Power Supply 101: A Reference Of Specifications

Other Power Supply Specifications And Certifications

In addition to power output, many other specifications and features go into making a high-quality power supply. I have had many systems over the years. My experience has been that if a brownout occurs in a room with several systems running, the systems with higher-quality power supplies and higher output ratings are far more likely to make it through the power disturbances unscathed, whereas others choke.

High-quality power supplies also help protect your systems. A high-quality power supply from a vendor such as PC Power and Cooling will not be damaged if any of the following conditions occur:

  • A 100% power outage of any duration
  • A brownout of any kind
  • A spike of up to 2500V applied directly to the AC input (for example, a lightning strike or a lightning simulation test)

Decent power supplies have an extremely low current leakage to ground of less than 500 microamps. This safety feature is important if your outlet has a missing or an improperly wired ground line.

As you can see, these specifications are fairly tough and are certainly representative of a high-quality power supply. Make sure that your supply can meet these specifications.

You can also use many other criteria to evaluate a power supply. The power supply is a component many users ignore when shopping for a PC, so it is one that some system vendors choose to skimp on. After all, a dealer is far more likely to be able to increase the price of a computer by spending money on additional memory or a larger hard drive than by installing a better power supply.

When buying a computer (or a replacement power supply), learn as much as possible about the power supply. Many consumers are intimidated by the vocabulary and statistics found in a typical power supply specification. Here are some of the most common parameters found on power supply specification sheets, along with their meanings:

  • Mean Time Between Failures (MTBF) or Mean Time To Failure (MTTF)—The (calculated) average interval, in hours, that the power supply is expected to operate before failing. Power supplies typically have MTBF ratings (such as 100 000 hours or more) that are clearly not the result of real-time empirical testing. In fact, manufacturers use published standards to calculate the results based on the failure rates of the power supply’s individual components. MTBF figures for power supplies often include the load to which the power supply was subjected (in the form of a percentage) and the temperature of the environment in which the tests were performed.
  • Input Range (or Operating Range)—The range of voltages that the power supply is prepared to accept from the AC power source. For 120 V AC power, an input range of 90 V–135 V is common; for 240 V power, a 180 V–270 V range is typical.
  • Peak Inrush Current—The greatest amount of current drawn by the power supply at a given moment immediately after it is turned on, expressed in terms of amps at a particular voltage. The lower the current, the less thermal shock the system experiences.
  • Hold-Up Time—The amount of time (in milliseconds) that a power supply can maintain output within the specified voltage ranges after a loss of input power. This enables your PC to continue running without resetting or rebooting if a brief interruption in AC power occurs. Values of 15–30 milliseconds are common for today’s power supplies, and the higher (longer), the better. The Power Supply Design Guide for Desktop Platform Form Factors specification calls for a minimum of 16 ms hold-up time. The hold-up time is also greatly affected by the load on the power supply. The hold-up specification is normally listed as the minimum time measured under the maximum load. As the load is reduced, hold-up times should increase proportionately. For example, if a 1000 W PSU has a 20 ms hold-up time specification (measured under a 1000 W load), then under a 500 W (half) load I’d expect that to double, and under a 250 W load I’d expect it to double again. This is in fact one of the reasons I’ve always been a proponent of specifying higher output PSUs than are strictly necessary when building systems.
  • Transient Response—The amount of time (in microseconds) a power supply takes to bring its output back to the specified voltage ranges after a steep change in the output current. In other words, the amount of time it takes for the output power levels to stabilize after a device in the system starts or stops drawing power. Power supplies sample the current being used by the computer at regular intervals. When a device stops drawing power during one of these intervals (such as when a floppy drive stops spinning), the power supply might supply too high a voltage to the output for a brief time. This excess voltage is called overshoot, and the transient response is the time that it takes for the voltage to return to the specified level. This is seen as a spike in voltage by the system and can cause glitches and lockups. Once a major problem that came with switching power supplies, overshoot has been greatly reduced in recent years. Transient response values are sometimes expressed in time intervals, and at other times they are expressed in terms of a particular output change, such as “power output levels stay within regulation during output changes of up to 20%.”
  • Overvoltage Protection—Defines the trip points for each output at which the power supply shuts down or squelches that output. Values can be expressed as a percentage (for example, 120% for +3.3 and +5 V) or as raw voltages (for example, +4.6 V for the +3.3 V output and +7.0 V for the +5 V output).
  • Maximum Load Current—The largest amount of current (in amps) that safely can be delivered through a particular output. Values are expressed as individual amperages for each output voltage. With these figures, you can calculate not only the total amount of power the power supply can supply, but also how many devices using those various voltages it can support.
  • Minimum Load Current—The smallest amount of current (in amps) that must be drawn from a particular output for that output to function. If the current drawn from an output falls below the minimum, the power supply could be damaged or automatically shut down.
  • Load Regulation (or Voltage Load Regulation)—When the current drawn from a particular output increases or decreases, the voltage changes slightly as well—usually increasing as the current rises. Load regulation is the change in the voltage for a particular output as it transitions from its minimum load to its maximum load (or vice versa). Values, expressed in terms of a +/– percentage, typically range from +/–1% to +/–5% for the +3.3 V, +5 V, and +12 V outputs.
  • Line Regulation—The change in output voltage as the AC input voltage transitions from the lowest to the highest value of the input range. A power supply should be capable of handling any AC voltage in its input range with a change in its output of 1% or less.
  • Efficiency—The ratio of power input to power output, expressed in terms of a percentage. Values of 65%–85% are common for power supplies today. The remaining 15%–35% of the power input is converted to heat during the AC/DC conversion process. Although greater efficiency means less heat inside the computer (always a good thing) and lower electric bills, it should not be emphasized at the expense of precision, stability, and durability, as evidenced in the supply’s load regulation and other parameters.
  • Ripple (or Ripple and Noise, or AC Ripple, or PARD [Periodic and Random Deviation])—The average voltage of all AC effects on the power supply outputs, usually measured in millivolts peak-to-peak or as a percentage of the nominal output voltage. The lower this figure, the better. Higher-quality units are typically rated at 1% ripple (or less), which if expressed in volts would be 1% of the output. Consequently, for +5 V that would be 0.05 V or 50 mV (millivolts). Ripple can be caused by internal switching transients, rectified line frequency bleed-through, or other random noise.

