Issues with PXI-8335

[Ed S]

We are using PXI-8335s as PCI bus extenders connecting a primary CompactPCI backplane to two secondary CompactPCI backplanes, each located 300 feet from the primary. We currently have 15 such systems running in the field, and, for the most part, they run for extended periods of time with no problem whatsoever. There are, however, two problems that will occasionally pop up:

1) A link will refuse to initialize on system power-up causing the system to hang. Power and Link indicators will be illuminated on the PXI-8335s at either end of the link; the TX indicator will be lit at the primary end and the RX indicator lit at the secondary end. Power cycling of at least the primary chassis and frequently both primary and secondary chassis is necessary (so
metimes repeatedly) to restore operation. We use timing relays to ensure that the primary chassis is not powered until the secondary chassis have been powered for a fixed number of seconds. Is there either a specified or recommended delay in the power sequencing of the primary vs. secondary chassis? Is there some other means of ensuring that the links initialize reliably?

2) The vast majority of links, once established, run forever. There are a small number of links that will fail at intervals of tens of hours. Admittedly not bad overall, but we'd really like to get all of the links running reliably forever. The two leading suspects here are the quality of power in the failing chassis and the quality of the interconnecting fiber optic cable. While buying a prefabricated cable assembly would go a long way to addressing the cable as a source of problems, that option is not open to us; the cables must be pulled and terminated in place. What testing criteria (e.g. maximum end-to
-end attenuation @ 850 nm) do we need to specify to the people installing the cable to ensure that the PXI-8335s at either end communicate reliably?

Thanks in advance to any and all who can help shed light on these questions.

[Patrick R.]

Greetings,

Regarding the first issue, as long as the power to the secondary chassis is applied before the power to the primary chassis, there should not be any issues with the initialization of the system. The PCI bus master of the computer should see it's own PCI slots and the PXI slots as if they are one after the link is established. This occurs during power-up of the system. Therefore, I don't think varying the delay time would change the performance that you are seeing. One thing that you might want to consider though: it may be beneficial to check the revision of your MXI devices and send them back to NI for upgrade if they are not the most recent. You should be able to find the revision letter as a part of the Part Number;(186943x-02 (PCI) and 186897
x-02 (PXI) where x is the revision). You may choose to request support from the web at ni.com/support and contact NI by phone or e-mail.

Although there are not any cable attenuation recommendations for fiber, it does appear that there may be an issue with the cabling. Do you know whether or not the system functions without fail if it is placed within a few meters with a different, prefabricated cable? Also, you mentioned the quality of power in the failing chassis. Do you feel that the source of power is dirty or are you sharing the same power as devices that require a large amount of power? These two issues could potentially be eliminated by running an isolated chassis at a closer distance.

Have a good day,

Patrick R.

[Ed S]

Patrick R,

Thank you for your response. I have additional information based on your comments.

Regarding the issue of failure to establish a link at power up, experimentation on our development/test system has revealed another related anomaly that, unlike the problem described in my first post, is easily repeatable. I will try to follow this up through the support channels that you suggested. I will also enquire with them about the revision level of our boards (the part number given on our PXI-8335 boards is 184583G-02).

The additional experimentation on our test system also suggests that our second issue (failures in established links) is not a power supply issue and only indirectly a cable issue. I can produce failures with relatively short, prefabri
cated cables; these failures follow particular boards as they are swapped throughout the system. The results that I am seeing make me suspect that at least certain pairs of PXI-8335 boards are intolerant of losses of more than 2-3 db from transmitter to receiver. At the moment, this seems to be the most promising direction for me to investigate.


Regards,

Ed S