Non-contact detection technology for LED chip/device packaging defects

1. Introduction
In recent years, with the decline in manufacturing costs and the breakthrough of technical bottlenecks such as luminous efficiency and light decay, China's LED lighting industry has entered a stage of accelerated development, and the application market has grown rapidly, which has led to a huge market for LED packaging products and spawned thousands of LED packaging companies, making China the world's largest producer of LED packaging. The annual output value of LED packaging products has increased from 9.9 billion yuan in 2004 and 14 billion yuan in 2006 to 18.5 billion yuan in 2008, and the annual output has exceeded one trillion units [1][2]. If the scrap/defective rate of LED packaging is 0.1%, then hundreds of millions of scrap/defective LEDs may be produced out of the trillion LED packaging products in the country each year, causing direct economic losses of nearly one billion yuan.
In order to ensure the quality of packaging, LED packaging companies ensure the quality of LED packaging through microscopic inspection before packaging and sub-inspection after packaging. Microscopic inspection before packaging is to use a microscope to conduct an artificial appearance inspection of the raw material chip before packaging, observe whether there is mechanical damage and pitting on the surface of the chip material, whether the chip size and electrode size meet the process requirements, whether the electrode pattern is complete, and remove unqualified chips to prevent them from flowing into the next process and producing defective products; after packaging, the sorting inspection is to use an automatic spectrophotometer to check the optical and electrical parameters of the finished packaged product after the packaging is completed, and classify them according to the test results, and then package them. Obviously, the microscopic inspection before packaging and the sorting inspection after packaging can only distinguish the defective products produced in the packaging from the genuine products, or classify the genuine products according to the parameters, and cannot improve the yield of the packaging.
For modern fully automatic packaging lines, any slight difference in itself will quickly have a direct impact on the quality of the packaged products. Therefore, under the condition of the full popularization of fully automatic packaging lines, it has become an inevitable requirement to actively conduct online real-time detection of packaging quality during the packaging production process to improve the packaging level and ensure the packaging quality. Due to the small size of LED chips, high packaging process requirements, and fast packaging production speed, it is difficult to conduct real-time quality detection and control during the packaging process.

2. Analysis of the characteristics of LED packaging process
In order to conduct real-time online detection of the chip/package quality during the LED packaging process, it is necessary to first understand the process characteristics of LED packaging and the parameter characteristics of LED.
2.1 The process of LED packaging
The task of LED packaging is to connect the external leads to the electrodes of the LED chip, protect the LED chip, and improve the light extraction efficiency. The packaging forms of LEDs are varied, and the corresponding dimensions are mainly used according to different application occasions. The bracket-type full epoxy encapsulation is currently the largest and highest-yield form, so it should also be the key breakthrough object for online detection of LED packaging product quality.
The main process of bracket-type full epoxy encapsulation is [4], firstly, the LED chip is inspected and expanded under a microscope, and silver glue is applied to the center of the reflective bowl of each LED bracket in a group of brackets with connecting ribs and the back electrode of the chip (i.e., glue application and glue preparation process), then the LED chip is sucked up by a vacuum nozzle and placed in the center of the reflective bowl of the bracket, and the back electrode of the chip and the bracket are fixed together by sintering (i.e., solid crystal process); the electrode lead is connected to the LED chip by pressure welding to complete the connection of the internal and external leads of the product (i.e., pressure welding process); the optical epoxy glue is vacuum-defoamed and then poured into the LED molding mold, and then the bracket is pressed into the LED molding mold as a whole (i.e., glue filling process), the epoxy glue is subjected to high-temperature curing, annealing and cooling, and demolding after curing (i.e., curing process), and finally the connecting ribs of the LED bracket are cut off (as shown in Figure 1), and finally, sorting and packaging are carried out.
2.2 Analysis of the characteristics of LED packaging process
From the perspective of LED packaging process, the chip expansion, glue preparation, and crystal point process may cause damage to the chip, affecting all the optical and electrical characteristics of the LED; in the process of solid crystal and pressure welding of the bracket, chip dislocation, poor contact of the internal electrode, or cold welding or welding stress of the external electrode lead may occur. Chip dislocation affects the distribution and efficiency of the output light field, and poor contact or cold welding of the internal and external electrodes will increase the contact resistance of the LED; in the process of glue filling and epoxy curing, bubbles and thermal stress may be generated, affecting the output light efficiency of the LED.
Therefore, it can be seen that both the LED chip and the packaging process will affect its optical and electrical characteristics, so the final quality of the LED is a comprehensive reflection of each process link. To improve the quality of its packaged products, it is necessary to conduct real-time detection and adjust the process parameters of each production process link to minimize defective and waste products.
Due to the delicate, complex and high-speed characteristics of the packaging process, conventional contact measurement is almost difficult to achieve quality detection in packaging, and non-contact measurement is the most promising means.

