“The development of sensor semiconductor technology has increasingly become a typical way to improve sensor integration. In many cases, it has laid a solid foundation for improving the integration of special-purpose MEMS (Micro Electro Mechanical System) sensors.
The development of sensor semiconductor technology has increasingly become a typical way to improve sensor integration. In many cases, it has laid a solid foundation for improving the integration of special-purpose MEMS (Micro Electro Mechanical System) sensors.
This article introduces the packaging structure of a MEMS photothermal sensor and the assembly details of a system-in-package (SIP), involving an infrared sensor structure based on semiconductor technology. Sensor packaging and its physical interaction with the sensor chip are one of the main factors affecting the overall performance of the system. This article will focus on these physical elements.
The package structure discussed in this article is a cavity grid array (LGA). The structural characteristics and physical characteristics of the materials involved must match the optical signal processing of the sensor and the electrical signal processing performance of the built-in application-specific integrated circuit (ASIC) controller.
From the perspective of concept and design, the special organic substrate design, the molded cavity structure and the silicon-based infrared filter window are the main characteristics of the optical sensor system. At the end of this article, the sensor performance and photoelectric characterization test report is given, including the FFOV (full field of view) test results of two packages with different infrared light window sizes.
Nowadays, photothermal detectors are widely used in various functions such as somatosensory detection, temperature measurement, people counting and smoke detection, covering multiple markets such as construction, security, home appliances, industry and consumption.
The photothermal detector market has five major growth points in the future: portable spot temperature measurement, somatosensory detection, smart buildings, heating, ventilation and air conditioning (HVAC) and other media temperature measurement, and population statistics.
Every object generates heat radiation, and the radiation intensity is related to its own temperature. According to Stephen Boltzmann’s law, the relationship between the temperature of an object and the radiation energy is fixed. As the temperature rises, the wavelength of the radiation peak starts to become shorter: the radiation peak of 300K (room temperature) light is 10 um wavelength, The radiation peak of sunlight (6000K) is 500nm wavelength, which belongs to the visible light frequency domain.
After absorbing incident infrared radiation, the photothermal detector uses a pyroelectric mechanism to convert electromagnetic wave energy into electrical signals, such as pyroelectric voltage, Seebeck pyroelectric effect 3, resistance or pyroelectric voltage).
Modern semiconductor technology, especially MEMS manufacturing technology, can produce very efficient uncooled infrared detectors. Because thermal isolation can be achieved, the sensitivity of the sensor is very high, and the volume is small, the response time is very fast, and the mass production of semiconductors Ways 5 and 6 can reduce the price of MEMS sensors. In order to improve the efficiency of the sensor system, MEMS sensors must be matched with packages and optical units with similar performance.
Certain physical components of the sensor, such as the encapsulation housing and the light window that allows infrared radiation to reach the sensor, also play a role in protecting the peripheral circuits and interconnections. In some cases, the filter window can improve the response spectrum of the sensor and prevent visible light radiation from affecting the sensor’s performance. The filter window material is usually a silicon-based interference filter.
The physical location of this optical interface is located on the upper surface of the package, opposite to the surface where the leads connecting the sensor and the PCB circuit board are located.
This article introduces a system-in-package (SIP) that integrates an infrared sensor and an ASIC chip in a package with filter function, focusing on the relevant characteristics of the package, including material characteristics, optical performance, and overall system sensitivity. This is a cavity grid array (LGA) package concept with an integrated infrared filter window. We have designed, produced prototypes, and performed characterization tests. The sensor field of view ranges from 80° to 110°, depending on the geometric size of the light window. Finally, we also studied the impact of packaging on sensor sensitivity.
This innovative package is designed for MEMS infrared sensors based on micro-machined thermopiles and can encapsulate different types of infrared sensors. When the photosensitive area of the sensor is different, just recalculate the geometric dimensions of the package without modifying the package design and materials.
