Constant PCA failing / reset - PC build issue?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • jpdw
    Automated Home Guru
    • Oct 2007
    • 169

    Constant PCA failing / reset - PC build issue?

    I'm still building / tweaking my Cortex PC to drive to as low-consumption as possible. It's an EPIA EN1200, currently in an available ATX chassis.

    Till now, it's been powered by a spare 400w ATX PSU but I am trying to change this to a silent 60w PSU set, complete with laptop-style brick, from linitx (this one).

    But, either Cortex, the motherboard or PCA fails whenever I try it.

    It continually reports hardware errors and tries resetting the PCIF then cycles... and cycles. Swap the 400w ATX PSU back and it's all fine again.

    The power source is the ONLY thing that's different. The only other device drawing power from the active PSU is a 2.5inch SATA HD, so I can't be overloading the PSU.

    My best guess is that is a ground-related issue causing the PCD problems , but I'm at a loss to work out why - I've even tried leaving the ATX PSU in the chassis with mains cable connected in case the chassis needed a route to mains earth maintained. The rest of the Idranet connected parts are either idranet-only or supposed to be isolated so the change of PC PSU shouldn't affect it.

    Below is the output that cortex shows... sorry it's long, but you'll get the idea that it just repeats....

    Has anyone got any ideas or seen anything similar ?

    Edit: I'm using UNshielded twisted pair for the Idranet bus

    ----------------
    16:55:10 PCIF hard reset
    16:55:11 RS232 Purge buffers
    16:55:11 The hardware detected a break condition
    16:55:11 RS232 Purge buffers
    16:55:11 PCIF hard reset
    16:55:11 RS232 Purge buffers
    16:55:12 The hardware detected a break condition
    16:55:12 RS232 Purge buffers
    16:55:12 PCIF hard reset
    16:55:12 RS232 Purge buffers
    16:55:12 The hardware detected a break condition
    16:55:13 RS232 Purge buffers
    16:55:13 PCIF hard reset
    16:55:13 RS232 Purge buffers
    16:55:13 The hardware detected a break condition
    16:55:13 RS232 Purge buffers
    16:55:13 PCIF hard reset
    16:55:14 RS232 Purge buffers
    16:55:14 The hardware detected a break condition
    16:55:14 RS232 Purge buffers
    16:55:14 PCIF hard reset
    16:55:14 RS232 Purge buffers
    16:55:16 Invalid nmf 00
    16:55:16 Invalid nmf 60
    :: :: ::
    16:55:17 Invalid nmf E0
    16:55:17 Invalid nmf F8
    16:55:19 The hardware detected a framing error
    16:55:19 RS232 Purge buffers
    16:55:19 PCIF hard reset
    16:55:20 RS232 Purge buffers
    16:55:20 The hardware detected a break condition
    16:55:20 RS232 Purge buffers
    16:55:20 PCIF hard reset
    16:55:20 RS232 Purge buffers
    16:55:21 The hardware detected a break condition
    16:55:21 RS232 Purge buffers
    16:55:21 PCIF hard reset
    16:55:21 RS232 Purge buffers
    16:55:21 Invalid nmf 06
    16:55:21 Invalid nmf 78
    :: :: ::
    16:55:22 Invalid nmf 1E
    16:55:22 RS232 CHECKSUM ERROR 18E01E1898
    16:55:22 RS232 Purge buffers
    16:55:22 PCIF hard reset
    16:55:22 RS232 Purge buffers
    16:55:23 The hardware detected a break condition
    16:55:23 RS232 Purge buffers
    16:55:23 PCIF hard reset
    16:55:23 RS232 Purge buffers
    16:55:24 The hardware detected a break condition
    Last edited by jpdw; 1 November 2008, 06:19 PM. Reason: Added not about UTP
    Jon
  • Karam
    Automated Home Legend
    • Mar 2005
    • 863

    #2
    The problem appears to be at the RS232 side. I think it probably is a grounding issue or rather floating 0V line. Try common grounding of 0V on IDRANet supply and EPIA sides. Unshielded twisted pair for IDRANet bus should not have an adverse effect on IDRANet communications.

    Comment

    • KirasHome
      Automated Home Jr Member
      • Oct 2006
      • 37

      #3
      Hi jpdw,

      I'm using the same motherboard are you (EPIA EN12000G), in a Travla C158 Mini-ITX Case but with an External 120W Universal 110/240V 19V AC adapter and internal 120W 19V DC converter. My HDD is a 3.5in Samsung SATA. I've not seen any problems with it.

      Just a follow on thought from your comment about the earth, I'm wondering if the brick PSU is just struggling slightly to supply the power you need and that's affecting the RS232 voltages (http://en.wikipedia.org/wiki/RS232#Voltage_levels). Can you get a power meter on the setup? Or check the voltages on the RS232.

      Jamie.

