How We Test Power Supply Units

 Our methodology, testing equipment, and benchmarks.

The power supply unit (PSU) is the most important part of every electronic device, including, of course, computers. It is the heart of your system since it feeds energy to the other components. Consequently, if the PSU fails, everything else fails with it. This is the reason most experienced technicians start a failure investigation from the PSU before proceeding to the rest of the components. And, consequently, this is why you should pay extra attention to your choice of PSU and not make a decision based exclusively on price. After all, a good PSU will do its job for quite a long time, far outlasting the rest of your expensive system components.

To properly review a PSU, expensive equipment is required, and the reviewer needs to know not only how to operate it, but also have sufficient knowledge about electronics and a PSU's design. The knowledge part is especially crucial since not even the most expensive equipment can make a good PSU review if the reviewer doesn't know how to properly use it and what tests to conduct with it.

In our reviews, we examine the PSU's performance, noise, and temperature ratings, along with the build quality of the units. We also judge the PSU's individual components, cables, connectors, and even product specifications and packaging while providing a performance per dollar comparison.

Test Setup Overview

We use several fully equipped Chroma stations. They can deliver more than 4 kW of load and consist of two 63601-5 and one 63600-2 mainframes. The mainframes mentioned above host ten 63640-80-80 [400 W] electronic loads, along with a single 63610-80-20 [100 W x2] module.

Chroma loads are widely used by all PSU manufacturers and are pretty much the standard for PSU measurements. Finally, all of our equipment is controlled and monitored by a custom-made software suite that's highly sophisticated.

In addition to the Chroma loads, we also use two Chroma AC sources (6530 with 3kW and 61604 with 2kW max power), Kesight DSOX3024A and Rigol DS2072A oscilloscopes, several Picoscope 4444 oscilloscopes, and TC-08 thermocouple data loggers, two Fluke multimeters (models 289 and 175), a Keithley 2015 THD 6.5- digit bench DMM and four lab-grade 3-phase power analyzers (three N4L PPA1530 and a single PPA5530). We recently acquired some Rohde & Schwarz HMC8015 power analyzers.

To protect our Chroma AC sources, we use high-quality online (meaning that they always run off the battery providing the best possible protection and line filtering) UPS systems with 3000VA/2700W capacity each. Two from FSP (Champ Tower 3k), and the other two by Cyberpower (OLS3000E).

Our testing gear also includes hotboxes, which allow us to test PSUs at high ambient temperatures. Finally, we have a Class 1 Brüel & Kjaer 2250-L G4 Sound Analyzer equipped with a type 4955-A low-noise and free-field microphone that can measure down to 5 dB(A) (we also have a type 4189 microphone that features a 16.6-140 dBA-weighted dynamic range). We also have a Class 1 Brüel & Kjaer 2270 Sound Analyzer, equipped with a type 4955-A microphone. 

The infrared camera is a high-end Fluke model, the Ti480 Pro. Using an IR camera is not so straightforward, especially in a PSU, since you have to apply a special coating or paint to the parts you want to check. In any case, we also double confirm our findings through temperature probes. 

We have several soldering and desoldering stations that we use during the dismantling process of every PSU we test. Test results are one thing while checking out the build quality of a PSU is another. Finally, if we encounter any unusual results during the testing process, we examine the internals of a PSU to find out what is causing the issues


The 80 PLUS certification measures efficiency at 20-, 50- and 100-percent load of the PSU's max-rated capacity up to the Gold efficiency certifications. For the Platinum and Titanium levels, they also measure efficiency with 10 percent of the PSU's max-rated capacity load.

Simply put, if a PSU has an 80 PLUS certification, then it must have the equivalent efficiency required by the corresponding certification. However, 80 PLUS measures at a mere 23 °C (73.4 °F) ambient, whereas we measure efficiency at a higher ambient temperature. This means that, in many cases, a PSU that is certified to a certain efficiency category fails to deliver the same efficiency at higher temperatures in our tests.

Besides 80 PLUS, there is also the Cybenetics efficiency and noise measurements standard, testing with more than 1450 different load combinations at higher temperatures (30 °C +-2°C) for more accurate results. Cybenetics has evaluated more than 1,000 PSUs so far, providing full evaluation reports on its database.

