0

I need help quantifying Eye Diagram. I have to quantify Eye Diagram at the end of my Thesis for my Serial Link channel. Viewing an Eye Diagram, I can say probably whether it is bad or good. But for the hypothesis or any conclusion qualitative analysis is not enough. I need to have quantitative analysis. I know what parameters are there like jitter, eye width, height, crossing level etc. etc. But how to say up to which value of parameters the eye diagram is good. like how can I decide that like 35% BER or "x" ps jitter is tolerable for the serial link (Though it depends on circuit & signaling techniques etc...)? Simply to ask what separates a good ee diagram from a bad one?

If you have any note or suggestion or anything regarding my specific question will be appreciated also.

aguntuk
  • 13
  • 1
  • 5
  • 1
    35% BER is horrible. Usually BER is indicated by the factor of 10, such as 10^-6 or 10^-9. And usually you want BER < 10^-12 unless you also have FEC. – alex.forencich Jan 19 '18 at 17:46
  • @alex.forencich thank you for your reply. I just used 35% as an example. I want to know, is there any documentation for that to quantitatively decide a good eye diagram? Because not only BER, there are also many parameters also... – aguntuk Jan 19 '18 at 21:45
  • BER is what counts. Having some tolerance to jitter in the incoming signal and in the clock-recovery-PLL will be useful. Often the input signal drives a diffpair used as a limiter, to remove amplitude variation and noise. If the channel stores energy in resonant components (ESD diodes and bond-pad inductances, for example) life can be very bad, what with ISI. – analogsystemsrf Jan 20 '18 at 03:31

2 Answers2

1

It is common to use an eye mask to determine if an eye diagram is sufficiently open. Different standards use different masks.

EE_socal
  • 1,402
  • 7
  • 6
  • thanks a lot. Can you give me any good concrete reference where this mask is used for quantitative analysis of eye diagram to conclude that it is good or bad? And what is the measurement for eye mask? – aguntuk Jan 19 '18 at 21:48
  • Typically the eye mask is overlayed on top of the eye diagram. If any part of the eye touches the mask then it is considered a failure. This is usually done on a oscilloscope that captures the waveform and has eye masks that can selected and displayed. This kind of oscilloscope usually has very high bandwidth and differential probes and can be very expensive. Keysight is one company that makes this type of equipment so you can look on their web site to get an idea of how it works. – EE_socal Jan 19 '18 at 23:03
0

Ultimately, the most important factor for a serial link is the bit error rate or bit error ratio, because this ultimately determines the quality of the complete communications channel. However, the BER is a rather high-level parameter as it is a result of all of the components of the link interacting with each other. If you look at each end in isolation, then this is where you have to look at eye masks, Q factor, jitter and jitter tolerance, etc. to try to get an idea of how good the individual components will perform when assembled into a complete system. If you're working with an existing communications standard, then the standard should have some specifications that components would need to meet. If you want a single number that's not BER, then Q factor is probably a pretty good starting point. This may or may not be sufficient for what you're trying to accomplish.

alex.forencich
  • 40,922
  • 1
  • 69
  • 109