The present invention relates to a method for time synchronization tracking in broadband transmission systems. More particular, the invention relates to an algorithm for post-DFT time synchronization tracking in OFDM receivers. The present invention further relates to a system for performing a time synchronization tracking method within an OFDM receiver.
Orthogonal frequency division multiplexing (OFDM) has become a popular transmission method for high-speed wireless radio transmission, due to its potential for low complexity of transmitters and receivers, paired with robustness under severe multipath conditions. A more detailed discussion on OFDM is found in S. B. Weinstein and P. M. Ebert: Data transmission by frequency-division multiplexing using the discrete Fourier transform. IEEE Trans. Communication Technology, COM-19(5):628-634, October 1971.
In OFDM, a large number of closely-spaced orthogonal subcarriers are used to carry data. Each subcarrier is modulated with a linear modulation scheme (such as quadrature amplitude modulation (QAM) or phase shift keying) at a low symbol rate.
The orthogonality of the OFDM subcarriers allows for efficient modulator and demodulator implementation using inverse discrete Fourier transformation (IDFT) on the transmitter side for conversion of the signal into the time domain, and DFT on the receiver side for conversion back into the frequency domain.
Continuous reception of OFDM signals, such as in a receiver for the European digital terrestrial video broadcasting standard (DVB-T; Digital Video Broadcasting-Terrestrial; ETSI EN 300744, V1.5.1: Digital Video Broadcasting (DVB); “Framing Structure, channel coding and modulation for digital terrestrial television”, European Standard, European Telecommunications Standards Institute, 2004) requires continuous adaptation of the receiver sample time synchronization with respect to the transmitter sample timing which is referred to as time tracking, in order to prevent interference between subsequent OFDM symbols (inter-symbol interference—ISI) as well as inter-carrier interference (ICI) within individual OFDM symbols.
To avoid inter-symbol interference (ISI) in multipath fading channels, a guard interval is inserted prior to the IDFT block. During this interval, a cyclic prefix is transmitted which consists of the end of the IDFT output copied into the guard interval. If there is no multipath propagation, the receiver can select the time synchronization within a window that is the size of the cyclic prefix.
In multipath propagation environments, a transmitted signal reaches the receiver through multiple paths each of which may introduce a different delay, magnitude and phase thereby enlarging the transition time from one symbol to the next. Identifying the useful part of an OFDM symbol that contains minimum interference from adjacent symbols (inter-symbol interference) is a time synchronization task to be performed by the receiver. This task is critical to the overall receiver performance.
Time synchronization may be classified into two main categories: acquisition and tracking. Symbol time acquisition defines the task of initially finding the correct timing. Often, the symbol time acquisition is divided into two or more steps, where in the first step, coarse time synchronization is achieved. In the following steps, the time window is refined. For those successive steps, similar or identical algorithms that are used for tracking are often applied. Tracking defines the task of continuously adjusting the time window in the course of continuous reception to keep the time window at its optimum location.
Time tracking is crucial for the overall system performance. For OFDM, various methods for time tracking have been proposed. The known methods may be grouped into data assisted and non-data assisted tracking, and pre-DFT or post-DFT time tracking. Data assisted tracking makes use of known symbols in OFDM, e.g. reference symbols, also known as pilot symbols, or preambles, whereas non-data assisted tracking makes use of the correlation properties of the signal.
In DVB-T which is aimed at continuous reception, the standard does not define any preambles. Reference symbols are included in the multiplex, the standard defining so-called scattered pilots at every 12th carrier, and a smaller number of continual pilots that are present at fixed carrier locations.
Those pilot symbols are only accessible after DFT and only after some coarse time synchronization has already been established. Therefore, most initial time synchronization algorithms for DVB-T/H use the auto-correlation properties of the OFDM symbols with its cyclic extension for coarse symbol time estimation and then rely on the pilots for fine time synchronization and tracking.
Some pre-DFT, time-domain based, time tracking techniques that make use of the auto-correlation properties have been found to require relatively long averaging times to yield adequate results. Another disadvantage is that after the signal has been acquired those types of calculations are not required elsewhere in the receiver. Additionally, the performance under heavy multipath is not satisfying. Other known approaches aim to further improve the time domain correlation based method typically used for coarse time synchronization.
