An AGC method and apparatus based on Sigmoid variable step size normalization

Abstract

This invention discloses an AGC method and apparatus based on Sigmoid variable step size normalization. Belonging to the field of signal transmission, the method includes: nonlinearly compressing the output error at the current moment based on the Sigmoid function to obtain the base step size at the current moment; normalizing the base step size at the current moment according to the power of the input signal at the current moment to obtain the actual step size at the current moment; and determining the gain value at the next moment based on the actual step size at the current moment using an LMS algorithm incorporating a leakage factor, wherein the leakage factor is used to suppress gain drift. This invention achieves a dynamic balance between convergence speed, steady-state error, robustness, and abrupt change response capability.

Description

Technical Field

[0001] This invention belongs to the field of signal transmission technology, specifically relating to an AGC method and apparatus based on Sigmoid variable step size normalization. Background Technology

[0002] With the rapid development of communication technology, high-speed digital signal transmission is widely used in wireless communication, fiber optic communication, satellite communication, and high-speed wired interfaces. These systems typically face complex channel environments, including signal attenuation, multipath effects, and noise interference, leading to drastic variations in received signal strength. Therefore, the system needs adaptive adjustment capabilities to ensure the accuracy and stability of signal processing. Automatic Gain Control (AGC) is one of the core technologies for achieving this goal. AGC dynamically adjusts the gain of the received signal to ensure that the signal remains within a suitable amplitude range when entering subsequent processing modules (such as analog-to-digital converters and demodulators), thereby improving system stability, receiver sensitivity, and overall communication quality.

[0003] In receiver design, after passing through a low-noise amplifier and a variable gain amplifier, the signal is typically fed into an AGC (Automatic Gain Control) module. AGC detects the signal's power or amplitude and uses feedback to control the amplifier's gain, maintaining the output signal within a set range. This process is particularly important for analog-to-digital converters (ADCs) because ADCs have limited dynamic range; excessively strong input signals can lead to saturation distortion, while excessively weak signals reduce the signal-to-noise ratio. The introduction of AGC effectively resolves this contradiction, improving the system's adaptability to different input signal strengths and making it an indispensable component of high-speed communication links.

[0004] There are two main types of mainstream AGC implementations: analog AGC and digital AGC. A typical analog AGC structure consists of a closed loop formed by a voltage-controlled amplifier, a logarithmic compression amplifier, a power detector (rectifier / logarithmic amplifier), and an error amplifier. It can operate at extremely high sampling rates (>10GHz) with circuit response times on the order of nanoseconds. However, analog AGC is limited by device nonlinearity, temperature drift, and power consumption, making it difficult to achieve wide-range, programmable gain adjustment. In wide dynamic range (>80dB) and high-noise environments, it often experiences gain fluctuations, lockout, or saturation.

[0005] Digital AGC implements gain control in the digital domain. Since it does not use analog devices, it avoids a series of problems associated with analog AGC schemes. Common digital AGC schemes include fixed-gain multiplication, open-loop feedforward, least mean square (LMS), normalized LMS, and variable-step LMS. Fixed-gain multiplication and open-loop feedforward methods suffer from narrow operating ranges and sensitivity to error responses, respectively. Fixed-step AGC methods like LMS and normalized LMS require a trade-off between convergence speed and steady-state error. Variable-step LMS updates using a pure accumulation mode; after the signal reaches a steady state, the gain value is locked and cannot be automatically reset, resulting in a sluggish initial response to sudden signals and a slow tracking speed.

[0006] Therefore, there is an urgent need for an AGC method that can simultaneously improve convergence speed, reduce steady-state error, maintain robustness to power fluctuations, and improve response speed to burst signals. Summary of the Invention

[0007] This invention provides an AGC method and apparatus based on Sigmoid variable step size normalization, which can solve the above-mentioned technical problems.

[0008] In a first aspect, embodiments of the present invention provide an AGC method based on Sigmoid variable step size normalization, the method comprising:

[0009] Based on the Sigmoid function, the output error at the current time is nonlinearly compressed to obtain the basic step size at the current time.

[0010] Based on the power of the input signal at the current moment, the basic step size at the current moment is normalized to obtain the actual step size at the current moment;

[0011] The LMS algorithm based on the introduction of a leakage factor determines the gain value at the next time step according to the actual step size at the current time step, wherein the leakage factor is used to suppress gain drift.

[0012] Secondly, embodiments of the present invention provide an AGC device based on Sigmoid variable step size normalization, including a step size adjustment module, a normalization module, and a gain update module;

[0013] The step size adjustment module is used to perform nonlinear compression on the output error at the current time based on the Sigmoid function to obtain the basic step size at the current time.

