The voltages on the outputs from the power supply must be as close as their nominal values as possible. In other words, we want to see the +12 V outputs delivering +12 V and not with +13 V!
Voltages tend to drop with the increase in load. Switching mode power supplies are closed-loop systems, meaning that they are constantly reading the values on the output and reconfiguring the power supply on the fly to make sure that the outputs are always delivering their correct voltages.
A little difference of up to 5% for the positive voltages or up to 10% for the negative voltages is tolerable. See the table below. The -5 V voltage isn’t used anymore and was posted here just as a reference.
Output | Tolerance | Minimum | Maximum |
+12 V | ±5% | +11.40 V | +12.60 V |
+ 5 V | ±5% | +4.75 V | +5.25 V |
+5VSB | ±5% | +4.75 V | +5.25 V |
+3.3 V | ±5% | +3.14 V | +3.47 V |
-12 V | ±10% | -13.2 V | -10.8 V |
-5 V | ±10% | -5.25 V | -4.75 V |
Besides that, the power supply must be able to deliver a “clean” output. In a perfect world the voltages on the power supply outputs would draw a single horizontal line when seen on an oscilloscope. But in the real world they aren’t perfectly straight, they present a little oscillation, called ripple. On top of this oscillation you can see some spikes or noise. Ripple and noise together cannot exceed 120 mV on the +12 V outputs and 50 mV on the +5 V and +3.3 V outputs. These values are peak-to-peak values.
Let’s show you some examples for you to better understand this question. On Figure 32 we have the +12 V output of PC Power & Cooling Silencer 750 Quad delivering 750 W. Since our oscilloscope was adjusted at 0.02 V/div, this means that each green square represents 0.02 V (20 mV) in the y axis. Noise level as measured by our oscilloscope was at 50 mV, far from the 120 mV limit. Now compare Figure 32 with Figure 33. Figure 33 is the +12 V output of StarTech.com WattSmart 650 W delivering 650 W. Our oscilloscope measured 115.4 mV. Even though it was (barely) inside specs, we always want to see power supplies with ripple and noise at the lower values possible. Half the maximum allowed level is a good benchmark.
click to enlarge
Figure 32: Low noise level.
click to enlarge
Figure 33: High noise level.
Noise level is certainly something that most users aren’t aware of and only be analyzed thru reviews like the ones we post here on Hardware Secrets. The majority of websites don’t have an oscilloscope to perform their power supply reviews, publishing useless reviews (for a better discussion on this subject, take a look in our article Why 99% of Power Supply Reviews are Wrong).
