Tool/software:
please
regards,
Robin Liu
Hi Robin,
Ill have a look and get back to you tomorrow after validating the lengths.
Hi Robin,
Are the total lengths only the pre-channel or is it the pre-channel and post-channel.
I see the total length is in two buckets. 4 inches and 11 inches.
11 inches is a long trace and this will be on the higher end of the equalization, especially with 2 vias. This is a very risky length.
The 4 inches is a better length, but 4 vias in the signal path will lead to significant reflections especially at 10Gbps.
It may be helpful using the TI Trace budgeting tool here. Use the linear redriver source application tab and fill out all the info you can about the proposed PCB board.
Trace tool:
RX & TX trace length budgeting_Imperator.xlsx
password: ti
hi Vishesh,
M1, M2 is pre channel length, M3,M4,M5,M6 is post channel length, Via count in total in pre- channel + post channel;
4 inches and 11inches, yes. it is;
so need BU help judge if the 4inches and 11inches condition is ok to implement?
trace loss is around 0.936dB/inch
Hi Robin,
11 inches is usually a risky length so we will need to compensate by adding additional trace width and reducing the post-channel length.
You can adjust the parameters in the calculator tool to get a result that allows for 11 inches as a viable solution.
Through hole vias also increase risk, so I recommend having as few as possible.
Hi robin,
In our meeting today we discussed the platform that is used in the customer design. They used a 9in pre-channel trace in a previous generation with intel Arrow lake that was successful design. They are looking to push the trace length to 12 inch with the new platform on intel nova lake. I will reach out to our systems engineer to see if we have any documentation regrading max USB trace length for the intel nova lake.
Hi Robin,
I was able to get the PDG for the Novalake. In section 5.1.5.1 it describes the maximum trace length for USB Type-C.
hi Vishesh,
110mm = 4.3inches
so we redriver can add another about ~12inch,
so total can support 16inches?
Hi Robin,
I believe the 12ich approximation is coming from 12dB max gain provided by the redriver. The approximation is about 1dB/inch which is correct. However the assumption behind this approximation is that the trace is straight and has no vias. I believe 11inches pre-channel has a possibility of failing compliance due to the channel length + vias + curves in the traces. The is also a possibility we pass without issue, but I believe this is still a riskier option.
Additionally, Are you able to confirm with the SoC vendor what the EQ capability of the SoC is? The TUSB1142 can only compensate in the RX, so the SoC will also have to compensate for the 12inch channel in the TX directions.
Ex) white boxes represent where EQ can be applied.
hi Vishesh,
I think this case can be close, i already send you the layout to help check; thanks~