Disclosure of Invention
The invention aims to provide a novel frequency offset estimation and correction method in a TD-SCDMA system, which improves the frequency offset correction performance in large frequency offset.
The purpose of the invention is realized by the following steps:
the invention discloses a frequency offset estimation and correction method in a TD-SCDMA system, which is used for estimating and correcting the frequency offset existing between a received signal carrier and a local carrier, and comprises the following steps:
(1) The frequency offset of each user estimated by the previous subframe is used as the frequency offset estimation of the subframe to generate a training sequence generated locally to form a local training sequence with frequency offset, and then the local training sequence with frequency offset and the received training sequence with frequency offset are used for channel estimation;
(2) Processing the frequency offset of each user estimated by the previous subframe and a locally generated spreading and scrambling code composite sequence to form a spreading and scrambling code composite sequence with frequency offset, and performing joint detection by using the spreading and scrambling code composite sequence with frequency offset to estimate a symbol;
(3) Extracting symbol estimation of a certain user from the estimated symbols of all users, wherein the symbol estimation comprises a linear phase caused by the difference of the frequency offset of the current subframe and the previous subframe of the user, and the frequency offset difference can be estimated by utilizing the linear phase and is used for correcting the phase; the phase corrected symbol is output;
(4) And adding the estimated frequency deviation with the frequency deviation of the previous subframe, storing the frequency deviation as the frequency deviation estimation of the subframe, and using the frequency deviation estimation and correction of the next subframe.
The step (1) comprises the following steps:
(11) The training sequences with frequency offset of all users in a time slot are superposed after passing through a multipath channel to obtain a training sequence signal of the time slot:
e m =M fo h+n
wherein e m Is N m X 1 column vector, nm length of training sequence, h KW x 1 column vector formed by channel impulse response of all users in a time slot, K number of users, W length of channel impulse response of one user in chip unit, N N m Noise column vector of x 1, M fo Is a training sequence matrix with frequency offset, dimension N m ×KW;
Taking N m = KW, can obtain M fo :
Wherein, Δ f k And phi k Is the frequency offset and initial phase of the kth user, tc is the chip width, i is the position of the first chip of the training sequence in a time slot,is a transmitted training sequence;
(12) Using the frequency offset estimated by the last sub-frame as the frequency offset estimation of the sub-frame to obtain the local training sequence matrix estimation with frequency offset
Wherein
Is the
frequency offset 211 of the kth user estimated in the previous subframe;
(13) By using
Estimated channel
If frequency deviation of each user is delta f
k If the phase difference is very small, then
Approximating to form a circular matrix, thereby reducing the channel by Fourier transform
The amount of computation of the term.
The step (2) comprises the following steps:
(21) Superposing the data signals with frequency offset of all users in a time slot after passing through a multipath channel to obtain the user data signal of the time slot:
e d =A fo d+n
wherein e is d Is (N) s Q + W-1) × 1 column vector, ns isThe number of user symbols, Q is the spreading factor, d is the column vector formed by all user symbols, and the dimension is N s K × 1,n is N s K × 1 noisy column vector, A fo Is a system transmission matrix with dimensions of (N) s Q+W-1)×N s K;
A fo The structure of (a) is as follows:
A fo is a diagonal array of blocks, block V fo (n) Can be expressed as:
wherein, b k,fo (n) Is the convolution of the spreading scrambling code composite sequence of the k user with frequency offset and the channel impulse response of the k user, the dimension is (Q + W-1) x 1, and is expressed as:
wherein, c fo k Is the composite sequence of the spreading scrambling code of the k user with frequency deviation, and the dimension is (Q + W-1) xW, h k Is the channel impulse response of the kth user with dimension W × 1.
Locally generated spreading scrambling code composite sequence c of k user k Expressed as:
(22) The frequency deviation estimated by the previous subframe is used as the frequency deviation estimation of the subframe, thereby generating the frequency deviation-containing spread spectrum scrambling code composite sequence estimation
Wherein the content of the first and second substances,
is the frequency offset estimation of the k user of the previous subframe;
(23) Calculated from step (13)
Extracting the channel impulse response estimation of each user
K =1, …, K, and
and
synthesis of
V
fo (n) And systematic transmission matrix estimation with frequency offset
Minimum mean square error, MMSE, estimation of all user symbols
Comprises the following steps:
wherein R is n Is a noise correlation matrix, R d Is the correlation matrix of the user symbols.
