Fieldbus grounding and noise spikes [Archive]

IanVerhappen

October 3rd, 2003, 10:23 AM

Eric Byres wrote about industrial Ethernet:
> - multidrop: We are out of luck here...
The Beckhoff EtherCAT solution is a nice step away from the star configuration to the multi-drop. But it actually daisy-chaining, with the disadvantage that a node may not be removed or switched off as this disconnects the part of the network behind the node.
The ProfiNet V3 solution, with its built-in 4 port switch on every node, is similar. But this won't be on the market soon.
If anyone would invent a 3-port switch small enough to be contained inside a T-connector, we'd have something very similar to multi-drop wiring...

Rob Hulsebos
Eric:

Thanks for the GREAT posting!

The reference to the Rockwell paper was very informative.

On the use of Ethernet to the factory floor devices - I agree, 6 out of eight is a VERY good starting position. I hope the control vendors have lots of secret development work going on with this technology. I also agree it may be several years out, especially when considering the rate at which FF H1/ ProfiBus PA developed.

Regards,

Jim Reizner
Corporate Engineering
Procter & Gamble

PS: It may be uncommon, but our one noise experience with H1 was absolutely introduced into a properly shielded H1 cable. We have also had these problems with DeviceNet, in fact we have seen this in our labs with 4 to 20 when we put 480 volt cable right next to it (definitely NOT recommended). It just goes to show that anything can be broken if one tries hard enough, and our real-world industrial installations always have some unusual condition that causes "never before seen" problems. With FF H1 or any other signalling system, be smart and don't run it along with or next to 480 VAC cables.
----
Eric, thanks for your summary. Here are some additional items.
In 1999 I spoke with MTL about intrinsic safety barriers for Ethernet. The comment from their technical director was that MTL saw no technical barrier, but was only waiting for demand. As you know, MTL participated in the development of Foundation Fieldbus H1 and was instrumental in including IS in its design. They elected to not be part of the HSE design, but wait in the wings to see if HSE moved to the field. Maybe this is a chicken-and-egg problem, but I agree that the process industries will be slow to move HSE to the field even if IS barriers and HSE field devices (there are none) are offered.
For those who wonder why IS on Ethernet is not available now, there is a technical issue: zener diodes are the heart of IS barriers, but are frequency-sensitive. At Ethernet frequencies, they shunt the signal to ground. Instead, different and slightly more costly components must be used.
In 2000 I wanted to file a patent for multidropping 10/100BaseT Ethernet, but my research found that the basic technology was then being sold by Jetter AG as their JetWeb network technology, which itself is patented.
JetWeb is really a daisy-chain rather than multidrop, but the effect is the same. Every node is an Ethernet layer 2 switch. Messages not intended for the local node devices are shipped out the "uplink" port of the switch.
Sounds slow, but since switches operate at line speed it is not. You also get another switch advantage -- regeneration of the signal and another 100m distance. BTW, Broadcom sells a switch chip to do this for less than US$8.00. Cool stuff.
Dick Caro
----
Hi Ian
Dick and Jim are correct on a number of their comments regarding Ethernet on the plant floor. Dick points out that VFDs have little effect on 100Base-T wiring - most of this research was carried out by Bob Lounsbury at Rockwell and has been written up in the following paper:
http://www.ab.com/networks/enetpaper.html
The really interesting thing is that most noise problems are induced not the wiring, but in poorly shielded or grounded electronics. In fact I have seen a number of problems in fiber optic systems which should have no EMI issues in theory, but can deploy such poor electronics they are worse than UTP.
Jim points out that industrial systems need to include the following characteristics. The IEEE or industry has addressed a number of these already and I have noted references below:
- availability for loop power on two wires: See the IEEE 802.3af specification
- intrinsically safe: Still a issue according to what I've heard, but technically possible. I believe MTL is about to release some Ethernet Isolators.
- deterministic transmissions: Switches, 802.1p &Q and IEEE 1588 have dealt with this issue
- multidrop: We are out of luck here...
- be able to run a longer distance than 100 meters: We are out of luck here too, unless we go to fiber.
- be relatively noise-immune in industrial applications: See above
- low cost : Shouldn't be a problem...
- use appropriate industrial-grade connectors: Lots of those out there now. Check out ODVA or EIA/TIA TR49.7 for where this is going.
So we are batting 6 out of 8, good enough to start to see Ethernet appear in the plant floor devices. So far most of that action is in the automotive sector where Ethernet robots and welding devices are becoming Ethernet-based. I think process industries are still a few years away from that.
Regards
Eric Byres, P.Eng.