The thermopile is composed of N thermocouples in series, and the output voltage of the sensor is the voltage of a single thermocouple multiplied by N. A thermocouple is a temperature-sensitive element formed by interconnecting the two ends of two conductors of different materials. These two connecting ends are called the hot end and the cold end. According to the Seebeck thermoelectric effect 3, when the temperature of the hot and cold ends is different, a voltage difference ΔV will be generated between the two conductors. The following is the expression of this voltage difference:
V = NaT (1)
Where ?T is the temperature difference between the hot end and the cold end, and the Seebeck coefficient a is related to the conductor material.
In the micro-machined thermopile, the thermocouple legs are embedded in the dielectric film: the hot end is located in the suspended film, and the cold end is in the suspended film on the silicon substrate. This design is to optimize the temperature difference between the hot and cold ends to maximize Increase the output voltage. The output voltage is usually in the range of a few hundred microvolts, up to a few millivolts: therefore, the output signal needs to be appropriately amplified so that the back-end circuit can process the signal correctly.
The micro-mechanical thermopile sensor proposed in this paper is made up of p/n polysilicon thermocouples in series. The central aluminum plate is coated with a dielectric material and used as a radiation absorbing film. The photosensitive area of the sensor is 600mmX600mm. Figure 1 is a schematic diagram of the sensor layout. There is also an area on the physical package for integrating test sensors to measure sensor parameters during the characterization test. In order to reduce the chip size and optimize the position of the optical window, the advanced version will remove the test sensor.
Figure 1: The main body of the infrared sensor and the photosensitive area of the thermopile infrared sensor and the sensor integration area for testing
MEMS infrared sensors are usually electrically connected with an application specific integrated circuit (ASIC) to control the sensor and amplify the output signal. Therefore, we evaluated a system-in-package infrared sensor. In order to ensure that the incident infrared radiation reaches the photosensitive area of the sensor, and to avoid radiation noise caused by visible light flashes, for selected applications, we integrate an optional long-pass filter with an infrared wavelength of l> 5.5 µm on the system-in-package.
Within the wavelength range required by the presence detection sensor system, the total loss caused by the infrared long-pass filter is controlled within about 20%. For some main purposes, for example, installing a presence detection sensor or infrared temperature measurement on a device PCB board For sensors, energy losses of this magnitude are considered very limited. For other potential applications in the future, the interference filter in question will be replaced with a filter with a different transmission spectrum.
Figure 2: Transmission spectrum of the long-pass infrared filter integrated on the upper surface of the package
The package discussed in this article uses a silicon-based filter that usually has an interference layer integrated on both sides. You can also choose to install different types of filters to meet different application requirements, such as NDIR spectrometers.
Figure 3: Package layout of MEMS infrared sensor and ASIC
The design and development of the infrared sensor package adopts a common parallel layout, and the sensor and ASIC are placed side by side in the package (Figure 3).
An optical window is integrated on the upper surface of the package to select the wavelength component of infrared radiation. This optical window solution can prevent ambient light radiation from reaching the photosensitive area of the detector, thereby reducing the overall system noise. The polymer constituting the upper surface of the package and the cavity wall can be regarded as completely opaque to visible light-infrared radiation, and can be classified as an LCP material (liquid crystal polymer). Different applications can be installed with different filters, for example, NDIR spectrometer. As shown in Figure 3, the structural element includes two dies and bonding wires, the sensor and the signal processing circuit are interconnected, and then connected to the package substrate.
Figure 4: Rendering of the “small infrared light window” package and the “integrated infrared filter cap” package
Experimental setup and measurement
The photoelectric characteristics of the MEMS infrared sensor are characterized. The target object is a calibrated blackbody radiation source at -20°C to 160°C. The black body radiation source used is CI Systems’ SR-800R 4D/A, which has an area of 4 x 4 square inches and an emissivity of 0.99. During the characterization experiment, the sensor was placed at a distance of 5.0 cm from the surface of the black body in order to completely cover the field of view of the sensor.
