A chip trimming prediction method

Abstract

This disclosure discloses a chip tuning prediction method, comprising: S100: setting an upper limit for the number of tuning attempts of a first chip to be tuned; S200: measuring the signal frequency of the first chip to be tuned and obtaining a first measurement value; S300: archiving the first chip to be tuned based on the first measurement value, and tuning the first chip to be tuned according to the tuning prediction value of the chip's position, and recording the number of tuning attempts; S400: measuring the tuned signal frequency of the first chip to be tuned and obtaining a second measurement value; S500: comparing the second measurement value with a measurement standard value to test the first chip to be tuned. If the test passes, the first chip to be tuned is determined to be a good product, and the tuning prediction value of a second chip to be tuned in the same position as the first chip to be tuned is obtained based on the number of tuning attempts of the first chip to be tuned, and the tuning procedure is terminated.

Description

Technical Field

[0001] This disclosure pertains to the field of semiconductor testing, specifically relating to a chip trimming prediction method. Background Technology

[0002] With the rapid development of the chip design and manufacturing industry, the scale of semiconductor manufacturing is getting larger and larger, while the process dimensions are getting smaller and smaller. Even with increasingly advanced semiconductor production technologies and equipment, defects will always occur during the manufacturing process. Therefore, most chips with reference voltage, reference current, and frequency outputs have trimming circuits embedded in their design. During wafer probe testing and final packaging testing, these circuits are modified to correct any process changes that affect device parameters during wafer manufacturing.

[0003] There are two common tuning methods in chip testing. Figure 1 This is adjustment method one. This method measures the output signal and compares the measured actual value with the target value. After determining whether the actual value is too large or too small, the adjustment value is increased or decreased and then configured into the adjustment register, thereby adjusting the output signal. Figure 2 The second adjustment method measures the output signal and compares the measured actual value with the target value. If the actual value is too large or too small, the tester sends a corresponding adjustment signal (handshake signal) to the chip to adjust the chip signal.

[0004] In the first adjustment method, the adjustment value changes continuously during the adjustment process; generally, it is only increased or decreased by one. When dealing with multi-bit adjustment values, such as 4-bit adjustment values, the adjustment value is adjusted after each actual value measurement and then configured into the register. This operation may loop up to 16 times, resulting in 16 measurements and 16 register configurations, significantly increasing test time. Currently, common time optimization methods for adjustment method one include binary search and a statistical lookup table-based method for pre-setting initial adjustment values. The binary search method requires adjusting the adjustment value bit by bit from high to low, requiring 4-5 adjustments with a fixed test time. The lookup table method can pre-set adjustment values ​​more accurately, but it requires manual statistical analysis of measured data, increasing production complexity. Furthermore, this method cannot handle changes in adjustment performance caused by variations in process technology and chip characteristics between different batches of chips. In other words, the initial adjustment value preset by the statistical lookup table in the previous batch may not match that of the new batch, leading to inaccurate pre-setting.

[0005] The second adjustment method involves measuring the actual value, comparing it to the target value, executing the corresponding adjustment handshake protocol, and repeating the measurement and handshake operations until the measured value is adjusted to the required range. Alternatively, if the adjustment count reaches its limit and the measured value does not meet the pass requirement, the chip is considered a failure. However, if... Figure 2 As shown, each adjustment involves measurement and handshake protocol execution. If the number of adjustments reaches 10, the chip adjustment requires 11 signal measurements and 10 adjustment handshake protocols, which will lead to longer test time, reduced test efficiency, and increased test cost. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this disclosure is to provide a chip adjustment prediction method that can automatically adjust the predicted values ​​of each batch of products, thereby converging the chip prediction values ​​to the optimal range.

[0007] To achieve the above objectives, this disclosure provides the following technical solutions:

[0008] A chip trimming prediction method includes the following steps:

[0009] S100: Set the maximum number of adjustments required for the first chip to be adjusted;

[0010] S200: Measure the signal frequency of the first chip to be repaired and obtain the first measurement value;

[0011] S300: Archive the first chip to be repaired based on the first measurement value, and repair the first chip to be repaired according to the repair prediction value of the grade where the first chip to be repaired is located, and record the number of repairs.