----
Time for me to interject and ADD to Dick's comment below. Industry has for years used VFD to mean Variable Frequency Drive, and VSD to mean Variable Speed Drive (there is a difference). Unfortunately, this is one of those cases where a TLA (Three Letter Acronym) can have multiple meanings. The result is that we must use our noggins and put the term in context of the sentence around it or expand it if there is doubt. Another example that comes to mind is that AI can mean Artificial Intelligence as well as Analog Input.
Ian
BTW, a Variable Speed Drive is one in which an actuator is used to vary the effective gear ratio on a pump box between the motor and the pump while a VFD, actually alters the speed of the motor. The net effect to the pump is the same but the mechanics and actual equipment are significantly different so we cannot use Dick's VSD suggestion for varying speed pump drives.
-----Original Message-----
The "AC/DC Drives" industry has taken to name their AC motor drives as VFDs dating back to the original Reliance patent. Personally, I prefer VSD (Variable Speed Drive) that encompasses both AC and DC drives. We cannot change that any more than the Programmable Controller suppliers could claim the PC as their acronym after the introduction of the IBM PC in 1981. Since no company actually manufactures a Virtual Field Device, I am afraid all references to VFD will mean Variable Frequency Drive and the other must be VirtualFD or something. Anyway, for us process control types, let's all refer to these electronic frequency converters as VSD.
Dick Caro
---
We have the Relcom FBT3, and our experience with it is that is is so basic that it is only of value for the most simple of problems. We have had many problems where the FBT3 indicated no problem, but the device did not work (device needed to be sent back to the vendor for firmware upgrades, etc., etc.). In finding the source of electrical noise - well, the FBT3 is a less than optimal tool here. We used a oscilloscope on the line to look for noise problems, the National Instruments Fieldbus Simulator helped for some of the more difficult problems (but we found the National tool less than easy to learn). I know there are newer and better H1 tools available now. Emerson has a very nice document "Fieldbus Segment Checkout Procedure for Non-I.S. Applications" that discusses the use of an oscilloscope to check for proper power, grounding, and isolation - I do not know if Emerson considers this article propr! ietary. When we were experiencing our noise problems we used the FBT3 to no avail, and then used the Emerson document with a Fluke oscilloscope to check everything.

We did not have a noisy component in the H1 segment. The noise was being generated by the stairstep waveform of the VFD, transmitted down the standard non-shielded VFD cable, and coupled into the properly "everything" H1 cable. This was very easy to demonstrate, as moving the VFD and H1 segment cables a few feet apart made the problem go away. We replaced the non-shielded VFD cable with a shielded purpose-made VFD cable (Belden), and we had no problems even when the two cables were in close physical proximity.

Jim Reizner
----
First, thanks for sharing the problem with the group and opening up this thread.
Not to get defensive about your not being defensive, but the stated problem is and excellent example of the power this newsgroup can bring to the table. I am delighted to see all levels of responses (most simple to most complex) as that builds upon my total knowledge/skills base. It strengthens us all. I hope members do not take the comments to mean "don't offer your 'simple' solutions and continue to work through this as they have been doing. I have been teaching (and physically working in) I&C technology for years and even remember the few times that, regardless of my vaulted years of experience and expertise, I came across a simple switch I overlooked several times during troubleshooting. (The real dilemma I faced then was how to convince the operations staff that the problem was so deep, I could not begin to explain it, so they should just be grateful that I found it and all the hours they were shut down were unavoidable).
Of course I, like most, understand the cause may not be simple and will not be surprised if it takes someone on an intellectual parallel with Dick Caro to resolve it. I just think that it never hurts to go back to the basics, especially if more than one person is involved with the physical installation or testing. In the meantime, one move might be to get someone else to replicate the setup, if it is not too expensive, and see if the results are the same.
By the way, can we come up with another rubric for Variable Frequency Drives or change the one used for Virtual Field Device? I see some confusion in the future being generated by abbreviating those two items. Nothing major, just a thought.
Chuck Carter
----
We put the terminators in carbon steel NEMA12 grounded enclosures at "controller," and not out at a noisy end point in the field. He have read and followed the guidelines from the FF, from our DCS vendor, and most helpful, from Relcom (Relcom Fieldbus Guide and the Relcom Fieldbus Topology). Not to get defensive here (OK, maybe a little), but we are senior electrical engineers (OK, maybe we are just old) with dozens of years experience in Corporate Engineering in one of the larger companies in the world. If, with our best efforts, we have not been able to properly install the H1 segment in our research lab then there is something wrong with the complexity of this task.

