Eliminating EMI in motion systems
When working with grounds and shielding, it’s common to hear complaints ranging from “It’s just black art!” to “The rules change all the time!” and “There’s no way to understand it!” Though often repeated, these emotional statements are simply not true. Effective grounding and shields work on solid engineering principles. No wizard hats, no witchcraft — just sound mathematical and logical processes that can resolve or reduce electromagnetic interference (EMI) problems.
There are four paths that let undesired signals get into a system:
• Magnetic or inductive coupling
• Capacitive coupling from high-speed voltage changes (dV/dt)
• Direct coupling, and
• Radiative or RF coupling.
One may not be able to predict the exact path a noise signal takes just as one cannot predict the exact path of lightning. But one knows how to limit the chances of getting struck. Lightning is nothing more than one of the most powerful EMI events.
To understand why it seems “The rules change all the time,” one must look to history for the frequency response and sensitivity of the circuit. The radio-frequency (RF) sections of tube receivers from the 1960s have a combination of grounds at both ends. But their audio sections use grounds only at one end. For example, consumer FM tuners receive radio frequencies from 87.5 to 108.1 MHz in the U. S. This makes them sensitive to unwanted high frequencies. To keep these higher frequencies out of the tuner, the preferred approach is to ground both ends of the circuit with no break in the shields. This shorts out both magnetic and capacitive-coupled noise — more on that later.
However, audio is more susceptible to lower-frequency noise like that from 60-Hz power lines rather than to higher RF frequencies. The audio circuit’s singled-ended, relatively high-gain input makes it more sensitive to ground differentials — variations in voltage between ground points created by stray current flows through the grounding material. A wire connected from one ground point to another with a voltage difference lets current flow between the two points. This sets up a condition known as a ground loop, a common source of 60-cycle hum in audio equipment. It’s the ground differentials and low-frequency response that provide the key to properly shielding these circuits.
By grounding a shield at the signal source only, the path for current flow through the shield created by differences in ground voltage and, thus, the ground loop, is broken. However, the signal still needs a “ground” reference at the receiving end. This is handled by another wire other than the shield in the cable, and it is connected to the receiving equipment using a differential input. The rules have not changed, but the technology has.
In those sporadic cases with only a single-ended input with high gain, it may be necessary to isolate the signal ground from other grounds. By tying all grounds to a single reference point, ground currents are greatly reduced.
To say there’s no way to understand shielding and grounding simply isn’t true. Numerous books, articles, and white papers appear on the subject all the time. They usually include information on the coupling mechanisms and how to measure inductance, capacitance, mutual inductance, resistance, and field intensity — all measurable quantities. Manufacturers such as Kollmorgen, Radford, Va., typically provide connection recommendations on how to minimize EMI and other electrical noise issues.
Sometimes the victim of an EMI problem is also the source of the problem. It may be difficult trying to determine the source of the noise in circuits that contain feedback loops, as do most motor-speed controls and power supplies. In the case of power supplies or servoamplifiers, tracking down noise sources can become a daunting task. Current and voltage feedback in these types of devices use relatively high-gain stages that can easily feed back upon themselves. Care must be taken to preserve the integrity of the intended signal with robust rejection of undesired signals.
Many switch-mode power supplies contain both current and voltage feedback. Often, they operate at higher frequencies (typically 400 kHz and up) that can create capacitive and RF-type interference problems. Inadequate shielding or a poor layout of the voltage sense line may let interference from the pulse-width-modulated (PWM) output interfere with proper sensing. The power supply becomes both the source and victim of the problem.
The only solution here is to isolate the signals with layout and shielding changes. Shield the voltage source with a 360° coverage braid. The PWM edges can also be substantially tamed with either common or differential-mode inductors, depending on the cause. The inductors slow the rise and fall times of the wave, reducing dV/dt effects.
Shielding doesn’t stop the noise
Shields merely act as an isolation barrier to a signal. Using one barrier on the source of the noise and one barrier on the receiver typically delivers the greatest integrity.
Like a Band-Aid on a wound, shields are necessary; but one shouldn’t ignore the bleeding underneath. Ideally, there should be no signal to shield against. It’s important to understand what’s causing the noise and whether it can be resolved.
There are many techniques relating to noise that truly resolve the source of the problem, not just reduce noise levels. For example, the high dV/dt signals seen in PWMs may be impacted by slowing the rise time. The change in dV/dt may break any capacitance coupling factor should the rise time fall below the reactance level.'
RF signals
Unfortunately, it’s fairly easy to create RF signals. An electric arc jumping across a spark gap generates a wide range of frequencies. Relays are common culprits, as contact arcing may arise when the relay opens. The subsequent electromagnetic field generated by the spark may cause major interference problems within the equipment. In addition, the EMI noise is broadcast where its reception may interfere with proper operation of other nearby equipment.
