Hardware reliability test design example analysis

From a hardware perspective, reliability testing can be divided into two categories:

Reliability tests based on industry standards or national standards, such as electromagnetic compatibility tests, climate environment tests, mechanical environment tests, and safety tests.

The test items developed by the enterprise itself according to its product characteristics and understanding of quality, such as some fault simulation tests, voltage deviation tests, fast power on and off tests, etc.

The two types of reliability tests are introduced below.

1 Reliability test methods based on industry standards and national standards

Products are bound to be subjected to a lot of external stresses during their life cycle, and common stresses include business load, temperature, humidity, dust, air pressure, mechanical stress, etc. Various industry standards and national standard setters have given the stress level that a certain type of product will have under what application environment, and standard users must select the corresponding test conditions, i.e., stress level, based on the product's application environment and quality requirements. This selected stress level is essentially the product test specification.

During the product testing phase, we must apply the corresponding stress types and stress levels to enough test samples in a laboratory environment to examine the working stability of the product. For communication equipment, common test items include at least electromagnetic compatibility tests, safety tests, climate environment tests and mechanical environment tests, and the above four types of test items also include many test sub-items. For example, climate environment tests also include high temperature working tests, low temperature working tests, damp heat tests, temperature cycle tests, etc. There are many such test items, which will not be introduced in detail here. In general, all test items belong to specification compliance tests (i.e. PASS or FAIL tests), and the purpose of the test is to simulate the stress types and stress levels that the product will withstand during its life cycle to examine its working stability.

2 Reliability test methods designed by enterprises

Since the functions of network products vary greatly, the application scenarios may be various. The industry standards and national standards related to reliability testing generally only provide the test stress conditions for a certain type of product, and do not specify the working state or configuration combination under which the tested equipment is tested. Therefore, some test combinations may be omitted during test design. For example, for frame-type products, the line card type, line card installation position, message type, and system power configuration can be flexibly matched, which involves more test combinations, and there will inevitably be more extreme test combinations in these test combinations. For example, to verify the system heat dissipation performance of the frame, the worst test combination is to fully configure the line card board with maximum power on the frame with heat dissipation conditions; if the low-temperature working performance of a line card board is considered, the more extreme combination is to configure the least single board on the frame with the best heat dissipation conditions and the configured single board has the lowest power consumption, and the single board is placed in the slot with the best heat dissipation.

In short, when doing test design, it is necessary to break away from the limitations of traditional test specifications and test standards, and design tests from the perspective of product application to ensure that every hardware feature and hardware function of the product under the typical application combination, full configuration combination or extreme test combination is fully exposed to various test stresses. Only when the test in this link is guaranteed can the reliability of the product be guaranteed.

The following two examples illustrate how to design reliability testing methods based on product characteristics.

2.1 Example 1: Parallel bus test of packet processor external buffer

In order to cope with network bursts and manage traffic, the packet processor inside the network device is usually equipped with various random access memories (RAM) to cache packets. Since the packet processor and RAM are interconnected through a high-speed parallel bus, the operating clock frequency of the parallel bus may be as high as 800Mhz, and the number of signals is large and the topology is complex. With the increasing density of product devices, the product is likely to encounter serious signal quality problems such as crosstalk and switching synchronization noise (SSN). In response to the above problems, we need to conduct careful business design to fully expose the corresponding hardware circuits to adverse physical conditions to see whether they work stably.

Crosstalk is simply a kind of interference. Due to the internal and external routing of ASIC, the jump on a signal line will cause unexpected voltage noise interference to other signals. In order to improve the circuit working speed and reduce low power consumption, the amplitude of the signal is often very low. A small signal interference may cause a digital 0 or 1 level recognition error, which will have a great impact on the reliability of the system. When designing the test, it is necessary to impose a special service load on the device under test to make a large number of specific signal jumps on the tested bus, that is, to expose the bus to the largest possible crosstalk conditions, and use an oscilloscope to observe whether the signal quality of each bus is acceptable and whether the monitoring service is normal. Taking the 16-bit parallel bus as an example, in order to extreme the impact of this crosstalk, when designing the test message, 15 of the 16 signal lines (i.e., the attack signal line Agressor) have the same jump direction, that is, all 15 signal lines jump from 0 to 1 at the same time, and let another interfered signal line (i.e., Victim) jump from 1 to 0, so that all 16 lines have to traverse this situation.

