Background technology
In mobile communication, mobile device is in when start, often and have bigger frequency departure between the crystal oscillator of base station.Mobile device must carry out frequency compensation, by certain frequency adjustment process, frequency deviation adjust to acceptable among a small circle in.Because there is frequency drift in crystal oscillator, mobile device also must carry out frequency-tracking, and frequency deviation is maintained within the small range.In the TD-SCDMA system, mobile device is wanted correct demodulating data, just need make frequency deviation less than 0.01ppm, i.e. 200Hz.
No matter be frequency adjustment or frequency-tracking, its precondition all is accurately and apace to estimate frequency deviation, so frequency deviation estimating method is most important to the performance of mobile device.
At present, using training sequence to carry out in the method for frequency offset estimating, better performances be that maximum likelihood estimate is (referring to document [1] Marco Luise, Ruggero Reggiannini, " Carrier Frequency Recoveryin All-Digital Modems for Burst-Mode Transmissions ", IEEE Trans.on Commun., vol.43, no.2/3/4, Feb/Mar/Apr 1995pp.1169-1178).It carries out related operation to the training sequence data that receives and the training sequence of transmission, is only contained the related data of frequency deviation information, then related data is asked for the maximum of likelihood function.This method complexity is low, but operand is very big, and does not have simple closed solution.After approximate processing, can obtain a kind of maximum likelihood estimate of suboptimum.This sub-optimal algorithm has significantly reduced operand, is attainable.But basic maximum likelihood estimate is anti-multipath not, has bigger evaluated error under multi-path environment, and it can not be used for environment that frequency offset estimation accuracy is had relatively high expectations, as frequency accurate adjustment process and frequency-tracking process.Document [2] Michele Morelli and Umberto Mengali, " Frequency Estimation for the Downlink of the UMTS-TDD Component ", IEEETrans.On Wireless Commun., vol.1, No.4, if Oct.2002 has put down in writing and introduced channel impulse response in frequency offset estimating, just can resist multi-path influence, but these class methods often to be carried out complicated matrix operation, have both strengthened complexity, increased operand again, thereby be not easy to realize.
In the TD-SCDMA system, use training sequence to carry out frequency offset estimating, also must consider the characteristics of its training sequence.The training sequence of TD-SCDMA (midamble sign indicating number) can be the stack sum of a plurality of subcodes (user midamble sign indicating number), though the long user midamble sign indicating number of these 144 chips (chip) all is to be formed by a basic midamble sign indicating number (128chip) cyclic shift, receiving terminal does not also know to comprise which user midamble sign indicating number in the midamble data that transmitting terminal sends.When sending a plurality of user midamble sign indicating number,, will cause bigger evaluated error if directly use basic midamble sign indicating number to carry out maximal possibility estimation.Therefore, when estimating frequency deviation, also must estimate the composition of the midamble data of transmission earlier.
In frequency offset estimating, introduce channel impulse response though application number is 200410000119.5 patent, and do not needed to carry out complicated matrix operation.But its estimated channel impulse response, itself just has been subjected to the influence of frequency deviation.And when estimating the described reference signal of this patent, (be equivalent to the composition of estimating to send the midamble data), directly basic midamble sign indicating number and the channel impulse response with 128chip carries out circular convolution, only used the 128chip in the 144chip training sequence that system provides to carry out frequency offset estimating.These 2 have all reduced the accuracy that this method is estimated frequency deviation, thereby it can only be used for the frequency coarse adjustment stage, can not be used for the frequency accurate adjustment.
Summary of the invention
The object of the present invention is to provide a kind of frequency deviation estimating method and device, this method can accurately estimate frequency departure, can adapt to the system that is made of a plurality of subcodes, as the system of the training sequence of TD-SCDMA, have good anti-multipath performance, and simple, be easy to realize.
