Timing synchronization in a receiver employing orthogonal frequency division multiplexing

Timing synchronization of receiver using quadrature frequency divider multiplexer

The present invention is related to an electromagnetic signal receiver using multi-carrier modulation. More specifically, the present invention relates to the synchronization of receivers using Orthogonal Frequency Division Multiplexing in time intervals, and the time intervals that need to be synchronized are calculated by such receivers in the fast Fourier transform ( Fast Fourier Transform) time required.

Coded Orthogonal Frequency Division Multiplexing (Coded Orthogonal Frequency Division Multiplexing, COFDM technology has been proposed for digital audio and digital video broadcasting, and both of the aforementioned two broadcastings require limited bandwidth for efficient use, and a transmission method that reliably solves many effects. For example, the impulse response of a typical channel can be modeled by the sum of a plurality of Dirac Pulses with different delay lengths. Each pulse corresponds to a multiplication factor, where the amplitude of the pulse usually follows the Rayleigh Law, and such a pulse can be extended for several microseconds, causing uncoded transmission to become unreliable at high transmission rates. In addition to random noise, impulse noise, and attenuation, digital terrestrial transmissions with high transmission rates also include major difficulties such as Multipath Propagation and adjacent channel interference. Among them, adjacent frequencies Interference is a highly correlated signal change between adjacent frequencies. COFDM technology is particularly suitable for applications in the above-mentioned situations. In the actual COFDM arrangement, a relatively small amount of data is modulated onto a large number of carriers, and these carriers are quite close in frequency. The duration of the data symbol is increased by the same multiple as the number of carriers or sub-channels, so that the mutual interference between internal symbols is significantly reduced.

Figure 1 and Figure 2 are diagrams of multiple selection based on COFDM. Line 2 in Figure 1 is the spectrum of a single COFDM carrier or sub-channel. Figure 2 shows a set of carrier frequencies with overlapping waveforms, where orthogonality is satisfied. Two real number functions are orthogonal, if<maths><img file="TW363321B_D0001.tif" /></maths>

Where K is a constant, and if pq, K=0; if p=q, K0. According to the actual encoding and decoding of the COFDM signal, the fast Fourier transform (FFT) is quite dependent, which can be seen from the following equations.

Signal of given carrier c<maths><img file="TW363321B_D0002.tif" /></maths>

Where A_cIs the data at time t, ω_<i>c</i>Is the frequency of the carrier,<img file="TW363321B_D0003.tif" />Is the phase.

N carriers of a given COFDM signal<maths><img file="TW363321B_D0004.tif" /></maths>

Sampling a symbol period, then<maths><img file="TW363321B_D0005.tif" /></maths>

When the sampling frequency is 1/T, the generated signal is expressed as<maths><img file="TW363321B_D0006.tif" /></maths>

When sampling a data symbol period T=NT, and ω₀=0, the signal is<maths><img file="TW363321B_D0007.tif" /></maths>

Compare this formula with the general formula of Inverse Discrete Fourier Transform<maths><img file="TW363321B_D0008.tif" /></maths>

In formula (8)<img file="TW363321B_D0009.tif" />Is the input signal in the sampled frequency domain, and<i>S</i>_<i>s</i>(<i>kT</i>) Is the expression of the signal in the time domain. Increasing the size of the FFT can provide a longer symbol duration and improve the robustness of the system when the response exceeds the guard interval. But at the same time, the computational complexity of the system is also<i>N</i> bg₂<i>N</i>This is a practical limitation.

When the mutual interference between internal symbols occurs due to the relationship of the transmission channel, the orthogonality between the signals cannot be maintained. One solution is to appropriately sacrifice some of the transmitted energy and add a period of time that exceeds the channel-like memory and multipath delay before each symbol in the time domain. The guard interval selected in this way is large enough to absorb any internal inter-symbol interference, and this interval can be established by copying a certain part of the symbol to before the symbol. Such replication is a typical cyclic extension of the end of the symbol. Referring to Figure 3, the data symbol 4 has an active interval 6 which contains all the data transmitted by the symbol. The end part 8 of the action interval 6 is reproduced to the front end of the signal as the protection interval 10. The COFDM signal is represented by a solid line 12. It is also possible to cyclically copy the beginning part of the action interval 6 to the end of the symbol.

