Motion Control Resources
Troubleshooting Tips: Diagnosing and Mitigating Noise Spikes
by Kristin Lewotsky, Contributing Editor
Motion Control & Motor Association Posted 05/16/2012
Our panel of experts shares tips on discovering the source of noise spikes and ways to eliminate them.
Noise spikes can cause a range of problems in motion systems. At low voltages, spikes can introduce signal problems ranging from nuisance alarms and instrumentation malfunction to equipment damage or even bodily injury, due to the undesired effect caused by the noise (e.g. an actuator speeding up to an unspecified location).
In this troubleshooting article, we will focus on tips and techniques for discovering, mitigating, and preventing noise spikes. We queried some experts in the industry to see what tips they had to offer that might speed solving your own problems.
What are the most common causes of noise spikes?
Un-terminated transmission lines can produce noise spikes up to twice the driving voltage, causing various faults in the receiver logic and sometimes even the driver logic. With the fast rise times available now, very short transmission lines can still need proper termination. Remember, it is not the frequency of the signal but the rise/fall time that causes the problem here.
Mechanics with belt drives, especially polyurethane, make it very easy to inadvertently build your own Van de Graaff generator up to 20 to 40 kV. Properly designed bearings are insulating when spinning due to the film of oil. The discharges from these generators can get into lots of inputs if care is not taken. Belts with carbon black or the use of brushed to discharge the moving parts can eliminate this issue.
When plugging and unplugging circuits with power applied, you can have the power pin make before the ground pin does, and any ground-referenced I/O will jump up to the power rail while internal capacitors charge through the I/O. Hot-plugging inductive loads can cause significant sparks if the catch circuit is not on the motor side of the circuit when the connection is broken.
—Donald Labriola, President, QuickSilver Controls Inc.
The source of the noise in drives is usually the transistors chopping in the drive. If you have a shielded motor power cable, the conductor inside of the cable and the shield on the outside of the cable make a capacitor. When you're switching on the high-frequency switches, you can either charge up that capacitor or discharge it. If you leave both ends of the cable disconnected, you can measure high voltages on the shield even though it's not connected to anything. As you connect both ends of the shield, it then makes current in the shield.
—John Zagorski, Senior Product Support Engineer, Bosch
Operating MOSFETs without sufficient gate resistance to overcome their negative resistance due to inductance in the common source circuit causes UHF oscillations as they switch. These bursts of UHF occur during transition through the linear zone while switching. The bursts can be DC rectified by lots of the input circuits across the board to be detected as spikes. Without a fast scope and a good trigger to see where they are coming from, these can be interesting to find as the rectified glitches are at a much lower frequency.
Failure to use a good zone concept to mitigate the effects of EMC. No system can be perfect but it if a good attempt is made to adhere to the EMC rules the result is usually a fault free system.
—Carl Owens, Applications Engineer, Siemens Inc.
Power switching circuits can be good noise sources. Reverse recovery or snap recovery of the diodes can excite significantly higher-frequency components, which can afflict the adjacent circuitry. Proper snubbers can significantly reduce the noise, as can proper board layout to minimize inductance and loop area in the switching circuit.
Running traces under inductive components can magnetically couple switched signals into control signals. This can happen even for “shielded” inductors, but much more with unshielded inductors.
DC brushed motors make nice noise sources, every time the brush breaks contact. Proper filtering will help.
Input L filters with inductive inputs, if not properly designed, can cause nearly a 100% overshoot in the output voltage when the input is step-switched on, causing power-supply failures. The step response of an LC circuit is a damped sinusoid with a peak amplitude of the input step and a settling value equal to the input voltage. This means that without damping, the output voltage hits double on its way to settling. Adding in some damping can reduce the overshoot. A catch diode across the inductor can also do the same (if in a DC circuit).
Primary-secondary coupling from switching power supplies can cause noise if the output from the power supply does not have a return path to the primary circuit, such as an AC or DC return path).
Poor grounding of the system. This includes the shield of the cables.
—Bill Leang, Manager of Motion Engineering, Yaskawa America Inc.
How do you detect a noise spike?
During optimization of the drive tuning, you can see excessive noise on the encoders with a trace buffer function in the drive engineering software. This should be rectified by separating the encoder cable from higher voltage wiring and grounding the shielded encoder cable on both sides.
Most customers discover they have noise spikes after they try to trace one of two problems: Either their encoders are dying at irregular and short intervals, or their system malfunctions, again on irregular intervals with drive trips or inaccurate speed or position data.
—Brian Winter, Encoder Product Manager, Avtron Industrial Automation
Normally, the symptoms will be that I can't read the encoders, there are signals on the computer that I can see, I get big voltage spikes on the DC bus that I can't synchronize to the mains with the power supply, things like that.
