A Real Risk Case in Component Kitting: Recent BGA Issues - Refurbishment, Contamination, and Failed Solderability!

Word Explanation

• Cold solder joints: The solder joints seem to be bonded, but there is actually poor contact;
• Bubbles or voids: Reduce electrical conductivity and heat dissipation performance;
• Early failure: Accelerated fracture under high temperature and high stress environments.

About Us

• VGS(th): Threshold Voltage
• RDS(on): On-State Resistance
• IDSS: Drain-Source Leakage Current
• BVDSS: Drain-Source Breakdown Voltage
• VSD: Diode Forward Voltage Drop, etc.

PCI-Express (Peripheral Component Interconnect Express) Design Specifications for PCB

1. The trace length from the edge of the gold finger to the PCIe chip pins shall be limited to 4 inches (approximately 100mm).
2. PCIe PERP/N, PETP/N, and PECKP/N are three differential pairs. Ensure proper isolation: maintain a 20mil spacing between differential pairs and between differential pairs and all non-PCIe signals to minimize harmful crosstalk and Electromagnetic Interference (EMI). Avoid high-frequency signals on the reverse side of the chip and PCIe signal traces; a full ground plane (GND) is recommended.
3. The length difference between the two traces in a differential pair shall not exceed 5mil, with length matching required for every segment of the two traces. Use a 7mil trace width and 7mil spacing between the two traces in each differential pair.
4. When PCIe signal pairs change layers, place ground vias near the signal vias—1 to 3 ground vias per signal pair are recommended. Use 25/14 vias for PCIe differential pairs, and the two vias must be symmetrically placed.
5. PCIe requires AC coupling between the transmitter and receiver. The two AC coupling capacitors for a differential pair must have the same package size, be symmetrically placed, and located close to the gold finger. The recommended capacitance is 0.1μF, and through-hole packages are prohibited.
6. Signals such as SCL shall not route through the PCIe main chip.       

Why are different test frequencies/voltages used to test capacitance for different capacitance ranges?

The frequency setting of the instrument depends primarily on the parasitic components of the component. For more accurate component testing, the SRF (Self-Resonant Frequency) measurement frequency of the component should be avoided. Industry users set standards for different frequency points based on capacitance values ​​(Table 1). Capacitances above 10uF are considered within the tantalum capacitor range. Therefore, as the ceramic capacitance range begins to expand to the tantalum capacitor range, the industry applies the frequency standards for tantalum capacitor measurements to ceramic capacitors.

The applied voltage also depends on the capacitor's capacitance. Typically, a voltage of 1.0 ± 0.2 Vrms is applied for 10uF and below. However, for above 10uF, the applied voltage is 0.5 ± 0.2 Vrms. High-capacitance capacitors have extremely low impedance; therefore, to supply sufficient current for measurement, the power supply needs to provide more current than that supplied with 1.0 ± 0.2 Vrms. Therefore, by reducing the applied voltage, the power supply will be able to provide sufficient current to accurately measure high-capacitance capacitors.

Class Type Capacitance Frequency Voltage

Class I 1,000pF and under 1MHz ± 10% 0.5 ～ 5 Vrms Over 1,000pF 1kHz ± 10%

Class II 10uF and under 1kHz ± 10% 1.0 ± 0.2 Vrms Over 10uF 120Hz ± 20% 0.5 ± 0.2 Vrms

Nexperia 16474 models (MOS, transistors, ICs, IGBTs) , SPQ，MOQ and (COO) Certificate of Origin lookup

Nexperia 10911 diodes SPQ，MOQ and (COO) Certificate of Origin lookup

Nexperia 83 IGBTs and modules SPQ，MOQ and (COO) Certificate of Origin lookup

Nexperia 4244 logic ICs (74 series) SPQ，MOQ and (COO) Certificate of Origin lookup

Nexperia 1236 MOS (BUK series, etc.) SPQ，MOQ and (COO) Certificate of Origin lookup

This is an excerpt of the table starting with a model number; please download the attachment to view the complete data.

etc.

etc.

etc.

etc.

