Hi guys,I've observed some "weird" issue today. I'm pulling Test_out[4:0] out to my Logic so that I can read them as system status. When I do that I have intermittent link training issue. Disconnect the lines will make the problem go away. - Connect those LTSSM bits to my logic: intermittent, failed to nego every 5-6 times - Disconnect those LTSMM from my logic OR Register them first before feeding into my logic: issue goes away. Those LTSSM comes directly from HIP, any chance that lots of fanout/delay would cause internal statemachine malfunction? I repeatedly connect and disconnect the LTSSM from my logic several times and always be able to reproduced. Baffled by this thing !!! Thanks a bunch Jeff
I've tried to view the LTSSM when the link can't get up, LTSSM goes back and forth between 00000b (detect.quiet) and 00001b (detect.active) ---> what does it mean?
Further probe made me see that the Detect Sequence during Detect.Active substate is not carried out. Somehow, Mr Arria II doesn't output the Voltage stepping and detect the far-end device.... :| frustrated !!!
--- Quote Start --- Hi, Are you able to solve your intermittent issue? Also using Arria II device. I have similar intermittent ltssm issue during power-on. Regards --- Quote End --- Mine was an issue caused by the Reset controller in the Reference design. The reset controller has a statemachine which is asynchrnously reset by a signal fed directly from PLL's lock --> The statemachine goes to unknown state sometimes and the HIP can't be initialized properly. Exactly where I can't recall. Please trace it first if you still can't find it, I may need to dig up the code. Remember, reset controller, PLL's lock signal --> Resolved by a simple synchronization of Reset to the clock domain.
--- Quote Start --- Hi, Thanks, is it in the PCIE HIP verilog codes? Regards, --- Quote End --- I'm not sure where exactly. https://www.alteraforum.com/forum/attachment.php?attachmentid=7523 https://www.alteraforum.com/forum/attachment.php?attachmentid=7524 I can't confirm that you have the same issue. But here are the signals. Look at the last signals in the 2 waveforms. The states variable: ...|calibration|state Bad case: the state register goes into some wierd value, everything just stuck there Good case: the controller move on after reset is released. You can trace that signal
Hey,It seems like your is working. When the statemachine is able to run to that KICK_DELAY_OC state. In my case, when it's stuck, it's stuck right after the IDLE state and can't recover. Look at the "PowerOnStuck.jpg". The state is stuck at 00001 which is not a valid state. Jeff