For very high voltage applications, undamped signals can over/under shoot to voltage levels that violate the specifications of the devices they are driving or [variable frequency drives] may exceed the insulation specifications for the cable.
—Bill Allai, Motion Control Principal Engineer, National Instruments
One of our old noise immunity tests was a nice large relay wired to turn itself off when energized. Buzzing away when attached to an AC power line, this was a good initial test before formal testing.
Most of the time, [noise spikes] show up during system start-up. An alarm appears on the machine indicating abnormal behavior.
How do you isolate the source of the spike?
I know if I see a sine wave with six little glitches in it, that's an SCR drive somewhere. The noise generated from the drive that you see on the mains is always twice the PWM frequency. We normally run at 4 kHz—if I'm measuring the noise somewhere and I see that it's 8 kHz, I can be pretty sure that the noise is generated by the drive. Also, on the power supplies, there's a switcher transistor that's 10 kHz, so if I see 20 kHz noise, I suspect that it's from our power supply. If I see voltage on the waveform harmonics, I can guess that the transformer is being overloaded or something similar. I think it takes somebody who has seen these things before and can recognize the waveforms. It's more an art than a science.
A harmonics test for the supply can tell you if this is an issue external to the system. The next step is to look for components that may be faulty or not suppressed correctly. The final stage is to start correcting the design; for example, if the 24-V wires are run with 480 VAC motor starters, take the 24-V wiring outside of the 480 VAC conduit in the panel. Success is usually a combination of different improvements.
By going through the wiring system, you can typically find the problem during one of the preventative steps—maybe you find the cable shield is grounded on both ends, or the wiring doesn’t use the cable pairs correctly. Typically, after correcting each wiring issue, we will have the customer check the system.
Once you detect a spike, how do you eliminated/mitigate it?
To reduce the magnitude of this behavior you typically have a compromise of adding external components resulting in higher cost, additional labor, failure points, and lower system efficiency. Knowing what kind of component to add and where in the system to place it can be tricky, though. There are tons of papers on this kind of topic to Google.
Since it is usually in the cabinet wiring, a scope is needed to see the voltage and current spikes. In the past ten to fifteen years, we have become used to our software tools for optimizing the drives, but for seeing noise signals, a good quality scope is needed.
Noise can affect various parts of the system such as the field bus communication, encoder communication, converter communication, instrument performance, etc. In general, check for the grounding of shielded cables (ground on one side, both sides, neither side). Check for signal lines running in parallel with power lines and change them to be perpendicular or separate by at least 30 cm.
Noise spikes are very tricky—you can waste a lot of time trying to catch them, see them visually on a scope, hooking up a strip chart or computer to run for 24 hours a day, hoping to see the issue. Then you waste more time trying to prove what is causing the spike and trying to eliminate the spike, which often proves impossible. For noise spike issues on encoders, it is far better to simply observe the bad behavior of the system, and fix the encoder and wiring issues. It saves huge amounts of time.
Honestly, we aren’t nearly as concerned about the true source of the spikes as we are committed to getting them out of the encoder signals. In our environment, it is nearly impossible to ask a vendor to rework a drive to eliminate EMI spikes—we just have to live in the environment and teach the customer how to set up the encoder and wiring to work anyway.
How can you protect a system against noise spikes?
The conduit in panels should have good separation of low voltage (24VDC) from higher control voltage (110 or 220 VAC). For the contactor coils, a suppression circuit should be used if available. The encoder and motor cables should be shielded and grounded at both the drive and motor side. The braided shield should be clamped to the shield ground on the inverter with the braided shield surface in contact with the shield ground. The surface area of this grounding is important—more surface area in contact equals better mitigation for higher frequency noise.
Have the low voltage supply one panel and the high-voltage stuff on the other panel. What I always recommend is to put these big integrated ground strips to connect isolated panels in the back of the cabinet. That seems to help a lot of noise problems. I also make sure that the motor power cables, the shields, are connected by hard surface area connections. A lot of times people want to twist the shields and put them into some kind of a connector or a screw-down terminal strip. You can't get the high-frequency stuff through that terminal strip. You really have to have 360° large surface area connections for any high-frequency currents that you want to control.
Ferrite cores can be added to cables in the system as a quick fix. More involved fixes would be to clamp cable shields to metal surfaces.
You have to have a filter that can reduce the noise on the dirty side of the filter as well as keeping it from getting to the outside world. We depend on a good connection to the backplane with the filter because wires are not very good conductors for high-frequency currents. You have to have big surface areas. We use an unpainted metal backplane with our filter and drive mounted to the backplane, and the motor power shield connected with the high-surface-area connection on the drive. The main pathway for the high-frequency currents is through the case of the drive to the backplane and to the filter. If you have the backplane painted, that makes it very difficult to get a good surface area connection.