The transmission of COFDM data can be accomplished by the conventional general method shown in Figure 4. The serial data string 14 is converted into a series of parallel strings 16 in a serial-to-parallel converter 18. Each parallel string 16 is composed of x bits, which are used to represent a complex number, where x determines the signal constellation of its associated parallel string. After the external coding and interleaving of the block 20, the pilot carrier is inserted in the signal mapper 22 for synchronization and channel estimation at the receiver. There are usually two types of guided waves. The continuous guided waves are transmitted at the same position and the same amplitude in each symbol. At the receiving end, these guided waves are used for phase noise removal, automatic frequency control, and time/sampling synchronization. Scattered pilot carriers are scattered around the symbol, and the location of the scattered pilot waves often varies with the symbol. They are mainly useful for channel estimation. Then these complex numbers are modulated at the fundamental frequency by an inverse Fourier transform at block 24 and inserted into the guard interval at block 26. Then this discrete symbol is converted into an analog signal, usually by filtering out the low-pass (low-pass filtered), and upload and convert to radiofrequency in block 28. The signal is then transmitted through the channel 30 and received by the receiver 32. In the conventional technology, the receiver uses the reverse procedure of the transmission to obtain the transmitted information, and in particular, FFT is used to demodulate the signal.

A new COFDM application was proposed in the European Telecommunication Standard DRAFT (European Telecommunication Standard DRAFT) pr ETS 300 744 in May 1996, which defines the framing structure and channel of digital terrestrial television. Coding, and modulation. The definition of this specification is to enable digital near-earth television to provide moderate prevention of high-level sub-frequency and near-frequency interference under the existing analog transmission spectrum configuration. The flexible protection interval is clearly indicated so that the system can provide a variety of network settings while maintaining the hyperspectral effect and protecting the sub-frequency interference and near-frequency interference from the existing PAL/SECAM services. The European communication standards mentioned above define two modes of operation. One is a 2K mode suitable for single-transmitter operation and a small single-frequency network with limited transmission distance, and the other is an 8K mode that can be used for single-transmitter operation or a large single-frequency network. Quadrature Amplitude Modulation (Quadrature Amplitude Modulation, QAM) are supported, and they each have different internal bit rates to balance the bit transmission rate and robustness of the system. According to the Moving Picture Experts Group (MPEG), this system is pre-applied with a transport layer and is directly compatible with MPEG-2 encoded TV signals (SIO/IEC 13818).

In the aforementioned European communication standards, the data carrier in the COFDM framework can be quadrature phase shift keyed (QPSK), 16-QAM, 64-QAM, and uniform 16-QAM , Or use non-uniform 64-QAM with gray mapping.

When receiving COFDM transmission, an important problem is how to maintain synchronization under phase noise and turbulence. These phase noises and jitters are usually caused by upload conversion before transmission, download conversion at the receiver, or an oscillator (usually a voltage controlled oscillator) at the front end of the regulator. In addition to providing pilot carriers to assist in the synchronization of demodulation, the European communication standards do not specifically regulate synchronization, but are left to the executor to solve it.

There are basically two types of phase disturbance. One is the composition of noise that interferes with adjacent carriers in a multi-carrier system, which we call "foreign noise contribution" (FNC), and the other is the composition of noise that interferes with the self-carrier, which is called "self-contribution". "Own noise contribution" (own noise contribution).

Referring to Figure 5, the position of the ideal quadrant sample set (constellation samples) is marked with an "x" symbol 34. The effect of "external noise influence" is random, producing Guassian-like noise. The sample disturbed in this way is marked by circle 36 in Figure 5. The effect of "self-noise influence" is the common rotation of all quadrant points, as shown by the displacement between each x symbol 34 and the associated circle 36. This is called "common phase error" (common phase error). ). The "in-phase error" differs significantly from symbol to symbol, so every symbol cycle interval T_sIt needs to be recalculated. "In-phase error" can also be interpreted as the symbol period T_sThe average phase error.

In an actual system, in order for the receiver 32 to process data symbols, some mathematical processing is required for the complex signals representing the data symbols. Usually FFT is used. If you want to get the correct result, you need a special type of timing synchronization to map the interval between the FFT and the received data.

One of the main objects of the present invention is to provide an improved method for synchronizing the received data with the FFT window of the COFDM transmission signal.

Another object of the present invention is to improve the near-ground transmission of COFDM signals.

Another object of the present invention is to provide better COFDM signal receiving equipment.