3. Basic principles of non-contact detection
3.1 Photovoltaic characteristics of LED chips
The core of the light-emitting diode LED chip is the doped PN junction. When a forward working voltage VD is applied to it, the holes in the valence band are driven to pass through the PN junction into the N-type region, and the electrons in the conduction band are driven to cross the PN junction into the P-type region. The excess carriers will recombine near the junction, and emit light during the recombination process, thereby converting electrical energy into light energy. The nature of its light emission under current driving conditions is determined by the doping characteristics of PN, and the photoelectric characteristics of the photodiode PD are also determined by the doping characteristics of PN. Therefore, LED and PD are similar in nature. When the light beam is irradiated on the open circuit LED chip, the accumulation of photogenerated carriers, electrons and holes, will be generated at both ends of the PN junction of the LED chip, forming a photogenerated voltage VL. If the external circuit of this LED chip is short-circuited, the photogenerated carriers at both ends of its PN junction will flow in a directional manner to form a photogenerated current IL: [4][5]

In the formula: A is the PN junction area of ​​the chip, q is the electron charge, w is the barrier width of the PN junction, Ln and Lp are the diffusion lengths of electrons and holes respectively, β is the quantum yield (i.e. the number of electron-hole pairs generated per absorbed photon), and P is the average light intensity irradiated on the PN junction (i.e. the number of photons absorbed by the semiconductor material per unit area per unit time). They are:

Among them, μn and μp are the electron and hole mobility (related to the material itself, doping concentration and temperature), KB is the Boltzmann constant, T is the Kelvin temperature, τn and τp are the electron and hole carrier lifetimes (related to the material itself and temperature), α is the material absorption coefficient related to the semiconductor PN junction material itself, the doping concentration and the wavelength of the excitation light, d is the thickness of the PN junction, and P(x) is the excitation light intensity at position x in the PN junction.
From formulas (1) to (3), it can be seen that the photovoltaic characteristics of the LED chip are related to the structural parameters and material parameters of its PN junction, and these parameters are exactly the key parameters that determine the LED luminescence characteristics. Therefore, if the luminescence characteristics of an LED chip are good, its photovoltaic characteristics are also good, and vice versa. Therefore, this intrinsic connection between the luminescence characteristics and photovoltaic characteristics of the LED chip can be used to indirectly test its luminescence characteristics by testing its photovoltaic characteristics, judge the quality of the LED chip, and realize non-contact detection of its packaging quality.
3.2 Equivalent circuit of LED photovoltaic characteristics
For LEDs packaged in bracket type, during the packaging process, a set of brackets with connecting ribs are clamped on the packaging machine, and then the chip and the bracket are packaged together to form the bracket packaging structure shown in Figure 1. As can be seen from Figures 1 (b) and (c), the bracket, bracket connecting ribs, leads, silver glue and LED chip of the LED together form a complete external circuit short-circuit channel, which meets the working requirements of the photovoltaic effect. However, for the conventional detection method of LED packaging quality, such working conditions are completely impossible to carry out detection.
Since the actual LED is not a simple ideal PN junction, it not only includes the internal resistance, parallel resistance and series resistance of the PN junction, but also includes the bracket, bracket connecting ribs, leads and silver glue. Therefore, the photoelectric effect generated by the PN junction under external light generates a photoelectric current IL that is not completely equal to the photoelectric current IL1 flowing through the bracket. Therefore, the current flowing through the bracket is a comprehensive reflection of the LED photoelectric parameters.

