Low-Complexity Dispersion Estimation Method Applicable to Multiple Oversampling Factors

Abstract

The present invention relates to a low-complexity dispersion estimation method applicable to multiple oversampling factors, including: collecting an oversampled time-domain signal, processing the time-domain signal to obtain a frequency-domain signal sequence; constructing a low-complexity dispersion estimation model, where the low-complexity dispersion estimation model includes: a timing phase error detection model, a sequence zero-padding model, and a dispersion estimation model; inputting the frequency-domain signal sequence into the low-complexity dispersion estimation model to obtain an estimation result. The present invention is applicable to multiple sampling rates where the oversampling factor is greater than 1 and less than or equal to 2, with flexible application. It can still normally estimate dispersion at a low sampling rate, greatly reducing the power consumption and cost of the ADC.

Description

Technical Field

[0001] The present invention relates to the technical field of fiber optic communication coherent detection and digital signal processing, and particularly to a low-complexity dispersion estimation method applicable to multiple oversampling factors. Background Art

[0002] Coherent detection has become an important technology for long-distance fiber optic communication systems due to its advantages of high transmission rate, high sensitivity, and the ability to combine digital signal processing (DSP) to compensate for channel impairments. In addition, with the explosive growth of data center capacity, the traditional intensity modulation and direct detection (IMDD) technology is difficult to continue maintaining its high cost-performance advantage, and fiber optic communication technology based on coherent detection has become a strong candidate for short-distance interconnection in the next-generation data centers. Currently, the main factors restricting the further application of coherent detection are power consumption and cost. Whether for long-distance or short-distance applications, in application-specific integrated circuits (ASICs) using 14nm technology, analog-to-digital converters (ADCs) account for approximately 20% of the power consumption, and the higher the sampling ratio, the greater the power consumption and manufacturing cost of the ADC. Therefore, reducing the sampling ratio of the ADC can effectively reduce the power consumption and cost of high-speed fiber optic communication systems. In addition, the computational complexity of DSP algorithms also affects the power consumption of ASIC chips. If the computational complexity is too large, it will occupy a high amount of hardware resources and consume a large amount of power. Therefore, reducing the computational complexity of DSP is also very important.

[0003] As the transmission rate continues to increase, the degradation of the transmission performance of high-speed fiber optic communication systems caused by dispersion becomes more and more serious. At the receiving end of a high-speed fiber optic communication system based on coherent detection, in order to synchronize the two clocks at the transmitting end and the receiving end, a clock recovery (CR) module must be used to ensure that the receiving end samples at the optimal time point. However, when there is large dispersion, the performance of the CR module deteriorates significantly. For a Nyquist signal with a baud rate of 61 GBaud and a roll-off factor (ROF) of 0.1, the dispersion tolerance of the commonly used clock recovery algorithms based on two timing phase error detectors (TPEDs), Gardner and Godard, is about 0.48 ns / nm. Therefore, in a high-speed fiber optic communication system, it is necessary to use a relatively accurate dispersion compensation algorithm in front of the CR module to ensure that the subsequent residual dispersion is within the tolerance of the CR module. Placing a frequency-domain static dispersion compensation module in front of the CR module can compensate for a large amount of dispersion. However, the static dispersion compensation module is a blind algorithm, and directly using it will generate large errors. Therefore, the commonly used method is to estimate the dispersion magnitude using Godard's TPED and then transfer the estimated value to the static dispersion compensation module to achieve large dispersion compensation. Since Godard works in the frequency domain, the introduced additional computational complexity is low. However, Godard's TPED requires the oversampling factor (OSF) to be 2, so the power consumption and cost of the ADC are high. Moreover, the summation terms in Godard's calculation formula include irrelevant noise terms, resulting in poor noise resistance, a large block length for each fast Fourier transform (FFT), and high computational complexity. Therefore, the power consumption and cost of the entire high-speed fiber optic communication system are high. Summary of the Invention

[0004] The object of the present invention is to propose a low-complexity dispersion estimation method applicable to various oversampling factors to solve the problem of high power consumption and cost of the dispersion estimation method in current high-speed fiber optic transmission systems.

