With the rapid development of generative AI technology, low latency is particularly important in providing real-time data processing capabilities for AI and enhancing user application experience. It has become a necessary test item for end-users to verify optical modules. Therefore, the authenticity and accuracy of the optical module’s delay performance test results are extremely crucial.

However, in the past, the delay performance indicators measured by optical module manufacturers using foreign traffic analyzers often led to customer doubts due to inaccurate data. This sometimes resulted in negative outcomes, such as being at a disadvantage compared to competitors in terms of delay indicators. Consequently, many optical module manufacturers had to set up environments with oscilloscopes to test the real delay indicators of the modules (as oscilloscopes are physical layer testing instruments that can reflect physical layer performance indicators most accurately). However, setting up an oscilloscope testing environment is not only time-consuming and labor-intensive but also cannot test under various real traffic conditions. What’s more frustrating is that the delay data tested with oscilloscopes and foreign traffic analyzers are not only inconsistent but also vary greatly (in a real case, the same optical module had a delay result of 90ns tested with an oscilloscope, while a foreign traffic analyzer showed 220ns). Additionally, the delay test results of foreign traffic analyzers vary under different Ethernet application modes (such as 400GE, 200GE, 100GE).

To address this industry demand and pain point, Hywicom, based on a deep understanding of underlying implementations and international communication standards, has pioneered a delay testing solution for traffic analyzers based on the physical layer. Unlike the delay testing solution of foreign instrument manufacturers that timestamp at the IP layer packet generation, Hywicom’s testing principle and implementation method are: on the sending side, timestamp at the exit near the physical layer (i.e., the signal of the instrument’s physical layer connecting to the optical module), and on the receiving side, immediately extract the timestamp at the physical layer entry (i.e., the signal of the optical module entering the instrument’s physical layer). Delay calculation is then performed. Since all timestamp generation, extraction, and calculation occur near the physical layer, the delay performance and accuracy are not affected by Ethernet upper layer application modes, i.e., the delay measured is basically consistent under different Ethernet application modes such as 800GE, 400GE, 200GE, 100GE, 50GE, 25GE, 10GE. This is because the optical module itself is a physical layer device and is invisible to the upper Ethernet application modes, so the delay test results should not be affected by them.