Hello, and question about profibus rings

[strantor]

Hello Everybody, My first time here. I thought maybe one you of knowledgeable people could explain a question I have about profibus rings. We have a machine here that sends a profibus signal through several sets of sliprings and periodically has problems with communication. I suggested that we run another profibus wire from the end of the branch back to the PLC to create a redundant ring. My coworker replied that "it doesn't work that way" but couldn't explain further. I looked up on the internet about profibus and I ran across people saying that it is only possible via fiber and OLMs. Why? I don't get it! refer to my drawing below: [not drawn in picture: sliprings between PLC and I/O modules(1 set), and (2 sets) between every I/O module.] [we could install the long jumper pictured at bottom, only going only through the slipring after the PLC. I think the new jumper would be more reliable than the one currently going through all the sliprings] Thanks everybody!

[paulengr]

First off, fiber or Ethernet "rings" are not physically wired as rings. The issue of whether the medium is fiber or wire doesn't really matter. The ring concept is used when you have active devices (repeaters/switches/routers) within the network and everything is "daisy chained" together except that the last node is then connected to the first node. At the physical cabling level, there really isn't a ring at all. It's just a whole series of smaller networks with just 2 nodes on each network. But at the logical/software side of things, we can pretend that there is one big ring because we can make packets bounce from one node to the next until they reach the correct destination. If the nodes on the network are programmed to handle it, they might have multiple paths (as in a ring) and can navigate to a backup path if the primary one fails. The ultimate here is something like BGMP or RSTP; both of these protocols are designed to find a route in an arbitrary topology with almost no restrictions on the number of parallel/alternate paths and link qualities. But that's not what you have... Profibus is a type of twisted pair style communication. Almost all of these are at least loosely based on the RS-485 standard, no matter what the final specification is. The way that it works is that everything on the system is physically wired in parallel. Receivers maintain a high impedance connection to the cable. Transmitters are also high impedance when they are off. When they turn on, they put a voltage on the cable that is either positive (line 1 > line 2) or negative (line 1 < line 2) and the protocols are usually designed to be "balanced" (average voltage is nearly zero). Receivers don't care about the absolute voltages, only the direction (line 1 > line 2?). The key feature here is to maintain connections where needed, keep noise off the system (since it is low voltage and low frequency...no notch unlike Ethernet), and maintain the system impedance (resistance, capacitance, and inductance) within specification. Although you have drawn everything as if it is all connected by perfect wires (R=0, L=0, C=0), in the real world, Profibus wiring is NOT this way. It is really a transmission line. This means that since eddy currents exist in your wire, it has inductance. Since there is some space (however small) between the signal and return lines, you have a capacitor. And of course copper is a good but not perfect conductor of electricity. Here is much more detailed information on transmission lines: http://en.wikipedia.org/wiki/Transmission_line One of the interesting things about transmission lines is that we can simplify some of the details here. They all have a characteristic impedance (called Z0), and as long as we stay within the signal bandwidth that the line is capable of carrying (multimode or "0 order"), the signals will come out cleanly at the other end without worrying about the R's, L's, and C's. Whenever we tie two transmission lines together, or if we terminate (end) the transmission line, it is very important to maintain the same characteristic impedance (Z0 is the same). If there is a section of the line which has a different impedance (the R/L/C ratios are significantly different), the signal does not pass smoothly from one cable section to the next. Instead, some signal is passed on but some of it "reflects" and then starts travelling backwards down the cable in the opposite direction! So when the "forward" and "backward" signals mix at the receiver nodes, the output often becomes completely garbled and nothing works. There are some fuzzy details here dealing with how the C's and L's actually work in the real world, but even though this idea of a "reflection" is a bit of a simplification of how things really work, it very closely resembles how the real world works. In fact to find problems in long lines, you can use a tool called a TDR (time domain reflectometer) to actually measure the reflections and find cable faults over very large distances on transmission lines. As an example, at the end point of the cable, you effectively have two "waves" travelling down the cable on each line. They are almost identical in amplitude (voltage) but opposite in polarity. Placing a resistor across the lines allows both waves to enter and cancel each other out. Even if there's a reflection, it meets an identical reflection on the opposite line as it travels around through the resistor. If the resistor is not the right value, then the "forward" and "backward" portions are of different amplitudes. The really nasty part about all of this is that the problem isn't just as the end of the cable. As the two waves see each other on the cable, they add in different relative amounts, causing the receiver to see either amplification or attenuation of the intended (forward) signal. Worse yet, this signal can then travel down to the other end of the cable and back again, causing another pattern. As it oscillates back and forth, you get what is called "ringing" which looks like multiple ripples of the same signal repeating over and over again. With the right resistor however the forward and backward waveforms exactly cancel out, and the ringing or other interference effects are 100% cancelled out. Lesson learned: use the specified cable or equivalent, and don't forget the resistors when required. 14 gauge house wiring is NOT going to work as communication cable, nor is whatever shielded or unshielded stuff you have in your shop. Splices are important, too. In fact when you get to Ethernet levels (still a transmission line but bandwidth is now 100-200 MHz!), you may as well swear off splicing altogether. Proper termination procedures are absolutely critical (no stripping to an arbitrary length and just getting "close enough") since those little RJ-45's believe it or not are rated to 250 MHz. At roughly 6 MHz for standard video or 12 MHz for "HD" video, imagine how many radio/TV signals can theoretically fit on a single pair of CAT 5 cable, and why even something as innocent as pinching the cable in a door can completely disrupt communications. The same kinds of issues occur at slip ring interfaces because the impedance of the sliding joint is different from the cable impedance (the R's, L's, and C's are different), AND this value frequently changes as the slip rings mechanically alter the electrical characteristics. I'm not even aware of anyone building slip rings that are specifically "rated" for use in Profibus or any serial bus for that matter. What happens in reality is that usually they are "close enough" and if you keep the baud rate slow enough, you can usually manage to squeeze a signal across a slip ring even though it is not really meant for it, and that's what you have now. Now in your particular case, jumpering the ends specifically halves the characteristic impedance of the whole system. This will cause problems with communications at ALL nodes because the tendency will be for half of the signal power to get reflected back to the transmitter at the node, cutting your signal power in half across the whole system. You doubled your connections at the slip rings so you might indeed lower the amount of signal degradation at the slip rings but you are going to cause more problems to the overall system. On top of that, you've just created two very large loop antennas. Loop antennas are particularly sensitive to magnetic fields, and industrial plants are usually full of large sources of magnetic fields (motors). Even if you can overcome the impedance mismatch problem, you are also going to create a huge potential for interference from nearby sources of noise, regardless of whether the cable is shielded or not (shielding works against E-fields, not so much against M-fields). As a much better solution since the impedance of the slip rings doesn't match the Profibus spec anyways, you might try connecting two slip rings in parallel on the theory that the odds of both slip rings having signal losses simultaneously is less than a single slip ring. It may make it worse or again, it might make it better. If you are willing to walk away from slip rings altogether... I've gotten rid of slip rings by switching to wireless. Most of the time, if you are running on 24 VDC power, you can still use slip rings even if there's some "chatter" because the capacitors in a DC switching power supply allow you to "ride through" some of this. If you are particularly paranoid, Allen Bradley even sells a "DC UPS", but I've never used it. Just put the DC power on the moving part and use AC across the coupling (let the power supply deal with the chatter). Be sure to add appropriate filtering and surge suppression because it's still a noisy system. Then, transmit your signalling wirelessly. There are basically two approaches here. The most reliable one is inductive. This has a limited range (2-3 cm max.) but can transmit both power and communications. There are multiple vendors for this technology. Here is one: http://www.pepperl-fuchs.com/cps/rde/xchg/global/hs.xsl/2294_induktives__bertragungssystem_wis.htm?rdeCOQ=SID-058CFC44-9975AD86 The second option is outright "wireless" as in radio communication. Works well when it works. Works poorly when it doesn't. I am kind of partial to the stuff that Banner Engineering sells. It's relatively inexpensive, fairly rugged, and if you get the DX70 version, it's technician friendly (no "setup" other than setting a couple dials on the front). Of course if you are already using Ethernet (Profinet), then this is all moot and there are dozens of vendors. The third option is a variant on the second. If you can maintain line-of-sight, then you can use a laser transceiver. Again, if you have a geometry which works with this (linear motion, not rotary, unless you can locate it right at the axis of rotation) they tend to be more reliable than wireless. If not, then it's not really an option.