If the internal resistance RL of the lead frame is regarded as the load of the LED when illuminated, and the photocurrent IL generated by the photovoltaic effect of the PN junction is regarded as a constant current source, the equivalent circuit of the LED when illuminated is shown in Figure 2. That is, the LED working under the photovoltaic effect is equivalent to an ideal current source IL, an ideal diode D, and the corresponding equivalent series and parallel resistances Rsh and Rs. The equivalent parallel resistance Rsh includes the leakage resistance in the PN junction and the leakage resistance at the edge of the junction, and the equivalent series resistance Rs includes the body resistance Rs1 of the P region and the N region, the resistance of the electrode, and the contact resistance Rs2 between the electrode and the junction, and

IL1 is the load current flowing through the lead frame, and IF is the forward current flowing through the ideal diode D. It satisfies the relationship with the voltage VD across the diode:

Where Is is the reverse saturation current of the diode, and η is a parameter related to the PN junction current recombination mechanism, both of which are determined by the characteristics of the LED chip. Therefore, IF reflects the chip characteristics of the LED.

According to the equivalent circuit shown in FIG2 , the relationship between the photocurrent IL and the current IL1 flowing through the bracket can be obtained as follows:

It can be seen from formula (7) that for LED packaging products, the current IL1 on the external circuit consists of two parts, of which the numerator mainly reflects the intrinsic quality of the chip, while the denominator mainly reflects the quality of the device outside the chip (such as the many defects in the solid crystal glue connection and lead welding quality in the packaging process). Therefore, as long as the photocurrent on the connecting rib is detected, the packaging quality of the LED chip/device can be fully grasped.
4. Weak signal detection technology for non-contact online detection of LED packaging quality
4.1 System implementation principle
It can be seen from Figure 1 (b), (c) and formula (7) that after the LED is pressed and before the glue is poured, the short circuit necessary for the LED photovoltaic effect has been formed. Therefore, after the pressure welding and before the glue is poured, the photovoltaic effect of the LED can be used to detect the chip quality, solid crystal quality, and pressure welding quality, and timely pick out defective products for manual repair, and according to the detection results, the corresponding process parameters of the LED packaging production line are corrected in real time to further control the defective rate. After the epoxy encapsulation is completed and before the rib cutting, the photovoltaic effect of the LED can be used again to perform non-contact detection of the encapsulation effect, guide the real-time adjustment of the epoxy filling and curing process, and eliminate defective/waste products.
According to Figure 1 and formula (7), there are three keys to using the photovoltaic effect of LED for non-contact detection of chips/packages. First, a specific light beam is used to accurately illuminate the LED chip to provide the light excitation required for the photovoltaic effect in a non-contact manner; second, a special technical means is used to obtain the photocurrent in the bracket circuit in a non-contact manner; and third, the quality defects of the chip are judged based on the obtained photocurrent. For this purpose, the principle system shown in Figure 3 is used to realize non-contact detection of LEDs [5][6].