Technical scheme of the present invention is as follows:
The inventive method is to estimate the reception data that not influenced by frequency deviation by receiving data r
Carry out frequency offset estimating by basic maximum likelihood estimate then.Fig. 1 is the equivalent periodic line model of the principle of the invention.The midamble data of r for receiving among the figure,
Be the reception midamble data that not influenced by frequency deviation,
Be the midamble data that send, Δ f is that the frequency deviation on chip (chip) frequency is arrived in the frequency departure equivalence of local carrier and base station carrier,
For estimating frequency deviation, h influences channel impulse response, h for not being subjected to frequency deviation
FreqBe the original channel impulse response that influenced by frequency deviation Δ f.Therefore, r with
Between only differ frequency deviation information, as long as estimate
Just can estimate frequency deviation accurately by basic maximum estimated method.
The flow chart of the inventive method is seen Fig. 2, comprising the following step:
A) the midamble data r of a certain time slot of reception;
B) carry out channel estimating, obtain the original channel impulse response h that influenced by frequency deviation
Freq
C) detect the composition that sends the midamble data;
D) estimate the channel impulse response h not influenced by frequency deviation;
E) estimate the reception midamble data not influenced by frequency deviation
F) carry out the maximum likelihood frequency offset estimating, frequency deviation is estimated in output.
In the step b) of described method,, carry out channel estimating by following formula to the midamble data that receive and known basic midamble sign indicating number:
H in the formula
FreqBe the channel impulse response that influenced by frequency deviation, r is the midamble data that receive, and m is basic midamble sign indicating number, and fft and ifft represent fast fourier transform and inverse fast fourier transform respectively.
In the step c) of described method, owing to send each user midamble sign indicating number u that the midamble data comprise
kBe to form h by the same intervals cyclic shift by same basic midamble sign indicating number
FreqWill be by channel impulse response h corresponding to each user midamble sign indicating number
kForm each u according to the order of sequence
kH
kAt h
FreqIn the position be this u
kChannel estimation window.If sent certain u
k, each bar multipath of channel all can appear in k the channel estimation window, otherwise, the impulse response h in the window
kTo only determine by noise.Therefore can be by comparing the power of each window impulse response, judging in which window has signal, thereby detects the composition that sends the midamble data.
In the step d) of described method, comprise following steps:
D1) at h
FreqIn each has in the estimating window of signal, find out the impulse response of representing each multipath of channel;
The power of the impulse response of each multipath in each window when d2) estimating not influenced by frequency deviation;
The phase place of the impulse response of each multipath in each window when d3) estimating not influenced by frequency deviation;
In the step e) of described method, according to step c) and d) the result, estimate the reception midamble data not influenced by frequency deviation by following formula
1≤n≤N
In the formula, u
k(n-t
i) be the user midamble sign indicating number u of k estimating window correspondence
kT moves right
iChip, h
k(t
i) be the impulse response of i bar multipath in k the window of h, t
iBe the time delay of this multipath with respect to article one multipath, N is the length of midamble data.
In the step e) of described method, may further comprise the steps:
E1) calculate the related data y that is used for the maximum likelihood frequency offset estimating:
1≤n≤N
In the formula
*Conjugation is got in expression, and N is the length of midamble data.
E2) calculate the estimation frequency deviation
Promptly ask for likelihood function by maximum likelihood method
Maximum, it is peaked that likelihood function is got
It is exactly the estimated value of actual frequency deviation Δ f.
The suboptimum maximum likelihood estimate that can use document [1] to provide, to reduce operand, it estimates that frequency deviation is:
N is the length of midamble data in the formula.
Use the frequency deviation estimation device of the inventive method, form by following four parts:
F) signal receiving device: be used to receive the signal that the base station sends, and from receiving data sequence, intercept out the midamble data.
G) channel estimating apparatus: the midamble data that receive are carried out channel estimating, and it at first obtains the original channel impulse response h that influenced by frequency deviation
Freq, estimate the channel impulse response h that not influenced by frequency deviation then.
H) Midamble composition checkout gear: according to channel impulse response h
FreqPower in each estimating window detects and sends the user midamble sign indicating number u that comprises in the midamble data
k
I) signal estimation unit:, estimate the reception midamble data that not influenced by frequency deviation according to the user midamble sign indicating number that comprises in channel impulse response h that not influenced by frequency deviation and the midamble data
J) frequency deviation estimation device: frequency deviation is carried out maximal possibility estimation.
