File download problem

[Frank4d]

I am having a problem at work and am not sure whether it is a XP Pro, networking, or cabling issue so I'll post it here.

I have a XP Pro computer with P4 Xeon CPUs and a 10/100/1000 NIC that is plugged into an unmanaged Linksys 16 port switch. Also plugged into the switch is an avionics box (mission computer for a Navy helicopter) containing two DY4 Systems 182 and 712 PowerPC CPU cards (physically separate cards) with 10/100 NICs.

When the avionics box is powered on, both it's CPUs attempt to load the same VxWorks file from the PC via FTP. Only the avionics box is mis-configured... one CPU should be loading VxWorks for a DY4 712 board, and the other (incorrectly configured) CPU should be loading a VxWorks file for a DY4 182 board. Result is that neither CPU can load the file and time out error occurs for both.

When I built the ethernet cable between the avionics box and the Linksys switch, I cheated a little. I ran a CAT 5E cable and used the orange & green pairs for one CPU, and the blue & brown pairs for the other CPU. I don't THINK that should matter.... does it?

As an experiment, I bypassed my cable and used two separate CAT 5E cables... which partially solved the problem (the CPUs load VxWorks properly about 1/2 the time). So I thought it's a cable problem.

But then... I configured the DY4 182 card to load the proper VxWorks file from a different folder on the PC, and that also solves the problem 100%. Both CPUs always load VxWorks... with the original "shared" CAT 5E cable or with separate cables.

My guess is that I have two problems:
1. The PC can't handle two CPUs concurrently trying to download the same file via FTP at 100Mbps.
2. I need to re-wire my "shared" cable.

I would prefer a software fix. Can anyone tell be if my cabling breaks the rules?

 

[rvnguy]

Frank4d,

I am sorry that you are not receiving many/none responses to your query. You must know that you are out on the edge in the computing arena with this setup.

I have to ask; As you have stated that your dilema seems to be solved when you load the correct VxWorks file from a specified dir. 100% of the time cabling not with standing, do you still have a problem? and if so what is it?

It would appear to the lay(me) person that you have solved your problem


rvnguy
"I know everything..I just can't remember it all

 

[Frank4d]

While using the correct files for the two CPUs does solve the problem 100% of the time, I am still curious why changing the cabling seems to have ay least a partial effect. When I get to work I'll try some network analyzer tools.

 

[kmcferrin]

You have more than one problem. One of them is definitely the cabling. You MIGHT be able to split off the pairs from an 8-wire cable like you described if you are crimping them to the correct pins on the RJ45 connector. However, you need to do so at both ends. If your ethernet cable looks like a Y-cable then you will definitely run into problems, as your switch doesn't know that your cable has been split and is looking on pins 1 and 2 for transmitting and pins 3 and 6 receiving data (as opposed to two sets of pins). Your blue and brown pairs are coming out pins that aren't be used.

You're better off to just make a pair of standard ethernet cables than to try to "cheat" like you did. The only thing that you could be saving by using your method is a few inches of cable, but it's costing you a lot more headache.

 

[Frank4d]

The 1&2 and 3&6 pin wiring isn't a problem... that part is per spec. But I an stuck trying to adapt equipment to my employer's design requirements. On the PC end I have a 10/100 switch that has RJ-45 connectors. The other end of the five foot CAT 5E cables is connected through a panel that has connector pins about 0.050 square... then through another two feet of CAT 5E cable through a panel that has 0.020 round pins (military D38999 circular connector)... then through another 12 feet of CAT 5E cable and another set of 0.020 round pins (military D38999 circular connectors again)... to the target.

The wiring is correct from point to point, but is definitely not per the EIA-568 wiring standard and I suspect all those non-standard connectors are creating impedance bumps that are causing the problem.