The invention discloses a frequency offset estimation and correction method in a TD-SCDMA system, which is used for estimating and correcting the frequency offset existing between a received signal carrier and a local carrier, and comprises the following steps:
(1) Receiving a chip-level signal;
(2) Extracting from the chip-level signal a training sequence em and a user data part ed:
(3) Generating a local training sequence matrix M:
(4) Frequency offset estimated from the last subframe
K =1, …, K, which is the frequency offset estimation of the subframe:
(5) Using local training sequence matrices M and
k =1, …, K, producing a local training sequence matrix estimate with frequency offset
(6) Using training sequences e
m And
performing channel estimation to obtain a channel estimation value
(7) Generating a local spread spectrum scrambling code composite sequence matrix C k ,k=1,…,K:
(8) Using c
k K =1, …, K and
k =1, …, K, producing a local spread spectrum scrambling code composite sequence matrix estimate with frequency offset
k=1,…,K,n=1,…,N
s :
(9) From
Extracting channel estimation value of each user
K =1, … K; by using
k=1,…,K,n=1,…,N
s And
k =1, …, K production
k=1,…,K,n=1,…,N
s :
(10) By using
k=1,…,K,n=1,…,N
s Composition V
fo (n) ,n=1,…,N
s :
(11) By usingV
fo (n) ,n=1,…,N
s Forming system transmission matrix with frequency deviation
(12) By using
And e
d Performing joint detection to estimate user symbols
(13) By using
Performing frequency offset estimation to estimate the frequency offset difference between the current subframe and the previous subframe
k=1,…,K;
(14) By using
K =1, …, K pairs of user symbols
Carrying out frequency offset correction and outputting a corrected symbol;
(15) Adding the obtained frequency deviation and the estimated frequency deviation of the previous subframe to be used as the estimated frequency deviation of the next subframe:
k=1,…,K。
in short, the invention carries out channel estimation and joint detection by using the training sequence with frequency offset and the spreading scrambling code composite sequence with frequency offset, thereby realizing the following two points and further achieving the purpose of improving the frequency offset correction performance. First, a channel is estimated more accurately; and secondly, converting the estimated large frequency offset into a small frequency offset of two adjacent subframes.
Detailed Description
The system block diagram of the invention is shown in figure 2. The basic idea is as follows.
And taking the frequency offset 211 of each user estimated in the previous subframe as the frequency offset estimation of the subframe, substituting the frequency offset 211 into the training sequence 203 generated locally to form a local training sequence 204 with frequency offset, and performing channel estimation by using the local training sequence 204 with frequency offset and the received training sequence 201 with frequency offset. Generally, it can be considered that the frequency offset between two adjacent subframes is not changed much by 5 ms. Under such conditions, the estimated channel values 205 are more accurate than those estimated using the local training sequence without frequency offset.
Then, the frequency offset 211 of each user estimated in the previous sub-frame is substituted into the locally generated spreading and scrambling code composite sequence 207 to form a spreading and scrambling code composite sequence 208 with frequency offset, and then the spreading and scrambling code composite sequence with frequency offset is used for joint detection 206. The symbols thus estimated include the linear phase caused by the difference between the frequency offset of the present subframe and the frequency offset of the previous subframe. Since the difference between the frequency offset of the present subframe and the frequency offset of the previous subframe can be considered to be relatively small, the frequency compensation can be performed better by the frequency offset estimation and correction 210 at the following symbol level.
The present invention will be described in detail below.
First, channel estimation
Received training sequence signal e of a time slot m 201 is the superposition of the training sequences with frequency offset of all users in the time slot after passing through the multipath channel, which can be expressed as
e m =M fo h+n(1)
Wherein e m Is N m X 1 column vector, nm length of training sequence, h KW x 1 column vector formed by channel impulse response of all users in a time slot, K number of users, W length of channel impulse response of one user in chip unit, N N m Noise column vector of x 1, M fo Is a training sequence matrix with frequency deviation and dimension N m X KW. When designing a system, to simplify the calculation, generally take N m =KW。 M fo Can be expressed as
Wherein Δ f
k And phi
k Is the frequency offset and initial phase of the kth user, tc is the chip width, i is the position of the first chip of the training sequence in a time slot,
is the transmitted training sequence.
The locally generated training sequence matrix M203 is represented as
It can be seen that M is a circulant matrix, i.e. each row is cyclically shifted to the right by one element, i.e. the next row. Using the frequency offset estimated from the previous sub-frame as the sub-frameFrequency offset estimation, thereby producing a local training sequence matrix estimate with frequency offset
Wherein
The frequency offset 211 of the kth user estimated in the previous subframe.
By using
Estimated channel
Can be expressed as
And the channel estimated by the local training sequence matrix MCan be expressed as
When frequency deviation is delta f
k When the frequency offset is larger, under the condition that the frequency offset difference between two adjacent subframes is not large,
ratio of
Is accurate.