In the early 1970s, nearby automobiles would interfere with the reception of AM radios. The main cause of this electrical interference was the arc of the spark plugs as the magnetic field in the ignition coil collapsed with a factor of E = L di/dt. Any energy not recovered either created heat in the circuit or was transmitted as a radiated signal.
Radio signals have both magnetic and electric components that transmit at a ratio of di/dt and dV/dt. A derivative of Ohm’s Law states that circuit impedance equals the circuit voltage divided by circuit current, or Z = V/I. Substituting the change in voltage and current into that equation yields Z = dV/dt/di/dt. The dts cancel, leaving the formula Z = dV/di. Once the signal current and voltage rate exceed 377 Ω (the impedance of dry air), a signal transmits.
The automotive industry overcame this interference problem by adding a resistor in series with the spark plug. The spark’s dV/di no longer matches the impedance of air. As a result, it does not transmit enough power to generate EMI.
An understanding of the transmission and reception mechanism of a radiated signal makes it possible to devise techniques that effectively eliminate their associated problems.
First, it takes an antenna to transmit or receive a radiated signal. But any type of radiative surface can function as an antenna.
There are two fields present on any working antenna: an E-field that represents voltage and an H-field, which is magnetic. When they are in phase on the antenna, they will support each other and maintain their resonant relationship. This is sometimes referred to as sympathetic operation.
The area within one wavelength (λ) of the source is called the near field or inductive area. The RF signal within the near field shows a 90° shift between the E-field and the H-field, making a true analysis of the signal unreliable. However, the magnetic field is the primary affector; thus, a magnetic shield is used to block the transmission source and prevent reception.
Beyond 1 λ, in the far or radiative field area, the electric field has the predominant effect. Characteristics of the signal are plainly visible. For example, electromagnetic compatibility (EMC) labs measure signals 30 MHz and higher at distances over 10 m from the unit under test (UUT). Ten meters is the 1-λ wavelength for 30 MHz. Typically they use an E-field (voltage) antenna as a receptor.
Magnetic coupling
PWM signals from servoamplifiers and power supplies are often noise culprits, projecting their signals into unwanted areas. These unwanted signals have sufficient power to, in some cases, actually turn on nearby devices. For example, a high-magnetic field across a device may illuminate LEDs within the field. At the same time, any solid-state device connected to the same circuit may partially conduct at the wrong time.
An egregious example is the PWM output from servodrive systems. Typically, these outputs are sent via long cables to motors that effectively represent an inductive load in electrical terms. The combined effect of these factors generates a perfect storm for careless cable selection, grounding, shielding, and connections.
Fortunately, answers do exist to control this well-known combination of problems. The complex model of a motor amplifier, cable, and motor includes elements of inductance, resistance, capacitance, and current, along with forward and back-EMF voltages. This complex reactive load may not function as one thinks, but it is possible to use it to resolve the interference issues. In a majority of cases, this type of system is prone to the H-field or magnetic effect of the PWM signal. Interference is typically eliminated by surrounding this cable with a magnetic shield grounded at both the amplifier and motor terminals.
When a ground is not a ground
One of the most common direct-coupled noise sources arises when the ground used for reference or return is not referenced to earth as expected. This is especially prevalent in sensitive high-gain circuits. An example is a power system where a neutral line serves as the ground reference. Earth or safety ground and the neutral wire connect to the same potential in the power box. The difference is that neutral is a current-carrying conductor while an earth ground should never carry current under normal circumstances.
If the signal was monitored at this point, it’s most likely you’d see voltage fluctuations using the neutral line as reference, but you wouldn’t on the ground connection. If the power return line of the equipment was inadvertently connected to ground instead of neutral, it most likely would operate. But besides creating an electrical danger, it would likely put unintended noise on the ground line, negatively affecting all of the devices connected to that ground.
While the proper answer to this problem is to connect the system correctly, the power line inputs of the equipment should also contain filters that block sensitive devices from power-line disturbances. These power-line disturbances are quite common and filters from a number of different manufacturers sufficiently address this issue.
PWM simulation
The complex conditions for EMI noise are related but not limited to motor inductance, motor resistance, cable capacitance, shielding, and the capacitive effects of the motor windings. Add to this the unpredictable nature of what the PWM will do with back-EMF voltages and the possibilities become numerous. A model of the PWM technique can be illustrated by way of an electronic simulation program.
To start, the motor’s complex electrical qualities are modeled as an inductor in series with the resistance of the motor. Motor capacitance is lumped in with the cables, while one-third of the PWM bridge drive is modeled using IGBTs to simulate a single phase of the drive.