Two basic approaches for post-DFT based time tracking are known both using an estimate of the channel transfer function:
In a first approach, the estimated channel transfer function is transferred back into the time domain by means of an IDFT to obtain an estimate of the channel impulse response from the estimated channel transfer function. Afterwards an energy search is performed on the estimated channel impulse response. This method, however, is computationally intensive and requires additional memory.
An alternative approach for post-DFT based time tracking, therefore, is to calculate the average phase difference from one scattered pilot to the next thereby estimating the mean slope of the channel transfer function. This is based on the property of the DFT that a delay in time domain corresponds to a phase proportional to the carrier index and proportional to the delay in time domain. Therefore, in single paths channels, the time delay may be directly estimated from the slope. Such algorithms is described, e.g. by Young-Jae Ryu, Dong-Seog Han in “Timing phase estimator overcoming Rayleigh fading for OFDM systems”, IEEE Trans. Consumer Electronics, vol. 47, issue 3. August 2001, pp. 370-377, and by Hou-Shin Chen, Yumin Lee, in “Novel sampling clock offset estimation for DVB-T OFDM”, Proc. IEEE VTC 2003-Fall, vol. 4, pp. 2272-2276.
This simple method using the estimate of the mean value of the slope of the channel transfer function, while giving satisfactory results in channels with low delay spread, has been found to give no adequate results under heavy multipath conditions as can be experienced in single frequency networks (SFNs). Experiments have shown that this method does not withstand tests for guard interval utilization in SFNs. One reason for this is that the simple pilot phase slope based estimators target at moving the center of gravity of the channel impulse response to a certain position. However, if the channel impulse response is longer than half the guard interval, the simple method will essentially result in pulling the strong path into the middle of the guard interval, while the smaller path moves out of the guard window, causing inter-symbol interference.
An object of the invention is to provide an improved post-DFT algorithm for estimating a time tracking error in OFDM receivers such as to allow an effective time tracking even in single frequency networks. A more particular object of the invention is to devise a time tracking algorithm for OFDM receivers that supports longer channel impulse responses without causing inter-symbol interference.
This object is achieved by an algorithm as claimed in claim 1 and by a circuit arrangement as claimed in claim 13 or claim 14.
The reference-symbol based time synchronization tracking method for orthogonal frequency division multiplexing (OFDM) receivers according to the present invention operates on DFT output vectors and computes an indication of the direction where the OFDM symbol timing should be shifted to, and the OFDM symbol timing is shifted by enlarging or decreasing the number of samples which are removed between DFT input bocks, i.e. the block length of the guard interval removal is enlarged or reduced, depending on the shift direction. Alternatively to perform a discrete time shift to adapt the timing, the ADC clock may be controlled or the conversion ratio of a sample rate converter may be controlled to change the timing.
A plurality of frequency correlations is determined at equidistant frequency positions for each of a number of subsequent OFDM symbols, and the obtained frequency correlations are linearly combined to obtain the time synchronization offset indication.
The method according to the invention can be viewed as an enhancement of the simple pilot phase slope based estimators of Young-Jae Ryu, Dong-Seog Han, and Hou-Shin Chen, Yumin Lee mentioned above. In contrast to these slope based estimators the estimator according to the invention instead of aiming to move the center of gravity of the channel impulse response to a certain position targets to keep the channel impulse energy within certain time limits. The method according to the invention is beneficial in particular for long channel impulse responses with much energy on one end and little energy on the other end. The method of the invention advantageously supports longer channel impulse responses without causing inter-symbol interference.
Additional features and advantages of the present invention will be apparent from the following detailed description of specific embodiments which is given by way of example only and in which reference will be made to the accompanying drawings, wherein:
In the drawings the same or equivalent components are designated with equivalent reference numerals with the last two digits thereof being the same for similar or equivalent elements.
