WO2022213877A1

WO2022213877A1 - Loop iteration calibration method and test device using same

Info

Publication number: WO2022213877A1
Application number: PCT/CN2022/084406
Authority: WO
Inventors: 徐韡; 王贺春
Original assignee: 爱德万测试股份有限公司
Priority date: 2021-04-08
Filing date: 2022-03-31
Publication date: 2022-10-13
Also published as: CN115201735A

Abstract

A loop iteration calibration method and a test device (10) using same. The loop iteration calibration method comprises: an inner loop step (S10) and an outer loop step (S20). The inner loop step (S10) comprises: adjusting input power of a first known good unit by means of a first gradient algorithm according to target power and output power, and then adjusting the input power by means of a second gradient algorithm, the step size of the first gradient algorithm being greater than the step size of the second gradient algorithm; and taking a first average value for the input power adjusted by means of the second gradient algorithm. The outer loop step (S20) comprises: performing the inner loop step (S10) by using the second known good unit to obtain a second average value of the second known good unit; and averaging the first average value and the second average value to obtain a final average value, and calibrating the test device (10) according to the final average value. The effects of the present invention comprise: the number of tests can be reduced, the efficiency of calibration can be improved, and errors due to difference in working performance of multiple known good units or a random error of a single known good unit can be reduced.

Description

Cyclic iterative calibration method and test equipment using the same

This application claims the priority of Chinese patent application 202110379301X with the filing date of 2021/4/8. This application cites the full text of the above Chinese patent application.

technical field

The invention relates to a related technology of chip unit testing, in particular to a cyclic iterative calibration method and testing equipment using the same.

Background technique

In the production process of the chip unit, it is necessary to screen out the chip unit that meets the specification through the testing equipment. However, test equipment usually has errors due to minor environmental differences caused by factors such as interface load board, test card, hardware aging, or temperature. Therefore, it is necessary to use known good products that have been tested and meet the specifications After the unit (known good unit, KGU) calibrates the test equipment, the test screening of the chip unit is carried out.

For a power amplifier (PA) of a communication chip unit, when its power is located at the intersection of the saturation region and the linear region, the linearity changes greatly, so it is not easy to calibrate or test. In addition, the performance of multiple known-good units will not be exactly the same, and even a single known-good unit has random errors. Furthermore, if the input power is too large or the time is too long during testing or calibration, the temperature of the chip will rise, resulting in changes in working performance. The above-mentioned uncertain factors will lead to inefficiencies in calibration and testing, and create risks in production, so there is a technical need.

SUMMARY OF THE INVENTION

An object of the present invention is to provide a cyclic iterative calibration method and a test equipment using the same, which can effectively improve the efficiency of calibration.

Embodiments of the present invention provide a cyclic iterative calibration method and test equipment using the same. The test equipment includes an interface carrier board, a test card, and a workstation. Known good cells are provided on the interface carrier board. A workstation is used to execute a computer program to perform the cyclic iterative calibration method. The loop iteration calibration method includes an inner loop step and an outer loop step. The inner loop step includes: applying input power to the first known good unit through the workstation and the test card, and measuring the output power of the first known good unit; the workstation adjusts the input power through the first gradient algorithm according to the target power and the output power, Then, the input power is adjusted through a second gradient algorithm, and the step size of the first gradient algorithm is larger than that of the second gradient algorithm; and the workstation takes a first average value for the input power adjusted through the second gradient algorithm. The outer circulation step includes: performing the inner circulation step with the second known good unit to obtain a second average value of the second known good unit; and the workstation averages the first average value and the second average value to obtain a final average value , and calibrated against the final average.

In an embodiment provided by the present invention, the inner loop step further includes: when no input power is applied, maintaining the operating voltages of the first known good cell and the second known good cell through the test card, and turning off the first known good cell cell and amplifier module in a second known good cell.

In an embodiment provided by the present invention, the first gradient algorithm is an accelerated gradient algorithm, and the second gradient algorithm is a fixed gradient algorithm.