Switching synchronous noise is also an unexpected physical phenomenon that may occur in RAM high-speed parallel interfaces. When the IC drivers switch at the same time, a large current that changes instantly will be generated. When passing through the inductance on the return path, an AC voltage drop will be formed, thereby generating noise (called SSN), which may affect the signal level judgment at the signal receiving end. This is a very bad working state of the parallel bus, which poses a severe test to the high-speed signal conversion capability, driving capability, dynamic response of the power supply, and filtering design of the power supply of the signal driver. In order to verify whether the product works reliably under such working conditions, a special test load, that is, a special test message, must be added to the device under test (DUT).

Example:

If the bus under test is 16 bits wide, to make all 16 signal lines flip synchronously, the message content should be:

FFFF 0000 FFFF 0000

If the bus under test is 32 bits wide, to make all 32 signal lines flip synchronously, the test message content should be:

FFFF FFFF 0000 0000 FFFF FFFF 0000 0000

If the bus under test is 64 bits wide, to make all 64 signal lines flip synchronously, the test message content should be:

FFFF FFFF FFFF FFFF 0000 0000 0000 0000 FFFF FFFF FFFF FFFF 0000 0000 0000 0000

If the message exists in the business channel inside the DUT at the same time with the bus of the above bit width, the business test must load the above message to see whether the DUT UUT works normally under each message, and at the same time perform signal testing on the corresponding bus to see whether the signal is normal.

2.2 Example 2: Thermal Test

Thermal testing uses a multi-channel spot thermometer to measure the temperature distribution of key points or key components inside the product. The test results are the input conditions for calculating the device life (such as E-Cap) and predicting product reliability indicators. It is an important reliability activity in the product development process.

Generally speaking, thermal testing is mainly to verify whether the thermal design of the product meets the product's operating temperature range specifications. It is a laboratory benchmark test, which means that in order to ensure the consistency of the test results, strict requirements must be placed on the test environment, such as requiring the device under test to be free of heat sources and forced air cooling equipment within a certain range, and the surface cannot be covered with any foreign matter. But in fact, the working environment of many products is different from the above test environment:

Some products may be placed on a table or hung on a wall when in use. These devices basically rely on natural heat dissipation. Different installation methods will directly affect the thermal convection of the device, and then affect the temperature distribution inside the device. Therefore, different installation positions must be considered when testing such devices. If the device passes the thermal test on a table under laboratory conditions, it does not mean that the device will also pass the thermal test when hung on a wall.

Some network equipment is widely used in the Internet cafe industry, and it is common to stack several devices together. When doing thermal testing on similar products, it is necessary to consider whether the thermal testing of the product meets the requirements under this condition.

Some frame-type equipment may have certain dead angles in the air duct design due to the large number of slots. If the object under test is a service board, which can be inserted into multiple service card slots at will, the board under test must be placed in the slot with the worst heat dissipation during thermal testing, and a high-power service board supported by the specification must be inserted into the slot next to it, and then the board under test can be made to work as an auxiliary board and at full load, and thermal testing can be performed under this service configuration condition.

3 Conclusion

For different product forms, the hardware reliability test items may be different, but the basic idea of ​​the test is the same. The basic idea is to completely analyze the possible application environment of the test object, and to withstand possible working conditions including extreme working conditions in the possible application environment. Various stress conditions are created in the laboratory environment, and the working conditions of the equipment are changed. Every hardware feature and hardware function of the product is exposed to various extreme stresses one by one. Missing any test combination will inevitably affect the reliability of the product.