Jump to Articles from:   Issue 45       Issue 44       Issue 43       Issue 42       Issue 41       Issue 40       Issue 39       Issue 38       Issue 37       Issue 36       Issue 35       Issue 34       Issue 33       Issue 32       Issue 31       Issue 30       Issue 29       Issue 28       Issue 27       Issue 26       Issue 25       Issue 24       Issue 23       Issue 22       Issue 21       Issue 20       Issue 19       Issue 18       Issue 17       Issue 16       Issue 15       Issue 14       Issue 13       Issue 12       Issue 11       Issue 10       Issue 9       Issue 8       Issue 7       Issue 6       Issue 5       Issue 4       Issue 3       Issue 2       Issue 1

Industrial control networks place a greater burden on designers of system components than IT networks. Industrial control often requires fast and guaranteed (deterministic) throughput. Networks may also require a greater level of noise resistance and environmental sealing. By presenting test data and real-world results, the authors show what is needed, and what is possible, in industrial environments.
By Bob Lounsbury

Industrial Noise

Industrial control devices, including network components, must be designed to withstand plant floor generated noise and noise immunity and emissions as defined by European Communities Directives and other local authorities. A direct correlation between these doesn't always exist. Products should be designed to withstand external noise and limit potentially disruptive emissions. Table 1 shows a list of EU EMC Directives and Table 2 is a list of noise-generating devices common to manufacturing environments.

Noise Coupling (Conducted)

Integrators, engineers and designers can realize substantial savings by combining communications, instrumentation and power conductors into a single raceway. However, this practice leaves the cabling highly vulnerable to coupled noise. The impact of combining conductors with high dv/dt transients and a communications cable carrying < 100mV signals can be enormous. The coupled voltage (common mode) of a 10 Vrms RF signal is measured in tenths of volts. Further, for high dv/dt noises, the coupling of a 2 kV common mode transient can be a magnitude greater that the RF signals (hundredths of volts). In a transformer-coupled system such as Ethernet 100Base-T, the common-mode voltage is not an issue. However, the voltage produced by poor common-mode performance can cause relatively large differential voltages, which can be disruptive to communications. It's worth noting that most commercial Ethernet media access transformers only have an isolation voltage of 1500 volts. The voltage is 500 volts short of the required 2kV needed to pass the IEC surge and electrical fast transient burst (EFT) tests.

Optimizing Throughput

Optimizing throughput for an Ethernet system involves reducing unnecessary traffic and minimizing collisions. In a hub-based system these are normally high. A switch minimizes traffic by setting up independent, virtual connections so devices only receive traffic destined for the switched, eliminating the collision domain. Full duplex is another method of optimizing throughput. The use of switches and full duplex communications will effectively eliminate collisions. However, errors are more difficult to address because their sources are varied. In an office environment, most errors are caused by impedance mismatches which cause reflections. Assuming that all of the reflections are in phase, the sum of the reflected voltages (due to only the +/- 15% impedance tolerance) could be 0.47 Volts. For example, a 10-meter jumper at 85 ohms connected at both ends with connector at 115 ohms could produce reflections at each impedance discontinuity. At ~40 MHz, the reflections will be in phase and add. Figure 1 shows the effects of impedance mismatch on the link error rate.

Common-Mode Noise

Common-mode noise is the voltage (common-mode voltage, or CMV) present on all conductors with respect to ground. Due to the number of sources in industrial environments, communications systems are highly susceptible to common- mode noise related problems.

The CMRR (Common-Mode Rejection Ratio) of an industrial cabling system is critical. CMRR is the ratio, expressed in dB, of common-mode noise rejected and prevented from converting to a differential mode voltage.

The trace in Figure 2 is a real-life example of what happens when a 10 V RMS RF signal is applied through a coupling clamp to a cable. The 0.0938V RMS signal is coupled differentially to the near-end of the cable. The CMRR of this cable is roughly 40.6 dB at 40 MHz. With a 13dB link budget and the noise applied at the far end of the cable (receiver) the signal to noise ratio is 8dB. These measurements were taken with an optimized cable having 40.6dB CMRR. Other cables in the same classification measure 17 dB.

For a shielded twisted pair cable (STP), the CMRR is 43.4 dB as RF currents are conducted on the shield and not the cable. There is some capacitive coupling, though much less. Note that the overall coupling is lower with the STP cable, and attenuation is about 6dB (Figure 3). This can be attributed to the fact that the noise ingress has a different path/mechanism. Evaluating the cable's CMRR yields the same results. Figure 4 through Figure 6 show compensated CMRRs. All three meet the specification, but the study shows that with careful optimization, un-shielded twisted pair cables can approach the performance of shielded designs. Further, standard off-the-shelf UTP cables will not provide as high CMR as the industrially optimized UTP cables. The CMR Ref plot represents the performance limit line for the test system used to measure the CMR of the various cables.

The tests show that CMRR performance of optimized un-shielded twisted pair cables can be as good or better than shielded twisted pairs. The noise coupling mechanism for STP is primarily due to currents in the shield that are capacitively coupled to the pairs. The noise coupling mechanism for UTP is capacitively coupled to the pairs from the test clamp where the CMRR performance is a function of the UN-balances in the link. The cable capacitance UN-balance is major contributor to the overall system UN-balances. When the same cables are subjected to common-mode signals and the capacitance UN-balance becomes a variable, the amount of differential noise measured is high. Figure 7 shows the CMRR (dB) versus capacitance UN-balance. The trend is upward (less CMRR) as the capacitance UN-Balance increases.

