[moved] Is there any high-frequency effect Spectre cannot take account of?

[xiangx93]

I've designed a PAM4-Transmitter with a Nyquist-frequency of 10GHz.Its test result came out far from expected.Apart from the IC itself,I use S-parameters module to act as a low-pass filter to get a general influence of PCB traces&Cables&Connectors in the design process.
Even so,the outcome is much worse than what I got in Spectre.

I am wondering if this is from things like skin effect,dielectric loss or anything becomes severe only at high frequencies.Is there any way to precisely predict behavior of the circuit at ~10GHz?Or we can barely do anything about it?

 

[FvM]

I suppose, Specctre can "take account of" parasitic wiring effects. But it's a different question if you manage to predict it appropriately. The effects can be modeled.

We see a considerable amount of noise in the eye diagram, also a probably unwanted fs/2 component. But we can hardly decide if it's due to unsuitable signal probing or insufficient signal quality.