This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TUSB7320 Start-Up Issue

Other Parts Discussed in Thread: TUSB7320, TUSB7340

We are using the TUSB7320 in an embedded board application (not a PCIe adapter card), and cannot enumerate the device over PCIe.

Power-up sequence looks correct, with GRST# deasserted about 3 msec after both supplies are stable. PERST# is deasserted about 200msec after deassertion of GRST# and presence of ref_clk. Recommended 48MHz crystal topology is being used.

1. Can ref_clk be present before GRST# is deasserted? (Not clear from Figure 3-1 in datasheet.)

Being an embedded application, neither WAKE# nor CLKREQ# support is required. Both pins are unconnected in the design.

2. Is this an issue given the erratum regarding oscillation of the ESD structure? Could this affect our ability to enumerate the part over PCIe?

Both 3.3V (i/o) and 1.1V (core) supplies are derived from low-noise switching supplies (TPS54626PWP & TPS54320RHL). An LC similar to that of the Demo EVM (SLLU146C) is used on the analog supply. Proper decoupling and bulk capacitor recommendations have been followed throughout).

3. Could any supply noise affect our ability to enumerate over PCIe?

  • Hello,
    Can you provide your schematic?
    Yes, REF_CLk can be present before GRST# is de-asserted.
    If WAKE# and CLKREQ# are disconnected then I don't think this will prevent the TUSB7340 to be enumerated, however, by PCIe Spec, these signals must have a pull-up on the host side.

    Yes, supply noise can affect the link training. Can you provide a scope capture of the power-up sequence?
    Regards
  • After a bit of software debug , it looks like the chip is enumerating now.

    Elias - do you recommend pullups on the WAKE# and CLKREQ# pins?

    My first instinct is to leave them floating (since the erratum is related to a fast rising edge on either signal). This is an embedded design (TUSB7320 is on main board) with only a rudimentary PCIe implementation (SERDES TX/RX lanes, ref_clk, reset). Neither WAKE# nor CLKREQ# route to or are supported by the processor.

    Thanks...

    Dave

  • Hello,

    Pull-ups are not required if those terminals are unused.

    Just be sure to implement the errata's workaround if the scenario described there is going to be your case.

    Regards