① On the circuit board, manually soldered joints can be seen connected to movable wires, but their ability to withstand repeated vibrations is definitely not as good as connectors (even those with leads)
② This is a space board that has already been put into use and was only dismantled later
③ It uses chips from Xilinx, which has been acquired by AMD and has become a subsidiary of AMD
④ What are the standards for aerospace circuit board levels? Have any industry partners shared them

Recently, images of humanoid robots running on the track and performing at exhibitions have gone viral. However, many observant viewers noticed that the engineers were still holding remote controls! The debate over whether "remote control equals fraud" is intensifying online and in the media. Today, we'll specifically discuss the issue of remote control of humanoid robots and fraud. Behind this lies a cognitive gap between technological progress and public expectations.

What is a one-year-old baby like? Robots are like that now.

Just like a one-year-old learning to walk, they need adults to support and protect them, and if they fall, they need to be picked up immediately. Humanoid robots are similar: they can take a few steps and pick things up on their own, but they rely on a "remote control" as an "invisible crutch." For example, when humanoid robots are trained in factories, they can perform tasks such as inspecting car door locks and affixing car stickers, but they must first go through the three steps of "perception-planning-execution." Autonomous decision-making in open environments is something the world hasn't yet mastered. An experienced worker in the industry gave an analogy: a robot's "cerebellum" can control balance and joint movement, but its "brain" gets confused when faced with complex situations—just like a baby can imitate walking, but is prone to falling if not being led by an adult.

Remote control isn't fraud; it's a "learning tool" for growth.

Some say "using remote control is deception," but Academician Han Guangjie said that remote control is more like an "emergency button," allowing engineers to take over and ensure safety when the robot suddenly becomes confused. Whether humanoid robot manufacturers are fraudulent depends on their advertising. If they claim "fully autonomous" operation while relying on remote control, that's fraud. If they're honest and the remote control is a "learning tool" during development, like a parent helping a toddler learn to walk, that's the proper approach. Think about it: robots learn skills through "imitation-practice-autonomy." Tesla's Optimus and Zhiyuan Yuanzheng A2 do this, collecting data through human "hands-on" teaching to form an evolutionary loop of "teaching-training-autonomy." This isn't technological regression; it's an essential part of growth.

From "learning to walk at one year old" to "walking independently," there are still hurdles to overcome.

Humanoid robots need to overcome several obstacles to walk independently without a remote control. Firstly, technically, the A2 robot from Zhiyuan Robotics, capable of traversing 100 kilometers across provinces, relies on multi-sensor fusion of dual GPS, LiDAR, and infrared depth cameras. While it can recognize complex terrain and avoid crowds, it is prone to misjudging situations and requires alerts. Secondly, in terms of safety, human-robot mixed environments require safety measures. For example, CloudMinds uses a cloud-based brain system to ensure remote intervention when robots make mistakes. Thirdly, in terms of industry, humanoid robots entering factory workshops must first overcome challenges such as cross-scene migration and small-sample learning. Currently, the robotics industry is moving towards "Level 3," aiming to handle most scenarios autonomously and only call for human assistance in complex situations. Experts predict that achieving fully autonomous "Level 4 and Level 5" robots will require another 3-5 years of development.

Controversy is a wake-up call, but also an opportunity.

This debate has also served as a reminder to the industry: technology must allow for trial and error, but communication must be transparent. Manufacturers must be honest and not mislead with claims of "fully autonomous" technology; the public must also understand that new technologies, like toddlers learning to walk, need room to grow. With breakthroughs in technologies like multimodal perception and reinforcement learning, humanoid robots will eventually need to "retire" once they reach 5 or 6 years old. Currently, robots are in a critical "learning-to-walk" stage. Controversy isn't a bad thing; it forces the industry to become more standardized. Technology hasn't been idle either, constantly pushing forward. As Pang Jianxin said, the development of humanoid robots will take three steps: first, performing limited tasks in structured scenarios; second, learning to accompany and interact; and finally, entering the home. Currently, UBTECH's training at Geely's factory and the A2 robot's cross-province journey are both examples of "learning to walk." When robots can, as Lu Junguo said, accompany the elderly and care for children in the home, that will be true "independence."

The process of a robot learning to walk is similar to that of a toddler learning to walk. First, it needs support to walk, then it gradually learns to walk independently. It can't be rushed, and it can't be faked. Don't you agree? If we are more patient and accept its "dependency" like we do with a toddler learning to walk, and if companies focus on solid research and development, robots will eventually be able to "run" on their own, entering factories, shopping malls, and even our homes.

As we all know, the design of humanoid robots is in full swing in both China and the United States. So, what chips are used in China's humanoid robots?