In an embodiment provided by the present invention, in the inner loop step, the workstation adjusts the input power through the first gradient algorithm for no more than three times according to the target power.

In an embodiment provided by the present invention, in the inner loop step, the workstation obtains a first average value and a second average value for the input power after being adjusted for a predetermined number of times by the second gradient algorithm.

In the embodiment of the present invention, the first gradient algorithm is used to adjust the input power applied to the known good cells, and then the input power is adjusted by the second gradient algorithm. Approaching the target input power reduces the number of tests and improves the test efficiency. In addition, the embodiment of the present invention adopts the method of averaging the input power, which can reduce the error caused by the difference in working performance among multiple known good units or the random error of a single known good unit.

The above description is only an overview of the technical solutions of the present invention, in order to be able to understand the technical means of the present invention more clearly, it can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand , the following specific preferred embodiments, and in conjunction with the accompanying drawings, are described in detail as follows.

Description of drawings

FIG. 1 is a schematic diagram of a testing device provided by an embodiment of the present invention.

FIG. 2 is a schematic flowchart of a loop iterative calibration method provided by an embodiment of the present invention.

FIG. 3 is a schematic diagram of a power consumption current measurement value in the cyclic iterative calibration method provided by an embodiment of the present invention.

FIG. 4 is a schematic diagram of output power versus input power in a cyclic iterative calibration method provided by an embodiment of the present invention.

FIG. 5 is a schematic diagram of a waveform of a power consumption current in a cyclic iterative calibration method provided by an embodiment of the present invention.

Detailed ways

FIG. 1 is a schematic diagram of a testing device provided by an embodiment of the present invention. In this embodiment, the testing equipment 10 , which may also be referred to as a tester (tester), includes an interface carrier board 101 , a tester header 102 , and a workstation 103 . The test head 102 of the test equipment 10 may have various test cards, such as a device power supply card 1021 (device power supply card), a radio frequency test card 1022 (RF card), and a digital control test card 1023 (digital card), etc. Limited to this, each test card is used for different test tasks. The test head 102 is electrically connected to the device under test (DUT) through the interface carrier board 101, for example, the wiring of the interface carrier board 101 is electrically connected to the probes of the test card and the pins of the device under test, so as to connect the test equipment 10 to the pins of the device under test. Resources, such as current, voltage, frequency, etc., are delivered to the device under test. When the test equipment 10 is calibrated, a known-good unit KGU is used as the device under test. The known-good unit KGU is, for example, a chip unit that has been tested in a laboratory and passed the specification. Figure 1 takes 4 known good cells KGU as an example, but is not limited to this. The workstation 103 is, for example, a computer using a Windows or Linux system and having a memory, a processor, a display card, a screen, a human-machine interface, etc., and the processor executes the computer program stored in the memory to perform the cyclic iterative calibration method, but is not limited to this. In addition, it needs to be supplemented that the original known-good unit KGUs are usually in small quantities and cannot support the mass production requirements of automatic test equipment (ATE) (usually the number of known-good unit KGUs provided by customers is <10pcs ). As mass production begins, these original numbers of known-good unit KGUs cannot support the correction requirements of multiple testers, and the loss of known-good unit KGUs during normal mass production also needs to be considered. Therefore, in the case of ensuring that the original known good unit KGU characteristics/data are not offset, the known good unit KGU with a small original number is expanded/collected in a mass production environment to a certain number that can be applied to large-scale mass production A known good unit KGU is very necessary. Therefore, a high-consistency KGU expansion/collection method is required, including step 1. Before mass production is officially released, a certain position (site) of a certain carrier board (LB) needs to be determined as a reference (the corresponding position of the carrier board should be It has been confirmed in the laboratory that there is no problem and agreed with the customer) All subsequent steps will use this position; Step 2. Use the loop iterative correction method defined in the present invention to test the original known good unit KGU; Step 3. Pass the defined by the present invention The loop iterative correction method obtains the compensation value and uses it in the program; Step 4. Test the 500-1000pcs mass-produced chip unit in advance to obtain the overall characteristics of the current lot; Step 5. Use script or other tools to obtain the overall lot according to Step 4. characteristics, tightening the limits of the current program; and step 6. Re-test the entire lot and get a known-good unit KGU with high consistency. In step 5 (tightening the limit range), set reasonable settings according to the number of key test items to be compared. value Max/Min) to tighten the new limit range by 50%. If there are a total of 20 key test items to be compared, the number of known good unit KGUs produced in the final may be greatly reduced (due to the previously mastered The current lot mean characteristic, so it is not exponentially decreasing, but it may still decrease substantially). In step 6 (final KGU output), the result will be affected by the new limit setting in step 5 (how many test items need to tighten the limit and how much the limit needs to be tightened), through appropriate adjustment settings , it is recommended that the final output of KGU should be kept below 10%. If the final output of KGU is too high, it is necessary to increase the number of KGUs in the formal mass production correction program to eliminate errors caused by insufficient KGU consistency. Of course, the above combination is not fixed, and some experiments need to be done in the early stage of mass production to weigh the relationship between KGU yield and KGU consistency.