These and other objects of the present invention are all based on the symbol period as T_sThe data symbol string is decoded to achieve this. Each symbol defines an active interval (active interval) and a protection interval, where the protection interval is made from a part of the interval for remaking, and the protection interval tends to be made by the cyclic extension of the active interval. The data symbol string is in a data symbol period T_sThe long evaluation phase is sampled. The samples are paired in pairs, and the samples in the same pair are separated by the active area in the data string. The paired data fall in the first and second intervals, and the value of each interval is not greater than the protection interval. To determine a function for each pair of data, it is better to take the difference of two constituent samples and use it as the first output signal. Then, the distribution of the first output signal is measured and used as the second output signal. Next, with respect to the data string, the first interval and the second interval are placed together, and the aforementioned paired sampling, determination function, and distribution measurement steps are repeated. The processed result is regarded as the third output signal. Then the second output signal is compared with the third output signal, and the result is used as the fourth output signal. Then a rule determined in advance for the fourth output signal is detected. The rule indicates the boundary of the protection zone and the action zone. This pre-determined rule can be a pre-determined value, a maximum value, or a minimum value.

It is best to calculate the modulus of the sample set. According to one aspect of the present invention, only the real or imaginary part of the data string needs to be sampled.

F ratio (F ratio) is used to compare the second and third output signals of two comparison intervals, and the F ratio is usually evaluated by the logarithmic difference of the numerator and denominator of the F ratio. The F ratio of the peak of the subsequent comparison interval is also evaluated, and the peak of the comparison interval indicates the boundary of the protection interval. The determination of the crest may include a test of statistical validity.

In the first operation mode, the evaluation interval is a first fixed value, and in the second operation mode, which allows the adaptability of the signal conditions to be changed, the evaluation interval is the second value. According to a certain part of the present invention, the position distribution of the protection interval of the consecutive symbols of the second operation mode will be memorized, and the value of the evaluation interval will be adjusted according to the memorized distribution.

According to another part of the present invention, after the boundary of the guard interval is detected, a new set of samples are then sampled and received in an active interval, and it is completed before the boundary detection of the next symbol starts.

The present invention provides a demodulator for a frequency division multiplexed signal (frequency division multiplexed signal), wherein the signal has a symbol period of T_sThe data string constituted by the data symbols, and these symbols each define an action interval and a protection interval. The protection interval is a partial copy of the active interval, and the loop extension at the end of the interval is preferred. A timing synchronization circuit is provided in the demodulator to detect the boundary of the protection interval. This circuit has a first delay circuit (delay circuit), usually designed as a first-in-first-out memory (first-in-first-out, FIFO), and is connected to the input signal S_i, And this circuit should be able to store L samples, where L is the size of the FFT window frame, so that the input signal S_iDelay the interval of the FFT window frame size. Then the input signal is subtracted from the delayed version of the input signal, and then the modulus of the input signal or difference signal is calculated. This circuit also has a second delay circuit to delay the aforementioned modulus by a first predetermined interval. An adder/subtractor subtracts the delay modulus from the aforementioned modulus and adds a feedback signal from the register, and the register accumulates the result of the adder/subtractor. A lookup memory addressed by the output of the adder/subtractor stores the logarithmic value of the result of the adder/subtractor. The third delay circuit provides a delay amount of the second predetermined interval and has an input connected to the reference memory. A second subtractor has a first input connected to the reference memory; and a second input connected to the output of the aforementioned second delay circuit. A comparison circuit is connected to the output of the subtractor and compared with a threshold value.

A part of the present invention has a fourth delay circuit, which is connected to the aforementioned input signal S_i, This circuit can store less than L samples and provide a delay to the action interval. A selector is used to select the first delay circuit or the fourth delay circuit.

According to a further part of the present invention, there is a module for calculating the modulus of the input signal connected to the first delay circuit and the adder/subtractor.

According to another part of the present invention, a control circuit selects predetermined samples of the input signal for processing by the timing synchronization circuit, which allows the evaluation interval to be selected according to the data string symbol.

In order to have a better understanding of the objectives and other objectives of the present invention, detailed reference materials related to the present invention will be described below with the accompanying drawings as an example, in which: Figure 1 explains the frequency spectrum of a COFDM secondary frequency.

Figure 2 shows the frequency spectrum of a multi-carrier COFDM signal.

Figure 3 is a signal diagram based on COFDM and shows the format of data symbols.

Figure 4 is a block diagram of a COFDM system based on FFT.

Figure 5 shows some of the interference in the COFDM signal quadrant.