Advantage of the present invention is:
1. the present invention has introduced channel impulse response in frequency offset estimating, and has removed the influence of frequency deviation to channel impulse response, thereby has good anti-multipath performance.
2. the present invention has used whole midamble data that system provides, and frequency deviation is carried out maximal possibility estimation, thereby can reach very high estimated accuracy.
3. operand of the present invention is low, and is easy to realize.
The present invention can solve the frequency offset estimating problem of TD-SCDMA system in frequency accurate adjustment and frequency-tracking process, and can be applicable in other system with similar training sequence.
Embodiment
Below by embodiment method of the present invention and device are elaborated.Concrete implementation step in the TD-SCDMA system is as follows:
1. reception data
Receive the midamble data r of a certain time slot, the length of r is 144chip.
2. estimate the original channel impulse response that influenced by frequency deviation
Take out continuous 128chip data r ' among the r, be calculated as follows the original channel impulse response h that influenced by frequency deviation
Freq
Fft and ifft represent fast fourier transform and inverse fast fourier transform respectively in the formula, and m is basic midamble sign indicating number, and length is 128chip, h
FreqLength also be 128chip.
3. detect the composition that sends the midamble data
The training sequence that the TD-SCDMA system sends is by a plurality of user midamble sign indicating number u
kForm, these yards are basic midamble sign indicating number results of cyclic shift at certain intervals, referring to Fig. 3.Number of users K among the figure is 4, { m
1, m
2, Λ, m
128) be basic midamble sign indicating number, u
1, u
2, u
3And u
4Be respectively first to the 4th user midamble sign indicating number.The quantity K of the user midamble sign indicating number that basic midamble sign indicating number can generate is by default.Because each u
kAt h
FreqIn separately corresponding to a channel estimation window, K is the quantity of estimating window just, thereby claim that here K is a window number.It is long that shift intervals W is called window, and it and window number have following relation:
When estimating to send the composition of midamble data, because h
FreqBy corresponding to each u
kImpulse response h
kForm according to the order of sequence,
h
freq(n)=h
k(n-(k-1)W) (k-1)W+1≤n≤kW
If receive among the data r and comprise certain u
kComposition, each bar multipath of channel all can appear in k the channel estimation window, h
kTo include more powerful impulse response, otherwise, h
kWill be only by the noise decision, its power is less relatively, referring to Fig. 4.Condition among the figure is: the midmable data of transmission comprise u
1, u
2And u
3, channel has two multipaths, and window number K is 8, and the long W of window is 16chip.Corresponding to u
1, u
2, u
3 Channel estimation window 1,2,3 in two impulses that power is bigger are respectively arranged, and the power of impulse response is all less in other window.Therefore, estimate to send the composition of midamble data according to the following steps:
1), determines each user midamble sign indicating number u according to known window number K and the long W of window
kAt h
FreqIn the position of channel estimation window;
2) find out channel impulse response h in each window
kMaximum power value, and as the power of this window;
3) power of each window and thresholding are compared,, think that then this window has signal, promptly comprise u in the midamble data of Fa Songing as greater than thresholding
k, otherwise think this window no signal.
4. estimate the channel impulse response h not influenced by frequency deviation.
Referring to Fig. 5.Frequency deviation to the condition of channel impulse response influence is: the midmable data of transmission comprise u
1And u
4, channel has three multipaths, and window number K is 16, and the long W of window is 8chip.
Referring to Fig. 5 (a).During no frequency deviation, the phase place of each multipath of channel corresponding impulse response in each channel estimation window all should be identical with power ratio.
Referring to Fig. 5 (b).When having frequency deviation, each multipath is at h
FreqThe phase place of impulse response is no longer identical with power ratio in each window.
The channel impulse response h that estimation is not influenced by frequency deviation comprises following steps:
1) at h
FreqIn each has in the estimating window of signal, relatively the power of impulse response is selected the bigger impulse response of power, as the impulse response of representing each multipath of channel.