FIG. 2 is a schematic flowchart of a loop iterative calibration method provided by an embodiment of the present invention. FIG. 3 is a schematic diagram of a power consumption current measurement value in the cyclic iterative calibration method provided by an embodiment of the present invention. Please refer to FIG. 2 and FIG. 3 together. In this embodiment, as shown in FIG. 2 , the cyclic iterative calibration method performed by the testing device 10 includes an inner loop step S10 and an outer loop step S20 . Fig. 3 shows the calibration records CL1, CL2 and CL3 of the three inner loop steps S10, the horizontal axis represents the measurement of each known good unit KGU; the vertical axis represents the power consumption current measurement of the known good unit KGU The difference between the value and the target value of the current consumption, hereinafter referred to as ICC delta, in milliamps (mA). In this embodiment, the target power consumption current of the known good unit KGU is used as a reference for the target power, so as to calibrate the input power of the test equipment 10 to the known good unit KGU. As can be seen from Figure 3, as the number of adjustments of the input power increases, the ICC delta gradually decreases. However, the present invention is not limited thereto, and the test equipment 10 may also be calibrated according to the voltage of the known good cell KGU or the reference voltage, current or combination of different circuit nodes as the reference of the target power.

FIG. 4 is a schematic diagram of output power versus input power in a cyclic iterative calibration method provided by an embodiment of the present invention. Please refer to Figure 3 and Figure 4 together. It can be seen in the calibration record CL3 shown in Figure 3 that after the 4th measurement, the ICC delta oscillates around plus or minus 20mA. This is due to the input of the known good unit KGU. At higher powers, it is known that the good unit KGU operates in a nonlinear region (as shown in Figure 4, near point P2). Compared with operating in the linear region (as shown in Figure 4, near point P1), the power consumption current of the known good unit KGU operating in the nonlinear region is very sensitive and unstable, and even a slight change in the input power will affect the power consumption. It is known that the current consumption of a good cell KGU has a large effect.

Please continue to refer to Figure 2. In this embodiment, the inner loop step S10 includes steps S101 to S107. Next, steps S101 to S107 will be described.

The count value of the number of measurements is incremented by 1 (step S101).

The input power is applied to the known-good unit KGU through the workstation 103 and the test head 102, and the output power of the known-good unit KGU is measured (step S103).

