AGX THOR T5000 Mgbe QSFP28 25G Loss dB and length confirm

I have reviewed the Jetson_Thor_Series_Modules_DesignGuide_DG12084001_v1.5, which contains the description for MGBE 25G.


The length of C2M does not exceed 46mm. However, when I checked the agx thor T5000 reference design, the longest trace for RX was 73mm, and for TX it was 68mm. Both the length specifications did not meet the requirements of the design guide.


The length of the QSFP wiring in the original version also fails to meet the specifications. Could you please explain why?Thanks

Moving to Jetson AGX Thor forum.

Hi,

As described in Note2 for Table 7-25 of the Design Guide, the 46mm C2M trace length is based on EM370z. If you use a different PCB material or stackup, please ensure that your configuration meets the specified loss budget.

Thank you,

Hi,

I understand what you mean. As long as my dB loss is less than 4.15, then the trace length can exceed 46mm?
According to your statement, was the dB loss of the agx thor T5000 reference design not exceeding 4.15?
Does this mean that the agx thor T5000 reference design uses better materials than the EM370z (FR4)?

Hi,

Yes, the 46mm and 4.15dB in the Design Guide are based on EM370z material. If you use a lower-loss material, longer traces are acceptable as long as the channel insertion loss remains within the loss budget.

We are unable to share internal details about the PCB material used in the reference design.

Thank you,

Could you please tell me if NV has used an oscilloscope to test the eye diagrams of QSFP28 25G or 10G? Could you please tell us if the result is PASS?

Could you please provide any update for this topic?