First, be sure the encoder cable doesn’t closely parallel motor power/drive cables for long distances; and keep distance between motor/power cables and encoder cables. Use twisted-pair shielded cable (individual shielded pairs are even better), and ground the cable shield only on one end. Be sure to use differential signals (A, A/, B, B/, Z, Z/)—not single-ended (A,B,Z only). Then, be sure to wire the differential pairs (e.g. A, A/) in each twisted pair. This makes external noise spikes show up on both signals identically—and differential input circuits subtract A/ from A, meaning the noise is eliminated and ignored.
We use a combination of protection systems to keep noise spikes from killing our encoders. We use automotive grade voltage regulators on the power supply lines, allowing the encoders to tolerate spikes in excess of 50 V on 24-V devices. We add shunt diodes on the output lines, so spikes from induced voltages are directed around our circuits.
What are the most common mistakes engineers make when dealing with noise spikes?
Not considering EMC rules and looking at the failure as component failure.
Engineers don't design a machine from the beginning with noise in mind. They wait until they have a noise problem and then try to solve it and it's very difficult. The mistakes in design would be painting the back plane, not having a common ground for all the back panels as far as high-frequency stuff.
Many engineers and electricians are confused by encoder cable shield grounding. When you ground a motor/drive power cable, it should be grounded on both ends, because you are trying to keep the noise in, to drain away the EMI before it radiates, and there are no ground potential issues. But on the encoder, you only want to ground one end because ground differences between the encoder and the control cabinet can actually make noise on the encoder signal wires if you ground both ends of the shield. We see incorrect grounding of shielding on a large number of systems, or no grounding of the shield at all.
Assuming the servo amplifier or its cables to be the source of the noise spike when in reality it is the insufficient grounding of the cable shield that is influencing the noise of the system.
What was the most unusual cause of a noise spike that you ever discovered, and how did you find it?
A machine was stopping on nuisance alarms several times a day. It was caused by 60 Hz current flowing in the ground wire due to a high impedance ground fault somewhere in the building power supply.
During a commissioning of a mill there were problems with noise spikes on encoder signals. After looking at the conduit trays under the seal plates we found that 400 VAC wires were run in the same tray as low voltage PLC signal and encoder cable. The contractor only had one type of multi-conductor cable (white exterior insulation) so the motor fan wires were the same type cable as the 24-VDC I/O wiring.
I was at a place in Arkansas once and the sine waves were so distorted that could hardly recognize them as sine waves. Nothing I could find could explain the frequencies that I was seeing. The noise would all of a sudden disappear for no reason and then come back. Gradually, I turned off one thing at a time in the plant and measured the voltage—it was still there. It turned out that the noise was coming from a steel plant down the road. They had these induction furnaces and whenever they turned them on, they created noise on the three phase that made my machine crash.
Another strange one was a printing press where we kept blowing up power supplies and nobody could figure out why. Finally, we traced it back to as the paper was being pulled off of the role, static built up. You could measure the static charge on the paper at several hundred thousand volts.
About seven years ago, we had a new customer complain that he had a crane that was killing our encoder about every 30 days. It turned out that the installer hadn’t tightened the springs on the overhead power rail “shoes” for the crane. Our engineer spotted the sparks shooting from underneath the crane as it moved back and forth.
How do you decide whether it is small enough to ignore?
Typically, our noise looks like packets of energy at twice the PWM frequency, but if I zoom in on that little packet, it will be 3 or 4 MHz sine waves with decreasing amplitude. I don’t know how to specify that. I could ask what is the largest spike I can measure, but then it depends on you measuring device because a lot of times the scope probes, especially differential probes, only go to 2 or 3 MHz. If you have a really big spike, you might not be able to measure accurately how high it was or what it really looked like.
It's really hard to define whether or not this noise is going to cause a problem because it's not necessarily the peak, and it's not necessarily the 8 kHz but how much energy is in that spike, and that's very difficult to measure in the field.
To ensure system reliability, noise spikes should not be ignored [but] if the rest of the system is not affected adversely, then a spike can be considered “small enough” to be ignored.
Ascertaining the significance of the problem is usually a matter of taking measurements. The trick can be ensuring you are using the right measurement device to capture the full bandwidth of the signal of interest and not masking out information due to the fidelity of your measurements. Once armed with the data, the decision to "fix" anything becomes subjective as to the amount of margin the designer feels the system requires.
The problem is that there is always noise on these high-powered systems. It's really hard to quantify what's too much. You'll never get it to zero, so my measurement is that if the machine works, that's good enough.