When the number of times of adjusting the input power (for example, the value of the count value of the measurement times minus 1 or the number of times of recording the adjustment of the input power) is not more than 3 times, the workstation 103 adjusts the input power through the first gradient algorithm according to the target power and the output power; When the number of times of adjusting the input power is greater than three times, the workstation 103 adjusts the input power through the second gradient algorithm according to the target power and the output power, wherein the step size of the first gradient algorithm is larger than that of the second gradient algorithm (step S105 ) . In detail, as shown in Figure 3, the ICC delta is relatively large from the first to the fourth measurement, that is, in the process of adjusting the input power in the first few times, the fluctuation of the output power accounts for the smaller part of the ICC delta. It can be expected that the direction of adjusting the input power is unchanged in the first few times, that is, in the first few times, the input power needs to be increased or the input power is decreased. Therefore, the input power is adjusted by the first gradient algorithm with a larger step size. The step size refers to the parameters such as step size or learning rate used to determine the update amount of the current input power and the next input power. Then, as shown in Figure 3, the ICC delta is small in the fifth measurement and later, and the fluctuation of the output power begins to account for the main part of the ICC delta. Therefore, the input power is adjusted by the second gradient algorithm with a smaller step size. Wherein, the first gradient algorithm and the second gradient algorithm are, for example, accelerated gradient algorithms, such as algorithms such as Nesterov Accelerated Gradient, and the step size (learning rate) used by the first gradient algorithm is greater than the step size (learning rate) of the second gradient algorithm; Alternatively, the first gradient algorithm and the second gradient algorithm are, for example, a fixed gradient algorithm, wherein the second gradient algorithm adjusts the input power with a step size (learning rate) of 0.15dBm, and the step size used by the first gradient algorithm is greater than 0.15dBm, for example is twice that of the second gradient algorithm; alternatively, the first gradient algorithm and the second gradient algorithm can also be a combination of the above algorithms, for example, the first gradient algorithm is an accelerated gradient algorithm, and the second gradient algorithm is a fixed gradient algorithm, but not limited to this. In short, for fast convergence, a search with a larger step size is implemented in the first few loop iterations, and the next search is set by the slope of this time and the result of the previous one. It is expected that the KGU of known good cells should still be Works in the linear region and is not too affected by temperature from the start. While the search with smaller step size is performed after the search with larger step size (usually 3 steps are required), there are two reasons for this: a. Multiple searches may cause a significant increase in the KGU heat/temperature of known good cells, which Will lead to known good cell KGU performance degradation, and the search may exceed expectations and will no longer be suitable; b. The close target is usually at the boundary of the linear and saturated regions, making the pass rate too low, and the adjustment method is not suitable for edge cells. part. If the search is not correct, even if you increase the calibration time, you may oscillate and never converge. So by combining them for faster and more stable/repeatable calibration.

When the count value of the number of measurements is less than 20 times, return to step S101; when the count value of the number of measurements is equal to 20 times, the workstation 103 resets the count value of the number of measurements to zero, and filters the input power after the predetermined number of times adjusted by the second gradient algorithm And take the average value (step S107). For example, taking 20mA as the screening condition, and taking 10 adjustments for the above predetermined times as an example, in the calibration records CL1 and CL2 shown in Figure 3, the ICC delta after 10 adjustments is between 20mA, so All input power after 10 times of adjustment will be used for averaging; in calibration record CL3, the ICC delta measured at the 17th time is greater than 20mA, so the 17th input is filtered out of all the input power after 10 times of adjustment Power is not used and the rest of the input power is used for averaging. In other embodiments, it is also possible to return to step S101 when the count value of the measurement times is different, for example, preferably more than 10 times to reduce errors, but not limited to this.

Therefore, although it is known that the good unit KGU may be in different working ranges during calibration and have different output power curves as shown in the calibration records CL1 to CL3 in FIG. 3 , the inner loop step S10 provided by this embodiment passes through The method of taking the average value enables different output power curves to be applied, which reduces the need for manual operation and has good versatility.

Please continue to refer to Figure 2. In this embodiment, the outer loop step S20 includes the following steps: performing the above-mentioned inner loop step S10 on each known good unit KGU to obtain the average value of the input power of each known good unit KGU (step S201 ); The average value of the input power of the known good unit KGU is screened and the final average value is obtained, and calibration is performed according to the final average value (step S203). In detail, after obtaining multiple averages of the input power of all known good unit KGUs, the extreme values or unqualified values in the multiple averages are screened out and not used, and the remaining averages are finally taken. average value.

