Multi-channel power device reliability testing device and health monitoring method

Abstract

The invention discloses a multipath power device reliability test device and health monitoring method, the multipath power device reliability test device comprises a plurality of to-be-tested device test paths, and each test path comprises a to-be-tested device, a first switch, a second switch, a current-limiting resistor and a sampling resistor; the first switch is connected with a grid electrode of a to-be-tested device through a current-limiting resistor, a drain electrode of the to-be-tested device is grounded through a sampling resistor, a source electrode of the first switch is connected with the second switch, the first switch is provided with three contacts, the second switch is provided with three contacts, and for each test path, the current-limiting resistor is connected with the grid electrode of the to-be-tested device. The system can be switched among three test modes of HTGB, HTRB and Vth at will.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor testing, and particularly relates to a multi-channel power device reliability testing apparatus and a health monitoring method. Background Art

[0002] High Temperature Gate Bias (HTGB) and High Temperature Reverse Bias (HTRB) are common power device reliability testing methods. During the HTGB or HTRB experiment, in addition to observing the gate-source-drain current I gss or the zero gate voltage drain current I dss variation, the threshold voltage V th is also observed for any serious degradation.

[0003] Reliability testing requires a long time and a large number of samples. By building a multi-channel power device testing apparatus, the reliability verification of products in the laboratory and R & D stages can be accelerated.

[0004] However, the existing technical solutions have the following deficiencies:

[0005] The function of each test path is relatively single, and it does not have the ability to switch arbitrarily among the three test modes of HTGB, HTRB, and V th ;

[0006] During the HTGB or HTRB experiment, V th needs to be measured multiple times in the middle. Since the multi-channel testing apparatus usually multiplexes the signal source and the measurement unit through a switch, when each power device measures V th through mode switching, it will cause the interruption of HTGB or HTRB of all other power devices. The interruption time is long, and frequent switching affects the switch life and system stability. Summary of the Invention

[0007] In order to overcome the above-mentioned shortcomings and deficiencies of the prior art, the purpose of the present invention is to provide a multi-channel power device reliability testing apparatus and a health monitoring method.

[0008] The purpose of the present invention is achieved by the following technical solutions:

[0009] A reliability test device for a multi-channel power device, comprising: a plurality of test paths for devices under test, each test path including a device under test, a first switch, a second switch, a current-limiting resistor, and a sampling resistor; the first switch is connected to the gate of the device under test through the current-limiting resistor, the source of the device under test is grounded through the sampling resistor, the drain of the device under test is connected to the second switch, the first switch has three contacts, namely contact A1, contact B1, and contact C1, the second switch has three contacts, namely contact A2, contact B2, and contact C2, contact C2 is connected to the gate of the device under test, contact A2 is connected to the source of the device under test, contact A1, contact B1, and contact C1 are respectively connected to power supply VGG2, power supply VGG1, and the ground terminal, and contact B2 is connected to power supply VCC1.

[0010] Further,

[0011] HTGB test mode: The first switch is switched to contact B1, the second switch is switched to contact A2, power supply VGG1 is connected to the gate of the device under test through the current-limiting resistor, observe the change in the gate-source-drain current of the device under test, and calculate the gate-source-drain current by measuring the voltage across the sampling resistor.

[0012] HTRB test mode: The first switch is switched to contact C1, the second switch is switched to contact B2, power supply VCC1 is connected to the drain of the device under test, observe the change in the zero-gate-voltage drain current of the device under test, and measure the voltage across the sampling resistor to calculate the change in the zero-gate-voltage drain current.

[0013] V th Test mode: The first switch is switched to contact A1, the second switch is switched to contact C2, power supply VGG2 is connected to the gate of the device under test and a voltage is applied, gradually increase the voltage of power supply VGG2, observe the change in the source-drain current of the device under test until it increases to a preset value, and the difference between the voltage of power supply VGG2 and the voltage across the sampling resistor is V th , and only one device under test is in the V th test mode at the same time.

[0014] Further, for the HTGB test mode and the HTRB test mode, multiple devices can be in the same test mode simultaneously.

[0015] A health monitoring method based on the above-mentioned reliability test device for a multi-channel power device, including

[0016] Establish a multi-variable logistic regression model based on the gate-source-drain current, zero-gate-voltage drain current, threshold voltage in the test path, and the ambient temperature of the device under test;

[0017] According to the multivariate logistic regression model, calculate the health index of the device under test, determine the health status of the device under test, and determine whether to switch V according to the health index th test mode.

[0018] Furthermore, let the threshold voltage be V th The measurement interval is T0. If the health index of the device under test is high, it indicates that the health status of the device under test is stable. When the next T0 cycle arrives, the test path does not need to switch V th test mode.

[0019] Furthermore, the multivariate logistic regression model is specifically:

[0020] θ0, θ1, θ2, θ3 and θ4 are regression coefficients.