Power Factor Correction

In order to improve power line efficiency and to reduce harmonic distortion generation, the power factor of PC power supplies has come under examination. In particular, new standards are now mandatory in many European Union (EU) countries that require harmonics to be reduced below a specific amount. The circuitry required to do this is called power factor correction (PFC).

The power factor measures how effectively electrical power is being used and is expressed as a number between 0 and 1. A high power factor means that electrical power is being used effectively, whereas a low power factor indicates poor utilization of electrical power. To understand the power factor, you must understand how power is used.

Generally, two types of loads are placed on AC power lines:

  • Resistive—Power converted into heat, light, motion, or work
  • Inductive—Sustains an electromagnetic field, such as in a transformer or motor

A resistive load is often called working power and is measured in kilowatts (KW). An inductive load, on the other hand, is often called reactive power and is measured in kilovolt-amperes-reactive (KVAR). Working power and reactive power together make up apparent power, which is measured in kilovolt-amperes (KVA). The power factor is measured as the ratio of working power to apparent power, or working power/apparent power (KW/KVA). The ideal power factor is 1, where the working power and apparent power are the same.

The concept of a resistive load or working power is fairly easy to understand. For example, a light bulb that consumes 100 W of power generates 100 W worth of heat and light. This is a pure resistive load. An inductive load, on the other hand, is a little harder to understand. Think about a transformer, which has coil windings to generate an electromagnetic field and then induce current in another set of windings. A certain amount of power is required to saturate the windings and generate the magnetic field, even though no work is being done. A power transformer that is not connected to anything is a perfect example of a pure inductive load. An apparent power draw exists to generate the fields, but no working power exists because no actual work is being done.

When the transformer is connected to a load, it uses both working power and reactive power. In other words, power is consumed to do work (for example, if the transformer is powering a light bulb), and apparent power is used to maintain the electromagnetic field in the transformer windings. In an AC circuit, these loads can become out of sync or phase, meaning they don’t peak at the same time, which can generate harmonic distortions back down the power line. I’ve seen examples in which electric motors have caused distortions in television sets plugged into the same power circuit.

PFC usually involves adding capacitance to the circuit to maintain the inductive load without drawing additional power from the line. This makes the working power and apparent power the same, which results in a power factor of one. It usually isn’t just as simple as adding some capacitors to a circuit, although that can be done and is called passive power factor correction. Active power factor correction involves a more intelligent circuit designed to match the resistive and inductive loads so the electrical outlet sees them as the same.

