How to Avoid Power Supply PitfallsSmall details can make the difference between a quality power supply unit and a problematic oneWhen one looks at purchasing a power supply, it is often done with little research and using only the most basic information available. Consumers may only look for the following features:
To make an informed decision, it helps to understand a bit more about how a power supply is built. This document outlines some of the parameters that can drive quality and cost of a power supply. Rather than delving into complex design architectures and their pros and cons, this guide offers a straightforward explanation of common terms you'll encounter in reviews and testing, clarifying how these elements influence a power supply's real-world performance, overall level of quality, and reliability. Common Issues and Their CausesWhat is Ripple and Why High Ripple is Bad: Your home supplies alternating current (AC) power from the wall, but your computer components require directional current (DC). This is the purpose of your computer’s power supply; to convert AC to DC. When converting AC to DC, AC becomes a part of the DC output. In a switch mode supply the process of converting AC to DC, back to AC, and then finally back to DC again, can generate high-frequency noise and harmonics known as ripple. The smaller capacitors on the output side of a power supply help filter out as much ripple as possible. However, increasing their capacitance to filter out more ripple also raises the in-rush current because all these capacitors need to be fully charged to reach the desired output voltage. Output voltage regulation can be affected when power excursions, also known as transients, partially drain the capacitors, requiring them to recharge to maintain proper output voltage. This also stresses the switching components. So, it’s best to have the right balance of capacitance and an acceptable amount of filtering for ripple.
Poor Transient Response: Transient response refers to the behavior of output voltage as the PSU transitions from one load to another. It’s how we describe how a PSU responds to sudden load changes before settling into a steady state. For example, a sudden change in load from low to high may cause output voltage to drop. This can be normal behavior, but the question we ask is how long does the PSU stay in this state before the output voltage recovers to a nominal voltage? Quality Control Issues: Poor assembly practices and lack of quality control can result in premature failures. The manufacturing of power supplies remains largely manual, with automation reserved for premium manufacturers. Before reaching the bulk solder machine, components must be manually placed into the PCB. While these machines excel at consistent high-volume soldering, their output quality hinges entirely on proper input preparation. Low or No Power Factor Correction: Most PSUs in the EU and North America have power factor correction, and a high-power factor correction is required in order to achieve either an 80 PLUS or Cybenetics certification level of efficiency. However, some parts of the world do not require power factor correction. Additionally, some brands often “fake” their efficiency badges (this is when you see badge that look like 80 PLUS but say “85 PLUS” or “90 PLUS”, or the badge will look like an 80 PLUS badge, but there is no corresponding 80 PLUS report. Apparent power represents the total power supplied to the circuit and is measured in Volt Amps (VA). Real Power is measured in Watts (W) and represents the actual AC power that performs tangible work—such as powering a motor or illuminating a bulb. In the context of this article, it refers to the power that is ultimately converted into DC. Reactive power is considered problematic because it increases the overall current flowing through the power supply. Although it doesn't supply useful energy to the computer that the PSU is powering, it still adds to the current that power lines, transformers, and generators need to manage. This additional current results in more resistive losses on the power grid, generating heat and wasting energy.
Hold-Up Time Issues: Hold up time is a specification for how long the PSU can continue to deliver stable, regulated DC output power after there is an interruption of AC input power, which can happen when a brown out occurs. Regardless of which specification a PSU follows, it must maintain proper voltage levels during this hold up period (keeping +12V outputs above +11.4V, for instance) to safeguard the connected components from potential damage. Proper Fan Design and Fan Control: Fans used in power supplies need to have a higher static pressure specification (measured in mmH2O, or "millimeters of water column"), much like a radiator fan with dense, tightly packed blades, versus your typical case fan with its wider, more sweeping design which would focus on CFM (air measured in cubic feet per minute). Chinese Brand Components and Questionable Specifications: It’s no secret that, due to supply chain convenience, most power supply components are made in China. Not just finished goods, but the actual components used to manufacture these goods. Japanese capacitors? Made in China. German MOSFETs? Made in China. Korean ICs? Made in China. And this is all good as these parts follow strict quality control measures. But lately, we’ve seen a number of China-based brands producing products with datasheets that match their international brand counterparts, but in our testing we’ve often found that the specs of these Chinese brands do not align with their spec sheets. This is particularly problematic with Metal Oxide Semiconductor Field-Effect Transistors (MOSFETs). We’ve found that sometimes they run hotter, despite having the same RDS (on) (more on that below) as their more expensive counterpart. This challenges our ability to properly cool the device, while still maintaining a low-noise product. A higher operating temperature also increases the risk of thermal runaway. Thermal runaway is a self-reinforcing heating cycle. As a MOSFET conducts current, it dissipates heat from conduction and switching losses. As