Fig. 6 is a flowchart of a timing synchronization method according to a preferred embodiment of the present invention.

Figure 7 is a graph of the F ratio test under several data symbols for rough timing synchronization.

Figure 8 is a graph of the incomplete beta function under different degrees of freedom.

Figure 9 is a graph that helps to understand the statistical validity according to the present invention.

Fig. 10 is an electrical schematic diagram according to another embodiment of the present invention.

Fig. 11 is an electrical schematic diagram according to another embodiment of the present invention.

[Detailed Description of the Invention]

Referring again to FIGS. 3 and 4, according to the present invention, a statistical method is applied to the COFDM signal to find the end of the guard interval 10. This method will be explained by referring to the European communication standards mentioned above, but this method is suitable for multiplexing in any frequency interval with pre- and post-protection intervals. This method allows the receiver 32 to define the end of the protection interval under the condition that only the received sampled complex signal (solid line 12) and the size of the active interval 6 are given. This method relies on the reproduction of the guard interval 10 being the end part of the data symbol 4. In the receiver 32, the end of the guard interval 10 and the data symbol 4 will be different due to the channel noise and response and the error of the area oscillator. If the introduced errors are random, statistical methods can be used. According to the present invention, the received complex signal is sampled at almost the same frequency as the transmitting frequency of the transmitter. A difference signal is found from a pair of received samples separated by a time period as close as possible to the action zone 6. This period of time should be the same as the size of the FFT used (for example, 2048 or 8192 samples). make<i>S</i>_<i>i</i>=∣<i>s</i>_<i>i</i>∣-∣<i>s</i>_<i>i</i>-_{<i>fftsize</i>}｜　　　(10)

Where S_iIs the difference signal, s_iWith s_i-_fftsizeIt is the current and previous input signal, and the modulus is taken. The subscript i indicates the linear timing of the input value. Assuming that the input signal is random, then S_iIt will also be random. In the protection interval, s_iWith s_i-_fftsizeAlthough they are not the same, they are similar. The difference between the two is caused by the channel effect. So S_iWill be a random signal in a small distribution interval. In this article, the term "dispersion" refers only to the distribution of values, without any specific mathematical definition. Generally, the range of action of one symbol is not related to the range of action of the next symbol. Outside the protection zone, S_iWill be randomly scattered in a larger distribution interval. In order to find the end of the protection interval, we monitor the difference signal S_iThe distribution tries to find an effective upward part, and this upward part usually occurs at the boundary of the protection zone 10 and the action zone 6. The inventor also found that a huge decrease in the distribution interval can be observed at the beginning of the protection interval 10.

According to a preferred embodiment of the present invention, samples of the input signal are stored for more than at least one symbol period T_s. Difference signal S_iThe distribution of is calculated based on the samples in a block, this block is then moved back in a number of samples, n, and its distribution is recalculated. These two blocks are called "comparison blocks" here. The ratio of the current distribution of the first comparison block to the distribution of the previous comparison block is calculated. Then, the F-ratio effective test is used to determine the significant difference in the distribution of the two comparison blocks. Here the F ratio is defined as<maths><img file="TW363321B_D0010.tif" /></maths>

in<i>n</i>Is a positive integer,<i>i</i>Is the index of the input sample,<i>VAR</i>(<i>i</i>) Is the variance of the block value of the length N samples. The variance can be defined as<maths><img file="TW363321B_D0011.tif" /></maths>

Although the F-ratio effective test is used in this preferred embodiment, other functions that can give the change signal of the two distribution values can also be used. There are many such functions. One advantage of F ratio is that for random input signals, it has a known probability distribution, which is convenient for statistical performance analysis and system design. At the same time, the F ratio essentially normalizes the signal, making the result independent of the signal level.

This method is disclosed in Figure 6, where step 38 measures the first sample of the sample pair in the current evaluation interval. A time delay of action interval 6 (Figure 3) is executed in step 40. This delay can be achieved by using a digital delay, such as FIFO, or equivalently buffering a sample with a long active interval in the memory and accessing the appropriate unit of the memory. The second sample of the sample pair is measured in step 42, and the difference between the two samples is determined and stored in step 44. In the decision step 46, the end of the current block is tested. The size of the evaluation interval should not exceed the size of the protection interval, and it should be much smaller. If the end of the current block has not been reached, another sample is obtained in step 48, and the control flow is returned to step 38.