[0012] S400: Measure the signal frequency of the first chip to be tuned after tuning, and obtain the second measurement value;

[0013] S500: The first chip to be repaired is tested by comparing the second measured value with the measurement standard value. If the test is passed, the first chip to be repaired is determined to be a good product. The repair prediction value of the second chip to be repaired, which is in the same grade as the first chip to be repaired, is obtained based on the number of repairs of the first chip to be repaired. At the same time, the repair process is terminated.

[0014] Preferably, in step S500, the adjustment prediction value of the second chip to be adjusted is obtained by an exponential smoothing-based prediction method.

[0015] Preferably, the method further includes the following steps:

[0016] S600: In step S500, if the detection fails, the first chip to be repaired will be re-detected based on whether the number of repairs to be performed on the first chip to be repaired has reached the upper limit of the number of repairs set in step S100.

[0017] Preferably, the method further includes the following steps:

[0018] S700: In step S600, if the number of repairs to the first chip to be repaired reaches the upper limit of the number of repairs, the first chip to be repaired is determined to be defective and the repair process is terminated; if the number of repairs to the first chip to be repaired does not reach the upper limit of the number of repairs, the second measurement value of the first chip to be repaired is compared with the measurement standard value again, and the corresponding handshake protocol is executed according to the comparison result.

[0019] Preferably, the method further includes the following steps:

[0020] S800: Update the number of adjustments to the first chip to be adjusted according to the corresponding handshake protocol to obtain the actual number of adjustments to the first chip to be adjusted.

[0021] Preferably, the method further includes the following steps:

[0022] S900: Measure the signal frequency of the first chip to be adjusted after the number of adjustments has been updated, and obtain the third measurement value.

[0023] Preferably, the method further includes the following steps:

[0024] S1000: The third measured value is compared with the measurement standard value to perform a third test on the first chip to be repaired. If the test passes, the first chip to be repaired is determined to be a good product. Based on the actual number of repairs of the first chip to be repaired, the repair prediction value of the second chip to be repaired in the same grade as the first chip to be repaired is obtained, and the repair procedure is terminated.

[0025] Preferably, the method further includes the following steps:

[0026] S1100: In step S1000, if the third test fails, then steps S600 to S1000 are repeated.

[0027] This disclosure also provides a computer storage medium, including:

[0028] Memory is used to store multiple computer instructions;

[0029] A processor is used to execute computer instructions to implement the methods described in any of the preceding methods.

[0030] This disclosure also provides an electronic device, including:

[0031] Memory, processor, and computer programs stored in memory and executable on the processor, wherein,

[0032] When the processor executes the program, it implements the method described in any of the preceding methods.

[0033] Compared with the prior art, the beneficial effects of this disclosure are as follows:

[0034] 1. The method described in this disclosure can improve the efficiency of adjustment and reduce the adjustment time;

[0035] 2. The method described in this disclosure can reduce the number of manual data collection steps;

[0036] 3. The method described in this disclosure can automatically adjust the predicted values ​​of each batch of products so that the predicted values ​​converge to the optimal range. Attached Figure Description

[0037] Figure 1 This is a flowchart of the method for adjusting via configuration registers;

[0038] Figure 2 This is a flowchart illustrating the method of adjustment via handshake protocol;

[0039] Figure 3 This is a schematic diagram showing the relationship between the initial measurement frequency of chip B and the number of adjustments.

[0040] Figure 4 This is a flowchart of a chip tuning method based on an exponential smoothing prediction algorithm proposed in this disclosure;

[0041] Figure 5 This is a schematic diagram illustrating the predictive effect of exponential smoothing on the adjustment value. Detailed Implementation

[0042] The following will refer to the appendix. Figures 1 to 5 Specific embodiments of this disclosure are described in detail. While specific embodiments of this disclosure are shown in the accompanying drawings, it should be understood that this disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of this disclosure to those skilled in the art.

[0043] It should be noted that certain terms are used in the specification and claims to refer to specific components. Those skilled in the art will understand that different terms may be used to refer to the same component. This specification and claims do not distinguish components based on differences in terminology, but rather on differences in function. The terms "comprising" or "including" used throughout the specification and claims are open-ended and should be interpreted as "comprising but not limited to." The following descriptions are preferred embodiments for carrying out this disclosure; however, these descriptions are for the purpose of understanding the general principles of the specification and are not intended to limit the scope of this disclosure. The scope of protection of this disclosure is determined by the appended claims.

[0044] To facilitate understanding of the embodiments of this disclosure, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. The accompanying drawings do not constitute a limitation on the embodiments of this disclosure.