Dick:

As I noted previously, we have problems in the 300 KHz range with electrical noise radiated from servo drives. For a real simple test, take a simple AM radio and place it near some of your VFD / servo cables and see if you "hear" and EMF. A great paper on the subject of this can be found on the Belden web site, "Cable Alternatives for PMW AC Drive Applications, Paper No. PCIC -- 98."

Dick - I really like your idea of the major vendors looking at Ethernet for connection directly to the field device. I assume that some of the vendors are researching this, and that therefore none will respond to your question. H1 has some significant limitations based on the age of the technology and speed of the transmission. Ethernet can be essentially as fast as anyone wants.

I used to think that there was no possibility for Ethernet to ever be a system used to go directly to the field devices. From my perspective, field device signalling systems need to include the following characteristics:

- availability for loop power on two wires
- intrinsically safe
- deterministic transmissions
- multidrop
- be able to run a longer distance than 100 meters
- be relatively noise-immune in industrial applications
- low cost
- use appropriate industrial-grade connectors

I see that IEEE and the suppliers are chipping away at these requirements with revised Ethernet specifications and equipment. I think Ethernet (perhaps heavily modified from its present form) may just be the next system for field signalling interface for instruments. HSE, or an extension to it, may have a place here.

Regards,

Jim Reizner
Good discussion on VFD. I would check each and every physical layer component (Bricks, Megablock, Terminator and wiring) for noise. A simple network analyser such as FBT3 reading will pin point the noisy component which needs evaluation.
Sudhir Jain – Emerson Process Management
-----
Not to get too basic, but I assume the end terminator is also located in a relatively noise-free area such as the IJB. I have run across a few segments where someone thought the terminator had to be at the furthest point out, not at the end of the trunk.
Chuck Carter
-----
Jim, thanks for the re-assurance that at least some people know how to install instrumentation and properly ground shields.
My comments about 100BaseT Ethernet (used with HSE) being theoretically immune to VFD harmonics is not based on any experience I know about. If I were designing a de-modulator for 100BaseT using Cat5 or better cable, it would certainly use a simple high pass filter in order to discard low frequency noise. Since VFD harmonics typically reach only to a few tens of kHz, they would be filtered out of the 100 MHz baseband signal used for HSE. You just can't do that with H1 at its 31,250 Hz Manchester encoded signal.
There is nothing that would prevent FF HSE from being used in place of H1 except for very long connections (above 100m) and in those areas needing intrinsic safety, since an IS barrier for Ethernet has yet to be developed. The most significant inhibitor is that process control field instrumentation has not been developed for HSE. I think HSE would lower costs of the Fieldbus interface and the cabling. Current Foundation Fieldbus specifications would not need to be changed since HSE supports all of the H1 functionality.
I would love to hear a response from the instrument suppliers.
Dick Caro
-----
Dick:

Thanks for your message!

