Problem with BER on simulation of QPSK modulator

Hi, I am simulating a traveling wave IQ modulator and I have a problem with the calculation of BER. I just realized that a part of the received data is not correct.
The below figures are showing the received data and reference date generated by PSRBS element, as it can be seen for a period of time received data are not correct.
image

It should be noted that this part happens at different times for I and Q signals.
image

For now, I am using ignore options in EYE and VSA elements and ignore nearly half of the periods to make sure that part is not affecting my results. If I don’t use ignore options, my resulted eye diagrams would be like the following figure and my measured BER would be affected by the lines at the zero level.

I am attaching my file here:
IQ_60Gb_SPB8_SL2048_Test.icp (7.0 MB)

I will appreciate if anyone can help!

Hello @alireza.geravand.1,

Thank you for the question. I believe that this section with no signal corresponds to the beginning of the simulation, when the signal is still propagating through the circuit and has not yet reached the eye diagram and VSA elements.

It looks like the issue has something to do with the clock recovery. If you set signal reference input to false for the VSA and the eye diagram elements, it will move this section with no signal to the beginning of the output. If you then increase ignore start periods to 128 it will remove this part from your eye diagrams, without ignoring too much of the signal.

I hope this helps. Let me know if you have any questions.

Thank you @kjohnson for your helpful reply.

As you mentioned by disabling the signal reference input option that period of time will move to the beginning of the output and with a lower ignore start periods the problem of eye diagram would be solved.

But disabling signal reference input for VSA would affect the resulted constellation diagram. First following figure is the constellation diagram for the case with signal reference input and ignore start and end symbols equal 700. The second figure is the constellation diagram for the case without signal reference input with ignore start symbols equal to 128. Clearly, there is a difference here and I don’t know if it is related to signal reference input or automatic decision point feature or not.

signal reference input: True
ignore start symbols: 700
ignore end symbols: 700
image

signal reference input: False
ignore start symbols: 128
ignore end symbols: 8
image

I believe the difference in the constellation diagrams is due to the I and Q signals not being properly synchronized when you include the references due to incorrect delay compensation. We can look at the vector diagrams with reference inputs:
image
And without reference inputs:

We can see that the vector diagram without reference inputs is closer to what we would expect. The analyzers generally work well without the reference inputs, I would recommend that you not use them in your simulation.

As an aside, your current samples per bit setting (8) is a bit low, it would probably improve your results if you increased this to 64.

Please let me know if you have any other questions.

Thank you @kjohnson for your help.

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.