What is the appropriate bandwidth for an oscilloscope?

This is a common problem: I want to measure the 8Gbps PCIE (and SAS/SATA/USB/LVDS, etc.) serial bus. Can a 16G oscilloscope do it? Can a 12G oscilloscope do it? Can an 8G oscilloscope do it? For high-speed serial buses, what bandwidth is sufficient for the oscilloscope? The simplest way is to assume that the bus transmits signals with intervals of 0 and 1, that is, a square wave clock. The base frequency of the 8Gbps signal is 4GHz. According to the 5x bandwidth principle, the oscilloscope bandwidth should be 20GHz. In other words, multiply the data rate by 2.5, and that is the recommended bandwidth. Isn't it 3~5x bandwidth? Why not count 3x? How much money is different... The answer is: Of course, can a 12GHz oscilloscope measure an 8Gbps signal? Why not? In the eyes of testers, there is no such thing as whether it works or not, only whether it is accurate or not. As the popular saying now, talking about toxicity without considering the dose is just playing tricks. So how do you tell whether it is accurate or not? You will know after you try it yourself. I used BertScope to make a 12Gbps (SAS3 rate), PRBS7, 500mV serial signal, and sent it to MS073304DX, 33GHz oscilloscope using SMA cable. Then I used the bandwidth-limited method to perform an eye diagram test. 33GHz, 5.5x bandwidth, it looks very good. The measured rise time is 30.7ps.

25GHz, 4.2 times bandwidth, no change, the measured rise time is 30.5ps

20GHz, 3.3 times bandwidth, looks good, the measured rise time is 31.7ps

16GHz, 2.6 times bandwidth, the edge has obvious changes, and the measured rise time is 37ps

12Ghz, 2 times bandwidth, close to sine wave, the measured rise time is 42ps

8GHz, 1.3 times bandwidth, the line starts to become thicker, and the measured rise time is 52ps

6GHz, baseband bandwidth, barely readable, measured rise time is 140ps

5GHz, the bandwidth is totally insufficient, I can’t watch it…

After reading this, everyone should have a basic understanding of the bandwidth requirements of ordinary high-speed serial buses, right? If the accuracy requirement is high, 4-5 times the bandwidth or more is necessary. Starting from 3 times the bandwidth, the accuracy of the rise time begins to decrease, which will also affect the degradation of various parameters such as jitter. Now many high-speed serial bus consistency tests have index requirements for rise and fall time and jitter. So if you want to do consistency testing, try not to reduce the configuration. If you are just measuring the eye diagram, or the error tolerance of the test results is relatively high, then it is acceptable to reduce the bandwidth of the test system. So, can the 25.78Gbps electrical signal be tested with a 33GHz oscilloscope? Look~

Of course it is available. However, if you want to perform consistency testing, it is better to use an oscilloscope. After all, at such a high speed, the dynamic effective bit of a real-time oscilloscope is relatively low. Of course it is available. Is 70GHz bandwidth enough? Is it enough? For industry, a 33GHz oscilloscope basically covers most applications. Of course, if you need a higher bandwidth, please feel free to ask us for support.