[0045] In one embodiment, such as Figure 4 As shown, this disclosure provides a chip trimming prediction method, including the following steps:

[0046] S100: Set the maximum number of adjustments required for the first chip to be adjusted;

[0047] In this step, the chip to be repaired is designated as a type B chip, and the maximum number of repair attempts is set to 30. The distribution of repair attempts for type B chips is as follows: Figure 4 As shown, the number of adjustments is mainly concentrated at 20 times or less, and 30 times meets the testing requirements of most chips. If the upper limit of the number of adjustments is set too high, it means that for some defective products (fail, such as the chip adjustment module failure, which has no effect on the chip measurement parameters after executing the handshake protocol), more handshake signals need to be measured until the upper limit of the number of adjustments is reached before it is judged as a failure, which will waste a lot of time.

[0048] S200: Measure the signal frequency of the first chip to be repaired and obtain the first measurement value;

[0049] S300: Archive the first chip to be repaired based on the first measurement value, and repair the first chip to be repaired according to the repair prediction value of the grade where the first chip to be repaired is located, and record the number of repairs.

[0050] In this step, the signal frequency of the B-type chip is measured to be 135.812 kHz using a testing machine. This frequency is then used as the first measurement value, i.e., 135.812 kHz. Table 1 categorizes the different performance levels of the B-type chip, as shown below:

[0051] Table 1

[0052] gear Initial measurement frequency Predicted value symbol Predicted value <![CDATA[L1]]> ≤122KHz <![CDATA[L1:S n ]]> -12.40 <![CDATA[L2]]> 122~123KHz <![CDATA[L2:S n ]]> -9.39 <![CDATA[L3]]> 123~124KHz <![CDATA[L3:S n ]]> -7.50 <![CDATA[L4]]> 124~125KHz <![CDATA[L4:S n ]]> -5.84 <![CDATA[L5]]> 125~126KHz <![CDATA[L5:S n ]]> -4.16 <![CDATA[L6]]> 126~127KHz <![CDATA[L6:S n ]]> -2.37 <![CDATA[L pass ]]> 127~128KHz <![CDATA[L pass ;S pass ]]> 0.00 <![CDATA[L8]]> 128~129KHz <![CDATA[L8:S n ]]> 2.45 <![CDATA[L9]]> 129~130KHz <![CDATA[L9:S n ]]> 4.18 <![CDATA[L 10 ]]> 130~131KHz <![CDATA[L 10 :S n ]]> 5.76 <![CDATA[L 11 ]]> 131~132KHz <![CDATA[L n :S n ]]> 7.51 <![CDATA[L 12 ]]> 132~133KHz <![CDATA[L 12 :S n ]]> 9.28 <![CDATA[L 13 ]]> 133~134KHz <![CDATA[L 13 :S n ]]> 10.83 <![CDATA[L 14 ]]> 134~135KHz <![CDATA[L 14 :S n ]]> 12.81 <![CDATA[L 15 ]]> 135~136KHz <![CDATA[L 15 :S n ]]> 14.58 <![CDATA[L 16 ]]> 136~137KHz <![CDATA[L 16 :S n ]]> 16.33 <![CDATA[L 17 ]]> 137~138KHz <![CDATA[L 17 :S n ]]> 17.28 <![CDATA[L 18 ]]> ≥138KHz <![CDATA[L 18 :S n ]]> 19.28

[0053] According to Table 1, the chip to be repaired can be classified into category 15. If the chip to be repaired is the 2014th chip in this category, then the repair prediction value L of the 2014th chip in category 15 in Table 1 can be used. 15 :S 2014 The corresponding adjustment prediction value of 14.58 is used to adjust the chip to be adjusted. Specifically, when the tester adjusts the chip, it rounds the adjustment prediction value (rounding refers to taking the integer part; there is also a method of rounding the decimal part of 14.58 to 15, which is more accurate, but will execute the handshake protocol once more, and the handshake protocol execution time will be longer, so we consider taking the integer part directly here) to get 14. Therefore, the tester and the chip perform the handshake protocol B 14 times in a row, that is, the number of adjustments to the chip to be adjusted is recorded as 14.