AI Decision-Making Chips (The "Brain")

• International: NVIDIA Jetson Orin, Jetson AGX Thor.
• Chinese Brands: Rockchip RK3588, Horizon Robotics Journey A2000 (Huashan Series), among others, are making significant strides in this sector.

Motion Control Chips (The "Cerebellum")

• International: Texas Instruments (TI, U.S.) TMS320F28377D, STMicroelectronics (ST, Europe) STM32H7 series are widely adopted.
• Chinese Brands: HPMicro HPM6E00, GigaDevice GD32H75E, etc., meet the microsecond-level synchronous control requirements for multi-joint systems.

Visual Sensing Chips (The "Eyes")

• International: Sony (Japan) IMX series are common 2D vision CMOS chips.
• Chinese Brands: Orbbec Femto series iToF chips, OmniVision (China) global shutter image sensors, enabling high-precision 3D perception and dynamic image capture.

Communication Interface Chips (The "Data Neural Network")

• International: NXP S32K39, Infineon LAN9252 are widely used.
• Chinese Brands: Create Bright Technology EtherCAT slave control chips, Youtai Micro low-latency Ethernet chips, adapting to high-real-time transmission needs.

Power Management Chips (The "Energy Heart")

• International: Texas Instruments (TI, U.S.) BQ27546, Analog Devices (ADI) LTC3880 dominate the market.
• Chinese Brands: Eastchip Semiconductor DK87XXBD series, SGMICRO high-precision power management chips, suitable for wide-voltage and fast-charging scenarios.

Memory Chips (The "Memory Center")

• Chinese Brand: GigaDevice SPI NOR Flash GD25LX series, with a data throughput of 400MB/s, can meet the requirements of fast robot code reading and program startup.              

Statement Urging Nexperia Netherlands to Earnestly Respond to Communications and Resolve the Control Right Issue to Safeguard the Stability of the Global Supply Chain

Statement by Wingtech Technology on the Suspension of the Administrative Order by the Dutch Ministry of Economic Affairs

I. Our Company's Position on the Decision of the Dutch Ministry of Economic Affairs

II. The Dutch Ministry of Economic Affairs is Obligated to Thoroughly and Comprehensively Resolve the Nexperia Semiconductor Issue

III. Wingtech Technology's Complete Rights as a Shareholder and Its Legitimate Control Over Nexperia Must Be Restored

OnePcba conducted a full-process incoming inspection on a varistor in accordance with customer requirements. In addition to routine appearance screening, the laboratory focused on "anatomical" structural verification through Cross Section (section analysis) to accurately check the integrity of the internal structure of the component.

Section analysis is a core technology for micro-detection of electronic components: through steps such as resin encapsulation curing, precision grinding, and mirror polishing, the internal structure details are finally presented intuitively with a metallographic microscope (magnification 50~500x). Its advantage lies in capturing tiny defects that cannot be identified by the naked eye and conventional detection.

For varistors, the "safety valves" of circuits, the significance of section detection is particularly critical. As the core of voltage protection in scenarios such as automotive electronics and power modules, even micron-level cracks, bubbles, or voids inside them may become "invisible bombs" for electrical breakdown and functional failure.

The test results showed that multiple samples had obvious crack and void defects, and the specific risk mechanisms are as follows:

• Crack defects: Mainly caused by sintering stress in the production process, mechanical impact during transportation, or thermal stress caused by temperature fluctuations. Under high-voltage surge scenarios, arc discharge and local overheating are prone to occur at cracks, which may轻则 cause resistor functional failure, and重则 trigger circuit short circuits, threatening the safety of subsequent equipment.
• Void defects: Mostly caused by uneven shrinkage of resin materials during packaging and gas residue during curing. Voids will reduce local heat dissipation efficiency and cause electric field concentration effect, which accelerates device aging under repeated voltage shocks and significantly shortens service life.

In fields such as automotive electronics (e.g., BMS battery management systems, radar sensors) and medical electronics that require "zero tolerance" for reliability, such potential defects are completely unacceptable. Once the device fails, it may cause serious consequences such as vehicle safety failures and medical equipment shutdowns.

Every section analysis is an in-depth diagnosis of the "health status" of components; every test report of OnePcba is a quality protection wall built for the customer's supply chain. In the follow-up, we will continue to share special detection points of other components such as capacitors and chips, so stay tuned.