Figure 5: Experimental setup
Data was collected once with and without filters, and the signal-to-noise ratios were observed to be 1.6 and 2.36 respectively. When the filter is used, the sampling signal-to-noise ratio is reduced, which is caused by the light attenuation of the filter, and it is in full compliance with the frequency spectrum of Figure 2.
Figure 6: Sensitivity characterization of ceramic packaged sensors with and without infrared filters.
The system output is a digital signal. Under infrared radiation, the digital change of the least significant bit (lsb) represents the system output change. Under the condition that the package geometry is determined and the black body completely covers the field of view of the light window, the total sensitivity of the sensor under test is about 2000lsb/°C, and noise is found at 150lsb. The infrared long-pass filter can be selected, mainly to match the expected detection selectivity and the nature and size of the detectable object in front of the light window.
Figure 7: 3D-X-ray tomography image of the package with infrared silicon-based filter, where the filter has two metal reflective films, M1 and M2
As shown in Figure 7, two metal infrared filter films, M1 and M2, are placed on the MEMS infrared sensor to filter the incident radiation on the package surface. The wire bonding structure of the sensor and ASIC interconnection and the metal traces of the package substrate can also be seen in the 3D image.
Field of view (FOV) angle calculation
We usually define a field of view (FOV) parameter for the optical system to evaluate the size of the geometric space that the sensing system can detect. Any optical device can be defined as a half field of view (HFOV) with FOV = ±θ or a full field of view (FFOV) with FOV = θ. This article adopts the half-field definition of FOV = ±θ. In the geometric space evaluation, it is assumed that the refractive index of silicon is n = 3.44; the refractive index of air and vacuum is n = 1. The figure below shows the FOV calculation method for the cross-sectional structure of the package in question.
Figure 8: Cross-section diagram of FOV calculation principle
When calculating the angle of view, it is necessary to consider the refraction (or bending) that occurs when the light passes through the window.
Using the basic relationship of trigonometry, we find:
Where WO is the width of the package light window, WA is the width of the sensor’s photosensitive area, and Wt1+Wh1 is the width of the light path in air and silicon. The calculation method is shown in the following equation group:
Wt1=t1tgqS; (eq. 2a)
Among them, t1 and h1 are the geometrical vertical parameters of the package and the device itself, and qA and qS are the propagation angles of infrared rays in air and silicon, respectively. According to Snell’s law, the following equation gives the relationship between the two angles:
n1.sin (θ1) = n2.sin(θ2)(eq. 3)
n1 and n2 represent the refractive index of each material, θ1 and θ2 are the angles (counterclockwise) formed by the light propagating in each material and the surface normal, and assume that the refractive index of silicon n = 3.44, air/vacuum The refractive index n=1. Based on the above geometric assumptions, the expected field of view angle FFOV = 80°-82°. Then began the preliminary design of the cavity package, and manufactured two batches of prototypes in the package trial production line laboratory. In order to obtain different FFOV, we propose two different window designs. In order to verify the “T% = 0” condition of the packaging cavity wall material in the wavelength range of 1.0um -13.0um, the infrared transmittance value of the molded resin material was tested. The package structure is a system-in-package, in which the ASIC die and the MEMS infrared sensor are placed side by side, and the die is connected by wire bonding (WB), as shown in the figure below.
Figure 9: Package with infrared light window (left picture) and integrated infrared filter package (right picture), soldered on DIL 24 test board by surface mount technology (SMT)
Using the aforementioned black body radiation source, a characterization experiment was performed on the above two system packages at a distance of 22 cm from the top of the package.
Figure 10: Sensitivity comparison of a MEMS infrared sensor package with a small light window on the cap and a package with an infrared filter as a whole
After the experiment, at 22cm, no difference in sensitivity measurement between the small light window and the integrated infrared filter cap was observed, and the response time was the same. This distance is chosen to make the beam direction close to the infrared plane incident wave on the upper surface of the sensor. In order to carry out the FOV characterization experiment, in view of the normal condition that the sensor’s photosensitive area is placed in front of the black body, the sensor is installed on a rotating table from -90° to +90°.
