CA 02554193 2001-O1-05
CHANNEL ESTIMATION FOR TIME DIVISION
DUPLEX COMMUNICATION SYSTEMS
This application is a division of Canadian patent application Serial
No. 2,396,571 filed internationally on January 5, 2001 and entered nationally
on
July 5, 2002.
BACKGROUND
The invention generally relates to wireless communication systems. In
particular, the invention relates to channel estimation in a wireless
communication
system.
Figure 1 is an illustration of a wireless communication system 10. The
communication system 10 has base stations 12, to 125 which communicate with
user equipments (UEs) 14, to 143. Each base station 12, has an associated
operational area where it communicates with UEs 14, to 143 in its operational
area.
In some communication systems, such as code division multiple access
(CDMA) and time division duplex using code division multiple access
(TDD/CDMA), multiple communications are sent over the same frequency
spectrum. These communications are typically differentiated by their chip code
sequences. To more efficiently use the frequency spectrum, TDD/CDMA
communication systems use repeating frames divided into time slots for
communication. A communication sent in such a system will have one or multiple
associated chip codes and time slots assigned to it based on the
communication's
bandwidth.
Since multiple communications may be sent in the same frequency spectrum
and at the same time, a receiver in such a system must distinguish between the
multiple communications. One approach to detecting such signals is single user
detection. In single user detection, a receiver detects only the
communications
from a desired transmitter using a code associated with the desired
transmitter, and
treats signals of other transmitters as interference. Another approach is
referred to
CA 02554193 2001-O1-05
-2-
as joint detection. In joint detection, multiple communications are detected
simultaneously.
To utilize these detection techniques, it is desirable to have an estimation
of
the wireless channel in which each communication travels. In a typical TDD
system,
the channel estimation is performed using midamble sequences in communication
bursts.
A typical communication burst 16 has a midamble 20, a guard period 18 and
two data bursts 22, 24, as shown in Figure 2. The midamble 20 separates the
two
data bursts 22, 24 and the guard period 18 separates the communication bursts
16 to
allow for the difference in arrival times of bursts 16 transmitted from
different
transmitters. The two data bursts 22, 24 contain the communication burst's
data. The
midamble 20 contains a training sequence for use in channel estimation.
After a receiver receives a communication burst 16, it estimates the channel
using the received midamble sequence. When a receiver receives multiple bursts
16
in a time slot, it typically estimates the channel for each burst 16. One
approach for
such channel estimation for communication bursts 16 sent through multiple
channels is a Steiner Channel Estimator. Steiner Channel Estimation is
typically
used for uplink communications from multiple UEs, 14, to 143, where the
channel
estimator needs to estimate multiple channels.
"Optimum and Suboptimum Channel Estimation for the Uplink of CDMA
Mobile Radio Systems with Joint and Detection," by Steiner and Jung in
European
Transactions on Telecommunications and Related Technologies (1994), IT, AEI,
Milano, Vol. 5, No. 1, 39-50, discloses an approach for channel estimation.
One
approach uses a single cyclic correlator. Using the known transmitted midamble
sequences, a matrix M is constructed. The received midamble vector a is
multiplied
by the first column of M. The multiplication is performed by the cyclic
correlator
over P values and by shifting the values 2P-1 times. P is a period of the
midamble
codes.
In some situations, multiple bursts 16 experience the same wireless channel.
One case is a high data rate service, such as a 2 megabits per second (Mbps)
service.
CA 02554193 2001-O1-05
-3-
In such a system, a transmitter may transmit multiple bursts in a single time
slot.
Steiner estimation can be applied in such a case by averaging the estimated
channel
responses from all the bursts 16. However, this approach has a high
complexity.
Accordingly, it is desirable to have alternate approaches to channel
estimation.
SUMMARY
A single transmitter transmits K communication bursts in a shared spectrum in
a time slot of a time division duplex communication system. The system
associated
with N mid -amble sequences. Each burst has an associated midamble sequence. A
receiver receives a vector corresponding to the transmitted midamble sequences
of the
K communication bursts. A matrix having N identical right circulant matrix
blocks is
constructed based in part on the known N midamble sequences. The wireless
channel
between the transmitter and receiver is estimated based on in part one of the
N blocks
and the received vector.
The invention thus provides according to a first aspect, for a method for
estimating a wireless channel in a time division duplex communication system
using
code division multiple access. The system is associated with N midamble
sequences,
the wireless channel existing between a single transmitter and a single
receiver, the
single transmitter transmitting K communication bursts in a shared spectrum in
a time
slot, each burst having an associated midamble sequence of the N sequences,
the
receiver knowing the N midamble sequences and receiving a vector corresponding
to
the transmitted midamble sequences of the K communication bursts at the single
receiver. The method is characterized by constructing a matrix having N
identical right
circulant matrix blocks based in part on the known N midamble sequences, and
estimating the wireless channel based on in part one of the N blocks and the
received
vector.
