Clock frequency prediction method based on temperature and aging hybrid compensation

Abstract

The invention relates to the field of clock frequency compensation, in particular to a clock frequency prediction method based on temperature and aging mixed compensation. According to the method, linear fitting is carried out on the original observation points in the multiple aging rate estimation time periods, local linearization is carried out on the nonlinear clock frequency aging offset process, compared with an overall nonlinear fitting mode, the calculated amount is reduced, potential error accumulation caused by inaccurate models in overall nonlinear fitting can be prevented, and the accuracy of the overall nonlinear fitting is improved. And the fitting accuracy is improved. In addition, piecewise linear fitting can also adapt to dynamic changes of data, and prediction precision is improved. Meanwhile, the temperature correction values of the adjacent reference points are used for estimating the temperature correction values of the to-be-predicted points, and the accuracy is high.

Description

Technical Field

[0001] The present invention relates to the field of clock frequency compensation, and more particularly, to a clock frequency prediction method based on hybrid compensation of temperature and aging. Background Art

[0002] The clock frequency is the basic frequency used for timing and controlling operations in an electronic system. It determines the speed and rhythm of internal operations of the system and is very important for the stability and accuracy of the system. During long-term operation, the clock frequency may deviate due to factors such as aging and temperature changes. By estimating and predicting the clock frequency, these deviations can be detected in a timely manner, and corresponding maintenance measures, such as replacing components or adjusting the frequency, can be taken to maintain the normal operation of the system.

[0003] Generally, due to the combined effects of multiple physical mechanisms and engineering factors (such as voltage changes, mechanical stress, electromagnetic interference, and crystal oscillator differences), the influence of temperature on the clock frequency is non-linear or non-repeatable, that is, for the same clock hardware at the same temperature, the frequency increment is not necessarily the same; at the same time, the aging effect is also non-repeatable. Taking the electrical start-up time as the starting point for multiple repeated tests, at the same relative time point, the frequency change of the clock is not consistent.

[0004] The prior art CN106936425A discloses a clock frequency maintaining method. This method obtains the current temperature information and the current frequency control word, separates the current temperature signal and the current aging signal, and performs temperature prediction based on the current temperature signal and the current temperature information; performs aging prediction based on the current aging signal, thereby determining temperature and aging compensation, and further maintaining the stability and accuracy of the clock frequency output. However, both its temperature prediction and aging prediction are based on the Kalman model. The Kalman model relies on the linear assumption established by historical data, and the non-repeatability of temperature and aging will cause deviation in model prediction. Therefore, this method still has a large compensation error. Summary of the Invention

[0005] The object of the present invention is to improve the prediction accuracy of the clock frequency.

[0006] To achieve the above object, the present invention provides a clock frequency prediction method based on hybrid compensation of temperature and aging, and the method includes the following steps:

[0007] Step 1: After the electronic system is powered on, real-time monitor the temperature data of the environment, mark several reference points according to a preset temperature interval, and record the temperature value, time stamp, and average clock frequency corresponding to the reference points; synchronously collect the measured clock frequency values of the electronic system according to a preset measurement period, and record the temperature value and time stamp corresponding to each collection point to obtain several original observation points;

[0008] Step 2: Select several time periods with stable temperature changes as the aging rate estimation time periods. For any one of the aging rate estimation time periods, fit the original observation points therein using a first-order linear model, and use the slope term of the first-order linear model as the aging rate estimation value at the average time point of the aging rate estimation time period;

[0009] Step 3: Based on the aging rate estimation value, fit the relationship curve of the aging rate changing with time to obtain an aging rate curve, and perform integral processing on the aging rate curve to obtain an aging curve;

[0010] Step 4: Calculate the aging correction values of the point to be predicted and its adjacent reference points based on the aging curve. Subtract the corresponding aging correction value from the average clock frequency of the adjacent reference point to obtain the temperature correction value of the adjacent reference point, and calculate the temperature correction value of the point to be predicted based on the temperature correction value of the adjacent reference point;

[0011] Step 5: Add the aging correction value of the point to be predicted to the temperature correction value of the point to be predicted to obtain the clock frequency prediction value of the point to be predicted.