The standard simulation produces an indication of ringing on the leading edge of the PWM output. It happens at a substantially higher frequency than that of the PWM waveform and is usually only on the rising edge of the pulse.
Given this propensity to oscillate and the fact that each transistor will produce ringing, it’s easy to see how the frequencies involved can couple into a load that contains uneven resistance, inductance, and capacitance (RLC) characteristics.
Common-mode noise, as well as voltage spikes from the di/dt through the inductance of the motor winding, can damage insulation, producing an electrical failure in the motor or drive.
These oscillations are prevalent on the PWM of all transistors in the output power bridge. Any load imbalance results in current flow from the oscillation source to ground via this RLC circuit. If the inductance path to ground is a lower impedance, it usually results in some of those leading edge oscillations leaking onto the signal returns.
To correct this problem requires a threefold attack. First, reduce the emissions. Second, shield the receiver to break the coupling. And third, force a current path to ground via the intended connections.
One drive model places an inductor in series to slow the rise time of the PWM signal and reduce oscillations. Analysis of the PWM signal shows there is some improvement, but not enough to consider this an effective answer.
While they are easier to shield than the original oscillations, there are still some significant oscillations that could couple into the system. Reducing the amplitude and frequency of these oscillations further should be the goal here.
An alternative solution is to add a snubbing circuit. This often results in a well-controlled PWM pulse without the wild oscillating rising edge that can lead to EMI noise. This model includes both a common-mode inductor as well as a series mode matched to cable capacitance. The oscillation of this PWM edge is benign. The drawback of this design, however, is that it slows PWM power changes in the drive.
Now with a good connection to ground and a lower frequency of ringing, it’s significantly easier to eliminate the noise. A good shield and ground can minimize the EMI infiltration without significantly adding to the cost of the system.
However, don’t forget that cables also contribute to the oscillation effect. The combination of cable capacitance and motor inductance can produce a “tank circuit” that resonates at a frequency related to the rise time of the PWM. Should this happen, even accidentally, there may be catastrophic results.
As can be seen from the final model, both transmission and reception of EMI can be taken care of with shields, inductance, and capacitance in the right areas. This is the best answer for an existing design, with minimal impact on delivering a sound product.
Finally, manufacturers that are CE compliant should have an intimate understanding of EMI and EMC compliance. Drives displaying the CE mark should come with information about how to maintain compatibility as well as supply external products for special cases.
One final note: An EMI Checklist
This checklist collects the experience of Kollmorgen engineers who have spent many hours in the field solving noise-related problems. These guidelines can’t guarantee protection against noise issues, but can reduce the chance of problems with electrical noise.
1. Use approved cables or cabling. Many manufacturers have either designed or specified the best cables for use with their servosystems. Experience has shown that machine builders who use the specified power and feedback cables have far fewer problems than those who do not.
2. Use common-mode chokes on motor leads as required. Cables longer than 25 m (82 ft) may need motor common-mode chokes. Check the product documentation for details to see if this is the case for your installation.
3. Separate drive/motor power and signal cables. Bundle and route signal cables separately from motor/power cables. Either run the cables in separate conduits or maintain at least 100 mm (4 in.) between the signal and power bundles for drives under 20 A. Use 150 mm (6 in.) for drives up to 40 A and 200 mm (8 in.) for 80-A drives. If you are using a separate ac-power filter, maintain separation of leads entering and exiting the line power filter. Locate the filter as close as possible to the point where the incoming power enters the cabinet. If the system has internal power, maintain at least 100 mm (4 in.) of separation between line input power and output motor leads. If input power and motor leads must cross, cross them at a 90° angle.
4. Splice cables properly. If you need to divide or split cables, use connectors with metal backshells. Ensure that both shells connect along the full 360° of the shields. No portion of the cabling should be unshielded. Never divide a cable across a terminal strip.
5. Ensure good shield connections. For cables entering a cabinet, connect shields on all 360° of the cable. Never connect a simple “pigtail.”
6. Use differential inputs for analog signals. Noise susceptibility in analog signals is greatly reduced by using differential inputs. Normally, connect the output signal to the noninverting (+) differential input and the ground of the device generating the output to the inverting (–) differential input. Use twisted-pair, shielded signal lines, connecting shields on both ends.
7. Ensure good connections between cabinet components. Connect the back panel and cabinet door to the cabinet body using several conductive braids. Never rely on hinges or mounting bolts for ground connections. Provide an electrical connection across the entire back surface of the drive panel. Electrically conductive panels such as aluminum or galvanized steel are preferred. For painted and other coated metal panels, remove all coating behind the drive.
8. Ensure good ground connections. Connect from the cabinet to proper earth ground. Ground leads should be the same gauge as the leads to main power or one gauge smaller.