mrjava241
Newbie level 4
Hi All,
I have a one-off board with an unstable PCI bus (1 out of 30 boards built has this issue).
The design is as follows:
CPU->South Bridge (Ibex Peak)->PCIe->PCIe-to-PCI bridge (PEX8112)->PCI devices.
The PCI bus is running at 33Mhz. The problem is as follows:
On initial boot, all PCI devices are detected and operational. They will stay operational with no issues for 12++ hours while writing/reading in each PCI devices memory space.
When the board is power cycled, two of the PCI devices are dropped from the tree. The board has to be turned off for 5-10 minutes for it to return to a normal state (all PCI devices detected).
I've done the following:
When in the bad state, I checked the following signals:
PCI Clocks:
DST_PCIE_CLKN -> OK (PCIe clock signals going to the PCIe-to-PCI bridge)
DST_PCIE_CLKP -> OK
PCI_BRIDGE_A\PCLK0 -> OK
Checked if the following signals were pulled low:
PCIRST# -> HIGH
PCISERR# -> HIGH
PCIPERR# -> HIGH
MASTER_SYS_RESET# -> HIGH
Checked if PCIe link from south bridge to PCIe-PCI bridge was established -> Yes.
Checked PCI_CLK -> clock waveform @ 33Mhz.
I've also checked to ensure all PCI config pull up resistors were at their appropriate values -> Yes.
When the system goes into the bad state (missing PCI devices), I've scanned the JTAG chain and the missing PCI devices do appear in the JTAG list.
I've checked impedences w.r.t ground on the following signals.
PCI_ADDR 0-31, impedences ranging from 1.17M-1.31M. Nothing unusual found.
PCI_CBEN 0-3: impedences ranging from 1.26M - 1.28M. Nothing unusual found.
It seems like there is a power discharge / capacitance problem however I'm running out of ideas...
Has anyone seen a similar PCI issue or does anyone have any ideas?
I have a one-off board with an unstable PCI bus (1 out of 30 boards built has this issue).
The design is as follows:
CPU->South Bridge (Ibex Peak)->PCIe->PCIe-to-PCI bridge (PEX8112)->PCI devices.
The PCI bus is running at 33Mhz. The problem is as follows:
On initial boot, all PCI devices are detected and operational. They will stay operational with no issues for 12++ hours while writing/reading in each PCI devices memory space.
When the board is power cycled, two of the PCI devices are dropped from the tree. The board has to be turned off for 5-10 minutes for it to return to a normal state (all PCI devices detected).
I've done the following:
When in the bad state, I checked the following signals:
PCI Clocks:
DST_PCIE_CLKN -> OK (PCIe clock signals going to the PCIe-to-PCI bridge)
DST_PCIE_CLKP -> OK
PCI_BRIDGE_A\PCLK0 -> OK
Checked if the following signals were pulled low:
PCIRST# -> HIGH
PCISERR# -> HIGH
PCIPERR# -> HIGH
MASTER_SYS_RESET# -> HIGH
Checked if PCIe link from south bridge to PCIe-PCI bridge was established -> Yes.
Checked PCI_CLK -> clock waveform @ 33Mhz.
I've also checked to ensure all PCI config pull up resistors were at their appropriate values -> Yes.
When the system goes into the bad state (missing PCI devices), I've scanned the JTAG chain and the missing PCI devices do appear in the JTAG list.
I've checked impedences w.r.t ground on the following signals.
PCI_ADDR 0-31, impedences ranging from 1.17M-1.31M. Nothing unusual found.
PCI_CBEN 0-3: impedences ranging from 1.26M - 1.28M. Nothing unusual found.
It seems like there is a power discharge / capacitance problem however I'm running out of ideas...
Has anyone seen a similar PCI issue or does anyone have any ideas?