The bit error rate is stated to be within the standard, but the test duration is only a few tens of seconds. This kind of report is difficult to reproduce in the field. BER is a statistical quantity, and the conclusion cannot be separated from the sample size. It is also necessary to distinguish whether the error measured is before correction or after correction.

First calculate how many bits to transmit

To confirm with 95% confidence that the bit error rate is less than 1E-12, at least about 3×10^12 bits need to be transmitted. Converted into duration: about five minutes on the 10G side, about two minutes on the 25G side, and about half a minute on the 100G side to meet the sample size; otherwise, as long as one bit error is detected, the conclusion must be revised downwards. A report that is not long enough but gives beautiful numbers has limited interpretation value. For batch testing, it is even more important to write down both the duration and the speed clearly.

The number before correction and after correction are not the same

• First look at the bit error before correction and FEC correction count. Zero after correction does not mean that the link is healthy.
• Pay attention to uncorrectable code words. If they occur once, retest and check for jitter and reflection.
• Errored seconds and severe errored seconds are suitable for long-term observation and are closer to business experience than instantaneous readings.
• The statistical characteristics of instrument test patterns and real business traffic are different. The two sets of calibers cannot be mixed for judgment.

Confirm the link and instrument before reading

First perform a self-loop to verify that there is no problem with the instrument and adapter itself, and then install the device under test. Bent test leads and dirty connectors will give stable but erroneous low margin conclusions. For batch testing, one instrument and one set of wires are fixed. If the instrument is changed midway, it will not be merged into the same table.10G DAC high speed cableThis type of passive copper cable itself does not produce bit errors. The measured differences mostly come from the contact and peer equipment. The interpretation must be compared with the same piece.PCI-EX16 100G single optical port fiber optic network card (Intel E810 chip)The flow can be used as cross-validation on the business side.

Only by writing down the rate, code type, duration, before and after correction, and the instrument number can we make a judgment. The EB-LINK factory error code sampling test is based on this standard. The specific threshold can be confirmed by calling 0755-83179002.