[0054] It should be noted that the above-mentioned correction prediction values ​​are set based on the principle that the closer they are to the average number of corrections for that gear, the better. However, this data usually comes from the analysis of measured data, such as... Figure 3 , Figure 4 As shown, however, the data fluctuates after the first batch of tests and after changing batches, and the analysis results may not be accurate. Therefore, a prediction algorithm is used here to update the revised prediction values. By utilizing the learning and adjustment capabilities of the algorithm, it can be adapted to each batch of products. In the absence of statistical data on the products, the prediction values ​​of each grade can be set to 0 first, and then the algorithm will automatically adjust them.

[0055] It should be further noted that, in the absence of statistical data on previously tested chips (if this particular chip model has not been tested before), the initial predicted values ​​for each level of the program can be set to 0. The algorithm will then adjust these values ​​from 0 to more accurate predictions. The specific process is as follows: Figure 5 As shown.

[0056] S400: Measure the signal frequency of the first chip to be tuned after tuning, and obtain the second measurement value;

[0057] After adjustment, the signal frequency of the chip to be adjusted decreased compared to the first measured value of 135.812KHz. At this time, the signal frequency of the chip was measured again by the tester, and the measured frequency was 127.6KHz (this data is only for illustrative purposes and may be other frequencies). This value was then used as the second measured value.

[0058] S500: The first chip to be repaired is tested based on the comparison between the second measured value and the measurement standard value. If the test is passed, the first chip to be repaired is determined to be a good product. Based on the number of repairs of the first chip to be repaired, the repair prediction value of the second chip to be repaired in the same grade as the first chip to be repaired is obtained, and the repair process is terminated.

[0059] In this step, since the second measured value is located at the measurement standard value L pass The frequency range is between 127-128kHz, therefore the chip to be repaired is tested as a good product and passes the test. Simultaneously, the correct handshake repair count y of the chip to be repaired is recorded. 2014 (i.e., the modification number 14 recorded in step S300) The predicted number of adjustments for the 2015th chip, which is also in the 15th grade, can be obtained using the exponential smoothing-based prediction method. The exponential smoothing-based prediction method is a time series analysis prediction method developed based on the moving average method. Its characteristic is that it assigns different weights to past observations, that is, the weight of more recent observations is greater than the weight of more distant observations. Since the actual measured values ​​and adjustment values ​​of chips of the same model and batch are only affected by the chip's inherent quality during the measurement process, and the test data is independent of the test time and sequence, the time series of adjustment data for chips in the same grade will not have obvious trend changes. The exponential smoothing prediction algorithm only needs to be used once for prediction.

[0060] The prediction method based on exponential smoothing is specifically expressed as follows:

[0061] S n+1 =ay n +(1-a)S n

[0062] Among them, S n It is the predicted value for period n obtained by using the exponential smoothing algorithm for period n-1; S n+1 It is the predicted value for period n+1 obtained by using the exponential smoothing algorithm for period n; y n It is the actual observed value in period n; a is the smoothing constant, and its value range is [0,1].

[0063] This prediction method involves calculating the exponential smoothing value S. n+1 (The predicted value for the next period), combined with a certain time series forecasting model, is used to predict the future. The principle is that the predicted value S for the next period... n+1 These are all actual observed values ​​y from this period. n Compared with the forecast value S for this period n The weighted summation is given. In the formula, the smoothing constant *a* affects the algorithm's prediction convergence speed. A larger smoothing constant *a* results in greater weight for recent values, significantly influencing the predicted value for the next period and increasing the convergence speed. However, an excessively large smoothing constant *a* leads to larger fluctuations in the predicted value and a decrease in prediction accuracy. Since the number of chips in actual chip testing is large and the data distribution is relatively regular, selecting a smaller smoothing constant *a* is more beneficial for adjusting the predicted values ​​during testing.

[0064] The calculation process for obtaining the predicted number of adjustments for the 2015th chip in the 15th stage, based on the above formula, is as follows:

[0065] S 2014+1 =ay 2014 +(1-a)S 2014 ),

[0066] S 2015 =0.1*y 2014 +(1-0.1)S 2014 (Here, the smoothing constant a is set to 0.1 as an example)

[0067] S 2015 =0.1*14 + 0.9*14.58

[0068] =14.522

[0069] At this point, L is calculated. 15 The predicted number of adjustments for the 2015th chip is S. 2015 =14.522.