If the end of the current block has been reached, the distribution of the current block is measured in step 50 and regarded as one of the two comparison blocks of data. The test in decision step 52 determines whether a set of two comparison blocks has been evaluated. If the test is negative, the data of another block is obtained in step 54 and then the flow control returns to step 38. The other blocks of data do not need to be continuous with the block just calculated.

If the test result in step 52 is affirmative, in step 56 the F ratio of the set of two comparison blocks is calculated. The result of step 56 is sent to step 60 for peak detection. Peak detection can optionally include a statistical test of validity, as explained below.

If the peak is detected, the boundary of the protection interval can be established in step 62. The boundary of the protection interval is established for the synchronization of the FFT window frame required for further signal reconstruction. If the peak value is not detected, another part of the data string is taken out of the same block to repeat the above process.

Example 1:

Now please refer to Figure 7. A random number generator is used to generate a complex signal in accordance with the aforementioned European communication standard, and the signal is processed through a Ricean channel model and added white Guassian noise (SNR). =3.7) are transmitted together. Then use the methods mentioned above to analyze the data symbols. The six result symbols are shown in Figure 7. For the convenience of presentation, the F ratio is represented by the line 64 on the logarithmic coordinate. The reason for this is that the spikes 66, 68 at the beginning and the end of the protection interval The value is very large.

Although it is obvious from Figure 7 that there are many conventional peak detection methods that can find the end of the protection interval, statistical methods can answer the question more correctly: whether the samples of the two blocks have the same distribution . H₀It is the initial assumption that the distribution is the same and the observed peaks of F are only caused by random changes. If H₀If the probability is very small, it can be rejected. This phenomenon corresponds to the detection of the beginning or end of the protection interval. From the perspective of the way the COFDM symbols are constructed, the comparison block is completely within the protection interval or the active interval.₀Should be true, but if the comparison block crosses the start or end boundary of the protection interval, then H₀Should be false (false). If the comparison block of a random sample is taken from the same parent body, then the probability of F is<maths><img file="TW363321B_D0012.tif" /></maths>

one of them<i>I</i>() is an incomplete beta function,<maths><img file="TW363321B_D0013.tif" /></maths>

v₁And v₂It is used to estimate the number of degrees of freedom for the first and second distributions. In this example, if n>=N, then v1=v2=(N-1). The graph of this function is shown in Figure 8. From a statistical point of view, n should be large enough so that there is no overlap between the two blocks, that is, n>=N. If the two blocks overlap, the distribution of the second block will be calculated using the sample when calculating the distribution of the first block. This will reduce the degree of freedom and reduce the effectiveness of the result. It is already known that the result when n=N is acceptable.

The function Q() in equation (13) actually only gives the probability of one-tailed. When the value of F is very large or very small, H₀Will be rejected, so a two-tailed test is required. But in fact the probability of both sides is the same, so the probability of both sides should be twice the probability given by equation (13). But this will cause the probability value obtained when F&lt;1 to be greater than 1. Therefore, the probability p should be calculated as follows:<maths><img file="TW363321B_D0014.tif" /></maths>

And when p>1, p=2-p. This formula reflects H₀Feasibility. So if p is quite small, H₀Can be rejected, and with a certain degree of confidence, explain this phenomenon as the two comparison blocks come from two different sample spaces. The European communication standard stipulates that the block size N of the correlation algorithm should be 32. And N={32,64} has been successfully tested and feasible. Please refer to Figure 9 for the probability function graph obtained by using these values of N. In a preferred embodiment, H₀The rejection is set at p<=0.05.

An accurate implementation procedure of the present invention is to calculate F, then x, then an incomplete beta function, then p, and finally do a threshold test. This algorithm is not easy to implement in hardware because the beta function is very complicated. In the preferred embodiment, it is much simpler to provide the same result. The acceptance threshold and the parameter N are set, and the upper and lower limits of F are also given. Then we only need to calculate the value of F and compare it with the upper and lower limits. In order to easily determine the end of the protection interval, it should be safe to set F>1. Only the upper limit of F is required. In order to correctly calculate the limit value of F, root finding methods such as Newton-Raphson can be used. Typical values are given in Table 1.

<tables><img file="TW363321B_D0015.tif" /></tables>

This method has been successfully tested in the designated channel mode with white Gaussian noise (SNR=3.7).