[0005] To achieve the above object, the present invention provides the following solution:

[0006] A low-complexity dispersion estimation method applicable to various oversampling factors, comprising:

[0007] Collecting oversampled time-domain signals, processing the time-domain signals to obtain a frequency-domain signal sequence;

[0008] Constructing a low-complexity dispersion estimation model, the low-complexity dispersion estimation model including: a timing phase error detection model, a sequence zero-padding model, and a dispersion estimation model;

[0009] Inputting the frequency-domain signal sequence into the low-complexity dispersion estimation model to obtain an estimation result.

[0010] Optionally, obtaining the frequency-domain signal sequence includes: performing FFT transformation on the time-domain signal to obtain the frequency-domain signal sequence.

[0011] Optionally, the timing phase error detection model is:

[0012]

[0013] where ε is a real sequence, Y is the FFT result of the input complex signal sequence, n is the sequence index, N is the size of the FFT block, (·) * is to take the conjugate of the operand inside the parentheses, η is the oversampling factor, β is the roll-off factor, and Im(·) is to take the imaginary part of the operand inside the parentheses.

[0014] Optionally, constructing the zero-padding model for the sequence includes:

[0015] Obtaining the first clock component model of the timing phase error detection model and the frequency transfer function of dispersion, combining the frequency transfer function of dispersion with the first clock component model to obtain a second clock component model, and performing zero-padding on the second clock component model to obtain the zero-padding model for the sequence.

[0016] Optionally, the method for obtaining the zero-padding model for the sequence is:

[0017]

[0018] where CT is the clock component, λ is the optical wavelength, D is the dispersion coefficient, η is the oversampling factor, n is the sequence index, c is the speed of light, L is the fiber length, Z is the summation term of the zero-padded clock component, e is the natural base, π is pi, and M is N / 2.

[0019] Optionally, constructing the dispersion estimation model includes: performing an FFT form conversion on the zero-padding model for the sequence to obtain the dispersion estimation model.

[0020] Optionally, the method for obtaining the dispersion estimation model is:

[0021]

[0022] where CD is the dispersion estimation value, m is the index value, c is the speed of light, λ is the optical wavelength, B is the baud rate, and η is the oversampling factor.

[0023] Optionally, inputting the frequency-domain signal sequence into the low-complexity dispersion estimation model includes:

[0024] Inputting the frequency-domain signal sequence into the timing phase error detection model, calculating the conjugate product of the points at the frequency point indices in the frequency-domain signal sequence and the frequency points at the corresponding index intervals to obtain a complex sequence;

[0025] Input the complex sequence into the sequence zero-padding model, perform zero-padding processing on the complex sequence, and obtain a sequence with a target length;

[0026] Extract the FFT peak value of the sequence with the target length, further obtain the index position of the peak value, input the index position into the dispersion estimation model, and obtain the estimation result.

[0027] Optionally, after obtaining the sequence with the target length, it includes: performing FFT transformation on the sequence with the target length.

[0028] The beneficial effects of the present invention are as follows:

[0029] The present invention is applicable to various sampling rates with an oversampling factor greater than 1 and less than or equal to 2, has flexible applications, can still normally estimate dispersion at a low sampling rate, and greatly reduces the power consumption and cost of the ADC.

[0030] This method is based on the clock component CT in the TR module, does not introduce excessive computational complexity, and since the estimation process uses zero-padding operations, it does not introduce irrelevant noise terms, has strong anti-noise ability. When the oversampling factor is small, the number of effective terms for calculating CT becomes larger. Therefore, under the condition of ensuring a certain accuracy, the size N of the FFT block can take a lower value, greatly reducing the computational complexity and the occupation of hardware resources, and having good application prospects in high-speed optical fiber transmission systems such as short-distance metropolitan area networks and long-distance backbone networks. Description of the Drawings

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0032] Figure 1 It is a schematic structural diagram of a low-complexity dispersion estimation method applicable to various oversampling factors according to an embodiment of the present invention;

[0033] Figure 2 It is a curve graph of the estimation results of the Godard dispersion estimation method according to an embodiment of the present invention within the range of 32 ns / nm of the optical fiber dispersion value under the conditions of oversampling factors of 1.25, 1.5, and 2;

[0034] Figure 3 It is the |A k | spectral line graph when estimating dispersion by the Godard method under various oversampling factors according to an embodiment of the present invention;

[0035] Figure 4 The curve graph of the dispersion estimation results of the estimation method based on the new TPED proposed by the present invention in the present invention embodiment under various oversampling factors;