[0070] Since the 2014th chip has been adjusted 14 times, the corresponding adjustment prediction value S 2014 =14.58, and S was obtained after calculation. 2015 =14.522, S 2015 =14.522 is the predicted value for adjusting the 2015th chip in this batch. Therefore, it can be seen that after adjusting one chip, the predicted value of the next chip updated by the prediction algorithm will be adjusted according to the latest chip, and it will approach the number of adjustments (14) of the new chip. This reflects the ability of the prediction method based on exponential smoothing to automatically adjust and adapt to the new batch. From Figure 5 As can be seen, the chips in this price range require 14 or 15 adjustments, with a very small number requiring 16 adjustments. The average number of adjustments is calculated to be 14.62. With an initial adjustment value (default predicted value S0) of 0, the predicted value can approximate the actual value after testing approximately 30 chips. After the batch of data converges, the error between the predicted and actual values ​​is 0.58. This error mainly stems from the inclusion of 14 and 15 adjustments for this chip price range, with an average prediction value of 14.58. Rounding the predicted value yields 14 adjustments. Therefore, using an exponential smoothing-based prediction algorithm in the actual test can reduce the data measurement time for this chip price range by 13 times, significantly reducing the time spent on chip adjustments and improving overall testing efficiency.

[0071] In another embodiment, the method further includes the following steps:

[0072] S600: In step S500, if the detection fails, the first chip to be repaired will be re-detected based on whether the number of repairs to be performed on the first chip to be repaired has reached the upper limit of the number of repairs set in step S100.

[0073] In this embodiment, if the second measurement value of the first chip to be adjusted is 128.2KHz, which exceeds 127-128KHz, the detection fails. At this time, since the chip has been adjusted 14 times, which is less than the upper limit of the number of adjustments (30 times) set in step 1, the adjustment can continue.

[0074] In another embodiment, the method further includes the following steps:

[0075] S700: In step S600, if the number of repairs to the first chip to be repaired reaches the upper limit of the number of repairs, the first chip to be repaired is determined to be defective and the repair process is terminated; if the number of repairs to the first chip to be repaired does not reach the upper limit of the number of repairs, the second measurement value of the first chip to be repaired is compared with the measurement standard value again, and the corresponding handshake protocol is executed according to the comparison result.

[0076] In this embodiment, since the measured value of 128.2 kHz is greater than the standard measurement value of 127-128 kHz, it can be known that the actual signal frequency of the chip is higher than the target value. Therefore, the signal frequency output by the chip needs to be adjusted to a lower frequency. At this time, a handshake protocol B is required to reduce the chip output frequency (there are three handshake protocols during chip tuning: protocol A, protocol B, and protocol C. Protocol A indicates frequency adjustment to higher, that is, if the measured frequency signal is lower than the target value, protocol A is executed to adjust the chip output signal frequency to higher; protocol B indicates frequency adjustment to lower, that is, if the measured signal frequency is higher than the target value, protocol B is executed to adjust the chip output signal frequency to lower; protocol C indicates no frequency adjustment, that is, if the measured signal frequency is within the target value range, protocol C is executed to stop adjusting the chip output signal frequency).

[0077] In another embodiment, the method further includes the following steps:

[0078] S800: Update the number of adjustments to the first chip to be adjusted according to the corresponding handshake protocol to obtain the actual number of adjustments to the first chip to be adjusted.

[0079] In this embodiment, since the above embodiment executes handshake protocol B, the number of adjustments should be increased by 1 from the original 14. Therefore, the number of adjustments after the chip is updated, i.e. the actual number of adjustments, is 15.

[0080] In another embodiment, the method further includes the following steps:

[0081] S900: Measure the signal frequency of the first chip to be adjusted after the number of adjustments has been updated, and obtain the third measurement value.

[0082] In another embodiment, the method further includes the following steps:

[0083] S1000: The third measured value is compared with the measurement standard value to perform a third test on the first chip to be repaired. If the test passes, the first chip to be repaired is determined to be a good product. Based on the actual number of repairs of the first chip to be repaired, the repair prediction value of the second chip to be repaired in the same grade as the first chip to be repaired is obtained, and the repair procedure is terminated.

[0084] In another embodiment, the method further includes the following steps:

[0085] S1100: In step S1000, if the third test fails, then steps S600 to S1000 are repeated.