[0012] Wherein, the principle of the present invention is:

[0013] At any time point t, the actual value f(t, T) of the electronic system clock frequency includes three parts: the nominal value f0 of the clock frequency (i.e., the reference frequency without external factor interference), the frequency aging offset f c (t) that gradually accumulates over time, and the frequency temperature offset f c (T) that fluctuates with the current temperature T, that is: f(t, T) = f0 + f c (t) + f c (T) (1).

[0014] Wherein, f c (t) can also be divided into two parts: one part is the instantaneous clock frequency offset caused by the aging effect at the moment when the electronic system is powered on, and the magnitude of this part is fixed after the system is powered on; the other part is the aging offset that accumulates as the electronic system runs after power-on. In order to separate the two and enable the aging offset that accumulates as the electronic system runs to be directly calculated by fitting a curve, the present invention differentiates both sides of formula (1) with respect to time to obtain:

[0015]

[0016] Wherein, (The derivative of the frequency aging offset with respect to time) represents the aging offset rate, which is a function related to time and is denoted by K(t). Integrating K(t) gives S(t) (where S(t) = ∫K(t)dt), and then we have:

[0017] f c (t) = S(t) + C0;

[0018] where C0 is the integration constant, representing the instantaneous clock frequency offset, and S(t) is the aging correction value, representing the aging offset accumulated as the electronic system operates after power-on.

[0019] Therefore, formula (1) can be written as f(t,T) = f0 + f c (T) + S(t) + C0;

[0020] where, during a period with stable temperature change, the temperature change rate is close to 0. Therefore, the influence of temperature change on the frequency-temperature offset f c (T) can be ignored, that is, f c (T) is a constant during these periods. At the same time, f0 and C0 are also fixed constants. Therefore, formula (1) can be further written as:

[0021] f(t,T) = S(t) + C(T) (3);

[0022] where C(T) is a constant related only to temperature, which is defined as the temperature correction value, and C(T) = f0 + f c (T) + C0.

[0023] Therefore, if the aging correction value S(t) and the temperature correction value C(t) of the point to be predicted are obtained and added together, the clock frequency prediction value of the point to be predicted can be obtained.

[0024] For the aging correction value: The present invention first selects a certain number of aging rate estimation time periods, and fits the original observation points within these time periods using a first-order linear model. Since the temperature change is stable during this period, the change in the measured value of the system clock frequency can be considered to be caused only by the aging effect. Therefore, the first-order linear model obtained by fitting only reflects the change of the clock frequency with time, and its slope term can be used as the aging rate estimation value to reflect the aging rate. Then, the aging rate curve is obtained by fitting the aging rate estimation values at different time points, and further integrated to obtain the aging curve; substituting the time corresponding to the point to be predicted into the aging curve, the aging correction value can be obtained.

[0025] For the temperature correction value: In the present invention, first, the aging correction values of two adjacent reference points of the point to be predicted are obtained through the aging curve, then, based on formula (3), the average clock frequency of the adjacent reference points is subtracted by their corresponding aging correction values to obtain the temperature correction values of the adjacent reference points, and then the temperature correction value of the point to be predicted is calculated by using the linear interpolation method or the extrapolation method.

[0026] In the present invention, by performing linear fitting on the original observation points within multiple aging rate estimation time periods, the non-linear clock frequency aging offset process is locally linearized. Compared with the overall non-linear fitting method, not only the calculation amount is reduced, but also the potential error accumulation caused by inaccurate models in the overall non-linear fitting can be prevented, and the accuracy of the fitting is improved. Moreover, the piecewise linear fitting can also adapt to the dynamic changes of the data and improve the prediction accuracy. At the same time, the present invention estimates the temperature correction value of the point to be predicted by using the temperature correction values of the adjacent reference points, with a high accuracy rate.

