TW202240185A - Method of increasing accuracy of signal acquisition in chip testing equipment - Google Patents

Abstract

A method of increasing accuracy of signal acquisition in chip test equipment includes: selecting ideal transiting point(s); selecting sampling point(s), performing scan test(s) on each sampling point and each ideal transiting point, and storing scan test results; repeatedly performing the previous one step; building a statistical data model; acquiring real transiting point(s); repeatedly performing the previous three steps; calculating overall offset time of the signal; and compensating the overall offset time to a signal input terminal.

Description

A Method of Improving the Accuracy of Grabbing Signals in a Wafer Tester

The invention relates to the technical field of wafer testing machines, in particular to a method for improving the accuracy of grabbing signals in a wafer testing machine.

Since the paths of each channel of the chip testing machine are different, the time for the signal to reach the end of the channel will be different. Therefore, it is necessary to compensate for the slower signal transmission to ensure that the time for the signal to reach the end of the channel is consistent. This compensation needs to accurately capture the position of the signal change, but due to the uncertainty of the signal, the signal will be biased every time it is captured. As shown in Figure 1, when the signal rises or falls, the ideal signal is a straight line. As for the actual signal, as shown in Figure 2, in practical applications, there will be deviations every time the signal changes, which will cause accuracy problems when the repeatability is required to be highly consistent. There are two main reasons for the deviation of the signal in the chip tester every time it changes. One is that the chip tester is limited by the accuracy of the programmable logic array (FPGA) inside the chip tester and the devices on the transmission line, so that the signal actually occurs in the Jitter within a range, this factor is a circuit characteristic, it is difficult to eliminate. Second, because the minimum unit of the signal generated by the chip tester is a fixed value of 100 picoseconds (ps), the time for generating the signal is discontinuous. For example, ideally, rising or falling edges of 10 nanoseconds (ns), 10.1ns, and 10.2ns that are multiples of 100ps can be generated, but rising or falling edges of 10.05ns, 10.15ns, and 10.25ns cannot be generated. And due to the precision limitation of the hardware, 100ps itself is also a variable. The clock pulse of the chip tester system is 2ns, and the signal needs to take 20 steps in a period of 2ns. These 20 steps are different for each step of the signal due to the uncertainty of the smallest unit.

The current traditional method generally uses the average value, but the average value ignores the distribution of each signal, the signal obtained in this way is not optimal, and the position accuracy of capturing signal changes is low.

When capturing the signal, if any measurement result is used as the actual signal of the signal to process, it will inevitably cause deviation. Because this is only one measurement result, if it is measured at other locations and times, it will be another result. If the compensation is handled by randomly grabbing the signal at any time without optimization, when changing different test points to repeat the inspection, the edge placement accuracy (EPA) can only be within +/- -300Ps, and simple average processing, EPA can only reach +/-200ps, poor signal stability.

Therefore, a method for improving the accuracy of capturing signals in a chip testing machine is designed, taking all points in a time period, and performing multi-point and multiple scans on all points to obtain the positions of several real points, and then taking several real points. Averaging is used to cover the signal changes at various positions within a period, fully consider the signal distribution, and obtain the precise position of the signal change for compensation, thereby improving the signal capture accuracy.

In order to overcome the deficiencies of the prior art, the present invention provides a method for improving the accuracy of capturing signals in a wafer testing machine, which takes all points in a time period, and performs multi-point and multiple scans on all points to obtain the positions of several real points. Then take the average of several real points to cover the signal changes at various positions within a period, fully consider the distribution of the signal, and obtain the precise position of the signal change for compensation, thereby improving the signal capture accuracy.

In order to achieve the above object, a method for improving the accuracy of capturing signals in a chip testing machine is designed, including:

S1, taking 100 picoseconds (ps) as the time interval to take 21 points as the 21 ideal change points of the signal change (for example, marked as P _n );

S2, take the first ideal change point (for example, marked as P ₁ ) among the 21 ideal change points as the center, take 5 time points at intervals of 100 ps as sample points in the preceding and following periods, and for each sample point and the first ideal change point P ₁ to perform detection scans respectively, and record the results of each detection scan. The detected signal change is recorded as the first value (for example, marked as P), and the signal change that is not detected is recorded as the second value (for example, marked as F);

S3, repeating the detection scan in step S2 multiple times for each sample point and the _first ideal change point P1, and recording multiple detection scan results of the multiple detection scans;