      Comment

      • toscal
        Moderator
        • Oct 2005
        • 2061

        #4
        Certain models of EPIA mini-itx boards have had known grounding issues. My mini-itx is an MII 12000. I solved it by running a ground wire between the PSU board, its running on a brick style 80W PSU and converter board , and the motherboard. The problems I had was bad RS232 coms or continuous com chatter on the serial port, and also I had problems booting the machine up. Since adding this wire the machine has been OK.
        It might be worth checking the earth in the mains socket.
        Also the EPIA EN15000G I know uses about 30W to 50W depending on the type of power supply used, with just a hard drive connected. So I guess you should get similar figures with your 12000. It should be OK but it could be on the limit. What is the internal DC to DC PSU converter board rated at.
        Good luck
        Last edited by toscal; 3 November 2008, 10:19 AM.
        IF YOU CAN'T FIX IT WITH A HAMMER, YOU'VE GOT AN ELECTRICAL PROBLEM.
        Renovation Spain Blog

        Comment

        • jpdw
          Automated Home Guru
          • Oct 2007
          • 169

          #5
          Thanks all for the many comments....

          Originally posted by Karam View Post
          The problem appears to be at the RS232 side. I think it probably is a grounding issue or rather floating 0V line. Try common grounding of 0V on IDRANet supply and EPIA sides. Unshielded twisted pair for IDRANet bus should not have an adverse effect on IDRANet communications.
          Indeed, using UTP should definately not be the cause - I was only mentioning it so it was clear I wasn't using STP and grounding the shield all over the place (as noted in one of your guide documents)

          Comparing the 0v on the idranet bus with the 0v line on the PC shows virtually no difference between the two PSUs (~1.x mV for the ATX, ~2.2mV for the fanless set). Linking the two, similarly, seemed to make no difference (admitidly this test was done with a flying lead rather than making proper connections, so maybe it really would have worked -- all have another go at this one).

          Is the "common ground" pin of the RS232 connector used on the PCD? Isn't that supposed to make this ground connection?

          Originally posted by KirasHome View Post
          Just a follow on thought from your comment about the earth, I'm wondering if the brick PSU is just struggling slightly to supply the power you need
          It's supposed to be a 60w. Reading the wattage drawn from the mains, it's 26w with the ATX PSU and ~18w with the fanless set, so well within the 60

          Originally posted by KirasHome View Post
          and that's affecting the RS232 voltages (http://en.wikipedia.org/wiki/RS232#Voltage_levels). Can you get a power meter on the setup? Or check the voltages on the RS232.
          Only thing I can check at the moment is the levels on the ATX supply. The 5v is 4.88 and the -12v is -11.31 . I wonder if this is bit too low. (I assume the chipset will use the available -12v rather than synthesise it from the -5v).

          Originally posted by toscal View Post
          Certain models of EPIA mini-itx boards have had known grounding issues. My mini-itx is an MII 12000. I solved it by running a ground wire between the PSU board, its running on a brick style 80W PSU and converter board , and the motherboard. The problems I had was bad RS232 coms or continuous com chatter on the serial port, and also I had problems booting the machine up. Since adding this wire the machine has been OK.
          Sounds exactly like what I'm seeing. Isn't this what the various GND pins in the ATX cable should be doing? Can you be more specific about where on the motherboard you connected it (shield of the RS232? Common GND pin of the RS232 etc...?)

          Originally posted by toscal View Post
          It might be worth checking the earth in the mains socket.
          Ha ha.. I have a couple of concerns about the sockets in this room since I discovered the circuitous route the circuit takes... BUT I dont that this would be the cause of this problem... If the earth was dodgy / missing, wouldn't this affect BOTH the ATX PSU and the fanless?


          To investigate further I need to haul the PC up to my desk... but currently there's no idranet there, so I need to wait till it's warm enough that I can down the network and "borrow" the PSU/MPD and setup a mini test network. Give me an excuse to "play" at the weekend instead of feeding the babies... :-)

          Meanwhile it'll be back to burning the additional 10w of power to run the ATX PSU :-(
          Jon

          Comment

          • toscal
            Moderator
            • Oct 2005
            • 2061

            #6
            This is what I did.
            1. I cut a length of wire, about 10cm.
            2. Crimped a round terminal on to each end. Make sure the hole is large enough to fit the screws that attach the motherboard to the chassis.
            3. Attach one end to the screw nearest to the PS2 connector.
            4. Attach the other end to the ground on the DC to DC converter board, normally one of the screws attaching the board to the chassis will do.

            If I can I will see if I can get a photo or two for you.
            IF YOU CAN'T FIX IT WITH A HAMMER, YOU'VE GOT AN ELECTRICAL PROBLEM.
            Renovation Spain Blog

            Comment

            • jpdw
              Automated Home Guru
              • Oct 2007
              • 169

              #7
              Originally posted by toscal View Post
              This is what I did.
              1. I cut a length of wire, about 10cm.
              2. Crimped a round terminal on to each end. Make sure the hole is large enough to fit the screws that attach the motherboard to the chassis.
              3. Attach one end to the screw nearest to the PS2 connector.
              4. Attach the other end to the ground on the DC to DC converter board, normally one of the screws attaching the board to the chassis will do.