[0027] Further, in step one, the average clock frequency corresponding to the reference point is obtained in the following manner:

[0028] Taking the temperature value corresponding to the reference point as the reference point, and using the preset temperature window threshold as the half-width, a reference temperature range is symmetrically constructed;

[0029] Taking the timestamp corresponding to the reference point as the reference point, and using the preset time window threshold as the half-width, a reference time range is symmetrically constructed;

[0030] Selecting the original observation points whose temperature values satisfy the reference temperature range and whose timestamps satisfy the reference time range as the average observation points corresponding to the reference points, and taking the arithmetic mean of the measured clock frequency values of the average observation points as the average clock frequency corresponding to the reference points.

[0031] Among them, since the average observation points satisfy both the reference temperature range and the reference time range, their aging offsets and temperature offsets are both close to the reference points. Therefore, the arithmetic mean of the measured clock frequency values of the average observation points can generally reflect the clock frequency of the reference points, and even if there are errors in individual average observation points, they will not have too much impact on the final result, thus enhancing the reliability of the measurement.

[0032] Further, in step two, the time period with stable temperature change includes: the time period in which the temperature change range does not exceed the preset range threshold and the temperature change rate does not exceed the preset rate threshold.

[0033] Further, in step two, the average time point of the aging rate estimation time period is: the time point corresponding to the arithmetic mean of the timestamps of all the original observation points in the aging rate estimation time period.

[0034] Among them, the average time point of the aging rate estimation time period can reflect the central position of the aging rate estimation time period, which is beneficial to the construction of the subsequent aging rate curve.

[0035] Further, in step four, the adjacent reference points include a first adjacent reference point and a second adjacent reference point. Among them, the temperature value of the first adjacent reference point is less than the temperature value of the point to be predicted, and the temperature value of the second reference point is greater than the temperature value of the point to be predicted.

[0036] Among them, the temperature value of the point to be predicted in the present invention is located between the temperature values of the first reference point and the second reference point, so that the temperature correction value of the point to be predicted can be estimated by using the interpolation method subsequently. Compared with the extrapolation method, the accuracy of the estimation result of the interpolation method is higher.

[0037] Further, in step four, the calculation formula for obtaining the temperature correction value of the point to be predicted based on the temperature correction values of the adjacent reference points is:

[0038]

[0039] Among them, C(T p ) is the temperature correction value of the point to be predicted, T p is the temperature value of the point to be predicted, T1 is the temperature value of the first reference point, T2 is the temperature value of the second reference point, C(T1) is the temperature correction value of the first reference point, and C(T2) is the temperature correction value of the second reference point.

[0040] Further, in step five, the predicted clock frequency value of the point to be predicted includes: the nominal value of the clock frequency, the frequency aging offset, and the frequency temperature offset.

[0041] Further, the following relationship exists between the frequency aging offset of the point to be predicted and the aging correction value of the point to be predicted:

[0042] S(t p ) = f c (t p ) - C0;

[0043] Among them, f c (t p ) is the frequency aging offset of the point to be predicted, S(t p ) is the aging correction value of the point to be predicted, and C0 is the integration constant.

[0044] Further, the following relationship exists between the frequency temperature offset of the point to be predicted and the temperature correction value of the point to be predicted:

[0045] C(T p) = f c (T p ) + f0 + C0;

[0046] Wherein, C(T p ) is the temperature correction value of the point to be predicted, f c (T p ) is the frequency-temperature offset of the point to be predicted, and f0 is the nominal value of the clock frequency of the point to be predicted.

[0047] Furthermore, the measured value of the clock frequency includes system measurement errors, and the method further includes:

[0048] Step six: Subtract the system measurement error from the predicted value of the clock frequency of the point to be predicted to obtain the clock frequency compensation reference value of the point to be predicted.

[0049] Wherein, the system measurement error is caused by factors such as measurement instruments, methods or environments, and is difficult to avoid in actual operations. Therefore, the average value of the clock frequencies of adjacent reference points in the present invention includes the system measurement error, which ultimately causes the predicted value of the clock frequency of the point to be predicted to also include the system measurement error. It is necessary to subtract this error to obtain the clock frequency compensation reference value of the point to be predicted. Finally, the clock frequency of the electronic system is compensated based on the clock frequency compensation reference value of the point to be predicted.