In the formula (5), the following requirements are satisfied
The calculation amount is large. If frequency offset of each user is delta f
k If the phase difference is very small, then it will be
An approximation process is performed to form a circulant matrix, thereby reducing the amount of calculation by using FFT (Fourier transform). Assuming frequency offset estimation for each user
Average value of (2)
Is composed of
Will be provided with
Make the following approximation
Then
Becomes a circulant matrix that can be computed using a Fourier transform FFT
Second, joint detection
Received user data signal e of one time slot d 201 is the superposition of the data signals with frequency offset of all users in the time slot after passing through the multipath channel, which can be expressed as
e d =A fo d+n
Wherein e d Is (N) s Q + W-1) x 1, ns is the number of user symbols, Q is the spreading factor, d is the column vector of all user symbols, and the dimension is N s K × 1,n is N s K × 1 noise column vector, afo is the system transmission matrix, with dimensions of (N) s Q+W-1)×N s K. Afo has the following structure:
afo is a block diagonal matrix, block V fo (n) Can be expressed as
Wherein b is k,fo (n) Is the convolution of the spreading scrambling code composite sequence of the k user with frequency deviation and the channel impulse response of the k user, the dimension is (Q + W-1) multiplied by 1, which is expressed as
Wherein c is fo k Is the composite sequence of the spreading scrambling code of the k user with frequency deviation, and the dimension is (Q + W-1) xW, h k Is the channel impulse response of the kth user with dimension W × 1.
The locally generated spreading scrambling code composite sequence ck207 of the k-th user is denoted as
The
frequency deviation 211 estimated by the previous subframe is used as the frequency deviation estimation of the subframe, thereby generating the frequency deviation-containing spread spectrum scrambling code composite sequence estimation
Is shown as
Wherein
Frequency offset
estimate 211 for the k user of the previous subframe. Calculated from the first step
Extracting the channel impulse response estimation of each user
K =1, …, K, and
and
synthesis of
V
fo (n) And systematic transmission matrix estimation with frequency offset
MMSE estimation of all user symbols
Is composed of
Wherein R is n Is a noise correlation matrix, R d Is the correlation matrix of the user symbols. R n Can be estimated from measured data, R d Can be calculated by a priori knowledge.
Notably, the use of systematic transmission matrix estimation with frequency offset
To perform symbol estimation, the estimated symbol
The method includes a linear phase caused by the difference of the frequency offset of the current subframe and the previous subframe, but not a linear phase caused by the frequency offset of the current subframe.
Third, frequency offset estimation and correction
All user symbols estimated from the second step
Extracting symbol estimation of the k user
The linear phase caused by the frequency offset difference between the current subframe and the previous subframe of the kth user is included, and the frequency offset difference between the current subframe and the previous subframe can be generally considered to be small (< 200 Hz). The frequency offset 210 can be estimated by using the linear phase and corrected by using the frequency offset
Phase 210 of (a). The phase corrected
symbol 213 is output. And meanwhile, adding the estimated frequency deviation with the frequency deviation of the previous subframe, storing the frequency deviation as the frequency deviation estimation of the subframe, and using the frequency deviation estimation and correction of the next subframe. There are many documents and patents on how to perform the frequency offset estimation and
correction 210, which are not in the scope of the present invention.
The implementation steps of the technical scheme are further described in detail in the following steps with the combination of the accompanying drawings:
1. receiving a chip-level signal 201
2. The training sequence e is extracted from the received chip-level signal 201 m And a user data section e d 。
3. A local training sequence matrix M203 is generated, as in equation (3).
4. Frequency offset estimated from the last subframeK =1, …, K211 as the frequency offset estimate of the present subframe.
5. With M and
k =1, … K, producing a local training sequence matrix estimate with frequency offset
Such as formula 4.
6. With e
m And
channel estimation 202 is performed as shown in equation (5). Obtaining a channel estimate
7. Generating a composite sequence matrix c of local spreading scrambling codes k K =1, …, K207, as in formula (13).
8. By c
k K =1, …, K and
k =1, … K, producing a local spread spectrum scrambling code composite sequence matrix estimate with frequency offset
k=1,…,K,n=1,…
N s 208, as in formula (14).
9. From
Extracting channel estimation value of each user
k =1, … k. By using
k=1,…,K,n=1,…,N
s And
k =1, …, K production
k=1,…,K,n=1,…N
s Such as formula 15.
10. By usingk=1,…,K,n=1,…,N s Composition V fo (n) ,n=1,…,N s Such as formula (16).
11. By V
fo (n) ,n=1,…,N
s Forming system transmission matrix with frequency deviation
Such as formula (17).
12. By using
And e
d Performing joint detection to estimate user symbols
As in formula (18)
13. By using
Performing frequency offset
estimation 210 to estimate the frequency offset between the current subframe and the previous subframe
k=1,…,K
14. By using
K =1, …, K pairs of user symbols
Frequency offset
correction 210 is performed and corrected
symbol 213 is output.
The invention achieves the purpose of improving the frequency offset correction performance in large frequency offset through the following two points:
1. and the local training sequence with frequency offset is used for channel estimation, so that the channel estimation is more accurate.
2. And performing symbol estimation by using the local spread spectrum scrambling code composite sequence with the frequency offset, so that the estimated symbol comprises a linear phase caused by the frequency offset difference between the current subframe and the previous subframe, and further converting the estimated large frequency offset into the small frequency offset of the adjacent subframe.