S4. Establish a statistical data model based on the detection and scanning results. The columns of the statistical data model are sorted according to the time sequence of each sample point and the _first ideal change point P1, and the horizontal columns are sorted according to the order in which the signal changes are found in the scan. sort;

S5. Obtain the first real change point (for example, marked as V ₁ ) of the signal change at the ideal change point P ₁ according to the position of the signal change in the statistical data model;

S6, repeating steps S2-S5 to obtain the real change points of the 21 ideal change points P _n (for example, marked as V _n );

S7, calculating the overall deviation time of the signal (for example, marked as ∆t) according to the 21 ideal change points P _n and the 21 real change points V _n ;

S8, compensating the overall deviation time Δt at the signal input end of the wafer testing machine;

The method of obtaining the _first real change point V1 of the signal change in the step S5 is specifically as follows:

If the position of the signal change point is the same in multiple detection scans, then take this point as the first real change point V ₁ of the signal change; if the signal change point appears in multiple positions in multiple detection scans, then take the present The most frequent point is the first real change point V ₁ of signal change.

The number of repeated detection scans in step S3 is 99 times.

，其中∆t為該整體偏差時間， P _n為該21個理想變化點，且V _n為該21個真實變化點。 The calculation formula of the overall deviation time ∆t in the step S7 is

, where Δt is the overall deviation time, P _{n is} the 21 ideal change points, and V _{n is} the 21 real change points.

Compared with the prior art, the present invention takes all points in a time period, and performs multi-point and multiple scans on all points to obtain the positions of several real points, and then averages the several real points to cover the signal in one cycle In the case of various position changes, the distribution of the signal is fully considered, and the precise position of the signal change is obtained for compensation, thereby improving the accuracy of signal capture.

Embodiment one:

This embodiment is a method for improving the accuracy of capturing signals in a chip testing machine, which specifically includes the following steps:

S1, take 100 picoseconds (ps) as the time interval to take 21 points as the ideal change point P _n of the signal change, respectively P ₁ =10ns, P ₂ =10.1ns, P ₃ =10.2ns, P ₄ =10.3ns , P ₅ =10.4ns, P ₆ =10.5ns, P ₇ =10.6ns, P ₈ =10.7ns, P ₉ =10.8ns, P ₁₀ =10.9ns, P ₁₁ =11ns, P ₁₂ =11.1ns, P ₁₃ =11.2ns, P ₁₄ =11.3ns, P ₁₅ =11.4ns, P ₁₆ =11.5ns, P ₁₇ =11.6ns, P ₁₈ =11.7ns, P ₁₉ =11.8ns, P ₂₀ =11.9ns, P ₂₁ =12ns ;

S2, with the ideal change point P ₁ (10ns) as the center, take 5 time points at intervals of 100ps as sample points, respectively 9.5ns, 9.6ns, 9.7ns, 9.8ns, 9.9ns, 10.1 ns, 10.2ns, 10.3ns, 10.4ns, 10.5ns, each sample point and the ideal change point P ₁ (10ns) were detected and scanned respectively, and the detection and scanning results were recorded, and the detected signal change was recorded as the first value (e.g. marked as P), no signal change detected is recorded as a second value (e.g. marked as F);

S3, 99 detection scans are performed on the ideal change point P1 and ₁₀ sample points, and 11 scan results are obtained for each detection scan;

S4. Establish a statistical data model based on the scanning results. The columns of the statistical data model are sorted according to the time sequence of the ideal change point P ₁ and 10 sample points, and the horizontal columns are sorted according to the sequence of signal changes found in the scan. The obtained statistical data Model 300 is shown in FIG. 3 .

S5, in the resulting statistical model, it can be seen that the signal changes at a certain point. It can be seen from Figure 3 that the signal change point appears in multiple positions, and the point with the most occurrences is taken, and 10.1 ns is the real change point V ₁ of the signal change.

S6, repeating steps S2-S5 to obtain the real change points V _n corresponding to the 21 ideal change points P _n respectively, V ₂ =9.9, V ₃ =10, V ₄ =9.7, V ₅ =10.2, V ₆ = 10.3, V ₇ =10.4, V ₈ =10.4, V ₉ =10.5, V ₁₀ =10.6, V ₁₁ =10.7, V ₁₂ =10.8, V ₁₃ =10.9, V ₁₄ =11, V ₁₅ =11.2, V ₁₆ = 11.3, V ₁₇ =11.4, V ₁₈ =11.5, V ₁₉ =11.5, V ₂₀ =11.6, V ₂₁ =11.7, as shown in Fig. 4 .