A power supply with active power factor correction draws low distortion current from the AC source and has a power factor rating of 0.9 or greater. A nonpower factor-corrected supply draws highly distorted current and is sometimes referred to as a nonlinear load. The power factor of a noncorrected supply is typically 0.6–0.8. Therefore, only 60% of the apparent power consumed is actually doing real work!

Having a power supply with active PFC might or might not lower your electric bill (it depends on how your power is measured), but it definitely reduces the load on the building wiring. With PFC, all the power going into the supply is converted into actual work, and the wiring is less overworked. For example, if you ran a number of computers on a single breaker-controlled circuit and found that you were blowing the breaker periodically, you could switch to systems with active PFC power supplies and reduce the load on the wiring by up to 40%, meaning you would be less likely to blow the breaker.

The International Electrotechnical Commission (IEC) has released standards dealing with the low-frequency public supply system. The initial standards were 555.2 (Harmonics) and 555.3 (Flicker), but they have since been refined and are now available as IEC 1000-3-2 and IEC 1000-3-3, respectively. As governed by the EMC directive, most electrical devices sold within the member countries of the EU must meet the IEC standards. The IEC1000-3-2/3 standards became mandatory in 1997 and 1998.

Even if you don’t live in a country where PFC is required, I highly recommend specifying PC power supplies with active PFC. The 80 PLUS certification for highly efficient power supplies also includes a requirement that the power supply has active PFC. The main benefits of PFC supplies is that they do not overheat building wiring or distort the AC source waveform, which causes less interference on the line for other devices.

SLI-Ready and CrossFireX Certifications

Both Nvidia and AMD have certification programs that test and certify power supplies to be able to power systems with multiple graphics cards in either a Scalable Link Interface (SLI) or CrossFire configuration. This type of configuration puts extreme demands on the PSU, because it not only has to power what would normally be a high-end motherboard, CPU, and multiple drives in a RAID configuration, but also up to three video cards, which may be capable of drawing 300 watts or more each.

The certification process involves PSU manufacturers sending PSUs in for testing, whereby they are verified to supply sufficient power (and the proper type and number of connectors) to run the desired graphics hardware as well as the system. Power supplies that have passed either of these certifications are virtually guaranteed to produce high output and use high-quality design, engineering, and manufacturing. For more information on these certifications, as well as lists of certified PSUs, visit the following links.

Certified SLI-Ready Power Supplies— www.slizone.com/object/slizone_build_psu

CrossFire Certified Power Supplies—http://support.amd.com/us/certified/power-supplies

I recommend checking the lists or looking for the “NVIDIA SLI-Ready” or “AMD CrossFireX Technology” logos on a power supply as an excellent indicator of a high-power, high-quality unit.

Safety Certifications

Many agencies around the world certify electric and electronic components for safety and quality. The most commonly known agency in the United States is Underwriters Laboratories, Inc. (UL). UL standard #60950—Safety of Information Technology Equipment —covers power supplies and other PC components. You should always purchase power supplies and other devices that are UL-certified. It has often been said that, although not every good product is UL-certified, no bad products are.

In Canada, electric and electronic products are certified by the Canadian Standards Agency (CSA). The German equivalents are TÜV Rheinland and VDE, and NEMKO operating in Norway. These agencies are responsible for certification of products throughout Europe. Power supply manufacturers that sell to an international market should have products that are certified at least by UL, the CSA, and TÜV—if not by all the agencies listed, and more.

Apart from UL-type certifications, many power supply manufacturers, even the most reputable ones, claim that their products have a Class B certification from the Federal Communications Commission (FCC), meaning that they meet FCC standards for electromagnetic and radio frequency interference (EMI/RFI). This is a contentious point, however, because the FCC does not certify power supplies as individual components. Title 47 of the Code of Federal Regulations, Part 15, Section 15.101(c) states the following:

“The FCC does NOT currently authorize motherboards, cases, and internal power supplies. Vendor claims that they are selling ‘FCC-certified cases,’ ‘FCC-certified motherboards,’ or ‘FCC-certified internal power supplies’ are false.”

