baseband signal
is bi-phase modulated onto the carrier.
17