              If I can I will see if I can get a photo or two for you.
              Okay, that makes sense - I see now what you are commoning -- making sure the chassis ground is good.

              I'll give that a go later on.
              Jon

              Comment

              • jpdw
                Automated Home Guru
                • Oct 2007
                • 169

                #8
                Originally posted by Karam View Post
                Try common grounding of 0V on IDRANet supply and EPIA sides.
                Originally posted by toscal View Post
                This is what I did.
                1. I cut a length of wire, about 10cm.
                2. Crimped a round terminal on to each end. Make sure the hole is large enough to fit the screws that attach the motherboard to the chassis.
                3. Attach one end to the screw nearest to the PS2 connector.
                4. Attach the other end to the ground on the DC to DC converter board, normally one of the screws attaching the board to the chassis will do.

                If I can I will see if I can get a photo or two for you.
                Unfortunately it seems that neither of these (either separately or in combination) solves the problem.

                The bus voltages out of the DC/DC board are marginally below those from the ATX PSU -- but only by ~0.2v. (Measured under the load of the motherboard).

                I'm not sure what to try next. If I had a 'scope I'd be looking at the levels on the serial cable but I don't :-( . It's still bugging me that the only difference to the setup between working and not working is the PSU.
                Jon

                Comment

                • toscal
                  Moderator
                  • Oct 2005
                  • 2061

                  #9
                  Can you try another brick type psu. As you may have a faulty one. OR try one of a higher wattage, at least then you will know if its the PSU.
                  IF YOU CAN'T FIX IT WITH A HAMMER, YOU'VE GOT AN ELECTRICAL PROBLEM.
                  Renovation Spain Blog

                  Comment

                  • Gumby
                    Moderator
                    • May 2004
                    • 437

                    #10
                    Maplin do an ATX PSU checker device, may be this would help. If you are anywhere near Gloucestershire I can lend you one.
                    ----------------------
                    www.gumbrell.com

                    Comment

                    • Karam
                      Automated Home Legend
                      • Mar 2005
                      • 863

                      #11
                      Originally posted by jpdw View Post
                      I'm not sure what to try next. If I had a 'scope I'd be looking at the levels on the serial cable but I don't :-( . It's still bugging me that the only difference to the setup between working and not working is the PSU.
                      Did you notice whether you got RS232 errors reported when Cortex is static (not running network)?

                      When you commoned 0V did you also connect to earth? You may want to try this.

                      I suppose a possibility is that the DC/DC convertor (which I've just seen is a pcb) might be putting out high levels of emmissions which are somehow getting on the RS232 signal lines.

                      Comment

                      • jpdw
                        Automated Home Guru
                        • Oct 2007
                        • 169

                        #12
                        Originally posted by Karam View Post
                        Did you notice whether you got RS232 errors reported when Cortex is static (not running network)?

                        When you commoned 0V did you also connect to earth? You may want to try this.

                        I suppose a possibility is that the DC/DC convertor (which I've just seen is a pcb) might be putting out high levels of emmissions which are somehow getting on the RS232 signal lines.
                        The original trace is from when Cortex "just monitoring" the network -- ie it's not running the network but the comms window is listing idranet chatter from the reflexes running on the modules (ie decoding various temperatures etc). I guess this is what you mean as static.

                        When I tried the 0v link, I *didn't* also link to (mains) earth.

                        I wonder if the noise could be on the supply lines out of the DC/DC? Maybe I should add some decoupling to the motherboard supply between +5 & ground and +12 & ground in case.
                        Jon

                        Comment

                        • Gumby
                          Moderator
                          • May 2004
                          • 437

                          #13
                          I know this sounds obvious, but have you checked the PCA cable? I'm wondering if you could have a problem with the RS232 ground in the cable.
                          ----------------------
                          www.gumbrell.com

                          Comment

                          • jpdw
                            Automated Home Guru
                            • Oct 2007
                            • 169

                            #14
                            Originally posted by Gumby View Post
                            I know this sounds obvious, but have you checked the PCA cable? I'm wondering if you could have a problem with the RS232 ground in the cable.
                            No I haven't done that and yes, it is rather obvious. Oops. I assumed that the one Idratek supplied is of good quality and fully grounded :-)

                            I should have another one somewhere, just need to find the right box.
                            Jon

                            Comment

                            • Karam
                              Automated Home Legend
                              • Mar 2005
                              • 863

                              #15
                              It is only grounded if you ground it (via 0V). Try earth - via a capacitor if concerned.

                              Comment

                              Working...
                              X