[strantor]

paulengr, That is without a doubt the most informative and well written response that I have ever recieved in a forum. Thank you very much for your insight and your time. The WIS looks very intriguing and right off hand I can think of about 100 different things I could use that for, but unfortunately this isn't one of them. It's difficult to explain the layout of the machine, but all those I/O modules are on "mostly stationary" gimbles inside a large spinning tube. Plus there are drives on the gimbles (I didn't draw them in) which are also part of the network and they need bona fide network communication. laser is right out; can't shine through the spinning tube. I will look further into wireless. I brought that up casually before right after I started working here in a discussion about the issues with this machine but it was quickly shot down with "there's too much intereference here for that". Maybe if I do some research and present a logical case, they might let me try it out on 1 gimble first and see how that goes. I have seen bluetooth used in machinery before and it worked great (almost all the time) so I might start there and see what other (better) wireless options are available. All your information about impedance and transmission lines and such as pointed out to me that I don't know as much about my job as I probably should, and I am going to see if they can send me to class on it. Thanks again, Charlie

[paulengr]

Wireless "interference" is a funny thing. Often if you have a lot of RF, BROADBAND interference is the issue. A great example is a foundry...molten iron has it's electrons moving around it more or less in a highly energized "cloud" which makes it virtually radio-opaque. HOWEVER, if you use spread spectrum radios and stick to the low speed ones they usually penetrate right through even that, even over a decent amount of distance. The DX series radios I mentioned do this. So do the Aerocomm radios (www.mouser.com), and I'm sure there are several others. I strongly recommend 900 MHz gear for industrial applications especially if you don't need the speed. Why? Penetration. First, calculate the wavelength. See this site for instance: http://www.csgnetwork.com/freqwavelengthcalc.html Figure that an object will cause significant diffraction/interference at 10% of a wavelength, and it becomes pretty much opaque when it equals the wavelength. So for 2.4 GHz, the wavelength is 4.92". So 1/2" of steel plate is all that it takes to start impeding a radio signal, and by the time you get to the typical thickness of a structural column or concrete, it's becoming radio opaque. At 900 MHz, the wavelength increases to 13". So the signal starts to have attenuation at around 1.3" of steel plate...essentially only very solid concrete bulkheads are completely impervious to 900 MHz radio signals. So in a typical manufacturing plant, 2.4 GHz right from the start is likely to run into all sorts of reliability issues. Second, spread spectrum radios are not the total answer. For industrial purposes, you don't need "56 Mbps" for control purposes in many cases. That's why many control system protocols rarely reach more than 50-100 kbps. Going slower solves a lot of problems because the signal (whether spread spectrum or not) is concentrated and less likely to have interference. 802.11 equipment was specifically designed for office/residential use, not industrial plants or outdoors. It can be used in those applications but it was not designed or optimized for that use. Hence the reason that it usually results in poor performance. So again, those "slow" serial radios tend to work much better than radios that are optimized for internet connections on office networks.