For example, using 10 known good units KGU to perform the loop iterative calibration method, the outer loop step S20 includes: performing the above inner loop step S10 on the first known good unit KGU to obtain the input of the first known good unit KGU The first average value of power; then perform the above-mentioned inner loop step S10 on the second known good unit KGU to obtain the second average value of the input power of the second known good unit KGU; and so on until the tenth known good unit is obtained. The tenth average value of the input power of the good unit KGU. Then, from the first average to the tenth average, the extreme values with a large gap or the averages that do not meet the conditions are not used, and the remaining averages are taken as the final average. Finally, the workstation 103 is calibrated based on the final average. In other embodiments, it is also possible to use more than 10 known good unit KGUs, such as 20 or more.

FIG. 5 is a schematic diagram of a waveform of a power consumption current in a cyclic iterative calibration method provided by an embodiment of the present invention. During the process of performing the inner loop step S10 on the known good unit KGU, step S103 will be performed multiple times, that is, input power (such as waveform 4) to the known good unit KGU multiple times and measure the output power (power consumption current, such as waveform 1 to 3). In the time interval between the end of inputting power to the known good unit KGU and measuring the output power and before the next input power, such as between time T2 and T3 or between time T4 and T5 as shown in waveform 3, the known good unit KGU There will still be power consumption current for normal operation, and such power consumption current will also heat up the KGU of the known good unit. Therefore, in this embodiment, as shown in waveform 1, the device power test card 1021 is used to maintain the operating voltage VCC of the known good cell KGU, and the digital control test card 1023 is used to send a signal to the control module of the known good cell KGU , to turn off the amplifier module in the known good unit KGU after measuring the output power and before the next input power, thereby avoiding the test error caused by adjusting the working voltage, and reducing the temperature change of the amplifier module and changing the working characteristics. The above manner of turning off the amplifier module in the known-good unit KGU is not limited to this, and the amplifier module in the known-good unit KGU may be turned off externally through the interface carrier board 101 or the test head 102 or the like. In addition, it is not necessary to turn off the amplifier module in the known-good unit KGU every time the input power ends, it can also be selectively turned off only after a certain number of input power ends, such as waveform 1 shown in Figure 5, every time Turn off the power only once twice. In other embodiments, it may be turned off when the input power is adjusted using the first gradient algorithm, and not turned off when the input power is adjusted using the second gradient algorithm. In other embodiments, when there is no known good unit KGU input power, the waveform 2 shown in FIG. 5 can also be used, and the power supply VCC provided to the known good unit KGU is turned off or reduced by the device power test card 1021 , to reduce heating.

It should be noted that, in other embodiments, different times of adjusting the input power may also be used as the basis for the switching algorithm. For example, the number of times of adjusting the input power using the first gradient algorithm may also be less than 3 times or more than 3 times. In addition, the step size of the first gradient algorithm or the second gradient algorithm can also be adjusted. For example, the number of times the input power is adjusted using the first gradient algorithm is decreased and the step size is increased, or the number of times the input power is adjusted using the first gradient algorithm is increased and the step size is decreased. In addition, the number of times may not be used as the basis of the switching algorithm. For example, in step S105, when the measured output power approaches the target value and falls within a range (for example, referring to FIG. 3, when the ICC delta is less than 40mA), the next time The adjustment switches to adjust the input power using the second gradient algorithm. It should be noted that the basis of the above-mentioned switching gradient algorithm, the step size used by the gradient algorithm, and the predetermined number of times used for averaging can all be adjusted by the workstation 103 according to the measured results or actual needs, for example, by modifying the parameters used by the computer program. Or rewrite computer programs.

