Devicenet 3Sec Node Drop Issue With New Master

Hi all,

I am new here and this is my first post. I have already enjoyed reading many of the threads and articles.

At my job I am facing a very challenging issue with Devicenet. We have had to implement a new architecture using the Hilscher CIFX90 Devicenet master which is a mPCIe card. Most of our network is controlled via PLC on EtherCAT, but due to having to support some older devices that require explicit messaging we need the Devicenet master.

There are 8 slaves on the Devicenet, with 2 gateways communicating back to PLC for IO poll. 3 of the slaves are PCVs. Now the problem being experienced is one of those PCVs will drop off the network for 3 seconds and then reconnect, flagging an error in software. The interval it occurs varies, but it is usually every 7-30 minutes. We have tried changing node addresses on these 3, and noticed that whichever one is highest, that is the one that drops. Normally they are Node 5, 6, and 7, and problem was first seen with 7. If Node 5 is changed to Node 8, then Node 8 experiences the drop-off.

We have been working with the manufacturer Hilscher to address this problem. Some background on these CIFX cards. Hilscher decided they were going to make 1 chip the netX100 for their product line across all the different control systems i.e. Devicenet, EtherCAT, EthernetIP, etc. just change some connectors and firmware and in theory it should work. The firmware has to be downloaded to the card using, and then it requires building the network of master + slaves, and .EDS files are needed, this project is then downloaded to the card. This is a huge departure from the card it is replacing as that would just auto-scan the network, and the driver api is used to communicate. This extra configuration for the Hilscher CIFX is not a deal breaker, but requires extra work.

Anyway we were able to obtain a Netmeter device to diagnose what is going on. The first thing we noticed was the bus traffic is always maxed out near 98.6%! I setup a bench with just 3 of those PCVs (pressure control valve) to try to reproduce the issue, but was not able to. Just with those 3 devices, network bus is maxed at 98.6%, with ~33% per Node respectively. For comparison the old Devicenet master on the same bench bus traffic averaged 4.1-4.9%. Our suspicion was this Hilscher CIFX was scanning the bus at a high speed, so we connected an O scope that could read CAN traffic. We were seeing packets in the 14ns microsecond territory. Another thing that was odd is the CIFX master was scanning, irrespective if our software was running or not. Just in Windows at idle, with nothing running, the master was generating max bus traffic. We have asked the manufacturer to look into this. They recommended changing a setting for production inhibit to slow down the individual node, and while this worked, and there were no drops, it rendered the PCV useless reading a bunch of varying incorrect values every half second.

Onto the machine these slaves are running on. Using the Netmeter we were able to determine that the Shield is not grounded properly. The network is terminated properly at both ends with 120ohm resistor, and the resistance reading from end of network was 59.2 and 61 at the master, so pretty close to 60, which is why I don't think it is termination related. When we get access to it again, we will audit the grounding on the shield and 24v supply for Devicenet. My working theory right now is that not having the shield properly grounded is causing noise to disrupt the signal and the high scan rate of the CIFX master is a contributing factor. Note: This issue is not seen when we use the old Devicenet master.

I was hoping there might be someone on these forums with knowledge of Devicenet and / or Hilscher products that could point me in the right direction or had any ideas for test to run that would provide any insight on what is going on. A challenge I face is the market for Devicenet master cards is a desert, with only really 1 vendor, Hilscher. I've thought about alternatives like EtherCAT slave to Devicenet master gateway like Anybus because it supports explicit messaging, but I don't think we have the software time to redevelop.

Anyway, thank you for reading this and any suggestions that may result.
There are situations where you appear to provide too much information - but clearly made every effort to provide full details.
Also I had to google legacy kit just to remind myself what DeviceNet was about.

From the above, you appear to have self-diagnosed earthing fault. This obviously needs to be addressed as soon as you are able.
The second point is you may be better having slaves connected to the same technology as they were designed; Allen-Bradley 1784-U2DN is relatively expensive but is of the same era:

Let us know how you get on.
Thanks Oneye, I think I am going to get one of those 1784-U2DN with the NetDecoder software and that way I can do a traffic capture.

The problem I face with Devicenet is there isn't much support for it anymore