Yes, the H1 segment cable shield was properly grounded at one and only one end.
We tested the ground using a ground loop impedance tester (GLIT) and an AEMC 3711 clamp-on ground resistance tester, and found the ground to meet the requirements of the National Electrical Code (2002 NEC, Article 250 - Grounding). I don't remember the exact values, but it was less than 5 Ohms.
The issue of induced noise from VFD and servo cables has been long known and well-documented in industry. Some of the significant harmonics from these systems reach into the KHz range. We routinely have problems with metal detectors that are tuned to approximately 300 KHz from electrical noise induced by servo drive systems.
Dick, you note the knowledge that VFD harmonics have little or no effect on 100BaseT signals, at least theoretically. If you have the time and energy, reference to papers on this might be of interest to the FUN audience.
I wonder if anyone has real-world experience with either Ethernet or the HSE-flavor and servo / VFD's or other sources of electrical noise?
Regards,
Jim Reizner
Though we are not an end-user of FF devices, we have conducted a large amount of EMC testing on the FF devices design and manufactured at our facility. Many of these tests include FF devices within our company and from other vendors. During some of our initial testing, we used power supply conditioners and terminators that had built-in surge protection.
When conducting the Burst Transient Test (EN61000-4-4), we observed all devices on the segment under test go off-line. The power supply conditioner had an LED to indicate when the surge protection was being employed. When we monitor this LED and the supply voltage on the segment, we discovered that the surge protection was being enabled during the Burst event and was dropping the supply voltage below the 9 Vdc minimum needed to support the devices. In an other situation, we observed problems with the quick disconnect connectors used on in many fieldbus applications. For our situation, the connectors (Tee connectors) did maintain the shielding through the connector; however, the shield was brought through as a single conductor and the connector body was completely unshielded. During Radiated RF testing (EN61000-4-3/ENV50140), we had problems with the devices falling off the bus until the connectors were removed. Once the connectors were removed, we had zero failures.
Hopefully this information can help you with your problem.
Best Regards,
Stewart Thoeni
----
It is always great to have real experience documented. Thanks Jim!!! Before completely accepting your history, I would like to be sure that the shield on the H1 cable was properly grounded at one end only, and the quality of the ground point had been tested and found to be essentially at earth potential.
VFDs generate high frequency harmonics due to the pulse width modulation method they use to simulate an analog variable frequency AC power waveform. The envelope of the waveform looks to an induction motor as though it is AC, but the signal is really a high frequency variable amplitude waveform. We already know that the AC motor must be built to tolerate the additional heat energy generated within the windings due to their inductance to this high frequency component. Now, it appears that we must also be aware of the induced noise and its effect on H1 Fieldbus. The harmonic component of VFDs is known to be in the low kHz band, which is unfortunately exactly where H1 is. Analog 4-20 mA devices naturally filter out this high frequency noise signal.
I guess that this experience will go down in my book as a positive benefit for using Foundation Fieldbus HSE directly to field devices when intrinsic safety is not an issue. We already know that VFD harmonics have little or no effect (theoretically) on 100BaseT Ethernet signals.
Dick Caro
----
A perspective on the relative noise immunity of H1, qualitatively - not qualitatively.

Before our first H1 installation, we had concerns about the electrical noise immunity of the H1 system because of the digital nature of its signals. We were assured by the experts who had started up many systems that H1 was very noise immune, and that they had never experienced any noise problems with it.

At start-up, we experienced problems in one of the segments (sorry, I don't recall the specific problem other than to remember that it prevented that segment from being useful).

The H1 segment was properly grounded, shielded, etc. We traced down the problem to be electrical noise induced into a H1 segment from a 480 VAC three-phase VFD (variable frequency drive) motor cable. The VFD and the H1 cables were NOT run in conduit, and were within about three feet or so of one another for approximately a five foot run. We could easily confirm the problem, as simply moving the VFD cable farther away from the H1 cable made the problem go away. Because of the layout, moving the VFD cable was not an acceptable long-term solution, so we solved the problem by replacing the standard SO cord VFD cable with a purpose-made shielded VFD cable.

I know that a conventional 4 to 20 mA DC signal would not have had a problem with this level of electrical noise.

My point is not to incite H1, but rather to say that it IS a digital signal and that standard precautions that are in order for H1 as with any digital electrical signalling technology. H1 is certainly more noise-immune than many higher-speed systems, but even with H1 there comes a point where it cries "uncle."