To sum up, the embodiment of the present invention adopts the first gradient algorithm to adjust the input power applied to the known good cells, and then adjusts the input power through the second gradient algorithm, because the step size of the first gradient algorithm is larger than that of the second gradient algorithm. It can quickly approach the target input power to reduce the number of tests and improve the efficiency of the test. In addition, the embodiment of the present invention adopts the method of averaging multiple input powers adjusted by a second gradient algorithm with a small step size for a single known good unit, which can avoid fluctuations caused by the known good unit working in a nonlinear interval . In addition, in the embodiment of the present invention, the average value of the input power of each known good unit is taken as the final average value, which can reduce the working performance difference between multiple known good units or the random error of a single known good unit. error generated.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention in any form. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. The technical personnel, within the scope of the technical solution of the present invention, can make some changes or modifications to equivalent examples of equivalent changes by using the methods and technical contents disclosed above, provided that the content of the technical solution of the present invention is not departed from, Any simple modifications, equivalent changes and modifications made to the above embodiments according to the technical essence of the present invention still fall within the scope of the technical solutions of the present invention.

Claims

A test equipment using a cyclic iterative calibration method, characterized in that the test equipment includes an interface carrier board, a test card, and a workstation, the interface carrier board is provided with known-good cells, and the workstation is used to execute a computer a program to perform the cyclic iterative calibration method, the cyclic iterative calibration method comprising:

Inner loop steps, including:

Apply input power to the first known-good unit through the workstation and the test card, and measure the output power of the first known-good unit;

The workstation adjusts the input power through a first gradient algorithm according to the target power and the output power, and then adjusts the input power through a second gradient algorithm, where the step size of the first gradient algorithm is larger than that of the second gradient the step size of the algorithm; and

the workstation takes a first average of the input power adjusted by the second gradient algorithm; and

Outer loop steps, including:

performing the inner recycle step with a second known good unit to obtain a second average value of the second known good unit; and

The workstation averages the first average value and the second average value to obtain a final average value, and performs calibration according to the final average value.
The test apparatus of claim 1, wherein the inner loop step further comprises: maintaining the first known good cell and the second known good cell by the test card when the input power is not applied The operating voltage of the known good cell is known, and the amplifier modules in the first known good cell and the second known good cell are turned off.
The test equipment of claim 1, wherein the first gradient algorithm is an accelerated gradient algorithm, and the second gradient algorithm is a fixed gradient algorithm.
The test equipment according to claim 1, wherein, in the inner loop step, the workstation adjusts the input power through the first gradient algorithm for no more than three times according to the target power.
The test equipment according to claim 1, wherein in the inner loop step, the workstation takes the first average value of the input power after being adjusted for a predetermined number of times by the second gradient algorithm and the second average.
A cyclic iterative calibration method, suitable for test equipment, characterized in that the test equipment includes an interface carrier board, a test card, and a workstation, the interface carrier board is provided with a known-good unit, and the workstation is used for performing A computer program to perform the cyclic iterative calibration method, the cyclic iterative calibration method comprising:

Inner loop steps, including:

Apply input power to the first known-good unit through the workstation and the test card, and measure the output power of the first known-good unit;

The workstation adjusts the input power through a first gradient algorithm according to the target power and the output power, and then adjusts the input power through a second gradient algorithm, where the step size of the first gradient algorithm is larger than that of the second gradient the step size of the algorithm; and

the workstation takes a first average of the input power adjusted by the second gradient algorithm; and

Outer loop steps, including:

performing the inner recycle step with a second known good unit to obtain a second average value of the second known good unit; and

The workstation averages the first average value and the second average value to obtain a final average value, and performs calibration according to the final average value.
6. The loop iterative calibration method of claim 6, wherein the inner loop step further comprises: maintaining the first known good cell and the first known good cell through the test card when the input power is not applied The operating voltage of the second known good cell, and the amplifier modules in the first known good cell and the second known good cell are turned off.
The loop iterative calibration method according to claim 6, wherein the first gradient algorithm is an accelerated gradient algorithm, and the second gradient algorithm is a fixed gradient algorithm.
The loop iterative calibration method according to claim 6, wherein in the inner loop step, the workstation adjusts the input power through the first gradient algorithm according to the target power for no more than 3 times Second-rate.
The loop iterative calibration method according to claim 6, characterized in that, in the inner loop step, the workstation takes the first input power for the input power adjusted for a predetermined number of times by the second gradient algorithm average and the second average.