[0014] The normalization module is used to normalize the base step size at the current moment based on the power of the input signal at the current moment, so as to obtain the actual step size at the current moment.

[0015] The gain update module is used to determine the gain value at the next time step based on the actual step size at the current time step of the LMS algorithm with the leakage factor introduced, wherein the leakage factor is used to suppress gain drift.

[0016] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: The present invention uses the Sigmoid function to make the basic step size adapt nonlinearly with the error, providing a large step size in the early stage to improve the convergence speed, and switching to a small step size in the later stage to reduce the steady-state error. The normalization process improves the robustness to signal power fluctuations, and the introduction of a leakage factor in the LMS algorithm can effectively suppress long-term gain drift and improve the response capability to sudden signals; thus, a dynamic balance is achieved between convergence speed, steady-state error, robustness, and sudden change response capability. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the implementation of an AGC method based on Sigmoid variable step size normalization, provided in an embodiment of the present invention;

[0018] Figure 2 This is a schematic diagram of an AGC device based on Sigmoid variable step size normalization provided in an embodiment of the present invention. Detailed Implementation

[0019] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0020] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0021] It should also be understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0022] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0023] Furthermore, in the description of this invention and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0025] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0026] Example 1

[0027] Figure 1 The diagram shown illustrates an implementation flowchart of an AGC method based on Sigmoid variable step size normalization provided by an embodiment of the present invention. As an example and not a limitation, the method may include steps S101-S103, which are described below.

[0028] S101, based on the Sigmoid function, performs nonlinear compression on the output error at the current time step to obtain the basic step size at the current time step step.

[0029] In one example, to balance acquisition speed and steady-state accuracy, an error magnitude-based approach can be used. The Sigmoid function calculates the base step size at the current time step.

[0030] For example, the base step size can be updated using the following formula:

[0031] ,

[0032] in, For the first The base step size at any given time. , These are the minimum and maximum step size thresholds, respectively. It is a natural constant. For scaling parameters, This represents the Sigmoid function. For the first Output error at any given time.

[0033] Specifically, when When larger , near The basic step size can converge quickly; when When smaller, , Back to This reduces jitter during steady-state operation.

[0034] The Sigmoid function enables the basic step size to converge quickly when the error is large, and suppresses oscillations when the error is small, thereby improving steady-state accuracy.

[0035] S102, based on the power of the input signal at the current moment, normalize the basic step size at the current moment to obtain the actual step size at the current moment.

[0036] In one example, the actual step size can be calculated using the following formula:

[0037] ,

[0038] in, For the first The actual step size at each moment, For the first The base step size at any given time. For the input signal Power at any moment To prevent division by zero of constants.

[0039] For example, the power of the input signal at the current moment can satisfy the following formula:

[0040] ,

[0041] in, This is the smoothing coefficient for the moving average. For the input signal Power at any moment For the first The input signal at any given moment.

[0042] Normalization enables the step size to be updated adaptively based on the input power, preventing instability in step size updates caused by high power input.

[0043] If the normalized step size is directly processed using the Sigmoid function, the actual step size will be affected by both the error (via the Sigmoid function) and the input power (normalization). In scenarios with large errors and high power, such as when sudden strong interference causes both the Sigmoid function output value and the normalization denominator to be large, the step size command will be ambiguous, leading to momentary instability or sluggish response. This invention first uses the Sigmoid function to calculate the basic step size. The Sigmoid function can non-linearly map the error into a bounded basic step size, establishing a stable baseline. Then, the basic step size is normalized to determine the actual step size. Normalization allows for fine-tuning of the stable basic step size, avoiding direct conflict between the two strong variables of error and input power, making the dynamic update process of the step size smoother and more robust.

[0044] S103, based on the LMS algorithm with a leakage factor, determines the gain value for the next time step according to the actual step size at the current time step.

[0045] In one example, the leakage factor acts on the product of the gain and the step size.

[0046] For example, the gain value can be updated according to the following formula:

[0047] ,

[0048] in, For the first Gain value at time, As a leakage factor, For the first The actual step size at each moment, For the first Output error at time 10:00 For the first The input signal at any given moment.

[0049] Specifically, when In this case, the gain value will be slightly reduced in each iteration, which is numerically equivalent to adding a quadratic penalty term to the cost function. .

[0050] For example, the output error can be calculated using the following formula:

[0051] ,

[0052] in:

[0053] ,

[0054] in, Output amplitude for the target. For the first The actual output amplitude at any given moment.