[0086] Thus, all the above embodiments constitute the complete technical solution of this disclosure. Compared with existing methods, the method of this disclosure has the following technical effects:

[0087] 1. It can improve editing efficiency and reduce editing time;

[0088] Figure 2 The chip tuning method shown requires a measurement comparison after each handshake signal adjustment; however, this disclosure predicts the number of chip tuning cycles using an algorithm, as described in the above embodiments, and performs continuous handshake adjustments based on the predicted number of cycles, reducing the number of measurements performed after each handshake signal. In the example, for L... 15 A certain chip in the chip is divided into segments after the first frequency measurement, and the predicted number of adjustments is 14.58. Based on the predicted number of adjustments, the handshake signal B is executed 14 times in a row. During this process, the frequency is measured only once before and after the handshake signal is executed, and a total of 2 frequency measurements are required. However, the existing technology requires frequency measurement and comparison after each handshake signal execution, which requires a total of 15 frequency measurements.

[0089] In this example, this method can reduce the number of measurement operations for this chip by 13. The tuning time for this chip model is recorded as follows: one frequency measurement takes 12ms, and one handshake protocol execution takes 18ms. Using existing technology for L... 15 The time required for 14 adjustments to the chip is the sum of 15 frequency measurement times (12*15=180ms) and 14 handshake protocol times (18*14=252ms), totaling 434ms. Using this method, the time required is the sum of 2 measurement times (12*2=24ms) and 14 handshake protocol execution times (18*14=252ms), totaling 276ms. Therefore, the reduction in adjustment time for this chip is [percentage missing]. This can significantly improve the adjustment efficiency of this gear.

[0090] The method described in this disclosure, when applied to chip tuning of configuration registers, can achieve the same tuning effect as 14 tuning operations by using a prediction method, requiring only two measurement times and one configuration register time, thus reducing the time spent on 13 data measurements and 13 configuration register operations. This is achieved by predicting the chip's tuning value S. n The configuration register is used for coarse adjustment to bring the chip's actual measured value close to the target value. Then, the chip is fine-tuned an average of 0.58 times. The reduction in adjustment time for this chip at this level can be calculated. By analogy, it was found that the absolute error between the predicted adjustment value and the actual adjustment value of the other positions of the chip using the prediction algorithm does not exceed 1. It can be seen that the chip that uses the prediction algorithm to configure the register for adjustment has an average of no more than two adjustment times during the test process. This significantly improves the chip adjustment efficiency and reduces the production cost while ensuring the adjustment effect.

[0091] For tuning chips, the tuning time optimization effect is affected by the number of tunings required for the chip. The more tunings required, the more time will be reduced using this method. The reduction ratio is affected by the measurement frequency. Measuring high-frequency signals takes less time, and the lower the measurement frequency, the greater the reduction in tuning time.

[0092] 2. It can reduce the number of manual data collection steps;

[0093] Existing methods for establishing pre-repair survey tables based on statistical data from tested chips require further processing when inspecting the first wafer or the first batch of chips. Figure 2 The current program structure uses conventional methods for adjustment. Only after one batch of tests is completed, the test data is processed, a lookup table relating the initial measurement frequency to the number of adjustments is created, and the program calls the lookup table, can subsequent batches of tests continue. Therefore, the existing method is complex and cumbersome, requiring a pause between the first batch of tests and subsequent batches to process the data, which affects testing efficiency.

[0094] This method uses an exponential smoothing prediction algorithm to automatically adjust during testing, and its prediction effect on the number of adjustments for B-type chips is as follows: Figure 5 As shown, this data was obtained from actual chip measurements, by selecting a batch with an initial measurement frequency of L. 15 The actual number of adjustments to the chip (135-136kHz) was determined. The smoothing coefficient 'a' was set to 0.1, and the initial default prediction value S0 was set to 0. The exponential smoothing prediction algorithm was used for verification. This was based on data analysis and... Figure 5It can be seen that the chips in this range require 14 or 15 adjustments, with a very small number of chips requiring 16 adjustments, resulting in an average of 14.62 adjustments. After the data prediction converged for this batch, the average prediction value was 14.58, with an error of 0.58 between the predicted and actual values. This error mainly stems from the fact that the chips in this range underwent 14 or 15 adjustments. After testing approximately 30 chips, the predicted value can approach the actual value.

[0095] The predicted values ​​are updated in real time by recording the adjustment data during the test, which reduces the steps of manually organizing and looking up tables. The test does not require pausing or modifying the program, making it more convenient and easier to use.

[0096] 3. Automatic adjustment for each batch of products enables the predicted values ​​to converge to the optimal range.