In fact, an FCC certification can be issued collectively only to a base unit consisting of a computer case, motherboard, and power supply. Thus, a power supply purported to be FCC-certified was actually certified along with a particular case and motherboard—not necessarily the same case and motherboard you are using in your system. This does not mean, however, that the manufacturer is being deceitful or that the power supply is inferior. If anything, this means that when evaluating power supplies, you should place less weight on the FCC certification than on other factors, such as UL certification.

This thread is closed for comments
    Your comment
  • joytech22
    On the other hand, if you plug into a 240 V outlet and have the switch set for 120 V, you can cause damage.

    Did that when unboxing a computer, must have flipped the small red switch on the supply and boom, at the Windows XP loading bar the PSU exploded. lol.
  • cmcghee358
    Did I miss them covering efficiency and the whole 80 PLUS thing?

    I can't imagine as detailed as it is, omitting something like that...
  • cangelini
    cmcghee358Did I miss them covering efficiency and the whole 80 PLUS thing?I can't imagine as detailed as it is, omitting something like that...

    There's still one last part to go!
  • cmcghee358
    But the last part isn't for PSUs. It's just the last part in the series of PC components.
  • nikorr
    Thanx ...
  • neiroatopelcc
    I wonder how much the power_good signal prevents? is it just the powering of the cpu ?
    I recall once using two power supplies to power a sli board and accidently use a molex from the second supply to power a sli power connector on the motherboard - resulting in fans powering up if you powered the second psu even when the first wasn't on (and if you didn't, the geforces would screech due to lack of power)..... maybe that was just the creative yet rubbish asrock board design, but it certainly didn't need a power_good to power up the fans.

    ps. "Note: If you find that a system consistently fails to boot up properly the first time you turn on the switch, but that it subsequently boots up if you press the reset or Ctrl+Alt+Delete warm boot command, you likely have a problem with the Power_Good timing. You should install a new, higher-quality power supply and see whether that solves the problem."
    Could this explain why I only have 4-6GB memory at post, but 10GB after a quick power off and back on (didn't bother with a reset switch when designing case). Note that 10GB is still 2 short. It used to initialize 10GB - then power off and back on would provide the full amount. Running less than 6GB memory doesn't cause the error.
    Someone said I'd have to reseat the cpu, but maybe it's just that rubbish coolermaster power supply?
  • chesteracorgi
    Very informative and interesting. The part about single 12V vs. multiple 12V rails is important reading for system builders who opt for "safer" multiple 12V PSUs. With the current state of design of PSUs anyone planning a sli or Xfire rig is well advised to opt for the single 12V design rather than risk an imbalanced PSU that overvolts a component.
  • JohnnyLucky
    Great article. It's not just for beginners.
  • Reynod
    Compatibility Issues was a useful section.

    Overall very well written.

  • kd0frg
    awesome information! nice work!
  • elbert
    cmcghee358But the last part isn't for PSUs. It's just the last part in the series of PC components.
    covering efficiency and the whole 80 PLUS

    If you picked one of these books up you would want the efficiency to move them. Edition 17 was huge and very heavy. These books are already to thick for many to pick up with one hand. Scott Mueller's has published 20 editions of this book and most come with CD/DVD which may guide you to online information about the subject.

    Here is a link to his online forum.
  • digiex
    I'm just wondering what is the use of the floppy connector...

    Until unexpected glitch ruined the flashing if my motherboard, beyond this, I think the floppy connector is useless.
  • mayankleoboy1
    PSU: the most overlooked and underrated component
  • I fixed one of those non-compatible Dells way back with a standard PSU. Dell wanted £120 for a new PSU, I was suspicious, "how could they get away with that?". Checked online, found the incompatibility, dodged the bullet bought a PSU for £20 and an adapter for £5. Never bought Dell again nor recommend them.
  • A Bad Day
    This reminded me of a friend who bought a $5 no-name "600 watt" PSU for a +$900 rig.

    As soon as he turned on the computer, the PSU failed so badly that it exploded into flames and took out everything: motherboard, RAM, CPU, GPU, hard drive, CD drive, you name it.
  • grantmcconnaughey
    I've been reading this book lately. To me, this is absolutely the bible of PC hardware.
  • newbie_mcnoob
    I remember working on an old Dell Dimension 4100 series with the proprietary power supply and RIMM memory. I'm glad those got phased out.
  • hunter315
    In other words, it is far better to have a single 12 V rail that can supply 40 amps than two 12 V rails supplying 20 amps each because with the single rail you don’t have to worry which connectors derive power from which rail and then try to ensure that you don’t overload one or the other.