The invention provides according to a second aspect, for a receiver for use in
a wireless time division duplex communication system using code division
multiple
access. The system is associated with N midamble sequences. A single
transmitter in
the system transmits K communication bursts in a shared spectrum in a time
slot, each
CA 02554193 2001-O1-05
-4-
burst having an associated midamble sequence of the N sequences, the receiver
knowing the N midamble sequences, the receiver comprising an antenna for
receiving
the K communication bursts including a vector corresponding to the transmitted
midamble sequences of the bursts. The receiver is characterized by a channel
estimator
for constructing a matrix having N identical right circulant-matrix blocks
based in part
on the known N midamble sequences and estimates the wireless channel between
the
receiver and the single transmitter based on in part one of the N blocks and
the received
vector, and a data detector for recovering data from the received
communication bursts
using the estimated wireless channel.
According to a third aspect, the invention provides for a wireless spread
spectrum communication system using code division multiple access associated
with
N midamble sequences. The system communicates using communication bursts, each
burst having an associated midamble sequence, a base station comprising: a
data
generator for generating data; a plurality of modulation/spreading devices for
formatting the generated data into K communication bursts time multiplexed to
be in
a same time slot and is a shared spectrum; and an antenna for radiating the K
communication bursts, and a user equipment which comprises an antenna for
receiving
the K communication bursts including a vector corresponding to the transmitted
midamble sequences of the bursts. The system is characterized by the user
equipment
comprising a channel estimator for constructing a matrix having N identical
right
circulant matrix blocks based in part on the N midamble sequences and
estimating the
wireless channel between the base station and the user equipment based on in
part one
of the N blocks and the received vector, and a data detector for recovering
data from
the received communication bursts using the estimated wireless channel.
According to a fourth aspect, the invention provides for a method for
estimating
a wireless channel in a time division duplex communication system using code
division
multiple access, the wireless channel existing between a single transmitter
and a single
receiver, the single transmitter transmitting K communication bursts in a
shared
spectrum in a time slot, each burst having an associated midamble sequence,
and the
receiver knowing the midamble sequences of the K bursts. The method comprises:
CA 02554193 2001-O1-05
-5-
receiving a vector corresponding to the transmitted midamble sequences of the
K
communication bursts at the single receiver; constructing a matrix having K
right
circulant matrix blocks based in part on the known K midamble sequences; and
estimating the wireless channel based on in part the K block matrix and the
received
vector.
According to a fifth aspect, the invention provides for a receiver for use in
a
wireless time division duplex communication system using code division
multiple
access. A single transmitter in the system transmits K communication bursts in
a
shared spectrum in a time slot, each burst having an associated midamble
sequence, and
the receiver knowing the midamble sequences of the K bursts. The receiver
comprises:
an antenna for receiving the K communication bursts including a vector
corresponding
to the transmitted midamble sequences of the bursts; a channel estimator for
constructing a matrix having K right circulant-matrix blocks based in part on
the known
K midamble sequences and estimating the wireless channel between the receiver
and
the single transmitter based on in part the K block matrix and the received
vector; and
a data detector for recovering data from the received communication bursts
using the
estimated wireless channel.
According to a sixth aspect, the invention provides for a user equipment
communicating in a code division multiple access format associated with N
midamble
sequences and the user equipment receiving K communication bursts in a shared
spectrum in a time slot, each burst having an associated midamble sequence of
the N
sequences, the user equipment knowing the N midamble sequences and comprising
an
antenna for receiving the K communication bursts including a vector
corresponding to
the transmitted midamble sequences of the bursts. The user equipment is
characterized
by: a channel estimator for constructing a matrix having N identical right
circulant-
matrix blocks based in part on the known N midamble sequences and estimating
the
wireless channel between the receiver and the single transmitter based on in
part one
of the N blocks and the received vector; and a data detector for recovering
data from
the received communication bursts using the estimated wireless channel.
CA 02554193 2001-O1-05
-6-
BRIEF DESCRIPTION OF THE DRAWINGS)
Figure 1 is a wireless communication system.
Figure 2 is an illustration of a communication burst.
Figure 3 is a simplified multiburst transmitter and receiver.
Figure 4 is a flow chart of multiburst channel estimation.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS)
Figure 3 illustrates a simplified multicode transmitter 26 and receiver 28 in
a TDD/CDMA communication system. In a preferred application, such as a 2 Mbs
downlink service, the receiver 28 is in a UE 14, and the transmitter 26 is in
a base
station 12,, although the receiver 28 and transmitter 26 may be used in other
applications.
