MOV Modbus Design

I work in a refinery where we have two cokers with 30 MOVs in each coker. We communicate to and from these MOVs using RS485 2 wire + shield. This design has failed and we believe it is due to the overall length of the cable (From I/O room to the MOVs) in addition to not having the third conductor for signal grounding. I am working on a project to redesign this system to something more reliable.

Current Design:
There are 7 segments in each coker with four to five daisy chained MOVs in each segment. The first MOV in the segment is wired directly to the I/O room into an 8-port MOXA then to Delta V serial cards.

New Design:
We are stripping the MOVs 2 wire cable and replacing it with a 3 wire + shield cable, keeping the daisy chain design. Instead of routing the cable to the I/O control room, we are taking each segment to MOXAs MGATEs MB3170I-T (Modbus-Ethernet). These MGATEs will then be wired to cisco switch IE2000 (Ethernet-Fiber) via a CAT 6 ethernet cable. This Cisco switch is then wired to a fiber patch panel which relays are the signals to the I/O control room via fiber. We are hoping this helps with the length concerns we have.

I am curious if anyone has ever come across similar issues or have any advice to our project team. We are a few months away from implementing this design.
I assume you're not running at baud rate that exceeds the RS-485 spec. What are segment lengths and baud rates?

For what it's worth, Profibus DP is an engineered RS-485 network. I find Profi's DP segment cable length maximums more conservative than the standard RS-485 baud-vs-cable length values. If a vendor's engineered 485 network is more conservative than the general 485 spec (and by all reports very reliable when installed correctly) then it might pay to consider not pressing the spec's upper limit on distances at a given baud rate for a as-you-go network.

Converting to Ethernet/F.O. for part of the transmission link will provide isolation and some noise immunity, but you're still left with RS-485 on each segment and some latency at the conversion point.

Are you aware of the great app note, Ten Ways to Bulletproof RS-485 Interfaces? It's at this link:

Questions on the current setup

1. What does the 8 port Moxa do now? If Delta V has serial cards and the MOV's are serial, why the intermediate Moxa?

2. You didn't say what the failure mode is. Why do the networks not operate correctly? What type(s) of faults do you encounter?

3. Do the MOV's have a third terminal for 485 signal ground connection? Or is the proposed 3rd wire grounding to chassis/earth ground?

4. Are the daisy chained serial line field connection drops wired direct on the MOV with incoming and outgoing serial lines at the same point, or is the drop connection in a junction box with a stub line running to the MOV?

5. Any chance you're using a multipaircable with multiple twisted pairs some of which are not used? I've heard of 'resonance' in unused TP's that causes noise resulting in bad msgs that can be eliminated by terminating the unused pairs with a terminating resistor.

6. Does the master serial port provide any biasing for reliable tristating?

For what it's worth, I have solved several 485 multidrop noise/common mode issues with field devices that have no 3rd wire signal ground by putting an RS-485 repeater/isolator module at each device. Isolation is worth its weight in gold.