The light emitted by the semiconductor laser LD is focused and projected onto the LED chip to excite the LED to produce a photovoltaic effect. In the signal acquisition stage, the electromagnetic coupling method is used to obtain the current signal output by the LED under light to achieve non-contact measurement. Finally, the photocurrent is calculated and processed using formula (7) to judge the quality of the LED, find out the reasons that affect the packaging quality, and distinguish the factors of the chip and packaging.
Although the LED will produce a photovoltaic effect under light, its photovoltaic effect is much weaker than that of the photodiode PD as a photodetector. Therefore, its photocurrent IL is extremely weak, only in the order of microamperes. Therefore, non-contact acquisition of the photocurrent in the bracket circuit is the most technically difficult key. Although the electromagnetic coupling method can realize non-contact measurement of the LED photocurrent, the electromagnetic coupling method will also couple into the spatial electromagnetic field. These external electromagnetic field noises and interferences are much stronger than the photocurrent IL. Therefore, it is very difficult to extract the very weak photocurrent IL from the strong external electromagnetic field signal. For this reason, a combination of anti-aliasing filtering and phase-locked amplification is used to achieve the purpose of separating the photocurrent IL from the strong environmental noise.
4.2 System Verification Experiment
Using the principle system shown in Figure 3, a test platform was built to conduct a principle verification experiment on a series of bracket-type LED packaging products. The experimental conditions were the finished product group of bracket-type LED packaging after the epoxy package was demolded but the connecting ribs had not been cut off. The main experiments included a comprehensive qualitative experiment on the system detection effect, a simulation experiment on the influence of chip bonding misalignment on the LED output photocurrent, and a simulation experiment on the influence of lead welding quality on the LED output photocurrent, etc. [4][5].
4.2.1 Comparative experiment of LEDs with different chips
Figure 4 shows the comparative experimental results of LEDs with different chips. Figures 4 (a), (b), and (c) are comparative experiments of three LEDs with different chips under the same conditions, and Figure 4 (d) is the output result without LED (equivalent to the result of pure environmental noise). It can be seen from Figure 4 that the differences between different chips are fully reflected; and from Table 1, it can be seen that the results of 30 experimental repetitions are extremely consistent. In addition, it can be seen from Figure 4 that the detection time of each LED is only 5 milliseconds. If calculated based on a 1:1 signal duty cycle, without considering mechanical movement and inertia, this method can achieve a detection speed of 100 LEDs per second purely from an electrical processing perspective.

4.2.2 Simulation experiment of the influence of LED chip die dislocation
When the die-bonding position is deviated, the chip will deviate from the center of the epoxy lens. At this time, the incident laser beam will be deflected after passing through the lens and cannot be fully focused on the chip, resulting in a weakening of the total light intensity P received by the chip. It can be seen from formula (7) that the change of the incident light intensity P will cause a linear change of IL1. Therefore, the signal intensity output by the system can also reflect the quality of the die-bonding. For this reason, the die-bonding deviation is simulated by adjusting the intensity of the laser light source irradiating the LED. The experimental results are shown in Figure 5, which are completely consistent with formula (7).

4.2.3 Simulation results of the influence of lead welding quality
In the equivalent circuit shown in Figure 2, Rs2 and the load RL are connected in series. Since the resistance of the electrode and the contact resistance Rs2 between the electrode and the junction are difficult to measure directly, the experiment simulates the influence of the contact resistance Rs on the detection results by connecting different load resistances RL in series. The test results are shown in Figure 6. As shown in Figure 6, as the external load RL increases, the current flowing through the load becomes smaller and smaller. Both experiments and theories show that a slight change in the contact resistance Rs will cause a large change in the current IL1 flowing through the bracket. For a fully functional LED chip, the series resistance Rs of the LED can be calculated by measuring the photogenerated current IL1 flowing through the bracket. If the series resistance value is infinite, there may be problems with silver glue debonding, solder leakage or wire breakage between the chip and the electrode. If the series resistance is greatly different from the series resistance in the normal connection state, there may be other welding problems between the chip and the electrode, such as cold soldering, repeated soldering, etc. Therefore, by analyzing the value of the photogenerated current flowing through the bracket, the electrical connection state between the chip and the lead bracket during the LED packaging process can be detected.

5. Conclusion
Since the output of LED packaging in China is very large, real-time online detection of LED packaging quality on large-scale packaging production lines can effectively replace and improve the backward situation of manual visual inspection currently adopted by large-scale packaging production enterprises, and effectively reduce the defective/scrap rate. To this end, we fully utilize the characteristics of LED with photovoltaic effect similar to PD, and the established relationship between LED chip/device packaging quality and photocurrent, and build a non-contact detection experimental platform for LED packaging quality. Through simulation experiments, it is proved that the influence of chip differences, solid crystal quality, and welding quality can be reflected through the characteristics of the output signal of the detector, and the detection discreteness is less than 10-6, and the detection speed can reach 100 pieces/second. On this basis, the actual detection prototype shown in Figure 7 [7] was also developed, and the system integration of the actual detection prototype and the packaging production line is in progress, as well as further quantitative research on LED parameters.