Figure 11: FOV characterization experiment results of infrared small light window package, integrated infrared filter package and large ceramic package of infrared sensor
In a large ceramic package, the FFOV angle of the infrared sensor is 109°±2°, which is smaller than the theoretical value of the Lambertian distribution (theoretically 120°), which may be caused by the silicon embedded structure of the MEMS. The FFOV angle of the small light window package is 88°. Using the same package rotation method, the FFOV of the integrated infrared filter molded package is 100°. In the last case, the asymmetry effect was observed because the cavity wall of the molded package was close to the sensor’s photosensitive area.
Package stress simulation
For specific absorbed power, high thermal isolation ensures the maximum temperature difference between the hot and cold ends, which is an important factor in obtaining large output voltage from the thermopile. The gas in the cavity can be selected using the MEMS package, and the pressure can be selected from 100Bar to 100mBar. The thermal conductivity of the gas affects the temperature conduction speed and the temperature difference between the hot and cold ends of the thermopile, which in turn affects the output voltage change and the sensor efficiency.
MEMS packaging is achieved through wire bonding technology between wafers. The MEMS sensor system is mainly composed of a silicon microstructure manufactured by a surface micromachining process. Usually two or more wafers (die) are stacked and soldered in a silicon-based package with a glass material compound solder. .
There is a silicon protective cap with a thickness of about 150um on the sensor, which itself has a natural infrared wavelength filtering function for the radiation incident on the surface of the sensor. Of course, the infrared transmission spectrum of the silicon protective cap deteriorates the optical performance of the sensor in the 1-13um wavelength infrared region12, depending on the silicon characteristics.
Sensor development requires the integration of MEMS silicon caps on the sensor wafer. We simulated the entire sensor system consisting of infrared sensors, silicon caps, ASICs and packages. Because the die is stacked and mounted on the packaging substrate, the sensor microstructure and the packaging structure are integrated. Therefore, the packaging has an impact on the sensor signal performance. In addition to the stress received during the working process, critical situations may also occur during the manufacturing process, especially the cooling process after the package is soldered to the PCB. Since the package is made of materials with different coefficients of thermal expansion (CTE), thermal gradients can cause warpage, which can lead to stress transfer to the sensor microstructure, thereby affecting the sensing performance.
A finite element 3D model was established with SolidWorks Simulation software to simulate the warpage of the silicon substrate carrying the sensor microstructure. The cooling simulation after soldering takes into account the soldering of the package on the reference PCB. Table 3 summarizes the thermal load and boundary conditions. Figure 12 is a finite element model.
Table 2 lists the characteristics of the materials used in the simulation.
Although it is known that the simulation results depend to a large extent on the material model and the characteristics of the materials used, considering the conventional practices in the packaging simulation literature, we still assume the purpose of the analysis and comparison, the available material data, and the static nature of the simulation performed , The isotropic elasticity of the material.
In order to reduce the calculation time, we consider creating a simplified model. However, because the ASIC is placed asymmetrically inside the package and there is a light window on the cap, it is necessary to simulate the entire model. For the upper and lower substrate layers of the package, the equivalent mechanical performance calculation method is as follows14:
Among them, Eeff is the effective Young’s modulus, and αeff is the effective thermal expansion coefficient, which are the Young’s modulus Ei, αi, Vi and CTE and the volume or area percentage of the constituent material. Figure 12 is the finite element model, and Figure 13 is the warpage simulation results on the sensor, ASIC, and substrate. The warpage w of the substrate carrying the sensor microstructure is defined as the difference between the maximum and minimum displacement z along the frame itself.
Table 2. Material characteristics
α1 = 60
α2 = 130
Tg = 114°C
Coresubstrate core substrate
α1 = 19
α2 = 5
Tg = 230°C
Copper copper layer