    Im quite disappointed to see tom's fell for the marketing BS of "a single rail is better than multiple rails". On a well designed unit it does not matter one bit, the design engineers already split the connectors so the rails were reasonably balanced, and the OCP threshold is set such that added together their theoretical current limit is more than the total limit of the 12 V source so you don't have to have your rails perfectly balanced to get the full power out of your unit.

    I wrote up a post on this a while ago, if anyone has any questions or anything they think should be added to it let me know.
    Single 12V rail or multiple 12V rails? The eternal question answered

    Also, you guys left the CPU off the +12 V part of your chart of what requires what voltages.
  • PreferLinux
    ChesteracorgiVery informative and interesting. The part about single 12V vs. multiple 12V rails is important reading for system builders who opt for "safer" multiple 12V PSUs. With the current state of design of PSUs anyone planning a sli or Xfire rig is well advised to opt for the single 12V design rather than risk an imbalanced PSU that overvolts a component.

    I guess it is better to be able to use the 12 V rail as an arc welder then? Because you could if you have a >1000 W single-rail PSU. Not to mention that it won't overvolt anything – how does a high power draw cause high voltages? It generally causes low voltages. And if the PSU is a decent one, the rails will be pretty well balanced, especially for SLI or Crossfire.
  • iam2thecrowe
    ChesteracorgiVery informative and interesting. The part about single 12V vs. multiple 12V rails is important reading for system builders who opt for "safer" multiple 12V PSUs. With the current state of design of PSUs anyone planning a sli or Xfire rig is well advised to opt for the single 12V design rather than risk an imbalanced PSU that overvolts a component.

    you couldn't be more wrong.
  • Onus
    Thank you, Hunter315. Maybe that was an issue 6-8 years ago, maybe? It no longer is; I cannot imagine any possibility of creating sufficiently unbalanced +12V rail loading creating the least difficulty for any reasonable PSU.
    Anyway, there were some interesting tidbits in there, like the specifics on the capacities of individual connectors. The connector type used is another area where I'd be certain the cheap PSUs skimp.
    Some of it was a trip down memory lane; I remember those load resistors in the hard drive cages of the IBM PC-AT.

    Edit: Oh, and your answer to the Eternal Question was very good, as was the short discussion which followed, touching on the safety issues associated with overly large +12V rails. Although I have said many of those things, I doubt I've expressed myself as clearly. Those who have read this chapter and this far into the comments would probably also like to click on your link and read that thread.
  • aidynphoenix
    It really doesn't matter weather you have a psu with a single 12v rail or several.
    what matters is the quality of the power supply and how well it performs.
    viewing a powesupply's voltages on a graph and seeing how far the voltages fluctuate and how smooth the ripple currents are is the most important thing.
    even if your powersupply is 1200w id rather have a psu that can get the job done, but stays as close as possible to the 12v specification, and the power is clean and has little spikes.
    i am a firm believer that poorly designed powersupply's is the #1 cause of premature hardisk failure aside from shipping damage and excessive heat.
    you know that when company's calculate their MTBF they are useing the cleanest power possible.
    I have had alot of experience with poor powersupply's and many HDD deaths.
    now im running a powersupply with a very smooth 12v ripple and high tolorance (thanks corsair)
    and have not had a hardisk die on me since.

    i order recommend 1 seasonic, 2 corsair, 3 antec.. i do not recommend cooler-master.. or any other brand.
  • neiroatopelcc
    neiroatopelcc"Note: If you find that a system consistently fails to boot up properly the first time you turn on the switch, but that it subsequently boots up if you press the reset or Ctrl+Alt+Delete warm boot command, you likely have a problem with the Power_Good timing. You should install a new, higher-quality power supply and see whether that solves the problem." Could this explain why I only have 4-6GB memory at post, but 10GB after a quick power off and back on (didn't bother with a reset switch when designing case). Note that 10GB is still 2 short.

    To answer my own question, as nobody else probably will, I tested it. Installed a reset switch yesterday - well two wires I could shortcircuit anyway. Only effect was that my mainboard thought overclocking had failed. It didn't fix the memory issue. So this is a fat NO. PSU power good timing isn't my issue.
  • ojas

    Any idea when the paperback comes out??? Looking to buy it...