2) calculate the power ratio of representing the impulse response of each multipath in each window, each power ratio is averaged, estimate the performance number of each multipath impulse response in each window when not influenced by frequency deviation.
3) estimate the phase place of the impulse response of each multipath in each window.
At first, calculate the phase difference of the impulse response of each multipath in each window.For example the phase difference of i bar multipath impulse response in k1 and k2 window is
Δ
i,k1,k2=
i,k1-
i,k2
And calculate θ by following formula
I, k1, k2:
In the formula, W is the length of channel estimation window.
Again to each θ
I, k1, k2Value averages, and obtains the chip of the unit angular frequency increment θ that is caused by frequency deviation.Estimate the phase place of each multipath at last by following formula
In the formula, t
iBe the time delay of i bar multipath with respect to article one multipath.
According to above result, be calculated as follows the channel impulse response h that not influenced by frequency deviation:
In the formula, h
k(t
i) be the impulse response of i bar multipath in k the window of h, t
iBe the time delay of this multipath with respect to article one multipath, p
I, kBe impulse response power,
It is the impulse response phase place.
5. estimate the reception midamble data not influenced by frequency deviation
To each u
kEach bar multipath component, superpose by following formula, thereby estimate the reception midamble data that not influenced by frequency deviation
1≤n≤144
In the formula, u
k(n-t
i) be the user midamble sign indicating number u of k estimating window correspondence
kT moves right
iChip, h
k(t
i) be the impulse response of i bar multipath in k the window of h, t
iIt is this multipath reaches the footpath with respect to head time delay.
6. carry out the maximum likelihood frequency offset estimating, frequency deviation is estimated in output.
At first calculate the related data y that is used for the maximum likelihood frequency offset estimating:
1≤n≤144
In the formula
*Conjugation is got in expression.
Calculate then and make maximum likelihood function
Get peaked estimation frequency deviation
Can obtain approximate maximal possibility estimation frequency deviation by following formula:
N is 144 in the formula,
*Conjugation is got in expression.Need to prove, though what the present invention is directed to is that to send the midamble data be a plurality of u
kThe situation of stack, but it is equally applicable to send the midamble data and only contains a u
kSituation, at this moment θ is 0,
The single u that also only comprises
kEach multipath component.
Referring to Fig. 6.Frequency deviation estimation device of the present invention comprises signal receiving device, channel estimating apparatus, midamble composition checkout gear, signal estimation unit and frequency deviation estimation device.When carrying out frequency offset estimating, signal receiver 1 receives the signal that send the base station, intercepts out midamble data r from receiving data sequence, and sends channel estimator 2 and correlator 9 to.Channel estimator 2 estimates the original channel impulse response h that influenced by frequency deviation
Freq, and send midamble composition detector 7 and multipath detector 3 to.Midamble composition detector 7 detects and sends each user midamble sign indicating number u that data comprise
k, notify multipath detector 3 and signal estimator 8 then.Multipath detector 3 is used to detect each multipath of channel at each u
kChannel estimation window in corresponding impulse response, and the information of each impulse response sent to power estimator 4 and phase estimating device 5, power estimator 4 is used to remove the influence of frequency deviation to the impulse responding power, and phase estimating device 5 is used to remove the influence of frequency deviation to phase place.The power of each impulse response and phase place are sent to channel estimator 6, and channel estimator 6 will calculate the channel impulse response h that not influenced by frequency deviation.Signal estimator 8 estimates the reception midamble data that not influenced by frequency deviation according to the composition of midamble data and the channel impulse response that not influenced by frequency deviation
Be sent to correlator 9, r carries out conjugate multiplication with the actual reception data, obtains related data y.By maximum likelihood frequency offset estimator 10, calculate the maximum likelihood estimator of frequency deviation at last, and frequency deviation is estimated in output
The present invention also can be used for other mobile communication system of using similar training sequence, and the UMTS TDD system as high spreading rate (3.84MHz) can reach the same effect of the present invention.
Embodiments of the invention only are in order to describe the present invention, and are not limitation of the invention.Those of ordinary skills know that any distortion that does not break away from spirit of the present invention all belongs to scope of patent protection of the present invention with variation.