[0097] Because chips can vary in quality due to manufacturing processes and other issues, even with the same initial measurements, the number of adjustments required for each batch may differ. Therefore, using a lookup table based on test data from a single batch is not entirely suitable for all chip batches, as the lookup table may be biased, leading to inaccurate predictions. This method, however, updates the predicted values ​​based on the measured data during testing and adjusts the predicted values ​​to converge with the new batch after a chip batch change, resulting in more accurate predictions and improving the prediction performance across different batches.

[0098] In another embodiment, this disclosure also provides a computer storage medium, comprising:

[0099] Memory is used to store multiple computer instructions;

[0100] A processor is used to execute computer instructions to implement the methods described in any of the preceding methods.

[0101] In another embodiment, this disclosure also provides an electronic device, including:

[0102] Memory, processor, and computer programs stored in memory and executable on the processor, wherein,

[0103] When the processor executes the program, it implements the method described in any of the preceding methods.

[0104] The above description, using specific embodiments, is merely for the purpose of aiding understanding and is not intended to limit the scope of this disclosure. Any modifications or substitutions made by those skilled in the art within the scope of the technology disclosed herein should be included within the scope of this disclosure.

Claims

1. A chip trimming prediction method, comprising the following steps: S100: Set the maximum number of adjustments required for the first chip to be adjusted; S200: Measure the signal frequency of the first chip to be repaired and obtain the first measurement value; S300: Archive the first chip to be repaired based on the first measurement value, and repair the first chip to be repaired according to the repair prediction value of the grade where the first chip to be repaired is located, and record the number of repairs. in, The adjusted forecast value is obtained using a forecasting method based on exponential smoothing. The prediction method based on exponential smoothing is specifically expressed as follows: S n+1 =is n +(1-a)S n， Among them, S n It is the predicted value for period n obtained by using the exponential smoothing algorithm for period n-1; S n+1 It is the predicted value for period n+1 obtained by using the exponential smoothing algorithm for period n; y n It is the actual observed value in period n; a is the smoothing constant, and its value range is [0,1]. S400: Measure the signal frequency of the first chip to be tuned after tuning, and obtain the second measurement value; S5 00: The first chip to be repaired is tested by comparing the second measured value with the measurement standard value. If the test is passed, the first chip to be repaired is determined to be a good product. Based on the number of repairs of the first chip to be repaired, the repair prediction value of the second chip to be repaired in the same grade as the first chip to be repaired is obtained, and the repair process is terminated.

2. The method according to claim 1, wherein, In step S500, the adjustment prediction value of the second chip to be adjusted is obtained by an exponential smoothing-based prediction method.

3. The method according to claim 1, wherein, The method further includes the following steps: S600: In step S500, if the detection fails, the first chip to be repaired will be re-detected based on whether the number of repairs to be performed on the first chip to be repaired has reached the upper limit of the number of repairs set in step S100.

4. The method according to claim 3, wherein, The method further includes the following steps: S700: In step S600, if the number of repairs to the first chip to be repaired reaches the upper limit of the number of repairs, the first chip to be repaired is determined to be defective, and the repair process is terminated. If the number of adjustments to the first chip to be adjusted has not reached the upper limit, the second measurement value of the first chip to be adjusted is compared with the measurement standard value again, and the corresponding handshake protocol is executed according to the comparison result.

5. The method according to claim 4, wherein, The method further includes the following steps: S800: Update the number of adjustments to the first chip to be adjusted according to the corresponding handshake protocol to obtain the actual number of adjustments to the first chip to be adjusted.

6. The method according to claim 5, wherein, The method further includes the following steps: S900: Measure the signal frequency of the first chip to be adjusted after the number of adjustments has been updated, and obtain the third measurement value.

7. The method according to claim 6, wherein, The method further includes the following steps: S1000: The third measured value is compared with the measurement standard value to perform a third test on the first chip to be repaired. If the test passes, the first chip to be repaired is determined to be a good product. Based on the actual number of repairs of the first chip to be repaired, the repair prediction value of the second chip to be repaired in the same grade as the first chip to be repaired is obtained, and the repair procedure is terminated.

8. The method according to claim 7, wherein, The method further includes the following steps: S1100: In step S1000, if the third test fails, then steps S600 to S1000 are repeated.

9. A computer storage medium, comprising: Memory is used to store multiple computer instructions; A processor for executing computer instructions to implement the method of any one of claims 1 to 8.

10. An electronic device, comprising: Memory, processor, and computer programs stored in memory and executable on the processor, wherein, When the processor executes the program, it implements the method described in any one of claims 1 to 8.

Images