[0021] Furthermore, based on a large sample of historical data combined with the mechanism of power devices, determine the regression coefficients θ0, θ1, θ2, θ3 and θ4.

[0022] Furthermore, the large sample of historical data includes health data and abnormal data.

[0023] Furthermore, the mechanism of power devices includes gate reliability and key parameter degradation mechanism.

[0024] Furthermore, θ1 and θ3 are negative values.

[0025] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0026] (1) For each test path, it can be arbitrarily switched among three test modes: HTGB, HTRB, and V th ;

[0027] (2) By establishing a multivariate linear regression model based on machine learning to predict the health index of the device, for power devices with a high health index, reduce the measurement of Vth during mode switching, maximize the continuity of the DUT in the HTGB or HTRB measurement mode, and reduce the impact of frequent switching on the switch life and system stability. Description of the Drawings

[0028] Figure 1 is the structural diagram of the multi-channel power device reliability test device of the present invention.

[0029] Figure 2 is the schematic diagram of the health monitoring method of the present invention. Detailed Embodiments

[0030] The following combines embodiments to further elaborate on the present invention in detail, but the implementation manners of the present invention are not limited thereto.

[0031] As Figure 1 and Figure 2 shown, a reliability test device for multi-channel power devices can arbitrarily switch among three test modes, namely HTGB, HTRB, and V th for each test path. The specific device includes:

[0032] n test paths for devices under test DUT1 - DUTn. Each test path includes a device under test, a first switch, a second switch, a current-limiting resistor, and a measuring resistor. Taking the test path of device under test DUT1 as an example, it includes a first switch S11, a second switch S12, a current-limiting resistor RG, and a sampling resistor RL.

[0033] The first switch S11 is connected to the gate of the device under test through the current-limiting resistor RG. The source of the device under test is grounded through the sampling resistor RL. The drain of the device under test is connected to the second switch S12. The first switch S11 has three contacts, namely contact A1, contact B1, and contact C1. The second switch S12 has three contacts, namely contact A2, contact B2, and contact C2. The contact C2 is connected to the gate of the device under test. The contact A2 is connected to the source of the device under test. The contacts A1, B1, and C1 are respectively connected to the power supply VGG2, the power supply VGG1, and the ground terminal GND. The contact B2 is connected to the power supply VCC1, and the other end of the power supply VCC1 is grounded.

[0034] The specific process of realizing the three test modes is as follows:

[0035] HTGB test mode: For the device under test, its first switch S11 is switched to contact B1, and the second switch S12 is switched to contact A2. The power supply VGG1 is connected to the gate of the device under test DUT1 through the current-limiting resistor RG and outputs a specific voltage. Observe the change in the gate-source-drain current I gss of the device under test DUT1. The circuit where the sampling resistor RL is located is the sampling circuit. After measuring the voltage across the sampling resistor RL, calculate I gss . Multiple devices can be in the HTGB test mode simultaneously.

[0036] HTRB test mode: The first switch S11 is switched to contact C1, and the second switch S12 is switched to contact B2. The power supply VCC1 is connected to the drain of the device under test DUT1. Observe the change in the zero-gate-voltage drain current of the device under test. Measure the voltage across the sampling resistor to calculate the change in the zero-gate-voltage drain current. Multiple devices can be in the HTRB test mode simultaneously.

[0037] V thTest mode: The first switch S11 is switched to contact A1, the second switch S12 is switched to contact C2, the power supply VGG2 is connected to the gate of the device under test DUT1 and a voltage is applied. Gradually increase the voltage of the power supply VGG2 and observe the change in the source-drain current of the device under test until it increases to a preset value. The difference between the voltage of the power supply VGG2 and the voltage across the sampling resistor is V th , and only one device under test is in the V th test mode at the same time.

[0038] A health detection method based on a multivariable linear regression model for a multi-channel power device reliability test device, which analyzes the gate-source-drain current I gss or the zero gate voltage leakage current I dss changes and the V th changes, judges the health status of the power device and predicts its health index. For power devices with a stable current health status and a high health index, reduce the mode switching to measure V th .

[0039] Specifically include:

[0040] Establish a multivariable linear regression model based on the gate-source-drain current, zero gate voltage leakage current, threshold voltage in the detection test path, and the ambient temperature of the device under test;

[0041] Calculate the health index H of the device under test according to the multivariable logistic regression model.

[0042] The multivariable linear regression model is specifically:

[0043] Use the cross-entropy loss function For each coefficient θ j , calculate the gradient of the loss function: where I gss is the gate-source-drain current, I dss is the zero gate voltage leakage current, T is the ambient temperature, i represents the sample number, N is the number of samples, y is the health of the known sample, and the four input features of the gate-source-drain current, zero gate voltage leakage current, threshold voltage, and temperature are used as a feature vector X = [I gss , I dss , V th , T], represents the jth feature of the X vector in the ith sample.