the junction temperature rises, key parameters (like RDS (on), threshold voltage, leakage current) change. These changes increase power loss even further, causing more heating. Inevitably, the MOSFET may exceed its Safe Operating Area (SOA), leading to destruction. RDS (on) is the drain-to-source resistance of a MOSFET when it is fully turned on (saturated). It determines how much the MOSFET resists current flow during conduction. The higher the ACTUAL RDS (on), the hotter the transistor. Not Using Soft Termination MLCCs in “High Risk” Locations Can Cause Premature Failures: Multilayer Ceramic Capacitor (MLCC) is a widely used surface mount technology (SMT) capacitor found in electronic devices. These passive components store electrical energy and are mainly utilized for decoupling, filtering, bypassing, and timing tasks within circuits. For power supply units with modular PCBs that use MLCCs, it is recommended to exclusively use soft termination MLCCs. This precaution is necessary because the insertion and removal of modular cables can cause the PCB to flex, potentially damaging the MLCC or its connection to the board. Why Are Some Power Supplies “230V Only": By Watt’s law, lower voltage requires higher current for the same power, so 100–127V regions such as North America, Japan, Taiwan, and other areas need bulkier, more heat-resistant rectifiers, which raises costs. If you're designing for countries where residential voltages are 220V or higher, there's no need for a rectifier that can handle as much current. Nevertheless, brownouts can reduce voltage significantly below what is considered normal. Hence, it's crucial to have proper protections in place to prevent the power supply from failing catastrophically if the input current surpasses the bridge rectifier's capacity. Why You Can’t Find a 2000W+ PSU That Works in Your American Home: Residential voltages vary widely: Japan uses 100V; Taiwan, Cuba and many U.S. outlets measure about 115V (nominally 120V); parts of South America and the Caribbean use 127V. Poor Cable QualityAn often-overlooked aspect of power supplies are the cables that are included, the materials used to construct them, and how well they are constructed. The following three points are the biggest pitfalls we’ve seen when “cheaper cables” are provided. Aluminum Wires: Copper-Clad Aluminum (CCA) wire is an aluminum wire with a copper coating, while pure copper wire is made entirely of copper. Pure copper is superior in terms of electrical conductivity, flexibility, and resistance to heat and corrosion. CCA is cheaper, but it is not recommended for many applications, especially where electrical safety and performance are critical, due to its higher resistance (55 to 60% higher for aluminum versus copper wire of the same gauge) and lower melting point. Copper is also more flexible and less prone to breaking after repeated bending.
Incorrect Wire Gauge for Intended Application: Copper isn’t cheap. The thicker the copper, the more expensive the wire. The problem is that thinner copper has more resistance than thicker copper. Unfortunately, wire gauge is not often marked on cables, and a customer may perceive a smaller gauge wire used in a cable as a better-quality cable because they have better flexibility. Lower Rated Temperature Rating for a PVC Jacket: Polyvinyl Chloride (PVC), also known as a plastic polymer, insulation softens and can deteriorate over time when exposed to temperatures exceeding its rated limit. The breakdown of the PVC insulation can compromise the wire's ability to safely carry electrical current. Over time, this can lead to cracks, shorts, and potentially even electrical fires. Better Parts Do Mean Lower NoiseToroid Coils Inductors vs. Rod Coils: A toroid coil inductor consists of a ferromagnetic ring wound with copper wire. A rod coil inductor uses a cylindrical core wrapped along its length. The only real benefit of a rod core is lower material and labor cost. Toroid cores, by contrast, generate less audible noise. The magnetic forces within do not cause bending in the core—only compression or tension—and their circular design offers better mechanical stability.
The impacts of rod coils tend not to be evident in the lab. I once worked on a project that used rod coils on the output stage. Lab testing on our Chroma didn’t detect audible noise. But real-world usage told a different story: complaints came pouring in, and we quickly switched to toroids. Using Sendust as a Ferromagnetic Material for Inductors: Sendust is a magnetic alloy that was created as an alternative to iron powder and other magnetic core materials used in inductors and transformers. It is made up of 85% iron, 9% silicon, and 6% aluminum. Sendust is highly regarded because it has lower eddy current losses and does not produce mechanical vibrations when exposed to magnetic fields. In our industry, the primary drawback is its cost. Additionally, since Sendust is a sintered compound, it tends to be more brittle than other materials. Consequently, Sendust inductors might need to be slightly larger to match the energy storage capacity of an iron core, so space limitations in smaller designs should be considered. Using Split-Windings as Opposed to Single-Windings for the Power Factor Correction Inductor: Once again, we have an optimal solution that requires additional space. However, if space permits, we highly recommend using split-windings. As the term suggests, a single-wound inductor consists of a single continuous copper wire coil wrapped around the core. A split-wound inductor, also called a bifilar-wound or common-mode choke, divides the copper wire into two separate windings around the core. This configuration effectively blocks common-mode noise, making it perfect for use as a PFC choke since it helps reduce electromagnetic interference (EMI) and radio frequency interference (RFI). Common-mode noise often arises from parasitic capacitances between the MOSFETs and ground. Even though the split-wound inductor is larger, its cost should not significantly exceed that of a single-wound inductor, provided the manufacturing plant uses an automated process to produce them.
|