Jim Reizner
Corporate Engineering
The Procter and Gamble Company
Hi Carl,
Interesting noise problem you have - I assume that the 2000 MV is really 2000 mV as 2000 MV would destroy the instrument electronics.
One question I have "how did you measure it". If the spike is 2 V (2000 mV) & common mode then because FF is a balanced feed then it should not be affected by as little as a
2 V spike. Another point is - maybe the spike is way larger than the 2 V you see & is beyond the CMRR of FF. If the spike is indeed 2 V normal/differential mode (DM) it would cause a "comms error" especially if a series of them occurred during a H1 packet. The question is how does a spike like this get into a balanced comms feed (DM). It either must be enormous or there is some thing wrong with your grounding and/or shield wiring. Are all the shields connected as per the FF recommendations - NO double grounding points etc etc.
As an overall solution with any transient problem - we always have 3 areas to consider :-
1) find the "source" & suppress it with proper techniques (always the BEST economic approach as there may be more than one "victim" in the plant)
2) minimize the path from "source" to "victim" with proper cabling layout, grounding & shielding techniques
3) suppress at the "victim" as a last resort with SPDs or some filtering approach.
Hope this is all helpful with your noise problems - maybe a more detailed layout of your wiring might help if more is needed.
Regards,
Allen Tighe
Carl,
I, with the help of some others, have done testing on the impact of noise on Foundation Fieldbus. One area that you may wish to look at is VFD drives. It appears that more companies are using tray cable. I strongly suggest that Teck cable be used for VFD drives (between the motors and the controllers). And ensure that the armour is properly grounded at both ends.
For our testing, the Foundation Fieldbus cable used was Type A with 4 pair individual shield with overall shield. During this testing, only one pair was used with that pair shield properly grounded at one end.
There was no difference even with connecting the three unused pairs and shields all together and grounding that at one end. Was also able to generate sufficient noise by having a "poor" ground for the teck armour.
To mimic field conditions, we had approx 50 meters of cable in various configurations.
In both cases, the noise seen was very disruptive on the Fieldbus communications. Even with an air separation of 12 inches (to simulate spacing in a cable tray) between the tray or poorly grounded tech and the Type A Foundation Fieldbus cable, the communications was disrupted. Using harmonic filters and toroid coils did not help.
Our testing also showed that there was no impact due to noise from zero-crossing controllers (e.g.: high pressure sodium lamps, electric heat tracing controllers)with the power running in tray cable and the Type A Foundation Fieldbus cable used.
All other testing did not produce any noise that caused problems.
I would also suggest checking the grounding. I am not sure what your field wiring looks like (field junction boxes ... surge suppressors ... local ground ... blocks used) but it may be possible that noise is entering in from ground in the field. With more details, I may be able to offer a few suggestions.
I have seen much documentation on the types of wire to use for Foundation Fieldbus, how to calculate the loading, lift-off voltage, details of the communications ... but nothing on practical wiring guidelines for refinery wiring (homeruns, spurs) and recommended wiring techniques to avoid the problem that you are having. Has anyone ever seen any publications of this nature?
Glenn
Obviously this guy is measuring mV (millivilots), not MV (Megavolts) ie 2V spike. Segregation, cable type, and compliant installation are critical - noted in the foundation fundamentals course I attended. Be interesting to know exactly how the 'spikes' are being measured.
Phil
-----Original Message-----
From: ISON Graham
Sent: Tuesday, 23 September 2003 08:53
I presume he means mV.
----
Carl,
I would like to know something about your fieldbus wiring system to learn how 2000 mV can be get on shielded twisted pair cable. From testing we have done, I know that to get that kind of differential noise on the wires, the surge on the shield has to be several kilovolts. Is your fieldbus power supply ground isolated?
Maris Graube
We don't monitor our segments continuously, but when we do we never see any significant noise spikes, even on segments where the Relcom tester flags as having a noise problem. Noise is typically high frequency (relative to the FF square wave) and of fairly uniform amplitude.
On occasion we do see a device go to "standby" for no apparent reason (maybe one out of 400 per year); I think in most cases it's been a four-wire device (e.g., coriolis meter) where its power has been cycled due to a power system glitch.
Our home run cables are in tray, and most of the spurs are in rigid galvanized conduit. We are in a relatively humid climate (Ohio) so we do get moisture in devices and junction boxes.
John D Rezabek
-----
Where capacitors are unable to clamp the spikes, it maybe useful to consider using tranzorbs or MOVs (metal oxide varistors) instead. they are faster in response time. also, depending on network, the issue maybe trying to actually limit the CMRR (common mode rejection ratio) and/or the TMRR (Transverse Mode Rejection Ration). this can technique can also be consider on the power supply rails as well (but needs to be sized accordingly).
Hope this helps
Aris