[0050] One or more technical solutions provided by the present invention have at least the following technical effects or advantages:

[0051] By linearly fitting the original observation points within multiple aging rate estimation time periods, the present invention localizes the non-linear clock frequency aging offset process, which not only reduces the calculation amount compared with the overall non-linear fitting method, but also prevents the potential error accumulation caused by inaccurate models in the overall non-linear fitting, improving the accuracy of the fitting. Moreover, the piecewise linear fitting can also adapt to the dynamic changes of the data, improving the prediction accuracy. At the same time, the present invention estimates the temperature correction value of the point to be predicted by using the temperature correction values of adjacent reference points, with high accuracy. Description of the Drawings

[0052] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, and constitute a part of the present invention, but do not limit the embodiments of the present invention;

[0053] Figure 1 is a schematic flowchart of the clock frequency prediction method based on temperature and aging hybrid compensation in the present invention. Detailed Embodiments

[0054] To more clearly understand the above objects, features, and advantages of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other.

[0055] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention may be implemented in other ways different from those described within the scope hereof. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0056] Embodiment 1

[0057] Please refer to Figure 1 , Embodiment 1 of the present invention provides a clock frequency prediction method based on temperature and aging mixed compensation. The method includes the following steps:

[0058] Step 1: After the electronic system is powered on, the temperature data of the environment is monitored in real time. Several reference points are marked at a preset temperature interval, and the temperature values, timestamps, and average clock frequencies corresponding to the reference points are recorded. The measured values of the clock frequency of the electronic system are synchronously collected at a preset measurement period, and the temperature values and timestamps corresponding to each collection point are recorded to obtain several original observation points.

[0059] Step 2: Select several time periods with stable temperature changes as the aging rate estimation time periods. For any one of the aging rate estimation time periods, the original observation points therein are fitted using a first-order linear model, and the slope term of the first-order linear model is used as the aging rate estimation value of the average time point of the aging rate estimation time period.

[0060] Step 3: Based on the aging rate estimation value, fit the relationship curve of the aging rate changing with time to obtain an aging rate curve, and perform integral processing on the aging rate curve to obtain an aging curve.

[0061] Step 4: Calculate the aging correction values of the point to be predicted and its adjacent reference points based on the aging curve. Subtract the corresponding aging correction value from the average clock frequency of the adjacent reference point to obtain the temperature correction value of the adjacent reference point, and calculate the temperature correction value of the point to be predicted based on the temperature correction value of the adjacent reference point.

[0062] Step 5: Add the aging correction value of the point to be predicted to the temperature correction value of the point to be predicted to obtain the clock frequency prediction value of the point to be predicted.

[0063] Among them, in step one, for the specific preset temperature interval and preset measurement period, those skilled in the art can determine according to the actual situation. For example, the preset temperature interval can be 0.5 °C, 1 °C, etc., and the preset measurement period can be 15 minutes, 30 minutes, etc. The measured value of the clock frequency can be determined by methods such as the frequency meter measurement method and the oscilloscope measurement method. The present invention does not make any limitations in this regard.

[0064] Among them, in step one, the average value of the clock frequency corresponding to the reference point is obtained by the following method:

[0065] Taking the temperature value corresponding to the reference point as the reference point, and using the preset temperature window threshold as the half-width, symmetrically construct a reference temperature range;

[0066] Taking the timestamp corresponding to the reference point as the reference point, and using the preset time window threshold as the half-width, symmetrically construct a reference time range;

[0067] Select the original observation points whose temperature values satisfy the reference temperature range and whose timestamps satisfy the reference time range as the average observation points corresponding to the reference point, and use the arithmetic mean of the measured values of the clock frequencies of the average observation points as the average value of the clock frequency corresponding to the reference point.