In our reviews, we measure efficiency with a super light load of 2% of the PSU's max-rated capacity, at light loads from 20W up to 80W, and at normal loads (10%-110%).

The ATX specification also states that the efficiency of the 5VSB rail should be measured, too. In the table below, you will find the minimum 5VSB efficiency levels that the ATX specification recommends.

Recommended System DC And AC Power Consumption

≤0.225W< 0.5W to meet 2013 ErP Lot 6 requirement (100V~240V)
≤0.45W< 1W to meet ErP Lot 6 requirement (100V~240V)
≤2.75W< 5W to meet 2014 ErP Lot 3 requirement (100V~240V)

Testing in Standby Mode

In 2010, the European Union released a guideline on Energy Related Products (ErP Lot 6), which states that every electronic device should have below 1W power consumption in standby mode. In 2013, this limit was further reduced to 0.5W. The same year, the EU also released the ErP Lot 3 guideline for computers and computer servers.

This is why we measure the consumption of a PSU in standby mode, which is something that would be difficult without our monitoring software since the readings at such low consumption levels have significant fluctuations. We have to average them over a significant period of time to provide enough accuracy.

Ripple Voltage

Ripple represents the AC fluctuations (periodic) and noise (random) found in the DC rails of a PSU. Ripple significantly decreases the life span of capacitors since it increases their temperature; a 10 °C increase can cut into a capacitor's life span by 50 percent. Ripple also plays an important role in overall system stability, especially when it is overclocked. 

The ripple limits, according to the ATX specification, are 120mV for the +12V and -12V rails, and 50mV for the remaining rails (5V, 3.3V, and 5VSB). Nonetheless, in modern PSUs, we expect to find a much lower ripple. It should be just a small fraction in high-end platforms with quality components and the proper amount of filtering capacitors. Below, you will find a schematic that analyzes a ripple waveform.

In the above schematic, four AC components can be identified:
  • Low-frequency ripple associated with AC mains frequency.
  • High-frequency ripple due to PWM of the main switches.
  • Switching noise that has the same frequency with switching PWM.
  • Non-periodic random noise that is not related to any of the above.

Hold-Up Time And Power Good Signal

Hold-up time represents the amount of time, usually measured in milliseconds, that a PSU can maintain output regulations as defined by the ATX specification without input power. Put simply, hold-up time is the amount of time that the system can continue to run without shutting down or rebooting during a power interruption.

According to the ATX spec, the PWR_OK is a “power good” signal. This signal should be asserted high, at 5V, by the power supply to indicate that the +12V, 5V, and 3.3V outputs are within the regulation thresholds and that sufficient mains energy is stored by the APFC converter to guarantee continuous power operation within specification for at least 17 ms. Conversely, PWR_OK should be de-asserted to a low state, 0V, when any of the +12V, 5V, or 3.3V output voltages falls below its under voltage threshold, or when mains power has been removed for a time sufficiently long such that power supply operation cannot be guaranteed. The AC loss to PWR_OK minimum hold-up time is set at 16 ms, a lower period than the hold-up time described in the first paragraph and ATX spec sets also a PWR_OK inactive to DC loss delay which should be more than 1 ms. This means that in any case, the AC loss to PWR_OK hold-up should be lower than the overall hold-up time of the PSU and this ensures that in no case the power supply will continue sending a power good signal, while any of the +12V, 5V, and 3.3V rails are out of spec.

The ATX specification sets the minimum hold-up time to 17 ms with the maximum continuous output load. In many cases, manufacturers use smaller capacitors in the APFC converter, resulting in a measurement of less than 17 ms. Manufacturers do this mostly to cut production costs, as these capacitors are expensive. The smaller bulk capacitors also improve efficiency by a little bit.

Measuring the hold-up time is a dangerous procedure since you have to connect an oscilloscope to the main grid. Unless you are taking the right precautions, you never want to do this; it's dangerous and you could harm yourself and your equipment!

Post a Comment

Previous Post Next Post