For the PDF version of this article, click here.
Changes in power supply design can cause big headaches for design engineers, even when a customer requests only a few changes to the design of an already tested and proven power supply. Simple redesigns often have more roadblocks than anticipated. And simple changes can require extensive troubleshooting to ensure that the device meets all codes and specifications. Such troubleshooting often requires specific skills and knowledge of electromagnetic compliance (EMC) that fall outside many design engineers' core competencies.
This article examines how unintended antennas are created during power supply redesign and provides guidelines to remediate unwanted noise problems associated with those antennas. All noise currents have a return path to their source. Minimizing or redirecting this path to keep noise from radiating into the environment is the overall objective.
At Sypris Test & Measurement (Billerica, Mass.), many customers request power supply redesigns to reconfigure the instrument from Class A (commercial environment) to Class B (residential environment). A typical scenario might include a pre-designed power supply that already fits a customer's application. Only a few minor changes are necessary to deliver on time and meet specifications. The original design passed all Federal Communications Commission (FCC)/Common Market Europe (CE) noise requirements. In addition, the power supply is a big seller in the marketplace, everyone is using it and there are only a few customer complaints. So, modifying the existing power supply might appear to be a win-win proposition for the customer and the power supply designer.
Not so fast. To qualify the power supply according to Class B specifications, for example, the mechanical box in which it is housed will need to be adjusted. Since we cannot control a customer's box, it's essential to locate and resolve problems in the redesign of the power source early in the process, because remediation of radiated and conducted noise becomes increasingly difficult as the implementation process proceeds.
What follows are some common design and troubleshooting problems experienced by customers prior to working with our professional EMC laboratory. While each noise problem is unique and requires a different solution, certain guidelines are universal in every power source redesign.
Improper Grounding Causes Radiated Noise
There are two different types of grounding: single-point and multipoint grounding. Power supply manufacturers typically use single-point grounding because this method is ideal for dc and low-frequency noise. However, single-point grounding is effective only at frequencies below 100 kHz. Therefore, a multipoint method is necessary to properly ground the power supply to meet the requirements for conducted emissions in the 150-kHz to 30-MHz range and requirements for radiated emissions above 30 MHz.
The effect of multipoint grounding is illustrated by measuring the radiated emissions of a power supply unit in accordance with CISPR22 Class B limits (Figs. 1 and 2). The power supply under test operates from an input of 100 Vac to 220 Vac and generates 3.3 Vdc output at 60 A and 5 Vdc output at 20 A. In this test, radiated emissions were measured from 30 MHz to 300 MHz using a horizontally oriented antenna positioned 3 m from the power supply under test. Test data was run through looped-back data cables.
Fig. 1 displays the radiated emissions measurements for the case where the power supply uses a single-point ground at the line filter. In Fig. 2, the same power supply was tested using a multipoint ground connected through four securing screws and at the line filter. In these plots, CISPR22 Class A and Class B limits are shown on channels 1 and 2, respectively.
With the single-point ground, the power supply under test met CISPR Class A limits for radiated emissions, but exceeded the more stringent Class B limits. However, when a multipoint grounding system was implemented in the same power supply, emissions were reduced below the Class B limits. Note that the grounding of the power supply only affected frequencies under 60 MHz. That is because this power supply had little noise above 60 MHz. The rest of the noise was due to the logic circuitry not affected by the power ground.
When customers transition from Class A to Class B, I've seen numerous misguided attempts to solve the grounding problem. Sometimes a customer will attempt to use a filter to reduce noise. Yet, this device is not related to the grounding problem, and noise problems can continue.
Other times they add front-end capacitance. Capacitors draw a large transient current to charge, which might saturate the inductor in the filter, making it useless to suppress harmonics in the surge current. The result: Noise is reflected back to the power source due to mismatched impedance.
Another result of improper grounding is capacitive coupling from the case of the switching component, creating a common-mode conduction that forms a loop from the case to the supply. This can be reduced by grounding the heatsink. A floating switching component coupled to the heatsink is typically the source of the problem. Placing four screws in each corner and substantial posts secured to a pad — not just an annular ring connection — is a good solution. This method is effective below 1 GHz.
A basic oversight that leads to noise is failure to twist all lead wires and returns together. This cancels out the magnetic fields they produce, thereby eliminating noise.
Customers often overlook the circuit board as a noise source. Let's suppose a large noise loop is created. The problem lies in the return path for the noise. Typically, customers look to add more layers to the circuit board to reduce noise. This is an expensive method. The simple remedy is running a ground plane adjacent to each power or signal layer so that you don't have to worry about the noise returns for any conductor. An even less costly solution is to run a ground conductor next to a high current switching or signal conductor, reducing the noise loop.
Filter Design Becomes Inadequate
Oftentimes, when using an existing power supply in a new application, the original filter design doesn't block all the logic circuitry noise as the power supply changes from Class A to Class B. Filters are designed to reduce the impedance mismatch between the low impedance input line and the high impedance of the input caps. The solution is to design a filter that takes into account a broad range of possible noise frequencies, covering conducted emissions at 150 kHz to 30 MHz and radiated emissions from 30 MHz to 1 GHz. As a general rule: The higher the capacitance of the input, the more inductive impedance necessary in the filter.
The effect of adding a line filter for conducted emissions is reflected in the measurements shown in Figs. 3 and 4. Conducted emissions were measured in accordance with CISPR22 requirements with Class B limits identified in the plots. Input to the power supply was 240 Vac, 50 Hz. In Fig. 3, the measurements were taken with no line filter connected while in Fig. 4 a line filter is installed. As shown in the plots, adding the line filter reduced conducted emissions to the point where the power supply passes CISPR Class B limits.
Note that the line filter is in a metal case and the case is grounded to the equipment chassis. The input power is filtered immediately upon entry to the equipment chassis. This way, no noise from the logic circuitry or power supply can get on the power cord unless it passes through the filter. The line filter was one with a 100-Vac to 220-Vac, 6-A rating specifically for switching power supplies, providing a higher serial impedance than a typical line filter. To choose a filter, look at the frequency versus attenuation curves for the different ferrite materials in the line filters. Pick a filter that has attenuation in the frequency range where your noise emissions need correction.
Overall Strategy for Reducing Noise
A solid ground plane on any given circuit board can provide the return for all circuits. This layer keeps any noise loops small and reduces noise effects on the control circuitry. The proper design does not allow for splitting these planes. However, often it is necessary to have separate planes for primary and secondary circuits at 50 Hz/60 Hz. The noise currents need to flow across these gaps to keep the return loops small. To accomplish this, connect the two planes for only high frequencies by placing a surface-mount capacitor (with proper dc hipot characteristics) across the gap, near any conductor that crosses the two planes. Now the noise return follows the source, reducing the noise loop.
Troubleshooting the source of EMC violations in power supplies is a labor-intensive task, and minor changes to EMC-compliant units can produce big problems for designers. Based on my experience, problems with noise are almost certain to appear in new designs. Working with an EMC specialist early on in the design phase will significantly reduce testing time, cost and time-to-market.
“A Handbook Series on Electromagnetic Interference and Compatibility,” Volume 4, Filters and Power Conditioning, Interference Control Technologies Inc., Gainesville, Va.
Archambeault, Bruce R. PCB Design for Real-World EMI Control, Kluwer Academic Publishers.