The transmitter 26 sends data over a wireless radio channel 30. The data is
sent in K communication bursts. Data generators 32, to 32K in the transmitter
26
generate data to be communicated to the receiver 28. Modulation/spreading and
training sequence insertion devices 34, to 34K spread the data and make the
spread
reference data time-multiplexed with a midamble training sequence in the
appropriate assigned time slot and codes for spreading the data, producing the
K
communication bursts. Typical values of K for a base station 12, transmitting
downlink bursts are from 1 to 16. The communication bursts are combined by a
combiner 48 and modulated by a modulator 36 to radio frequency (RF). An
antenna
38 radiates the RF signal through the wireless radio channel 30 to an antenna
40 of
the receiver 28. The type of modulation used for the transmitted communication
can be any of those known to those skilled in the art, such as binary phase
shift
keying (BPSK) or quadrature phase shift keying (QPSK).
2S The antenna 40 of the receiver 28 receives various radio frequency signals.
The received signals are demodulated by a demodulator 42 to produce a baseband
signal. The baseband signal is processed, such as by a channel estimation
device 44
and a data detection device 46, in the time slot and with the appropriate
codes
assigned to the transmitted communication bursts. The data detection device 46
CA 02554193 2001-O1-05
_7-
may be a multiuser detector or a single user detector. The channel estimation
device 44 uses the midamble training sequence component in the baseband signal
to
provide channel information, such as channel impulse responses. The channel
information is used by the data detection device 46 to estimate the
transmitted data
of the received communication bursts as hard symbols.
To illustrate one implementation of multiburst channel estimation, the
following midamble type is used, although multiburst channel estimation is
applicable to other midamble types. The K midamble codes, m~k~ , where k =
1...K,
are derived as time shifted versions of a periodic single basic midamble code,
m,~ ,
of period P chips. The length of each midamble code is Lm = P + W -1. W is the
length of the user channel impulse response. Typical values for L"' are 256
and
512 chips. W is the length of the user channel impulse response. Although the
following discussion is based on each burst having a different midamble code,
some
midambles may have the same code. As, a result, the analysis is based on N
midamble codes, N < K. Additionally, the system may have a maximum number of
acceptable midamble codes N. The receiver 28 in such a system may estimate the
channel for the N maximum number of codes, even if less than N codes are
transmitted.
The elements of m~ take values from the integer set { 1, -1 ~. The sequence
m,, is first converted to a complex sequence mp~i~ = j' ~ mp~i~ , where i = 1~
~ ~ P .
The m~k~ are obtained by picking K sub-sequences of length Lm from a 2P long
sequence formed by concatenating two periods of mp . The i'h element of m~k~
is
CA 02554193 2001-O1-05
_g_
related to mP by Equation 1.
m; k) =mp~~K-k~W+i], for 1Si < P-~K-k~W
=mP~i-P+~K-k~W], fore-~K-k~W<-i__<P+W-1
Equation 1
Thus, the starting point of m(k~,k = 1...K shifts to the right by W chips as k
increases from 1 to K.
The combined received midamble sequences are a superposition of the K
convolutions. The 7~" convolution represents the convolution of m~k) with h(k)
.
h(k) is the channel response of the k'" user. The preceding data field in the
burst
corrupts the first ~W-1~ chips of the received midamble. Hence, for the
purpose
of channel estimation, only the last P of L", chips are used to estimate the
channel.
Multiburst channel estimation will be explained in conjunction with the flow
chart of Figure 4. To solve for the individual channel responses h(k) ,
Equation 2 is
used.
(I)
mP '.. n1(K-pw+I i rrt(K_,)w "' m(x-z)w+~ i i mw "' m~ _h rw
I I I
rnl "' n1(K-pw+z i m(K-I)w+I "' m(K-z)w+z i i mw+I "' m2 h(2) rw+I
I I I
rw+z
1n2 ~ ~ ~ YIZ(K-1)W+3 j ri1(K-1)W+2 ' ~- YI1(K-2)GV+3 j ~ ~ ~ j mW+2 "- )n3
I I I
I I I ,
I . . . I I . . . ,
I I I
rrlKw-I ~ rn(K-1)W ~ m(K_I)W-I ~.. m(K-2)W i i mw-I mP h(K) rL.n
Equation 2
rw~ ~ ~rL,u are the received combined chips of the midamble sequences. The m
values are the elements of mP
CA 02554193 2001-O1-05
-9-
Equation 2 may also be rewritten in shorthand as Equation 3.
K
M~k~ h~k~ = r Equation 3
k-1
Each M~k~ is a KW-by-W matrix. r is the received midamble chip responses.
When all the bursts travel through the same channel, h~l~- ~ ~h~k~ can be
replaced by
h as in Equation 4, 50.
K _
Mtk~ h = r Equation 4
k=1
G is defined as per Equation S.
G = [M~'~,..., M~k~,..., M~K~] Equation 5
As a result, G is a KW-by-KW matrix. Since G is a right circulant matrix,
Equation
4 can be rewritten using K identical right circulant matrix blocks B, as per
Equation
6, 52.