[0055] Since the gain is required to have memoryless or self-resetting characteristics, if we directly use the traditional LMS algorithm and adopt the weight accumulation update mode: Determine the gain value based on the actual step size and update the item. The cumulative effect will lead to gain If the signal is weak or the error is zero for a long time, the gain will remain at a historical value, or it will drift slowly under noise disturbances and fail to reset automatically, thus prolonging the gain optimization time. Therefore, this invention introduces a leakage factor to add a forgetting term to the update formula, which enables the gain to decay automatically and gently, suppressing gain accumulation and drift caused by the lack of continuous error drive or noise, and continuously driving the gain value update.

[0056] For example, the cost function of the LMS algorithm can be expressed as:

[0057] ,

[0058] in, For the first The cost at time step 1, i.e., the cost function at time step 2. Output value at time.

[0059] Specifically, the formula for updating the gain value is constructed by minimizing the cost function through gradient descent, and the gradient of the cost function is... The minimum time.

[0060] Although the leakage factor is directly used as the gradient term Introducing weights into the LMS algorithm can, to some extent, avoid local minima or enhance exploratory behavior, but due to dynamic signals... The resulting disturbances are difficult to predict. When negative, they can cause gain sign flipping, which conflicts with the basic requirement of AGC to stabilize the gain within a reasonable positive range, easily leading to system oscillation or saturation. This invention uses the leakage factor directly as the weight of the product of gain and step size (both of which are positive values), thus avoiding gain sign flipping. It also allows the adjustment of the leakage factor to be independent of the previous error adjustment (handled by Sigmoid) and input power adjustment (handled by normalization), thereby accurately balancing the ability to track new signals and the ability to forget old states and prevent drift.

[0061] This invention utilizes the Sigmoid function to make the base step size adapt nonlinearly with the error. It provides a large step size in the early stage to improve the convergence speed, and then switches to a small step size in the later stage to reduce steady-state error. Normalization processing improves robustness to signal power fluctuations, and the introduction of a leakage factor in the LMS algorithm can effectively suppress long-term gain drift and improve the response capability to sudden signals. Thus, a dynamic balance is achieved between convergence speed, steady-state error, robustness, and sudden change response capability.

[0062] Example 2

[0063] Figure 2 The diagram shown illustrates the structure of an AGC device based on Sigmoid variable step size normalization according to an embodiment of the present invention. As an example and not a limitation, the device 200 may include a step size adjustment module 210, a normalization module 220, and a gain update module 230.

[0064] For example, the step size adjustment module 210 is used to perform nonlinear compression on the output error at the current moment based on the Sigmoid function to obtain the basic step size at the current moment; the normalization module 220 is used to normalize the basic step size at the current moment according to the power of the input signal at the current moment to obtain the actual step size at the current moment; the gain update module 230 is used to determine the gain value at the next moment based on the actual step size at the current moment using the LMS algorithm with the leakage factor introduced, wherein the leakage factor is used to suppress gain drift.

[0065] In one possible implementation, the basic step size satisfies the following formula:

[0066] ,

[0067] in, For the first The base step size at any given time. , These are the minimum and maximum step size thresholds, respectively. It is a natural constant. For scaling parameters, This represents the Sigmoid function. For the first Output error at any given time.

[0068] In one possible implementation, the actual step size satisfies the following formula:

[0069] ,

[0070] in, For the first The actual step size at each moment, For the first The base step size at any given time. For the input signal Power at any moment To prevent division by zero of constants.

[0071] In one possible implementation, the leakage factor acts on the product of the actual step size and the gain value.

[0072] In one example, the gain value satisfies the following formula:

[0073] ,

[0074] in, For the first Gain value at time, As a leakage factor, For the first The actual step size at each moment, For the first Output error at time 10:00 For the first The input signal at any given moment.

[0075] This invention utilizes the Sigmoid function to make the base step size adapt nonlinearly with the error. It provides a large step size in the early stage to improve the convergence speed, and then switches to a small step size in the later stage to reduce steady-state error. Normalization processing improves robustness to signal power fluctuations, and the introduction of a leakage factor in the LMS algorithm can effectively suppress long-term gain drift and improve the response capability to sudden signals. Thus, a dynamic balance is achieved between convergence speed, steady-state error, robustness, and sudden change response capability.

[0076] To better illustrate the beneficial effects of the present invention, the following simulation experiments were conducted:

[0077] In the simulation experiment, the traditional LMS method, the traditional variable step size LMS method, the normalized LMS method, and the AGC method provided by this invention were compared in terms of convergence time, steady-state mean square error (MSE), gain ripple, dynamic response time, and output signal-to-noise ratio (SNR). Specific data can be found in Table 1 below.