Channel transfer function estimates at the described positions are then used to compute an auto correlation function in frequency direction (163), at a specific set of small frequency offsets; only the imaginary parts of these correlations are needed. A weighted sum of the obtained set of correlation values is computed (164), resulting in a scalar value. This value indicates where the timing should be shifted to. Multiple averages in time direction are computed, and the sign of the value indicates the direction of the next time shift operation to be performed. In the simplest form of the algorithm, the amount of the time shift is constant, i.e. constitutes a preset value, but it is also possible to exploit the absolute value of the indicator output to control the amount of the time shift.
The weighting coefficients effectively define a high pass filter whose frequency response is shown in
The zero region determines the time range where the channel impulse response is allowed to have energy. If some energy of the channel impulse response falls beyond the flat region, as is illustrated in
The above considerations may be expressed in mathematical form as follows:
Let F(·) denote the Fourier transform and F−1(·) the inverse Fourier transform.
The channel impulse response h(t) is the inverse Fourier transform of the channel transfer function H(f):
h(t)=F−1(H(f))H(f)=F(h(t)).
Thus, the squared channel impulse response is the inverse Fourier transform of the autocorrelation function of the channel transfer function:
|h(t)|2=h(t)·h*(t)=F−1(H(f)*H*(−f)).
An integration over the weighted squared channel impulse response corresponds, in the frequency domain, to an integration of the weighted autocorrelation function (ACF) of the channel transfer function,
∫h(t)·h*(t)·w(t)dt=∫[H(f)*H*(−f)]W(f)df,
wherein the weighting applied to this autocorrelation function is the Fourier transform of the weighting applied to the impulse response:
W(f)=F(w(t))w(t)=F−1(W(f)).
The weighting applied to the squared impulse response, w(t), has a shape as shown in
Evaluating the ACF of the channel transfer function over this weighting function W(f) yields an indication of the direction where the timing should be shifted to in order to have all energy of the channel impulse response located in the zero portion of the weighting filter response, as may be seen from
I=∫[H(f)*H*(−f)]·W(f)df.
The sign of the value I indicates the direction where to shift the timing.
Given the above considerations an embodiment of the inventive method will be described in detail with reference to
c(l,k): transmitted symbols before IDFT
y(l,k): received symbols at DFT output
l: OFDM symbol index in time direction,
k: OFDM subcarrier index in frequency direction.
H(l,k): estimated channel transfer function values.
With the pilot symbols extracted, direct channel estimations are computed (162) at the pilot positions wherein the applicable pairs (l,k) are determined by the pilot pattern as shown in
H
direct(l,k)=y(l,k)/c(l,k).
Further, time-interpolated channel estimates are computed wherein the applicable pairs (l,k) are determined by the pilot pattern as shown in
with O being the set of offset positions in time direction where direct channel estimates are available and an being a set of interpolation coefficients.
The combined set of direct and time-interpolated channel estimates, H(l,kchannel), within the OFDM symbol has to be chosen such that exploited channel estimates are equidistantly spaced in frequency direction (all integers):
k
channel
=k
0
+n·K
nε{n
1
, n
1+1, n1+2, . . . , n2},
with K being the constant spacing between two adjacent frequency positions where a channel estimate is available, and k0 being a constant offset. Preferably, K is chosen to be smaller than a ratio of the DFT length to the guard interval length.
Subsequently, frequency correlations are computed (163) for OFDM symbol l,
for a small number of values, nε{1, 2, . . . nc}.
A sum of weighted frequency correlations is calculated (164),
with W(n) being the weighting pattern with a frequency response as shown in
Steps 162 to 164 are repeated for a plurality of pilot symbols, and averaging of the calculated sums in time direction yields an indicator
from where a sign may be extract that gives the direction of the timing shift to be performed.
It is also possible to reverse the steps of weighting and averaging, i.e. to first calculate an average for each of a set of frequency correlations obtained for a number of subsequent pilot symbols prior to determine a weighted sum from the averaged frequency correlations to obtain the indicator. This process of weighting and averaging effectively constitutes a linear combination operation for the obtained frequency correlations, wherein only the non-zero weights are used.
In another embodiment of the method according to the invention a part of the time tracking unit described in conjunction with
In
The arrangement of
The arrangement of
Number | Date | Country | Kind |
---|---|---|---|
08100784.1 | Jan 2008 | EP | regional |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/IB09/50096 | 1/12/2009 | WO | 00 | 7/15/2010 |