[0068] For example, assume that the temperature value corresponding to a reference point is T i , and the timestamp is t i , then the reference temperature range is (T i -T tol , T i +T tol ), where T tol is the preset temperature window threshold, and the reference time range is (t i -t tol , t i +t tol ), where t tol is the preset time window threshold. The calculation method of the average value of the clock frequency corresponding to the reference point is:

[0069]

[0070] Among them, is the average value of the clock frequency corresponding to the reference point, f(t j , T j ) is the measured value of the clock frequency of the average observation point, and n is the number of the average observation points.

[0071] Among them, in step two, the time period with stable temperature change includes: the time period during which the temperature change range does not exceed the preset range threshold and the temperature change rate does not exceed the preset rate threshold. For the length k of the specific aging rate estimation time period, those skilled in the art can determine it according to the actual situation. Assume that the first-order linear model obtained by fitting is f k (t) = K k ·t + c (c is a constant term), then the slope term K k is used as the aging rate estimation value at the average time point t k of the aging rate estimation time period.

[0072] In step two, the average time point of the aging rate estimation time period is: the time point corresponding to the arithmetic mean of the timestamps of all the original observation points in the aging rate estimation time period. That is where t r is the timestamp of the original observation point in the aging rate estimation time period, and N is the number of the original observation points in the aging rate estimation time period.

[0073] Among them, in step three, when fitting the aging rate curve, different curve types can be selected according to the actual situation, including but not limited to constants, polynomial curves or exponential curves. In this embodiment, an exponential curve is selected:

[0074] K(t) = b·e c·t + a, where a, b, c are constants to be fitted.

[0075] Integrating the above exponential curve to obtain the aging curve:

[0076]

[0077] Among them, in step four, the adjacent reference points include the first adjacent reference point and the second adjacent reference point. Among them, the temperature value of the first adjacent reference point is less than the temperature value of the point to be predicted, and the temperature value of the second reference point is greater than the temperature value of the point to be predicted.

[0078] For example, if t1 is the timestamp of the first reference point, t2 is the timestamp of the second reference point, and t p is the timestamp of the point to be predicted, substitute t1, t2, and t p into the above aging curve respectively to obtain the aging correction values S(t1), S(t2), and S(t p ) of the first reference point, the second reference point, and the point to be predicted. Then the aging correction values C(T1) and C(T2) of the first reference point and the second reference point are respectively:

[0079]

[0080] Among them, and are respectively the average clock frequencies corresponding to the first reference point and the first reference point.

[0081] Among them, in step four, the calculation formula for obtaining the temperature correction value of the point to be predicted based on the temperature correction values of the adjacent reference points is:

[0082]

[0083] Among them, C(T p ) is the temperature correction value of the point to be predicted, T p is the temperature value of the point to be predicted, T1 is the temperature value of the first reference point, and T2 is the temperature value of the second reference point.

[0084] Therefore, the predicted clock frequency value of the point to be predicted

[0085] Among them, the measured clock frequency value includes system measurement errors, and the method further includes:

[0086] Step six: Subtract the system measurement error δ from the predicted clock frequency value of the point to be predicted to obtain the clock frequency compensation reference value of the point to be predicted.

[0087] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0088] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A clock frequency prediction method based on mixed compensation of temperature and aging, characterized in that The method includes the following steps: Step 1: After the electronic system is powered on, the temperature data of the environment is monitored in real time. Several reference points are marked at preset temperature intervals, and the temperature values, timestamps, and average clock frequencies corresponding to the reference points are recorded. The measured clock frequency values of the electronic system are synchronously collected at preset measurement periods, and the temperature values and timestamps corresponding to each collection point are recorded to obtain several original observation points; Step 2: Select several time periods with stable temperature changes as the aging rate estimation time periods. For any one of the aging rate estimation time periods, the original observation points therein are fitted with a first-order linear model, and the slope term of the first-order linear model is used as the aging rate estimation value of the average time point of the aging rate estimation time period; Step 3: Based on the aging rate estimation value, fit the relationship curve of the aging rate changing with time to obtain an aging rate curve, and perform integral processing on the aging rate curve to obtain an aging curve; Step 4: Based on the aging curve, calculate the aging correction values of the point to be predicted and its adjacent reference points. Subtract the corresponding aging correction value from the average clock frequency of the adjacent reference point to obtain the temperature correction value of the adjacent reference point, and calculate the temperature correction value of the point to be predicted based on the temperature correction value of the adjacent reference point; Step 5: Add the aging correction value of the point to be predicted to the temperature correction value of the point to be predicted to obtain the clock frequency prediction value of the point to be predicted.