[0036] Figure 5 When the oversampling factors in the present invention embodiment are 2, 1.5, and 1.25, the |A k | spectral line graph when the new method proposed by the present invention estimates dispersion;

[0037] Figure 6 The curve graph of the change of the estimation accuracy compared with the FFT block size when the commonly used Godard method for estimating dispersion in the present invention embodiment and the new method proposed by the present invention are under their respective workable oversampling factors;

[0038] Figure 7 When the Godard method for estimating dispersion in the present invention embodiment and the new method proposed by the present invention are respectively under the oversampling factors of 2 and 1.25 and N = 2096, the |A k | spectral line graph. Detailed implementation manners

[0039] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0040] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0041] As Figure 1 shown, the structural schematic diagram of the low-complexity dispersion estimation method applicable to multiple oversampling factors provided by the present invention embodiment includes a first FFT operator 1, a sequence division module 2, a conjugate multiplication operator 3, a zero-padding module 4, a second FFT operator 5, a peak searching module 6, a dispersion estimation value calculation module 7, and a sum and take imaginary part operator 8. 1-7 constitute the method module proposed by the present invention, and 1, 2, 3, and 8 constitute the proposed TPED. It can be seen that there is a common part between the method proposed by the present invention and the TPED for executing TR, and the introduced additional computational complexity is relatively low.

[0042] The specific zero-padding process of the zero-padding module is as follows: zero-padding at both ends in the manner of [0..0, Z, 0..0], or zero-padding at a single end in the manner of [0..0, Z] and [Z, 0..0]. The zero-padding method at both ends is used when deriving the formula.

[0043] The peak searching module searches for peaks in the result sequence, which includes: the peak module obtains the maximum value by calculating the modulus of the obtained complex sequence, and at the same time can obtain the sequence position n0 where the maximum value is located (the range is 1 to the length of the complex sequence). m is in the range of [-M / 2, M / 2 - 1], and then find the value of the n0th point within this range (for example, n0 = 1 means this value is -M / 2), and this value is the peak index position m.

[0044] The summation and imaginary part extraction operator 8 has nothing to do with the dispersion estimation module of this embodiment. It can be seen from the given method diagram. However, since the method of this embodiment is based on the proposed TPED, and the TPED itself includes an operation of extracting the imaginary part. In order to clearly show the relationship between the TPED and the dispersion estimation module, it is therefore drawn in Figure 1 Complex numbers include real and imaginary parts. Extracting the imaginary part is to obtain the imaginary component of the complex number.

[0045] The low-complexity dispersion estimation method disclosed in this embodiment applicable to multiple oversampling factors, combined with Figure 1 includes the following steps:

[0046] Collect the time-domain signal of the oversampling factor, process the time-domain signal to obtain the frequency-domain signal sequence; construct a low-complexity dispersion estimation model, and the low-complexity dispersion estimation model includes: a timing phase error detection model, a sequence zero-padding model, and a dispersion estimation model; input the frequency-domain signal sequence into the low-complexity dispersion estimation model to obtain the estimation result, specifically:

[0047] The input time-domain signal with an oversampling factor greater than 1 is transformed to the frequency domain through FFT, and the frequency points with the frequency point index and the corresponding index interval of round(n+(1 - 1 / η)N) are divided, and the conjugate product of the two is calculated to obtain a complex sequence with a length of Zero-padding is performed at both ends or one end of this sequence to obtain a sequence with a length of M, perform FFT on this sequence, and search for peaks in the result sequence to obtain the peak index position m, and obtain the dispersion estimation result according to formula (8);

[0048] Among them, ceil(·), floor(·), and round(·) respectively represent rounding up, rounding down, and rounding to the nearest integer of the operand in the parentheses, N represents the size of the FFT block, β represents ROF, and η represents OSF.

[0049] This method is based on the following new TPED (timing phase error detector) calculation formula,

[0050]

[0051] Among them, ε is a real sequence, Y represents the FFT result of the input complex signal sequence, n represents the sequence index, N represents the size of the FFT block, (·) * represents taking the conjugate of the operand inside the parentheses, η represents the OSF, β represents the ROF, and Im(·) represents taking the imaginary part of the operand inside the parentheses.