The distribution equation in equation (12) requires a multiplier when fabricated with silicon semiconductors. The calculation of the F value is a division, in which the (N-1) normalization constant can be removed when the size of the two blocks is the same. In silicon circuits, the use of precise multiplication and division is quite expensive, so in a preferred embodiment, a simplified version that can be implemented is provided. This simplified version will provide a less precise but still usable F value. It can be assumed that S_iThe average value of is zero, so there is no need to calculate the average value from the samples in the block. Doing so also increases the degree of freedom from (N-1) to N. Rather than using the standard square sum formula to calculate the variance, it is better to use the absolute value of the mean error to estimate the distribution. In this case, the formula of VAR(i) becomes<maths><img file="TW363321B_D0016.tif" /></maths>

If the size of the two blocks is the same, (1/N) can be removed in the calculation of F. Up to this point, there is still a need for division and squaring between the two distributions. This can be solved with a logarithm of base 2. Substituting equation (16) into equation (11) to get<maths><img file="TW363321B_D0017.tif" /></maths>

Take the base 2 logarithm to get log<i>F</i>=2(log<i>s</i>_<i>a</i>-log<i>s</i>_<i>b</i>)=<i>y</i>　　　<b>(18)</b>

In this way, you only need to calculate y and compare the upper limit of y with the logarithmic value of F with the base 2 as the logarithmic value. This comparison can be achieved by subtracting the logarithm of the upper limit of F to the base 2 with 2 (log2sa-log2sb), and comparing the result with 0. The constant multiple of 2 can be moved to the upper and lower limits.

If the number is stored in fixed point fractions, the base 2 logarithm calculation is quite straightforward in hardware. Decimals can be divided into exponent and mantissa parts: x=A2^B. Take 2 as the base logarithm to get logx=logA+B. Since A is a decimal number, the logarithm can actually be found by looking up the table, and the exponent B can be known by looking up the position of the most significant bit (MSB) (because s_aAnd s_bAll are positive numbers).

The required calculations can be simplified in this way to only need to use addition and subtraction arithmetic operations. When using this method, the limit value also needs to be recalculated with v1=v2=N. In practice, the significance level of a particular application can be given empirically, preferably p=0.05.

Those with experience in this technical field should be able to understand that various estimates of distribution can be used without departing from the spirit of the present invention, such as standard deviation, skew, various moments, and bar statistics. Histograms or other calculations well known in this technical field.

In the first alternative embodiment of the present invention, the above-mentioned methods are used, and these methods all use the real or imaginary part of the signal instead of modulus. This embodiment achieves hardware economy.

In the second alternative embodiment of the present invention, we optimize the n parameter of equation (11). At the end of the protection interval, the two blocks extend the conversion to the active interval, resulting in a defined increase in the distribution. Setting any n>2 will have the disadvantage of a significant increase in the value of several consecutive points when the following block crosses the boundary. This small problem can be solved by introducing a dead period after the boundary is measured. In other words, when a peak-shaped protrusion is detected, a group of samples with the same number as the FFT window frame size is received first, and no further attempts to identify the position of another peak-shaped protrusion are made. This failure interval has the additional benefit of not introducing false peaks. When a larger value of n is used, peak-like protrusions 66, 68 will increase (Figure 7), but H₀The noise F signal remains the same.

Example 2:

The maximum F-Spike height as a function of n has been systematically combined with the background variations of F as the quantitative side. The results are listed in Table 2.

<tables><img file="TW363321B_D0018.tif" /></tables>

Table 2 is developed using the first five frames of the signal analyzed in Figure 7. The statistics in the columns (2) and (3) of Table 2 are made by removing all the points with F>=3.0, so as to exclude the peak-shaped protrusions from the calculation. Otherwise, these peaks will affect the calculated mean and standard deviation, even if these peaks come from different statistical spaces.

The result shows that F's background variation (variation), F_sdAffected by n, it increases to about 0.28. This may be caused by overlapping blocks. For example, when N=64 and n<64, some values of the blocks used to calculate the distribution are the same, so the blocks will be related. In order to test this theory, estimate F_sdWhen n>N, the results are shown in Table 3

<tables><img file="TW363321B_D0019.tif" /></tables><tables><img file="TW363321B_D0020.tif" /></tables>

When n>=N/2, the correlation becomes linear. If F is calculated from every sample to once every n samples, this correlation will decrease. But doing so is risky for a small protection interval. There will be no first block that falls into the protection interval, and the second block is completely within the active interval.