The capacitance UN-Balance affects the link performance in a non-noise environment. The bit error rate (BER) will increase as the capacitance UN-Balance in increased. Figure 8 shows the BER performance as a function of capacitance UN-Balance.

The increased BER is caused by the distortion in the MLT3 encoded signaling on the wire. The effect of noise will be greater for a poorly balanced system. That is to say that when common mode noise is applied to a differential system where the system is poorly balanced the amount of differential noise voltage will be greater for a given (fixed) amount of coupling. With 24 pf of capacitance added to the link the waveform is vastly distorted as shown in Figure 9.

Test data suggests that the key to noise immunity is a highly balanced differential system. Any device, cable, connector, patch panel or jumper that affects the capacitance UN-Balance will have negative effects on the system noise and BER performance. The trace and component layout of a network interface could easily have 10-pf to 20-pf differences in TX+/RX+ and TX-/RX-. Significant degradation to the signal quality could result if compounded by both TX and RX of two different devices connected by TP cables.

Table 3 is a comparison of noise measured in a motor drive controller application-to-noise coupling measured in EFT noise. Although this is only one comparison, other tests have shown that lab testing to EMC directives is usually more difficult to survive than most of the real-world applications. Both field and lab tests show that the higher performance cables (based on EIA/TIA category standards ratings) have greater common mode voltages coupled to the conductors. Cables that exhibit lower CMRRs will have a greater differential noise voltage as seen by the receiver.

A Real World Application

Ethernet cable tests were conducted on a robot used to spot weld sheet metal parts. Electro Magnetic Interference (EMI) is generated at the instance of the spot weld arc. In addition, large magnetic fields are generated in the secondary weld conductors. The motor drives for the robot arm and conveyors produce conducted noise currents and local ground currents.

Four-pair Ethernet cables were tightly strapped to the high-current secondary of the welding conductors and routed in close proximity to the weld tip to provide magnetic and radiated noises during welding. Figure 10 shows the routing of the cables. An analyzer was used to generate 100Mbit NRZ data that was sent to a pair of Fast Ethernet transceivers for encoding and decoding into MLT3 type data. The telecommunications analyzer provided bit-level error detection. Performance statistics were recorded for each of five cables. Surprisingly, zero data errors were recorded for all five.

The routing of the Ethernet cables, in both drives testing and the robotic application, was in violation of most published wiring practices, including the National Electric Codes in the USA. This was purposely done to simulate worst-case conditions, including a marginal cable installation.

The coupled voltages were greater for higher performance cables (based on EIA/TIA category standards ratings) than lower ones. The greater the cable bandwidth, the greater the noise power in the system. For industrial environments, limiting the cable bandwidth can enhance the signal-to-noise performance. If most of the signal energy for Ethernet is below 62.5 MHz, reducing the cable bandwidth to below 100 MHz can provide up to 10-dB increase in the signal-to-noise performance.

Connectors

Connectors play a vital role in the overall performance of Ethernet. Tests have shown that not all commercially available RJ45 connectors are suitable for industrial environments and their performance in vibration and repetitive shock tests vary considerably. It is important to note that fretting due to micro contact movement and oxidation will eventually cause contact failure and connector designs which minimize movement are preferable when vibration and shock are present.

Temperature

High-temperature environments may cause problems with a cable's electrical performance. As the temperature increases, so does the DC resistance of the copper, affecting the attenuation of the cable. Low temperatures can cause PVC jackets to fail.

Conclusions:

To provide near-deterministic performancenetwork traffic must be minimized. This requires optimizing the network using guidelines, selecting proper materials, reducing external interference, enhancing the infrastructure and minimizing its deficiencies. Equally important is the isolation of control networks through firewalls and security measures.

Some conditions negatively affect performance and increase error rates. Deficiencies in electrical properties can add further errors. The effects of worst-case impedance and capacitance unbalanced, coupled with the effects of temperature, can degrade system performance by 7-14%.

In testing, off-the-shelf shielded and (unmodified) unshielded Ethernet cables failed to meet one or more of the targeted industrial performance requirements.

CMR should be as high as is feasible. Tests indicate that the Common Mode Rejection Ratio (CMRR) for industrial cables should be minimum 40dB. To achieve high CMRR and minimize residual error rates the capacitance UN-balance of the cable must be kept below 0.5pf/m, and the cabling impedance should be controlled to less than +/-5% from 1_32 MHz. Minimizing impedance tolerance helps decrease residual errors.

Off-the-shelf cables can meet the capacitance UN-balance of 0.5pf/m but fail to meet impedance tolerances. Unshielded cables can be optimized to provide acceptable performance. Industrially-optimized unshielded cables should offer the industrial user a lower cost and less complex infrastructure.

A cable's ability to meet standards targeted by this study is typically not in correlation with its ANSI/TIA/EIA rating.

Some off-the-shelf RJ45 connectors performed adequately when tested against the recognized shock and vibration specifications, others did not. Engineers targeting industrial environments should carefully select RJ45 connectors and consider enhanced products such as IP67-rated connectors.

Switches should be employed, and careful consideration given to the operational/environmental limitations of the electronics selected. The selection of electronics is application-dependent.

Bob Lounsbury is a Physical Layer Engineer at Rockwell Automation and Jim Westerman is Marketing Manager at Anister Inc