[0078] Table 1

[0079]

[0080] Specifically, convergence time is the number of sampling points (or time) required for the AGC method to converge to the steady-state error threshold from the start of operation; steady-state MSE can be used to measure the long-term error level of the adaptive system, and the smaller the value, the better; gain fluctuation can measure the magnitude of AGC steady-state jitter, and the larger the fluctuation, the lower the tolerance to noise and model mismatch, which may lead to amplitude jitter of subsequent signals, decrease in SNR, and increase in power consumption; dynamic recovery value refers to the number of sampling points (or time) required for AGC to reconverge from the new error to the steady-state error threshold after a sudden change in the amplitude of the input signal. It can measure the tracking ability of AGC in sudden environments and directly determines whether the system can maintain stable output amplitude, low bit error rate, and controllable power consumption during power fluctuations; the higher the output SNR, the better the system suppresses errors in low SNR environments.

[0081] As can be seen from Table 1 above, the present invention has the minimum convergence time, steady-state error, gain fluctuation, and dynamic recovery time, and the output signal-to-noise ratio is higher. This indicates that the present invention uses the Sigmoid function to make the basic step size change nonlinearly with the error, providing a large step size in the early stage to improve the convergence speed, and switching to a small step size in the later stage to reduce the steady-state error. The normalization process improves the robustness to signal power fluctuations, achieving a dynamic balance between convergence speed, steady-state error, robustness, and sudden change response capability.

[0082] Furthermore, referring to Table 1 above, it can be seen that compared to the traditional LMS method, the increases in gain fluctuation and dynamic recovery value of the present invention are greater than the sum of the increases of other traditional improved LMS methods (such as variable step size and normalization) compared to the traditional LMS method. This indicates that the present invention, by introducing a leakage factor into the LMS algorithm, can effectively suppress long-term gain drift caused by the Sigmoid function and normalization, thereby improving the response capability to sudden signals.

[0083] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

Claims

1. An AGC method based on Sigmoid variable step size normalization, characterized in that, include: Based on the Sigmoid function, the output error at the current time is nonlinearly compressed to obtain the basic step size at the current time. Based on the power of the input signal at the current moment, the basic step size at the current moment is normalized to obtain the actual step size at the current moment; The LMS algorithm based on the introduction of a leakage factor determines the gain value at the next time step according to the actual step size at the current time step, wherein the leakage factor is used to suppress gain drift. The basic step size satisfies the following formula: ， in, For the first The base step size at any given time. , These are the minimum and maximum step size thresholds, respectively. It is a natural constant. For scaling parameters, This represents the Sigmoid function. For the first Output error at any given time.

2. The AGC method according to claim 1, characterized in that, The actual step size satisfies the following formula: ， in, For the first The actual step size at each moment, For the first The base step size at any given time. For the input signal Power at any moment To prevent division by zero of constants.

3. The AGC method according to claim 1, characterized in that, The leakage factor acts on the product of the actual step size and the gain value.

4. The AGC method according to claim 3, characterized in that, The gain value satisfies the following formula: ， in, For the first Gain value at time, As a leakage factor, For the first The actual step size at each moment, For the first Output error at time 10:00 For the first The input signal at any given moment.

5. An AGC device based on Sigmoid variable step size normalization, characterized in that, Includes a step size adjustment module, a normalization module, and a gain update module; The step size adjustment module is used to perform nonlinear compression on the output error at the current time based on the Sigmoid function to obtain the basic step size at the current time. The normalization module is used to normalize the base step size at the current moment based on the power of the input signal at the current moment, so as to obtain the actual step size at the current moment. The gain update module is used to determine the gain value at the next time step based on the actual step size at the current time step of the LMS algorithm with the leakage factor introduced, wherein the leakage factor is used to suppress gain drift. The basic step size satisfies the following formula: ， in, For the first The base step size at any given time. , These are the minimum and maximum step size thresholds, respectively. It is a natural constant. For scaling parameters, This represents the Sigmoid function. For the first Output error at any given time.

6. The AGC device according to claim 5, characterized in that, The actual step size at the current moment satisfies the following formula: ， in, For the first The actual step size at each moment, For the first The base step size at any given time. For the input signal Power at any moment To prevent division by zero of constants.

7. The AGC device according to claim 5, characterized in that, The leakage factor acts on the product of the actual step size and the gain value.

8. The AGC device according to claim 7, characterized in that, The gain value satisfies the following formula: ， in, For the first Gain value at time, As a leakage factor, For the first The actual step size at each moment, For the first Output error at time 10:00 For the first The input signal at any given moment.

Images