[0052] The size of the clock component CT of this TPED can be expressed by the following formula.

[0053]

[0054] Among them, CT is the clock component, η is the oversampling factor, n is the sequence index, β is the ROF, Y is the FFT result of the input complex signal sequence, N is the size of the FFT block, (·) * is to take the conjugate of the operand inside the parentheses.

[0055] For the sequence Y in the above formula n the influence of dispersion is not considered, |(·)| represents taking the modulus value of the operand inside the parentheses, and the frequency-domain transfer function of dispersion is as follows:

[0056]

[0057] Among them, F sa represents the sampling rate, satisfying F sa = η·B, where B represents the baud rate, c represents the speed of light, n represents the sequence index, λ represents the optical wavelength, L represents the fiber length, D represents the dispersion coefficient, N represents the size of the FFT block, H CD [] represents the CD frequency-domain transfer function, π represents pi, j represents the imaginary unit, and e represents the natural base.

[0058] After considering the influence of dispersion on Y n , CT is written as the following formula,

[0059]

[0060] Among them, X n represents the FFT result of the input sequence affected by dispersion.

[0061] Satisfies After sorting, the size of CT can be expressed as the following formula,

[0062]

[0063] Among them Padding zeros to the above formula to make the number of summation terms M, which does not affect the size of CT, as shown in the following formula,

[0064]

[0065] Wherein, CT is the clock component, λ is the optical wavelength, D is the dispersion coefficient, η is the oversampling factor, n is the sequence index, c is the speed of light, L is the optical fiber length,

[0066] Z is the summation term of the zero-padded clock component, e is the natural base, π is the pi, and M is N / 2.

[0067] The magnitude of CT can be expressed as an M-point FFT form with respect to Z n :

[0068]

[0069] Wherein, A m is the value of the complex sequence in the frequency domain of Z n at index m.

[0070] Wherein CD = D·L, so the CD estimation formula is as follows:

[0071]

[0072] Wherein, CD is the dispersion estimation value, m is the index value, c is the speed of light, λ is the optical wavelength, B is the baud rate, and η is the oversampling factor.

[0073] It can be seen therefrom that only by performing a peak search operation on |A k | (-M / 2 ≤ k ≤ M / 2 - 1) to obtain the index value m at the peak position, the magnitude of CD can be estimated;

[0074] Wherein, |A k | is the modulus of the complex sequence in the frequency domain of Z n .

[0075] In Figure 2 shows the estimation result curves of the Godard dispersion estimation method under the conditions of oversampling factors of 1.25, 1.5, and 2 for the optical fiber dispersion value within the range of 32 ns / nm, wherein the transmitted signal is a 61 GBaud dual-polarization Nyquist signal, the roll-off factor β = 0.1, the FFT block size N = 4096, the speed of light c = 299792458 m / s, and the wavelength λ = 1550 nm. Figure 3 Corresponding to Figure 2 in the case of an oversampling factor of 1.25, the |A k | spectral lines of the Godard estimation method at different dispersion values are normalized according to the highest value of the spectral line at a dispersion value of 0. Figure 4It shows the estimation result curves of the dispersion estimation method based on the new TPED proposed in the present invention under the conditions of oversampling factors of 1.25, 1.5, and 2 for the fiber dispersion value within the range of 32 ns / nm. The transmitted signal is a 61 GBaud dual-polarization Nyquist signal, the roll-off factor β = 0.1, the FFT block size N = 4096, the speed of light c = 299792458 m / s, and the wavelength λ = 1550 nm. Figure 5 Corresponding to Figure 4 When the oversampling factor is 1.25, the |A k | spectral lines of the proposed new method at different dispersion values are normalized according to the highest value of the spectral line when the dispersion value is 0. It can be seen that the Godard method can achieve a relatively accurate estimation result when the oversampling factor is 2 and can find the accurate peak index. However, when the oversampling factor is reduced to 1.5 and even 1.25, the Godard method fails, and the |A k | spectral lines are submerged by noise, and the peak position is blurred. The method based on the new TPED proposed by us will not have such a problem. For the cases of oversampling factors of 2, 1.5, and 1.25, relatively accurate estimation results can be obtained, the peak position is obvious, and the correct peak position can be found. Therefore, the proposed new method has the advantage of being applicable to multiple oversampling factors.