[0044] Initialize θ0, θ1, θ2, θ3, and θ4 to small random numbers. According to the gate reliability of the power device and the degradation mechanism of key parameters, I gss and V thAn upward trend indicates device failure. Therefore, θ1 and θ3 are negative values. Based on a large amount of historical data such as healthy data and abnormal data, combined with the mechanisms of power devices such as gate reliability and the degradation mechanisms of key parameters, the regression coefficients θ0, θ1, θ2, θ3, and θ4 are established.

[0045] After the model is established, the current detected value of the DUT is substituted into the model to calculate the health index H (range 0 - 1). Assume V th The measurement interval is T0. If the DUT health index is high, such as close to 1, and the health state is stable, then when the next T0 cycle arrives, no mode switching measurement is performed for V th , the first switch and the second switch do not act, and the continuity of the DUT in the HTGB or HTRB measurement mode is maintained to the greatest extent without being frequently interrupted.

[0046] Furthermore, when the health index is high, specifically between 0.9 - 1, and the health state is stable, then when the next T0 cycle arrives, no mode switching and measurement are performed for V th , the switches Si1 and Si2 do not act, and the continuity of the DUT in the HTGB or HTRB measurement mode is maintained to the greatest extent without being frequently interrupted.

[0047] Furthermore, the test duration of HTGB, HTRB, and V th is 1000 hours, and the sampling period can be set. According to the product characteristics, the sampling period ranges from several seconds to several minutes.

[0048] Furthermore, the on-resistances of the first switch and the second switch are low, usually within 1 ohm, and the switch has a long service life with a switch switching times of more than 100,000 times.

[0049] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited by the described embodiments. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent substitution methods and are all included in the protection scope of the present invention.

Claims

1. A reliability test device for a multi-channel power device, characterized in that Including: Multiple test paths for devices under test, each test path including a device under test, a first switch, a second switch, a current-limiting resistor, and a sampling resistor; the first switch is connected to the gate of the device under test through the current-limiting resistor, the source of the device under test is grounded through the sampling resistor, the drain of the device under test is connected to the second switch, the first switch has three contacts, namely contact A1, contact B1, and contact C1, the second switch has three contacts, namely contact A2, contact B2, and contact C2, contact C2 is connected to the gate of the device under test, contact A2 is connected to the source of the device under test, contact A1, contact B1, and contact C1 are respectively connected to power supply VGG2, power supply VGG1, and the ground terminal, and contact B2 is connected to power supply VCC1.

2. The multi-channel power device reliability test device according to claim 1, characterized in that HTGB test mode: The first switch is switched to contact B1, the second switch is switched to contact A2, power supply VGG1 is connected to the gate of the device under test through the current-limiting resistor, observe the change of the gate-source-drain current of the device under test, and calculate the gate-source-drain current by measuring the voltage across the sampling resistor; HTRB test mode: The first switch is switched to contact C1, the second switch is switched to contact B2, power supply VCC1 is connected to the drain of the device under test, observe the change of the zero-gate-voltage drain current of the device under test, and measure the voltage across the sampling resistor to calculate the change of the zero-gate-voltage drain current; V th Test mode: The first switch is switched to contact A1, the second switch is switched to contact C2, the power supply VGG2 is connected to the gate of the device under test and a voltage is applied. The voltage of the power supply VGG2 is gradually increased, and the change in the source-drain current of the device under test is observed until it increases to a preset value. The difference between the power supply voltage and the voltage across the sampling resistor is V th , and only one device under test is in V th test mode at the same time.

3. The multi-channel power device reliability test apparatus according to claim 2, wherein For the HTGB test mode and the HTRB test mode, multiple devices can be in the same test mode simultaneously.

4. A health detection method for a multi-channel power device reliability test device according to any one of claims 1-3, characterized in that Based on the gate-source-drain current, zero-gate-voltage drain current, threshold voltage in the test path, and the ambient temperature of the device under test, establish a multi-variable logistic regression model; According to the multivariate logistic regression model, calculate the health index of the device under test, determine the health status of the device under test, and determine whether to switch the V th test mode.

5. The health monitoring method according to claim 4, characterized in that, set Threshold voltage V th The measurement interval is T0. If the health index of the device under test is high, it indicates that the health state of the device under test is stable. When the next T0 cycle arrives, the test path does not need to switch V th Test mode.

6. The health monitoring method according to claim 4, wherein The specific multi-variable logistic regression model is: θ0, θ1, θ2, θ3 and θ4 are regression coefficients.

7. The health monitoring method according to claim 5, characterized in that Based on a large sample of historical data combined with the mechanism of power devices, determine the regression coefficients θ0, θ1, θ2, θ3, and θ4.

8. The health monitoring method according to claim 7, wherein The large sample of historical data includes healthy data and abnormal data.

9. The health monitoring method according to claim 7, wherein The mechanism of power devices includes gate reliability and key parameter degradation mechanism.

10. The health monitoring method according to claim 7, characterized in that θ1 and θ3 are negative values.

Images