B
K B
M~k~ : = D Equation 6
k=l '
B
B is a W-by-W right circulant matrix. The number of B-blocks is K. Using
Equation 6, Equation 4 can be rewritten as Equation 7.
Dh = r Equation 7
CA 02554193 2001-O1-05
-10-
Equation 7 describes an over-determined system with dimensions KW-by-W. One
approach to solve Equation 7 is a least squares solution, 54. The least
squares
solution of Equation 7 is given by Equation 8.
h = (DHD) I DHY Equation 8
DH is the hermitian of D.
Applying Equation 6 to Equation 8 results in Equation 9.
~DHD~ 1 - K ~BHB~ 1 Equation 9
The received vector ~ of dimension KW can be decomposed as per Equation 10.
Y~
Y~
Y = Equation 10
rk
The dimension of ~k is W. Substituting Equations 9 and 10 into Equation 8, the
least-squares solution for the channel coefficients per Equation 11 results.
K
h = ~BHB~ ~ BH K ~ ~k = ~BHB~ IBHt~k Equation 11
k-1
i~k represents the average of the segments of r . Since B is a square matrix,
Equation 11 becomes Equation 12.
h = B-~~k Equation 12
CA 02554193 2001-O1-05
-11-
Since B is a right circulant matrix and the inverse of a right circulant
matrix is also
right circulant, the channel estimator can be implemented by a single cyclic
correlator of dimension 57, or by a discrete fourier transform (DFT) solution.
A W point DFT method is as follows. Since B is right circulant, Equation 13
can be used.
B = DW' ~ A~ ~ Dw Equation 13
DW is the W point DFT matrix as per Equation 14.
~o .ro ~o ~o ~.o
W W W W "' W
w w' WZ W3 ... ~,(W-1)
0 2 4 6 2(W-1)
.'.
Dw W W 3 W 6 W9 ~3(w-~) Equation 14
= W W W "' W
W
W(w-1)WZ(W-') W3(W-')... ~(W-1)(w-1)
A ~ is a diagonal matrix whose main diagonal is the DFT of the first column of
B, as
per Equation 15.
A~ = diag(Dw(B(:,1))) Equation 15
_2 ~r
W = e-' w . Thus, DW is the DFT operator so that DWx represents the W point
DFT
of the vector x . By substituting Equation 13 into Equation 12 and using
DW' - W , results in Equation 16.
h = CDw* ~ ~ ~ A~ ~ DwJ r Equation 16
Dw is the element-by-element complex conjugate of DW.
CA 02554193 2001-O1-05
-I2-
Alternately, an equivalent form that expresses ~ in terms of AR instead of
A~, can be derived. AR is a diagonal matrix whose main diagonal is the DFT of
the
first row of B per Equation 17.
AR = diag(DW~B~I,:~)) Equation 17
S Since the transpose of B, BT, is also right circulant and that its first
column is the
first row of B, BT can be expressed by Equation 18.
BT = DW' ~ AR ~ DW Equation 18
Using Equation 18 and that DW = DW,AR = AR and that for any invertible matrix
A, (AT ) ~ _ (A-')T , B can be expressed as per Equation 19.
I0 B = DW ~ AR ~ Dw' Equation 19
Substituting Equation 19 into Equation I2 and that DW' = D~ results in
Equation
W
20.
h = ~ DW ~ A R ~ ~ DwJ r Equation 20
Equations 16 or 20 can be used to solve for h . Since all DFTs are of length
W, the
15 complexity in solving the equations is dramatically reduced.
An approach using a single cycle correlator is as follows. Since B-' is the
inverse of a right circulant matrix, it can be written as Equation 21.
CA 02554193 2001-O1-05
-13-
T~ TP w T3 Tz
Tz T, w T4 T3
B-1 = : : ~ ~ : : Equation 21
T = ~
T w-, T w-z T, T
"' w
Tw Tw-, '-' Tz T,
The first row of the matrix T is equal to the inverse DFT of the main diagonal
of
A R . Thus, the matrix T is completely determined by A R .
The taps of the channel response h are obtained successively by an inner
product of successive rows of T with the average of W-length segments of the
received vector r . The successive rows of T are circularly right shifted
versions of
the previous row. Using registers to generate the inner product, the first
register
holds the averaged segments of r , and the second register is a shift register
that
holds the first row of the matrix T. The second register is circularly shifted
at a
certain clock rate. At each cycle of the clock, a new element of h is
determined by
the inner product of the vectors stored in the two registers. It is
advantageous to
shift the first row of the matrix T rather than the received midambles. As a
result,
no extra storage is required for the midambles. The midambles continue to
reside
in the received buffer that holds the entire burst. Since the correlator
length is only
W, a significant reduction in complexity of estimating the channel is
achieved.
