Forum Discussion
New ContributorUnable to Generate Valid PAM4 Eye Diagram and BER Measurements on Stratix 10 SI E-Tile Native PHY
TTK in Quartus 18.1 Eye Viewer is really old and the newer Quartus would resolved most of the problem. I would advise to use the latest Quartus.
The E-Tile PAM4 link must use true differential TX+/TX− and RX+/RX− signaling; terminating RX− with 50 Ω and applying only 20 mVpp to RX+ is not sufficient for reliable 51-Gbps reception. A locked PLL/CDR does not guarantee valid PAM4 eye separation, so use direct differential SMA loopback first, PRBS13Q/PRBS31Q, initial adaptation with valid traffic, and disable continuous adaptation before BER/Eye Viewer. A full-red heatmap and undefined eye dimensions usually indicate invalid/weak differential data, incorrect mapping/equalization, or a Toolkit issue—not simply insufficient captured bits.
New ContributorThanks for the help so far.
I have another doubt regarding the SMA channels on the E-tile:
- My experimental setup uses single-ended signaling. Can I terminate Tx− and Rx− with 50 Ω and proceed using Tx+ and Rx+ only?
- What is the minimum acceptable peak-to-peak voltage level for the Rx input? Does the Rx input need to be differential, or can it operate with a single-ended input?
- We are getting around 500 mVpp on Tx+, which we process through some channel. After this conversion, we are getting approximately 20 mVpp, which we plan to feed into Rx+. Would this voltage level be sufficient for the Rx to reliably detect the signal?
Frequent ContributorTTK in Quartus 18.1 Eye Viewer is really old and the newer Quartus would resolved most of the problem. I would advise to use the latest Quartus.
The E-Tile PAM4 link must use true differential TX+/TX− and RX+/RX− signaling; terminating RX− with 50 Ω and applying only 20 mVpp to RX+ is not sufficient for reliable 51-Gbps reception. A locked PLL/CDR does not guarantee valid PAM4 eye separation, so use direct differential SMA loopback first, PRBS13Q/PRBS31Q, initial adaptation with valid traffic, and disable continuous adaptation before BER/Eye Viewer. A full-red heatmap and undefined eye dimensions usually indicate invalid/weak differential data, incorrect mapping/equalization, or a Toolkit issue—not simply insufficient captured bits.