[0076] Figure 6 It shows the absolute value error curves estimated when the Godard method has an oversampling factor of 2, the proposed method has oversampling factors of 2 and 1.25, and N varies from 2096 to 4096 under the condition that the fiber dispersion value is 16 ns / nm. The transmitted signal is a 61 GBaud dual-polarization Nyquist signal, β = 0.1, c = 299792458 m / s, and λ = 1550 nm. Figure 7 It shows the |A k | spectral lines of the Godard method and the proposed new method when the oversampling factors are 2 and 1.25 respectively and N = 2096, which are normalized according to the highest value of the spectral line of the proposed method when the oversampling rate is 1.25. It can be seen that as the FFT block decreases, the estimation effect of the Godard method gradually deteriorates. When N = 2096, the peak position is blurred, and the estimation error has exceeded the tolerance range of 0.48 ns / nm of Godard-TPED, so it cannot be used in the actual system at this time. However, when N = 2096, the peak position of the proposed method is obvious, and the dispersion value can still be accurately estimated. It can be predicted that for the tolerance range of 0.48 ns / nm, the proposed method can further reduce the FFT block size, and the oversampling factor is 1.25. Therefore, the proposed method can not only reduce the computational complexity but also reduce the ADC sampling rate, greatly reducing the power consumption and cost of the system.

[0077] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A low-complexity dispersion estimation method applicable to multiple oversampling factors, characterized in that, Including: Collecting an oversampled time-domain signal, processing the time-domain signal to obtain a frequency-domain signal sequence; Constructing a low-complexity dispersion estimation model, the low-complexity dispersion estimation model including: a timing phase error detection model, a sequence zero-padding model, and a dispersion estimation model; Constructing the sequence zero-padding model includes: Obtaining a first clock component model of the timing phase error detection model and a frequency transfer function of dispersion, combining the frequency transfer function of dispersion with the first clock component model to obtain a second clock component model, and zero-padding the second clock component model to obtain the sequence zero-padding model; The method for obtaining the sequence zero-padding model is: Where CT is the clock component, λ is the optical wavelength, D is the dispersion coefficient, η is the oversampling factor, n is the sequence index, c is the speed of light, L is the fiber length, Z is the sum term of the zero-padded clock component, e is the natural base, π is pi, and M is N / 2; Inputting the frequency-domain signal sequence into the low-complexity dispersion estimation model to obtain an estimation result: Inputting the frequency-domain signal sequence into the timing phase error detection model, calculating the conjugate product of the point with the frequency point index in the frequency-domain signal sequence and the frequency point with the corresponding index interval to obtain a complex sequence; Inputting the complex sequence into the sequence zero-padding model, performing zero-padding processing on the complex sequence to obtain a target-length sequence; Extracting the FFT peak of the target-length sequence, further obtaining the index position of the peak, and inputting the index position into the dispersion estimation model to obtain the estimation result.

2. The low-complexity dispersion estimation method applicable to multiple oversampling factors according to claim 1, wherein Obtaining the frequency-domain signal sequence includes: performing an FFT transform on the time-domain signal to obtain the frequency-domain signal sequence.

3. The low-complexity dispersion estimation method applicable to multiple oversampling factors according to claim 1, characterized in that The timing phase error detection model is: where ε is a real sequence, Y is the FFT result of the input complex signal sequence, n is the sequence index, N is the size of the FFT block, (·) * denotes taking the conjugate of the operand inside the parentheses, η is the oversampling factor, β is the roll-off factor, and Im(·) denotes taking the imaginary part of the operand inside the parentheses.

4. The low-complexity dispersion estimation method applicable to multiple oversampling factors according to claim 1, wherein, Constructing the dispersion estimation model includes: performing an FFT form conversion on the sequence zero-padding model to obtain the dispersion estimation model.

5. The low-complexity dispersion estimation method applicable to multiple oversampling factors according to claim 4, characterized in that The method for obtaining the dispersion estimation model is: Where CD is the dispersion estimation value, m is the index value, c is the speed of light, λ is the optical wavelength, B is the baud rate, and η is the oversampling factor.

6. The low-complexity dispersion estimation method applicable to multiple oversampling factors according to claim 1, characterized in that After obtaining the target-length sequence, it includes: performing an FFT transform on the target-length sequence.

Images