2. The clock frequency prediction method based on temperature and aging hybrid compensation according to claim 1, wherein In Step 1, the average clock frequency corresponding to the reference point is obtained through the following method: Taking the temperature value corresponding to the reference point as the reference point, symmetrically construct a reference temperature range with the preset temperature window threshold as the half-width; Taking the timestamp corresponding to the reference point as the reference point, symmetrically construct a reference time range with the preset time window threshold as the half-width; Select the original observation points whose temperature values satisfy the reference temperature range and whose timestamps satisfy the reference time range as the average observation points corresponding to the reference points, and use the arithmetic mean of the measured clock frequency values of the average observation points as the average clock frequency corresponding to the reference point.

3. A clock frequency prediction method based on temperature and aging hybrid compensation according to claim 1, characterized in that In Step 2, the time period with stable temperature change includes: a time period in which the temperature change range does not exceed the preset range threshold and the temperature change rate does not exceed the preset rate threshold.

4. A clock frequency prediction method based on temperature and aging hybrid compensation according to claim 1, characterized in that In Step 2, the average time point of the aging rate estimation time period is: the time point corresponding to the arithmetic mean of the timestamps of all the original observation points in the aging rate estimation time period.

5. A clock frequency prediction method based on temperature and aging hybrid compensation according to claim 1, characterized in that In Step 4, the adjacent reference points include a first adjacent reference point and a second adjacent reference point, where the temperature value of the first adjacent reference point is less than the temperature value of the point to be predicted, and the temperature value of the second reference point is greater than the temperature value of the point to be predicted.

6. The clock frequency prediction method based on temperature and aging hybrid compensation according to claim 5, wherein In Step 4, the calculation formula for calculating the temperature correction value of the point to be predicted based on the temperature correction value of the adjacent reference point is: Among them, C(T p ) is the temperature correction value of the point to be predicted, T p is the temperature value of the point to be predicted, T1 is the temperature value of the first reference point, T2 is the temperature value of the second reference point, C(T1) is the temperature correction value of the first reference point, and C(T2) is the temperature correction value of the second reference point.

7. A clock frequency prediction method based on temperature and aging hybrid compensation according to claim 1, characterized in that In Step 5, the clock frequency prediction value of the point to be predicted includes: the nominal value of the clock frequency, the frequency aging offset, and the frequency temperature offset.

8. A clock frequency prediction method based on temperature and aging hybrid compensation according to claim 7, characterized in that The frequency aging offset of the to-be-predicted point and the aging correction value of the to-be-predicted point have the following relationship: S(t p ) = f c (t p ) - C0; Among them, f c (t p ) is the frequency aging offset of the point to be predicted, S(t p ) is the aging correction value of the point to be predicted, and C0 is the integration constant.

9. A clock frequency prediction method based on temperature and aging hybrid compensation according to claim 8, characterized in that The frequency-temperature offset of the to-be-predicted point and the temperature correction value of the to-be-predicted point have the following relationship: C(T p ) = f c (T p ) + f0 + C0; Among them, C(T p ) is the temperature correction value of the point to be predicted, f c (T p ) is the frequency-temperature offset of the point to be predicted, and f0 is the nominal value of the clock frequency of the point to be predicted.

10. A clock frequency prediction method based on temperature and aging hybrid compensation according to claim 1, characterized in that, The measured value of the clock frequency contains a system measurement error, and the method further includes: Step Six: Subtract the system measurement error from the predicted value of the clock frequency of the to-be-predicted point to obtain the clock frequency compensation reference value of the to-be-predicted point.

Images