； S7, calculate the overall deviation time ∆t of the signal according to the ideal change point P _n and the real change point V _n ,

;

S8. Compensating the overall deviation time Δt at the signal input terminal of the wafer testing machine.

In step S1, the minimum unit of the chip tester signal is 100 ps, and the signal has a cycle of 2 ns, so taking 21 points can cover the position where the signal appears in the entire cycle. In this embodiment, 21 points between 10-12 ns are selected to cover all positions where the signal appears within a signal period.

In step S2, the reason for taking 5 points on the left and right is that 10*100ps=1ns, which completely covers the possible range of the current signal. In this step, the comparison unit of the channel in the chip testing machine is used to detect whether the signal rises at 11 time points.

In step S4, the columns of the statistical data model are sorted according to the order in which the signal changes were detected in the scans, rather than in the order in which the scans were detected. For example, in Figure 3, the first detection scan finds that the level change position is at 10.1 ns, and the second detection scan finds that the level change position is at 9.9 ns, then in the statistical data model, the result of the second detection scan should be Arranged in the first line before the first detection scan, it is convenient to build statistical data model.

In step S5, if the position of the signal change point is the same in multiple detection scans, this point is taken as the real change point V _n of the signal change.

In step S6, in the actual application process, the signal may appear at any time step within a cycle, and the probability of appearing at any time step is equal. Therefore, in order to meet the requirement of the minimum position deviation of 21 steps, the time deviation of each point is averaged, and used as the compensation value of the time signal deviation in the channel.

After compensating for the overall deviation time ∆t in step S8, test the time edge placement accuracy (EPA) of the channel of the wafer testing machine, and the measured EPA is +/-120ps, which improves the signal capture compared with the simple average processing of the traditional method. Take precision.

The present invention takes all points in a time period, and performs multi-point and multiple scans on all points to obtain the positions of several real points, and then takes the average of several real points to cover the situation that the signal changes in each position within a cycle, fully Considering the distribution of the signal, the precise position of the signal change is obtained for compensation, thereby improving the accuracy of signal capture.

300: Statistical Data Modeling P: first value F: second value

[Figure 1] is a schematic diagram of ideal signal changes in the wafer testing machine; [Figure 2] is a schematic diagram of the actual signal changes in the chip testing machine; [Fig. 3] is the data statistical model obtained in step S4 of the first embodiment of the present invention; and [ FIG. 4 ] is a schematic diagram of the ideal change point, the real change point and their difference obtained in step S6 and step S7 of the embodiment of the present invention.

300: Statistical Data Modeling

P: first value

F: second value

Claims (3)

A method for improving the accuracy of capturing signals in a wafer testing machine, comprising the following steps: S1, take 21 points as the 21 ideal change points of the signal change with a time interval of 100 picoseconds; S2, taking a first ideal change point in the 21 ideal change points as the center, taking 5 time points at intervals of 100 picoseconds as sample points, for each sample point and the first ideal change point The change points are detected and scanned respectively, and the detection and scanning results are recorded. The detected signal change is recorded as a first value, and the signal change is not detected as a second value; S3. Repeat the detection scan in step S2 multiple times for each sample point and the first ideal change point, and record multiple detection scan results of the multiple detection scans; S4. Establish a statistical data model according to the detection and scanning results. The columns of the statistical data model are sorted according to the time order of each sample point and the first ideal change point, and the horizontal columns are sorted according to the sequence of signal changes found in the scan. sort; S5. Obtain a first real change point of the signal change at the first ideal change point according to the position of the signal change in the statistical data model; S6, repeating the steps S2-S5 to obtain 21 real change points corresponding to the 21 ideal change points; S7, calculating the overall deviation time of the signal according to the 21 ideal change points and the 21 real change points; and S8, compensating the overall deviation time at the signal input end of a chip testing machine, Wherein the first real change point of the signal change obtained in the step S5 includes: If the position of the signal change point is the same in multiple detection scans, then take this point as the first real change point of the signal change; if the signal change point appears in multiple positions in multiple detection scans, take the most frequent occurrence The point of is the first real change point of the signal change. The method for improving the accuracy of capturing signals in a wafer testing machine as claimed in item 1, wherein the number of repeated detection scans in step S3 is 99 times. Such as the method for improving the accuracy of capturing signals in the chip testing machine of claim 1, wherein the calculation formula of the overall deviation time in the step S7 is:

, where Δt is the overall deviation time, P _{n is} the 21 ideal change points, and V _{n is} the 21 real change points.

Images