A third alternative embodiment of the present invention is disclosed in FIG. 10, in which a schematic diagram shows a time synchronization circuit 70. This circuit receives a complex input signal 72 and includes a circuit module 74 whose modulus is taken from the input of the node 83. The circuit module 74 confirms that the value to be processed later is a number without positive or negative. The input of the circuit module 74 is the difference signal generated by the subtractor 75. The input of this subtractor is the input signal 72 and the delayed version of the signal 72 processed by the delay circuit 79. We suggest to make this delay circuit with a FIFO 77 of length L, and the L is the window frame size of the FFT. According to the foregoing, this circuit can also be operated when the input signal 72 is a real number, an imaginary number, a complex number, or even the absolute value of a complex number. When the input signal 72 is a real number or an imaginary number, the circuit module 74 can be changed. The circuit can be any conventional circuit that removes the sign of the output of the subtractor 75, or the sign of the output can be set to the same circuit, so that it can monotonically accumulate the output, that is, it has Unipolar output. The output of the circuit module 74 is exhaustively clocked into a digital delay, and the FIFO 78 is best implemented here. When the FIFO 78 is full, a SIG1 80 signal is displayed, and the FIFO The output of 78 can be used as indicated by the AND gate 82. An adder-subtractor circuit 84 is also connected to the node 76, and its output is stored in the register 86. The delayed version of the output of the adder-subtractor circuit 84 can be obtained by the register 86 and fed back at the line 88 as the second input of the adder-subtractor circuit 84. If SIG1 80 is instructed, an output version of circuit module 74 is subtracted from the signal at node 76. Here, the output of the circuit module 74 is delayed by a predetermined interval N, where N is the number of samples in the comparison block.

The signal on line 88 is an index of a lookup table, preferably implemented in read-only-memory (ROM), as shown in ROM 90. The address of ROM 90 contains the base 2 logarithm of the signal on line 88, and this signal will then appear at node 92. Node 92 is connected to subtractor 94 and a delay circuit (represented by FIFO 98), and is used to calculate the denominator of the middle term of equation (17).

The subtractor 92 produces an output, which is used to compare with a predetermined threshold F_LIMITThe logarithmic value of 2 is the base for comparison in the comparison circuit 106. For the sake of simplification, an adder 108 is connected to the comparator 110 to represent it here. When the boundary of the protection zone is detected, the output signal SYNC 12 is displayed.

Although it cannot be implemented in the current preferred embodiment, it is feasible to dynamically set the size of the FIFO 77, so that the size of the measured interval can be adjusted according to the mode of operation. For the calculation of its distribution, this can be done simply by storing the value of the node 92 in the RAM 114.

Referring to FIG. 11, in the fourth alternative embodiment of the present invention, components similar to those in the embodiment shown in FIG. 10 have the same reference numerals. The timing synchronization circuit 116 is similar to the timing synchronization circuit 70, except that the delay circuit 79 is changed to a FIFO 77 and another FIFO 100, and the multiplexer 102 is used to select one. FIFO 77 and 100 have the same delay, but their storage sizes are different. The FIFO 100 provides storage space for the number of samples of the FFT window frame size, and is usually selected in the first mode of operation. For example, when a channel is to be acquired, the entire symbol needs to be calculated in order to locate the boundary of the guard interval. In European communication standards, at least 8K of data storage is required, as well as a commensurate resource requirement. In the next action, the approximate position of the boundary of the protection interval can be known from the historical record of the previous symbol. Therefore, in the second operation mode, only a much smaller interval needs to be checked, in order to check the correct position of the boundary of the protection interval. The number of samples required to calculate the distribution can be kept small, preferably 32 or 64, so a much smaller FIFO 77 can be used to store the calculated value. The resources saved here can be used for other functions of the demodulator, and the memory used by the larger FIFO 100 can also be used for other purposes.

The control block 81 selectively moves the evaluation interval forward, and the previous movement is relative to the symbol boundary of the consecutive symbol data string. In addition, the block 81 can also be used to delay the dead period. The last moving evaluation interval will cross the boundary of the current symbol protection interval, and the synchronization can be determined at this time. Choose the size of the evaluation interval to minimize the use of memory, but still maintain a sufficient size to achieve the validity of the evaluation interval statistics. The size of the evaluation interval and the FIFO 77 can be configured dynamically or statically.

The structure of the present invention is disclosed based on the above description, but the present invention is not limited to the details described in the invention, and the application of the present invention also includes various modifications and changes made under the spirit of the scope of the following patent applications.